diff --git a/.config/hakari.toml b/.config/hakari.toml
index 12d2d1bf9c..15b939e86f 100644
--- a/.config/hakari.toml
+++ b/.config/hakari.toml
@@ -4,7 +4,7 @@
 hakari-package = "workspace_hack"
 
 # Format for `workspace-hack = ...` lines in other Cargo.tomls. Requires cargo-hakari 0.9.8 or above.
-dep-format-version = "3"
+dep-format-version = "4"
 
 # Setting workspace.resolver = "2" in the root Cargo.toml is HIGHLY recommended.
 # Hakari works much better with the new feature resolver.
diff --git a/.github/PULL_REQUEST_TEMPLATE/release-pr.md b/.github/PULL_REQUEST_TEMPLATE/release-pr.md
index a848077e6a..1e18fd5d44 100644
--- a/.github/PULL_REQUEST_TEMPLATE/release-pr.md
+++ b/.github/PULL_REQUEST_TEMPLATE/release-pr.md
@@ -10,6 +10,7 @@
 <!-- List everything that should be done **before** release, any issues / setting changes / etc -->
 
 ### Checklist after release
+- [ ] Make sure instructions from PRs included in this release and labeled `manual_release_instructions` are executed (either by you or by people who wrote them).
 - [ ] Based on the merged commits write release notes and open a PR into `website` repo ([example](https://github.com/neondatabase/website/pull/219/files))
 - [ ] Check [#dev-production-stream](https://neondb.slack.com/archives/C03F5SM1N02) Slack channel
 - [ ] Check [stuck projects page](https://console.neon.tech/admin/projects?sort=last_active&order=desc&stuck=true)
diff --git a/.github/actions/allure-report/action.yml b/.github/actions/allure-report/action.yml
index e35cbb20fd..9a1037064a 100644
--- a/.github/actions/allure-report/action.yml
+++ b/.github/actions/allure-report/action.yml
@@ -45,12 +45,12 @@ runs:
       shell: bash -euxo pipefail {0}
       run: |
         if [ "${{ inputs.action }}" != "store" ] && [ "${{ inputs.action }}" != "generate" ]; then
-          echo 2>&1 "Unknown inputs.action type '${{ inputs.action }}'; allowed 'generate' or 'store' only"
+          echo >&2 "Unknown inputs.action type '${{ inputs.action }}'; allowed 'generate' or 'store' only"
           exit 1
         fi
 
         if [ -z "${{ inputs.test_selection }}" ] && [ "${{ inputs.action }}" == "store" ]; then
-          echo 2>&1 "inputs.test_selection must be set for 'store' action"
+          echo >&2 "inputs.test_selection must be set for 'store' action"
           exit 2
         fi
 
diff --git a/.github/actions/download/action.yml b/.github/actions/download/action.yml
index eb34d4206a..d3f9bc0414 100644
--- a/.github/actions/download/action.yml
+++ b/.github/actions/download/action.yml
@@ -37,7 +37,7 @@ runs:
             echo 'SKIPPED=true' >> $GITHUB_OUTPUT
             exit 0
           else
-            echo 2>&1 "Neither s3://${BUCKET}/${PREFIX}/${FILENAME} nor its version from previous attempts exist"
+            echo >&2 "Neither s3://${BUCKET}/${PREFIX}/${FILENAME} nor its version from previous attempts exist"
             exit 1
           fi
         fi
diff --git a/.github/actions/neon-branch-create/action.yml b/.github/actions/neon-branch-create/action.yml
index 7ee43a3587..f1eea34ab9 100644
--- a/.github/actions/neon-branch-create/action.yml
+++ b/.github/actions/neon-branch-create/action.yml
@@ -58,7 +58,7 @@ runs:
         done
 
         if [ -z "${branch_id}" ] || [ "${branch_id}" == "null" ]; then
-          echo 2>&1 "Failed to create branch after 10 attempts, the latest response was: ${branch}"
+          echo >&2 "Failed to create branch after 10 attempts, the latest response was: ${branch}"
           exit 1
         fi
 
@@ -122,7 +122,7 @@ runs:
         done
 
         if [ -z "${password}" ] || [ "${password}" == "null" ]; then
-          echo 2>&1 "Failed to reset password after 10 attempts, the latest response was: ${reset_password}"
+          echo >&2 "Failed to reset password after 10 attempts, the latest response was: ${reset_password}"
           exit 1
         fi
 
diff --git a/.github/actions/neon-branch-delete/action.yml b/.github/actions/neon-branch-delete/action.yml
index 5689093e2e..f8cd351dd9 100644
--- a/.github/actions/neon-branch-delete/action.yml
+++ b/.github/actions/neon-branch-delete/action.yml
@@ -48,7 +48,7 @@ runs:
         done
 
         if [ -z "${branch_id}" ] || [ "${branch_id}" == "null" ]; then
-          echo 2>&1 "Failed to delete branch after 10 attempts, the latest response was: ${deleted_branch}"
+          echo >&2 "Failed to delete branch after 10 attempts, the latest response was: ${deleted_branch}"
           exit 1
         fi
       env:
diff --git a/.github/actions/run-python-test-set/action.yml b/.github/actions/run-python-test-set/action.yml
index 11f5c78f19..115f555913 100644
--- a/.github/actions/run-python-test-set/action.yml
+++ b/.github/actions/run-python-test-set/action.yml
@@ -202,7 +202,7 @@ runs:
         prefix: latest
 
     - name: Create Allure report
-      if: success() || failure()
+      if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report
       with:
         action: store
diff --git a/.github/actions/upload/action.yml b/.github/actions/upload/action.yml
index 291a2cf3b0..63973dfbe7 100644
--- a/.github/actions/upload/action.yml
+++ b/.github/actions/upload/action.yml
@@ -23,7 +23,7 @@ runs:
         mkdir -p $(dirname $ARCHIVE)
 
         if [ -f ${ARCHIVE} ]; then
-          echo 2>&1 "File ${ARCHIVE} already exist. Something went wrong before"
+          echo >&2 "File ${ARCHIVE} already exist. Something went wrong before"
           exit 1
         fi
 
@@ -33,10 +33,10 @@ runs:
         elif [ -f ${SOURCE} ]; then
           time tar -cf ${ARCHIVE} --zstd ${SOURCE}
         elif ! ls ${SOURCE} > /dev/null 2>&1; then
-          echo 2>&1 "${SOURCE} does not exist"
+          echo >&2 "${SOURCE} does not exist"
           exit 2
         else
-          echo 2>&1 "${SOURCE} is neither a directory nor a file, do not know how to handle it"
+          echo >&2 "${SOURCE} is neither a directory nor a file, do not know how to handle it"
           exit 3
         fi
 
diff --git a/.github/ansible/prod.ap-southeast-1.hosts.yaml b/.github/ansible/prod.ap-southeast-1.hosts.yaml
index c185086eef..9c53733491 100644
--- a/.github/ansible/prod.ap-southeast-1.hosts.yaml
+++ b/.github/ansible/prod.ap-southeast-1.hosts.yaml
@@ -17,7 +17,7 @@ storage:
           kind: "LayerAccessThreshold"
           period: "10m"
           threshold: &default_eviction_threshold "24h"
-      evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
+        evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
       remote_storage:
         bucket_name: "{{ bucket_name }}"
         bucket_region: "{{ bucket_region }}"
diff --git a/.github/ansible/prod.eu-central-1.hosts.yaml b/.github/ansible/prod.eu-central-1.hosts.yaml
index 0a0f974ea4..3186519ca8 100644
--- a/.github/ansible/prod.eu-central-1.hosts.yaml
+++ b/.github/ansible/prod.eu-central-1.hosts.yaml
@@ -17,7 +17,7 @@ storage:
           kind: "LayerAccessThreshold"
           period: "10m"
           threshold: &default_eviction_threshold "24h"
-      evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
+        evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
       remote_storage:
         bucket_name: "{{ bucket_name }}"
         bucket_region: "{{ bucket_region }}"
diff --git a/.github/ansible/prod.us-east-1.hosts.yaml b/.github/ansible/prod.us-east-1.hosts.yaml
new file mode 100644
index 0000000000..b5b2b076bb
--- /dev/null
+++ b/.github/ansible/prod.us-east-1.hosts.yaml
@@ -0,0 +1,50 @@
+storage:
+  vars:
+    bucket_name: neon-prod-storage-us-east-1
+    bucket_region: us-east-1
+    console_mgmt_base_url: http://neon-internal-api.aws.neon.tech
+    broker_endpoint: http://storage-broker-lb.theta.us-east-1.internal.aws.neon.tech:50051
+    pageserver_config_stub:
+      pg_distrib_dir: /usr/local
+      metric_collection_endpoint: http://neon-internal-api.aws.neon.tech/billing/api/v1/usage_events
+      metric_collection_interval: 10min
+      disk_usage_based_eviction:
+        max_usage_pct: 85 # TODO: decrease to 80 after all pageservers are below 80
+        min_avail_bytes: 0
+        period: "10s"
+      tenant_config:
+        eviction_policy:
+          kind: "LayerAccessThreshold"
+          period: "10m"
+          threshold: &default_eviction_threshold "24h"
+        evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
+      remote_storage:
+        bucket_name: "{{ bucket_name }}"
+        bucket_region: "{{ bucket_region }}"
+        prefix_in_bucket: "pageserver/v1"
+    safekeeper_s3_prefix: safekeeper/v1/wal
+    hostname_suffix: ""
+    remote_user: ssm-user
+    ansible_aws_ssm_region: us-east-1
+    ansible_aws_ssm_bucket_name: neon-prod-storage-us-east-1
+    console_region_id: aws-us-east-1
+    sentry_environment: production
+
+  children:
+    pageservers:
+      hosts:
+        pageserver-0.us-east-1.aws.neon.tech:
+          ansible_host: i-085222088b0d2e0c7
+        pageserver-1.us-east-1.aws.neon.tech:
+          ansible_host: i-0969d4f684d23a21e
+        pageserver-2.us-east-1.aws.neon.tech:
+          ansible_host: i-05dee87895da58dad
+
+    safekeepers:
+      hosts:
+        safekeeper-0.us-east-1.aws.neon.tech:
+          ansible_host: i-04ce739e88793d864
+        safekeeper-1.us-east-1.aws.neon.tech:
+          ansible_host: i-0e9e6c9227fb81410
+        safekeeper-2.us-east-1.aws.neon.tech:
+          ansible_host: i-072f4dd86a327d52f
diff --git a/.github/ansible/prod.us-east-2.hosts.yaml b/.github/ansible/prod.us-east-2.hosts.yaml
index 4427bb344e..3062475b20 100644
--- a/.github/ansible/prod.us-east-2.hosts.yaml
+++ b/.github/ansible/prod.us-east-2.hosts.yaml
@@ -17,7 +17,7 @@ storage:
           kind: "LayerAccessThreshold"
           period: "10m"
           threshold: &default_eviction_threshold "24h"
-      evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
+        evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
       remote_storage:
         bucket_name: "{{ bucket_name }}"
         bucket_region: "{{ bucket_region }}"
diff --git a/.github/ansible/prod.us-west-2.hosts.yaml b/.github/ansible/prod.us-west-2.hosts.yaml
index 53626b4f59..9cf847bcb1 100644
--- a/.github/ansible/prod.us-west-2.hosts.yaml
+++ b/.github/ansible/prod.us-west-2.hosts.yaml
@@ -17,7 +17,7 @@ storage:
           kind: "LayerAccessThreshold"
           period: "10m"
           threshold: &default_eviction_threshold "24h"
-      evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
+        evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
       remote_storage:
         bucket_name: "{{ bucket_name }}"
         bucket_region: "{{ bucket_region }}"
@@ -34,7 +34,7 @@ storage:
     pageservers:
       hosts:
         pageserver-0.us-west-2.aws.neon.tech:
-          ansible_host: i-0d9f6dfae0e1c780d 
+          ansible_host: i-0d9f6dfae0e1c780d
         pageserver-1.us-west-2.aws.neon.tech:
           ansible_host: i-0c834be1dddba8b3f
         pageserver-2.us-west-2.aws.neon.tech:
@@ -49,5 +49,5 @@ storage:
         safekeeper-1.us-west-2.aws.neon.tech:
           ansible_host: i-074682f9d3c712e7c
         safekeeper-2.us-west-2.aws.neon.tech:
-          ansible_host: i-042b7efb1729d7966 
-          
+          ansible_host: i-042b7efb1729d7966
+
diff --git a/.github/ansible/staging.eu-central-1.hosts.yaml b/.github/ansible/staging.eu-central-1.hosts.yaml
new file mode 100644
index 0000000000..db1d1adcff
--- /dev/null
+++ b/.github/ansible/staging.eu-central-1.hosts.yaml
@@ -0,0 +1,47 @@
+storage:
+  vars:
+    bucket_name: neon-dev-storage-eu-central-1
+    bucket_region: eu-central-1
+    # We only register/update storage in one preview console and manually copy to other instances
+    console_mgmt_base_url: http://neon-internal-api.helium.aws.neon.build
+    broker_endpoint: http://storage-broker-lb.alpha.eu-central-1.internal.aws.neon.build:50051
+    pageserver_config_stub:
+      pg_distrib_dir: /usr/local
+      metric_collection_endpoint: http://neon-internal-api.helium.aws.neon.build/billing/api/v1/usage_events
+      metric_collection_interval: 10min
+      disk_usage_based_eviction:
+        max_usage_pct: 80
+        min_avail_bytes: 0
+        period: "10s"
+      tenant_config:
+        eviction_policy:
+          kind: "LayerAccessThreshold"
+          period: "20m"
+          threshold: &default_eviction_threshold "20m"
+      evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
+      remote_storage:
+        bucket_name: "{{ bucket_name }}"
+        bucket_region: "{{ bucket_region }}"
+        prefix_in_bucket: "pageserver/v1"
+    safekeeper_s3_prefix: safekeeper/v1/wal
+    hostname_suffix: ""
+    remote_user: ssm-user
+    ansible_aws_ssm_region: eu-central-1
+    ansible_aws_ssm_bucket_name: neon-dev-storage-eu-central-1
+    console_region_id: aws-eu-central-1
+    sentry_environment: staging
+
+  children:
+    pageservers:
+      hosts:
+        pageserver-0.eu-central-1.aws.neon.build:
+          ansible_host: i-011f93ec26cfba2d4
+
+    safekeepers:
+      hosts:
+        safekeeper-0.eu-central-1.aws.neon.build:
+          ansible_host: i-0ff026d27babf8ddd
+        safekeeper-1.eu-central-1.aws.neon.build:
+          ansible_host: i-03983a49ee54725d9
+        safekeeper-2.eu-central-1.aws.neon.build:
+          ansible_host: i-0bd025ecdb61b0db3
diff --git a/.github/ansible/staging.eu-west-1.hosts.yaml b/.github/ansible/staging.eu-west-1.hosts.yaml
index 34c8e77280..39f5613935 100644
--- a/.github/ansible/staging.eu-west-1.hosts.yaml
+++ b/.github/ansible/staging.eu-west-1.hosts.yaml
@@ -17,7 +17,7 @@ storage:
           kind: "LayerAccessThreshold"
           period: "20m"
           threshold: &default_eviction_threshold "20m"
-      evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
+        evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
       remote_storage:
         bucket_name: "{{ bucket_name }}"
         bucket_region: "{{ bucket_region }}"
diff --git a/.github/ansible/staging.us-east-2.hosts.yaml b/.github/ansible/staging.us-east-2.hosts.yaml
index 94f2be83a4..fb218c443d 100644
--- a/.github/ansible/staging.us-east-2.hosts.yaml
+++ b/.github/ansible/staging.us-east-2.hosts.yaml
@@ -17,7 +17,7 @@ storage:
           kind: "LayerAccessThreshold"
           period: "20m"
           threshold: &default_eviction_threshold "20m"
-      evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
+        evictions_low_residence_duration_metric_threshold: *default_eviction_threshold
       remote_storage:
         bucket_name: "{{ bucket_name }}"
         bucket_region: "{{ bucket_region }}"
@@ -48,9 +48,9 @@ storage:
       hosts:
         safekeeper-0.us-east-2.aws.neon.build:
           ansible_host: i-027662bd552bf5db0
-        safekeeper-1.us-east-2.aws.neon.build:
-          ansible_host: i-0171efc3604a7b907
         safekeeper-2.us-east-2.aws.neon.build:
           ansible_host: i-0de0b03a51676a6ce
+        safekeeper-3.us-east-2.aws.neon.build:
+          ansible_host: i-05f8ba2cda243bd18
         safekeeper-99.us-east-2.aws.neon.build:
           ansible_host: i-0d61b6a2ea32028d5
diff --git a/.github/helm-values/dev-eu-central-1-alpha.neon-storage-broker.yaml b/.github/helm-values/dev-eu-central-1-alpha.neon-storage-broker.yaml
new file mode 100644
index 0000000000..aaa1ec59b4
--- /dev/null
+++ b/.github/helm-values/dev-eu-central-1-alpha.neon-storage-broker.yaml
@@ -0,0 +1,52 @@
+# Helm chart values for neon-storage-broker
+podLabels:
+  neon_env: staging
+  neon_service: storage-broker
+
+# Use L4 LB
+service:
+  # service.annotations -- Annotations to add to the service
+  annotations:
+    service.beta.kubernetes.io/aws-load-balancer-type: external  # use newer AWS Load Balancer Controller
+    service.beta.kubernetes.io/aws-load-balancer-nlb-target-type: ip
+    service.beta.kubernetes.io/aws-load-balancer-scheme: internal  # deploy LB to private subnet
+    # assign service to this name at external-dns
+    external-dns.alpha.kubernetes.io/hostname: storage-broker-lb.alpha.eu-central-1.internal.aws.neon.build
+  # service.type -- Service type
+  type: LoadBalancer
+  # service.port -- broker listen port
+  port: 50051
+
+ingress:
+  enabled: false
+
+metrics:
+  enabled: false
+
+extraManifests:
+  - apiVersion: operator.victoriametrics.com/v1beta1
+    kind: VMServiceScrape
+    metadata:
+      name: "{{ include \"neon-storage-broker.fullname\" . }}"
+      labels:
+        helm.sh/chart: neon-storage-broker-{{ .Chart.Version }}
+        app.kubernetes.io/name: neon-storage-broker
+        app.kubernetes.io/instance: neon-storage-broker
+        app.kubernetes.io/version: "{{ .Chart.AppVersion }}"
+        app.kubernetes.io/managed-by: Helm
+      namespace: "{{ .Release.Namespace }}"
+    spec:
+      selector:
+        matchLabels:
+          app.kubernetes.io/name: "neon-storage-broker"
+      endpoints:
+        - port: broker
+          path: /metrics
+          interval: 10s
+          scrapeTimeout: 10s
+      namespaceSelector:
+        matchNames:
+          - "{{ .Release.Namespace }}"
+
+settings:
+  sentryEnvironment: "staging"
diff --git a/.github/helm-values/dev-eu-west-1-zeta.neon-proxy-scram.yaml b/.github/helm-values/dev-eu-west-1-zeta.neon-proxy-scram.yaml
index 2307856464..a7d8587ec2 100644
--- a/.github/helm-values/dev-eu-west-1-zeta.neon-proxy-scram.yaml
+++ b/.github/helm-values/dev-eu-west-1-zeta.neon-proxy-scram.yaml
@@ -7,13 +7,13 @@ deploymentStrategy:
     maxSurge: 100%
     maxUnavailable: 50%
 
-# Delay the kill signal by 7 days (7 * 24 * 60 * 60)
+# Delay the kill signal by 5 minutes (5 * 60)
 # The pod(s) will stay in Terminating, keeps the existing connections
 # but doesn't receive new ones
 containerLifecycle:
   preStop:
     exec:
-      command: ["/bin/sh", "-c", "sleep 604800"]
+      command: ["/bin/sh", "-c", "sleep 300"]
 terminationGracePeriodSeconds: 604800
 
 image:
@@ -23,6 +23,7 @@ settings:
   authBackend: "console"
   authEndpoint: "http://neon-internal-api.aws.neon.build/management/api/v2"
   domain: "*.eu-west-1.aws.neon.build"
+  otelExporterOtlpEndpoint: "https://otel-collector.zeta.eu-west-1.internal.aws.neon.build"
   sentryEnvironment: "staging"
   wssPort: 8443
   metricCollectionEndpoint: "http://neon-internal-api.aws.neon.build/billing/api/v1/usage_events"
diff --git a/.github/helm-values/dev-us-east-2-beta.neon-proxy-link.yaml b/.github/helm-values/dev-us-east-2-beta.neon-proxy-link.yaml
index feca05aff6..893e0fab10 100644
--- a/.github/helm-values/dev-us-east-2-beta.neon-proxy-link.yaml
+++ b/.github/helm-values/dev-us-east-2-beta.neon-proxy-link.yaml
@@ -9,6 +9,7 @@ settings:
   authEndpoint: "https://console.stage.neon.tech/authenticate_proxy_request/"
   uri: "https://console.stage.neon.tech/psql_session/"
   domain: "pg.neon.build"
+  otelExporterOtlpEndpoint: "https://otel-collector.beta.us-east-2.internal.aws.neon.build"
   sentryEnvironment: "staging"
   metricCollectionEndpoint: "http://neon-internal-api.aws.neon.build/billing/api/v1/usage_events"
   metricCollectionInterval: "1min"
diff --git a/.github/helm-values/dev-us-east-2-beta.neon-proxy-scram-legacy.yaml b/.github/helm-values/dev-us-east-2-beta.neon-proxy-scram-legacy.yaml
index feee1b369a..77f6cf080e 100644
--- a/.github/helm-values/dev-us-east-2-beta.neon-proxy-scram-legacy.yaml
+++ b/.github/helm-values/dev-us-east-2-beta.neon-proxy-scram-legacy.yaml
@@ -1,6 +1,22 @@
 # Helm chart values for neon-proxy-scram.
 # This is a YAML-formatted file.
 
+deploymentStrategy:
+  type: RollingUpdate
+  rollingUpdate:
+    maxSurge: 100%
+    maxUnavailable: 50%
+
+# Delay the kill signal by 5 minutes (5 * 60)
+# The pod(s) will stay in Terminating, keeps the existing connections
+# but doesn't receive new ones
+containerLifecycle:
+  preStop:
+    exec:
+      command: ["/bin/sh", "-c", "sleep 300"]
+terminationGracePeriodSeconds: 604800
+
+
 image:
   repository: neondatabase/neon
 
@@ -8,6 +24,7 @@ settings:
   authBackend: "console"
   authEndpoint: "http://neon-internal-api.aws.neon.build/management/api/v2"
   domain: "*.cloud.stage.neon.tech"
+  otelExporterOtlpEndpoint: "https://otel-collector.beta.us-east-2.internal.aws.neon.build"
   sentryEnvironment: "staging"
   wssPort: 8443
   metricCollectionEndpoint: "http://neon-internal-api.aws.neon.build/billing/api/v1/usage_events"
diff --git a/.github/helm-values/dev-us-east-2-beta.neon-proxy-scram.yaml b/.github/helm-values/dev-us-east-2-beta.neon-proxy-scram.yaml
index 2a8f028f3b..2510d624cd 100644
--- a/.github/helm-values/dev-us-east-2-beta.neon-proxy-scram.yaml
+++ b/.github/helm-values/dev-us-east-2-beta.neon-proxy-scram.yaml
@@ -7,15 +7,16 @@ deploymentStrategy:
     maxSurge: 100%
     maxUnavailable: 50%
 
-# Delay the kill signal by 7 days (7 * 24 * 60 * 60)
+# Delay the kill signal by 5 minutes (5 * 60)
 # The pod(s) will stay in Terminating, keeps the existing connections
 # but doesn't receive new ones
 containerLifecycle:
   preStop:
     exec:
-      command: ["/bin/sh", "-c", "sleep 604800"]
+      command: ["/bin/sh", "-c", "sleep 300"]
 terminationGracePeriodSeconds: 604800
 
+
 image:
   repository: neondatabase/neon
 
@@ -24,6 +25,7 @@ settings:
   authEndpoint: "http://neon-internal-api.aws.neon.build/management/api/v2"
   domain: "*.us-east-2.aws.neon.build"
   extraDomains: ["*.us-east-2.postgres.zenith.tech", "*.us-east-2.retooldb-staging.com"]
+  otelExporterOtlpEndpoint: "https://otel-collector.beta.us-east-2.internal.aws.neon.build"
   sentryEnvironment: "staging"
   wssPort: 8443
   metricCollectionEndpoint: "http://neon-internal-api.aws.neon.build/billing/api/v1/usage_events"
diff --git a/.github/helm-values/preview-template.neon-proxy-scram.yaml b/.github/helm-values/preview-template.neon-proxy-scram.yaml
new file mode 100644
index 0000000000..f4bd418e28
--- /dev/null
+++ b/.github/helm-values/preview-template.neon-proxy-scram.yaml
@@ -0,0 +1,67 @@
+# Helm chart values for neon-proxy-scram.
+# This is a YAML-formatted file.
+
+deploymentStrategy:
+  type: RollingUpdate
+  rollingUpdate:
+    maxSurge: 100%
+    maxUnavailable: 50%
+
+image:
+  repository: neondatabase/neon
+
+settings:
+  authBackend: "console"
+  authEndpoint: "http://neon-internal-api.${PREVIEW_NAME}.aws.neon.build/management/api/v2"
+  domain: "*.cloud.${PREVIEW_NAME}.aws.neon.build"
+  sentryEnvironment: "staging"
+  wssPort: 8443
+  metricCollectionEndpoint: "http://neon-internal-api.${PREVIEW_NAME}.aws.neon.build/billing/api/v1/usage_events"
+  metricCollectionInterval: "1min"
+
+# -- Additional labels for neon-proxy pods
+podLabels:
+  neon_service: proxy-scram
+  neon_env: test
+  neon_region: ${PREVIEW_NAME}.eu-central-1
+
+
+exposedService:
+  annotations:
+    service.beta.kubernetes.io/aws-load-balancer-type: external
+    service.beta.kubernetes.io/aws-load-balancer-nlb-target-type: ip
+    service.beta.kubernetes.io/aws-load-balancer-scheme: internet-facing
+    external-dns.alpha.kubernetes.io/hostname: cloud.${PREVIEW_NAME}.aws.neon.build
+  httpsPort: 443
+
+#metrics:
+#  enabled: true
+#  serviceMonitor:
+#    enabled: true
+#    selector:
+#      release: kube-prometheus-stack
+
+extraManifests:
+  - apiVersion: operator.victoriametrics.com/v1beta1
+    kind: VMServiceScrape
+    metadata:
+      name: "{{ include \"neon-proxy.fullname\" . }}"
+      labels:
+        helm.sh/chart: neon-proxy-{{ .Chart.Version }}
+        app.kubernetes.io/name: neon-proxy
+        app.kubernetes.io/instance: "{{ include \"neon-proxy.fullname\" . }}"
+        app.kubernetes.io/version: "{{ .Chart.AppVersion }}"
+        app.kubernetes.io/managed-by: Helm
+      namespace: "{{ .Release.Namespace }}"
+    spec:
+      selector:
+        matchLabels:
+          app.kubernetes.io/name: "neon-proxy"
+      endpoints:
+        - port: http
+          path: /metrics
+          interval: 10s
+          scrapeTimeout: 10s
+      namespaceSelector:
+        matchNames:
+          - "{{ .Release.Namespace }}"
diff --git a/.github/helm-values/prod-ap-southeast-1-epsilon.neon-proxy-scram.yaml b/.github/helm-values/prod-ap-southeast-1-epsilon.neon-proxy-scram.yaml
index 5a98217bae..6088d62fba 100644
--- a/.github/helm-values/prod-ap-southeast-1-epsilon.neon-proxy-scram.yaml
+++ b/.github/helm-values/prod-ap-southeast-1-epsilon.neon-proxy-scram.yaml
@@ -7,13 +7,13 @@ deploymentStrategy:
     maxSurge: 100%
     maxUnavailable: 50%
 
-# Delay the kill signal by 7 days (7 * 24 * 60 * 60)
+# Delay the kill signal by 5 minutes (5 * 60)
 # The pod(s) will stay in Terminating, keeps the existing connections
 # but doesn't receive new ones
 containerLifecycle:
   preStop:
     exec:
-      command: ["/bin/sh", "-c", "sleep 604800"]
+      command: ["/bin/sh", "-c", "sleep 300"]
 terminationGracePeriodSeconds: 604800
 
 
diff --git a/.github/helm-values/prod-eu-central-1-gamma.neon-proxy-scram.yaml b/.github/helm-values/prod-eu-central-1-gamma.neon-proxy-scram.yaml
index a9ee49d82f..7d26f2e02f 100644
--- a/.github/helm-values/prod-eu-central-1-gamma.neon-proxy-scram.yaml
+++ b/.github/helm-values/prod-eu-central-1-gamma.neon-proxy-scram.yaml
@@ -7,13 +7,13 @@ deploymentStrategy:
     maxSurge: 100%
     maxUnavailable: 50%
 
-# Delay the kill signal by 7 days (7 * 24 * 60 * 60)
+# Delay the kill signal by 5 minutes (5 * 60)
 # The pod(s) will stay in Terminating, keeps the existing connections
 # but doesn't receive new ones
 containerLifecycle:
   preStop:
     exec:
-      command: ["/bin/sh", "-c", "sleep 604800"]
+      command: ["/bin/sh", "-c", "sleep 300"]
 terminationGracePeriodSeconds: 604800
 
 
diff --git a/.github/helm-values/prod-us-east-1-theta.neon-proxy-scram.yaml b/.github/helm-values/prod-us-east-1-theta.neon-proxy-scram.yaml
new file mode 100644
index 0000000000..1c7e646810
--- /dev/null
+++ b/.github/helm-values/prod-us-east-1-theta.neon-proxy-scram.yaml
@@ -0,0 +1,69 @@
+# Helm chart values for neon-proxy-scram.
+# This is a YAML-formatted file.
+
+deploymentStrategy:
+  type: RollingUpdate
+  rollingUpdate:
+    maxSurge: 100%
+    maxUnavailable: 50%
+
+# Delay the kill signal by 5 minutes (5 * 60)
+# The pod(s) will stay in Terminating, keeps the existing connections
+# but doesn't receive new ones
+containerLifecycle:
+  preStop:
+    exec:
+      command: ["/bin/sh", "-c", "sleep 300"]
+terminationGracePeriodSeconds: 604800
+
+image:
+  repository: neondatabase/neon
+
+settings:
+  authBackend: "console"
+  authEndpoint: "http://neon-internal-api.aws.neon.tech/management/api/v2"
+  domain: "*.us-east-1.aws.neon.tech"
+  # *.us-east-1.retooldb.com hasn't been delegated yet.
+  extraDomains: ["*.us-east-1.postgres.vercel-storage.com"]
+  sentryEnvironment: "production"
+  wssPort: 8443
+  metricCollectionEndpoint: "http://neon-internal-api.aws.neon.tech/billing/api/v1/usage_events"
+  metricCollectionInterval: "10min"
+
+podLabels:
+  neon_service: proxy-scram
+  neon_env: prod
+  neon_region: us-east-1
+
+exposedService:
+  annotations:
+    service.beta.kubernetes.io/aws-load-balancer-type: external
+    service.beta.kubernetes.io/aws-load-balancer-nlb-target-type: ip
+    service.beta.kubernetes.io/aws-load-balancer-scheme: internet-facing
+    external-dns.alpha.kubernetes.io/hostname: us-east-1.aws.neon.tech
+  httpsPort: 443
+
+extraManifests:
+  - apiVersion: operator.victoriametrics.com/v1beta1
+    kind: VMServiceScrape
+    metadata:
+      name: "{{ include \"neon-proxy.fullname\" . }}"
+      labels:
+        helm.sh/chart: neon-proxy-{{ .Chart.Version }}
+        app.kubernetes.io/name: neon-proxy
+        app.kubernetes.io/instance: "{{ include \"neon-proxy.fullname\" . }}"
+        app.kubernetes.io/version: "{{ .Chart.AppVersion }}"
+        app.kubernetes.io/managed-by: Helm
+      namespace: "{{ .Release.Namespace }}"
+    spec:
+      selector:
+        matchLabels:
+          app.kubernetes.io/name: "neon-proxy"
+      endpoints:
+        - port: http
+          path: /metrics
+          interval: 10s
+          scrapeTimeout: 10s
+      namespaceSelector:
+        matchNames:
+          - "{{ .Release.Namespace }}"
diff --git a/.github/helm-values/prod-us-east-1-theta.neon-storage-broker.yaml b/.github/helm-values/prod-us-east-1-theta.neon-storage-broker.yaml
new file mode 100644
index 0000000000..7c16911b5e
--- /dev/null
+++ b/.github/helm-values/prod-us-east-1-theta.neon-storage-broker.yaml
@@ -0,0 +1,52 @@
+# Helm chart values for neon-storage-broker
+podLabels:
+  neon_env: production
+  neon_service: storage-broker
+
+# Use L4 LB
+service:
+  # service.annotations -- Annotations to add to the service
+  annotations:
+    service.beta.kubernetes.io/aws-load-balancer-type: external  # use newer AWS Load Balancer Controller
+    service.beta.kubernetes.io/aws-load-balancer-nlb-target-type: ip
+    service.beta.kubernetes.io/aws-load-balancer-scheme: internal  # deploy LB to private subnet
+    # assign service to this name at external-dns
+    external-dns.alpha.kubernetes.io/hostname: storage-broker-lb.theta.us-east-1.internal.aws.neon.tech
+  # service.type -- Service type
+  type: LoadBalancer
+  # service.port -- broker listen port
+  port: 50051
+
+ingress:
+  enabled: false
+
+metrics:
+  enabled: false
+
+extraManifests:
+  - apiVersion: operator.victoriametrics.com/v1beta1
+    kind: VMServiceScrape
+    metadata:
+      name: "{{ include \"neon-storage-broker.fullname\" . }}"
+      labels:
+        helm.sh/chart: neon-storage-broker-{{ .Chart.Version }}
+        app.kubernetes.io/name: neon-storage-broker
+        app.kubernetes.io/instance: neon-storage-broker
+        app.kubernetes.io/version: "{{ .Chart.AppVersion }}"
+        app.kubernetes.io/managed-by: Helm
+      namespace: "{{ .Release.Namespace }}"
+    spec:
+      selector:
+        matchLabels:
+          app.kubernetes.io/name: "neon-storage-broker"
+      endpoints:
+        - port: broker
+          path: /metrics
+          interval: 10s
+          scrapeTimeout: 10s
+      namespaceSelector:
+        matchNames:
+          - "{{ .Release.Namespace }}"
+
+settings:
+  sentryEnvironment: "production"
diff --git a/.github/helm-values/prod-us-east-2-delta.neon-proxy-scram.yaml b/.github/helm-values/prod-us-east-2-delta.neon-proxy-scram.yaml
index 239a9911c7..ae239fd3c1 100644
--- a/.github/helm-values/prod-us-east-2-delta.neon-proxy-scram.yaml
+++ b/.github/helm-values/prod-us-east-2-delta.neon-proxy-scram.yaml
@@ -7,13 +7,13 @@ deploymentStrategy:
     maxSurge: 100%
     maxUnavailable: 50%
 
-# Delay the kill signal by 7 days (7 * 24 * 60 * 60)
+# Delay the kill signal by 5 minutes (5 * 60)
 # The pod(s) will stay in Terminating, keeps the existing connections
 # but doesn't receive new ones
 containerLifecycle:
   preStop:
     exec:
-      command: ["/bin/sh", "-c", "sleep 604800"]
+      command: ["/bin/sh", "-c", "sleep 300"]
 terminationGracePeriodSeconds: 604800
 
 
diff --git a/.github/helm-values/prod-us-west-2-eta.neon-proxy-scram-legacy.yaml b/.github/helm-values/prod-us-west-2-eta.neon-proxy-scram-legacy.yaml
index a186fb833f..7378e8abda 100644
--- a/.github/helm-values/prod-us-west-2-eta.neon-proxy-scram-legacy.yaml
+++ b/.github/helm-values/prod-us-west-2-eta.neon-proxy-scram-legacy.yaml
@@ -7,13 +7,13 @@ deploymentStrategy:
     maxSurge: 100%
     maxUnavailable: 50%
 
-# Delay the kill signal by 7 days (7 * 24 * 60 * 60)
+# Delay the kill signal by 5 minutes (5 * 60)
 # The pod(s) will stay in Terminating, keeps the existing connections
 # but doesn't receive new ones
 containerLifecycle:
   preStop:
     exec:
-      command: ["/bin/sh", "-c", "sleep 604800"]
+      command: ["/bin/sh", "-c", "sleep 300"]
 terminationGracePeriodSeconds: 604800
 
 
diff --git a/.github/helm-values/prod-us-west-2-eta.neon-proxy-scram.yaml b/.github/helm-values/prod-us-west-2-eta.neon-proxy-scram.yaml
index c987ae236a..d9d458f081 100644
--- a/.github/helm-values/prod-us-west-2-eta.neon-proxy-scram.yaml
+++ b/.github/helm-values/prod-us-west-2-eta.neon-proxy-scram.yaml
@@ -7,13 +7,13 @@ deploymentStrategy:
     maxSurge: 100%
     maxUnavailable: 50%
 
-# Delay the kill signal by 7 days (7 * 24 * 60 * 60)
+# Delay the kill signal by 5 minutes (5 * 60)
 # The pod(s) will stay in Terminating, keeps the existing connections
 # but doesn't receive new ones
 containerLifecycle:
   preStop:
     exec:
-      command: ["/bin/sh", "-c", "sleep 604800"]
+      command: ["/bin/sh", "-c", "sleep 300"]
 terminationGracePeriodSeconds: 604800
 
 
diff --git a/.github/workflows/benchmarking.yml b/.github/workflows/benchmarking.yml
index 028fe8d8ad..a5a27e59a8 100644
--- a/.github/workflows/benchmarking.yml
+++ b/.github/workflows/benchmarking.yml
@@ -30,7 +30,7 @@ defaults:
 
 concurrency:
   # Allow only one workflow per any non-`main` branch.
-  group: ${{ github.workflow }}-${{ github.ref }}-${{ github.ref == 'refs/heads/main' && github.sha || 'anysha' }}
+  group: ${{ github.workflow }}-${{ github.ref_name }}-${{ github.ref_name == 'main' && github.sha || 'anysha' }}
   cancel-in-progress: true
 
 jobs:
@@ -42,7 +42,7 @@ jobs:
       DEFAULT_PG_VERSION: 14
       TEST_OUTPUT: /tmp/test_output
       BUILD_TYPE: remote
-      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref == 'refs/heads/main' ) }}
+      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref_name == 'main' ) }}
       PLATFORM: "neon-staging"
 
     runs-on: [ self-hosted, us-east-2, x64 ]
@@ -92,7 +92,7 @@ jobs:
         api_key: ${{ secrets.NEON_STAGING_API_KEY }}
 
     - name: Create Allure report
-      if: success() || failure()
+      if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report
       with:
         action: generate
@@ -174,7 +174,7 @@ jobs:
       DEFAULT_PG_VERSION: 14
       TEST_OUTPUT: /tmp/test_output
       BUILD_TYPE: remote
-      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref == 'refs/heads/main' ) }}
+      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref_name == 'main' ) }}
       PLATFORM: ${{ matrix.platform }}
 
     runs-on: [ self-hosted, us-east-2, x64 ]
@@ -226,7 +226,7 @@ jobs:
             CONNSTR=${{ secrets.BENCHMARK_RDS_POSTGRES_CONNSTR }}
             ;;
           *)
-            echo 2>&1 "Unknown PLATFORM=${PLATFORM}. Allowed only 'neon-captest-reuse', 'neon-captest-new', 'neon-captest-freetier', 'rds-aurora', or 'rds-postgres'"
+            echo >&2 "Unknown PLATFORM=${PLATFORM}. Allowed only 'neon-captest-reuse', 'neon-captest-new', 'neon-captest-freetier', 'rds-aurora', or 'rds-postgres'"
             exit 1
             ;;
         esac
@@ -282,7 +282,7 @@ jobs:
         api_key: ${{ secrets.NEON_STAGING_API_KEY }}
 
     - name: Create Allure report
-      if: success() || failure()
+      if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report
       with:
         action: generate
@@ -305,7 +305,7 @@ jobs:
     #
     # *_CLICKBENCH_CONNSTR: Genuine ClickBench DB with ~100M rows
     # *_CLICKBENCH_10M_CONNSTR: DB with the first 10M rows of ClickBench DB
-    if: success() || failure()
+    if: ${{ !cancelled() }}
     needs: [ generate-matrices, pgbench-compare ]
 
     strategy:
@@ -317,7 +317,7 @@ jobs:
       DEFAULT_PG_VERSION: 14
       TEST_OUTPUT: /tmp/test_output
       BUILD_TYPE: remote
-      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref == 'refs/heads/main' ) }}
+      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref_name == 'main' ) }}
       PLATFORM: ${{ matrix.platform }}
 
     runs-on: [ self-hosted, us-east-2, x64 ]
@@ -356,7 +356,7 @@ jobs:
             CONNSTR=${{ secrets.BENCHMARK_RDS_POSTGRES_CLICKBENCH_10M_CONNSTR }}
             ;;
           *)
-            echo 2>&1 "Unknown PLATFORM=${PLATFORM}. Allowed only 'neon-captest-reuse', 'rds-aurora', or 'rds-postgres'"
+            echo >&2 "Unknown PLATFORM=${PLATFORM}. Allowed only 'neon-captest-reuse', 'rds-aurora', or 'rds-postgres'"
             exit 1
             ;;
         esac
@@ -379,7 +379,7 @@ jobs:
         BENCHMARK_CONNSTR: ${{ steps.set-up-connstr.outputs.connstr }}
 
     - name: Create Allure report
-      if: success() || failure()
+      if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report
       with:
         action: generate
@@ -401,7 +401,7 @@ jobs:
     # We might change it after https://github.com/neondatabase/neon/issues/2900.
     #
     # *_TPCH_S10_CONNSTR: DB generated with scale factor 10 (~10 GB)
-    if: success() || failure()
+    if: ${{ !cancelled() }}
     needs: [ generate-matrices, clickbench-compare ]
 
     strategy:
@@ -413,7 +413,7 @@ jobs:
       DEFAULT_PG_VERSION: 14
       TEST_OUTPUT: /tmp/test_output
       BUILD_TYPE: remote
-      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref == 'refs/heads/main' ) }}
+      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref_name == 'main' ) }}
       PLATFORM: ${{ matrix.platform }}
 
     runs-on: [ self-hosted, us-east-2, x64 ]
@@ -452,7 +452,7 @@ jobs:
             CONNSTR=${{ secrets.BENCHMARK_RDS_POSTGRES_TPCH_S10_CONNSTR }}
             ;;
           *)
-            echo 2>&1 "Unknown PLATFORM=${PLATFORM}. Allowed only 'neon-captest-reuse', 'rds-aurora', or 'rds-postgres'"
+            echo >&2 "Unknown PLATFORM=${PLATFORM}. Allowed only 'neon-captest-reuse', 'rds-aurora', or 'rds-postgres'"
             exit 1
             ;;
         esac
@@ -475,7 +475,7 @@ jobs:
         BENCHMARK_CONNSTR: ${{ steps.set-up-connstr.outputs.connstr }}
 
     - name: Create Allure report
-      if: success() || failure()
+      if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report
       with:
         action: generate
@@ -491,7 +491,7 @@ jobs:
         SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}
 
   user-examples-compare:
-    if: success() || failure()
+    if: ${{ !cancelled() }}
     needs: [ generate-matrices, tpch-compare ]
 
     strategy:
@@ -503,7 +503,7 @@ jobs:
       DEFAULT_PG_VERSION: 14
       TEST_OUTPUT: /tmp/test_output
       BUILD_TYPE: remote
-      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref == 'refs/heads/main' ) }}
+      SAVE_PERF_REPORT: ${{ github.event.inputs.save_perf_report || ( github.ref_name == 'main' ) }}
       PLATFORM: ${{ matrix.platform }}
 
     runs-on: [ self-hosted, us-east-2, x64 ]
@@ -542,7 +542,7 @@ jobs:
             CONNSTR=${{ secrets.BENCHMARK_USER_EXAMPLE_RDS_POSTGRES_CONNSTR }}
             ;;
           *)
-            echo 2>&1 "Unknown PLATFORM=${PLATFORM}. Allowed only 'neon-captest-reuse', 'rds-aurora', or 'rds-postgres'"
+            echo >&2 "Unknown PLATFORM=${PLATFORM}. Allowed only 'neon-captest-reuse', 'rds-aurora', or 'rds-postgres'"
             exit 1
             ;;
         esac
@@ -565,7 +565,7 @@ jobs:
         BENCHMARK_CONNSTR: ${{ steps.set-up-connstr.outputs.connstr }}
 
     - name: Create Allure report
-      if: success() || failure()
+      if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report
       with:
         action: generate
diff --git a/.github/workflows/build_and_test.yml b/.github/workflows/build_and_test.yml
index c096aef4a9..e5ba7aa3eb 100644
--- a/.github/workflows/build_and_test.yml
+++ b/.github/workflows/build_and_test.yml
@@ -13,7 +13,7 @@ defaults:
 
 concurrency:
   # Allow only one workflow per any non-`main` branch.
-  group: ${{ github.workflow }}-${{ github.ref }}-${{ github.ref == 'refs/heads/main' && github.sha || 'anysha' }}
+  group: ${{ github.workflow }}-${{ github.ref_name }}-${{ github.ref_name == 'main' && github.sha || 'anysha' }}
   cancel-in-progress: true
 
 env:
@@ -111,8 +111,21 @@ jobs:
       - name: Get postgres headers
         run: make postgres-headers -j$(nproc)
 
-      - name: Run cargo clippy
-        run: ./run_clippy.sh
+      # cargo hack runs the given cargo subcommand (clippy in this case) for all feature combinations.
+      # This will catch compiler & clippy warnings in all feature combinations.
+      # TODO: use cargo hack for build and test as well, but, that's quite expensive.
+      # NB: keep clippy args in sync with ./run_clippy.sh
+      - run: |
+          CLIPPY_COMMON_ARGS="$( source .neon_clippy_args; echo "$CLIPPY_COMMON_ARGS")"
+          if [ "$CLIPPY_COMMON_ARGS" = "" ]; then
+            echo "No clippy args found in .neon_clippy_args"
+            exit 1
+          fi
+          echo "CLIPPY_COMMON_ARGS=${CLIPPY_COMMON_ARGS}" >> $GITHUB_ENV
+      - name: Run cargo clippy (debug)
+        run: cargo hack --feature-powerset clippy $CLIPPY_COMMON_ARGS
+      - name: Run cargo clippy (release)
+        run: cargo hack --feature-powerset clippy --release $CLIPPY_COMMON_ARGS
 
       # Use `${{ !cancelled() }}` to run quck tests after the longer clippy run
       - name: Check formatting
@@ -368,7 +381,7 @@ jobs:
           build_type: ${{ matrix.build_type }}
           test_selection: performance
           run_in_parallel: false
-          save_perf_report: ${{ github.ref == 'refs/heads/main' }}
+          save_perf_report: ${{ github.ref_name == 'main' }}
         env:
           VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
           PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
@@ -541,7 +554,7 @@ jobs:
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/base:pinned
       options: --init
-    needs: [ push-docker-hub, tag ]
+    needs: [ promote-images, tag ]
     steps:
       - name: Set PR's status to pending and request a remote CI test
         run: |
@@ -584,8 +597,7 @@ jobs:
   neon-image:
     runs-on: [ self-hosted, gen3, large ]
     needs: [ tag ]
-    # https://github.com/GoogleContainerTools/kaniko/issues/2005
-    container: gcr.io/kaniko-project/executor:v1.7.0-debug
+    container: gcr.io/kaniko-project/executor:v1.9.2-debug
     defaults:
       run:
         shell: sh -eu {0}
@@ -597,11 +609,32 @@ jobs:
           submodules: true
           fetch-depth: 0
 
-      - name: Configure ECR login
-        run: echo "{\"credsStore\":\"ecr-login\"}" > /kaniko/.docker/config.json
+      - name: Configure ECR and Docker Hub login
+        run: |
+          DOCKERHUB_AUTH=$(echo -n "${{ secrets.NEON_DOCKERHUB_USERNAME }}:${{ secrets.NEON_DOCKERHUB_PASSWORD }}" | base64)
+          echo "::add-mask::${DOCKERHUB_AUTH}"
+
+          cat <<-EOF > /kaniko/.docker/config.json
+            {
+              "auths": {
+                "https://index.docker.io/v1/": {
+                  "auth": "${DOCKERHUB_AUTH}"
+                }
+              },
+              "credHelpers": {
+                "369495373322.dkr.ecr.eu-central-1.amazonaws.com": "ecr-login"
+              }
+            }
+          EOF
 
       - name: Kaniko build neon
-        run: /kaniko/executor --reproducible --snapshotMode=redo --skip-unused-stages --cache=true --cache-repo 369495373322.dkr.ecr.eu-central-1.amazonaws.com/cache --context . --build-arg GIT_VERSION=${{ github.sha }} --destination 369495373322.dkr.ecr.eu-central-1.amazonaws.com/neon:${{needs.tag.outputs.build-tag}}
+        run:
+          /kaniko/executor --reproducible --snapshot-mode=redo --skip-unused-stages --cache=true
+                           --cache-repo 369495373322.dkr.ecr.eu-central-1.amazonaws.com/cache
+                           --context .
+                           --build-arg GIT_VERSION=${{ github.sha }}
+                           --destination 369495373322.dkr.ecr.eu-central-1.amazonaws.com/neon:${{needs.tag.outputs.build-tag}}
+                           --destination neondatabase/neon:${{needs.tag.outputs.build-tag}}
 
       # Cleanup script fails otherwise - rm: cannot remove '/nvme/actions-runner/_work/_temp/_github_home/.ecr': Permission denied
       - name: Cleanup ECR folder
@@ -652,7 +685,7 @@ jobs:
   compute-tools-image:
     runs-on: [ self-hosted, gen3, large ]
     needs: [ tag ]
-    container: gcr.io/kaniko-project/executor:v1.7.0-debug
+    container: gcr.io/kaniko-project/executor:v1.9.2-debug
     defaults:
       run:
         shell: sh -eu {0}
@@ -661,18 +694,41 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v1 # v3 won't work with kaniko
 
-      - name: Configure ECR login
-        run: echo "{\"credsStore\":\"ecr-login\"}" > /kaniko/.docker/config.json
+      - name: Configure ECR and Docker Hub login
+        run: |
+          DOCKERHUB_AUTH=$(echo -n "${{ secrets.NEON_DOCKERHUB_USERNAME }}:${{ secrets.NEON_DOCKERHUB_PASSWORD }}" | base64)
+          echo "::add-mask::${DOCKERHUB_AUTH}"
+
+          cat <<-EOF > /kaniko/.docker/config.json
+            {
+              "auths": {
+                "https://index.docker.io/v1/": {
+                  "auth": "${DOCKERHUB_AUTH}"
+                }
+              },
+              "credHelpers": {
+                "369495373322.dkr.ecr.eu-central-1.amazonaws.com": "ecr-login"
+              }
+            }
+          EOF
 
       - name: Kaniko build compute tools
-        run: /kaniko/executor --reproducible --snapshotMode=redo --skip-unused-stages --cache=true --cache-repo 369495373322.dkr.ecr.eu-central-1.amazonaws.com/cache --context . --build-arg GIT_VERSION=${{ github.sha }} --dockerfile Dockerfile.compute-tools --destination 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:${{needs.tag.outputs.build-tag}}
+        run:
+          /kaniko/executor --reproducible --snapshot-mode=redo --skip-unused-stages --cache=true
+                           --cache-repo 369495373322.dkr.ecr.eu-central-1.amazonaws.com/cache
+                           --context .
+                           --build-arg GIT_VERSION=${{ github.sha }}
+                           --dockerfile Dockerfile.compute-tools
+                           --destination 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:${{needs.tag.outputs.build-tag}}
+                           --destination neondatabase/compute-tools:${{needs.tag.outputs.build-tag}}
 
+      # Cleanup script fails otherwise - rm: cannot remove '/nvme/actions-runner/_work/_temp/_github_home/.ecr': Permission denied
       - name: Cleanup ECR folder
         run: rm -rf ~/.ecr
 
   compute-node-image:
     runs-on: [ self-hosted, gen3, large ]
-    container: gcr.io/kaniko-project/executor:v1.7.0-debug
+    container: gcr.io/kaniko-project/executor:v1.9.2-debug
     needs: [ tag ]
     strategy:
       fail-fast: false
@@ -689,12 +745,36 @@ jobs:
           submodules: true
           fetch-depth: 0
 
-      - name: Configure ECR login
-        run: echo "{\"credsStore\":\"ecr-login\"}" > /kaniko/.docker/config.json
+      - name: Configure ECR and Docker Hub login
+        run: |
+          DOCKERHUB_AUTH=$(echo -n "${{ secrets.NEON_DOCKERHUB_USERNAME }}:${{ secrets.NEON_DOCKERHUB_PASSWORD }}" | base64)
+          echo "::add-mask::${DOCKERHUB_AUTH}"
+
+          cat <<-EOF > /kaniko/.docker/config.json
+            {
+              "auths": {
+                "https://index.docker.io/v1/": {
+                  "auth": "${DOCKERHUB_AUTH}"
+                }
+              },
+              "credHelpers": {
+                "369495373322.dkr.ecr.eu-central-1.amazonaws.com": "ecr-login"
+              }
+            }
+          EOF
 
       - name: Kaniko build compute node with extensions
-        run: /kaniko/executor --reproducible --snapshotMode=redo --skip-unused-stages --cache=true --cache-repo 369495373322.dkr.ecr.eu-central-1.amazonaws.com/cache  --context . --build-arg GIT_VERSION=${{ github.sha }} --build-arg PG_VERSION=${{ matrix.version }} --dockerfile Dockerfile.compute-node --destination 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-${{ matrix.version }}:${{needs.tag.outputs.build-tag}}
+        run:
+          /kaniko/executor --reproducible --snapshot-mode=redo --skip-unused-stages --cache=true
+                           --cache-repo 369495373322.dkr.ecr.eu-central-1.amazonaws.com/cache
+                           --context .
+                           --build-arg GIT_VERSION=${{ github.sha }}
+                           --build-arg PG_VERSION=${{ matrix.version }}
+                           --dockerfile Dockerfile.compute-node
+                           --destination 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-${{ matrix.version }}:${{needs.tag.outputs.build-tag}}
+                           --destination neondatabase/compute-node-${{ matrix.version }}:${{needs.tag.outputs.build-tag}}
 
+      # Cleanup script fails otherwise - rm: cannot remove '/nvme/actions-runner/_work/_temp/_github_home/.ecr': Permission denied
       - name: Cleanup ECR folder
         run: rm -rf ~/.ecr
 
@@ -786,13 +866,11 @@ jobs:
     runs-on: [ self-hosted, gen3, small ]
     needs: [ tag, test-images, vm-compute-node-image ]
     container: golang:1.19-bullseye
-    if: github.event_name != 'workflow_dispatch'
+    # Don't add if-condition here.
+    # The job should always be run because we have dependant other jobs that shouldn't be skipped
 
     steps:
       - name: Install Crane & ECR helper
-        if: |
-          (github.ref_name == 'main' || github.ref_name == 'release') &&
-          github.event_name != 'workflow_dispatch'
         run: |
           go install github.com/google/go-containerregistry/cmd/crane@31786c6cbb82d6ec4fb8eb79cd9387905130534e # v0.11.0
           go install github.com/awslabs/amazon-ecr-credential-helper/ecr-login/cli/docker-credential-ecr-login@69c85dc22db6511932bbf119e1a0cc5c90c69a7f # v0.6.0
@@ -802,10 +880,15 @@ jobs:
           mkdir /github/home/.docker/
           echo "{\"credsStore\":\"ecr-login\"}" > /github/home/.docker/config.json
 
+      - name: Copy vm-compute-node images to Docker Hub
+        run: |
+          crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v14:${{needs.tag.outputs.build-tag}} vm-compute-node-v14
+          crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v15:${{needs.tag.outputs.build-tag}} vm-compute-node-v15
+
       - name: Add latest tag to images
         if: |
           (github.ref_name == 'main' || github.ref_name == 'release') &&
-          github.event_name != 'workflow_dispatch'
+           github.event_name != 'workflow_dispatch'
         run: |
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/neon:${{needs.tag.outputs.build-tag}} latest
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:${{needs.tag.outputs.build-tag}} latest
@@ -814,50 +897,10 @@ jobs:
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v15:${{needs.tag.outputs.build-tag}} latest
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v15:${{needs.tag.outputs.build-tag}} latest
 
-      - name: Cleanup ECR folder
-        run: rm -rf ~/.ecr
-
-  push-docker-hub:
-    runs-on: [ self-hosted, dev, x64 ]
-    needs: [ promote-images, tag ]
-    container: golang:1.19-bullseye
-
-    steps:
-      - name: Install Crane & ECR helper
-        run: |
-          go install github.com/google/go-containerregistry/cmd/crane@31786c6cbb82d6ec4fb8eb79cd9387905130534e # v0.11.0
-          go install github.com/awslabs/amazon-ecr-credential-helper/ecr-login/cli/docker-credential-ecr-login@69c85dc22db6511932bbf119e1a0cc5c90c69a7f # v0.6.0
-
-      - name: Configure ECR login
-        run: |
-          mkdir /github/home/.docker/
-          echo "{\"credsStore\":\"ecr-login\"}" > /github/home/.docker/config.json
-
-      - name: Pull neon image from ECR
-        run: crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/neon:${{needs.tag.outputs.build-tag}} neon
-
-      - name: Pull compute tools image from ECR
-        run: crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:${{needs.tag.outputs.build-tag}} compute-tools
-
-      - name: Pull compute node v14 image from ECR
-        run: crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v14:${{needs.tag.outputs.build-tag}} compute-node-v14
-
-      - name: Pull vm compute node v14 image from ECR
-        run: crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v14:${{needs.tag.outputs.build-tag}} vm-compute-node-v14
-
-      - name: Pull compute node v15 image from ECR
-        run: crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v15:${{needs.tag.outputs.build-tag}} compute-node-v15
-
-      - name: Pull vm compute node v15 image from ECR
-        run: crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v15:${{needs.tag.outputs.build-tag}} vm-compute-node-v15
-
-      - name: Pull rust image from ECR
-        run: crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned rust
-
       - name: Push images to production ECR
         if: |
           (github.ref_name == 'main' || github.ref_name == 'release') &&
-          github.event_name != 'workflow_dispatch'
+           github.event_name != 'workflow_dispatch'
         run: |
           crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/neon:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/neon:latest
           crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:latest
@@ -872,28 +915,12 @@ jobs:
           echo "" > /github/home/.docker/config.json
           crane auth login -u ${{ secrets.NEON_DOCKERHUB_USERNAME }} -p ${{ secrets.NEON_DOCKERHUB_PASSWORD }} index.docker.io
 
-      - name: Push neon image to Docker Hub
-        run: crane push neon neondatabase/neon:${{needs.tag.outputs.build-tag}}
+      - name: Push vm-compute-node to Docker Hub
+        run: |
+          crane push vm-compute-node-v14 neondatabase/vm-compute-node-v14:${{needs.tag.outputs.build-tag}}
+          crane push vm-compute-node-v15 neondatabase/vm-compute-node-v15:${{needs.tag.outputs.build-tag}}
 
-      - name: Push compute tools image to Docker Hub
-        run: crane push compute-tools neondatabase/compute-tools:${{needs.tag.outputs.build-tag}}
-
-      - name: Push compute node v14 image to Docker Hub
-        run: crane push compute-node-v14 neondatabase/compute-node-v14:${{needs.tag.outputs.build-tag}}
-
-      - name: Push vm compute node v14 image to Docker Hub
-        run: crane push vm-compute-node-v14 neondatabase/vm-compute-node-v14:${{needs.tag.outputs.build-tag}}
-
-      - name: Push compute node v15 image to Docker Hub
-        run: crane push compute-node-v15 neondatabase/compute-node-v15:${{needs.tag.outputs.build-tag}}
-
-      - name: Push vm compute node v15 image to Docker Hub
-        run: crane push vm-compute-node-v15 neondatabase/vm-compute-node-v15:${{needs.tag.outputs.build-tag}}
-
-      - name: Push rust image to Docker Hub
-        run: crane push rust neondatabase/rust:pinned
-
-      - name: Add latest tag to images in Docker Hub
+      - name: Push latest tags to Docker Hub
         if: |
           (github.ref_name == 'main' || github.ref_name == 'release') &&
           github.event_name != 'workflow_dispatch'
@@ -913,7 +940,7 @@ jobs:
     container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/ansible:pinned
     # We need both storage **and** compute images for deploy, because control plane picks the compute version based on the storage version.
     # If it notices a fresh storage it may bump the compute version. And if compute image failed to build it may break things badly
-    needs: [ push-docker-hub, tag, regress-tests ]
+    needs: [ promote-images, tag, regress-tests ]
     if: |
       contains(github.event.pull_request.labels.*.name, 'deploy-test-storage') &&
       github.event_name != 'workflow_dispatch'
@@ -947,7 +974,7 @@ jobs:
   deploy:
     runs-on: [ self-hosted, gen3, small ]
     container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/ansible:latest
-    needs: [ push-docker-hub, tag, regress-tests ]
+    needs: [ promote-images, tag, regress-tests ]
     if: ( github.ref_name == 'main' || github.ref_name == 'release' ) && github.event_name != 'workflow_dispatch'
     steps:
       - name: Fix git ownership
@@ -984,7 +1011,7 @@ jobs:
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
       options: --init
-    needs: [ push-docker-hub, tag, regress-tests ]
+    needs: [ promote-images, tag, regress-tests ]
     if: github.ref_name == 'release' && github.event_name != 'workflow_dispatch'
     steps:
       - name: Promote compatibility snapshot for the release
@@ -1007,7 +1034,7 @@ jobs:
 
             S3_KEY=$(aws s3api list-objects-v2 --bucket ${BUCKET} --prefix ${OLD_PREFIX} | jq -r '.Contents[].Key' | grep ${FILENAME} | sort --version-sort | tail -1 || true)
             if [ -z "${S3_KEY}" ]; then
-              echo 2>&1 "Neither s3://${BUCKET}/${OLD_PREFIX}/${FILENAME} nor its version from previous attempts exist"
+              echo >&2 "Neither s3://${BUCKET}/${OLD_PREFIX}/${FILENAME} nor its version from previous attempts exist"
               exit 1
             fi
 
diff --git a/.github/workflows/deploy-dev.yml b/.github/workflows/deploy-dev.yml
index b080a29f7c..5d1c6e0e16 100644
--- a/.github/workflows/deploy-dev.yml
+++ b/.github/workflows/deploy-dev.yml
@@ -48,7 +48,8 @@ jobs:
         shell: bash
     strategy:
       matrix:
-        target_region: [ eu-west-1, us-east-2 ]
+        # TODO(sergey): Fix storage deploy in eu-central-1
+        target_region: [ eu-west-1, us-east-2]
     environment:
       name: dev-${{ matrix.target_region }}
     steps:
@@ -133,6 +134,53 @@ jobs:
   
       - name: Cleanup helm folder
         run: rm -rf ~/.cache
+
+  deploy-preview-proxy-new:
+    runs-on: [ self-hosted, gen3, small ]
+    container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/ansible:pinned
+    if: inputs.deployProxy
+    defaults:
+      run:
+        shell: bash
+    strategy:
+      matrix:
+        include:
+          - target_region:  eu-central-1
+            target_cluster: dev-eu-central-1-alpha
+    environment:
+      name: dev-${{ matrix.target_region }}
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v3
+        with:
+          submodules: true
+          fetch-depth: 0
+          ref: ${{ inputs.branch }}
+  
+      - name: Configure AWS Credentials
+        uses: aws-actions/configure-aws-credentials@v1-node16
+        with:
+          role-to-assume: arn:aws:iam::369495373322:role/github-runner
+          aws-region: eu-central-1
+          role-skip-session-tagging: true
+          role-duration-seconds: 1800
+  
+      - name: Configure environment
+        run: |
+          helm repo add neondatabase https://neondatabase.github.io/helm-charts
+          aws --region ${{ matrix.target_region }} eks update-kubeconfig --name  ${{ matrix.target_cluster }}
+  
+      - name: Re-deploy preview proxies
+        run: |
+          DOCKER_TAG=${{ inputs.dockerTag }}
+          for PREVIEW_NAME in helium argon krypton xenon radon oganesson hydrogen nitrogen oxygen fluorine chlorine; do
+            export PREVIEW_NAME
+            envsubst <.github/helm-values/preview-template.neon-proxy-scram.yaml >preview-${PREVIEW_NAME}.neon-proxy-scram.yaml
+            helm upgrade neon-proxy-scram-${PREVIEW_NAME} neondatabase/neon-proxy --namespace neon-proxy-${PREVIEW_NAME} --create-namespace --install --atomic -f preview-${PREVIEW_NAME}.neon-proxy-scram.yaml --set image.tag=${DOCKER_TAG} --set settings.sentryUrl=${{ secrets.SENTRY_URL_PROXY }} --wait --timeout 15m0s
+          done
+
+      - name: Cleanup helm folder
+        run: rm -rf ~/.cache
   
   deploy-storage-broker-new:
     runs-on: [ self-hosted, gen3, small ]
@@ -148,6 +196,8 @@ jobs:
             target_cluster: dev-us-east-2-beta
           - target_region:  eu-west-1
             target_cluster: dev-eu-west-1-zeta
+          - target_region:  eu-central-1
+            target_cluster: dev-eu-central-1-alpha
     environment:
       name: dev-${{ matrix.target_region }}
     steps:
diff --git a/.github/workflows/deploy-prod.yml b/.github/workflows/deploy-prod.yml
index 6096ac8ab9..9fa31b3225 100644
--- a/.github/workflows/deploy-prod.yml
+++ b/.github/workflows/deploy-prod.yml
@@ -49,7 +49,7 @@ jobs:
         shell: bash
     strategy:
       matrix:
-        target_region: [ us-east-2, us-west-2, eu-central-1, ap-southeast-1 ]
+        target_region: [ us-east-2, us-west-2, eu-central-1, ap-southeast-1, us-east-1 ]
     environment:
       name: prod-${{ matrix.target_region }}
     steps:
@@ -97,6 +97,10 @@ jobs:
             target_cluster: prod-ap-southeast-1-epsilon
             deploy_link_proxy: false
             deploy_legacy_scram_proxy: false
+          - target_region: us-east-1
+            target_cluster: prod-us-east-1-theta
+            deploy_link_proxy: false
+            deploy_legacy_scram_proxy: false
     environment:
       name: prod-${{ matrix.target_region }}
     steps:
@@ -147,6 +151,8 @@ jobs:
             target_cluster: prod-eu-central-1-gamma
           - target_region: ap-southeast-1
             target_cluster: prod-ap-southeast-1-epsilon
+          - target_region: us-east-1
+            target_cluster: prod-us-east-1-theta
     environment:
       name: prod-${{ matrix.target_region }}
     steps:
diff --git a/.github/workflows/neon_extra_builds.yml b/.github/workflows/neon_extra_builds.yml
index ef4c293e31..1196881541 100644
--- a/.github/workflows/neon_extra_builds.yml
+++ b/.github/workflows/neon_extra_builds.yml
@@ -12,7 +12,7 @@ defaults:
 
 concurrency:
   # Allow only one workflow per any non-`main` branch.
-  group: ${{ github.workflow }}-${{ github.ref }}-${{ github.ref == 'refs/heads/main' && github.sha || 'anysha' }}
+  group: ${{ github.workflow }}-${{ github.ref_name }}-${{ github.ref_name == 'main' && github.sha || 'anysha' }}
   cancel-in-progress: true
 
 env:
diff --git a/.github/workflows/pg_clients.yml b/.github/workflows/pg_clients.yml
index 9f57519589..224b7b4a6d 100644
--- a/.github/workflows/pg_clients.yml
+++ b/.github/workflows/pg_clients.yml
@@ -14,7 +14,7 @@ on:
 
 concurrency:
   # Allow only one workflow per any non-`main` branch.
-  group: ${{ github.workflow }}-${{ github.ref }}-${{ github.ref == 'refs/heads/main' && github.sha || 'anysha' }}
+  group: ${{ github.workflow }}-${{ github.ref_name }}-${{ github.ref_name == 'main' && github.sha || 'anysha' }}
   cancel-in-progress: true
 
 jobs:
diff --git a/.neon_clippy_args b/.neon_clippy_args
new file mode 100644
index 0000000000..25e09c61a6
--- /dev/null
+++ b/.neon_clippy_args
@@ -0,0 +1,4 @@
+# * `-A unknown_lints` – do not warn about unknown lint suppressions
+#                        that people with newer toolchains might use
+# * `-D warnings`      - fail on any warnings (`cargo` returns non-zero exit status)
+export CLIPPY_COMMON_ARGS="--locked --workspace --all-targets -- -A unknown_lints -D warnings"
diff --git a/Cargo.lock b/Cargo.lock
index 5b99e93e76..5f3a83ce2d 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -64,28 +64,77 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4b46cbb362ab8752921c97e041f5e366ee6297bd428a31275b9fcf1e380f7299"
 
 [[package]]
-name = "anyhow"
-version = "1.0.68"
+name = "anstream"
+version = "0.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2cb2f989d18dd141ab8ae82f64d1a8cdd37e0840f73a406896cf5e99502fab61"
+checksum = "9e579a7752471abc2a8268df8b20005e3eadd975f585398f17efcfd8d4927371"
+dependencies = [
+ "anstyle",
+ "anstyle-parse",
+ "anstyle-query",
+ "anstyle-wincon",
+ "colorchoice",
+ "is-terminal",
+ "utf8parse",
+]
+
+[[package]]
+name = "anstyle"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "41ed9a86bf92ae6580e0a31281f65a1b1d867c0cc68d5346e2ae128dddfa6a7d"
+
+[[package]]
+name = "anstyle-parse"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e765fd216e48e067936442276d1d57399e37bce53c264d6fefbe298080cb57ee"
+dependencies = [
+ "utf8parse",
+]
+
+[[package]]
+name = "anstyle-query"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5ca11d4be1bab0c8bc8734a9aa7bf4ee8316d462a08c6ac5052f888fef5b494b"
+dependencies = [
+ "windows-sys 0.48.0",
+]
+
+[[package]]
+name = "anstyle-wincon"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4bcd8291a340dd8ac70e18878bc4501dd7b4ff970cfa21c207d36ece51ea88fd"
+dependencies = [
+ "anstyle",
+ "windows-sys 0.48.0",
+]
+
+[[package]]
+name = "anyhow"
+version = "1.0.70"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7de8ce5e0f9f8d88245311066a578d72b7af3e7088f32783804676302df237e4"
 dependencies = [
  "backtrace",
 ]
 
 [[package]]
 name = "archery"
-version = "0.4.0"
+version = "0.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0a8da9bc4c4053ee067669762bcaeea6e241841295a2b6c948312dad6ef4cc02"
+checksum = "b6cd774058b1b415c4855d8b86436c04bf050c003156fe24bc326fb3fe75c343"
 dependencies = [
  "static_assertions",
 ]
 
 [[package]]
 name = "asn1-rs"
-version = "0.5.1"
+version = "0.5.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "cf6690c370453db30743b373a60ba498fc0d6d83b11f4abfd87a84a075db5dd4"
+checksum = "7f6fd5ddaf0351dff5b8da21b2fb4ff8e08ddd02857f0bf69c47639106c0fff0"
 dependencies = [
  "asn1-rs-derive",
  "asn1-rs-impl",
@@ -105,7 +154,7 @@ checksum = "726535892e8eae7e70657b4c8ea93d26b8553afb1ce617caee529ef96d7dee6c"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
  "synstructure",
 ]
 
@@ -117,46 +166,47 @@ checksum = "2777730b2039ac0f95f093556e61b6d26cebed5393ca6f152717777cec3a42ed"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
 name = "async-stream"
-version = "0.3.3"
+version = "0.3.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dad5c83079eae9969be7fadefe640a1c566901f05ff91ab221de4b6f68d9507e"
+checksum = "cd56dd203fef61ac097dd65721a419ddccb106b2d2b70ba60a6b529f03961a51"
 dependencies = [
  "async-stream-impl",
  "futures-core",
+ "pin-project-lite",
 ]
 
 [[package]]
 name = "async-stream-impl"
-version = "0.3.3"
+version = "0.3.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "10f203db73a71dfa2fb6dd22763990fa26f3d2625a6da2da900d23b87d26be27"
+checksum = "16e62a023e7c117e27523144c5d2459f4397fcc3cab0085af8e2224f643a0193"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.15",
 ]
 
 [[package]]
 name = "async-trait"
-version = "0.1.64"
+version = "0.1.68"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1cd7fce9ba8c3c042128ce72d8b2ddbf3a05747efb67ea0313c635e10bda47a2"
+checksum = "b9ccdd8f2a161be9bd5c023df56f1b2a0bd1d83872ae53b71a84a12c9bf6e842"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.15",
 ]
 
 [[package]]
 name = "atomic-polyfill"
-version = "1.0.1"
+version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d299f547288d6db8d5c3a2916f7b2f66134b15b8c1ac1c4357dd3b8752af7bb2"
+checksum = "c314e70d181aa6053b26e3f7fbf86d1dfff84f816a6175b967666b3506ef7289"
 dependencies = [
  "critical-section",
 ]
@@ -187,13 +237,13 @@ dependencies = [
  "aws-http",
  "aws-sdk-sso",
  "aws-sdk-sts",
- "aws-smithy-async",
- "aws-smithy-client",
- "aws-smithy-http",
- "aws-smithy-http-tower",
+ "aws-smithy-async 0.51.0",
+ "aws-smithy-client 0.51.0",
+ "aws-smithy-http 0.51.0",
+ "aws-smithy-http-tower 0.51.0",
  "aws-smithy-json",
- "aws-smithy-types",
- "aws-types",
+ "aws-smithy-types 0.51.0",
+ "aws-types 0.51.0",
  "bytes",
  "hex",
  "http",
@@ -206,15 +256,29 @@ dependencies = [
  "zeroize",
 ]
 
+[[package]]
+name = "aws-credential-types"
+version = "0.55.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f4232d3729eefc287adc0d5a8adc97b7d94eefffe6bbe94312cc86c7ab6b06ce"
+dependencies = [
+ "aws-smithy-async 0.55.1",
+ "aws-smithy-types 0.55.1",
+ "fastrand",
+ "tokio",
+ "tracing",
+ "zeroize",
+]
+
 [[package]]
 name = "aws-endpoint"
 version = "0.51.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6ca8f374874f6459aaa88dc861d7f5d834ca1ff97668eae190e97266b5f6c3fb"
 dependencies = [
- "aws-smithy-http",
- "aws-smithy-types",
- "aws-types",
+ "aws-smithy-http 0.51.0",
+ "aws-smithy-types 0.51.0",
+ "aws-types 0.51.0",
  "http",
  "regex",
  "tracing",
@@ -226,9 +290,9 @@ version = "0.51.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "78d41e19e779b73463f5f0c21b3aacc995f4ba783ab13a7ae9f5dfb159a551b4"
 dependencies = [
- "aws-smithy-http",
- "aws-smithy-types",
- "aws-types",
+ "aws-smithy-http 0.51.0",
+ "aws-smithy-types 0.51.0",
+ "aws-types 0.51.0",
  "bytes",
  "http",
  "http-body",
@@ -248,15 +312,15 @@ dependencies = [
  "aws-http",
  "aws-sig-auth",
  "aws-sigv4",
- "aws-smithy-async",
+ "aws-smithy-async 0.51.0",
  "aws-smithy-checksums",
- "aws-smithy-client",
+ "aws-smithy-client 0.51.0",
  "aws-smithy-eventstream",
- "aws-smithy-http",
- "aws-smithy-http-tower",
- "aws-smithy-types",
+ "aws-smithy-http 0.51.0",
+ "aws-smithy-http-tower 0.51.0",
+ "aws-smithy-types 0.51.0",
  "aws-smithy-xml",
- "aws-types",
+ "aws-types 0.51.0",
  "bytes",
  "bytes-utils",
  "http",
@@ -275,13 +339,13 @@ dependencies = [
  "aws-endpoint",
  "aws-http",
  "aws-sig-auth",
- "aws-smithy-async",
- "aws-smithy-client",
- "aws-smithy-http",
- "aws-smithy-http-tower",
+ "aws-smithy-async 0.51.0",
+ "aws-smithy-client 0.51.0",
+ "aws-smithy-http 0.51.0",
+ "aws-smithy-http-tower 0.51.0",
  "aws-smithy-json",
- "aws-smithy-types",
- "aws-types",
+ "aws-smithy-types 0.51.0",
+ "aws-types 0.51.0",
  "bytes",
  "http",
  "tokio-stream",
@@ -297,14 +361,14 @@ dependencies = [
  "aws-endpoint",
  "aws-http",
  "aws-sig-auth",
- "aws-smithy-async",
- "aws-smithy-client",
- "aws-smithy-http",
- "aws-smithy-http-tower",
+ "aws-smithy-async 0.51.0",
+ "aws-smithy-client 0.51.0",
+ "aws-smithy-http 0.51.0",
+ "aws-smithy-http-tower 0.51.0",
  "aws-smithy-query",
- "aws-smithy-types",
+ "aws-smithy-types 0.51.0",
  "aws-smithy-xml",
- "aws-types",
+ "aws-types 0.51.0",
  "bytes",
  "http",
  "tower",
@@ -318,20 +382,20 @@ checksum = "12cbe7b2be9e185c1fbce27fc9c41c66b195b32d89aa099f98768d9544221308"
 dependencies = [
  "aws-sigv4",
  "aws-smithy-eventstream",
- "aws-smithy-http",
- "aws-types",
+ "aws-smithy-http 0.51.0",
+ "aws-types 0.51.0",
  "http",
  "tracing",
 ]
 
 [[package]]
 name = "aws-sigv4"
-version = "0.51.0"
+version = "0.51.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "03ff4cff8c4a101962d593ba94e72cd83891aecd423f0c6e3146bff6fb92c9e3"
+checksum = "5c0b2658d2cb66dbf02f0e8dee80810ef1e0ca3530ede463e0ef994c301087d1"
 dependencies = [
  "aws-smithy-eventstream",
- "aws-smithy-http",
+ "aws-smithy-http 0.51.0",
  "bytes",
  "form_urlencoded",
  "hex",
@@ -356,14 +420,26 @@ dependencies = [
  "tokio-stream",
 ]
 
+[[package]]
+name = "aws-smithy-async"
+version = "0.55.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "88573bcfbe1dcfd54d4912846df028b42d6255cbf9ce07be216b1bbfd11fc4b9"
+dependencies = [
+ "futures-util",
+ "pin-project-lite",
+ "tokio",
+ "tokio-stream",
+]
+
 [[package]]
 name = "aws-smithy-checksums"
 version = "0.51.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cc227e36e346f45298288359f37123e1a92628d1cec6b11b5eb335553278bd9e"
 dependencies = [
- "aws-smithy-http",
- "aws-smithy-types",
+ "aws-smithy-http 0.51.0",
+ "aws-smithy-types 0.51.0",
  "bytes",
  "crc32c",
  "crc32fast",
@@ -383,10 +459,10 @@ version = "0.51.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ff28d553714f8f54cd921227934fc13a536a1c03f106e56b362fd57e16d450ad"
 dependencies = [
- "aws-smithy-async",
- "aws-smithy-http",
- "aws-smithy-http-tower",
- "aws-smithy-types",
+ "aws-smithy-async 0.51.0",
+ "aws-smithy-http 0.51.0",
+ "aws-smithy-http-tower 0.51.0",
+ "aws-smithy-types 0.51.0",
  "bytes",
  "fastrand",
  "http",
@@ -400,13 +476,33 @@ dependencies = [
  "tracing",
 ]
 
+[[package]]
+name = "aws-smithy-client"
+version = "0.55.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b2f52352bae50d3337d5d6151b695d31a8c10ebea113eca5bead531f8301b067"
+dependencies = [
+ "aws-smithy-async 0.55.1",
+ "aws-smithy-http 0.55.1",
+ "aws-smithy-http-tower 0.55.1",
+ "aws-smithy-types 0.55.1",
+ "bytes",
+ "fastrand",
+ "http",
+ "http-body",
+ "pin-project-lite",
+ "tokio",
+ "tower",
+ "tracing",
+]
+
 [[package]]
 name = "aws-smithy-eventstream"
 version = "0.51.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d7ea0df7161ce65b5c8ca6eb709a1a907376fa18226976e41c748ce02ccccf24"
 dependencies = [
- "aws-smithy-types",
+ "aws-smithy-types 0.51.0",
  "bytes",
  "crc32fast",
 ]
@@ -418,7 +514,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bf58ed4fefa61dbf038e5421a521cbc2c448ef69deff0ab1d915d8a10eda5664"
 dependencies = [
  "aws-smithy-eventstream",
- "aws-smithy-types",
+ "aws-smithy-types 0.51.0",
  "bytes",
  "bytes-utils",
  "futures-core",
@@ -434,13 +530,49 @@ dependencies = [
  "tracing",
 ]
 
+[[package]]
+name = "aws-smithy-http"
+version = "0.55.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "03bcc02d7ed9649d855c8ce4a735e9848d7b8f7568aad0504c158e3baa955df8"
+dependencies = [
+ "aws-smithy-types 0.55.1",
+ "bytes",
+ "bytes-utils",
+ "futures-core",
+ "http",
+ "http-body",
+ "hyper",
+ "once_cell",
+ "percent-encoding",
+ "pin-project-lite",
+ "pin-utils",
+ "tracing",
+]
+
 [[package]]
 name = "aws-smithy-http-tower"
 version = "0.51.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "20c96d7bd35e7cf96aca1134b2f81b1b59ffe493f7c6539c051791cbbf7a42d3"
 dependencies = [
- "aws-smithy-http",
+ "aws-smithy-http 0.51.0",
+ "bytes",
+ "http",
+ "http-body",
+ "pin-project-lite",
+ "tower",
+ "tracing",
+]
+
+[[package]]
+name = "aws-smithy-http-tower"
+version = "0.55.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "da88b3a860f65505996c29192d800f1aeb9480440f56d63aad33a3c12045017a"
+dependencies = [
+ "aws-smithy-http 0.55.1",
+ "aws-smithy-types 0.55.1",
  "bytes",
  "http",
  "http-body",
@@ -455,7 +587,7 @@ version = "0.51.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d8324ba98c8a94187723cc16c37aefa09504646ee65c3d2c3af495bab5ea701b"
 dependencies = [
- "aws-smithy-types",
+ "aws-smithy-types 0.51.0",
 ]
 
 [[package]]
@@ -464,7 +596,7 @@ version = "0.51.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "83834ed2ff69ea6f6657baf205267dc2c0abe940703503a3e5d60ce23be3d306"
 dependencies = [
- "aws-smithy-types",
+ "aws-smithy-types 0.51.0",
  "urlencoding",
 ]
 
@@ -480,6 +612,19 @@ dependencies = [
  "time",
 ]
 
+[[package]]
+name = "aws-smithy-types"
+version = "0.55.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cd0afc731fd1417d791f9145a1e0c30e23ae0beaab9b4814017708ead2fc20f1"
+dependencies = [
+ "base64-simd",
+ "itoa",
+ "num-integer",
+ "ryu",
+ "time",
+]
+
 [[package]]
 name = "aws-smithy-xml"
 version = "0.51.0"
@@ -495,10 +640,10 @@ version = "0.51.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "05701d32da168b44f7ee63147781aed8723e792cc131cb9b18363b5393f17f70"
 dependencies = [
- "aws-smithy-async",
- "aws-smithy-client",
- "aws-smithy-http",
- "aws-smithy-types",
+ "aws-smithy-async 0.51.0",
+ "aws-smithy-client 0.51.0",
+ "aws-smithy-http 0.51.0",
+ "aws-smithy-types 0.51.0",
  "http",
  "rustc_version",
  "tracing",
@@ -506,10 +651,26 @@ dependencies = [
 ]
 
 [[package]]
-name = "axum"
-version = "0.6.4"
+name = "aws-types"
+version = "0.55.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e5694b64066a2459918d8074c2ce0d5a88f409431994c2356617c8ae0c4721fc"
+checksum = "b9b082e329d9a304d39e193ad5c7ab363a0d6507aca6965e0673a746686fb0cc"
+dependencies = [
+ "aws-credential-types",
+ "aws-smithy-async 0.55.1",
+ "aws-smithy-client 0.55.1",
+ "aws-smithy-http 0.55.1",
+ "aws-smithy-types 0.55.1",
+ "http",
+ "rustc_version",
+ "tracing",
+]
+
+[[package]]
+name = "axum"
+version = "0.6.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3b32c5ea3aabaf4deb5f5ced2d688ec0844c881c9e6c696a8b769a05fc691e62"
 dependencies = [
  "async-trait",
  "axum-core",
@@ -529,16 +690,15 @@ dependencies = [
  "serde",
  "sync_wrapper",
  "tower",
- "tower-http",
  "tower-layer",
  "tower-service",
 ]
 
 [[package]]
 name = "axum-core"
-version = "0.3.2"
+version = "0.3.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1cae3e661676ffbacb30f1a824089a8c9150e71017f7e1e38f2aa32009188d34"
+checksum = "759fa577a247914fd3f7f76d62972792636412fbfd634cd452f6a385a74d2d2c"
 dependencies = [
  "async-trait",
  "bytes",
@@ -584,6 +744,16 @@ version = "0.21.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a4a4ddaa51a5bc52a6948f74c06d20aaaddb71924eab79b8c97a8c556e942d6a"
 
+[[package]]
+name = "base64-simd"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "339abbe78e73178762e23bea9dfd08e697eb3f3301cd4be981c0f78ba5859195"
+dependencies = [
+ "outref",
+ "vsimd",
+]
+
 [[package]]
 name = "bincode"
 version = "1.3.3"
@@ -595,9 +765,9 @@ dependencies = [
 
 [[package]]
 name = "bindgen"
-version = "0.61.0"
+version = "0.65.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8a022e58a142a46fea340d68012b9201c094e93ec3d033a944a24f8fd4a4f09a"
+checksum = "cfdf7b466f9a4903edc73f95d6d2bcd5baf8ae620638762244d3f60143643cc5"
 dependencies = [
  "bitflags",
  "cexpr",
@@ -606,12 +776,13 @@ dependencies = [
  "lazycell",
  "log",
  "peeking_take_while",
+ "prettyplease 0.2.4",
  "proc-macro2",
  "quote",
  "regex",
  "rustc-hash",
  "shlex",
- "syn",
+ "syn 2.0.15",
  "which",
 ]
 
@@ -623,18 +794,18 @@ checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
 
 [[package]]
 name = "block-buffer"
-version = "0.10.3"
+version = "0.10.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "69cce20737498f97b993470a6e536b8523f0af7892a4f928cceb1ac5e52ebe7e"
+checksum = "3078c7629b62d3f0439517fa394996acacc5cbc91c5a20d8c658e77abd503a71"
 dependencies = [
  "generic-array",
 ]
 
 [[package]]
 name = "bstr"
-version = "1.2.0"
+version = "1.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b7f0778972c64420fdedc63f09919c8a88bda7b25135357fd25a5d9f3257e832"
+checksum = "c3d4260bcc2e8fc9df1eac4919a720effeb63a3f0952f5bf4944adfa18897f09"
 dependencies = [
  "memchr",
  "once_cell",
@@ -702,9 +873,9 @@ checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
 
 [[package]]
 name = "chrono"
-version = "0.4.23"
+version = "0.4.24"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "16b0a3d9ed01224b22057780a37bb8c5dbfe1be8ba48678e7bf57ec4b385411f"
+checksum = "4e3c5919066adf22df73762e50cffcde3a758f2a848b113b586d1f86728b673b"
 dependencies = [
  "iana-time-zone",
  "num-integer",
@@ -742,9 +913,9 @@ dependencies = [
 
 [[package]]
 name = "clang-sys"
-version = "1.4.0"
+version = "1.6.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fa2e27ae6ab525c3d369ded447057bca5438d86dc3a68f6faafb8269ba82ebf3"
+checksum = "c688fc74432808e3eb684cae8830a86be1d66a2bd58e1f248ed0960a590baf6f"
 dependencies = [
  "glob",
  "libc",
@@ -765,30 +936,38 @@ dependencies = [
 
 [[package]]
 name = "clap"
-version = "4.1.4"
+version = "4.2.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f13b9c79b5d1dd500d20ef541215a6423c75829ef43117e1b4d17fd8af0b5d76"
+checksum = "9b802d85aaf3a1cdb02b224ba472ebdea62014fccfcb269b95a4d76443b5ee5a"
 dependencies = [
- "bitflags",
+ "clap_builder",
  "clap_derive",
- "clap_lex 0.3.1",
- "is-terminal",
  "once_cell",
+]
+
+[[package]]
+name = "clap_builder"
+version = "4.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "14a1a858f532119338887a4b8e1af9c60de8249cd7bafd68036a489e261e37b6"
+dependencies = [
+ "anstream",
+ "anstyle",
+ "bitflags",
+ "clap_lex 0.4.1",
  "strsim",
- "termcolor",
 ]
 
 [[package]]
 name = "clap_derive"
-version = "4.1.0"
+version = "4.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "684a277d672e91966334af371f1a7b5833f9aa00b07c84e92fbce95e00208ce8"
+checksum = "3f9644cd56d6b87dbe899ef8b053e331c0637664e9e21a33dfcdc36093f5c5c4"
 dependencies = [
  "heck",
- "proc-macro-error",
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.15",
 ]
 
 [[package]]
@@ -802,12 +981,9 @@ dependencies = [
 
 [[package]]
 name = "clap_lex"
-version = "0.3.1"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "783fe232adfca04f90f56201b26d79682d4cd2625e0bc7290b95123afe558ade"
-dependencies = [
- "os_str_bytes",
-]
+checksum = "8a2dd5a6fe8c6e3502f568a6353e5273bbb15193ad9a89e457b9970798efbea1"
 
 [[package]]
 name = "close_fds"
@@ -829,6 +1005,12 @@ dependencies = [
  "unicode-width",
 ]
 
+[[package]]
+name = "colorchoice"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "acbf1af155f9b9ef647e42cdc158db4b64a1b61f743629225fde6f3e0be2a7c7"
+
 [[package]]
 name = "comfy-table"
 version = "6.1.4"
@@ -859,7 +1041,7 @@ version = "0.1.0"
 dependencies = [
  "anyhow",
  "chrono",
- "clap 4.1.4",
+ "clap 4.2.2",
  "compute_api",
  "futures",
  "hyper",
@@ -879,6 +1061,7 @@ dependencies = [
  "tracing-subscriber",
  "tracing-utils",
  "url",
+ "utils",
  "workspace_hack",
 ]
 
@@ -920,7 +1103,7 @@ name = "control_plane"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "clap 4.1.4",
+ "clap 4.2.2",
  "comfy-table",
  "git-version",
  "nix",
@@ -956,15 +1139,15 @@ dependencies = [
 
 [[package]]
 name = "core-foundation-sys"
-version = "0.8.3"
+version = "0.8.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5827cebf4670468b8772dd191856768aedcb1b0278a04f989f7766351917b9dc"
+checksum = "e496a50fda8aacccc86d7529e2c1e0892dbd0f898a6b5645b5561b89c3210efa"
 
 [[package]]
 name = "cpufeatures"
-version = "0.2.5"
+version = "0.2.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "28d997bd5e24a5928dd43e46dc529867e207907fe0b239c3477d924f7f2ca320"
+checksum = "280a9f2d8b3a38871a3c8a46fb80db65e5e5ed97da80c4d08bf27fb63e35e181"
 dependencies = [
  "libc",
 ]
@@ -1031,9 +1214,9 @@ checksum = "6548a0ad5d2549e111e1f6a11a6c2e2d00ce6a3dafe22948d67c2b443f775e52"
 
 [[package]]
 name = "crossbeam-channel"
-version = "0.5.6"
+version = "0.5.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c2dd04ddaf88237dc3b8d8f9a3c1004b506b54b3313403944054d23c0870c521"
+checksum = "a33c2bf77f2df06183c3aa30d1e96c0695a313d4f9c453cc3762a6db39f99200"
 dependencies = [
  "cfg-if",
  "crossbeam-utils",
@@ -1041,9 +1224,9 @@ dependencies = [
 
 [[package]]
 name = "crossbeam-deque"
-version = "0.8.2"
+version = "0.8.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "715e8152b692bba2d374b53d4875445368fdf21a94751410af607a5ac677d1fc"
+checksum = "ce6fd6f855243022dcecf8702fef0c297d4338e226845fe067f6341ad9fa0cef"
 dependencies = [
  "cfg-if",
  "crossbeam-epoch",
@@ -1052,22 +1235,22 @@ dependencies = [
 
 [[package]]
 name = "crossbeam-epoch"
-version = "0.9.13"
+version = "0.9.14"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "01a9af1f4c2ef74bb8aa1f7e19706bc72d03598c8a570bb5de72243c7a9d9d5a"
+checksum = "46bd5f3f85273295a9d14aedfb86f6aadbff6d8f5295c4a9edb08e819dcf5695"
 dependencies = [
  "autocfg",
  "cfg-if",
  "crossbeam-utils",
- "memoffset 0.7.1",
+ "memoffset 0.8.0",
  "scopeguard",
 ]
 
 [[package]]
 name = "crossbeam-utils"
-version = "0.8.14"
+version = "0.8.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4fb766fa798726286dbbb842f174001dab8abc7b627a1dd86e0b7222a95d929f"
+checksum = "3c063cd8cc95f5c377ed0d4b49a4b21f632396ff690e8470c29b3359b346984b"
 dependencies = [
  "cfg-if",
 ]
@@ -1109,9 +1292,9 @@ dependencies = [
 
 [[package]]
 name = "cxx"
-version = "1.0.89"
+version = "1.0.94"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bc831ee6a32dd495436e317595e639a587aa9907bef96fe6e6abc290ab6204e9"
+checksum = "f61f1b6389c3fe1c316bf8a4dccc90a38208354b330925bce1f74a6c4756eb93"
 dependencies = [
  "cc",
  "cxxbridge-flags",
@@ -1121,9 +1304,9 @@ dependencies = [
 
 [[package]]
 name = "cxx-build"
-version = "1.0.89"
+version = "1.0.94"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "94331d54f1b1a8895cd81049f7eaaaef9d05a7dcb4d1fd08bf3ff0806246789d"
+checksum = "12cee708e8962df2aeb38f594aae5d827c022b6460ac71a7a3e2c3c2aae5a07b"
 dependencies = [
  "cc",
  "codespan-reporting",
@@ -1131,31 +1314,31 @@ dependencies = [
  "proc-macro2",
  "quote",
  "scratch",
- "syn",
+ "syn 2.0.15",
 ]
 
 [[package]]
 name = "cxxbridge-flags"
-version = "1.0.89"
+version = "1.0.94"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "48dcd35ba14ca9b40d6e4b4b39961f23d835dbb8eed74565ded361d93e1feb8a"
+checksum = "7944172ae7e4068c533afbb984114a56c46e9ccddda550499caa222902c7f7bb"
 
 [[package]]
 name = "cxxbridge-macro"
-version = "1.0.89"
+version = "1.0.94"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "81bbeb29798b407ccd82a3324ade1a7286e0d29851475990b612670f6f5124d2"
+checksum = "2345488264226bf682893e25de0769f3360aac9957980ec49361b083ddaa5bc5"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.15",
 ]
 
 [[package]]
 name = "darling"
-version = "0.14.2"
+version = "0.14.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b0dd3cd20dc6b5a876612a6e5accfe7f3dd883db6d07acfbf14c128f61550dfa"
+checksum = "7b750cb3417fd1b327431a470f388520309479ab0bf5e323505daf0290cd3850"
 dependencies = [
  "darling_core",
  "darling_macro",
@@ -1163,27 +1346,27 @@ dependencies = [
 
 [[package]]
 name = "darling_core"
-version = "0.14.2"
+version = "0.14.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a784d2ccaf7c98501746bf0be29b2022ba41fd62a2e622af997a03e9f972859f"
+checksum = "109c1ca6e6b7f82cc233a97004ea8ed7ca123a9af07a8230878fcfda9b158bf0"
 dependencies = [
  "fnv",
  "ident_case",
  "proc-macro2",
  "quote",
  "strsim",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
 name = "darling_macro"
-version = "0.14.2"
+version = "0.14.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7618812407e9402654622dd402b0a89dff9ba93badd6540781526117b92aab7e"
+checksum = "a4aab4dbc9f7611d8b55048a3a16d2d010c2c8334e46304b40ac1cc14bf3b48e"
 dependencies = [
  "darling_core",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -1217,9 +1400,9 @@ dependencies = [
 
 [[package]]
 name = "der-parser"
-version = "8.1.0"
+version = "8.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "42d4bc9b0db0a0df9ae64634ac5bdefb7afcb534e182275ca0beadbe486701c1"
+checksum = "dbd676fbbab537128ef0278adb5576cf363cff6aa22a7b24effe97347cfab61e"
 dependencies = [
  "asn1-rs",
  "displaydoc",
@@ -1248,7 +1431,7 @@ checksum = "3bf95dc3f046b9da4f2d51833c0d3547d8564ef6910f5c1ed130306a75b92886"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -1268,9 +1451,9 @@ dependencies = [
 
 [[package]]
 name = "enum-map"
-version = "2.4.2"
+version = "2.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "50c25992259941eb7e57b936157961b217a4fc8597829ddef0596d6c3cd86e1a"
+checksum = "988f0d17a0fa38291e5f41f71ea8d46a5d5497b9054d5a759fae2cbb819f2356"
 dependencies = [
  "enum-map-derive",
 ]
@@ -1283,7 +1466,7 @@ checksum = "2a4da76b3b6116d758c7ba93f7ec6a35d2e2cf24feda76c6e38a375f4d5c59f2"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -1304,7 +1487,7 @@ dependencies = [
  "darling",
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -1322,13 +1505,13 @@ dependencies = [
 
 [[package]]
 name = "errno"
-version = "0.2.8"
+version = "0.3.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f639046355ee4f37944e44f60642c6f3a7efa3cf6b78c78a0d989a8ce6c396a1"
+checksum = "4bcfec3a70f97c962c307b2d2c56e358cf1d00b558d74262b5f929ee8cc7e73a"
 dependencies = [
  "errno-dragonfly",
  "libc",
- "winapi",
+ "windows-sys 0.48.0",
 ]
 
 [[package]]
@@ -1360,23 +1543,23 @@ checksum = "4443176a9f2c162692bd3d352d745ef9413eec5782a80d8fd6f8a1ac692a07f7"
 
 [[package]]
 name = "fastrand"
-version = "1.8.0"
+version = "1.9.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a7a407cfaa3385c4ae6b23e84623d48c2798d06e3e6a1878f7f59f17b3f86499"
+checksum = "e51093e27b0797c359783294ca4f0a911c270184cb10f85783b118614a1501be"
 dependencies = [
  "instant",
 ]
 
 [[package]]
 name = "filetime"
-version = "0.2.19"
+version = "0.2.21"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4e884668cd0c7480504233e951174ddc3b382f7c2666e3b7310b5c4e7b0c37f9"
+checksum = "5cbc844cecaee9d4443931972e1289c8ff485cb4cc2767cb03ca139ed6885153"
 dependencies = [
  "cfg-if",
  "libc",
- "redox_syscall",
- "windows-sys 0.42.0",
+ "redox_syscall 0.2.16",
+ "windows-sys 0.48.0",
 ]
 
 [[package]]
@@ -1421,9 +1604,9 @@ dependencies = [
 
 [[package]]
 name = "futures"
-version = "0.3.26"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "13e2792b0ff0340399d58445b88fd9770e3489eff258a4cbc1523418f12abf84"
+checksum = "23342abe12aba583913b2e62f22225ff9c950774065e4bfb61a19cd9770fec40"
 dependencies = [
  "futures-channel",
  "futures-core",
@@ -1436,9 +1619,9 @@ dependencies = [
 
 [[package]]
 name = "futures-channel"
-version = "0.3.26"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2e5317663a9089767a1ec00a487df42e0ca174b61b4483213ac24448e4664df5"
+checksum = "955518d47e09b25bbebc7a18df10b81f0c766eaf4c4f1cccef2fca5f2a4fb5f2"
 dependencies = [
  "futures-core",
  "futures-sink",
@@ -1446,15 +1629,15 @@ dependencies = [
 
 [[package]]
 name = "futures-core"
-version = "0.3.26"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ec90ff4d0fe1f57d600049061dc6bb68ed03c7d2fbd697274c41805dcb3f8608"
+checksum = "4bca583b7e26f571124fe5b7561d49cb2868d79116cfa0eefce955557c6fee8c"
 
 [[package]]
 name = "futures-executor"
-version = "0.3.26"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e8de0a35a6ab97ec8869e32a2473f4b1324459e14c29275d14b10cb1fd19b50e"
+checksum = "ccecee823288125bd88b4d7f565c9e58e41858e47ab72e8ea2d64e93624386e0"
 dependencies = [
  "futures-core",
  "futures-task",
@@ -1463,32 +1646,32 @@ dependencies = [
 
 [[package]]
 name = "futures-io"
-version = "0.3.26"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bfb8371b6fb2aeb2d280374607aeabfc99d95c72edfe51692e42d3d7f0d08531"
+checksum = "4fff74096e71ed47f8e023204cfd0aa1289cd54ae5430a9523be060cdb849964"
 
 [[package]]
 name = "futures-macro"
-version = "0.3.26"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "95a73af87da33b5acf53acfebdc339fe592ecf5357ac7c0a7734ab9d8c876a70"
+checksum = "89ca545a94061b6365f2c7355b4b32bd20df3ff95f02da9329b34ccc3bd6ee72"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.15",
 ]
 
 [[package]]
 name = "futures-sink"
-version = "0.3.26"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f310820bb3e8cfd46c80db4d7fb8353e15dfff853a127158425f31e0be6c8364"
+checksum = "f43be4fe21a13b9781a69afa4985b0f6ee0e1afab2c6f454a8cf30e2b2237b6e"
 
 [[package]]
 name = "futures-task"
-version = "0.3.26"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dcf79a1bf610b10f42aea489289c5a2c478a786509693b80cd39c44ccd936366"
+checksum = "76d3d132be6c0e6aa1534069c705a74a5997a356c0dc2f86a47765e5617c5b65"
 
 [[package]]
 name = "futures-timer"
@@ -1498,9 +1681,9 @@ checksum = "e64b03909df88034c26dc1547e8970b91f98bdb65165d6a4e9110d94263dbb2c"
 
 [[package]]
 name = "futures-util"
-version = "0.3.26"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9c1d6de3acfef38d2be4b1f543f553131788603495be83da675e180c8d6b7bd1"
+checksum = "26b01e40b772d54cf6c6d721c1d1abd0647a0106a12ecaa1c186273392a69533"
 dependencies = [
  "futures-channel",
  "futures-core",
@@ -1516,9 +1699,9 @@ dependencies = [
 
 [[package]]
 name = "generic-array"
-version = "0.14.6"
+version = "0.14.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bff49e947297f3312447abdca79f45f4738097cc82b06e72054d2223f601f1b9"
+checksum = "85649ca51fd72272d7821adaf274ad91c288277713d9c18820d8499a7ff69e9a"
 dependencies = [
  "typenum",
  "version_check",
@@ -1526,20 +1709,22 @@ dependencies = [
 
 [[package]]
 name = "getrandom"
-version = "0.2.8"
+version = "0.2.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c05aeb6a22b8f62540c194aac980f2115af067bfe15a0734d7277a768d396b31"
+checksum = "c85e1d9ab2eadba7e5040d4e09cbd6d072b76a557ad64e797c2cb9d4da21d7e4"
 dependencies = [
  "cfg-if",
+ "js-sys",
  "libc",
  "wasi",
+ "wasm-bindgen",
 ]
 
 [[package]]
 name = "gimli"
-version = "0.27.1"
+version = "0.27.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "221996f774192f0f718773def8201c4ae31f02616a54ccfc2d358bb0e5cefdec"
+checksum = "ad0a93d233ebf96623465aad4046a8d3aa4da22d4f4beba5388838c8a434bbb4"
 
 [[package]]
 name = "git-version"
@@ -1560,7 +1745,7 @@ dependencies = [
  "proc-macro-hack",
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -1571,9 +1756,9 @@ checksum = "d2fabcfbdc87f4758337ca535fb41a6d701b65693ce38287d856d1674551ec9b"
 
 [[package]]
 name = "h2"
-version = "0.3.15"
+version = "0.3.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5f9f29bc9dda355256b2916cf526ab02ce0aeaaaf2bad60d65ef3f12f11dd0f4"
+checksum = "17f8a914c2987b688368b5138aa05321db91f4090cf26118185672ad588bce21"
 dependencies = [
  "bytes",
  "fnv",
@@ -1638,7 +1823,7 @@ dependencies = [
  "atomic-polyfill",
  "hash32",
  "rustc_version",
- "spin 0.9.4",
+ "spin 0.9.8",
  "stable_deref_trait",
 ]
 
@@ -1666,6 +1851,12 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "hermit-abi"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fed44880c466736ef9a5c5b5facefb5ed0785676d0c02d612db14e54f0d84286"
+
 [[package]]
 name = "hex"
 version = "0.4.3"
@@ -1677,9 +1868,9 @@ dependencies = [
 
 [[package]]
 name = "hex-literal"
-version = "0.3.4"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7ebdb29d2ea9ed0083cd8cece49bbd968021bd99b0849edb4a9a7ee0fdf6a4e0"
+checksum = "6fe2267d4ed49bc07b63801559be28c718ea06c4738b7a03c94df7386d2cde46"
 
 [[package]]
 name = "hmac"
@@ -1703,9 +1894,9 @@ dependencies = [
 
 [[package]]
 name = "http"
-version = "0.2.8"
+version = "0.2.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "75f43d41e26995c17e71ee126451dd3941010b0514a81a9d11f3b341debc2399"
+checksum = "bd6effc99afb63425aff9b05836f029929e345a6148a14b7ecd5ab67af944482"
 dependencies = [
  "bytes",
  "fnv",
@@ -1723,12 +1914,6 @@ dependencies = [
  "pin-project-lite",
 ]
 
-[[package]]
-name = "http-range-header"
-version = "0.3.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0bfe8eed0a9285ef776bb792479ea3834e8b94e13d615c2f66d03dd50a435a29"
-
 [[package]]
 name = "httparse"
 version = "1.8.0"
@@ -1759,9 +1944,9 @@ dependencies = [
 
 [[package]]
 name = "hyper"
-version = "0.14.23"
+version = "0.14.26"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "034711faac9d2166cb1baf1a2fb0b60b1f277f8492fd72176c17f3515e1abd3c"
+checksum = "ab302d72a6f11a3b910431ff93aae7e773078c769f0a3ef15fb9ec692ed147d4"
 dependencies = [
  "bytes",
  "futures-channel",
@@ -1774,7 +1959,7 @@ dependencies = [
  "httpdate",
  "itoa",
  "pin-project-lite",
- "socket2",
+ "socket2 0.4.9",
  "tokio",
  "tower-service",
  "tracing",
@@ -1790,10 +1975,10 @@ dependencies = [
  "http",
  "hyper",
  "log",
- "rustls",
+ "rustls 0.20.8",
  "rustls-native-certs",
  "tokio",
- "tokio-rustls",
+ "tokio-rustls 0.23.4",
 ]
 
 [[package]]
@@ -1823,16 +2008,16 @@ dependencies = [
 
 [[package]]
 name = "iana-time-zone"
-version = "0.1.53"
+version = "0.1.56"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "64c122667b287044802d6ce17ee2ddf13207ed924c712de9a66a5814d5b64765"
+checksum = "0722cd7114b7de04316e7ea5456a0bbb20e4adb46fd27a3697adb812cff0f37c"
 dependencies = [
  "android_system_properties",
  "core-foundation-sys",
  "iana-time-zone-haiku",
  "js-sys",
  "wasm-bindgen",
- "winapi",
+ "windows",
 ]
 
 [[package]]
@@ -1863,9 +2048,9 @@ dependencies = [
 
 [[package]]
 name = "indexmap"
-version = "1.9.2"
+version = "1.9.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1885e79c1fc4b10f0e172c475f458b7f7b93061064d98c3293e98c5ba0c8b399"
+checksum = "bd070e393353796e801d209ad339e89596eb4c8d430d18ede6a1cced8fafbd99"
 dependencies = [
  "autocfg",
  "hashbrown 0.12.3",
@@ -1903,30 +2088,31 @@ dependencies = [
 
 [[package]]
 name = "io-lifetimes"
-version = "1.0.4"
+version = "1.0.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e7d6c6f8c91b4b9ed43484ad1a938e393caf35960fce7f82a040497207bd8e9e"
+checksum = "9c66c74d2ae7e79a5a8f7ac924adbe38ee42a859c6539ad869eb51f0b52dc220"
 dependencies = [
+ "hermit-abi 0.3.1",
  "libc",
- "windows-sys 0.42.0",
+ "windows-sys 0.48.0",
 ]
 
 [[package]]
 name = "ipnet"
-version = "2.7.1"
+version = "2.7.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "30e22bd8629359895450b59ea7a776c850561b96a3b1d31321c1949d9e6c9146"
+checksum = "12b6ee2129af8d4fb011108c73d99a1b83a85977f23b82460c0ae2e25bb4b57f"
 
 [[package]]
 name = "is-terminal"
-version = "0.4.2"
+version = "0.4.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "28dfb6c8100ccc63462345b67d1bbc3679177c75ee4bf59bf29c8b1d110b8189"
+checksum = "adcf93614601c8129ddf72e2d5633df827ba6551541c6d8c59520a371475be1f"
 dependencies = [
- "hermit-abi 0.2.6",
+ "hermit-abi 0.3.1",
  "io-lifetimes",
- "rustix",
- "windows-sys 0.42.0",
+ "rustix 0.37.11",
+ "windows-sys 0.48.0",
 ]
 
 [[package]]
@@ -1940,9 +2126,9 @@ dependencies = [
 
 [[package]]
 name = "itoa"
-version = "1.0.5"
+version = "1.0.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fad582f4b9e86b6caa621cabeb0963332d92eea04729ab12892c2533951e6440"
+checksum = "453ad9f582a441959e5f0d088b02ce04cfe8d51a8eaf077f12ac6d3e94164ca6"
 
 [[package]]
 name = "js-sys"
@@ -1955,11 +2141,11 @@ dependencies = [
 
 [[package]]
 name = "jsonwebtoken"
-version = "8.2.0"
+version = "8.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "09f4f04699947111ec1733e71778d763555737579e44b85844cae8e1940a1828"
+checksum = "6971da4d9c3aa03c3d8f3ff0f4155b534aad021292003895a469716b2a230378"
 dependencies = [
- "base64 0.13.1",
+ "base64 0.21.0",
  "pem",
  "ring",
  "serde",
@@ -2001,9 +2187,9 @@ checksum = "830d08ce1d1d941e6b30645f1a0eb5643013d835ce3779a5fc208261dbe10f55"
 
 [[package]]
 name = "libc"
-version = "0.2.139"
+version = "0.2.141"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "201de327520df007757c1f0adce6e827fe8562fbc28bfd9c15571c66ca1f5f79"
+checksum = "3304a64d199bb964be99741b7a14d26972741915b3649639149b2479bb46f4b5"
 
 [[package]]
 name = "libloading"
@@ -2030,6 +2216,12 @@ version = "0.1.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f051f77a7c8e6957c0696eac88f26b0117e54f52d3fc682ab19397a8812846a4"
 
+[[package]]
+name = "linux-raw-sys"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d59d8c75012853d2e872fb56bc8a2e53718e2cafe1a4c823143141c6d90c322f"
+
 [[package]]
 name = "lock_api"
 version = "0.4.9"
@@ -2122,9 +2314,9 @@ dependencies = [
 
 [[package]]
 name = "mime"
-version = "0.3.16"
+version = "0.3.17"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2a60c7ce501c71e03a9c9c0d35b861413ae925bd979cc7a4e30d060069aaac8d"
+checksum = "6877bb514081ee2a7ff5ef9de3281f14a4dd4bceac4c09388074a6b5df8a139a"
 
 [[package]]
 name = "mime_guess"
@@ -2144,23 +2336,23 @@ checksum = "68354c5c6bd36d73ff3feceb05efa59b6acb7626617f4962be322a825e61f79a"
 
 [[package]]
 name = "miniz_oxide"
-version = "0.6.4"
+version = "0.6.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f2e212582ede878b109755efd0773a4f0f4ec851584cf0aefbeb4d9ecc114822"
+checksum = "b275950c28b37e794e8c55d88aeb5e139d0ce23fdbbeda68f8d7174abdf9e8fa"
 dependencies = [
  "adler",
 ]
 
 [[package]]
 name = "mio"
-version = "0.8.5"
+version = "0.8.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e5d732bc30207a6423068df043e3d02e0735b155ad7ce1a6f76fe2baa5b158de"
+checksum = "5b9d9a46eff5b4ff64b45a9e316a6d1e0bc719ef429cbec4dc630684212bfdf9"
 dependencies = [
  "libc",
  "log",
  "wasi",
- "windows-sys 0.42.0",
+ "windows-sys 0.45.0",
 ]
 
 [[package]]
@@ -2193,15 +2385,6 @@ dependencies = [
  "minimal-lexical",
 ]
 
-[[package]]
-name = "nom8"
-version = "0.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ae01545c9c7fc4486ab7debaf2aad7003ac19431791868fb2e8066df97fad2f8"
-dependencies = [
- "memchr",
-]
-
 [[package]]
 name = "notify"
 version = "5.1.0"
@@ -2290,9 +2473,9 @@ dependencies = [
 
 [[package]]
 name = "once_cell"
-version = "1.17.0"
+version = "1.17.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6f61fba1741ea2b3d6a1e3178721804bb716a68a6aeba1149b5d52e3d464ea66"
+checksum = "b7e5500299e16ebb147ae15a00a942af264cf3688f47923b8fc2cd5858f23ad3"
 
 [[package]]
 name = "oorandom"
@@ -2357,8 +2540,8 @@ dependencies = [
  "futures-util",
  "opentelemetry",
  "prost",
- "tonic",
- "tonic-build",
+ "tonic 0.8.3",
+ "tonic-build 0.8.4",
 ]
 
 [[package]]
@@ -2410,9 +2593,9 @@ dependencies = [
 
 [[package]]
 name = "os_info"
-version = "3.6.0"
+version = "3.7.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5c424bc68d15e0778838ac013b5b3449544d8133633d8016319e7e05a820b8c0"
+checksum = "006e42d5b888366f1880eda20371fedde764ed2213dc8496f49622fa0c99cd5e"
 dependencies = [
  "log",
  "serde",
@@ -2421,9 +2604,15 @@ dependencies = [
 
 [[package]]
 name = "os_str_bytes"
-version = "6.4.1"
+version = "6.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9b7820b9daea5457c9f21c69448905d723fbd21136ccf521748f23fd49e723ee"
+checksum = "ceedf44fb00f2d1984b0bc98102627ce622e083e49a5bacdb3e514fa4238e267"
+
+[[package]]
+name = "outref"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4030760ffd992bef45b0ae3f10ce1aba99e33464c90d14dd7c039884963ddc7a"
 
 [[package]]
 name = "overload"
@@ -2441,7 +2630,7 @@ dependencies = [
  "byteorder",
  "bytes",
  "chrono",
- "clap 4.1.4",
+ "clap 4.2.2",
  "close_fds",
  "const_format",
  "consumption_metrics",
@@ -2492,6 +2681,7 @@ dependencies = [
  "tenant_size_model",
  "thiserror",
  "tokio",
+ "tokio-io-timeout",
  "tokio-postgres",
  "tokio-tar",
  "tokio-util",
@@ -2516,6 +2706,8 @@ dependencies = [
  "serde",
  "serde_json",
  "serde_with",
+ "strum",
+ "strum_macros",
  "utils",
  "workspace_hack",
 ]
@@ -2538,7 +2730,7 @@ checksum = "9069cbb9f99e3a5083476ccb29ceb1de18b9118cafa53e90c9551235de2b9521"
 dependencies = [
  "cfg-if",
  "libc",
- "redox_syscall",
+ "redox_syscall 0.2.16",
  "smallvec",
  "windows-sys 0.45.0",
 ]
@@ -2566,9 +2758,9 @@ checksum = "478c572c3d73181ff3c2539045f6eb99e5491218eae919370993b890cdbdd98e"
 
 [[package]]
 name = "petgraph"
-version = "0.6.2"
+version = "0.6.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e6d5014253a1331579ce62aa67443b4a658c5e7dd03d4bc6d302b94474888143"
+checksum = "4dd7d28ee937e54fe3080c91faa1c3a46c06de6252988a7f4592ba2310ef22a4"
 dependencies = [
  "fixedbitset",
  "indexmap",
@@ -2609,7 +2801,7 @@ checksum = "069bdb1e05adc7a8990dce9cc75370895fbe4e3d58b9b73bf1aee56359344a55"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -2703,14 +2895,14 @@ dependencies = [
  "futures",
  "once_cell",
  "pq_proto",
- "rustls",
+ "rustls 0.20.8",
  "rustls-pemfile",
  "serde",
  "thiserror",
  "tokio",
  "tokio-postgres",
  "tokio-postgres-rustls",
- "tokio-rustls",
+ "tokio-rustls 0.23.4",
  "tracing",
  "workspace_hack",
 ]
@@ -2767,7 +2959,6 @@ dependencies = [
  "pin-project-lite",
  "postgres-protocol",
  "rand",
- "serde",
  "thiserror",
  "tokio",
  "tracing",
@@ -2776,36 +2967,22 @@ dependencies = [
 
 [[package]]
 name = "prettyplease"
-version = "0.1.23"
+version = "0.1.25"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e97e3215779627f01ee256d2fad52f3d95e8e1c11e9fc6fd08f7cd455d5d5c78"
+checksum = "6c8646e95016a7a6c4adea95bafa8a16baab64b583356217f2c85db4a39d9a86"
 dependencies = [
  "proc-macro2",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
-name = "proc-macro-error"
-version = "1.0.4"
+name = "prettyplease"
+version = "0.2.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "da25490ff9892aab3fcf7c36f08cfb902dd3e71ca0f9f9517bea02a73a5ce38c"
-dependencies = [
- "proc-macro-error-attr",
- "proc-macro2",
- "quote",
- "syn",
- "version_check",
-]
-
-[[package]]
-name = "proc-macro-error-attr"
-version = "1.0.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a1be40180e52ecc98ad80b184934baf3d0d29f979574e439af5a55274b35f869"
+checksum = "1ceca8aaf45b5c46ec7ed39fff75f57290368c1846d33d24a122ca81416ab058"
 dependencies = [
  "proc-macro2",
- "quote",
- "version_check",
+ "syn 2.0.15",
 ]
 
 [[package]]
@@ -2816,9 +2993,9 @@ checksum = "dc375e1527247fe1a97d8b7156678dfe7c1af2fc075c9a4db3690ecd2a148068"
 
 [[package]]
 name = "proc-macro2"
-version = "1.0.50"
+version = "1.0.56"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ef7d57beacfaf2d8aee5937dab7b7f28de3cb8b1828479bb5de2a7106f2bae2"
+checksum = "2b63bdb0cd06f1f4dedf69b254734f9b45af66e4a031e42a7480257d9898b435"
 dependencies = [
  "unicode-ident",
 ]
@@ -2833,7 +3010,7 @@ dependencies = [
  "byteorder",
  "hex",
  "lazy_static",
- "rustix",
+ "rustix 0.36.12",
 ]
 
 [[package]]
@@ -2854,9 +3031,9 @@ dependencies = [
 
 [[package]]
 name = "prost"
-version = "0.11.6"
+version = "0.11.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "21dc42e00223fc37204bd4aa177e69420c604ca4a183209a8f9de30c6d934698"
+checksum = "0b82eaa1d779e9a4bc1c3217db8ffbeabaae1dca241bf70183242128d48681cd"
 dependencies = [
  "bytes",
  "prost-derive",
@@ -2864,9 +3041,9 @@ dependencies = [
 
 [[package]]
 name = "prost-build"
-version = "0.11.6"
+version = "0.11.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a3f8ad728fb08fe212df3c05169e940fbb6d9d16a877ddde14644a983ba2012e"
+checksum = "119533552c9a7ffacc21e099c24a0ac8bb19c2a2a3f363de84cd9b844feab270"
 dependencies = [
  "bytes",
  "heck",
@@ -2875,35 +3052,34 @@ dependencies = [
  "log",
  "multimap",
  "petgraph",
- "prettyplease",
+ "prettyplease 0.1.25",
  "prost",
  "prost-types",
  "regex",
- "syn",
+ "syn 1.0.109",
  "tempfile",
  "which",
 ]
 
 [[package]]
 name = "prost-derive"
-version = "0.11.6"
+version = "0.11.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8bda8c0881ea9f722eb9629376db3d0b903b462477c1aafcb0566610ac28ac5d"
+checksum = "e5d2d8d10f3c6ded6da8b05b5fb3b8a5082514344d56c9f871412d29b4e075b4"
 dependencies = [
  "anyhow",
  "itertools",
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
 name = "prost-types"
-version = "0.11.6"
+version = "0.11.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a5e0526209433e96d83d750dd81a99118edbc55739e7e61a46764fd2ad537788"
+checksum = "213622a1460818959ac1181aaeb2dc9c7f63df720db7d788b3e24eacd1983e13"
 dependencies = [
- "bytes",
  "prost",
 ]
 
@@ -2918,7 +3094,7 @@ dependencies = [
  "bstr",
  "bytes",
  "chrono",
- "clap 4.1.4",
+ "clap 4.2.2",
  "consumption_metrics",
  "futures",
  "git-version",
@@ -2948,20 +3124,21 @@ dependencies = [
  "reqwest-tracing",
  "routerify",
  "rstest",
- "rustls",
+ "rustls 0.20.8",
  "rustls-pemfile",
  "scopeguard",
  "serde",
  "serde_json",
  "sha2",
- "socket2",
+ "socket2 0.5.2",
  "sync_wrapper",
  "thiserror",
  "tls-listener",
  "tokio",
  "tokio-postgres",
  "tokio-postgres-rustls",
- "tokio-rustls",
+ "tokio-rustls 0.23.4",
+ "tokio-util",
  "tracing",
  "tracing-opentelemetry",
  "tracing-subscriber",
@@ -2969,16 +3146,16 @@ dependencies = [
  "url",
  "utils",
  "uuid",
- "webpki-roots",
+ "webpki-roots 0.23.0",
  "workspace_hack",
  "x509-parser",
 ]
 
 [[package]]
 name = "quote"
-version = "1.0.23"
+version = "1.0.26"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8856d8364d252a14d474036ea1358d63c9e6965c8e5c1885c18f73d70bff9c7b"
+checksum = "4424af4bf778aae2051a77b60283332f386554255d722233d09fbfc7e30da2fc"
 dependencies = [
  "proc-macro2",
 ]
@@ -3015,9 +3192,9 @@ dependencies = [
 
 [[package]]
 name = "rayon"
-version = "1.6.1"
+version = "1.7.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6db3a213adf02b3bcfd2d3846bb41cb22857d131789e01df434fb7e7bc0759b7"
+checksum = "1d2df5196e37bcc87abebc0053e20787d73847bb33134a69841207dd0a47f03b"
 dependencies = [
  "either",
  "rayon-core",
@@ -3025,9 +3202,9 @@ dependencies = [
 
 [[package]]
 name = "rayon-core"
-version = "1.10.2"
+version = "1.11.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "356a0625f1954f730c0201cdab48611198dc6ce21f4acff55089b5a78e6e835b"
+checksum = "4b8f95bd6966f5c87776639160a66bd8ab9895d9d4ab01ddba9fc60661aebe8d"
 dependencies = [
  "crossbeam-channel",
  "crossbeam-deque",
@@ -3057,10 +3234,19 @@ dependencies = [
 ]
 
 [[package]]
-name = "regex"
-version = "1.7.1"
+name = "redox_syscall"
+version = "0.3.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "48aaa5748ba571fb95cd2c85c09f629215d3a6ece942baa100950af03a34f733"
+checksum = "567664f262709473930a4bf9e51bf2ebf3348f2e748ccc50dea20646858f8f29"
+dependencies = [
+ "bitflags",
+]
+
+[[package]]
+name = "regex"
+version = "1.7.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8b1f693b24f6ac912f4893ef08244d70b6067480d2f1a46e950c9691e6749d1d"
 dependencies = [
  "aho-corasick",
  "memchr",
@@ -3078,9 +3264,9 @@ dependencies = [
 
 [[package]]
 name = "regex-syntax"
-version = "0.6.28"
+version = "0.6.29"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "456c603be3e8d448b072f410900c09faf164fbce2d480456f50eea6e25f9c848"
+checksum = "f162c6dd7b008981e4d40210aca20b4bd0f9b60ca9271061b07f78537722f2e1"
 
 [[package]]
 name = "remote_storage"
@@ -3090,8 +3276,8 @@ dependencies = [
  "async-trait",
  "aws-config",
  "aws-sdk-s3",
- "aws-smithy-http",
- "aws-types",
+ "aws-smithy-http 0.51.0",
+ "aws-types 0.55.1",
  "hyper",
  "metrics",
  "once_cell",
@@ -3110,9 +3296,9 @@ dependencies = [
 
 [[package]]
 name = "reqwest"
-version = "0.11.14"
+version = "0.11.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "21eed90ec8570952d53b772ecf8f206aa1ec9a3d76b2521c56c42973f2d91ee9"
+checksum = "27b71749df584b7f4cac2c426c127a7c785a5106cc98f7a8feb044115f0fa254"
 dependencies = [
  "base64 0.21.0",
  "bytes",
@@ -3132,27 +3318,27 @@ dependencies = [
  "once_cell",
  "percent-encoding",
  "pin-project-lite",
- "rustls",
+ "rustls 0.20.8",
  "rustls-pemfile",
  "serde",
  "serde_json",
  "serde_urlencoded",
  "tokio",
- "tokio-rustls",
+ "tokio-rustls 0.23.4",
  "tower-service",
  "url",
  "wasm-bindgen",
  "wasm-bindgen-futures",
  "web-sys",
- "webpki-roots",
+ "webpki-roots 0.22.6",
  "winreg",
 ]
 
 [[package]]
 name = "reqwest-middleware"
-version = "0.2.0"
+version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4a1c03e9011a8c59716ad13115550469e081e2e9892656b0ba6a47c907921894"
+checksum = "99c50db2c7ccd815f976473dd7d0bde296f8c3b77c383acf4fc021cdcf10852b"
 dependencies = [
  "anyhow",
  "async-trait",
@@ -3165,11 +3351,12 @@ dependencies = [
 
 [[package]]
 name = "reqwest-tracing"
-version = "0.4.0"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b739d87a6b2cf4743968ad2b4cef648fbe0204c19999509824425babb2097bce"
+checksum = "8a71d77945a1c5ae9604f0504901e77a1e2e71f2932b1cb8103078179ca62ff8"
 dependencies = [
  "async-trait",
+ "getrandom",
  "opentelemetry",
  "reqwest",
  "reqwest-middleware",
@@ -3208,18 +3395,18 @@ dependencies = [
 
 [[package]]
 name = "rpds"
-version = "0.12.0"
+version = "0.13.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "66262ea963eff99163e6b741fbc3417a52cc13074728c1047e9911789df9b000"
+checksum = "9bd6ce569b15c331b1e5fd8cf6adb0bf240678b5f0cdc4d0f41e11683f6feba9"
 dependencies = [
  "archery",
 ]
 
 [[package]]
 name = "rstest"
-version = "0.16.0"
+version = "0.17.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b07f2d176c472198ec1e6551dc7da28f1c089652f66a7b722676c2238ebc0edf"
+checksum = "de1bb486a691878cd320c2f0d319ba91eeaa2e894066d8b5f8f117c000e9d962"
 dependencies = [
  "futures",
  "futures-timer",
@@ -3229,23 +3416,23 @@ dependencies = [
 
 [[package]]
 name = "rstest_macros"
-version = "0.16.0"
+version = "0.17.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7229b505ae0706e64f37ffc54a9c163e11022a6636d58fe1f3f52018257ff9f7"
+checksum = "290ca1a1c8ca7edb7c3283bd44dc35dd54fdec6253a3912e201ba1072018fca8"
 dependencies = [
  "cfg-if",
  "proc-macro2",
  "quote",
  "rustc_version",
- "syn",
+ "syn 1.0.109",
  "unicode-ident",
 ]
 
 [[package]]
 name = "rustc-demangle"
-version = "0.1.21"
+version = "0.1.22"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7ef03e0a2b150c7a90d01faf6254c9c48a41e95fb2a8c2ac1c6f0d2b9aefc342"
+checksum = "d4a36c42d1873f9a77c53bde094f9664d9891bc604a45b4798fd2c389ed12e5b"
 
 [[package]]
 name = "rustc-hash"
@@ -3273,16 +3460,30 @@ dependencies = [
 
 [[package]]
 name = "rustix"
-version = "0.36.7"
+version = "0.36.12"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d4fdebc4b395b7fbb9ab11e462e20ed9051e7b16e42d24042c776eca0ac81b03"
+checksum = "e0af200a3324fa5bcd922e84e9b55a298ea9f431a489f01961acdebc6e908f25"
 dependencies = [
  "bitflags",
  "errno",
  "io-lifetimes",
  "libc",
- "linux-raw-sys",
- "windows-sys 0.42.0",
+ "linux-raw-sys 0.1.4",
+ "windows-sys 0.45.0",
+]
+
+[[package]]
+name = "rustix"
+version = "0.37.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "85597d61f83914ddeba6a47b3b8ffe7365107221c2e557ed94426489fefb5f77"
+dependencies = [
+ "bitflags",
+ "errno",
+ "io-lifetimes",
+ "libc",
+ "linux-raw-sys 0.3.1",
+ "windows-sys 0.48.0",
 ]
 
 [[package]]
@@ -3297,6 +3498,18 @@ dependencies = [
  "webpki",
 ]
 
+[[package]]
+name = "rustls"
+version = "0.21.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "07180898a28ed6a7f7ba2311594308f595e3dd2e3c3812fa0a80a47b45f17e5d"
+dependencies = [
+ "log",
+ "ring",
+ "rustls-webpki",
+ "sct",
+]
+
 [[package]]
 name = "rustls-native-certs"
 version = "0.6.2"
@@ -3319,16 +3532,26 @@ dependencies = [
 ]
 
 [[package]]
-name = "rustversion"
-version = "1.0.11"
+name = "rustls-webpki"
+version = "0.100.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5583e89e108996506031660fe09baa5011b9dd0341b89029313006d1fb508d70"
+checksum = "d6207cd5ed3d8dca7816f8f3725513a34609c0c765bf652b8c3cb4cfd87db46b"
+dependencies = [
+ "ring",
+ "untrusted",
+]
+
+[[package]]
+name = "rustversion"
+version = "1.0.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4f3208ce4d8448b3f3e7d168a73f5e0c43a61e32930de3bceeccedb388b6bf06"
 
 [[package]]
 name = "ryu"
-version = "1.0.12"
+version = "1.0.13"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7b4b9743ed687d4b4bcedf9ff5eaa7398495ae14e61cba0a295704edbc7decde"
+checksum = "f91339c0467de62360649f8d3e185ca8de4224ff281f66000de5eb2a77a79041"
 
 [[package]]
 name = "safekeeper"
@@ -3340,7 +3563,7 @@ dependencies = [
  "byteorder",
  "bytes",
  "chrono",
- "clap 4.1.4",
+ "clap 4.2.2",
  "const_format",
  "crc32c",
  "fs2",
@@ -3413,9 +3636,9 @@ checksum = "d29ab0c6d3fc0ee92fe66e2d99f700eab17a8d57d1c1d3b748380fb20baa78cd"
 
 [[package]]
 name = "scratch"
-version = "1.0.3"
+version = "1.0.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ddccb15bcce173023b3fedd9436f882a0739b8dfb45e4f6b6002bee5929f61b2"
+checksum = "1792db035ce95be60c3f8853017b3999209281c24e2ba5bc8e59bf97a0c590c1"
 
 [[package]]
 name = "sct"
@@ -3452,33 +3675,33 @@ dependencies = [
 
 [[package]]
 name = "semver"
-version = "1.0.16"
+version = "1.0.17"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "58bc9567378fc7690d6b2addae4e60ac2eeea07becb2c64b9f218b53865cba2a"
+checksum = "bebd363326d05ec3e2f532ab7660680f3b02130d780c299bca73469d521bc0ed"
 
 [[package]]
 name = "sentry"
-version = "0.29.2"
+version = "0.30.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a6097dc270a9c4555c5d6222ed243eaa97ff38e29299ed7c5cb36099033c604e"
+checksum = "b5ce6d3512e2617c209ec1e86b0ca2fea06454cd34653c91092bf0f3ec41f8e3"
 dependencies = [
  "httpdate",
  "reqwest",
- "rustls",
+ "rustls 0.20.8",
  "sentry-backtrace",
  "sentry-contexts",
  "sentry-core",
  "sentry-panic",
  "tokio",
  "ureq",
- "webpki-roots",
+ "webpki-roots 0.22.6",
 ]
 
 [[package]]
 name = "sentry-backtrace"
-version = "0.29.2"
+version = "0.30.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9d92d1e4d591534ae4f872d6142f3b500f4ffc179a6aed8a3e86c7cc96d10a6a"
+checksum = "0e7fe408d4d1f8de188a9309916e02e129cbe51ca19e55badea5a64899399b1a"
 dependencies = [
  "backtrace",
  "once_cell",
@@ -3488,9 +3711,9 @@ dependencies = [
 
 [[package]]
 name = "sentry-contexts"
-version = "0.29.2"
+version = "0.30.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3afa877b1898ff67dd9878cf4bec4e53cef7d3be9f14b1fc9e4fcdf36f8e4259"
+checksum = "5695096a059a89973ec541062d331ff4c9aeef9c2951416c894f0fff76340e7d"
 dependencies = [
  "hostname",
  "libc",
@@ -3502,9 +3725,9 @@ dependencies = [
 
 [[package]]
 name = "sentry-core"
-version = "0.29.2"
+version = "0.30.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fc43eb7e4e3a444151a0fe8a0e9ce60eabd905dae33d66e257fa26f1b509c1bd"
+checksum = "5b22828bfd118a7b660cf7a155002a494755c0424cebb7061e4743ecde9c7dbc"
 dependencies = [
  "once_cell",
  "rand",
@@ -3515,9 +3738,9 @@ dependencies = [
 
 [[package]]
 name = "sentry-panic"
-version = "0.29.2"
+version = "0.30.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ccab4fab11e3e63c45f4524bee2e75cde39cdf164cb0b0cbe6ccd1948ceddf66"
+checksum = "1f4ced2a7a8c14899d58eec402d946f69d5ed26a3fc363a7e8b1e5cb88473a01"
 dependencies = [
  "sentry-backtrace",
  "sentry-core",
@@ -3525,9 +3748,9 @@ dependencies = [
 
 [[package]]
 name = "sentry-types"
-version = "0.29.2"
+version = "0.30.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f63708ec450b6bdcb657af760c447416d69c38ce421f34e5e2e9ce8118410bc7"
+checksum = "360ee3270f7a4a1eee6c667f7d38360b995431598a73b740dfe420da548d9cc9"
 dependencies = [
  "debugid",
  "getrandom",
@@ -3542,35 +3765,44 @@ dependencies = [
 
 [[package]]
 name = "serde"
-version = "1.0.152"
+version = "1.0.160"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bb7d1f0d3021d347a83e556fc4683dea2ea09d87bccdf88ff5c12545d89d5efb"
+checksum = "bb2f3770c8bce3bcda7e149193a069a0f4365bda1fa5cd88e03bca26afc1216c"
 dependencies = [
  "serde_derive",
 ]
 
 [[package]]
 name = "serde_derive"
-version = "1.0.152"
+version = "1.0.160"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "af487d118eecd09402d70a5d72551860e788df87b464af30e5ea6a38c75c541e"
+checksum = "291a097c63d8497e00160b166a967a4a79c64f3facdd01cbd7502231688d77df"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.15",
 ]
 
 [[package]]
 name = "serde_json"
-version = "1.0.91"
+version = "1.0.96"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "877c235533714907a8c2464236f5c4b2a17262ef1bd71f38f35ea592c8da6883"
+checksum = "057d394a50403bcac12672b2b18fb387ab6d289d957dab67dd201875391e52f1"
 dependencies = [
  "itoa",
  "ryu",
  "serde",
 ]
 
+[[package]]
+name = "serde_spanned"
+version = "0.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0efd8caf556a6cebd3b285caf480045fcc1ac04f6bd786b09a6f11af30c4fcf4"
+dependencies = [
+ "serde",
+]
+
 [[package]]
 name = "serde_urlencoded"
 version = "0.7.1"
@@ -3585,9 +3817,9 @@ dependencies = [
 
 [[package]]
 name = "serde_with"
-version = "2.2.0"
+version = "2.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "30d904179146de381af4c93d3af6ca4984b3152db687dacb9c3c35e86f39809c"
+checksum = "331bb8c3bf9b92457ab7abecf07078c13f7d270ba490103e84e8b014490cd0b0"
 dependencies = [
  "base64 0.13.1",
  "chrono",
@@ -3601,14 +3833,14 @@ dependencies = [
 
 [[package]]
 name = "serde_with_macros"
-version = "2.2.0"
+version = "2.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a1966009f3c05f095697c537312f5415d1e3ed31ce0a56942bac4c771c5c335e"
+checksum = "859011bddcc11f289f07f467cc1fe01c7a941daa4d8f6c40d4d1c92eb6d9319c"
 dependencies = [
  "darling",
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -3650,9 +3882,9 @@ checksum = "43b2853a4d09f215c24cc5489c992ce46052d359b5109343cbafbf26bc62f8a3"
 
 [[package]]
 name = "signal-hook"
-version = "0.3.14"
+version = "0.3.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a253b5e89e2698464fc26b545c9edceb338e18a89effeeecfea192c3025be29d"
+checksum = "732768f1176d21d09e076c23a93123d40bba92d50c4058da34d45c8de8e682b9"
 dependencies = [
  "libc",
  "signal-hook-registry",
@@ -3671,9 +3903,9 @@ dependencies = [
 
 [[package]]
 name = "signal-hook-registry"
-version = "1.4.0"
+version = "1.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e51e73328dc4ac0c7ccbda3a494dfa03df1de2f46018127f60c693f2648455b0"
+checksum = "d8229b473baa5980ac72ef434c4415e70c4b5e71b423043adb4ba059f89c99a1"
 dependencies = [
  "libc",
 ]
@@ -3698,9 +3930,9 @@ checksum = "7bd3e3206899af3f8b12af284fafc038cc1dc2b41d1b89dd17297221c5d225de"
 
 [[package]]
 name = "slab"
-version = "0.4.7"
+version = "0.4.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4614a76b2a8be0058caa9dbbaf66d988527d86d003c11a94fbd335d7661edcef"
+checksum = "6528351c9bc8ab22353f9d776db39a20288e8d6c37ef8cfe3317cf875eecfc2d"
 dependencies = [
  "autocfg",
 ]
@@ -3713,14 +3945,24 @@ checksum = "a507befe795404456341dfab10cef66ead4c041f62b8b11bbb92bffe5d0953e0"
 
 [[package]]
 name = "socket2"
-version = "0.4.7"
+version = "0.4.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "02e2d2db9033d13a1567121ddd7a095ee144db4e1ca1b1bda3419bc0da294ebd"
+checksum = "64a4a911eed85daf18834cfaa86a79b7d266ff93ff5ba14005426219480ed662"
 dependencies = [
  "libc",
  "winapi",
 ]
 
+[[package]]
+name = "socket2"
+version = "0.5.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6d283f86695ae989d1e18440a943880967156325ba025f05049946bff47bcc2b"
+dependencies = [
+ "libc",
+ "windows-sys 0.48.0",
+]
+
 [[package]]
 name = "spin"
 version = "0.5.2"
@@ -3729,9 +3971,9 @@ checksum = "6e63cff320ae2c57904679ba7cb63280a3dc4613885beafb148ee7bf9aa9042d"
 
 [[package]]
 name = "spin"
-version = "0.9.4"
+version = "0.9.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7f6002a767bff9e83f8eeecf883ecb8011875a21ae8da43bffb817a57e78cc09"
+checksum = "6980e8d7511241f8acf4aebddbb1ff938df5eebe98691418c4468d0b72a96a67"
 dependencies = [
  "lock_api",
 ]
@@ -3755,7 +3997,7 @@ dependencies = [
  "anyhow",
  "async-stream",
  "bytes",
- "clap 4.1.4",
+ "clap 4.2.2",
  "const_format",
  "futures",
  "futures-core",
@@ -3769,8 +4011,8 @@ dependencies = [
  "prost",
  "tokio",
  "tokio-stream",
- "tonic",
- "tonic-build",
+ "tonic 0.9.1",
+ "tonic-build 0.9.1",
  "tracing",
  "utils",
  "workspace_hack",
@@ -3808,7 +4050,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "rustversion",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -3825,9 +4067,20 @@ checksum = "8fb1df15f412ee2e9dfc1c504260fa695c1c3f10fe9f4a6ee2d2184d7d6450e2"
 
 [[package]]
 name = "syn"
-version = "1.0.107"
+version = "1.0.109"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1f4064b5b16e03ae50984a5a8ed5d4f8803e6bc1fd170a3cda91a1be4b18e3f5"
+checksum = "72b64191b275b66ffe2469e8af2c1cfe3bafa67b529ead792a6d0160888b4237"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "unicode-ident",
+]
+
+[[package]]
+name = "syn"
+version = "2.0.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a34fcf3e8b60f57e6a14301a2e916d323af98b0ea63c599441eec8558660c822"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -3848,7 +4101,7 @@ checksum = "f36bdaa60a83aca3921b5259d5400cbf5e90fc51931376a9bd4a0eb79aa7210f"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
  "unicode-xid",
 ]
 
@@ -3865,24 +4118,24 @@ dependencies = [
 
 [[package]]
 name = "task-local-extensions"
-version = "0.1.3"
+version = "0.1.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4167afbec18ae012de40f8cf1b9bf48420abb390678c34821caa07d924941cc4"
+checksum = "ba323866e5d033818e3240feeb9f7db2c4296674e4d9e16b97b7bf8f490434e8"
 dependencies = [
- "tokio",
+ "pin-utils",
 ]
 
 [[package]]
 name = "tempfile"
-version = "3.4.0"
+version = "3.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "af18f7ae1acd354b992402e9ec5864359d693cd8a79dcbef59f76891701c1e95"
+checksum = "b9fbec84f381d5795b08656e4912bec604d162bff9291d6189a78f4c8ab87998"
 dependencies = [
  "cfg-if",
  "fastrand",
- "redox_syscall",
- "rustix",
- "windows-sys 0.42.0",
+ "redox_syscall 0.3.5",
+ "rustix 0.37.11",
+ "windows-sys 0.45.0",
 ]
 
 [[package]]
@@ -3922,7 +4175,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8901a55b0a7a06ebc4a674dcca925170da8e613fa3b163a1df804ed10afb154d"
 dependencies = [
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -3933,38 +4186,39 @@ checksum = "222a222a5bfe1bba4a77b45ec488a741b3cb8872e5e499451fd7d0129c9c7c3d"
 
 [[package]]
 name = "thiserror"
-version = "1.0.38"
+version = "1.0.40"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6a9cd18aa97d5c45c6603caea1da6628790b37f7a34b6ca89522331c5180fed0"
+checksum = "978c9a314bd8dc99be594bc3c175faaa9794be04a5a5e153caba6915336cebac"
 dependencies = [
  "thiserror-impl",
 ]
 
 [[package]]
 name = "thiserror-impl"
-version = "1.0.38"
+version = "1.0.40"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1fb327af4685e4d03fa8cbcf1716380da910eeb2bb8be417e7f9fd3fb164f36f"
+checksum = "f9456a42c5b0d803c8cd86e73dd7cc9edd429499f37a3550d286d5e86720569f"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.15",
 ]
 
 [[package]]
 name = "thread_local"
-version = "1.1.4"
+version = "1.1.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5516c27b78311c50bf42c071425c560ac799b11c30b31f87e3081965fe5e0180"
+checksum = "3fdd6f064ccff2d6567adcb3873ca630700f00b5ad3f060c25b5dcfd9a4ce152"
 dependencies = [
+ "cfg-if",
  "once_cell",
 ]
 
 [[package]]
 name = "time"
-version = "0.3.17"
+version = "0.3.20"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a561bf4617eebd33bca6434b988f39ed798e527f51a1e797d0ee4f61c0a38376"
+checksum = "cd0cbfecb4d19b5ea75bb31ad904eb5b9fa13f21079c3b92017ebdf4999a5890"
 dependencies = [
  "itoa",
  "serde",
@@ -3980,9 +4234,9 @@ checksum = "2e153e1f1acaef8acc537e68b44906d2db6436e2b35ac2c6b42640fff91f00fd"
 
 [[package]]
 name = "time-macros"
-version = "0.2.6"
+version = "0.2.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d967f99f534ca7e495c575c62638eebc2898a8c84c119b89e250477bc4ba16b2"
+checksum = "fd80a657e71da814b8e5d60d3374fc6d35045062245d80224748ae522dd76f36"
 dependencies = [
  "time-core",
 ]
@@ -4008,9 +4262,9 @@ dependencies = [
 
 [[package]]
 name = "tinyvec_macros"
-version = "0.1.0"
+version = "0.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "cda74da7e1a664f795bb1f8a87ec406fb89a02522cf6e50620d016add6dbbf5c"
+checksum = "1f3ccbac311fea05f86f61904b462b55fb3df8837a366dfc601a0161d0532f20"
 
 [[package]]
 name = "tls-listener"
@@ -4023,26 +4277,25 @@ dependencies = [
  "pin-project-lite",
  "thiserror",
  "tokio",
- "tokio-rustls",
+ "tokio-rustls 0.23.4",
 ]
 
 [[package]]
 name = "tokio"
-version = "1.25.0"
+version = "1.27.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c8e00990ebabbe4c14c08aca901caed183ecd5c09562a12c824bb53d3c3fd3af"
+checksum = "d0de47a4eecbe11f498978a9b29d792f0d2692d1dd003650c24c76510e3bc001"
 dependencies = [
  "autocfg",
  "bytes",
  "libc",
- "memchr",
  "mio",
  "num_cpus",
  "pin-project-lite",
  "signal-hook-registry",
- "socket2",
+ "socket2 0.4.9",
  "tokio-macros",
- "windows-sys 0.42.0",
+ "windows-sys 0.45.0",
 ]
 
 [[package]]
@@ -4057,13 +4310,13 @@ dependencies = [
 
 [[package]]
 name = "tokio-macros"
-version = "1.8.2"
+version = "2.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d266c00fde287f55d3f1c3e96c500c362a2b8c695076ec180f27918820bc6df8"
+checksum = "61a573bdc87985e9d6ddeed1b3d864e8a302c847e40d647746df2f1de209d1ce"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.15",
 ]
 
 [[package]]
@@ -4084,7 +4337,7 @@ dependencies = [
  "pin-project-lite",
  "postgres-protocol",
  "postgres-types",
- "socket2",
+ "socket2 0.4.9",
  "tokio",
  "tokio-util",
 ]
@@ -4097,10 +4350,10 @@ checksum = "606f2b73660439474394432239c82249c0d45eb5f23d91f401be1e33590444a7"
 dependencies = [
  "futures",
  "ring",
- "rustls",
+ "rustls 0.20.8",
  "tokio",
  "tokio-postgres",
- "tokio-rustls",
+ "tokio-rustls 0.23.4",
 ]
 
 [[package]]
@@ -4109,16 +4362,26 @@ version = "0.23.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c43ee83903113e03984cb9e5cebe6c04a5116269e900e3ddba8f068a62adda59"
 dependencies = [
- "rustls",
+ "rustls 0.20.8",
  "tokio",
  "webpki",
 ]
 
 [[package]]
-name = "tokio-stream"
-version = "0.1.11"
+name = "tokio-rustls"
+version = "0.24.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d660770404473ccd7bc9f8b28494a811bc18542b915c0855c51e8f419d5223ce"
+checksum = "e0d409377ff5b1e3ca6437aa86c1eb7d40c134bfec254e44c830defa92669db5"
+dependencies = [
+ "rustls 0.21.0",
+ "tokio",
+]
+
+[[package]]
+name = "tokio-stream"
+version = "0.1.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8fb52b74f05dbf495a8fba459fdc331812b96aa086d9eb78101fa0d4569c3313"
 dependencies = [
  "futures-core",
  "pin-project-lite",
@@ -4133,7 +4396,7 @@ dependencies = [
  "filetime",
  "futures-core",
  "libc",
- "redox_syscall",
+ "redox_syscall 0.2.16",
  "tokio",
  "tokio-stream",
  "xattr",
@@ -4153,9 +4416,9 @@ dependencies = [
 
 [[package]]
 name = "tokio-util"
-version = "0.7.4"
+version = "0.7.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0bb2e075f03b3d66d8d8785356224ba688d2906a371015e225beeb65ca92c740"
+checksum = "5427d89453009325de0d8f342c9490009f76e999cb7672d77e46267448f7e6b2"
 dependencies = [
  "bytes",
  "futures-core",
@@ -4167,33 +4430,36 @@ dependencies = [
 
 [[package]]
 name = "toml"
-version = "0.5.11"
+version = "0.7.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f4f7f0dd8d50a853a531c426359045b1998f04219d88799810762cd4ad314234"
+checksum = "b403acf6f2bb0859c93c7f0d967cb4a75a7ac552100f9322faf64dc047669b21"
 dependencies = [
  "serde",
+ "serde_spanned",
+ "toml_datetime",
+ "toml_edit",
 ]
 
 [[package]]
 name = "toml_datetime"
-version = "0.5.1"
+version = "0.6.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4553f467ac8e3d374bc9a177a26801e5d0f9b211aa1673fb137a403afd1c9cf5"
+checksum = "3ab8ed2edee10b50132aed5f331333428b011c99402b5a534154ed15746f9622"
 dependencies = [
  "serde",
 ]
 
 [[package]]
 name = "toml_edit"
-version = "0.17.1"
+version = "0.19.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a34cc558345efd7e88b9eda9626df2138b80bb46a7606f695e751c892bc7dac6"
+checksum = "239410c8609e8125456927e6707163a3b1fdb40561e4b803bc041f466ccfdc13"
 dependencies = [
  "indexmap",
- "itertools",
- "nom8",
  "serde",
+ "serde_spanned",
  "toml_datetime",
+ "winnow",
 ]
 
 [[package]]
@@ -4218,10 +4484,7 @@ dependencies = [
  "pin-project",
  "prost",
  "prost-derive",
- "rustls-native-certs",
- "rustls-pemfile",
  "tokio",
- "tokio-rustls",
  "tokio-stream",
  "tokio-util",
  "tower",
@@ -4231,17 +4494,62 @@ dependencies = [
  "tracing-futures",
 ]
 
+[[package]]
+name = "tonic"
+version = "0.9.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "38bd8e87955eb13c1986671838177d6792cdc52af9bffced0d2c8a9a7f741ab3"
+dependencies = [
+ "async-stream",
+ "async-trait",
+ "axum",
+ "base64 0.21.0",
+ "bytes",
+ "futures-core",
+ "futures-util",
+ "h2",
+ "http",
+ "http-body",
+ "hyper",
+ "hyper-timeout",
+ "percent-encoding",
+ "pin-project",
+ "prost",
+ "rustls-native-certs",
+ "rustls-pemfile",
+ "tokio",
+ "tokio-rustls 0.24.0",
+ "tokio-stream",
+ "tower",
+ "tower-layer",
+ "tower-service",
+ "tracing",
+]
+
 [[package]]
 name = "tonic-build"
 version = "0.8.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5bf5e9b9c0f7e0a7c027dcfaba7b2c60816c7049171f679d99ee2ff65d0de8c4"
 dependencies = [
- "prettyplease",
+ "prettyplease 0.1.25",
  "proc-macro2",
  "prost-build",
  "quote",
- "syn",
+ "syn 1.0.109",
+]
+
+[[package]]
+name = "tonic-build"
+version = "0.9.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0f60a933bbea70c95d633c04c951197ddf084958abaa2ed502a3743bdd8d8dd7"
+dependencies = [
+ "prettyplease 0.1.25",
+ "proc-macro2",
+ "prost-build",
+ "quote",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -4264,25 +4572,6 @@ dependencies = [
  "tracing",
 ]
 
-[[package]]
-name = "tower-http"
-version = "0.3.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f873044bf02dd1e8239e9c1293ea39dad76dc594ec16185d0a1bf31d8dc8d858"
-dependencies = [
- "bitflags",
- "bytes",
- "futures-core",
- "futures-util",
- "http",
- "http-body",
- "http-range-header",
- "pin-project-lite",
- "tower",
- "tower-layer",
- "tower-service",
-]
-
 [[package]]
 name = "tower-layer"
 version = "0.3.2"
@@ -4300,7 +4589,7 @@ name = "trace"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "clap 4.1.4",
+ "clap 4.2.2",
  "pageserver_api",
  "utils",
  "workspace_hack",
@@ -4327,7 +4616,7 @@ checksum = "4017f8f45139870ca7e672686113917c71c7a6e02d4924eda67186083c03081a"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
 ]
 
 [[package]]
@@ -4340,6 +4629,16 @@ dependencies = [
  "valuable",
 ]
 
+[[package]]
+name = "tracing-error"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d686ec1c0f384b1277f097b2f279a2ecc11afe8c133c1aabf036a27cb4cd206e"
+dependencies = [
+ "tracing",
+ "tracing-subscriber",
+]
+
 [[package]]
 name = "tracing-futures"
 version = "0.2.5"
@@ -4473,15 +4772,15 @@ dependencies = [
 
 [[package]]
 name = "unicode-bidi"
-version = "0.3.10"
+version = "0.3.13"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d54675592c1dbefd78cbd98db9bacd89886e1ca50692a0692baefffdeb92dd58"
+checksum = "92888ba5573ff080736b3648696b70cafad7d250551175acbaa4e0385b3e1460"
 
 [[package]]
 name = "unicode-ident"
-version = "1.0.6"
+version = "1.0.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "84a22b9f218b40614adcb3f4ff08b703773ad44fa9423e4e0d346d5db86e4ebc"
+checksum = "e5464a87b239f13a63a501f2701565754bae92d243d4bb7eb12f6d57d2269bf4"
 
 [[package]]
 name = "unicode-normalization"
@@ -4519,10 +4818,10 @@ dependencies = [
  "base64 0.13.1",
  "log",
  "once_cell",
- "rustls",
+ "rustls 0.20.8",
  "url",
  "webpki",
- "webpki-roots",
+ "webpki-roots 0.22.6",
 ]
 
 [[package]]
@@ -4549,6 +4848,12 @@ version = "0.7.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "09cc8ee72d2a9becf2f2febe0205bbed8fc6615b7cb429ad062dc7b7ddd036a9"
 
+[[package]]
+name = "utf8parse"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "711b9620af191e0cdc7468a8d14e709c3dcdb115b36f838e601583af800a370a"
+
 [[package]]
 name = "utils"
 version = "0.1.0"
@@ -4559,6 +4864,7 @@ dependencies = [
  "bincode",
  "byteorder",
  "bytes",
+ "chrono",
  "criterion",
  "futures",
  "heapless",
@@ -4570,6 +4876,7 @@ dependencies = [
  "nix",
  "once_cell",
  "pin-project-lite",
+ "pq_proto",
  "rand",
  "regex",
  "routerify",
@@ -4584,6 +4891,7 @@ dependencies = [
  "thiserror",
  "tokio",
  "tracing",
+ "tracing-error",
  "tracing-subscriber",
  "url",
  "uuid",
@@ -4592,9 +4900,9 @@ dependencies = [
 
 [[package]]
 name = "uuid"
-version = "1.3.0"
+version = "1.3.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1674845326ee10d37ca60470760d4288a6f80f304007d92e5c53bab78c9cfd79"
+checksum = "5b55a3fef2a1e3b3a00ce878640918820d3c51081576ac657d23af9fc7928fdb"
 dependencies = [
  "getrandom",
  "serde",
@@ -4612,12 +4920,18 @@ version = "0.9.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "49874b5167b65d7193b8aba1567f5c7d93d001cafc34600cee003eda787e483f"
 
+[[package]]
+name = "vsimd"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5c3082ca00d5a5ef149bb8b555a72ae84c9c59f7250f013ac822ac2e49b19c64"
+
 [[package]]
 name = "wal_craft"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "clap 4.1.4",
+ "clap 4.2.2",
  "env_logger",
  "log",
  "once_cell",
@@ -4629,12 +4943,11 @@ dependencies = [
 
 [[package]]
 name = "walkdir"
-version = "2.3.2"
+version = "2.3.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "808cf2735cd4b6866113f648b791c6adc5714537bc222d9347bb203386ffda56"
+checksum = "36df944cda56c7d8d8b7496af378e6b16de9284591917d307c9b4d313c44e698"
 dependencies = [
  "same-file",
- "winapi",
  "winapi-util",
 ]
 
@@ -4675,7 +4988,7 @@ dependencies = [
  "once_cell",
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
  "wasm-bindgen-shared",
 ]
 
@@ -4709,7 +5022,7 @@ checksum = "2aff81306fcac3c7515ad4e177f521b5c9a15f2b08f4e32d823066102f35a5f6"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 1.0.109",
  "wasm-bindgen-backend",
  "wasm-bindgen-shared",
 ]
@@ -4749,6 +5062,15 @@ dependencies = [
  "webpki",
 ]
 
+[[package]]
+name = "webpki-roots"
+version = "0.23.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "aa54963694b65584e170cf5dc46aeb4dcaa5584e652ff5f3952e56d66aff0125"
+dependencies = [
+ "rustls-webpki",
+]
+
 [[package]]
 name = "which"
 version = "4.4.0"
@@ -4791,19 +5113,28 @@ version = "0.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"
 
+[[package]]
+name = "windows"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e686886bc078bc1b0b600cac0147aadb815089b6e4da64016cbd754b6342700f"
+dependencies = [
+ "windows-targets 0.48.0",
+]
+
 [[package]]
 name = "windows-sys"
 version = "0.42.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5a3e1820f08b8513f676f7ab6c1f99ff312fb97b553d30ff4dd86f9f15728aa7"
 dependencies = [
- "windows_aarch64_gnullvm",
- "windows_aarch64_msvc",
- "windows_i686_gnu",
- "windows_i686_msvc",
- "windows_x86_64_gnu",
- "windows_x86_64_gnullvm",
- "windows_x86_64_msvc",
+ "windows_aarch64_gnullvm 0.42.2",
+ "windows_aarch64_msvc 0.42.2",
+ "windows_i686_gnu 0.42.2",
+ "windows_i686_msvc 0.42.2",
+ "windows_x86_64_gnu 0.42.2",
+ "windows_x86_64_gnullvm 0.42.2",
+ "windows_x86_64_msvc 0.42.2",
 ]
 
 [[package]]
@@ -4812,65 +5143,140 @@ version = "0.45.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "75283be5efb2831d37ea142365f009c02ec203cd29a3ebecbc093d52315b66d0"
 dependencies = [
- "windows-targets",
+ "windows-targets 0.42.2",
+]
+
+[[package]]
+name = "windows-sys"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "677d2418bec65e3338edb076e806bc1ec15693c5d0104683f2efe857f61056a9"
+dependencies = [
+ "windows-targets 0.48.0",
 ]
 
 [[package]]
 name = "windows-targets"
-version = "0.42.1"
+version = "0.42.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8e2522491fbfcd58cc84d47aeb2958948c4b8982e9a2d8a2a35bbaed431390e7"
+checksum = "8e5180c00cd44c9b1c88adb3693291f1cd93605ded80c250a75d472756b4d071"
 dependencies = [
- "windows_aarch64_gnullvm",
- "windows_aarch64_msvc",
- "windows_i686_gnu",
- "windows_i686_msvc",
- "windows_x86_64_gnu",
- "windows_x86_64_gnullvm",
- "windows_x86_64_msvc",
+ "windows_aarch64_gnullvm 0.42.2",
+ "windows_aarch64_msvc 0.42.2",
+ "windows_i686_gnu 0.42.2",
+ "windows_i686_msvc 0.42.2",
+ "windows_x86_64_gnu 0.42.2",
+ "windows_x86_64_gnullvm 0.42.2",
+ "windows_x86_64_msvc 0.42.2",
+]
+
+[[package]]
+name = "windows-targets"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7b1eb6f0cd7c80c79759c929114ef071b87354ce476d9d94271031c0497adfd5"
+dependencies = [
+ "windows_aarch64_gnullvm 0.48.0",
+ "windows_aarch64_msvc 0.48.0",
+ "windows_i686_gnu 0.48.0",
+ "windows_i686_msvc 0.48.0",
+ "windows_x86_64_gnu 0.48.0",
+ "windows_x86_64_gnullvm 0.48.0",
+ "windows_x86_64_msvc 0.48.0",
 ]
 
 [[package]]
 name = "windows_aarch64_gnullvm"
-version = "0.42.1"
+version = "0.42.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8c9864e83243fdec7fc9c5444389dcbbfd258f745e7853198f365e3c4968a608"
+checksum = "597a5118570b68bc08d8d59125332c54f1ba9d9adeedeef5b99b02ba2b0698f8"
+
+[[package]]
+name = "windows_aarch64_gnullvm"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "91ae572e1b79dba883e0d315474df7305d12f569b400fcf90581b06062f7e1bc"
 
 [[package]]
 name = "windows_aarch64_msvc"
-version = "0.42.1"
+version = "0.42.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4c8b1b673ffc16c47a9ff48570a9d85e25d265735c503681332589af6253c6c7"
+checksum = "e08e8864a60f06ef0d0ff4ba04124db8b0fb3be5776a5cd47641e942e58c4d43"
+
+[[package]]
+name = "windows_aarch64_msvc"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b2ef27e0d7bdfcfc7b868b317c1d32c641a6fe4629c171b8928c7b08d98d7cf3"
 
 [[package]]
 name = "windows_i686_gnu"
-version = "0.42.1"
+version = "0.42.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "de3887528ad530ba7bdbb1faa8275ec7a1155a45ffa57c37993960277145d640"
+checksum = "c61d927d8da41da96a81f029489353e68739737d3beca43145c8afec9a31a84f"
+
+[[package]]
+name = "windows_i686_gnu"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "622a1962a7db830d6fd0a69683c80a18fda201879f0f447f065a3b7467daa241"
 
 [[package]]
 name = "windows_i686_msvc"
-version = "0.42.1"
+version = "0.42.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bf4d1122317eddd6ff351aa852118a2418ad4214e6613a50e0191f7004372605"
+checksum = "44d840b6ec649f480a41c8d80f9c65108b92d89345dd94027bfe06ac444d1060"
+
+[[package]]
+name = "windows_i686_msvc"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4542c6e364ce21bf45d69fdd2a8e455fa38d316158cfd43b3ac1c5b1b19f8e00"
 
 [[package]]
 name = "windows_x86_64_gnu"
-version = "0.42.1"
+version = "0.42.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c1040f221285e17ebccbc2591ffdc2d44ee1f9186324dd3e84e99ac68d699c45"
+checksum = "8de912b8b8feb55c064867cf047dda097f92d51efad5b491dfb98f6bbb70cb36"
+
+[[package]]
+name = "windows_x86_64_gnu"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ca2b8a661f7628cbd23440e50b05d705db3686f894fc9580820623656af974b1"
 
 [[package]]
 name = "windows_x86_64_gnullvm"
-version = "0.42.1"
+version = "0.42.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "628bfdf232daa22b0d64fdb62b09fcc36bb01f05a3939e20ab73aaf9470d0463"
+checksum = "26d41b46a36d453748aedef1486d5c7a85db22e56aff34643984ea85514e94a3"
+
+[[package]]
+name = "windows_x86_64_gnullvm"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7896dbc1f41e08872e9d5e8f8baa8fdd2677f29468c4e156210174edc7f7b953"
 
 [[package]]
 name = "windows_x86_64_msvc"
-version = "0.42.1"
+version = "0.42.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "447660ad36a13288b1db4d4248e857b510e8c3a225c822ba4fb748c0aafecffd"
+checksum = "9aec5da331524158c6d1a4ac0ab1541149c0b9505fde06423b02f5ef0106b9f0"
+
+[[package]]
+name = "windows_x86_64_msvc"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1a515f5799fe4961cb532f983ce2b23082366b898e52ffbce459c86f67c8378a"
+
+[[package]]
+name = "winnow"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ae8970b36c66498d8ff1d66685dc86b91b29db0c7739899012f63a63814b4b28"
+dependencies = [
+ "memchr",
+]
 
 [[package]]
 name = "winreg"
@@ -4886,12 +5292,11 @@ name = "workspace_hack"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "byteorder",
  "bytes",
  "chrono",
- "clap 4.1.4",
+ "clap 4.2.2",
+ "clap_builder",
  "crossbeam-utils",
- "digest",
  "either",
  "fail",
  "futures",
@@ -4901,7 +5306,6 @@ dependencies = [
  "futures-sink",
  "futures-util",
  "hashbrown 0.12.3",
- "indexmap",
  "itertools",
  "libc",
  "log",
@@ -4916,16 +5320,18 @@ dependencies = [
  "regex-syntax",
  "reqwest",
  "ring",
- "rustls",
+ "rustls 0.20.8",
  "scopeguard",
  "serde",
  "serde_json",
- "socket2",
- "syn",
+ "socket2 0.4.9",
+ "syn 1.0.109",
+ "syn 2.0.15",
  "tokio",
- "tokio-rustls",
+ "tokio-rustls 0.23.4",
  "tokio-util",
- "tonic",
+ "toml_datetime",
+ "toml_edit",
  "tower",
  "tracing",
  "tracing-core",
@@ -4935,12 +5341,11 @@ dependencies = [
 
 [[package]]
 name = "x509-parser"
-version = "0.14.0"
+version = "0.15.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e0ecbeb7b67ce215e40e3cc7f2ff902f94a223acf44995934763467e7b1febc8"
+checksum = "bab0c2f54ae1d92f4fcb99c0b7ccf0b1e3451cbd395e5f115ccbdbcb18d4f634"
 dependencies = [
  "asn1-rs",
- "base64 0.13.1",
  "data-encoding",
  "der-parser",
  "lazy_static",
@@ -4968,15 +5373,15 @@ checksum = "4d25c75bf9ea12c4040a97f829154768bbbce366287e2dc044af160cd79a13fd"
 
 [[package]]
 name = "yasna"
-version = "0.5.1"
+version = "0.5.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "aed2e7a52e3744ab4d0c05c20aa065258e84c49fd4226f5191b2ed29712710b4"
+checksum = "e17bb3549cc1321ae1296b9cdc2698e2b6cb1992adfa19a8c72e5b7a738f44cd"
 dependencies = [
  "time",
 ]
 
 [[package]]
 name = "zeroize"
-version = "1.5.7"
+version = "1.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c394b5bd0c6f669e7275d9c20aa90ae064cb22e75a1cad54e1b34088034b149f"
+checksum = "2a0956f1ba7c7909bfb66c2e9e4124ab6f6482560f6628b5aaeba39207c9aad9"
diff --git a/Cargo.toml b/Cargo.toml
index 679605dc1d..f4872433cd 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -24,10 +24,10 @@ atty = "0.2.14"
 aws-config = { version = "0.51.0", default-features = false, features=["rustls"] }
 aws-sdk-s3 = "0.21.0"
 aws-smithy-http = "0.51.0"
-aws-types = "0.51.0"
+aws-types = "0.55"
 base64 = "0.13.0"
 bincode = "1.3"
-bindgen = "0.61"
+bindgen = "0.65"
 bstr = "1.0"
 byteorder = "1.4"
 bytes = "1.0"
@@ -50,7 +50,7 @@ git-version = "0.3"
 hashbrown = "0.13"
 hashlink = "0.8.1"
 hex = "0.4"
-hex-literal = "0.3"
+hex-literal = "0.4"
 hmac = "0.12.1"
 hostname = "0.3.1"
 humantime = "2.1"
@@ -80,18 +80,18 @@ reqwest = { version = "0.11", default-features = false, features = ["rustls-tls"
 reqwest-tracing = { version = "0.4.0", features = ["opentelemetry_0_18"] }
 reqwest-middleware = "0.2.0"
 routerify = "3"
-rpds = "0.12.0"
+rpds = "0.13"
 rustls = "0.20"
 rustls-pemfile = "1"
 rustls-split = "0.3"
 scopeguard = "1.1"
-sentry = { version = "0.29", default-features = false, features = ["backtrace", "contexts", "panic", "rustls", "reqwest" ] }
+sentry = { version = "0.30", default-features = false, features = ["backtrace", "contexts", "panic", "rustls", "reqwest" ] }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1"
 serde_with = "2.0"
 sha2 = "0.10.2"
 signal-hook = "0.3"
-socket2 = "0.4.4"
+socket2 = "0.5"
 strum = "0.24"
 strum_macros = "0.24"
 svg_fmt = "0.4.1"
@@ -106,17 +106,18 @@ tokio-postgres-rustls = "0.9.0"
 tokio-rustls = "0.23"
 tokio-stream = "0.1"
 tokio-util = { version = "0.7", features = ["io"] }
-toml = "0.5"
-toml_edit = { version = "0.17", features = ["easy"] }
-tonic = {version = "0.8", features = ["tls", "tls-roots"]}
+toml = "0.7"
+toml_edit = "0.19"
+tonic = {version = "0.9", features = ["tls", "tls-roots"]}
 tracing = "0.1"
+tracing-error = "0.2.0"
 tracing-opentelemetry = "0.18.0"
 tracing-subscriber = { version = "0.3", features = ["env-filter"] }
 url = "2.2"
 uuid = { version = "1.2", features = ["v4", "serde"] }
 walkdir = "2.3.2"
-webpki-roots = "0.22.5"
-x509-parser = "0.14"
+webpki-roots = "0.23"
+x509-parser = "0.15"
 
 ## TODO replace this with tracing
 env_logger = "0.10"
@@ -154,9 +155,9 @@ workspace_hack = { version = "0.1", path = "./workspace_hack/" }
 ## Build dependencies
 criterion = "0.4"
 rcgen = "0.10"
-rstest = "0.16"
+rstest = "0.17"
 tempfile = "3.4"
-tonic-build = "0.8"
+tonic-build = "0.9"
 
 # This is only needed for proxy's tests.
 # TODO: we should probably fork `tokio-postgres-rustls` instead.
diff --git a/Dockerfile.compute-node b/Dockerfile.compute-node
index 3473487444..229e09aa98 100644
--- a/Dockerfile.compute-node
+++ b/Dockerfile.compute-node
@@ -12,7 +12,7 @@ FROM debian:bullseye-slim AS build-deps
 RUN apt update &&  \
     apt install -y git autoconf automake libtool build-essential bison flex libreadline-dev \
     zlib1g-dev libxml2-dev libcurl4-openssl-dev libossp-uuid-dev wget pkg-config libssl-dev \
-    libicu-dev libxslt1-dev
+    libicu-dev libxslt1-dev liblz4-dev libzstd-dev
 
 #########################################################################################
 #
@@ -24,8 +24,13 @@ FROM build-deps AS pg-build
 ARG PG_VERSION
 COPY vendor/postgres-${PG_VERSION} postgres
 RUN cd postgres && \
-    ./configure CFLAGS='-O2 -g3' --enable-debug --with-openssl --with-uuid=ossp --with-icu \
-    --with-libxml --with-libxslt && \
+    export CONFIGURE_CMD="./configure CFLAGS='-O2 -g3' --enable-debug --with-openssl --with-uuid=ossp \
+    --with-icu --with-libxml --with-libxslt --with-lz4" && \
+    if [ "${PG_VERSION}" != "v14" ]; then \
+        # zstd is available only from PG15
+        export CONFIGURE_CMD="${CONFIGURE_CMD} --with-zstd"; \
+    fi && \
+    eval $CONFIGURE_CMD && \
     make MAKELEVEL=0 -j $(getconf _NPROCESSORS_ONLN) -s install && \
     make MAKELEVEL=0 -j $(getconf _NPROCESSORS_ONLN) -s -C contrib/ install && \
     # Install headers
@@ -60,6 +65,7 @@ RUN apt update && \
 
 # SFCGAL > 1.3 requires CGAL > 5.2, Bullseye's libcgal-dev is 5.2
 RUN wget https://gitlab.com/Oslandia/SFCGAL/-/archive/v1.3.10/SFCGAL-v1.3.10.tar.gz -O SFCGAL.tar.gz && \
+    echo "4e39b3b2adada6254a7bdba6d297bb28e1a9835a9f879b74f37e2dab70203232 SFCGAL.tar.gz" | sha256sum --check && \
     mkdir sfcgal-src && cd sfcgal-src && tar xvzf ../SFCGAL.tar.gz --strip-components=1 -C . && \
     cmake . && make -j $(getconf _NPROCESSORS_ONLN) && \
     DESTDIR=/sfcgal make install -j $(getconf _NPROCESSORS_ONLN) && \
@@ -68,6 +74,7 @@ RUN wget https://gitlab.com/Oslandia/SFCGAL/-/archive/v1.3.10/SFCGAL-v1.3.10.tar
 ENV PATH "/usr/local/pgsql/bin:$PATH"
 
 RUN wget https://download.osgeo.org/postgis/source/postgis-3.3.2.tar.gz -O postgis.tar.gz && \
+    echo "9a2a219da005a1730a39d1959a1c7cec619b1efb009b65be80ffc25bad299068 postgis.tar.gz" | sha256sum --check && \
     mkdir postgis-src && cd postgis-src && tar xvzf ../postgis.tar.gz --strip-components=1 -C . && \
     ./autogen.sh && \
     ./configure --with-sfcgal=/usr/local/bin/sfcgal-config && \
@@ -84,6 +91,7 @@ RUN wget https://download.osgeo.org/postgis/source/postgis-3.3.2.tar.gz -O postg
     echo 'trusted = true' >> /usr/local/pgsql/share/extension/address_standardizer_data_us.control
 
 RUN wget https://github.com/pgRouting/pgrouting/archive/v3.4.2.tar.gz -O pgrouting.tar.gz && \
+    echo "cac297c07d34460887c4f3b522b35c470138760fe358e351ad1db4edb6ee306e pgrouting.tar.gz" | sha256sum --check && \
     mkdir pgrouting-src && cd pgrouting-src && tar xvzf ../pgrouting.tar.gz --strip-components=1 -C . && \
     mkdir build && \
     cd build && \
@@ -104,6 +112,7 @@ RUN apt update && \
     apt install -y ninja-build python3-dev libncurses5 binutils clang
 
 RUN wget https://github.com/plv8/plv8/archive/refs/tags/v3.1.5.tar.gz -O plv8.tar.gz && \
+    echo "1e108d5df639e4c189e1c5bdfa2432a521c126ca89e7e5a969d46899ca7bf106 plv8.tar.gz" | sha256sum --check && \
     mkdir plv8-src && cd plv8-src && tar xvzf ../plv8.tar.gz --strip-components=1 -C . && \
     export PATH="/usr/local/pgsql/bin:$PATH" && \
     make DOCKER=1 -j $(getconf _NPROCESSORS_ONLN) install && \
@@ -125,11 +134,13 @@ COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 # packaged cmake is too old
 RUN wget https://github.com/Kitware/CMake/releases/download/v3.24.2/cmake-3.24.2-linux-x86_64.sh \
       -q -O /tmp/cmake-install.sh \
+      && echo "739d372726cb23129d57a539ce1432453448816e345e1545f6127296926b6754 /tmp/cmake-install.sh" | sha256sum --check \
       && chmod u+x /tmp/cmake-install.sh \
       && /tmp/cmake-install.sh --skip-license --prefix=/usr/local/ \
       && rm /tmp/cmake-install.sh
 
 RUN wget https://github.com/uber/h3/archive/refs/tags/v4.1.0.tar.gz -O h3.tar.gz && \
+    echo "ec99f1f5974846bde64f4513cf8d2ea1b8d172d2218ab41803bf6a63532272bc h3.tar.gz" | sha256sum --check && \
     mkdir h3-src && cd h3-src && tar xvzf ../h3.tar.gz --strip-components=1 -C . && \
     mkdir build && cd build && \
     cmake .. -DCMAKE_BUILD_TYPE=Release && \
@@ -139,6 +150,7 @@ RUN wget https://github.com/uber/h3/archive/refs/tags/v4.1.0.tar.gz -O h3.tar.gz
     rm -rf build
 
 RUN wget https://github.com/zachasme/h3-pg/archive/refs/tags/v4.1.2.tar.gz -O h3-pg.tar.gz && \
+    echo "c135aa45999b2ad1326d2537c1cadef96d52660838e4ca371706c08fdea1a956 h3-pg.tar.gz" | sha256sum --check && \
     mkdir h3-pg-src && cd h3-pg-src && tar xvzf ../h3-pg.tar.gz --strip-components=1 -C . && \
     export PATH="/usr/local/pgsql/bin:$PATH" && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -156,6 +168,7 @@ FROM build-deps AS unit-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN wget https://github.com/df7cb/postgresql-unit/archive/refs/tags/7.7.tar.gz -O postgresql-unit.tar.gz && \
+    echo "411d05beeb97e5a4abf17572bfcfbb5a68d98d1018918feff995f6ee3bb03e79 postgresql-unit.tar.gz" | sha256sum --check && \
     mkdir postgresql-unit-src && cd postgresql-unit-src && tar xvzf ../postgresql-unit.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -176,6 +189,7 @@ FROM build-deps AS vector-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN wget https://github.com/pgvector/pgvector/archive/refs/tags/v0.4.0.tar.gz -O pgvector.tar.gz && \
+    echo "b76cf84ddad452cc880a6c8c661d137ddd8679c000a16332f4f03ecf6e10bcc8 pgvector.tar.gz" | sha256sum --check && \
     mkdir pgvector-src && cd pgvector-src && tar xvzf ../pgvector.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -192,6 +206,7 @@ COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 # 9742dab1b2f297ad3811120db7b21451bca2d3c9 made on 13/11/2021
 RUN wget https://github.com/michelp/pgjwt/archive/9742dab1b2f297ad3811120db7b21451bca2d3c9.tar.gz -O pgjwt.tar.gz && \
+    echo "cfdefb15007286f67d3d45510f04a6a7a495004be5b3aecb12cda667e774203f pgjwt.tar.gz" | sha256sum --check && \
     mkdir pgjwt-src && cd pgjwt-src && tar xvzf ../pgjwt.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     echo 'trusted = true' >> /usr/local/pgsql/share/extension/pgjwt.control
@@ -206,6 +221,7 @@ FROM build-deps AS hypopg-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN wget https://github.com/HypoPG/hypopg/archive/refs/tags/1.3.1.tar.gz -O hypopg.tar.gz && \
+    echo "e7f01ee0259dc1713f318a108f987663d60f3041948c2ada57a94b469565ca8e hypopg.tar.gz" | sha256sum --check && \
     mkdir hypopg-src && cd hypopg-src && tar xvzf ../hypopg.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -221,6 +237,7 @@ FROM build-deps AS pg-hashids-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN wget https://github.com/iCyberon/pg_hashids/archive/refs/tags/v1.2.1.tar.gz -O pg_hashids.tar.gz && \
+    echo "74576b992d9277c92196dd8d816baa2cc2d8046fe102f3dcd7f3c3febed6822a pg_hashids.tar.gz" | sha256sum --check && \
     mkdir pg_hashids-src && cd pg_hashids-src && tar xvzf ../pg_hashids.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
@@ -236,6 +253,7 @@ FROM build-deps AS rum-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN wget https://github.com/postgrespro/rum/archive/refs/tags/1.3.13.tar.gz -O rum.tar.gz && \
+    echo "6ab370532c965568df6210bd844ac6ba649f53055e48243525b0b7e5c4d69a7d rum.tar.gz" | sha256sum --check && \
     mkdir rum-src && cd rum-src && tar xvzf ../rum.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
@@ -251,11 +269,28 @@ FROM build-deps AS pgtap-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN wget https://github.com/theory/pgtap/archive/refs/tags/v1.2.0.tar.gz -O pgtap.tar.gz && \
+    echo "9c7c3de67ea41638e14f06da5da57bac6f5bd03fea05c165a0ec862205a5c052 pgtap.tar.gz" | sha256sum --check && \
     mkdir pgtap-src && cd pgtap-src && tar xvzf ../pgtap.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     echo 'trusted = true' >> /usr/local/pgsql/share/extension/pgtap.control
 
+#########################################################################################
+#
+# Layer "ip4r-pg-build"
+# compile ip4r extension
+#
+#########################################################################################
+FROM build-deps AS ip4r-pg-build
+COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
+
+RUN wget https://github.com/RhodiumToad/ip4r/archive/refs/tags/2.4.1.tar.gz -O ip4r.tar.gz && \
+    echo "78b9f0c1ae45c22182768fe892a32d533c82281035e10914111400bf6301c726 ip4r.tar.gz" | sha256sum --check && \
+    mkdir ip4r-src && cd ip4r-src && tar xvzf ../ip4r.tar.gz --strip-components=1 -C . && \
+    make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
+    make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
+    echo 'trusted = true' >> /usr/local/pgsql/share/extension/ip4r.control
+
 #########################################################################################
 #
 # Layer "prefix-pg-build"
@@ -266,6 +301,7 @@ FROM build-deps AS prefix-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN wget https://github.com/dimitri/prefix/archive/refs/tags/v1.2.9.tar.gz -O prefix.tar.gz && \
+    echo "38d30a08d0241a8bbb8e1eb8f0152b385051665a8e621c8899e7c5068f8b511e prefix.tar.gz" | sha256sum --check && \
     mkdir prefix-src && cd prefix-src && tar xvzf ../prefix.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -281,6 +317,7 @@ FROM build-deps AS hll-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN wget https://github.com/citusdata/postgresql-hll/archive/refs/tags/v2.17.tar.gz -O hll.tar.gz && \
+    echo "9a18288e884f197196b0d29b9f178ba595b0dfc21fbf7a8699380e77fa04c1e9 hll.tar.gz" | sha256sum --check && \
     mkdir hll-src && cd hll-src && tar xvzf ../hll.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -296,6 +333,7 @@ FROM build-deps AS plpgsql-check-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN wget https://github.com/okbob/plpgsql_check/archive/refs/tags/v2.3.2.tar.gz -O plpgsql_check.tar.gz && \
+    echo "9d81167c4bbeb74eebf7d60147b21961506161addc2aee537f95ad8efeae427b plpgsql_check.tar.gz" | sha256sum --check && \
     mkdir plpgsql_check-src && cd plpgsql_check-src && tar xvzf ../plpgsql_check.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
@@ -315,6 +353,7 @@ ENV PATH "/usr/local/pgsql/bin:$PATH"
 RUN apt-get update && \
     apt-get install -y cmake && \
     wget https://github.com/timescale/timescaledb/archive/refs/tags/2.10.1.tar.gz -O timescaledb.tar.gz && \
+    echo "6fca72a6ed0f6d32d2b3523951ede73dc5f9b0077b38450a029a5f411fdb8c73 timescaledb.tar.gz" | sha256sum --check && \
     mkdir timescaledb-src && cd timescaledb-src && tar xvzf ../timescaledb.tar.gz --strip-components=1 -C . && \
     ./bootstrap -DSEND_TELEMETRY_DEFAULT:BOOL=OFF -DUSE_TELEMETRY:BOOL=OFF -DAPACHE_ONLY:BOOL=ON && \
     cd build && \
@@ -323,7 +362,39 @@ RUN apt-get update && \
     echo "trusted = true" >> /usr/local/pgsql/share/extension/timescaledb.control
 
 #########################################################################################
-# 
+#
+# Layer "pg-hint-plan-pg-build"
+# compile pg_hint_plan extension
+#
+#########################################################################################
+FROM build-deps AS pg-hint-plan-pg-build
+COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
+
+ARG PG_VERSION
+ENV PATH "/usr/local/pgsql/bin:$PATH"
+
+RUN case "${PG_VERSION}" in \
+      "v14") \
+        export PG_HINT_PLAN_VERSION=14_1_4_1 \
+        export PG_HINT_PLAN_CHECKSUM=c3501becf70ead27f70626bce80ea401ceac6a77e2083ee5f3ff1f1444ec1ad1 \
+        ;; \
+      "v15") \
+        export PG_HINT_PLAN_VERSION=15_1_5_0 \
+        export PG_HINT_PLAN_CHECKSUM=564cbbf4820973ffece63fbf76e3c0af62c4ab23543142c7caaa682bc48918be \
+        ;; \
+      *) \
+        echo "Export the valid PG_HINT_PLAN_VERSION variable" && exit 1 \
+        ;; \
+    esac && \
+    wget https://github.com/ossc-db/pg_hint_plan/archive/refs/tags/REL${PG_HINT_PLAN_VERSION}.tar.gz -O pg_hint_plan.tar.gz && \
+    echo "${PG_HINT_PLAN_CHECKSUM} pg_hint_plan.tar.gz" | sha256sum --check && \
+    mkdir pg_hint_plan-src && cd pg_hint_plan-src && tar xvzf ../pg_hint_plan.tar.gz --strip-components=1 -C . && \
+    make -j $(getconf _NPROCESSORS_ONLN) && \
+    make install -j $(getconf _NPROCESSORS_ONLN) && \
+    echo "trusted = true" >> /usr/local/pgsql/share/extension/pg_hint_plan.control
+
+#########################################################################################
+#
 # Layer "rust extensions"
 # This layer is used to build `pgx` deps
 #
@@ -351,7 +422,7 @@ RUN curl -sSO https://static.rust-lang.org/rustup/dist/$(uname -m)-unknown-linux
 USER root
 
 #########################################################################################
-# 
+#
 # Layer "pg-jsonschema-pg-build"
 # Compile "pg_jsonschema" extension
 #
@@ -359,15 +430,17 @@ USER root
 
 FROM rust-extensions-build AS pg-jsonschema-pg-build
 
-# there is no release tag yet, but we need it due to the superuser fix in the control file
+# caeab60d70b2fd3ae421ec66466a3abbb37b7ee6 made on 06/03/2023
+# there is no release tag yet, but we need it due to the superuser fix in the control file, switch to git tag after release >= 0.1.5
 RUN wget https://github.com/supabase/pg_jsonschema/archive/caeab60d70b2fd3ae421ec66466a3abbb37b7ee6.tar.gz -O pg_jsonschema.tar.gz && \
+    echo "54129ce2e7ee7a585648dbb4cef6d73f795d94fe72f248ac01119992518469a4 pg_jsonschema.tar.gz" | sha256sum --check && \
     mkdir pg_jsonschema-src && cd pg_jsonschema-src && tar xvzf ../pg_jsonschema.tar.gz --strip-components=1 -C . && \
     sed -i 's/pgx = "0.7.1"/pgx = { version = "0.7.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
     cargo pgx install --release && \
     echo "trusted = true" >> /usr/local/pgsql/share/extension/pg_jsonschema.control
 
 #########################################################################################
-# 
+#
 # Layer "pg-graphql-pg-build"
 # Compile "pg_graphql" extension
 #
@@ -375,11 +448,13 @@ RUN wget https://github.com/supabase/pg_jsonschema/archive/caeab60d70b2fd3ae421e
 
 FROM rust-extensions-build AS pg-graphql-pg-build
 
+# b4988843647450a153439be367168ed09971af85 made on 22/02/2023 (from remove-pgx-contrib-spiext branch)
 # Currently pgx version bump to >= 0.7.2  causes "call to unsafe function" compliation errors in
 # pgx-contrib-spiext. There is a branch that removes that dependency, so use it. It is on the
 # same 1.1 version we've used before.
-RUN git clone -b remove-pgx-contrib-spiext --single-branch https://github.com/yrashk/pg_graphql && \
-    cd pg_graphql && \
+RUN wget https://github.com/yrashk/pg_graphql/archive/b4988843647450a153439be367168ed09971af85.tar.gz -O pg_graphql.tar.gz && \
+    echo "0c7b0e746441b2ec24187d0e03555faf935c2159e2839bddd14df6dafbc8c9bd pg_graphql.tar.gz" | sha256sum --check && \
+    mkdir pg_graphql-src && cd pg_graphql-src && tar xvzf ../pg_graphql.tar.gz --strip-components=1 -C . && \
     sed -i 's/pgx = "~0.7.1"/pgx = { version = "0.7.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
     sed -i 's/pgx-tests = "~0.7.1"/pgx-tests = "0.7.3"/g' Cargo.toml && \
     cargo pgx install --release && \
@@ -396,8 +471,10 @@ RUN git clone -b remove-pgx-contrib-spiext --single-branch https://github.com/yr
 
 FROM rust-extensions-build AS pg-tiktoken-pg-build
 
-RUN git clone --depth=1 --single-branch https://github.com/kelvich/pg_tiktoken && \
-    cd pg_tiktoken && \
+# 801f84f08c6881c8aa30f405fafbf00eec386a72 made on 10/03/2023
+RUN wget https://github.com/kelvich/pg_tiktoken/archive/801f84f08c6881c8aa30f405fafbf00eec386a72.tar.gz -O pg_tiktoken.tar.gz && \
+    echo "52f60ac800993a49aa8c609961842b611b6b1949717b69ce2ec9117117e16e4a pg_tiktoken.tar.gz" | sha256sum --check && \
+    mkdir pg_tiktoken-src && cd pg_tiktoken-src && tar xvzf ../pg_tiktoken.tar.gz --strip-components=1 -C . && \
     cargo pgx install --release && \
     echo "trusted = true" >> /usr/local/pgsql/share/extension/pg_tiktoken.control
 
@@ -423,10 +500,12 @@ COPY --from=hypopg-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pg-hashids-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=rum-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pgtap-pg-build /usr/local/pgsql/ /usr/local/pgsql/
+COPY --from=ip4r-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=prefix-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=hll-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=plpgsql-check-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=timescaledb-pg-build /usr/local/pgsql/ /usr/local/pgsql/
+COPY --from=pg-hint-plan-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY pgxn/ pgxn/
 
 RUN make -j $(getconf _NPROCESSORS_ONLN) \
@@ -491,13 +570,17 @@ COPY --from=compute-tools --chown=postgres /home/nonroot/target/release-line-deb
 # Install:
 # libreadline8 for psql
 # libicu67, locales for collations (including ICU and plpgsql_check)
+# liblz4-1 for lz4
 # libossp-uuid16 for extension ossp-uuid
 # libgeos, libgdal, libsfcgal1, libproj and libprotobuf-c1 for PostGIS
 # libxml2, libxslt1.1 for xml2
+# libzstd1 for zstd
 RUN apt update &&  \
     apt install --no-install-recommends -y \
+        gdb \
         locales \
         libicu67 \
+        liblz4-1 \
         libreadline8 \
         libossp-uuid16 \
         libgeos-c1v5 \
@@ -507,7 +590,8 @@ RUN apt update &&  \
         libsfcgal1 \
         libxml2 \
         libxslt1.1 \
-        gdb && \
+        libzstd1 \
+        procps && \
     rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/* && \
     localedef -i en_US -c -f UTF-8 -A /usr/share/locale/locale.alias en_US.UTF-8
 
diff --git a/Dockerfile.vm-compute-node b/Dockerfile.vm-compute-node
index 957166ecd1..aabb3c9953 100644
--- a/Dockerfile.vm-compute-node
+++ b/Dockerfile.vm-compute-node
@@ -54,7 +54,7 @@ RUN set -e \
 
 RUN set -e \
 	&& echo "::sysinit:cgconfigparser -l /etc/cgconfig.conf -s 1664" >> /etc/inittab \
-	&& CONNSTR="dbname=neondb user=cloud_admin sslmode=disable" \
+	&& CONNSTR="dbname=postgres user=cloud_admin sslmode=disable" \
 	&& ARGS="--auto-restart --cgroup=neon-postgres --pgconnstr=\"$CONNSTR\"" \
 	&& echo "::respawn:su vm-informant -c '/usr/local/bin/vm-informant $ARGS'" >> /etc/inittab
 
diff --git a/README.md b/README.md
index 55df67f6c7..810937aff7 100644
--- a/README.md
+++ b/README.md
@@ -147,15 +147,15 @@ Created an initial timeline 'de200bd42b49cc1814412c7e592dd6e9' at Lsn 0/16B5A50
 Setting tenant 9ef87a5bf0d92544f6fafeeb3239695c as a default one
 
 # start postgres compute node
-> ./target/debug/neon_local pg start main
-Starting new postgres (v14) main on timeline de200bd42b49cc1814412c7e592dd6e9 ...
+> ./target/debug/neon_local endpoint start main
+Starting new endpoint main (PostgreSQL v14) on timeline de200bd42b49cc1814412c7e592dd6e9 ...
 Extracting base backup to create postgres instance: path=.neon/pgdatadirs/tenants/9ef87a5bf0d92544f6fafeeb3239695c/main port=55432
-Starting postgres node at 'host=127.0.0.1 port=55432 user=cloud_admin dbname=postgres'
+Starting postgres at 'host=127.0.0.1 port=55432 user=cloud_admin dbname=postgres'
 
 # check list of running postgres instances
-> ./target/debug/neon_local pg list
- NODE  ADDRESS          TIMELINE                          BRANCH NAME  LSN        STATUS
- main  127.0.0.1:55432  de200bd42b49cc1814412c7e592dd6e9  main         0/16B5BA8  running
+> ./target/debug/neon_local endpoint list
+ ENDPOINT  ADDRESS          TIMELINE                          BRANCH NAME  LSN        STATUS
+ main      127.0.0.1:55432  de200bd42b49cc1814412c7e592dd6e9  main         0/16B5BA8  running
 ```
 
 2. Now, it is possible to connect to postgres and run some queries:
@@ -184,14 +184,14 @@ Created timeline 'b3b863fa45fa9e57e615f9f2d944e601' at Lsn 0/16F9A00 for tenant:
 (L) ┗━ @0/16F9A00: migration_check [b3b863fa45fa9e57e615f9f2d944e601]
 
 # start postgres on that branch
-> ./target/debug/neon_local pg start migration_check --branch-name migration_check
-Starting new postgres migration_check on timeline b3b863fa45fa9e57e615f9f2d944e601 ...
+> ./target/debug/neon_local endpoint start migration_check --branch-name migration_check
+Starting new endpoint migration_check (PostgreSQL v14) on timeline b3b863fa45fa9e57e615f9f2d944e601 ...
 Extracting base backup to create postgres instance: path=.neon/pgdatadirs/tenants/9ef87a5bf0d92544f6fafeeb3239695c/migration_check port=55433
-Starting postgres node at 'host=127.0.0.1 port=55433 user=cloud_admin dbname=postgres'
+Starting postgres at 'host=127.0.0.1 port=55433 user=cloud_admin dbname=postgres'
 
 # check the new list of running postgres instances
-> ./target/debug/neon_local pg list
- NODE             ADDRESS          TIMELINE                          BRANCH NAME      LSN        STATUS
+> ./target/debug/neon_local endpoint list
+ ENDPOINT         ADDRESS          TIMELINE                          BRANCH NAME      LSN        STATUS
  main             127.0.0.1:55432  de200bd42b49cc1814412c7e592dd6e9  main             0/16F9A38  running
  migration_check  127.0.0.1:55433  b3b863fa45fa9e57e615f9f2d944e601  migration_check  0/16F9A70  running
 
diff --git a/compute_tools/Cargo.toml b/compute_tools/Cargo.toml
index f315d2b7d9..21226249cf 100644
--- a/compute_tools/Cargo.toml
+++ b/compute_tools/Cargo.toml
@@ -28,4 +28,5 @@ tracing-utils.workspace = true
 url.workspace = true
 
 compute_api.workspace = true
+utils.workspace = true
 workspace_hack.workspace = true
diff --git a/compute_tools/src/bin/compute_ctl.rs b/compute_tools/src/bin/compute_ctl.rs
index d61eae5f7a..36dbc382b5 100644
--- a/compute_tools/src/bin/compute_ctl.rs
+++ b/compute_tools/src/bin/compute_ctl.rs
@@ -34,7 +34,7 @@ use std::fs::File;
 use std::panic;
 use std::path::Path;
 use std::process::exit;
-use std::sync::{Arc, Condvar, Mutex};
+use std::sync::{mpsc, Arc, Condvar, Mutex};
 use std::{thread, time::Duration};
 
 use anyhow::{Context, Result};
@@ -45,12 +45,12 @@ use url::Url;
 
 use compute_api::responses::ComputeStatus;
 
-use compute_tools::compute::{ComputeNode, ComputeState};
+use compute_tools::compute::{ComputeNode, ComputeState, ParsedSpec};
+use compute_tools::configurator::launch_configurator;
 use compute_tools::http::api::launch_http_server;
 use compute_tools::logger::*;
 use compute_tools::monitor::launch_monitor;
 use compute_tools::params::*;
-use compute_tools::pg_helpers::*;
 use compute_tools::spec::*;
 
 fn main() -> Result<()> {
@@ -73,29 +73,29 @@ fn main() -> Result<()> {
     // Try to use just 'postgres' if no path is provided
     let pgbin = matches.get_one::<String>("pgbin").unwrap();
 
-    let mut spec = Default::default();
-    let mut spec_set = false;
+    let spec;
     let mut live_config_allowed = false;
     match spec_json {
         // First, try to get cluster spec from the cli argument
         Some(json) => {
-            spec = serde_json::from_str(json)?;
-            spec_set = true;
+            spec = Some(serde_json::from_str(json)?);
         }
         None => {
             // Second, try to read it from the file if path is provided
             if let Some(sp) = spec_path {
                 let path = Path::new(sp);
                 let file = File::open(path)?;
-                spec = serde_json::from_reader(file)?;
-                spec_set = true;
+                spec = Some(serde_json::from_reader(file)?);
             } else if let Some(id) = compute_id {
                 if let Some(cp_base) = control_plane_uri {
                     live_config_allowed = true;
-                    if let Ok(s) = get_spec_from_control_plane(cp_base, id) {
-                        spec = s;
-                        spec_set = true;
-                    }
+                    spec = match get_spec_from_control_plane(cp_base, id) {
+                        Ok(s) => s,
+                        Err(e) => {
+                            error!("cannot get response from control plane: {}", e);
+                            panic!("neither spec nor confirmation that compute is in the Empty state was received");
+                        }
+                    };
                 } else {
                     panic!("must specify both --control-plane-uri and --compute-id or none");
                 }
@@ -109,11 +109,15 @@ fn main() -> Result<()> {
     };
 
     let mut new_state = ComputeState::new();
-    if spec_set {
-        new_state.spec = spec;
+    let spec_set;
+    if let Some(spec) = spec {
+        let pspec = ParsedSpec::try_from(spec).map_err(|msg| anyhow::anyhow!(msg))?;
+        new_state.pspec = Some(pspec);
+        spec_set = true;
+    } else {
+        spec_set = false;
     }
     let compute_node = ComputeNode {
-        start_time: Utc::now(),
         connstr: Url::parse(connstr).context("cannot parse connstr as a URL")?,
         pgdata: pgdata.to_string(),
         pgbin: pgbin.to_string(),
@@ -142,33 +146,21 @@ fn main() -> Result<()> {
         }
     }
 
-    // We got all we need, fill in the state.
+    // We got all we need, update the state.
     let mut state = compute.state.lock().unwrap();
-    let pageserver_connstr = state
-        .spec
-        .cluster
-        .settings
-        .find("neon.pageserver_connstring")
-        .expect("pageserver connstr should be provided");
-    let storage_auth_token = state.spec.storage_auth_token.clone();
-    let tenant = state
-        .spec
-        .cluster
-        .settings
-        .find("neon.tenant_id")
-        .expect("tenant id should be provided");
-    let timeline = state
-        .spec
-        .cluster
-        .settings
-        .find("neon.timeline_id")
-        .expect("tenant id should be provided");
-    let startup_tracing_context = state.spec.startup_tracing_context.clone();
+    let pspec = state.pspec.as_ref().expect("spec must be set");
+    let startup_tracing_context = pspec.spec.startup_tracing_context.clone();
+
+    // Record for how long we slept waiting for the spec.
+    state.metrics.wait_for_spec_ms = Utc::now()
+        .signed_duration_since(state.start_time)
+        .to_std()
+        .unwrap()
+        .as_millis() as u64;
+    // Reset start time to the actual start of the configuration, so that
+    // total startup time was properly measured at the end.
+    state.start_time = Utc::now();
 
-    state.pageserver_connstr = pageserver_connstr;
-    state.storage_auth_token = storage_auth_token;
-    state.tenant = tenant;
-    state.timeline = timeline;
     state.status = ComputeStatus::Init;
     compute.state_changed.notify_all();
     drop(state);
@@ -198,6 +190,8 @@ fn main() -> Result<()> {
 
     // Launch remaining service threads
     let _monitor_handle = launch_monitor(&compute).expect("cannot launch compute monitor thread");
+    let _configurator_handle =
+        launch_configurator(&compute).expect("cannot launch configurator thread");
 
     // Start Postgres
     let mut delay_exit = false;
@@ -239,10 +233,25 @@ fn main() -> Result<()> {
         thread::sleep(Duration::from_secs(30));
     }
 
-    info!("shutting down tracing");
     // Shutdown trace pipeline gracefully, so that it has a chance to send any
-    // pending traces before we exit.
-    tracing_utils::shutdown_tracing();
+    // pending traces before we exit. Shutting down OTEL tracing provider may
+    // hang for quite some time, see, for example:
+    // - https://github.com/open-telemetry/opentelemetry-rust/issues/868
+    // - and our problems with staging https://github.com/neondatabase/cloud/issues/3707#issuecomment-1493983636
+    //
+    // Yet, we want computes to shut down fast enough, as we may need a new one
+    // for the same timeline ASAP. So wait no longer than 2s for the shutdown to
+    // complete, then just error out and exit the main thread.
+    info!("shutting down tracing");
+    let (sender, receiver) = mpsc::channel();
+    let _ = thread::spawn(move || {
+        tracing_utils::shutdown_tracing();
+        sender.send(()).ok()
+    });
+    let shutdown_res = receiver.recv_timeout(Duration::from_millis(2000));
+    if shutdown_res.is_err() {
+        error!("timed out while shutting down tracing, exiting anyway");
+    }
 
     info!("shutting down");
     exit(exit_code.unwrap_or(1))
diff --git a/compute_tools/src/checker.rs b/compute_tools/src/checker.rs
index b8413de516..b6a287bdeb 100644
--- a/compute_tools/src/checker.rs
+++ b/compute_tools/src/checker.rs
@@ -1,12 +1,28 @@
 use anyhow::{anyhow, Result};
-use postgres::Client;
 use tokio_postgres::NoTls;
 use tracing::{error, instrument};
 
 use crate::compute::ComputeNode;
 
+/// Update timestamp in a row in a special service table to check
+/// that we can actually write some data in this particular timeline.
+/// Create table if it's missing.
 #[instrument(skip_all)]
-pub fn create_writability_check_data(client: &mut Client) -> Result<()> {
+pub async fn check_writability(compute: &ComputeNode) -> Result<()> {
+    // Connect to the database.
+    let (client, connection) = tokio_postgres::connect(compute.connstr.as_str(), NoTls).await?;
+    if client.is_closed() {
+        return Err(anyhow!("connection to postgres closed"));
+    }
+
+    // The connection object performs the actual communication with the database,
+    // so spawn it off to run on its own.
+    tokio::spawn(async move {
+        if let Err(e) = connection.await {
+            error!("connection error: {}", e);
+        }
+    });
+
     let query = "
     CREATE TABLE IF NOT EXISTS health_check (
         id serial primary key,
@@ -15,31 +31,15 @@ pub fn create_writability_check_data(client: &mut Client) -> Result<()> {
     INSERT INTO health_check VALUES (1, now())
         ON CONFLICT (id) DO UPDATE
          SET updated_at = now();";
-    let result = client.simple_query(query)?;
-    if result.len() < 2 {
-        return Err(anyhow::format_err!("executed  {} queries", result.len()));
-    }
-    Ok(())
-}
-
-#[instrument(skip_all)]
-pub async fn check_writability(compute: &ComputeNode) -> Result<()> {
-    let (client, connection) = tokio_postgres::connect(compute.connstr.as_str(), NoTls).await?;
-    if client.is_closed() {
-        return Err(anyhow!("connection to postgres closed"));
-    }
-    tokio::spawn(async move {
-        if let Err(e) = connection.await {
-            error!("connection error: {}", e);
-        }
-    });
-
-    let result = client
-        .simple_query("UPDATE health_check SET updated_at = now() WHERE id = 1;")
-        .await?;
-
-    if result.len() != 1 {
-        return Err(anyhow!("statement can't be executed"));
+
+    let result = client.simple_query(query).await?;
+
+    if result.len() != 2 {
+        return Err(anyhow::format_err!(
+            "expected 2 query results, but got {}",
+            result.len()
+        ));
     }
+
     Ok(())
 }
diff --git a/compute_tools/src/compute.rs b/compute_tools/src/compute.rs
index 689aa6ef43..b6bc234beb 100644
--- a/compute_tools/src/compute.rs
+++ b/compute_tools/src/compute.rs
@@ -26,18 +26,18 @@ use chrono::{DateTime, Utc};
 use postgres::{Client, NoTls};
 use tokio_postgres;
 use tracing::{info, instrument, warn};
+use utils::id::{TenantId, TimelineId};
+use utils::lsn::Lsn;
 
 use compute_api::responses::{ComputeMetrics, ComputeStatus};
 use compute_api::spec::ComputeSpec;
 
-use crate::checker::create_writability_check_data;
 use crate::config;
 use crate::pg_helpers::*;
 use crate::spec::*;
 
 /// Compute node info shared across several `compute_ctl` threads.
 pub struct ComputeNode {
-    pub start_time: DateTime<Utc>,
     // Url type maintains proper escaping
     pub connstr: url::Url,
     pub pgdata: String,
@@ -65,30 +65,23 @@ pub struct ComputeNode {
 
 #[derive(Clone, Debug)]
 pub struct ComputeState {
+    pub start_time: DateTime<Utc>,
     pub status: ComputeStatus,
     /// Timestamp of the last Postgres activity
     pub last_active: DateTime<Utc>,
     pub error: Option<String>,
-    pub spec: ComputeSpec,
-    pub tenant: String,
-    pub timeline: String,
-    pub pageserver_connstr: String,
-    pub storage_auth_token: Option<String>,
-
+    pub pspec: Option<ParsedSpec>,
     pub metrics: ComputeMetrics,
 }
 
 impl ComputeState {
     pub fn new() -> Self {
         Self {
+            start_time: Utc::now(),
             status: ComputeStatus::Empty,
             last_active: Utc::now(),
             error: None,
-            spec: ComputeSpec::default(),
-            tenant: String::new(),
-            timeline: String::new(),
-            pageserver_connstr: String::new(),
-            storage_auth_token: None,
+            pspec: None,
             metrics: ComputeMetrics::default(),
         }
     }
@@ -100,6 +93,49 @@ impl Default for ComputeState {
     }
 }
 
+#[derive(Clone, Debug)]
+pub struct ParsedSpec {
+    pub spec: ComputeSpec,
+    pub tenant_id: TenantId,
+    pub timeline_id: TimelineId,
+    pub pageserver_connstr: String,
+    pub storage_auth_token: Option<String>,
+}
+
+impl TryFrom<ComputeSpec> for ParsedSpec {
+    type Error = String;
+    fn try_from(spec: ComputeSpec) -> Result<Self, String> {
+        let pageserver_connstr = spec
+            .cluster
+            .settings
+            .find("neon.pageserver_connstring")
+            .ok_or("pageserver connstr should be provided")?;
+        let storage_auth_token = spec.storage_auth_token.clone();
+        let tenant_id: TenantId = spec
+            .cluster
+            .settings
+            .find("neon.tenant_id")
+            .ok_or("tenant id should be provided")
+            .map(|s| TenantId::from_str(&s))?
+            .or(Err("invalid tenant id"))?;
+        let timeline_id: TimelineId = spec
+            .cluster
+            .settings
+            .find("neon.timeline_id")
+            .ok_or("timeline id should be provided")
+            .map(|s| TimelineId::from_str(&s))?
+            .or(Err("invalid timeline id"))?;
+
+        Ok(ParsedSpec {
+            spec,
+            pageserver_connstr,
+            storage_auth_token,
+            tenant_id,
+            timeline_id,
+        })
+    }
+}
+
 impl ComputeNode {
     pub fn set_status(&self, status: ComputeStatus) {
         let mut state = self.state.lock().unwrap();
@@ -125,14 +161,15 @@ impl ComputeNode {
     // Get basebackup from the libpq connection to pageserver using `connstr` and
     // unarchive it to `pgdata` directory overriding all its previous content.
     #[instrument(skip(self, compute_state))]
-    fn get_basebackup(&self, compute_state: &ComputeState, lsn: &str) -> Result<()> {
+    fn get_basebackup(&self, compute_state: &ComputeState, lsn: Lsn) -> Result<()> {
+        let spec = compute_state.pspec.as_ref().expect("spec must be set");
         let start_time = Utc::now();
 
-        let mut config = postgres::Config::from_str(&compute_state.pageserver_connstr)?;
+        let mut config = postgres::Config::from_str(&spec.pageserver_connstr)?;
 
         // Use the storage auth token from the config file, if given.
         // Note: this overrides any password set in the connection string.
-        if let Some(storage_auth_token) = &compute_state.storage_auth_token {
+        if let Some(storage_auth_token) = &spec.storage_auth_token {
             info!("Got storage auth token from spec file");
             config.password(storage_auth_token);
         } else {
@@ -141,14 +178,8 @@ impl ComputeNode {
 
         let mut client = config.connect(NoTls)?;
         let basebackup_cmd = match lsn {
-            "0/0" => format!(
-                "basebackup {} {}",
-                &compute_state.tenant, &compute_state.timeline
-            ), // First start of the compute
-            _ => format!(
-                "basebackup {} {} {}",
-                &compute_state.tenant, &compute_state.timeline, lsn
-            ),
+            Lsn(0) => format!("basebackup {} {}", spec.tenant_id, spec.timeline_id), // First start of the compute
+            _ => format!("basebackup {} {} {}", spec.tenant_id, spec.timeline_id, lsn),
         };
         let copyreader = client.copy_out(basebackup_cmd.as_str())?;
 
@@ -172,7 +203,7 @@ impl ComputeNode {
     // Run `postgres` in a special mode with `--sync-safekeepers` argument
     // and return the reported LSN back to the caller.
     #[instrument(skip(self, storage_auth_token))]
-    fn sync_safekeepers(&self, storage_auth_token: Option<String>) -> Result<String> {
+    fn sync_safekeepers(&self, storage_auth_token: Option<String>) -> Result<Lsn> {
         let start_time = Utc::now();
 
         let sync_handle = Command::new(&self.pgbin)
@@ -209,7 +240,7 @@ impl ComputeNode {
             .unwrap()
             .as_millis() as u64;
 
-        let lsn = String::from(String::from_utf8(sync_output.stdout)?.trim());
+        let lsn = Lsn::from_str(String::from_utf8(sync_output.stdout)?.trim())?;
 
         Ok(lsn)
     }
@@ -218,33 +249,85 @@ impl ComputeNode {
     /// safekeepers sync, basebackup, etc.
     #[instrument(skip(self, compute_state))]
     pub fn prepare_pgdata(&self, compute_state: &ComputeState) -> Result<()> {
-        let spec = &compute_state.spec;
+        #[derive(Clone)]
+        enum Replication {
+            Primary,
+            Static { lsn: Lsn },
+            HotStandby,
+        }
+
+        let pspec = compute_state.pspec.as_ref().expect("spec must be set");
+        let spec = &pspec.spec;
         let pgdata_path = Path::new(&self.pgdata);
 
+        let hot_replica = if let Some(option) = spec.cluster.settings.find_ref("hot_standby") {
+            if let Some(value) = &option.value {
+                anyhow::ensure!(option.vartype == "bool");
+                matches!(value.as_str(), "on" | "yes" | "true")
+            } else {
+                false
+            }
+        } else {
+            false
+        };
+
+        let replication = if hot_replica {
+            Replication::HotStandby
+        } else if let Some(lsn) = spec.cluster.settings.find("recovery_target_lsn") {
+            Replication::Static {
+                lsn: Lsn::from_str(&lsn)?,
+            }
+        } else {
+            Replication::Primary
+        };
+
         // Remove/create an empty pgdata directory and put configuration there.
         self.create_pgdata()?;
-        config::write_postgres_conf(&pgdata_path.join("postgresql.conf"), spec)?;
+        config::write_postgres_conf(&pgdata_path.join("postgresql.conf"), &pspec.spec)?;
 
-        info!("starting safekeepers syncing");
-        let lsn = self
-            .sync_safekeepers(compute_state.storage_auth_token.clone())
-            .with_context(|| "failed to sync safekeepers")?;
-        info!("safekeepers synced at LSN {}", lsn);
+        // Syncing safekeepers is only safe with primary nodes: if a primary
+        // is already connected it will be kicked out, so a secondary (standby)
+        // cannot sync safekeepers.
+        let lsn = match &replication {
+            Replication::Primary => {
+                info!("starting safekeepers syncing");
+                let lsn = self
+                    .sync_safekeepers(pspec.storage_auth_token.clone())
+                    .with_context(|| "failed to sync safekeepers")?;
+                info!("safekeepers synced at LSN {}", lsn);
+                lsn
+            }
+            Replication::Static { lsn } => {
+                info!("Starting read-only node at static LSN {}", lsn);
+                *lsn
+            }
+            Replication::HotStandby => {
+                info!("Initializing standby from latest Pageserver LSN");
+                Lsn(0)
+            }
+        };
 
         info!(
             "getting basebackup@{} from pageserver {}",
-            lsn, &compute_state.pageserver_connstr
+            lsn, &pspec.pageserver_connstr
         );
-        self.get_basebackup(compute_state, &lsn).with_context(|| {
+        self.get_basebackup(compute_state, lsn).with_context(|| {
             format!(
                 "failed to get basebackup@{} from pageserver {}",
-                lsn, &compute_state.pageserver_connstr
+                lsn, &pspec.pageserver_connstr
             )
         })?;
 
         // Update pg_hba.conf received with basebackup.
         update_pg_hba(pgdata_path)?;
 
+        match &replication {
+            Replication::Primary | Replication::Static { .. } => {}
+            Replication::HotStandby => {
+                add_standby_signal(pgdata_path)?;
+            }
+        }
+
         Ok(())
     }
 
@@ -306,19 +389,61 @@ impl ComputeNode {
         };
 
         // Proceed with post-startup configuration. Note, that order of operations is important.
-        handle_roles(&compute_state.spec, &mut client)?;
-        handle_databases(&compute_state.spec, &mut client)?;
-        handle_role_deletions(&compute_state.spec, self.connstr.as_str(), &mut client)?;
-        handle_grants(&compute_state.spec, self.connstr.as_str(), &mut client)?;
-        create_writability_check_data(&mut client)?;
-        handle_extensions(&compute_state.spec, &mut client)?;
+        let spec = &compute_state.pspec.as_ref().expect("spec must be set").spec;
+        handle_roles(spec, &mut client)?;
+        handle_databases(spec, &mut client)?;
+        handle_role_deletions(spec, self.connstr.as_str(), &mut client)?;
+        handle_grants(spec, self.connstr.as_str(), &mut client)?;
+        handle_extensions(spec, &mut client)?;
 
         // 'Close' connection
         drop(client);
 
         info!(
             "finished configuration of compute for project {}",
-            compute_state.spec.cluster.cluster_id
+            spec.cluster.cluster_id
+        );
+
+        Ok(())
+    }
+
+    // We could've wrapped this around `pg_ctl reload`, but right now we don't use
+    // `pg_ctl` for start / stop, so this just seems much easier to do as we already
+    // have opened connection to Postgres and superuser access.
+    #[instrument(skip(self, client))]
+    fn pg_reload_conf(&self, client: &mut Client) -> Result<()> {
+        client.simple_query("SELECT pg_reload_conf()")?;
+        Ok(())
+    }
+
+    /// Similar to `apply_config()`, but does a bit different sequence of operations,
+    /// as it's used to reconfigure a previously started and configured Postgres node.
+    #[instrument(skip(self))]
+    pub fn reconfigure(&self) -> Result<()> {
+        let spec = self.state.lock().unwrap().pspec.clone().unwrap().spec;
+
+        // Write new config
+        let pgdata_path = Path::new(&self.pgdata);
+        config::write_postgres_conf(&pgdata_path.join("postgresql.conf"), &spec)?;
+
+        let mut client = Client::connect(self.connstr.as_str(), NoTls)?;
+        self.pg_reload_conf(&mut client)?;
+
+        // Proceed with post-startup configuration. Note, that order of operations is important.
+        handle_roles(&spec, &mut client)?;
+        handle_databases(&spec, &mut client)?;
+        handle_role_deletions(&spec, self.connstr.as_str(), &mut client)?;
+        handle_grants(&spec, self.connstr.as_str(), &mut client)?;
+        handle_extensions(&spec, &mut client)?;
+
+        // 'Close' connection
+        drop(client);
+
+        let unknown_op = "unknown".to_string();
+        let op_id = spec.operation_uuid.as_ref().unwrap_or(&unknown_op);
+        info!(
+            "finished reconfiguration of compute node for operation {}",
+            op_id
         );
 
         Ok(())
@@ -327,19 +452,20 @@ impl ComputeNode {
     #[instrument(skip(self))]
     pub fn start_compute(&self) -> Result<std::process::Child> {
         let compute_state = self.state.lock().unwrap().clone();
+        let spec = compute_state.pspec.as_ref().expect("spec must be set");
         info!(
             "starting compute for project {}, operation {}, tenant {}, timeline {}",
-            compute_state.spec.cluster.cluster_id,
-            compute_state.spec.operation_uuid.as_ref().unwrap(),
-            compute_state.tenant,
-            compute_state.timeline,
+            spec.spec.cluster.cluster_id,
+            spec.spec.operation_uuid.as_deref().unwrap_or("None"),
+            spec.tenant_id,
+            spec.timeline_id,
         );
 
         self.prepare_pgdata(&compute_state)?;
 
         let start_time = Utc::now();
 
-        let pg = self.start_postgres(compute_state.storage_auth_token.clone())?;
+        let pg = self.start_postgres(spec.storage_auth_token.clone())?;
 
         self.apply_config(&compute_state)?;
 
@@ -352,7 +478,7 @@ impl ComputeNode {
                 .unwrap()
                 .as_millis() as u64;
             state.metrics.total_startup_ms = startup_end_time
-                .signed_duration_since(self.start_time)
+                .signed_duration_since(compute_state.start_time)
                 .to_std()
                 .unwrap()
                 .as_millis() as u64;
diff --git a/compute_tools/src/configurator.rs b/compute_tools/src/configurator.rs
new file mode 100644
index 0000000000..a07fd0b8cd
--- /dev/null
+++ b/compute_tools/src/configurator.rs
@@ -0,0 +1,54 @@
+use std::sync::Arc;
+use std::thread;
+
+use anyhow::Result;
+use tracing::{error, info, instrument};
+
+use compute_api::responses::ComputeStatus;
+
+use crate::compute::ComputeNode;
+
+#[instrument(skip(compute))]
+fn configurator_main_loop(compute: &Arc<ComputeNode>) {
+    info!("waiting for reconfiguration requests");
+    loop {
+        let state = compute.state.lock().unwrap();
+        let mut state = compute.state_changed.wait(state).unwrap();
+
+        if state.status == ComputeStatus::ConfigurationPending {
+            info!("got configuration request");
+            state.status = ComputeStatus::Configuration;
+            compute.state_changed.notify_all();
+            drop(state);
+
+            let mut new_status = ComputeStatus::Failed;
+            if let Err(e) = compute.reconfigure() {
+                error!("could not configure compute node: {}", e);
+            } else {
+                new_status = ComputeStatus::Running;
+                info!("compute node configured");
+            }
+
+            // XXX: used to test that API is blocking
+            // std::thread::sleep(std::time::Duration::from_millis(10000));
+
+            compute.set_status(new_status);
+        } else if state.status == ComputeStatus::Failed {
+            info!("compute node is now in Failed state, exiting");
+            break;
+        } else {
+            info!("woken up for compute status: {:?}, sleeping", state.status);
+        }
+    }
+}
+
+pub fn launch_configurator(compute: &Arc<ComputeNode>) -> Result<thread::JoinHandle<()>> {
+    let compute = Arc::clone(compute);
+
+    Ok(thread::Builder::new()
+        .name("compute-configurator".into())
+        .spawn(move || {
+            configurator_main_loop(&compute);
+            info!("configurator thread is exited");
+        })?)
+}
diff --git a/compute_tools/src/http/api.rs b/compute_tools/src/http/api.rs
index cea45dc596..4468f6f5e4 100644
--- a/compute_tools/src/http/api.rs
+++ b/compute_tools/src/http/api.rs
@@ -3,7 +3,7 @@ use std::net::SocketAddr;
 use std::sync::Arc;
 use std::thread;
 
-use crate::compute::{ComputeNode, ComputeState};
+use crate::compute::{ComputeNode, ComputeState, ParsedSpec};
 use compute_api::requests::ConfigurationRequest;
 use compute_api::responses::{ComputeStatus, ComputeStatusResponse, GenericAPIError};
 
@@ -18,8 +18,15 @@ use tracing_utils::http::OtelName;
 
 fn status_response_from_state(state: &ComputeState) -> ComputeStatusResponse {
     ComputeStatusResponse {
-        tenant: state.tenant.clone(),
-        timeline: state.timeline.clone(),
+        start_time: state.start_time,
+        tenant: state
+            .pspec
+            .as_ref()
+            .map(|pspec| pspec.tenant_id.to_string()),
+        timeline: state
+            .pspec
+            .as_ref()
+            .map(|pspec| pspec.timeline_id.to_string()),
         status: state.status,
         last_active: state.last_active,
         error: state.error.clone(),
@@ -79,7 +86,10 @@ async fn routes(req: Request<Body>, compute: &Arc<ComputeNode>) -> Response<Body
             let res = crate::checker::check_writability(compute).await;
             match res {
                 Ok(_) => Response::new(Body::from("true")),
-                Err(e) => Response::new(Body::from(e.to_string())),
+                Err(e) => {
+                    error!("check_writability failed: {}", e);
+                    Response::new(Body::from(e.to_string()))
+                }
             }
         }
 
@@ -135,6 +145,12 @@ async fn handle_configure_request(
     let spec_raw = String::from_utf8(body_bytes.to_vec()).unwrap();
     if let Ok(request) = serde_json::from_str::<ConfigurationRequest>(&spec_raw) {
         let spec = request.spec;
+
+        let parsed_spec = match ParsedSpec::try_from(spec) {
+            Ok(ps) => ps,
+            Err(msg) => return Err((msg, StatusCode::PRECONDITION_FAILED)),
+        };
+
         // XXX: wrap state update under lock in code blocks. Otherwise,
         // we will try to `Send` `mut state` into the spawned thread
         // bellow, which will cause error:
@@ -143,14 +159,14 @@ async fn handle_configure_request(
         // ```
         {
             let mut state = compute.state.lock().unwrap();
-            if state.status != ComputeStatus::Empty {
+            if state.status != ComputeStatus::Empty && state.status != ComputeStatus::Running {
                 let msg = format!(
                     "invalid compute status for configuration request: {:?}",
                     state.status.clone()
                 );
                 return Err((msg, StatusCode::PRECONDITION_FAILED));
             }
-            state.spec = spec;
+            state.pspec = Some(parsed_spec);
             state.status = ComputeStatus::ConfigurationPending;
             compute.state_changed.notify_all();
             drop(state);
diff --git a/compute_tools/src/http/openapi_spec.yaml b/compute_tools/src/http/openapi_spec.yaml
index bdb09d4a6b..cc8f074a50 100644
--- a/compute_tools/src/http/openapi_spec.yaml
+++ b/compute_tools/src/http/openapi_spec.yaml
@@ -152,11 +152,14 @@ components:
       type: object
       description: Compute startup metrics.
       required:
+        - wait_for_spec_ms
         - sync_safekeepers_ms
         - basebackup_ms
         - config_ms
         - total_startup_ms
       properties:
+        wait_for_spec_ms:
+          type: integer
         sync_safekeepers_ms:
           type: integer
         basebackup_ms:
@@ -181,6 +184,13 @@ components:
         - status
         - last_active
       properties:
+        start_time:
+          type: string
+          description: |
+            Time when compute was started. If initially compute was started in the `empty`
+            state and then provided with valid spec, `start_time` will be reset to the
+            moment, when spec was received.
+          example: "2022-10-12T07:20:50.52Z"
         status:
           $ref: '#/components/schemas/ComputeStatus'
         last_active:
diff --git a/compute_tools/src/lib.rs b/compute_tools/src/lib.rs
index aee6b53e6a..24811f75ee 100644
--- a/compute_tools/src/lib.rs
+++ b/compute_tools/src/lib.rs
@@ -4,6 +4,7 @@
 //!
 pub mod checker;
 pub mod config;
+pub mod configurator;
 pub mod http;
 #[macro_use]
 pub mod logger;
diff --git a/compute_tools/src/pg_helpers.rs b/compute_tools/src/pg_helpers.rs
index bb787d0506..40dbea6907 100644
--- a/compute_tools/src/pg_helpers.rs
+++ b/compute_tools/src/pg_helpers.rs
@@ -94,6 +94,7 @@ impl PgOptionsSerialize for GenericOptions {
 
 pub trait GenericOptionsSearch {
     fn find(&self, name: &str) -> Option<String>;
+    fn find_ref(&self, name: &str) -> Option<&GenericOption>;
 }
 
 impl GenericOptionsSearch for GenericOptions {
@@ -103,6 +104,12 @@ impl GenericOptionsSearch for GenericOptions {
         let op = ops.iter().find(|s| s.name == name)?;
         op.value.clone()
     }
+
+    /// Lookup option by name, returning ref
+    fn find_ref(&self, name: &str) -> Option<&GenericOption> {
+        let ops = self.as_ref()?;
+        ops.iter().find(|s| s.name == name)
+    }
 }
 
 pub trait RoleExt {
diff --git a/compute_tools/src/spec.rs b/compute_tools/src/spec.rs
index 2350113c39..bf3c407202 100644
--- a/compute_tools/src/spec.rs
+++ b/compute_tools/src/spec.rs
@@ -1,38 +1,121 @@
+use std::fs::File;
 use std::path::Path;
 use std::str::FromStr;
 
-use anyhow::Result;
+use anyhow::{anyhow, bail, Result};
 use postgres::config::Config;
 use postgres::{Client, NoTls};
-use tracing::{info, info_span, instrument, span_enabled, warn, Level};
+use reqwest::StatusCode;
+use tracing::{error, info, info_span, instrument, span_enabled, warn, Level};
 
 use crate::config;
 use crate::params::PG_HBA_ALL_MD5;
 use crate::pg_helpers::*;
 
+use compute_api::responses::{ControlPlaneComputeStatus, ControlPlaneSpecResponse};
 use compute_api::spec::{ComputeSpec, Database, PgIdent, Role};
 
+// Do control plane request and return response if any. In case of error it
+// returns a bool flag indicating whether it makes sense to retry the request
+// and a string with error message.
+fn do_control_plane_request(
+    uri: &str,
+    jwt: &str,
+) -> Result<ControlPlaneSpecResponse, (bool, String)> {
+    let resp = reqwest::blocking::Client::new()
+        .get(uri)
+        .header("Authorization", jwt)
+        .send()
+        .map_err(|e| {
+            (
+                true,
+                format!("could not perform spec request to control plane: {}", e),
+            )
+        })?;
+
+    match resp.status() {
+        StatusCode::OK => match resp.json::<ControlPlaneSpecResponse>() {
+            Ok(spec_resp) => Ok(spec_resp),
+            Err(e) => Err((
+                true,
+                format!("could not deserialize control plane response: {}", e),
+            )),
+        },
+        StatusCode::SERVICE_UNAVAILABLE => {
+            Err((true, "control plane is temporarily unavailable".to_string()))
+        }
+        StatusCode::BAD_GATEWAY => {
+            // We have a problem with intermittent 502 errors now
+            // https://github.com/neondatabase/cloud/issues/2353
+            // It's fine to retry GET request in this case.
+            Err((true, "control plane request failed with 502".to_string()))
+        }
+        // Another code, likely 500 or 404, means that compute is unknown to the control plane
+        // or some internal failure happened. Doesn't make much sense to retry in this case.
+        _ => Err((
+            false,
+            format!(
+                "unexpected control plane response status code: {}",
+                resp.status()
+            ),
+        )),
+    }
+}
+
 /// Request spec from the control-plane by compute_id. If `NEON_CONSOLE_JWT`
 /// env variable is set, it will be used for authorization.
-pub fn get_spec_from_control_plane(base_uri: &str, compute_id: &str) -> Result<ComputeSpec> {
+pub fn get_spec_from_control_plane(
+    base_uri: &str,
+    compute_id: &str,
+) -> Result<Option<ComputeSpec>> {
     let cp_uri = format!("{base_uri}/management/api/v2/computes/{compute_id}/spec");
-    let jwt: String = match std::env::var("NEON_CONSOLE_JWT") {
+    let jwt: String = match std::env::var("NEON_CONTROL_PLANE_TOKEN") {
         Ok(v) => v,
         Err(_) => "".to_string(),
     };
+    let mut attempt = 1;
+    let mut spec: Result<Option<ComputeSpec>> = Ok(None);
+
     info!("getting spec from control plane: {}", cp_uri);
 
-    // TODO: check the response. We should distinguish cases when it's
-    // - network error, then retry
-    // - no spec for compute yet, then wait
-    // - compute id is unknown or any other error, then bail out
-    let spec = reqwest::blocking::Client::new()
-        .get(cp_uri)
-        .header("Authorization", jwt)
-        .send()?
-        .json()?;
+    // Do 3 attempts to get spec from the control plane using the following logic:
+    // - network error -> then retry
+    // - compute id is unknown or any other error -> bail out
+    // - no spec for compute yet (Empty state) -> return Ok(None)
+    // - got spec -> return Ok(Some(spec))
+    while attempt < 4 {
+        spec = match do_control_plane_request(&cp_uri, &jwt) {
+            Ok(spec_resp) => match spec_resp.status {
+                ControlPlaneComputeStatus::Empty => Ok(None),
+                ControlPlaneComputeStatus::Attached => {
+                    if let Some(spec) = spec_resp.spec {
+                        Ok(Some(spec))
+                    } else {
+                        bail!("compute is attached, but spec is empty")
+                    }
+                }
+            },
+            Err((retry, msg)) => {
+                if retry {
+                    Err(anyhow!(msg))
+                } else {
+                    bail!(msg);
+                }
+            }
+        };
 
-    Ok(spec)
+        if let Err(e) = &spec {
+            error!("attempt {} to get spec failed with: {}", attempt, e);
+        } else {
+            return spec;
+        }
+
+        attempt += 1;
+        std::thread::sleep(std::time::Duration::from_millis(100));
+    }
+
+    // All attempts failed, return error.
+    spec
 }
 
 /// It takes cluster specification and does the following:
@@ -63,6 +146,21 @@ pub fn update_pg_hba(pgdata_path: &Path) -> Result<()> {
     Ok(())
 }
 
+/// Create a standby.signal file
+pub fn add_standby_signal(pgdata_path: &Path) -> Result<()> {
+    // XXX: consider making it a part of spec.json
+    info!("adding standby.signal");
+    let signalfile = pgdata_path.join("standby.signal");
+
+    if !signalfile.exists() {
+        info!("created standby.signal");
+        File::create(signalfile)?;
+    } else {
+        info!("reused pre-existing standby.signal");
+    }
+    Ok(())
+}
+
 /// Given a cluster spec json and open transaction it handles roles creation,
 /// deletion and update.
 #[instrument(skip_all)]
diff --git a/control_plane/src/bin/neon_local.rs b/control_plane/src/bin/neon_local.rs
index a9b66f479a..09278e1726 100644
--- a/control_plane/src/bin/neon_local.rs
+++ b/control_plane/src/bin/neon_local.rs
@@ -7,7 +7,8 @@
 //!
 use anyhow::{anyhow, bail, Context, Result};
 use clap::{value_parser, Arg, ArgAction, ArgMatches, Command};
-use control_plane::compute::ComputeControlPlane;
+use control_plane::endpoint::ComputeControlPlane;
+use control_plane::endpoint::Replication;
 use control_plane::local_env::LocalEnv;
 use control_plane::pageserver::PageServerNode;
 use control_plane::safekeeper::SafekeeperNode;
@@ -106,8 +107,9 @@ fn main() -> Result<()> {
             "start" => handle_start_all(sub_args, &env),
             "stop" => handle_stop_all(sub_args, &env),
             "pageserver" => handle_pageserver(sub_args, &env),
-            "pg" => handle_pg(sub_args, &env),
             "safekeeper" => handle_safekeeper(sub_args, &env),
+            "endpoint" => handle_endpoint(sub_args, &env),
+            "pg" => bail!("'pg' subcommand has been renamed to 'endpoint'"),
             _ => bail!("unexpected subcommand {sub_name}"),
         };
 
@@ -470,10 +472,17 @@ fn handle_timeline(timeline_match: &ArgMatches, env: &mut local_env::LocalEnv) -
             let mut cplane = ComputeControlPlane::load(env.clone())?;
             println!("Importing timeline into pageserver ...");
             pageserver.timeline_import(tenant_id, timeline_id, base, pg_wal, pg_version)?;
-            println!("Creating node for imported timeline ...");
             env.register_branch_mapping(name.to_string(), tenant_id, timeline_id)?;
 
-            cplane.new_node(tenant_id, name, timeline_id, None, None, pg_version)?;
+            println!("Creating endpoint for imported timeline ...");
+            cplane.new_endpoint(
+                tenant_id,
+                name,
+                timeline_id,
+                None,
+                pg_version,
+                Replication::Primary,
+            )?;
             println!("Done");
         }
         Some(("branch", branch_match)) => {
@@ -521,10 +530,10 @@ fn handle_timeline(timeline_match: &ArgMatches, env: &mut local_env::LocalEnv) -
     Ok(())
 }
 
-fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
-    let (sub_name, sub_args) = match pg_match.subcommand() {
-        Some(pg_subcommand_data) => pg_subcommand_data,
-        None => bail!("no pg subcommand provided"),
+fn handle_endpoint(ep_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
+    let (sub_name, sub_args) = match ep_match.subcommand() {
+        Some(ep_subcommand_data) => ep_subcommand_data,
+        None => bail!("no endpoint subcommand provided"),
     };
 
     let mut cplane = ComputeControlPlane::load(env.clone())?;
@@ -546,7 +555,7 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
             table.load_preset(comfy_table::presets::NOTHING);
 
             table.set_header([
-                "NODE",
+                "ENDPOINT",
                 "ADDRESS",
                 "TIMELINE",
                 "BRANCH NAME",
@@ -554,39 +563,39 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
                 "STATUS",
             ]);
 
-            for ((_, node_name), node) in cplane
-                .nodes
+            for (endpoint_id, endpoint) in cplane
+                .endpoints
                 .iter()
-                .filter(|((node_tenant_id, _), _)| node_tenant_id == &tenant_id)
+                .filter(|(_, endpoint)| endpoint.tenant_id == tenant_id)
             {
-                let lsn_str = match node.lsn {
-                    None => {
-                        // -> primary node
-                        // Use the LSN at the end of the timeline.
-                        timeline_infos
-                            .get(&node.timeline_id)
-                            .map(|bi| bi.last_record_lsn.to_string())
-                            .unwrap_or_else(|| "?".to_string())
-                    }
-                    Some(lsn) => {
-                        // -> read-only node
+                let lsn_str = match endpoint.replication {
+                    Replication::Static(lsn) => {
+                        // -> read-only endpoint
                         // Use the node's LSN.
                         lsn.to_string()
                     }
+                    _ => {
+                        // -> primary endpoint or hot replica
+                        // Use the LSN at the end of the timeline.
+                        timeline_infos
+                            .get(&endpoint.timeline_id)
+                            .map(|bi| bi.last_record_lsn.to_string())
+                            .unwrap_or_else(|| "?".to_string())
+                    }
                 };
 
                 let branch_name = timeline_name_mappings
-                    .get(&TenantTimelineId::new(tenant_id, node.timeline_id))
+                    .get(&TenantTimelineId::new(tenant_id, endpoint.timeline_id))
                     .map(|name| name.as_str())
                     .unwrap_or("?");
 
                 table.add_row([
-                    node_name.as_str(),
-                    &node.address.to_string(),
-                    &node.timeline_id.to_string(),
+                    endpoint_id.as_str(),
+                    &endpoint.address.to_string(),
+                    &endpoint.timeline_id.to_string(),
                     branch_name,
                     lsn_str.as_str(),
-                    node.status(),
+                    endpoint.status(),
                 ]);
             }
 
@@ -597,10 +606,10 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
                 .get_one::<String>("branch-name")
                 .map(|s| s.as_str())
                 .unwrap_or(DEFAULT_BRANCH_NAME);
-            let node_name = sub_args
-                .get_one::<String>("node")
-                .map(|node_name| node_name.to_string())
-                .unwrap_or_else(|| format!("{branch_name}_node"));
+            let endpoint_id = sub_args
+                .get_one::<String>("endpoint_id")
+                .map(String::to_string)
+                .unwrap_or_else(|| format!("ep-{branch_name}"));
 
             let lsn = sub_args
                 .get_one::<String>("lsn")
@@ -618,15 +627,34 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
                 .copied()
                 .context("Failed to parse postgres version from the argument string")?;
 
-            cplane.new_node(tenant_id, &node_name, timeline_id, lsn, port, pg_version)?;
+            let hot_standby = sub_args
+                .get_one::<bool>("hot-standby")
+                .copied()
+                .unwrap_or(false);
+
+            let replication = match (lsn, hot_standby) {
+                (Some(lsn), false) => Replication::Static(lsn),
+                (None, true) => Replication::Replica,
+                (None, false) => Replication::Primary,
+                (Some(_), true) => anyhow::bail!("cannot specify both lsn and hot-standby"),
+            };
+
+            cplane.new_endpoint(
+                tenant_id,
+                &endpoint_id,
+                timeline_id,
+                port,
+                pg_version,
+                replication,
+            )?;
         }
         "start" => {
             let port: Option<u16> = sub_args.get_one::<u16>("port").copied();
-            let node_name = sub_args
-                .get_one::<String>("node")
-                .ok_or_else(|| anyhow!("No node name was provided to start"))?;
+            let endpoint_id = sub_args
+                .get_one::<String>("endpoint_id")
+                .ok_or_else(|| anyhow!("No endpoint ID was provided to start"))?;
 
-            let node = cplane.nodes.get(&(tenant_id, node_name.to_string()));
+            let endpoint = cplane.endpoints.get(endpoint_id.as_str());
 
             let auth_token = if matches!(env.pageserver.pg_auth_type, AuthType::NeonJWT) {
                 let claims = Claims::new(Some(tenant_id), Scope::Tenant);
@@ -636,9 +664,23 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
                 None
             };
 
-            if let Some(node) = node {
-                println!("Starting existing postgres {node_name}...");
-                node.start(&auth_token)?;
+            let hot_standby = sub_args
+                .get_one::<bool>("hot-standby")
+                .copied()
+                .unwrap_or(false);
+
+            if let Some(endpoint) = endpoint {
+                match (&endpoint.replication, hot_standby) {
+                    (Replication::Static(_), true) => {
+                        bail!("Cannot start a node in hot standby mode when it is already configured as a static replica")
+                    }
+                    (Replication::Primary, true) => {
+                        bail!("Cannot start a node as a hot standby replica, it is already configured as primary node")
+                    }
+                    _ => {}
+                }
+                println!("Starting existing endpoint {endpoint_id}...");
+                endpoint.start(&auth_token)?;
             } else {
                 let branch_name = sub_args
                     .get_one::<String>("branch-name")
@@ -658,32 +700,46 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
                     .get_one::<u32>("pg-version")
                     .copied()
                     .context("Failed to `pg-version` from the argument string")?;
+
+                let replication = match (lsn, hot_standby) {
+                    (Some(lsn), false) => Replication::Static(lsn),
+                    (None, true) => Replication::Replica,
+                    (None, false) => Replication::Primary,
+                    (Some(_), true) => anyhow::bail!("cannot specify both lsn and hot-standby"),
+                };
+
                 // when used with custom port this results in non obvious behaviour
                 // port is remembered from first start command, i e
                 // start --port X
                 // stop
                 // start <-- will also use port X even without explicit port argument
-                println!("Starting new postgres (v{pg_version}) {node_name} on timeline {timeline_id} ...");
+                println!("Starting new endpoint {endpoint_id} (PostgreSQL v{pg_version}) on timeline {timeline_id} ...");
 
-                let node =
-                    cplane.new_node(tenant_id, node_name, timeline_id, lsn, port, pg_version)?;
-                node.start(&auth_token)?;
+                let ep = cplane.new_endpoint(
+                    tenant_id,
+                    endpoint_id,
+                    timeline_id,
+                    port,
+                    pg_version,
+                    replication,
+                )?;
+                ep.start(&auth_token)?;
             }
         }
         "stop" => {
-            let node_name = sub_args
-                .get_one::<String>("node")
-                .ok_or_else(|| anyhow!("No node name was provided to stop"))?;
+            let endpoint_id = sub_args
+                .get_one::<String>("endpoint_id")
+                .ok_or_else(|| anyhow!("No endpoint ID was provided to stop"))?;
             let destroy = sub_args.get_flag("destroy");
 
-            let node = cplane
-                .nodes
-                .get(&(tenant_id, node_name.to_string()))
-                .with_context(|| format!("postgres {node_name} is not found"))?;
-            node.stop(destroy)?;
+            let endpoint = cplane
+                .endpoints
+                .get(endpoint_id.as_str())
+                .with_context(|| format!("postgres endpoint {endpoint_id} is not found"))?;
+            endpoint.stop(destroy)?;
         }
 
-        _ => bail!("Unexpected pg subcommand '{sub_name}'"),
+        _ => bail!("Unexpected endpoint subcommand '{sub_name}'"),
     }
 
     Ok(())
@@ -802,7 +858,7 @@ fn handle_safekeeper(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> Resul
 }
 
 fn handle_start_all(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> anyhow::Result<()> {
-    // Postgres nodes are not started automatically
+    // Endpoints are not started automatically
 
     broker::start_broker_process(env)?;
 
@@ -836,10 +892,10 @@ fn handle_stop_all(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<
 fn try_stop_all(env: &local_env::LocalEnv, immediate: bool) {
     let pageserver = PageServerNode::from_env(env);
 
-    // Stop all compute nodes
+    // Stop all endpoints
     match ComputeControlPlane::load(env.clone()) {
         Ok(cplane) => {
-            for (_k, node) in cplane.nodes {
+            for (_k, node) in cplane.endpoints {
                 if let Err(e) = node.stop(false) {
                     eprintln!("postgres stop failed: {e:#}");
                 }
@@ -872,7 +928,9 @@ fn cli() -> Command {
         .help("Name of the branch to be created or used as an alias for other services")
         .required(false);
 
-    let pg_node_arg = Arg::new("node").help("Postgres node name").required(false);
+    let endpoint_id_arg = Arg::new("endpoint_id")
+        .help("Postgres endpoint id")
+        .required(false);
 
     let safekeeper_id_arg = Arg::new("id").help("safekeeper id").required(false);
 
@@ -919,6 +977,12 @@ fn cli() -> Command {
         .help("Specify Lsn on the timeline to start from. By default, end of the timeline would be used.")
         .required(false);
 
+    let hot_standby_arg = Arg::new("hot-standby")
+        .value_parser(value_parser!(bool))
+        .long("hot-standby")
+        .help("If set, the node will be a hot replica on the specified timeline")
+        .required(false);
+
     Command::new("Neon CLI")
         .arg_required_else_help(true)
         .version(GIT_VERSION)
@@ -1026,37 +1090,39 @@ fn cli() -> Command {
                 )
         )
         .subcommand(
-            Command::new("pg")
+            Command::new("endpoint")
                 .arg_required_else_help(true)
                 .about("Manage postgres instances")
                 .subcommand(Command::new("list").arg(tenant_id_arg.clone()))
                 .subcommand(Command::new("create")
-                    .about("Create a postgres compute node")
-                    .arg(pg_node_arg.clone())
+                    .about("Create a compute endpoint")
+                    .arg(endpoint_id_arg.clone())
                     .arg(branch_name_arg.clone())
                     .arg(tenant_id_arg.clone())
                     .arg(lsn_arg.clone())
                     .arg(port_arg.clone())
                     .arg(
                         Arg::new("config-only")
-                            .help("Don't do basebackup, create compute node with only config files")
+                            .help("Don't do basebackup, create endpoint directory with only config files")
                             .long("config-only")
                             .required(false))
                     .arg(pg_version_arg.clone())
+                    .arg(hot_standby_arg.clone())
                 )
                 .subcommand(Command::new("start")
-                    .about("Start a postgres compute node.\n This command actually creates new node from scratch, but preserves existing config files")
-                    .arg(pg_node_arg.clone())
+                    .about("Start postgres.\n If the endpoint doesn't exist yet, it is created.")
+                    .arg(endpoint_id_arg.clone())
                     .arg(tenant_id_arg.clone())
                     .arg(branch_name_arg)
                     .arg(timeline_id_arg)
                     .arg(lsn_arg)
                     .arg(port_arg)
                     .arg(pg_version_arg)
+                    .arg(hot_standby_arg)
                 )
                 .subcommand(
                     Command::new("stop")
-                    .arg(pg_node_arg)
+                    .arg(endpoint_id_arg)
                     .arg(tenant_id_arg)
                     .arg(
                         Arg::new("destroy")
@@ -1068,6 +1134,13 @@ fn cli() -> Command {
                 )
 
         )
+        // Obsolete old name for 'endpoint'. We now just print an error if it's used.
+        .subcommand(
+            Command::new("pg")
+                .hide(true)
+                .arg(Arg::new("ignore-rest").allow_hyphen_values(true).num_args(0..).required(false))
+                .trailing_var_arg(true)
+        )
         .subcommand(
             Command::new("start")
                 .about("Start page server and safekeepers")
diff --git a/control_plane/src/compute.rs b/control_plane/src/endpoint.rs
similarity index 61%
rename from control_plane/src/compute.rs
rename to control_plane/src/endpoint.rs
index bc81107706..7d3485518f 100644
--- a/control_plane/src/compute.rs
+++ b/control_plane/src/endpoint.rs
@@ -25,107 +25,117 @@ use crate::postgresql_conf::PostgresConf;
 //
 pub struct ComputeControlPlane {
     base_port: u16,
-    pageserver: Arc<PageServerNode>,
-    pub nodes: BTreeMap<(TenantId, String), Arc<PostgresNode>>,
+
+    // endpoint ID is the key
+    pub endpoints: BTreeMap<String, Arc<Endpoint>>,
+
     env: LocalEnv,
+    pageserver: Arc<PageServerNode>,
 }
 
 impl ComputeControlPlane {
-    // Load current nodes with ports from data directories on disk
-    // Directory structure has the following layout:
-    // pgdatadirs
-    // |- tenants
-    // |  |- <tenant_id>
-    // |  |   |- <node name>
+    // Load current endpoints from the endpoints/ subdirectories
     pub fn load(env: LocalEnv) -> Result<ComputeControlPlane> {
         let pageserver = Arc::new(PageServerNode::from_env(&env));
 
-        let mut nodes = BTreeMap::default();
-        let pgdatadirspath = &env.pg_data_dirs_path();
-
-        for tenant_dir in fs::read_dir(pgdatadirspath)
-            .with_context(|| format!("failed to list {}", pgdatadirspath.display()))?
+        let mut endpoints = BTreeMap::default();
+        for endpoint_dir in fs::read_dir(env.endpoints_path())
+            .with_context(|| format!("failed to list {}", env.endpoints_path().display()))?
         {
-            let tenant_dir = tenant_dir?;
-            for timeline_dir in fs::read_dir(tenant_dir.path())
-                .with_context(|| format!("failed to list {}", tenant_dir.path().display()))?
-            {
-                let node = PostgresNode::from_dir_entry(timeline_dir?, &env, &pageserver)?;
-                nodes.insert((node.tenant_id, node.name.clone()), Arc::new(node));
-            }
+            let ep = Endpoint::from_dir_entry(endpoint_dir?, &env, &pageserver)?;
+            endpoints.insert(ep.name.clone(), Arc::new(ep));
         }
 
         Ok(ComputeControlPlane {
             base_port: 55431,
-            pageserver,
-            nodes,
+            endpoints,
             env,
+            pageserver,
         })
     }
 
     fn get_port(&mut self) -> u16 {
         1 + self
-            .nodes
+            .endpoints
             .values()
-            .map(|node| node.address.port())
+            .map(|ep| ep.address.port())
             .max()
             .unwrap_or(self.base_port)
     }
 
-    pub fn new_node(
+    pub fn new_endpoint(
         &mut self,
         tenant_id: TenantId,
         name: &str,
         timeline_id: TimelineId,
-        lsn: Option<Lsn>,
         port: Option<u16>,
         pg_version: u32,
-    ) -> Result<Arc<PostgresNode>> {
+        replication: Replication,
+    ) -> Result<Arc<Endpoint>> {
         let port = port.unwrap_or_else(|| self.get_port());
-        let node = Arc::new(PostgresNode {
+
+        let ep = Arc::new(Endpoint {
             name: name.to_owned(),
             address: SocketAddr::new("127.0.0.1".parse().unwrap(), port),
             env: self.env.clone(),
             pageserver: Arc::clone(&self.pageserver),
             timeline_id,
-            lsn,
+            replication,
             tenant_id,
             pg_version,
         });
 
-        node.create_pgdata()?;
-        node.setup_pg_conf()?;
+        ep.create_pgdata()?;
+        ep.setup_pg_conf()?;
 
-        self.nodes
-            .insert((tenant_id, node.name.clone()), Arc::clone(&node));
+        self.endpoints.insert(ep.name.clone(), Arc::clone(&ep));
 
-        Ok(node)
+        Ok(ep)
     }
 }
 
 ///////////////////////////////////////////////////////////////////////////////
 
-#[derive(Debug)]
-pub struct PostgresNode {
-    pub address: SocketAddr,
-    name: String,
-    pub env: LocalEnv,
-    pageserver: Arc<PageServerNode>,
-    pub timeline_id: TimelineId,
-    pub lsn: Option<Lsn>, // if it's a read-only node. None for primary
-    pub tenant_id: TenantId,
-    pg_version: u32,
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub enum Replication {
+    // Regular read-write node
+    Primary,
+    // if recovery_target_lsn is provided, and we want to pin the node to a specific LSN
+    Static(Lsn),
+    // Hot standby; read-only replica.
+    // Future versions may want to distinguish between replicas with hot standby
+    // feedback and other kinds of replication configurations.
+    Replica,
 }
 
-impl PostgresNode {
+#[derive(Debug)]
+pub struct Endpoint {
+    /// used as the directory name
+    name: String,
+    pub tenant_id: TenantId,
+    pub timeline_id: TimelineId,
+    // Some(lsn) if this is a read-only endpoint anchored at 'lsn'. None for the primary.
+    pub replication: Replication,
+
+    // port and address of the Postgres server
+    pub address: SocketAddr,
+    pg_version: u32,
+
+    // These are not part of the endpoint as such, but the environment
+    // the endpoint runs in.
+    pub env: LocalEnv,
+    pageserver: Arc<PageServerNode>,
+}
+
+impl Endpoint {
     fn from_dir_entry(
         entry: std::fs::DirEntry,
         env: &LocalEnv,
         pageserver: &Arc<PageServerNode>,
-    ) -> Result<PostgresNode> {
+    ) -> Result<Endpoint> {
         if !entry.file_type()?.is_dir() {
             anyhow::bail!(
-                "PostgresNode::from_dir_entry failed: '{}' is not a directory",
+                "Endpoint::from_dir_entry failed: '{}' is not a directory",
                 entry.path().display()
             );
         }
@@ -135,7 +145,7 @@ impl PostgresNode {
         let name = fname.to_str().unwrap().to_string();
 
         // Read config file into memory
-        let cfg_path = entry.path().join("postgresql.conf");
+        let cfg_path = entry.path().join("pgdata").join("postgresql.conf");
         let cfg_path_str = cfg_path.to_string_lossy();
         let mut conf_file = File::open(&cfg_path)
             .with_context(|| format!("failed to open config file in {}", cfg_path_str))?;
@@ -156,18 +166,26 @@ impl PostgresNode {
             fs::read_to_string(pg_version_path).unwrap_or_else(|_| DEFAULT_PG_VERSION.to_string());
         let pg_version = u32::from_str(&pg_version_str)?;
 
-        // parse recovery_target_lsn, if any
-        let recovery_target_lsn: Option<Lsn> =
-            conf.parse_field_optional("recovery_target_lsn", &context)?;
+        // parse recovery_target_lsn and primary_conninfo into Recovery Target, if any
+        let replication = if let Some(lsn_str) = conf.get("recovery_target_lsn") {
+            Replication::Static(Lsn::from_str(lsn_str)?)
+        } else if let Some(slot_name) = conf.get("primary_slot_name") {
+            let slot_name = slot_name.to_string();
+            let prefix = format!("repl_{}_", timeline_id);
+            assert!(slot_name.starts_with(&prefix));
+            Replication::Replica
+        } else {
+            Replication::Primary
+        };
 
         // ok now
-        Ok(PostgresNode {
+        Ok(Endpoint {
             address: SocketAddr::new("127.0.0.1".parse().unwrap(), port),
             name,
             env: env.clone(),
             pageserver: Arc::clone(pageserver),
             timeline_id,
-            lsn: recovery_target_lsn,
+            replication,
             tenant_id,
             pg_version,
         })
@@ -269,7 +287,7 @@ impl PostgresNode {
     }
 
     // Write postgresql.conf with default configuration
-    // and PG_VERSION file to the data directory of a new node.
+    // and PG_VERSION file to the data directory of a new endpoint.
     fn setup_pg_conf(&self) -> Result<()> {
         let mut conf = PostgresConf::new();
         conf.append("max_wal_senders", "10");
@@ -289,7 +307,7 @@ impl PostgresNode {
         // walproposer panics when basebackup is invalid, it is pointless to restart in this case.
         conf.append("restart_after_crash", "off");
 
-        // Configure the node to fetch pages from pageserver
+        // Configure the Neon Postgres extension to fetch pages from pageserver
         let pageserver_connstr = {
             let config = &self.pageserver.pg_connection_config;
             let (host, port) = (config.host(), config.port());
@@ -302,50 +320,83 @@ impl PostgresNode {
         conf.append("neon.pageserver_connstring", &pageserver_connstr);
         conf.append("neon.tenant_id", &self.tenant_id.to_string());
         conf.append("neon.timeline_id", &self.timeline_id.to_string());
-        if let Some(lsn) = self.lsn {
-            conf.append("recovery_target_lsn", &lsn.to_string());
-        }
 
         conf.append_line("");
-        // Configure backpressure
-        // - Replication write lag depends on how fast the walreceiver can process incoming WAL.
-        //   This lag determines latency of get_page_at_lsn. Speed of applying WAL is about 10MB/sec,
-        //   so to avoid expiration of 1 minute timeout, this lag should not be larger than 600MB.
-        //   Actually latency should be much smaller (better if < 1sec). But we assume that recently
-        //   updates pages are not requested from pageserver.
-        // - Replication flush lag depends on speed of persisting data by checkpointer (creation of
-        //   delta/image layers) and advancing disk_consistent_lsn. Safekeepers are able to
-        //   remove/archive WAL only beyond disk_consistent_lsn. Too large a lag can cause long
-        //   recovery time (in case of pageserver crash) and disk space overflow at safekeepers.
-        // - Replication apply lag depends on speed of uploading changes to S3 by uploader thread.
-        //   To be able to restore database in case of pageserver node crash, safekeeper should not
-        //   remove WAL beyond this point. Too large lag can cause space exhaustion in safekeepers
-        //   (if they are not able to upload WAL to S3).
-        conf.append("max_replication_write_lag", "15MB");
-        conf.append("max_replication_flush_lag", "10GB");
+        // Replication-related configurations, such as WAL sending
+        match &self.replication {
+            Replication::Primary => {
+                // Configure backpressure
+                // - Replication write lag depends on how fast the walreceiver can process incoming WAL.
+                //   This lag determines latency of get_page_at_lsn. Speed of applying WAL is about 10MB/sec,
+                //   so to avoid expiration of 1 minute timeout, this lag should not be larger than 600MB.
+                //   Actually latency should be much smaller (better if < 1sec). But we assume that recently
+                //   updates pages are not requested from pageserver.
+                // - Replication flush lag depends on speed of persisting data by checkpointer (creation of
+                //   delta/image layers) and advancing disk_consistent_lsn. Safekeepers are able to
+                //   remove/archive WAL only beyond disk_consistent_lsn. Too large a lag can cause long
+                //   recovery time (in case of pageserver crash) and disk space overflow at safekeepers.
+                // - Replication apply lag depends on speed of uploading changes to S3 by uploader thread.
+                //   To be able to restore database in case of pageserver node crash, safekeeper should not
+                //   remove WAL beyond this point. Too large lag can cause space exhaustion in safekeepers
+                //   (if they are not able to upload WAL to S3).
+                conf.append("max_replication_write_lag", "15MB");
+                conf.append("max_replication_flush_lag", "10GB");
 
-        if !self.env.safekeepers.is_empty() {
-            // Configure the node to connect to the safekeepers
-            conf.append("synchronous_standby_names", "walproposer");
+                if !self.env.safekeepers.is_empty() {
+                    // Configure Postgres to connect to the safekeepers
+                    conf.append("synchronous_standby_names", "walproposer");
 
-            let safekeepers = self
-                .env
-                .safekeepers
-                .iter()
-                .map(|sk| format!("localhost:{}", sk.pg_port))
-                .collect::<Vec<String>>()
-                .join(",");
-            conf.append("neon.safekeepers", &safekeepers);
-        } else {
-            // We only use setup without safekeepers for tests,
-            // and don't care about data durability on pageserver,
-            // so set more relaxed synchronous_commit.
-            conf.append("synchronous_commit", "remote_write");
+                    let safekeepers = self
+                        .env
+                        .safekeepers
+                        .iter()
+                        .map(|sk| format!("localhost:{}", sk.pg_port))
+                        .collect::<Vec<String>>()
+                        .join(",");
+                    conf.append("neon.safekeepers", &safekeepers);
+                } else {
+                    // We only use setup without safekeepers for tests,
+                    // and don't care about data durability on pageserver,
+                    // so set more relaxed synchronous_commit.
+                    conf.append("synchronous_commit", "remote_write");
 
-            // Configure the node to stream WAL directly to the pageserver
-            // This isn't really a supported configuration, but can be useful for
-            // testing.
-            conf.append("synchronous_standby_names", "pageserver");
+                    // Configure the node to stream WAL directly to the pageserver
+                    // This isn't really a supported configuration, but can be useful for
+                    // testing.
+                    conf.append("synchronous_standby_names", "pageserver");
+                }
+            }
+            Replication::Static(lsn) => {
+                conf.append("recovery_target_lsn", &lsn.to_string());
+            }
+            Replication::Replica => {
+                assert!(!self.env.safekeepers.is_empty());
+
+                // TODO: use future host field from safekeeper spec
+                // Pass the list of safekeepers to the replica so that it can connect to any of them,
+                // whichever is availiable.
+                let sk_ports = self
+                    .env
+                    .safekeepers
+                    .iter()
+                    .map(|x| x.pg_port.to_string())
+                    .collect::<Vec<_>>()
+                    .join(",");
+                let sk_hosts = vec!["localhost"; self.env.safekeepers.len()].join(",");
+
+                let connstr = format!(
+                    "host={} port={} options='-c timeline_id={} tenant_id={}' application_name=replica replication=true",
+                    sk_hosts,
+                    sk_ports,
+                    &self.timeline_id.to_string(),
+                    &self.tenant_id.to_string(),
+                );
+
+                let slot_name = format!("repl_{}_", self.timeline_id);
+                conf.append("primary_conninfo", connstr.as_str());
+                conf.append("primary_slot_name", slot_name.as_str());
+                conf.append("hot_standby", "on");
+            }
         }
 
         let mut file = File::create(self.pgdata().join("postgresql.conf"))?;
@@ -358,21 +409,27 @@ impl PostgresNode {
     }
 
     fn load_basebackup(&self, auth_token: &Option<String>) -> Result<()> {
-        let backup_lsn = if let Some(lsn) = self.lsn {
-            Some(lsn)
-        } else if !self.env.safekeepers.is_empty() {
-            // LSN 0 means that it is bootstrap and we need to download just
-            // latest data from the pageserver. That is a bit clumsy but whole bootstrap
-            // procedure evolves quite actively right now, so let's think about it again
-            // when things would be more stable (TODO).
-            let lsn = self.sync_safekeepers(auth_token, self.pg_version)?;
-            if lsn == Lsn(0) {
-                None
-            } else {
-                Some(lsn)
+        let backup_lsn = match &self.replication {
+            Replication::Primary => {
+                if !self.env.safekeepers.is_empty() {
+                    // LSN 0 means that it is bootstrap and we need to download just
+                    // latest data from the pageserver. That is a bit clumsy but whole bootstrap
+                    // procedure evolves quite actively right now, so let's think about it again
+                    // when things would be more stable (TODO).
+                    let lsn = self.sync_safekeepers(auth_token, self.pg_version)?;
+                    if lsn == Lsn(0) {
+                        None
+                    } else {
+                        Some(lsn)
+                    }
+                } else {
+                    None
+                }
+            }
+            Replication::Static(lsn) => Some(*lsn),
+            Replication::Replica => {
+                None // Take the latest snapshot available to start with
             }
-        } else {
-            None
         };
 
         self.do_basebackup(backup_lsn)?;
@@ -380,8 +437,12 @@ impl PostgresNode {
         Ok(())
     }
 
+    pub fn endpoint_path(&self) -> PathBuf {
+        self.env.endpoints_path().join(&self.name)
+    }
+
     pub fn pgdata(&self) -> PathBuf {
-        self.env.pg_data_dir(&self.tenant_id, &self.name)
+        self.endpoint_path().join("pgdata")
     }
 
     pub fn status(&self) -> &str {
@@ -443,12 +504,11 @@ impl PostgresNode {
     }
 
     pub fn start(&self, auth_token: &Option<String>) -> Result<()> {
-        // Bail if the node already running.
         if self.status() == "running" {
-            anyhow::bail!("The node is already running");
+            anyhow::bail!("The endpoint is already running");
         }
 
-        // 1. We always start compute node from scratch, so
+        // 1. We always start Postgres from scratch, so
         // if old dir exists, preserve 'postgresql.conf' and drop the directory
         let postgresql_conf_path = self.pgdata().join("postgresql.conf");
         let postgresql_conf = fs::read(&postgresql_conf_path).with_context(|| {
@@ -466,12 +526,12 @@ impl PostgresNode {
         // 3. Load basebackup
         self.load_basebackup(auth_token)?;
 
-        if self.lsn.is_some() {
+        if self.replication != Replication::Primary {
             File::create(self.pgdata().join("standby.signal"))?;
         }
 
-        // 4. Finally start the compute node postgres
-        println!("Starting postgres node at '{}'", self.connstr());
+        // 4. Finally start postgres
+        println!("Starting postgres at '{}'", self.connstr());
         self.pg_ctl(&["start"], auth_token)
     }
 
@@ -480,7 +540,7 @@ impl PostgresNode {
         // use immediate shutdown mode, otherwise,
         // shutdown gracefully to leave the data directory sane.
         //
-        // Compute node always starts from scratch, so stop
+        // Postgres is always started from scratch, so stop
         // without destroy only used for testing and debugging.
         //
         if destroy {
@@ -489,7 +549,7 @@ impl PostgresNode {
                 "Destroying postgres data directory '{}'",
                 self.pgdata().to_str().unwrap()
             );
-            fs::remove_dir_all(self.pgdata())?;
+            fs::remove_dir_all(self.endpoint_path())?;
         } else {
             self.pg_ctl(&["stop"], &None)?;
         }
diff --git a/control_plane/src/lib.rs b/control_plane/src/lib.rs
index 6829479ad5..a773b8dcc3 100644
--- a/control_plane/src/lib.rs
+++ b/control_plane/src/lib.rs
@@ -9,7 +9,7 @@
 
 mod background_process;
 pub mod broker;
-pub mod compute;
+pub mod endpoint;
 pub mod local_env;
 pub mod pageserver;
 pub mod postgresql_conf;
diff --git a/control_plane/src/local_env.rs b/control_plane/src/local_env.rs
index 8cc6329ce6..2b1eec7c4b 100644
--- a/control_plane/src/local_env.rs
+++ b/control_plane/src/local_env.rs
@@ -200,14 +200,8 @@ impl LocalEnv {
         self.neon_distrib_dir.join("storage_broker")
     }
 
-    pub fn pg_data_dirs_path(&self) -> PathBuf {
-        self.base_data_dir.join("pgdatadirs").join("tenants")
-    }
-
-    pub fn pg_data_dir(&self, tenant_id: &TenantId, branch_name: &str) -> PathBuf {
-        self.pg_data_dirs_path()
-            .join(tenant_id.to_string())
-            .join(branch_name)
+    pub fn endpoints_path(&self) -> PathBuf {
+        self.base_data_dir.join("endpoints")
     }
 
     // TODO: move pageserver files into ./pageserver
@@ -427,7 +421,7 @@ impl LocalEnv {
             }
         }
 
-        fs::create_dir_all(self.pg_data_dirs_path())?;
+        fs::create_dir_all(self.endpoints_path())?;
 
         for safekeeper in &self.safekeepers {
             fs::create_dir_all(SafekeeperNode::datadir_path_by_id(self, safekeeper.id))?;
diff --git a/control_plane/src/pageserver.rs b/control_plane/src/pageserver.rs
index 094069e4c0..75991045a4 100644
--- a/control_plane/src/pageserver.rs
+++ b/control_plane/src/pageserver.rs
@@ -359,8 +359,8 @@ impl PageServerNode {
                 .transpose()
                 .context("Failed to parse 'trace_read_requests' as bool")?,
             eviction_policy: settings
-                .get("eviction_policy")
-                .map(|x| serde_json::from_str(x))
+                .remove("eviction_policy")
+                .map(serde_json::from_str)
                 .transpose()
                 .context("Failed to parse 'eviction_policy' json")?,
             min_resident_size_override: settings
@@ -368,6 +368,9 @@ impl PageServerNode {
                 .map(|x| x.parse::<u64>())
                 .transpose()
                 .context("Failed to parse 'min_resident_size_override' as integer")?,
+            evictions_low_residence_duration_metric_threshold: settings
+                .remove("evictions_low_residence_duration_metric_threshold")
+                .map(|x| x.to_string()),
         };
         if !settings.is_empty() {
             bail!("Unrecognized tenant settings: {settings:?}")
@@ -445,6 +448,9 @@ impl PageServerNode {
                     .map(|x| x.parse::<u64>())
                     .transpose()
                     .context("Failed to parse 'min_resident_size_override' as an integer")?,
+                evictions_low_residence_duration_metric_threshold: settings
+                    .get("evictions_low_residence_duration_metric_threshold")
+                    .map(|x| x.to_string()),
             })
             .send()?
             .error_from_body()?;
diff --git a/control_plane/src/postgresql_conf.rs b/control_plane/src/postgresql_conf.rs
index 34dc769e78..638575eb82 100644
--- a/control_plane/src/postgresql_conf.rs
+++ b/control_plane/src/postgresql_conf.rs
@@ -13,7 +13,7 @@ use std::io::BufRead;
 use std::str::FromStr;
 
 /// In-memory representation of a postgresql.conf file
-#[derive(Default)]
+#[derive(Default, Debug)]
 pub struct PostgresConf {
     lines: Vec<String>,
     hash: HashMap<String, String>,
diff --git a/docker-compose/compute_wrapper/var/db/postgres/specs/spec.json b/docker-compose/compute_wrapper/var/db/postgres/specs/spec.json
index 10ae0b0ecf..565e5e368e 100644
--- a/docker-compose/compute_wrapper/var/db/postgres/specs/spec.json
+++ b/docker-compose/compute_wrapper/var/db/postgres/specs/spec.json
@@ -28,11 +28,6 @@
                 "value": "replica",
                 "vartype": "enum"
             },
-            {
-                "name": "hot_standby",
-                "value": "on",
-                "vartype": "bool"
-            },
             {
                 "name": "wal_log_hints",
                 "value": "on",
diff --git a/libs/compute_api/src/responses.rs b/libs/compute_api/src/responses.rs
index 43289a5e3e..c409563b56 100644
--- a/libs/compute_api/src/responses.rs
+++ b/libs/compute_api/src/responses.rs
@@ -1,7 +1,9 @@
 //! Structs representing the JSON formats used in the compute_ctl's HTTP API.
 
 use chrono::{DateTime, Utc};
-use serde::{Serialize, Serializer};
+use serde::{Deserialize, Serialize, Serializer};
+
+use crate::spec::ComputeSpec;
 
 #[derive(Serialize, Debug)]
 pub struct GenericAPIError {
@@ -12,8 +14,9 @@ pub struct GenericAPIError {
 #[derive(Serialize, Debug)]
 #[serde(rename_all = "snake_case")]
 pub struct ComputeStatusResponse {
-    pub tenant: String,
-    pub timeline: String,
+    pub start_time: DateTime<Utc>,
+    pub tenant: Option<String>,
+    pub timeline: Option<String>,
     pub status: ComputeStatus,
     #[serde(serialize_with = "rfc3339_serialize")]
     pub last_active: DateTime<Utc>,
@@ -43,6 +46,8 @@ pub enum ComputeStatus {
     Init,
     // Compute is configured and running.
     Running,
+    // New spec is being applied.
+    Configuration,
     // Either startup or configuration failed,
     // compute will exit soon or is waiting for
     // control-plane to terminate it.
@@ -59,8 +64,29 @@ where
 /// Response of the /metrics.json API
 #[derive(Clone, Debug, Default, Serialize)]
 pub struct ComputeMetrics {
+    pub wait_for_spec_ms: u64,
     pub sync_safekeepers_ms: u64,
     pub basebackup_ms: u64,
     pub config_ms: u64,
     pub total_startup_ms: u64,
 }
+
+/// Response of the `/computes/{compute_id}/spec` control-plane API.
+/// This is not actually a compute API response, so consider moving
+/// to a different place.
+#[derive(Deserialize, Debug)]
+pub struct ControlPlaneSpecResponse {
+    pub spec: Option<ComputeSpec>,
+    pub status: ControlPlaneComputeStatus,
+}
+
+#[derive(Deserialize, Clone, Copy, Debug, PartialEq, Eq)]
+#[serde(rename_all = "snake_case")]
+pub enum ControlPlaneComputeStatus {
+    // Compute is known to control-plane, but it's not
+    // yet attached to any timeline / endpoint.
+    Empty,
+    // Compute is attached to some timeline / endpoint and
+    // should be able to start with provided spec.
+    Attached,
+}
diff --git a/libs/compute_api/src/spec.rs b/libs/compute_api/src/spec.rs
index 37fe133b68..f771910329 100644
--- a/libs/compute_api/src/spec.rs
+++ b/libs/compute_api/src/spec.rs
@@ -15,7 +15,10 @@ pub type PgIdent = String;
 #[derive(Clone, Debug, Default, Deserialize)]
 pub struct ComputeSpec {
     pub format_version: f32,
-    pub timestamp: String,
+
+    // The control plane also includes a 'timestamp' field in the JSON document,
+    // but we don't use it for anything. Serde will ignore missing fields when
+    // deserializing it.
     pub operation_uuid: Option<String>,
     /// Expected cluster state at the end of transition process.
     pub cluster: Cluster,
diff --git a/libs/consumption_metrics/Cargo.toml b/libs/consumption_metrics/Cargo.toml
index f26aa2fbc5..3f290821c2 100644
--- a/libs/consumption_metrics/Cargo.toml
+++ b/libs/consumption_metrics/Cargo.toml
@@ -4,13 +4,12 @@ version = "0.1.0"
 edition = "2021"
 license = "Apache-2.0"
 
-# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
-
 [dependencies]
-anyhow = "1.0.68"
-chrono = { version = "0.4", default-features = false, features = ["clock", "serde"] }
-rand = "0.8.3"
-serde = "1.0.152"
-serde_with = "2.1.0"
-utils = { version = "0.1.0", path = "../utils" }
-workspace_hack = { version = "0.1.0", path = "../../workspace_hack" }
+anyhow.workspace = true
+chrono.workspace = true
+rand.workspace = true
+serde.workspace = true
+serde_with.workspace = true
+utils.workspace = true
+
+workspace_hack.workspace = true
diff --git a/libs/pageserver_api/Cargo.toml b/libs/pageserver_api/Cargo.toml
index 7709da1072..f97ec54e91 100644
--- a/libs/pageserver_api/Cargo.toml
+++ b/libs/pageserver_api/Cargo.toml
@@ -7,6 +7,7 @@ license.workspace = true
 [dependencies]
 serde.workspace = true
 serde_with.workspace = true
+serde_json.workspace = true
 const_format.workspace = true
 anyhow.workspace = true
 bytes.workspace = true
@@ -14,6 +15,7 @@ byteorder.workspace = true
 utils.workspace = true
 postgres_ffi.workspace = true
 enum-map.workspace = true
-serde_json.workspace = true
+strum.workspace = true
+strum_macros.workspace = true
 
 workspace_hack.workspace = true
diff --git a/libs/pageserver_api/src/models.rs b/libs/pageserver_api/src/models.rs
index 98a4b56858..15c37b9453 100644
--- a/libs/pageserver_api/src/models.rs
+++ b/libs/pageserver_api/src/models.rs
@@ -7,6 +7,7 @@ use std::{
 use byteorder::{BigEndian, ReadBytesExt};
 use serde::{Deserialize, Serialize};
 use serde_with::{serde_as, DisplayFromStr};
+use strum_macros;
 use utils::{
     history_buffer::HistoryBufferWithDropCounter,
     id::{NodeId, TenantId, TimelineId},
@@ -18,11 +19,23 @@ use anyhow::bail;
 use bytes::{BufMut, Bytes, BytesMut};
 
 /// A state of a tenant in pageserver's memory.
-#[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
+#[derive(
+    Clone,
+    PartialEq,
+    Eq,
+    serde::Serialize,
+    serde::Deserialize,
+    strum_macros::Display,
+    strum_macros::EnumString,
+    strum_macros::EnumVariantNames,
+    strum_macros::AsRefStr,
+    strum_macros::IntoStaticStr,
+)]
+#[serde(tag = "slug", content = "data")]
 pub enum TenantState {
-    // This tenant is being loaded from local disk
+    /// This tenant is being loaded from local disk
     Loading,
-    // This tenant is being downloaded from cloud storage.
+    /// This tenant is being downloaded from cloud storage.
     Attaching,
     /// Tenant is fully operational
     Active,
@@ -31,15 +44,7 @@ pub enum TenantState {
     Stopping,
     /// A tenant is recognized by the pageserver, but can no longer be used for
     /// any operations, because it failed to be activated.
-    Broken,
-}
-
-pub mod state {
-    pub const LOADING: &str = "loading";
-    pub const ATTACHING: &str = "attaching";
-    pub const ACTIVE: &str = "active";
-    pub const STOPPING: &str = "stopping";
-    pub const BROKEN: &str = "broken";
+    Broken { reason: String, backtrace: String },
 }
 
 impl TenantState {
@@ -49,17 +54,26 @@ impl TenantState {
             Self::Attaching => true,
             Self::Active => false,
             Self::Stopping => false,
-            Self::Broken => false,
+            Self::Broken { .. } => false,
         }
     }
 
-    pub fn as_str(&self) -> &'static str {
+    pub fn broken_from_reason(reason: String) -> Self {
+        let backtrace_str: String = format!("{}", std::backtrace::Backtrace::force_capture());
+        Self::Broken {
+            reason,
+            backtrace: backtrace_str,
+        }
+    }
+}
+
+impl std::fmt::Debug for TenantState {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
         match self {
-            TenantState::Loading => state::LOADING,
-            TenantState::Attaching => state::ATTACHING,
-            TenantState::Active => state::ACTIVE,
-            TenantState::Stopping => state::STOPPING,
-            TenantState::Broken => state::BROKEN,
+            Self::Broken { reason, backtrace } if !reason.is_empty() => {
+                write!(f, "Broken due to: {reason}. Backtrace:\n{backtrace}")
+            }
+            _ => write!(f, "{self}"),
         }
     }
 }
@@ -121,6 +135,7 @@ pub struct TenantCreateRequest {
     // For now, this field is not even documented in the openapi_spec.yml.
     pub eviction_policy: Option<serde_json::Value>,
     pub min_resident_size_override: Option<u64>,
+    pub evictions_low_residence_duration_metric_threshold: Option<String>,
 }
 
 #[serde_as]
@@ -167,6 +182,7 @@ pub struct TenantConfigRequest {
     // For now, this field is not even documented in the openapi_spec.yml.
     pub eviction_policy: Option<serde_json::Value>,
     pub min_resident_size_override: Option<u64>,
+    pub evictions_low_residence_duration_metric_threshold: Option<String>,
 }
 
 impl TenantConfigRequest {
@@ -188,6 +204,7 @@ impl TenantConfigRequest {
             trace_read_requests: None,
             eviction_policy: None,
             min_resident_size_override: None,
+            evictions_low_residence_duration_metric_threshold: None,
         }
     }
 }
@@ -615,6 +632,7 @@ impl PagestreamBeMessage {
 #[cfg(test)]
 mod tests {
     use bytes::Buf;
+    use serde_json::json;
 
     use super::*;
 
@@ -665,4 +683,57 @@ mod tests {
             assert!(msg == reconstructed);
         }
     }
+
+    #[test]
+    fn test_tenantinfo_serde() {
+        // Test serialization/deserialization of TenantInfo
+        let original_active = TenantInfo {
+            id: TenantId::generate(),
+            state: TenantState::Active,
+            current_physical_size: Some(42),
+            has_in_progress_downloads: Some(false),
+        };
+        let expected_active = json!({
+            "id": original_active.id.to_string(),
+            "state": {
+                "slug": "Active",
+            },
+            "current_physical_size": 42,
+            "has_in_progress_downloads": false,
+        });
+
+        let original_broken = TenantInfo {
+            id: TenantId::generate(),
+            state: TenantState::Broken {
+                reason: "reason".into(),
+                backtrace: "backtrace info".into(),
+            },
+            current_physical_size: Some(42),
+            has_in_progress_downloads: Some(false),
+        };
+        let expected_broken = json!({
+            "id": original_broken.id.to_string(),
+            "state": {
+                "slug": "Broken",
+                "data": {
+                    "backtrace": "backtrace info",
+                    "reason": "reason",
+                }
+            },
+            "current_physical_size": 42,
+            "has_in_progress_downloads": false,
+        });
+
+        assert_eq!(
+            serde_json::to_value(&original_active).unwrap(),
+            expected_active
+        );
+
+        assert_eq!(
+            serde_json::to_value(&original_broken).unwrap(),
+            expected_broken
+        );
+        assert!(format!("{:?}", &original_broken.state).contains("reason"));
+        assert!(format!("{:?}", &original_broken.state).contains("backtrace info"));
+    }
 }
diff --git a/libs/postgres_ffi/build.rs b/libs/postgres_ffi/build.rs
index 66221af522..f7e39751ef 100644
--- a/libs/postgres_ffi/build.rs
+++ b/libs/postgres_ffi/build.rs
@@ -5,7 +5,7 @@ use std::path::PathBuf;
 use std::process::Command;
 
 use anyhow::{anyhow, Context};
-use bindgen::callbacks::ParseCallbacks;
+use bindgen::callbacks::{DeriveInfo, ParseCallbacks};
 
 #[derive(Debug)]
 struct PostgresFfiCallbacks;
@@ -20,7 +20,7 @@ impl ParseCallbacks for PostgresFfiCallbacks {
 
     // Add any custom #[derive] attributes to the data structures that bindgen
     // creates.
-    fn add_derives(&self, name: &str) -> Vec<String> {
+    fn add_derives(&self, derive_info: &DeriveInfo) -> Vec<String> {
         // This is the list of data structures that we want to serialize/deserialize.
         let serde_list = [
             "XLogRecord",
@@ -31,7 +31,7 @@ impl ParseCallbacks for PostgresFfiCallbacks {
             "ControlFileData",
         ];
 
-        if serde_list.contains(&name) {
+        if serde_list.contains(&derive_info.name) {
             vec![
                 "Default".into(), // Default allows us to easily fill the padding fields with 0.
                 "Serialize".into(),
diff --git a/libs/postgres_ffi/src/lib.rs b/libs/postgres_ffi/src/lib.rs
index 492ec9748a..b8eb469cb0 100644
--- a/libs/postgres_ffi/src/lib.rs
+++ b/libs/postgres_ffi/src/lib.rs
@@ -95,10 +95,13 @@ pub fn generate_wal_segment(
     segno: u64,
     system_id: u64,
     pg_version: u32,
+    lsn: Lsn,
 ) -> Result<Bytes, SerializeError> {
+    assert_eq!(segno, lsn.segment_number(WAL_SEGMENT_SIZE));
+
     match pg_version {
-        14 => v14::xlog_utils::generate_wal_segment(segno, system_id),
-        15 => v15::xlog_utils::generate_wal_segment(segno, system_id),
+        14 => v14::xlog_utils::generate_wal_segment(segno, system_id, lsn),
+        15 => v15::xlog_utils::generate_wal_segment(segno, system_id, lsn),
         _ => Err(SerializeError::BadInput),
     }
 }
diff --git a/libs/postgres_ffi/src/pg_constants.rs b/libs/postgres_ffi/src/pg_constants.rs
index 09678353af..6bc89ed37e 100644
--- a/libs/postgres_ffi/src/pg_constants.rs
+++ b/libs/postgres_ffi/src/pg_constants.rs
@@ -195,6 +195,7 @@ pub const FIRST_NORMAL_OBJECT_ID: u32 = 16384;
 
 pub const XLOG_CHECKPOINT_SHUTDOWN: u8 = 0x00;
 pub const XLOG_CHECKPOINT_ONLINE: u8 = 0x10;
+pub const XLP_FIRST_IS_CONTRECORD: u16 = 0x0001;
 pub const XLP_LONG_HEADER: u16 = 0x0002;
 
 /* From fsm_internals.h */
diff --git a/libs/postgres_ffi/src/xlog_utils.rs b/libs/postgres_ffi/src/xlog_utils.rs
index 272c4d6dcc..4d7bb61883 100644
--- a/libs/postgres_ffi/src/xlog_utils.rs
+++ b/libs/postgres_ffi/src/xlog_utils.rs
@@ -270,6 +270,11 @@ impl XLogPageHeaderData {
         use utils::bin_ser::LeSer;
         XLogPageHeaderData::des_from(&mut buf.reader())
     }
+
+    pub fn encode(&self) -> Result<Bytes, SerializeError> {
+        use utils::bin_ser::LeSer;
+        self.ser().map(|b| b.into())
+    }
 }
 
 impl XLogLongPageHeaderData {
@@ -328,22 +333,32 @@ impl CheckPoint {
     }
 }
 
-//
-// Generate new, empty WAL segment.
-// We need this segment to start compute node.
-//
-pub fn generate_wal_segment(segno: u64, system_id: u64) -> Result<Bytes, SerializeError> {
+/// Generate new, empty WAL segment, with correct block headers at the first
+/// page of the segment and the page that contains the given LSN.
+/// We need this segment to start compute node.
+pub fn generate_wal_segment(segno: u64, system_id: u64, lsn: Lsn) -> Result<Bytes, SerializeError> {
     let mut seg_buf = BytesMut::with_capacity(WAL_SEGMENT_SIZE);
 
     let pageaddr = XLogSegNoOffsetToRecPtr(segno, 0, WAL_SEGMENT_SIZE);
+
+    let page_off = lsn.block_offset();
+    let seg_off = lsn.segment_offset(WAL_SEGMENT_SIZE);
+
+    let first_page_only = seg_off < XLOG_BLCKSZ;
+    let (shdr_rem_len, infoflags) = if first_page_only {
+        (seg_off, pg_constants::XLP_FIRST_IS_CONTRECORD)
+    } else {
+        (0, 0)
+    };
+
     let hdr = XLogLongPageHeaderData {
         std: {
             XLogPageHeaderData {
                 xlp_magic: XLOG_PAGE_MAGIC as u16,
-                xlp_info: pg_constants::XLP_LONG_HEADER,
+                xlp_info: pg_constants::XLP_LONG_HEADER | infoflags,
                 xlp_tli: PG_TLI,
                 xlp_pageaddr: pageaddr,
-                xlp_rem_len: 0,
+                xlp_rem_len: shdr_rem_len as u32,
                 ..Default::default() // Put 0 in padding fields.
             }
         },
@@ -357,6 +372,33 @@ pub fn generate_wal_segment(segno: u64, system_id: u64) -> Result<Bytes, Seriali
 
     //zero out the rest of the file
     seg_buf.resize(WAL_SEGMENT_SIZE, 0);
+
+    if !first_page_only {
+        let block_offset = lsn.page_offset_in_segment(WAL_SEGMENT_SIZE) as usize;
+        let header = XLogPageHeaderData {
+            xlp_magic: XLOG_PAGE_MAGIC as u16,
+            xlp_info: if page_off >= pg_constants::SIZE_OF_PAGE_HEADER as u64 {
+                pg_constants::XLP_FIRST_IS_CONTRECORD
+            } else {
+                0
+            },
+            xlp_tli: PG_TLI,
+            xlp_pageaddr: lsn.page_lsn().0,
+            xlp_rem_len: if page_off >= pg_constants::SIZE_OF_PAGE_HEADER as u64 {
+                page_off as u32
+            } else {
+                0u32
+            },
+            ..Default::default() // Put 0 in padding fields.
+        };
+        let hdr_bytes = header.encode()?;
+
+        debug_assert!(seg_buf.len() > block_offset + hdr_bytes.len());
+        debug_assert_ne!(block_offset, 0);
+
+        seg_buf[block_offset..block_offset + hdr_bytes.len()].copy_from_slice(&hdr_bytes[..]);
+    }
+
     Ok(seg_buf.freeze())
 }
 
diff --git a/libs/postgres_ffi/wal_craft/src/lib.rs b/libs/postgres_ffi/wal_craft/src/lib.rs
index 969befc8e7..88ae41c636 100644
--- a/libs/postgres_ffi/wal_craft/src/lib.rs
+++ b/libs/postgres_ffi/wal_craft/src/lib.rs
@@ -6,9 +6,8 @@ use postgres::Client;
 use postgres_ffi::{WAL_SEGMENT_SIZE, XLOG_BLCKSZ};
 use postgres_ffi::{XLOG_SIZE_OF_XLOG_RECORD, XLOG_SIZE_OF_XLOG_SHORT_PHD};
 use std::cmp::Ordering;
-use std::fs;
 use std::path::{Path, PathBuf};
-use std::process::{Command, Stdio};
+use std::process::Command;
 use std::time::Instant;
 use tempfile::{tempdir, TempDir};
 
@@ -95,12 +94,6 @@ impl Conf {
 
     pub fn start_server(&self) -> Result<PostgresServer> {
         info!("Starting Postgres server in {:?}", self.datadir);
-        let log_file = fs::File::create(self.datadir.join("pg.log")).with_context(|| {
-            format!(
-                "Failed to create pg.log file in directory {}",
-                self.datadir.display()
-            )
-        })?;
         let unix_socket_dir = tempdir()?; // We need a directory with a short name for Unix socket (up to 108 symbols)
         let unix_socket_dir_path = unix_socket_dir.path().to_owned();
         let server_process = self
@@ -110,9 +103,7 @@ impl Conf {
             .arg(unix_socket_dir_path.as_os_str())
             .arg("-D")
             .arg(self.datadir.as_os_str())
-            .args(["-c", "logging_collector=on"]) // stderr will mess up with tests output
             .args(REQUIRED_POSTGRES_CONFIG.iter().flat_map(|cfg| ["-c", cfg]))
-            .stderr(Stdio::from(log_file))
             .spawn()?;
         let server = PostgresServer {
             process: server_process,
@@ -121,7 +112,7 @@ impl Conf {
                 let mut c = postgres::Config::new();
                 c.host_path(&unix_socket_dir_path);
                 c.user("postgres");
-                c.connect_timeout(Duration::from_millis(1000));
+                c.connect_timeout(Duration::from_millis(10000));
                 c
             },
         };
diff --git a/libs/pq_proto/Cargo.toml b/libs/pq_proto/Cargo.toml
index 76b71729ed..b286eb0358 100644
--- a/libs/pq_proto/Cargo.toml
+++ b/libs/pq_proto/Cargo.toml
@@ -10,7 +10,6 @@ byteorder.workspace = true
 pin-project-lite.workspace = true
 postgres-protocol.workspace = true
 rand.workspace = true
-serde.workspace = true
 tokio.workspace = true
 tracing.workspace = true
 thiserror.workspace = true
diff --git a/libs/pq_proto/src/lib.rs b/libs/pq_proto/src/lib.rs
index ed0239072a..2143ad2530 100644
--- a/libs/pq_proto/src/lib.rs
+++ b/libs/pq_proto/src/lib.rs
@@ -6,15 +6,10 @@ pub mod framed;
 
 use byteorder::{BigEndian, ReadBytesExt};
 use bytes::{Buf, BufMut, Bytes, BytesMut};
-use postgres_protocol::PG_EPOCH;
-use serde::{Deserialize, Serialize};
-use std::{
-    borrow::Cow,
-    collections::HashMap,
-    fmt, io, str,
-    time::{Duration, SystemTime},
-};
-use tracing::{trace, warn};
+use std::{borrow::Cow, collections::HashMap, fmt, io, str};
+
+// re-export for use in utils pageserver_feedback.rs
+pub use postgres_protocol::PG_EPOCH;
 
 pub type Oid = u32;
 pub type SystemId = u64;
@@ -664,7 +659,7 @@ fn write_cstr(s: impl AsRef<[u8]>, buf: &mut BytesMut) -> Result<(), ProtocolErr
 }
 
 /// Read cstring from buf, advancing it.
-fn read_cstr(buf: &mut Bytes) -> Result<Bytes, ProtocolError> {
+pub fn read_cstr(buf: &mut Bytes) -> Result<Bytes, ProtocolError> {
     let pos = buf
         .iter()
         .position(|x| *x == 0)
@@ -939,175 +934,10 @@ impl<'a> BeMessage<'a> {
     }
 }
 
-/// Feedback pageserver sends to safekeeper and safekeeper resends to compute.
-/// Serialized in custom flexible key/value format. In replication protocol, it
-/// is marked with NEON_STATUS_UPDATE_TAG_BYTE to differentiate from postgres
-/// Standby status update / Hot standby feedback messages.
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
-pub struct PageserverFeedback {
-    /// Last known size of the timeline. Used to enforce timeline size limit.
-    pub current_timeline_size: u64,
-    /// LSN last received and ingested by the pageserver.
-    pub last_received_lsn: u64,
-    /// LSN up to which data is persisted by the pageserver to its local disc.
-    pub disk_consistent_lsn: u64,
-    /// LSN up to which data is persisted by the pageserver on s3; safekeepers
-    /// consider WAL before it can be removed.
-    pub remote_consistent_lsn: u64,
-    pub replytime: SystemTime,
-}
-
-// NOTE: Do not forget to increment this number when adding new fields to PageserverFeedback.
-// Do not remove previously available fields because this might be backwards incompatible.
-pub const PAGESERVER_FEEDBACK_FIELDS_NUMBER: u8 = 5;
-
-impl PageserverFeedback {
-    pub fn empty() -> PageserverFeedback {
-        PageserverFeedback {
-            current_timeline_size: 0,
-            last_received_lsn: 0,
-            remote_consistent_lsn: 0,
-            disk_consistent_lsn: 0,
-            replytime: SystemTime::now(),
-        }
-    }
-
-    // Serialize PageserverFeedback using custom format
-    // to support protocol extensibility.
-    //
-    // Following layout is used:
-    // char - number of key-value pairs that follow.
-    //
-    // key-value pairs:
-    // null-terminated string - key,
-    // uint32 - value length in bytes
-    // value itself
-    //
-    // TODO: change serialized fields names once all computes migrate to rename.
-    pub fn serialize(&self, buf: &mut BytesMut) {
-        buf.put_u8(PAGESERVER_FEEDBACK_FIELDS_NUMBER); // # of keys
-        buf.put_slice(b"current_timeline_size\0");
-        buf.put_i32(8);
-        buf.put_u64(self.current_timeline_size);
-
-        buf.put_slice(b"ps_writelsn\0");
-        buf.put_i32(8);
-        buf.put_u64(self.last_received_lsn);
-        buf.put_slice(b"ps_flushlsn\0");
-        buf.put_i32(8);
-        buf.put_u64(self.disk_consistent_lsn);
-        buf.put_slice(b"ps_applylsn\0");
-        buf.put_i32(8);
-        buf.put_u64(self.remote_consistent_lsn);
-
-        let timestamp = self
-            .replytime
-            .duration_since(*PG_EPOCH)
-            .expect("failed to serialize pg_replytime earlier than PG_EPOCH")
-            .as_micros() as i64;
-
-        buf.put_slice(b"ps_replytime\0");
-        buf.put_i32(8);
-        buf.put_i64(timestamp);
-    }
-
-    // Deserialize PageserverFeedback message
-    // TODO: change serialized fields names once all computes migrate to rename.
-    pub fn parse(mut buf: Bytes) -> PageserverFeedback {
-        let mut rf = PageserverFeedback::empty();
-        let nfields = buf.get_u8();
-        for _ in 0..nfields {
-            let key = read_cstr(&mut buf).unwrap();
-            match key.as_ref() {
-                b"current_timeline_size" => {
-                    let len = buf.get_i32();
-                    assert_eq!(len, 8);
-                    rf.current_timeline_size = buf.get_u64();
-                }
-                b"ps_writelsn" => {
-                    let len = buf.get_i32();
-                    assert_eq!(len, 8);
-                    rf.last_received_lsn = buf.get_u64();
-                }
-                b"ps_flushlsn" => {
-                    let len = buf.get_i32();
-                    assert_eq!(len, 8);
-                    rf.disk_consistent_lsn = buf.get_u64();
-                }
-                b"ps_applylsn" => {
-                    let len = buf.get_i32();
-                    assert_eq!(len, 8);
-                    rf.remote_consistent_lsn = buf.get_u64();
-                }
-                b"ps_replytime" => {
-                    let len = buf.get_i32();
-                    assert_eq!(len, 8);
-                    let raw_time = buf.get_i64();
-                    if raw_time > 0 {
-                        rf.replytime = *PG_EPOCH + Duration::from_micros(raw_time as u64);
-                    } else {
-                        rf.replytime = *PG_EPOCH - Duration::from_micros(-raw_time as u64);
-                    }
-                }
-                _ => {
-                    let len = buf.get_i32();
-                    warn!(
-                        "PageserverFeedback parse. unknown key {} of len {len}. Skip it.",
-                        String::from_utf8_lossy(key.as_ref())
-                    );
-                    buf.advance(len as usize);
-                }
-            }
-        }
-        trace!("PageserverFeedback parsed is {:?}", rf);
-        rf
-    }
-}
-
 #[cfg(test)]
 mod tests {
     use super::*;
 
-    #[test]
-    fn test_replication_feedback_serialization() {
-        let mut rf = PageserverFeedback::empty();
-        // Fill rf with some values
-        rf.current_timeline_size = 12345678;
-        // Set rounded time to be able to compare it with deserialized value,
-        // because it is rounded up to microseconds during serialization.
-        rf.replytime = *PG_EPOCH + Duration::from_secs(100_000_000);
-        let mut data = BytesMut::new();
-        rf.serialize(&mut data);
-
-        let rf_parsed = PageserverFeedback::parse(data.freeze());
-        assert_eq!(rf, rf_parsed);
-    }
-
-    #[test]
-    fn test_replication_feedback_unknown_key() {
-        let mut rf = PageserverFeedback::empty();
-        // Fill rf with some values
-        rf.current_timeline_size = 12345678;
-        // Set rounded time to be able to compare it with deserialized value,
-        // because it is rounded up to microseconds during serialization.
-        rf.replytime = *PG_EPOCH + Duration::from_secs(100_000_000);
-        let mut data = BytesMut::new();
-        rf.serialize(&mut data);
-
-        // Add an extra field to the buffer and adjust number of keys
-        if let Some(first) = data.first_mut() {
-            *first = PAGESERVER_FEEDBACK_FIELDS_NUMBER + 1;
-        }
-
-        data.put_slice(b"new_field_one\0");
-        data.put_i32(8);
-        data.put_u64(42);
-
-        // Parse serialized data and check that new field is not parsed
-        let rf_parsed = PageserverFeedback::parse(data.freeze());
-        assert_eq!(rf, rf_parsed);
-    }
-
     #[test]
     fn test_startup_message_params_options_escaped() {
         fn split_options(params: &StartupMessageParams) -> Vec<Cow<'_, str>> {
diff --git a/libs/remote_storage/tests/pagination_tests.rs b/libs/remote_storage/tests/pagination_tests.rs
index eb52409c44..86a6888f98 100644
--- a/libs/remote_storage/tests/pagination_tests.rs
+++ b/libs/remote_storage/tests/pagination_tests.rs
@@ -99,7 +99,11 @@ struct S3WithTestBlobs {
 #[async_trait::async_trait]
 impl AsyncTestContext for MaybeEnabledS3 {
     async fn setup() -> Self {
-        utils::logging::init(utils::logging::LogFormat::Test).expect("logging init failed");
+        utils::logging::init(
+            utils::logging::LogFormat::Test,
+            utils::logging::TracingErrorLayerEnablement::Disabled,
+        )
+        .expect("logging init failed");
         if env::var(ENABLE_REAL_S3_REMOTE_STORAGE_ENV_VAR_NAME).is_err() {
             info!(
                 "`{}` env variable is not set, skipping the test",
@@ -204,12 +208,7 @@ async fn upload_s3_data(
             let data = format!("remote blob data {i}").into_bytes();
             let data_len = data.len();
             task_client
-                .upload(
-                    Box::new(std::io::Cursor::new(data)),
-                    data_len,
-                    &blob_path,
-                    None,
-                )
+                .upload(std::io::Cursor::new(data), data_len, &blob_path, None)
                 .await?;
 
             Ok::<_, anyhow::Error>((blob_prefix, blob_path))
diff --git a/libs/tracing-utils/Cargo.toml b/libs/tracing-utils/Cargo.toml
index 8c3d3f9063..b285c9b5b0 100644
--- a/libs/tracing-utils/Cargo.toml
+++ b/libs/tracing-utils/Cargo.toml
@@ -14,4 +14,5 @@ tokio = { workspace = true, features = ["rt", "rt-multi-thread"] }
 tracing.workspace = true
 tracing-opentelemetry.workspace = true
 tracing-subscriber.workspace = true
-workspace_hack = { version = "0.1", path = "../../workspace_hack" }
+
+workspace_hack.workspace = true
diff --git a/libs/utils/Cargo.toml b/libs/utils/Cargo.toml
index 391bc52a80..8239ffff57 100644
--- a/libs/utils/Cargo.toml
+++ b/libs/utils/Cargo.toml
@@ -11,6 +11,7 @@ async-trait.workspace = true
 anyhow.workspace = true
 bincode.workspace = true
 bytes.workspace = true
+chrono.workspace = true
 heapless.workspace = true
 hex = { workspace = true, features = ["serde"] }
 hyper = { workspace = true, features = ["full"] }
@@ -27,14 +28,16 @@ signal-hook.workspace = true
 thiserror.workspace = true
 tokio.workspace = true
 tracing.workspace = true
-tracing-subscriber = { workspace = true, features = ["json"] }
+tracing-error.workspace = true
+tracing-subscriber = { workspace = true, features = ["json", "registry"] }
 rand.workspace = true
 serde_with.workspace = true
 strum.workspace = true
 strum_macros.workspace = true
 url.workspace = true
-uuid = { version = "1.2", features = ["v4", "serde"] }
+uuid.workspace = true
 
+pq_proto.workspace = true
 metrics.workspace = true
 workspace_hack.workspace = true
 
diff --git a/libs/utils/src/http/endpoint.rs b/libs/utils/src/http/endpoint.rs
index 616f2b8468..b11aef9892 100644
--- a/libs/utils/src/http/endpoint.rs
+++ b/libs/utils/src/http/endpoint.rs
@@ -76,6 +76,7 @@ where
 
         let log_quietly = method == Method::GET;
         async move {
+            let cancellation_guard = RequestCancelled::warn_when_dropped_without_responding();
             if log_quietly {
                 debug!("Handling request");
             } else {
@@ -87,7 +88,11 @@ where
             // Usage of the error handler also means that we expect only the `ApiError` errors to be raised in this call.
             //
             // Panics are not handled separately, there's a `tracing_panic_hook` from another module to do that globally.
-            match (self.0)(request).await {
+            let res = (self.0)(request).await;
+
+            cancellation_guard.disarm();
+
+            match res {
                 Ok(response) => {
                     let response_status = response.status();
                     if log_quietly && response_status.is_success() {
@@ -105,6 +110,38 @@ where
     }
 }
 
+/// Drop guard to WARN in case the request was dropped before completion.
+struct RequestCancelled {
+    warn: Option<tracing::Span>,
+}
+
+impl RequestCancelled {
+    /// Create the drop guard using the [`tracing::Span::current`] as the span.
+    fn warn_when_dropped_without_responding() -> Self {
+        RequestCancelled {
+            warn: Some(tracing::Span::current()),
+        }
+    }
+
+    /// Consume the drop guard without logging anything.
+    fn disarm(mut self) {
+        self.warn = None;
+    }
+}
+
+impl Drop for RequestCancelled {
+    fn drop(&mut self) {
+        if let Some(span) = self.warn.take() {
+            // the span has all of the info already, but the outer `.instrument(span)` has already
+            // been dropped, so we need to manually re-enter it for this message.
+            //
+            // this is what the instrument would do before polling so it is fine.
+            let _g = span.entered();
+            warn!("request was dropped before completing");
+        }
+    }
+}
+
 async fn prometheus_metrics_handler(_req: Request<Body>) -> Result<Response<Body>, ApiError> {
     SERVE_METRICS_COUNT.inc();
 
diff --git a/libs/utils/src/http/json.rs b/libs/utils/src/http/json.rs
index 40e61e3d0c..8981fdd1dd 100644
--- a/libs/utils/src/http/json.rs
+++ b/libs/utils/src/http/json.rs
@@ -1,9 +1,7 @@
-use std::fmt::Display;
-
 use anyhow::Context;
 use bytes::Buf;
 use hyper::{header, Body, Request, Response, StatusCode};
-use serde::{Deserialize, Serialize, Serializer};
+use serde::{Deserialize, Serialize};
 
 use super::error::ApiError;
 
@@ -33,12 +31,3 @@ pub fn json_response<T: Serialize>(
         .map_err(|e| ApiError::InternalServerError(e.into()))?;
     Ok(response)
 }
-
-/// Serialize through Display trait.
-pub fn display_serialize<S, F>(z: &F, s: S) -> Result<S::Ok, S::Error>
-where
-    S: Serializer,
-    F: Display,
-{
-    s.serialize_str(&format!("{}", z))
-}
diff --git a/libs/utils/src/id.rs b/libs/utils/src/id.rs
index b27c5cda35..20b601f68d 100644
--- a/libs/utils/src/id.rs
+++ b/libs/utils/src/id.rs
@@ -265,6 +265,26 @@ impl fmt::Display for TenantTimelineId {
     }
 }
 
+impl FromStr for TenantTimelineId {
+    type Err = anyhow::Error;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        let mut parts = s.split('/');
+        let tenant_id = parts
+            .next()
+            .ok_or_else(|| anyhow::anyhow!("TenantTimelineId must contain tenant_id"))?
+            .parse()?;
+        let timeline_id = parts
+            .next()
+            .ok_or_else(|| anyhow::anyhow!("TenantTimelineId must contain timeline_id"))?
+            .parse()?;
+        if parts.next().is_some() {
+            anyhow::bail!("TenantTimelineId must contain only tenant_id and timeline_id");
+        }
+        Ok(TenantTimelineId::new(tenant_id, timeline_id))
+    }
+}
+
 // Unique ID of a storage node (safekeeper or pageserver). Supposed to be issued
 // by the console.
 #[derive(Clone, Copy, Eq, Ord, PartialEq, PartialOrd, Hash, Debug, Serialize, Deserialize)]
diff --git a/libs/utils/src/lib.rs b/libs/utils/src/lib.rs
index d4176911ac..82701ed4b0 100644
--- a/libs/utils/src/lib.rs
+++ b/libs/utils/src/lib.rs
@@ -54,6 +54,10 @@ pub mod measured_stream;
 pub mod serde_percent;
 pub mod serde_regex;
 
+pub mod pageserver_feedback;
+
+pub mod tracing_span_assert;
+
 /// use with fail::cfg("$name", "return(2000)")
 #[macro_export]
 macro_rules! failpoint_sleep_millis_async {
diff --git a/libs/utils/src/logging.rs b/libs/utils/src/logging.rs
index f770622a60..2b8c852d86 100644
--- a/libs/utils/src/logging.rs
+++ b/libs/utils/src/logging.rs
@@ -1,6 +1,7 @@
 use std::str::FromStr;
 
 use anyhow::Context;
+use once_cell::sync::Lazy;
 use strum_macros::{EnumString, EnumVariantNames};
 
 #[derive(EnumString, EnumVariantNames, Eq, PartialEq, Debug, Clone, Copy)]
@@ -23,24 +24,81 @@ impl LogFormat {
     }
 }
 
-pub fn init(log_format: LogFormat) -> anyhow::Result<()> {
-    let default_filter_str = "info";
+static TRACING_EVENT_COUNT: Lazy<metrics::IntCounterVec> = Lazy::new(|| {
+    metrics::register_int_counter_vec!(
+        "libmetrics_tracing_event_count",
+        "Number of tracing events, by level",
+        &["level"]
+    )
+    .expect("failed to define metric")
+});
 
+struct TracingEventCountLayer(&'static metrics::IntCounterVec);
+
+impl<S> tracing_subscriber::layer::Layer<S> for TracingEventCountLayer
+where
+    S: tracing::Subscriber,
+{
+    fn on_event(
+        &self,
+        event: &tracing::Event<'_>,
+        _ctx: tracing_subscriber::layer::Context<'_, S>,
+    ) {
+        let level = event.metadata().level();
+        let level = match *level {
+            tracing::Level::ERROR => "error",
+            tracing::Level::WARN => "warn",
+            tracing::Level::INFO => "info",
+            tracing::Level::DEBUG => "debug",
+            tracing::Level::TRACE => "trace",
+        };
+        self.0.with_label_values(&[level]).inc();
+    }
+}
+
+/// Whether to add the `tracing_error` crate's `ErrorLayer`
+/// to the global tracing subscriber.
+///
+pub enum TracingErrorLayerEnablement {
+    /// Do not add the `ErrorLayer`.
+    Disabled,
+    /// Add the `ErrorLayer` with the filter specified by RUST_LOG, defaulting to `info` if `RUST_LOG` is unset.
+    EnableWithRustLogFilter,
+}
+
+pub fn init(
+    log_format: LogFormat,
+    tracing_error_layer_enablement: TracingErrorLayerEnablement,
+) -> anyhow::Result<()> {
     // We fall back to printing all spans at info-level or above if
     // the RUST_LOG environment variable is not set.
-    let env_filter = tracing_subscriber::EnvFilter::try_from_default_env()
-        .unwrap_or_else(|_| tracing_subscriber::EnvFilter::new(default_filter_str));
+    let rust_log_env_filter = || {
+        tracing_subscriber::EnvFilter::try_from_default_env()
+            .unwrap_or_else(|_| tracing_subscriber::EnvFilter::new("info"))
+    };
 
-    let base_logger = tracing_subscriber::fmt()
-        .with_env_filter(env_filter)
-        .with_target(false)
-        .with_ansi(atty::is(atty::Stream::Stdout))
-        .with_writer(std::io::stdout);
-
-    match log_format {
-        LogFormat::Json => base_logger.json().init(),
-        LogFormat::Plain => base_logger.init(),
-        LogFormat::Test => base_logger.with_test_writer().init(),
+    // NB: the order of the with() calls does not matter.
+    // See https://docs.rs/tracing-subscriber/0.3.16/tracing_subscriber/layer/index.html#per-layer-filtering
+    use tracing_subscriber::prelude::*;
+    let r = tracing_subscriber::registry();
+    let r = r.with({
+        let log_layer = tracing_subscriber::fmt::layer()
+            .with_target(false)
+            .with_ansi(atty::is(atty::Stream::Stdout))
+            .with_writer(std::io::stdout);
+        let log_layer = match log_format {
+            LogFormat::Json => log_layer.json().boxed(),
+            LogFormat::Plain => log_layer.boxed(),
+            LogFormat::Test => log_layer.with_test_writer().boxed(),
+        };
+        log_layer.with_filter(rust_log_env_filter())
+    });
+    let r = r.with(TracingEventCountLayer(&TRACING_EVENT_COUNT).with_filter(rust_log_env_filter()));
+    match tracing_error_layer_enablement {
+        TracingErrorLayerEnablement::EnableWithRustLogFilter => r
+            .with(tracing_error::ErrorLayer::default().with_filter(rust_log_env_filter()))
+            .init(),
+        TracingErrorLayerEnablement::Disabled => r.init(),
     }
 
     Ok(())
@@ -157,3 +215,33 @@ impl std::fmt::Debug for PrettyLocation<'_, '_> {
         <Self as std::fmt::Display>::fmt(self, f)
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use metrics::{core::Opts, IntCounterVec};
+
+    use super::TracingEventCountLayer;
+
+    #[test]
+    fn tracing_event_count_metric() {
+        let counter_vec =
+            IntCounterVec::new(Opts::new("testmetric", "testhelp"), &["level"]).unwrap();
+        let counter_vec = Box::leak(Box::new(counter_vec)); // make it 'static
+        let layer = TracingEventCountLayer(counter_vec);
+        use tracing_subscriber::prelude::*;
+
+        tracing::subscriber::with_default(tracing_subscriber::registry().with(layer), || {
+            tracing::trace!("foo");
+            tracing::debug!("foo");
+            tracing::info!("foo");
+            tracing::warn!("foo");
+            tracing::error!("foo");
+        });
+
+        assert_eq!(counter_vec.with_label_values(&["trace"]).get(), 1);
+        assert_eq!(counter_vec.with_label_values(&["debug"]).get(), 1);
+        assert_eq!(counter_vec.with_label_values(&["info"]).get(), 1);
+        assert_eq!(counter_vec.with_label_values(&["warn"]).get(), 1);
+        assert_eq!(counter_vec.with_label_values(&["error"]).get(), 1);
+    }
+}
diff --git a/libs/utils/src/lsn.rs b/libs/utils/src/lsn.rs
index acf5ea28d7..0493d43088 100644
--- a/libs/utils/src/lsn.rs
+++ b/libs/utils/src/lsn.rs
@@ -62,29 +62,48 @@ impl Lsn {
     }
 
     /// Compute the offset into a segment
+    #[inline]
     pub fn segment_offset(self, seg_sz: usize) -> usize {
         (self.0 % seg_sz as u64) as usize
     }
 
     /// Compute LSN of the segment start.
+    #[inline]
     pub fn segment_lsn(self, seg_sz: usize) -> Lsn {
         Lsn(self.0 - (self.0 % seg_sz as u64))
     }
 
     /// Compute the segment number
+    #[inline]
     pub fn segment_number(self, seg_sz: usize) -> u64 {
         self.0 / seg_sz as u64
     }
 
     /// Compute the offset into a block
+    #[inline]
     pub fn block_offset(self) -> u64 {
         const BLCKSZ: u64 = XLOG_BLCKSZ as u64;
         self.0 % BLCKSZ
     }
 
+    /// Compute the block offset of the first byte of this Lsn within this
+    /// segment
+    #[inline]
+    pub fn page_lsn(self) -> Lsn {
+        Lsn(self.0 - self.block_offset())
+    }
+
+    /// Compute the block offset of the first byte of this Lsn within this
+    /// segment
+    #[inline]
+    pub fn page_offset_in_segment(self, seg_sz: usize) -> u64 {
+        (self.0 - self.block_offset()) - self.segment_lsn(seg_sz).0
+    }
+
     /// Compute the bytes remaining in this block
     ///
     /// If the LSN is already at the block boundary, it will return `XLOG_BLCKSZ`.
+    #[inline]
     pub fn remaining_in_block(self) -> u64 {
         const BLCKSZ: u64 = XLOG_BLCKSZ as u64;
         BLCKSZ - (self.0 % BLCKSZ)
diff --git a/libs/utils/src/pageserver_feedback.rs b/libs/utils/src/pageserver_feedback.rs
new file mode 100644
index 0000000000..a3b53201d3
--- /dev/null
+++ b/libs/utils/src/pageserver_feedback.rs
@@ -0,0 +1,214 @@
+use std::time::{Duration, SystemTime};
+
+use bytes::{Buf, BufMut, Bytes, BytesMut};
+use pq_proto::{read_cstr, PG_EPOCH};
+use serde::{Deserialize, Serialize};
+use serde_with::{serde_as, DisplayFromStr};
+use tracing::{trace, warn};
+
+use crate::lsn::Lsn;
+
+/// Feedback pageserver sends to safekeeper and safekeeper resends to compute.
+/// Serialized in custom flexible key/value format. In replication protocol, it
+/// is marked with NEON_STATUS_UPDATE_TAG_BYTE to differentiate from postgres
+/// Standby status update / Hot standby feedback messages.
+///
+/// serde Serialize is used only for human readable dump to json (e.g. in
+/// safekeepers debug_dump).
+#[serde_as]
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+pub struct PageserverFeedback {
+    /// Last known size of the timeline. Used to enforce timeline size limit.
+    pub current_timeline_size: u64,
+    /// LSN last received and ingested by the pageserver. Controls backpressure.
+    #[serde_as(as = "DisplayFromStr")]
+    pub last_received_lsn: Lsn,
+    /// LSN up to which data is persisted by the pageserver to its local disc.
+    /// Controls backpressure.
+    #[serde_as(as = "DisplayFromStr")]
+    pub disk_consistent_lsn: Lsn,
+    /// LSN up to which data is persisted by the pageserver on s3; safekeepers
+    /// consider WAL before it can be removed.
+    #[serde_as(as = "DisplayFromStr")]
+    pub remote_consistent_lsn: Lsn,
+    // Serialize with RFC3339 format.
+    #[serde(with = "serde_systemtime")]
+    pub replytime: SystemTime,
+}
+
+// NOTE: Do not forget to increment this number when adding new fields to PageserverFeedback.
+// Do not remove previously available fields because this might be backwards incompatible.
+pub const PAGESERVER_FEEDBACK_FIELDS_NUMBER: u8 = 5;
+
+impl PageserverFeedback {
+    pub fn empty() -> PageserverFeedback {
+        PageserverFeedback {
+            current_timeline_size: 0,
+            last_received_lsn: Lsn::INVALID,
+            remote_consistent_lsn: Lsn::INVALID,
+            disk_consistent_lsn: Lsn::INVALID,
+            replytime: *PG_EPOCH,
+        }
+    }
+
+    // Serialize PageserverFeedback using custom format
+    // to support protocol extensibility.
+    //
+    // Following layout is used:
+    // char - number of key-value pairs that follow.
+    //
+    // key-value pairs:
+    // null-terminated string - key,
+    // uint32 - value length in bytes
+    // value itself
+    //
+    // TODO: change serialized fields names once all computes migrate to rename.
+    pub fn serialize(&self, buf: &mut BytesMut) {
+        buf.put_u8(PAGESERVER_FEEDBACK_FIELDS_NUMBER); // # of keys
+        buf.put_slice(b"current_timeline_size\0");
+        buf.put_i32(8);
+        buf.put_u64(self.current_timeline_size);
+
+        buf.put_slice(b"ps_writelsn\0");
+        buf.put_i32(8);
+        buf.put_u64(self.last_received_lsn.0);
+        buf.put_slice(b"ps_flushlsn\0");
+        buf.put_i32(8);
+        buf.put_u64(self.disk_consistent_lsn.0);
+        buf.put_slice(b"ps_applylsn\0");
+        buf.put_i32(8);
+        buf.put_u64(self.remote_consistent_lsn.0);
+
+        let timestamp = self
+            .replytime
+            .duration_since(*PG_EPOCH)
+            .expect("failed to serialize pg_replytime earlier than PG_EPOCH")
+            .as_micros() as i64;
+
+        buf.put_slice(b"ps_replytime\0");
+        buf.put_i32(8);
+        buf.put_i64(timestamp);
+    }
+
+    // Deserialize PageserverFeedback message
+    // TODO: change serialized fields names once all computes migrate to rename.
+    pub fn parse(mut buf: Bytes) -> PageserverFeedback {
+        let mut rf = PageserverFeedback::empty();
+        let nfields = buf.get_u8();
+        for _ in 0..nfields {
+            let key = read_cstr(&mut buf).unwrap();
+            match key.as_ref() {
+                b"current_timeline_size" => {
+                    let len = buf.get_i32();
+                    assert_eq!(len, 8);
+                    rf.current_timeline_size = buf.get_u64();
+                }
+                b"ps_writelsn" => {
+                    let len = buf.get_i32();
+                    assert_eq!(len, 8);
+                    rf.last_received_lsn = Lsn(buf.get_u64());
+                }
+                b"ps_flushlsn" => {
+                    let len = buf.get_i32();
+                    assert_eq!(len, 8);
+                    rf.disk_consistent_lsn = Lsn(buf.get_u64());
+                }
+                b"ps_applylsn" => {
+                    let len = buf.get_i32();
+                    assert_eq!(len, 8);
+                    rf.remote_consistent_lsn = Lsn(buf.get_u64());
+                }
+                b"ps_replytime" => {
+                    let len = buf.get_i32();
+                    assert_eq!(len, 8);
+                    let raw_time = buf.get_i64();
+                    if raw_time > 0 {
+                        rf.replytime = *PG_EPOCH + Duration::from_micros(raw_time as u64);
+                    } else {
+                        rf.replytime = *PG_EPOCH - Duration::from_micros(-raw_time as u64);
+                    }
+                }
+                _ => {
+                    let len = buf.get_i32();
+                    warn!(
+                        "PageserverFeedback parse. unknown key {} of len {len}. Skip it.",
+                        String::from_utf8_lossy(key.as_ref())
+                    );
+                    buf.advance(len as usize);
+                }
+            }
+        }
+        trace!("PageserverFeedback parsed is {:?}", rf);
+        rf
+    }
+}
+
+mod serde_systemtime {
+    use std::time::SystemTime;
+
+    use chrono::{DateTime, Utc};
+    use serde::{Deserialize, Deserializer, Serializer};
+
+    pub fn serialize<S>(ts: &SystemTime, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: Serializer,
+    {
+        let chrono_dt: DateTime<Utc> = (*ts).into();
+        serializer.serialize_str(&chrono_dt.to_rfc3339())
+    }
+
+    pub fn deserialize<'de, D>(deserializer: D) -> Result<SystemTime, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        let time: String = Deserialize::deserialize(deserializer)?;
+        Ok(DateTime::parse_from_rfc3339(&time)
+            .map_err(serde::de::Error::custom)?
+            .into())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_replication_feedback_serialization() {
+        let mut rf = PageserverFeedback::empty();
+        // Fill rf with some values
+        rf.current_timeline_size = 12345678;
+        // Set rounded time to be able to compare it with deserialized value,
+        // because it is rounded up to microseconds during serialization.
+        rf.replytime = *PG_EPOCH + Duration::from_secs(100_000_000);
+        let mut data = BytesMut::new();
+        rf.serialize(&mut data);
+
+        let rf_parsed = PageserverFeedback::parse(data.freeze());
+        assert_eq!(rf, rf_parsed);
+    }
+
+    #[test]
+    fn test_replication_feedback_unknown_key() {
+        let mut rf = PageserverFeedback::empty();
+        // Fill rf with some values
+        rf.current_timeline_size = 12345678;
+        // Set rounded time to be able to compare it with deserialized value,
+        // because it is rounded up to microseconds during serialization.
+        rf.replytime = *PG_EPOCH + Duration::from_secs(100_000_000);
+        let mut data = BytesMut::new();
+        rf.serialize(&mut data);
+
+        // Add an extra field to the buffer and adjust number of keys
+        if let Some(first) = data.first_mut() {
+            *first = PAGESERVER_FEEDBACK_FIELDS_NUMBER + 1;
+        }
+
+        data.put_slice(b"new_field_one\0");
+        data.put_i32(8);
+        data.put_u64(42);
+
+        // Parse serialized data and check that new field is not parsed
+        let rf_parsed = PageserverFeedback::parse(data.freeze());
+        assert_eq!(rf, rf_parsed);
+    }
+}
diff --git a/libs/utils/src/tracing_span_assert.rs b/libs/utils/src/tracing_span_assert.rs
new file mode 100644
index 0000000000..b9f7986442
--- /dev/null
+++ b/libs/utils/src/tracing_span_assert.rs
@@ -0,0 +1,287 @@
+//! Assert that the current [`tracing::Span`] has a given set of fields.
+//!
+//! # Usage
+//!
+//! ```
+//! use tracing_subscriber::prelude::*;
+//! let registry = tracing_subscriber::registry()
+//!    .with(tracing_error::ErrorLayer::default());
+//!
+//! // Register the registry as the global subscriber.
+//! // In this example, we'll only use it as a thread-local subscriber.
+//! let _guard = tracing::subscriber::set_default(registry);
+//!
+//! // Then, in the main code:
+//!
+//! let span = tracing::info_span!("TestSpan", test_id = 1);
+//! let _guard = span.enter();
+//!
+//! // ... down the call stack
+//!
+//! use utils::tracing_span_assert::{check_fields_present, MultiNameExtractor};
+//! let extractor = MultiNameExtractor::new("TestExtractor", ["test", "test_id"]);
+//! match check_fields_present([&extractor]) {
+//!    Ok(()) => {},
+//!    Err(missing) => {
+//!        panic!("Missing fields: {:?}", missing.into_iter().map(|f| f.name() ).collect::<Vec<_>>());
+//!    }
+//! }
+//! ```
+//!
+//! Recommended reading: https://docs.rs/tracing-subscriber/0.3.16/tracing_subscriber/layer/index.html#per-layer-filtering
+//!
+
+use std::{
+    collections::HashSet,
+    fmt::{self},
+    hash::{Hash, Hasher},
+};
+
+pub enum ExtractionResult {
+    Present,
+    Absent,
+}
+
+pub trait Extractor: Send + Sync + std::fmt::Debug {
+    fn name(&self) -> &str;
+    fn extract(&self, fields: &tracing::field::FieldSet) -> ExtractionResult;
+}
+
+#[derive(Debug)]
+pub struct MultiNameExtractor<const L: usize> {
+    name: &'static str,
+    field_names: [&'static str; L],
+}
+
+impl<const L: usize> MultiNameExtractor<L> {
+    pub fn new(name: &'static str, field_names: [&'static str; L]) -> MultiNameExtractor<L> {
+        MultiNameExtractor { name, field_names }
+    }
+}
+impl<const L: usize> Extractor for MultiNameExtractor<L> {
+    fn name(&self) -> &str {
+        self.name
+    }
+    fn extract(&self, fields: &tracing::field::FieldSet) -> ExtractionResult {
+        if fields.iter().any(|f| self.field_names.contains(&f.name())) {
+            ExtractionResult::Present
+        } else {
+            ExtractionResult::Absent
+        }
+    }
+}
+
+struct MemoryIdentity<'a>(&'a dyn Extractor);
+
+impl<'a> MemoryIdentity<'a> {
+    fn as_ptr(&self) -> *const () {
+        self.0 as *const _ as *const ()
+    }
+}
+impl<'a> PartialEq for MemoryIdentity<'a> {
+    fn eq(&self, other: &Self) -> bool {
+        self.as_ptr() == other.as_ptr()
+    }
+}
+impl<'a> Eq for MemoryIdentity<'a> {}
+impl<'a> Hash for MemoryIdentity<'a> {
+    fn hash<H: Hasher>(&self, state: &mut H) {
+        self.as_ptr().hash(state);
+    }
+}
+impl<'a> fmt::Debug for MemoryIdentity<'a> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{:p}: {}", self.as_ptr(), self.0.name())
+    }
+}
+
+/// The extractor names passed as keys to [`new`].
+pub fn check_fields_present<const L: usize>(
+    must_be_present: [&dyn Extractor; L],
+) -> Result<(), Vec<&dyn Extractor>> {
+    let mut missing: HashSet<MemoryIdentity> =
+        HashSet::from_iter(must_be_present.into_iter().map(|r| MemoryIdentity(r)));
+    let trace = tracing_error::SpanTrace::capture();
+    trace.with_spans(|md, _formatted_fields| {
+        missing.retain(|extractor| match extractor.0.extract(md.fields()) {
+            ExtractionResult::Present => false,
+            ExtractionResult::Absent => true,
+        });
+        !missing.is_empty() // continue walking up until we've found all missing
+    });
+    if missing.is_empty() {
+        Ok(())
+    } else {
+        Err(missing.into_iter().map(|mi| mi.0).collect())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    use tracing_subscriber::prelude::*;
+
+    use super::*;
+
+    struct Setup {
+        _current_thread_subscriber_guard: tracing::subscriber::DefaultGuard,
+        tenant_extractor: MultiNameExtractor<2>,
+        timeline_extractor: MultiNameExtractor<2>,
+    }
+
+    fn setup_current_thread() -> Setup {
+        let tenant_extractor = MultiNameExtractor::new("TenantId", ["tenant_id", "tenant"]);
+        let timeline_extractor = MultiNameExtractor::new("TimelineId", ["timeline_id", "timeline"]);
+
+        let registry = tracing_subscriber::registry()
+            .with(tracing_subscriber::fmt::layer())
+            .with(tracing_error::ErrorLayer::default());
+
+        let guard = tracing::subscriber::set_default(registry);
+
+        Setup {
+            _current_thread_subscriber_guard: guard,
+            tenant_extractor,
+            timeline_extractor,
+        }
+    }
+
+    fn assert_missing(missing: Vec<&dyn Extractor>, expected: Vec<&dyn Extractor>) {
+        let missing: HashSet<MemoryIdentity> =
+            HashSet::from_iter(missing.into_iter().map(MemoryIdentity));
+        let expected: HashSet<MemoryIdentity> =
+            HashSet::from_iter(expected.into_iter().map(MemoryIdentity));
+        assert_eq!(missing, expected);
+    }
+
+    #[test]
+    fn positive_one_level() {
+        let setup = setup_current_thread();
+        let span = tracing::info_span!("root", tenant_id = "tenant-1", timeline_id = "timeline-1");
+        let _guard = span.enter();
+        check_fields_present([&setup.tenant_extractor, &setup.timeline_extractor]).unwrap();
+    }
+
+    #[test]
+    fn negative_one_level() {
+        let setup = setup_current_thread();
+        let span = tracing::info_span!("root", timeline_id = "timeline-1");
+        let _guard = span.enter();
+        let missing =
+            check_fields_present([&setup.tenant_extractor, &setup.timeline_extractor]).unwrap_err();
+        assert_missing(missing, vec![&setup.tenant_extractor]);
+    }
+
+    #[test]
+    fn positive_multiple_levels() {
+        let setup = setup_current_thread();
+
+        let span = tracing::info_span!("root");
+        let _guard = span.enter();
+
+        let span = tracing::info_span!("child", tenant_id = "tenant-1");
+        let _guard = span.enter();
+
+        let span = tracing::info_span!("grandchild", timeline_id = "timeline-1");
+        let _guard = span.enter();
+
+        check_fields_present([&setup.tenant_extractor, &setup.timeline_extractor]).unwrap();
+    }
+
+    #[test]
+    fn negative_multiple_levels() {
+        let setup = setup_current_thread();
+
+        let span = tracing::info_span!("root");
+        let _guard = span.enter();
+
+        let span = tracing::info_span!("child", timeline_id = "timeline-1");
+        let _guard = span.enter();
+
+        let missing = check_fields_present([&setup.tenant_extractor]).unwrap_err();
+        assert_missing(missing, vec![&setup.tenant_extractor]);
+    }
+
+    #[test]
+    fn positive_subset_one_level() {
+        let setup = setup_current_thread();
+        let span = tracing::info_span!("root", tenant_id = "tenant-1", timeline_id = "timeline-1");
+        let _guard = span.enter();
+        check_fields_present([&setup.tenant_extractor]).unwrap();
+    }
+
+    #[test]
+    fn positive_subset_multiple_levels() {
+        let setup = setup_current_thread();
+
+        let span = tracing::info_span!("root");
+        let _guard = span.enter();
+
+        let span = tracing::info_span!("child", tenant_id = "tenant-1");
+        let _guard = span.enter();
+
+        let span = tracing::info_span!("grandchild", timeline_id = "timeline-1");
+        let _guard = span.enter();
+
+        check_fields_present([&setup.tenant_extractor]).unwrap();
+    }
+
+    #[test]
+    fn negative_subset_one_level() {
+        let setup = setup_current_thread();
+        let span = tracing::info_span!("root", timeline_id = "timeline-1");
+        let _guard = span.enter();
+        let missing = check_fields_present([&setup.tenant_extractor]).unwrap_err();
+        assert_missing(missing, vec![&setup.tenant_extractor]);
+    }
+
+    #[test]
+    fn negative_subset_multiple_levels() {
+        let setup = setup_current_thread();
+
+        let span = tracing::info_span!("root");
+        let _guard = span.enter();
+
+        let span = tracing::info_span!("child", timeline_id = "timeline-1");
+        let _guard = span.enter();
+
+        let missing = check_fields_present([&setup.tenant_extractor]).unwrap_err();
+        assert_missing(missing, vec![&setup.tenant_extractor]);
+    }
+
+    #[test]
+    fn tracing_error_subscriber_not_set_up() {
+        // no setup
+
+        let span = tracing::info_span!("foo", e = "some value");
+        let _guard = span.enter();
+
+        let extractor = MultiNameExtractor::new("E", ["e"]);
+        let missing = check_fields_present([&extractor]).unwrap_err();
+        assert_missing(missing, vec![&extractor]);
+    }
+
+    #[test]
+    #[should_panic]
+    fn panics_if_tracing_error_subscriber_has_wrong_filter() {
+        let r = tracing_subscriber::registry().with({
+            tracing_error::ErrorLayer::default().with_filter(
+                tracing_subscriber::filter::dynamic_filter_fn(|md, _| {
+                    if md.is_span() && *md.level() == tracing::Level::INFO {
+                        return false;
+                    }
+                    true
+                }),
+            )
+        });
+
+        let _guard = tracing::subscriber::set_default(r);
+
+        let span = tracing::info_span!("foo", e = "some value");
+        let _guard = span.enter();
+
+        let extractor = MultiNameExtractor::new("E", ["e"]);
+        let missing = check_fields_present([&extractor]).unwrap_err();
+        assert_missing(missing, vec![&extractor]);
+    }
+}
diff --git a/pageserver/Cargo.toml b/pageserver/Cargo.toml
index 0bc7eba95e..ea81544cbe 100644
--- a/pageserver/Cargo.toml
+++ b/pageserver/Cargo.toml
@@ -52,6 +52,7 @@ sync_wrapper.workspace = true
 tokio-tar.workspace = true
 thiserror.workspace = true
 tokio = { workspace = true, features = ["process", "sync", "fs", "rt", "io-util", "time"] }
+tokio-io-timeout.workspace = true
 tokio-postgres.workspace = true
 tokio-util.workspace = true
 toml_edit = { workspace = true, features = [ "serde" ] }
diff --git a/pageserver/benches/bench_layer_map.rs b/pageserver/benches/bench_layer_map.rs
index 5edfa84d8a..ee5980212e 100644
--- a/pageserver/benches/bench_layer_map.rs
+++ b/pageserver/benches/bench_layer_map.rs
@@ -13,7 +13,7 @@ use std::time::Instant;
 
 use utils::lsn::Lsn;
 
-use criterion::{criterion_group, criterion_main, Criterion};
+use criterion::{black_box, criterion_group, criterion_main, Criterion};
 
 fn build_layer_map(filename_dump: PathBuf) -> LayerMap<LayerDescriptor> {
     let mut layer_map = LayerMap::<LayerDescriptor>::default();
@@ -114,7 +114,7 @@ fn bench_from_captest_env(c: &mut Criterion) {
     c.bench_function("captest_uniform_queries", |b| {
         b.iter(|| {
             for q in queries.clone().into_iter() {
-                layer_map.search(q.0, q.1);
+                black_box(layer_map.search(q.0, q.1));
             }
         });
     });
@@ -122,11 +122,11 @@ fn bench_from_captest_env(c: &mut Criterion) {
     // test with a key that corresponds to the RelDir entry. See pgdatadir_mapping.rs.
     c.bench_function("captest_rel_dir_query", |b| {
         b.iter(|| {
-            let result = layer_map.search(
+            let result = black_box(layer_map.search(
                 Key::from_hex("000000067F00008000000000000000000001").unwrap(),
                 // This LSN is higher than any of the LSNs in the tree
                 Lsn::from_str("D0/80208AE1").unwrap(),
-            );
+            ));
             result.unwrap();
         });
     });
@@ -183,7 +183,7 @@ fn bench_from_real_project(c: &mut Criterion) {
     group.bench_function("uniform_queries", |b| {
         b.iter(|| {
             for q in queries.clone().into_iter() {
-                layer_map.search(q.0, q.1);
+                black_box(layer_map.search(q.0, q.1));
             }
         });
     });
@@ -232,7 +232,7 @@ fn bench_sequential(c: &mut Criterion) {
     group.bench_function("uniform_queries", |b| {
         b.iter(|| {
             for q in queries.clone().into_iter() {
-                layer_map.search(q.0, q.1);
+                black_box(layer_map.search(q.0, q.1));
             }
         });
     });
diff --git a/pageserver/src/basebackup.rs b/pageserver/src/basebackup.rs
index 41fa0a67bb..c666fc785c 100644
--- a/pageserver/src/basebackup.rs
+++ b/pageserver/src/basebackup.rs
@@ -463,9 +463,13 @@ where
         let wal_file_path = format!("pg_wal/{}", wal_file_name);
         let header = new_tar_header(&wal_file_path, WAL_SEGMENT_SIZE as u64)?;
 
-        let wal_seg =
-            postgres_ffi::generate_wal_segment(segno, system_identifier, self.timeline.pg_version)
-                .map_err(|e| anyhow!(e).context("Failed generating wal segment"))?;
+        let wal_seg = postgres_ffi::generate_wal_segment(
+            segno,
+            system_identifier,
+            self.timeline.pg_version,
+            self.lsn,
+        )
+        .map_err(|e| anyhow!(e).context("Failed generating wal segment"))?;
         ensure!(wal_seg.len() == WAL_SEGMENT_SIZE);
         self.ar.append(&header, &wal_seg[..]).await?;
         Ok(())
diff --git a/pageserver/src/bin/pageserver.rs b/pageserver/src/bin/pageserver.rs
index ed23a18ee0..d843b01ed7 100644
--- a/pageserver/src/bin/pageserver.rs
+++ b/pageserver/src/bin/pageserver.rs
@@ -25,6 +25,7 @@ use pageserver::{
     virtual_file,
 };
 use postgres_backend::AuthType;
+use utils::logging::TracingErrorLayerEnablement;
 use utils::signals::ShutdownSignals;
 use utils::{
     auth::JwtAuth, logging, project_git_version, sentry_init::init_sentry, signals::Signal,
@@ -86,8 +87,19 @@ fn main() -> anyhow::Result<()> {
         }
     };
 
-    // Initialize logging, which must be initialized before the custom panic hook is installed.
-    logging::init(conf.log_format)?;
+    // Initialize logging.
+    //
+    // It must be initialized before the custom panic hook is installed below.
+    //
+    // Regarding tracing_error enablement: at this time, we only use the
+    // tracing_error crate to debug_assert that log spans contain tenant and timeline ids.
+    // See `debug_assert_current_span_has_tenant_and_timeline_id` in the timeline module
+    let tracing_error_layer_enablement = if cfg!(debug_assertions) {
+        TracingErrorLayerEnablement::EnableWithRustLogFilter
+    } else {
+        TracingErrorLayerEnablement::Disabled
+    };
+    logging::init(conf.log_format, tracing_error_layer_enablement)?;
 
     // mind the order required here: 1. logging, 2. panic_hook, 3. sentry.
     // disarming this hook on pageserver, because we never tear down tracing.
@@ -226,6 +238,7 @@ fn start_pageserver(
     );
     set_build_info_metric(GIT_VERSION);
     set_launch_timestamp_metric(launch_ts);
+    pageserver::preinitialize_metrics();
 
     // If any failpoints were set from FAILPOINTS environment variable,
     // print them to the log for debugging purposes
diff --git a/pageserver/src/config.rs b/pageserver/src/config.rs
index 19f0f22815..9e341230cf 100644
--- a/pageserver/src/config.rs
+++ b/pageserver/src/config.rs
@@ -6,6 +6,7 @@
 
 use anyhow::{anyhow, bail, ensure, Context, Result};
 use remote_storage::{RemotePath, RemoteStorageConfig};
+use serde::de::IntoDeserializer;
 use std::env;
 use storage_broker::Uri;
 use utils::crashsafe::path_with_suffix_extension;
@@ -62,7 +63,6 @@ pub mod defaults {
     pub const DEFAULT_CACHED_METRIC_COLLECTION_INTERVAL: &str = "1 hour";
     pub const DEFAULT_METRIC_COLLECTION_ENDPOINT: Option<reqwest::Url> = None;
     pub const DEFAULT_SYNTHETIC_SIZE_CALCULATION_INTERVAL: &str = "10 min";
-    pub const DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD: &str = "24 hour";
 
     ///
     /// Default built-in configuration file.
@@ -91,7 +91,6 @@ pub mod defaults {
 #cached_metric_collection_interval = '{DEFAULT_CACHED_METRIC_COLLECTION_INTERVAL}'
 #synthetic_size_calculation_interval = '{DEFAULT_SYNTHETIC_SIZE_CALCULATION_INTERVAL}'
 
-#evictions_low_residence_duration_metric_threshold = '{DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD}'
 
 #disk_usage_based_eviction = {{ max_usage_pct = .., min_avail_bytes = .., period = "10s"}}
 
@@ -108,6 +107,7 @@ pub mod defaults {
 #pitr_interval = '{DEFAULT_PITR_INTERVAL}'
 
 #min_resident_size_override = .. # in bytes
+#evictions_low_residence_duration_metric_threshold = '{DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD}'
 
 # [remote_storage]
 
@@ -182,9 +182,6 @@ pub struct PageServerConf {
     pub metric_collection_endpoint: Option<Url>,
     pub synthetic_size_calculation_interval: Duration,
 
-    // See the corresponding metric's help string.
-    pub evictions_low_residence_duration_metric_threshold: Duration,
-
     pub disk_usage_based_eviction: Option<DiskUsageEvictionTaskConfig>,
 
     pub test_remote_failures: u64,
@@ -257,8 +254,6 @@ struct PageServerConfigBuilder {
     metric_collection_endpoint: BuilderValue<Option<Url>>,
     synthetic_size_calculation_interval: BuilderValue<Duration>,
 
-    evictions_low_residence_duration_metric_threshold: BuilderValue<Duration>,
-
     disk_usage_based_eviction: BuilderValue<Option<DiskUsageEvictionTaskConfig>>,
 
     test_remote_failures: BuilderValue<u64>,
@@ -316,11 +311,6 @@ impl Default for PageServerConfigBuilder {
             .expect("cannot parse default synthetic size calculation interval")),
             metric_collection_endpoint: Set(DEFAULT_METRIC_COLLECTION_ENDPOINT),
 
-            evictions_low_residence_duration_metric_threshold: Set(humantime::parse_duration(
-                DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD,
-            )
-            .expect("cannot parse DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD")),
-
             disk_usage_based_eviction: Set(None),
 
             test_remote_failures: Set(0),
@@ -438,10 +428,6 @@ impl PageServerConfigBuilder {
         self.test_remote_failures = BuilderValue::Set(fail_first);
     }
 
-    pub fn evictions_low_residence_duration_metric_threshold(&mut self, value: Duration) {
-        self.evictions_low_residence_duration_metric_threshold = BuilderValue::Set(value);
-    }
-
     pub fn disk_usage_based_eviction(&mut self, value: Option<DiskUsageEvictionTaskConfig>) {
         self.disk_usage_based_eviction = BuilderValue::Set(value);
     }
@@ -525,11 +511,6 @@ impl PageServerConfigBuilder {
             synthetic_size_calculation_interval: self
                 .synthetic_size_calculation_interval
                 .ok_or(anyhow!("missing synthetic_size_calculation_interval"))?,
-            evictions_low_residence_duration_metric_threshold: self
-                .evictions_low_residence_duration_metric_threshold
-                .ok_or(anyhow!(
-                    "missing evictions_low_residence_duration_metric_threshold"
-                ))?,
             disk_usage_based_eviction: self
                 .disk_usage_based_eviction
                 .ok_or(anyhow!("missing disk_usage_based_eviction"))?,
@@ -721,12 +702,12 @@ impl PageServerConf {
                 "synthetic_size_calculation_interval" =>
                     builder.synthetic_size_calculation_interval(parse_toml_duration(key, item)?),
                 "test_remote_failures" => builder.test_remote_failures(parse_toml_u64(key, item)?),
-                "evictions_low_residence_duration_metric_threshold" => builder.evictions_low_residence_duration_metric_threshold(parse_toml_duration(key, item)?),
                 "disk_usage_based_eviction" => {
                     tracing::info!("disk_usage_based_eviction: {:#?}", &item);
                     builder.disk_usage_based_eviction(
-                    toml_edit::de::from_item(item.clone())
-                    .context("parse disk_usage_based_eviction")?)
+                        deserialize_from_item("disk_usage_based_eviction", item)
+                            .context("parse disk_usage_based_eviction")?
+                    )
                 },
                 "ondemand_download_behavior_treat_error_as_warn" => builder.ondemand_download_behavior_treat_error_as_warn(parse_toml_bool(key, item)?),
                 _ => bail!("unrecognized pageserver option '{key}'"),
@@ -827,18 +808,25 @@ impl PageServerConf {
 
         if let Some(eviction_policy) = item.get("eviction_policy") {
             t_conf.eviction_policy = Some(
-                toml_edit::de::from_item(eviction_policy.clone())
+                deserialize_from_item("eviction_policy", eviction_policy)
                     .context("parse eviction_policy")?,
             );
         }
 
         if let Some(item) = item.get("min_resident_size_override") {
             t_conf.min_resident_size_override = Some(
-                toml_edit::de::from_item(item.clone())
+                deserialize_from_item("min_resident_size_override", item)
                     .context("parse min_resident_size_override")?,
             );
         }
 
+        if let Some(item) = item.get("evictions_low_residence_duration_metric_threshold") {
+            t_conf.evictions_low_residence_duration_metric_threshold = Some(parse_toml_duration(
+                "evictions_low_residence_duration_metric_threshold",
+                item,
+            )?);
+        }
+
         Ok(t_conf)
     }
 
@@ -877,10 +865,6 @@ impl PageServerConf {
             cached_metric_collection_interval: Duration::from_secs(60 * 60),
             metric_collection_endpoint: defaults::DEFAULT_METRIC_COLLECTION_ENDPOINT,
             synthetic_size_calculation_interval: Duration::from_secs(60),
-            evictions_low_residence_duration_metric_threshold: humantime::parse_duration(
-                defaults::DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD,
-            )
-            .unwrap(),
             disk_usage_based_eviction: None,
             test_remote_failures: 0,
             ondemand_download_behavior_treat_error_as_warn: false,
@@ -938,6 +922,18 @@ where
     })
 }
 
+fn deserialize_from_item<T>(name: &str, item: &Item) -> anyhow::Result<T>
+where
+    T: serde::de::DeserializeOwned,
+{
+    // ValueDeserializer::new is not public, so use the ValueDeserializer's documented way
+    let deserializer = match item.clone().into_value() {
+        Ok(value) => value.into_deserializer(),
+        Err(item) => anyhow::bail!("toml_edit::Item '{item}' is not a toml_edit::Value"),
+    };
+    T::deserialize(deserializer).with_context(|| format!("deserializing item for node {name}"))
+}
+
 /// Configurable semaphore permits setting.
 ///
 /// Does not allow semaphore permits to be zero, because at runtime initially zero permits and empty
@@ -1004,9 +1000,10 @@ mod tests {
 
     use remote_storage::{RemoteStorageKind, S3Config};
     use tempfile::{tempdir, TempDir};
+    use utils::serde_percent::Percent;
 
     use super::*;
-    use crate::DEFAULT_PG_VERSION;
+    use crate::{tenant::config::EvictionPolicy, DEFAULT_PG_VERSION};
 
     const ALL_BASE_VALUES_TOML: &str = r#"
 # Initial configuration file created by 'pageserver --init'
@@ -1029,8 +1026,6 @@ cached_metric_collection_interval = '22200 s'
 metric_collection_endpoint = 'http://localhost:80/metrics'
 synthetic_size_calculation_interval = '333 s'
 
-evictions_low_residence_duration_metric_threshold = '444 s'
-
 log_format = 'json'
 
 "#;
@@ -1087,9 +1082,6 @@ log_format = 'json'
                 synthetic_size_calculation_interval: humantime::parse_duration(
                     defaults::DEFAULT_SYNTHETIC_SIZE_CALCULATION_INTERVAL
                 )?,
-                evictions_low_residence_duration_metric_threshold: humantime::parse_duration(
-                    defaults::DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD
-                )?,
                 disk_usage_based_eviction: None,
                 test_remote_failures: 0,
                 ondemand_download_behavior_treat_error_as_warn: false,
@@ -1144,7 +1136,6 @@ log_format = 'json'
                 cached_metric_collection_interval: Duration::from_secs(22200),
                 metric_collection_endpoint: Some(Url::parse("http://localhost:80/metrics")?),
                 synthetic_size_calculation_interval: Duration::from_secs(333),
-                evictions_low_residence_duration_metric_threshold: Duration::from_secs(444),
                 disk_usage_based_eviction: None,
                 test_remote_failures: 0,
                 ondemand_download_behavior_treat_error_as_warn: false,
@@ -1310,6 +1301,71 @@ trace_read_requests = {trace_read_requests}"#,
         Ok(())
     }
 
+    #[test]
+    fn eviction_pageserver_config_parse() -> anyhow::Result<()> {
+        let tempdir = tempdir()?;
+        let (workdir, pg_distrib_dir) = prepare_fs(&tempdir)?;
+
+        let pageserver_conf_toml = format!(
+            r#"pg_distrib_dir = "{}"
+metric_collection_endpoint = "http://sample.url"
+metric_collection_interval = "10min"
+id = 222
+
+[disk_usage_based_eviction]
+max_usage_pct = 80
+min_avail_bytes = 0
+period = "10s"
+
+[tenant_config]
+evictions_low_residence_duration_metric_threshold = "20m"
+
+[tenant_config.eviction_policy]
+kind = "LayerAccessThreshold"
+period = "20m"
+threshold = "20m"
+"#,
+            pg_distrib_dir.display(),
+        );
+        let toml: Document = pageserver_conf_toml.parse()?;
+        let conf = PageServerConf::parse_and_validate(&toml, &workdir)?;
+
+        assert_eq!(conf.pg_distrib_dir, pg_distrib_dir);
+        assert_eq!(
+            conf.metric_collection_endpoint,
+            Some("http://sample.url".parse().unwrap())
+        );
+        assert_eq!(
+            conf.metric_collection_interval,
+            Duration::from_secs(10 * 60)
+        );
+        assert_eq!(
+            conf.default_tenant_conf
+                .evictions_low_residence_duration_metric_threshold,
+            Duration::from_secs(20 * 60)
+        );
+        assert_eq!(conf.id, NodeId(222));
+        assert_eq!(
+            conf.disk_usage_based_eviction,
+            Some(DiskUsageEvictionTaskConfig {
+                max_usage_pct: Percent::new(80).unwrap(),
+                min_avail_bytes: 0,
+                period: Duration::from_secs(10),
+                #[cfg(feature = "testing")]
+                mock_statvfs: None,
+            })
+        );
+        match &conf.default_tenant_conf.eviction_policy {
+            EvictionPolicy::NoEviction => panic!("Unexpected eviction opolicy tenant settings"),
+            EvictionPolicy::LayerAccessThreshold(eviction_thresold) => {
+                assert_eq!(eviction_thresold.period, Duration::from_secs(20 * 60));
+                assert_eq!(eviction_thresold.threshold, Duration::from_secs(20 * 60));
+            }
+        }
+
+        Ok(())
+    }
+
     fn prepare_fs(tempdir: &TempDir) -> anyhow::Result<(PathBuf, PathBuf)> {
         let tempdir_path = tempdir.path();
 
diff --git a/pageserver/src/http/openapi_spec.yml b/pageserver/src/http/openapi_spec.yml
index 478e9d228a..95f6e96a5b 100644
--- a/pageserver/src/http/openapi_spec.yml
+++ b/pageserver/src/http/openapi_spec.yml
@@ -520,6 +520,43 @@ paths:
               schema:
                 $ref: "#/components/schemas/Error"
 
+  /v1/tenant/{tenant_id}/synthetic_size:
+    parameters:
+      - name: tenant_id
+        in: path
+        required: true
+        schema:
+          type: string
+          format: hex
+    get:
+      description: |
+        Calculate tenant's synthetic size
+      responses:
+        "200":
+          description: Tenant's synthetic size
+          content:
+            application/json:
+              schema:
+                $ref: "#/components/schemas/SyntheticSizeResponse"
+        "401":
+          description: Unauthorized Error
+          content:
+            application/json:
+              schema:
+                $ref: "#/components/schemas/UnauthorizedError"
+        "403":
+          description: Forbidden Error
+          content:
+            application/json:
+              schema:
+                $ref: "#/components/schemas/ForbiddenError"
+        "500":
+          description: Generic operation error
+          content:
+            application/json:
+              schema:
+                $ref: "#/components/schemas/Error"
+
   /v1/tenant/{tenant_id}/size:
     parameters:
       - name: tenant_id
@@ -829,12 +866,9 @@ components:
       type: object
       required:
         - id
-        - state
       properties:
         id:
           type: string
-        state:
-          type: string
         current_physical_size:
           type: integer
         has_in_progress_downloads:
@@ -951,6 +985,84 @@ components:
         latest_gc_cutoff_lsn:
           type: string
           format: hex
+
+    SyntheticSizeResponse:
+      type: object
+      required:
+        - id
+        - size
+        - segment_sizes
+        - inputs
+      properties:
+        id:
+          type: string
+          format: hex
+        size:
+          type: integer
+        segment_sizes:
+          type: array
+          items:
+            $ref: "#/components/schemas/SegmentSize"
+        inputs:
+          type: object
+          properties:
+            segments:
+              type: array
+              items:
+                $ref: "#/components/schemas/SegmentData"
+            timeline_inputs:
+              type: array
+              items:
+                $ref: "#/components/schemas/TimelineInput"
+
+    SegmentSize:
+      type: object
+      required:
+        - method
+        - accum_size
+      properties:
+        method:
+          type: string
+        accum_size:
+          type: integer
+
+    SegmentData:
+      type: object
+      required:
+        - segment
+      properties:
+        segment:
+          type: object
+          required:
+            - lsn
+          properties:
+            parent:
+              type: integer
+            lsn:
+              type: integer
+            size:
+              type: integer
+            needed:
+              type: boolean
+        timeline_id:
+          type: string
+          format: hex
+        kind:
+          type: string
+
+    TimelineInput:
+      type: object
+      required:
+        - timeline_id
+      properties:
+        ancestor_id:
+          type: string
+        ancestor_lsn:
+          type: string
+        timeline_id:
+          type: string
+          format: hex
+
     Error:
       type: object
       required:
diff --git a/pageserver/src/http/routes.rs b/pageserver/src/http/routes.rs
index 2db60f557d..b1251123b2 100644
--- a/pageserver/src/http/routes.rs
+++ b/pageserver/src/http/routes.rs
@@ -465,7 +465,7 @@ async fn tenant_list_handler(request: Request<Body>) -> Result<Response<Body>, A
         .iter()
         .map(|(id, state)| TenantInfo {
             id: *id,
-            state: *state,
+            state: state.clone(),
             current_physical_size: None,
             has_in_progress_downloads: Some(state.has_in_progress_downloads()),
         })
@@ -490,7 +490,7 @@ async fn tenant_status(request: Request<Body>) -> Result<Response<Body>, ApiErro
         let state = tenant.current_state();
         Ok(TenantInfo {
             id: tenant_id,
-            state,
+            state: state.clone(),
             current_physical_size: Some(current_physical_size),
             has_in_progress_downloads: Some(state.has_in_progress_downloads()),
         })
@@ -781,6 +781,19 @@ async fn tenant_create_handler(mut request: Request<Body>) -> Result<Response<Bo
 
     tenant_conf.min_resident_size_override = request_data.min_resident_size_override;
 
+    if let Some(evictions_low_residence_duration_metric_threshold) =
+        request_data.evictions_low_residence_duration_metric_threshold
+    {
+        tenant_conf.evictions_low_residence_duration_metric_threshold = Some(
+            humantime::parse_duration(&evictions_low_residence_duration_metric_threshold)
+                .with_context(bad_duration(
+                    "evictions_low_residence_duration_metric_threshold",
+                    &evictions_low_residence_duration_metric_threshold,
+                ))
+                .map_err(ApiError::BadRequest)?,
+        );
+    }
+
     let target_tenant_id = request_data
         .new_tenant_id
         .map(TenantId::from)
@@ -914,6 +927,19 @@ async fn update_tenant_config_handler(
 
     tenant_conf.min_resident_size_override = request_data.min_resident_size_override;
 
+    if let Some(evictions_low_residence_duration_metric_threshold) =
+        request_data.evictions_low_residence_duration_metric_threshold
+    {
+        tenant_conf.evictions_low_residence_duration_metric_threshold = Some(
+            humantime::parse_duration(&evictions_low_residence_duration_metric_threshold)
+                .with_context(bad_duration(
+                    "evictions_low_residence_duration_metric_threshold",
+                    &evictions_low_residence_duration_metric_threshold,
+                ))
+                .map_err(ApiError::BadRequest)?,
+        );
+    }
+
     let state = get_state(&request);
     mgr::set_new_tenant_config(state.conf, tenant_conf, tenant_id)
         .instrument(info_span!("tenant_config", tenant = ?tenant_id))
@@ -931,7 +957,7 @@ async fn handle_tenant_break(r: Request<Body>) -> Result<Response<Body>, ApiErro
         .await
         .map_err(|_| ApiError::Conflict(String::from("no active tenant found")))?;
 
-    tenant.set_broken("broken from test");
+    tenant.set_broken("broken from test".to_owned());
 
     json_response(StatusCode::OK, ())
 }
@@ -1175,6 +1201,37 @@ async fn handler_404(_: Request<Body>) -> Result<Response<Body>, ApiError> {
     )
 }
 
+#[cfg(feature = "testing")]
+async fn post_tracing_event_handler(mut r: Request<Body>) -> Result<Response<Body>, ApiError> {
+    #[derive(Debug, serde::Deserialize)]
+    #[serde(rename_all = "lowercase")]
+    enum Level {
+        Error,
+        Warn,
+        Info,
+        Debug,
+        Trace,
+    }
+    #[derive(Debug, serde::Deserialize)]
+    struct Request {
+        level: Level,
+        message: String,
+    }
+    let body: Request = json_request(&mut r)
+        .await
+        .map_err(|_| ApiError::BadRequest(anyhow::anyhow!("invalid JSON body")))?;
+
+    match body.level {
+        Level::Error => tracing::error!(?body.message),
+        Level::Warn => tracing::warn!(?body.message),
+        Level::Info => tracing::info!(?body.message),
+        Level::Debug => tracing::debug!(?body.message),
+        Level::Trace => tracing::trace!(?body.message),
+    }
+
+    json_response(StatusCode::OK, ())
+}
+
 pub fn make_router(
     conf: &'static PageServerConf,
     launch_ts: &'static LaunchTimestamp,
@@ -1315,5 +1372,9 @@ pub fn make_router(
             testing_api!("set tenant state to broken", handle_tenant_break),
         )
         .get("/v1/panic", |r| RequestSpan(always_panic_handler).handle(r))
+        .post(
+            "/v1/tracing/event",
+            testing_api!("emit a tracing event", post_tracing_event_handler),
+        )
         .any(handler_404))
 }
diff --git a/pageserver/src/import_datadir.rs b/pageserver/src/import_datadir.rs
index 39e434a023..936de35eb9 100644
--- a/pageserver/src/import_datadir.rs
+++ b/pageserver/src/import_datadir.rs
@@ -114,7 +114,7 @@ async fn import_rel(
     path: &Path,
     spcoid: Oid,
     dboid: Oid,
-    reader: &mut (impl AsyncRead + Send + Sync + Unpin),
+    reader: &mut (impl AsyncRead + Unpin),
     len: usize,
     ctx: &RequestContext,
 ) -> anyhow::Result<()> {
@@ -200,7 +200,7 @@ async fn import_slru(
     modification: &mut DatadirModification<'_>,
     slru: SlruKind,
     path: &Path,
-    reader: &mut (impl AsyncRead + Send + Sync + Unpin),
+    reader: &mut (impl AsyncRead + Unpin),
     len: usize,
     ctx: &RequestContext,
 ) -> anyhow::Result<()> {
@@ -612,8 +612,8 @@ async fn import_file(
     Ok(None)
 }
 
-async fn read_all_bytes(reader: &mut (impl AsyncRead + Send + Sync + Unpin)) -> Result<Bytes> {
+async fn read_all_bytes(reader: &mut (impl AsyncRead + Unpin)) -> Result<Bytes> {
     let mut buf: Vec<u8> = vec![];
     reader.read_to_end(&mut buf).await?;
-    Ok(Bytes::copy_from_slice(&buf[..]))
+    Ok(Bytes::from(buf))
 }
diff --git a/pageserver/src/lib.rs b/pageserver/src/lib.rs
index 278658eba3..04863886cb 100644
--- a/pageserver/src/lib.rs
+++ b/pageserver/src/lib.rs
@@ -44,6 +44,8 @@ pub const DELTA_FILE_MAGIC: u16 = 0x5A61;
 
 static ZERO_PAGE: bytes::Bytes = bytes::Bytes::from_static(&[0u8; 8192]);
 
+pub use crate::metrics::preinitialize_metrics;
+
 pub async fn shutdown_pageserver(exit_code: i32) {
     // Shut down the libpq endpoint task. This prevents new connections from
     // being accepted.
diff --git a/pageserver/src/metrics.rs b/pageserver/src/metrics.rs
index 1f31e5a8fb..deb20f21f8 100644
--- a/pageserver/src/metrics.rs
+++ b/pageserver/src/metrics.rs
@@ -1,12 +1,13 @@
 use metrics::core::{AtomicU64, GenericCounter};
 use metrics::{
     register_counter_vec, register_histogram, register_histogram_vec, register_int_counter,
-    register_int_counter_vec, register_int_gauge, register_int_gauge_vec, register_uint_gauge_vec,
-    Counter, CounterVec, Histogram, HistogramVec, IntCounter, IntCounterVec, IntGauge, IntGaugeVec,
-    UIntGauge, UIntGaugeVec,
+    register_int_counter_vec, register_int_gauge_vec, register_uint_gauge_vec, Counter, CounterVec,
+    Histogram, HistogramVec, IntCounter, IntCounterVec, IntGauge, IntGaugeVec, UIntGauge,
+    UIntGaugeVec,
 };
 use once_cell::sync::Lazy;
-use pageserver_api::models::state;
+use pageserver_api::models::TenantState;
+use strum::VariantNames;
 use utils::id::{TenantId, TimelineId};
 
 /// Prometheus histogram buckets (in seconds) for operations in the critical
@@ -147,15 +148,6 @@ static CURRENT_LOGICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
     .expect("failed to define current logical size metric")
 });
 
-// Metrics collected on tenant states.
-const TENANT_STATE_OPTIONS: &[&str] = &[
-    state::LOADING,
-    state::ATTACHING,
-    state::ACTIVE,
-    state::STOPPING,
-    state::BROKEN,
-];
-
 pub static TENANT_STATE_METRIC: Lazy<UIntGaugeVec> = Lazy::new(|| {
     register_uint_gauge_vec!(
         "pageserver_tenant_states_count",
@@ -213,6 +205,15 @@ static EVICTIONS_WITH_LOW_RESIDENCE_DURATION: Lazy<IntCounterVec> = Lazy::new(||
     .expect("failed to define a metric")
 });
 
+pub static UNEXPECTED_ONDEMAND_DOWNLOADS: Lazy<IntCounter> = Lazy::new(|| {
+    register_int_counter!(
+        "pageserver_unexpected_ondemand_downloads_count",
+        "Number of unexpected on-demand downloads. \
+         We log more context for each increment, so, forgo any labels in this metric.",
+    )
+    .expect("failed to define a metric")
+});
+
 /// Each [`Timeline`]'s  [`EVICTIONS_WITH_LOW_RESIDENCE_DURATION`] metric.
 #[derive(Debug)]
 pub struct EvictionsWithLowResidenceDuration {
@@ -265,6 +266,22 @@ impl EvictionsWithLowResidenceDuration {
         }
     }
 
+    pub fn change_threshold(
+        &mut self,
+        tenant_id: &str,
+        timeline_id: &str,
+        new_threshold: Duration,
+    ) {
+        if new_threshold == self.threshold {
+            return;
+        }
+        let mut with_new =
+            EvictionsWithLowResidenceDurationBuilder::new(self.data_source, new_threshold)
+                .build(tenant_id, timeline_id);
+        std::mem::swap(self, &mut with_new);
+        with_new.remove(tenant_id, timeline_id);
+    }
+
     // This could be a `Drop` impl, but, we need the `tenant_id` and `timeline_id`.
     fn remove(&mut self, tenant_id: &str, timeline_id: &str) {
         let Some(_counter) = self.counter.take() else {
@@ -342,11 +359,6 @@ pub static LIVE_CONNECTIONS_COUNT: Lazy<IntGaugeVec> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
-pub static NUM_ONDISK_LAYERS: Lazy<IntGauge> = Lazy::new(|| {
-    register_int_gauge!("pageserver_ondisk_layers", "Number of layers on-disk")
-        .expect("failed to define a metric")
-});
-
 // remote storage metrics
 
 /// NB: increment _after_ recording the current value into [`REMOTE_TIMELINE_CLIENT_CALLS_STARTED_HIST`].
@@ -377,6 +389,26 @@ static REMOTE_TIMELINE_CLIENT_CALLS_STARTED_HIST: Lazy<HistogramVec> = Lazy::new
     .expect("failed to define a metric")
 });
 
+static REMOTE_TIMELINE_CLIENT_BYTES_STARTED_COUNTER: Lazy<IntCounterVec> = Lazy::new(|| {
+    register_int_counter_vec!(
+        "pageserver_remote_timeline_client_bytes_started",
+        "Incremented by the number of bytes associated with a remote timeline client operation. \
+         The increment happens when the operation is scheduled.",
+        &["tenant_id", "timeline_id", "file_kind", "op_kind"],
+    )
+    .expect("failed to define a metric")
+});
+
+static REMOTE_TIMELINE_CLIENT_BYTES_FINISHED_COUNTER: Lazy<IntCounterVec> = Lazy::new(|| {
+    register_int_counter_vec!(
+        "pageserver_remote_timeline_client_bytes_finished",
+        "Incremented by the number of bytes associated with a remote timeline client operation. \
+         The increment happens when the operation finishes (regardless of success/failure/shutdown).",
+        &["tenant_id", "timeline_id", "file_kind", "op_kind"],
+    )
+    .expect("failed to define a metric")
+});
+
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
 pub enum RemoteOpKind {
     Upload,
@@ -597,7 +629,7 @@ pub struct TimelineMetrics {
     pub num_persistent_files_created: IntCounter,
     pub persistent_bytes_written: IntCounter,
     pub evictions: IntCounter,
-    pub evictions_with_low_residence_duration: EvictionsWithLowResidenceDuration,
+    pub evictions_with_low_residence_duration: std::sync::RwLock<EvictionsWithLowResidenceDuration>,
 }
 
 impl TimelineMetrics {
@@ -664,7 +696,9 @@ impl TimelineMetrics {
             num_persistent_files_created,
             persistent_bytes_written,
             evictions,
-            evictions_with_low_residence_duration,
+            evictions_with_low_residence_duration: std::sync::RwLock::new(
+                evictions_with_low_residence_duration,
+            ),
         }
     }
 }
@@ -683,6 +717,8 @@ impl Drop for TimelineMetrics {
         let _ = PERSISTENT_BYTES_WRITTEN.remove_label_values(&[tenant_id, timeline_id]);
         let _ = EVICTIONS.remove_label_values(&[tenant_id, timeline_id]);
         self.evictions_with_low_residence_duration
+            .write()
+            .unwrap()
             .remove(tenant_id, timeline_id);
         for op in STORAGE_TIME_OPERATIONS {
             let _ =
@@ -707,7 +743,7 @@ impl Drop for TimelineMetrics {
 pub fn remove_tenant_metrics(tenant_id: &TenantId) {
     let tid = tenant_id.to_string();
     let _ = TENANT_SYNTHETIC_SIZE_METRIC.remove_label_values(&[&tid]);
-    for state in TENANT_STATE_OPTIONS {
+    for state in TenantState::VARIANTS {
         let _ = TENANT_STATE_METRIC.remove_label_values(&[&tid, state]);
     }
 }
@@ -727,6 +763,8 @@ pub struct RemoteTimelineClientMetrics {
     remote_operation_time: Mutex<HashMap<(&'static str, &'static str, &'static str), Histogram>>,
     calls_unfinished_gauge: Mutex<HashMap<(&'static str, &'static str), IntGauge>>,
     calls_started_hist: Mutex<HashMap<(&'static str, &'static str), Histogram>>,
+    bytes_started_counter: Mutex<HashMap<(&'static str, &'static str), IntCounter>>,
+    bytes_finished_counter: Mutex<HashMap<(&'static str, &'static str), IntCounter>>,
 }
 
 impl RemoteTimelineClientMetrics {
@@ -737,6 +775,8 @@ impl RemoteTimelineClientMetrics {
             remote_operation_time: Mutex::new(HashMap::default()),
             calls_unfinished_gauge: Mutex::new(HashMap::default()),
             calls_started_hist: Mutex::new(HashMap::default()),
+            bytes_started_counter: Mutex::new(HashMap::default()),
+            bytes_finished_counter: Mutex::new(HashMap::default()),
             remote_physical_size_gauge: Mutex::new(None),
         }
     }
@@ -775,6 +815,7 @@ impl RemoteTimelineClientMetrics {
         });
         metric.clone()
     }
+
     fn calls_unfinished_gauge(
         &self,
         file_kind: &RemoteOpFileKind,
@@ -816,32 +857,125 @@ impl RemoteTimelineClientMetrics {
         });
         metric.clone()
     }
+
+    fn bytes_started_counter(
+        &self,
+        file_kind: &RemoteOpFileKind,
+        op_kind: &RemoteOpKind,
+    ) -> IntCounter {
+        // XXX would be nice to have an upgradable RwLock
+        let mut guard = self.bytes_started_counter.lock().unwrap();
+        let key = (file_kind.as_str(), op_kind.as_str());
+        let metric = guard.entry(key).or_insert_with(move || {
+            REMOTE_TIMELINE_CLIENT_BYTES_STARTED_COUNTER
+                .get_metric_with_label_values(&[
+                    &self.tenant_id.to_string(),
+                    &self.timeline_id.to_string(),
+                    key.0,
+                    key.1,
+                ])
+                .unwrap()
+        });
+        metric.clone()
+    }
+
+    fn bytes_finished_counter(
+        &self,
+        file_kind: &RemoteOpFileKind,
+        op_kind: &RemoteOpKind,
+    ) -> IntCounter {
+        // XXX would be nice to have an upgradable RwLock
+        let mut guard = self.bytes_finished_counter.lock().unwrap();
+        let key = (file_kind.as_str(), op_kind.as_str());
+        let metric = guard.entry(key).or_insert_with(move || {
+            REMOTE_TIMELINE_CLIENT_BYTES_FINISHED_COUNTER
+                .get_metric_with_label_values(&[
+                    &self.tenant_id.to_string(),
+                    &self.timeline_id.to_string(),
+                    key.0,
+                    key.1,
+                ])
+                .unwrap()
+        });
+        metric.clone()
+    }
+}
+
+#[cfg(test)]
+impl RemoteTimelineClientMetrics {
+    pub fn get_bytes_started_counter_value(
+        &self,
+        file_kind: &RemoteOpFileKind,
+        op_kind: &RemoteOpKind,
+    ) -> Option<u64> {
+        let guard = self.bytes_started_counter.lock().unwrap();
+        let key = (file_kind.as_str(), op_kind.as_str());
+        guard.get(&key).map(|counter| counter.get())
+    }
+
+    pub fn get_bytes_finished_counter_value(
+        &self,
+        file_kind: &RemoteOpFileKind,
+        op_kind: &RemoteOpKind,
+    ) -> Option<u64> {
+        let guard = self.bytes_finished_counter.lock().unwrap();
+        let key = (file_kind.as_str(), op_kind.as_str());
+        guard.get(&key).map(|counter| counter.get())
+    }
 }
 
 /// See [`RemoteTimelineClientMetrics::call_begin`].
 #[must_use]
-pub(crate) struct RemoteTimelineClientCallMetricGuard(Option<IntGauge>);
+pub(crate) struct RemoteTimelineClientCallMetricGuard {
+    /// Decremented on drop.
+    calls_unfinished_metric: Option<IntGauge>,
+    /// If Some(), this references the bytes_finished metric, and we increment it by the given `u64` on drop.
+    bytes_finished: Option<(IntCounter, u64)>,
+}
 
 impl RemoteTimelineClientCallMetricGuard {
-    /// Consume this guard object without decrementing the metric.
-    /// The caller vouches to do this manually, so that the prior increment of the gauge will cancel out.
+    /// Consume this guard object without performing the metric updates it would do on `drop()`.
+    /// The caller vouches to do the metric updates manually.
     pub fn will_decrement_manually(mut self) {
-        self.0 = None; // prevent drop() from decrementing
+        let RemoteTimelineClientCallMetricGuard {
+            calls_unfinished_metric,
+            bytes_finished,
+        } = &mut self;
+        calls_unfinished_metric.take();
+        bytes_finished.take();
     }
 }
 
 impl Drop for RemoteTimelineClientCallMetricGuard {
     fn drop(&mut self) {
-        if let RemoteTimelineClientCallMetricGuard(Some(guard)) = self {
+        let RemoteTimelineClientCallMetricGuard {
+            calls_unfinished_metric,
+            bytes_finished,
+        } = self;
+        if let Some(guard) = calls_unfinished_metric.take() {
             guard.dec();
         }
+        if let Some((bytes_finished_metric, value)) = bytes_finished {
+            bytes_finished_metric.inc_by(*value);
+        }
     }
 }
 
+/// The enum variants communicate to the [`RemoteTimelineClientMetrics`] whether to
+/// track the byte size of this call in applicable metric(s).
+pub(crate) enum RemoteTimelineClientMetricsCallTrackSize {
+    /// Do not account for this call's byte size in any metrics.
+    /// The `reason` field is there to make the call sites self-documenting
+    /// about why they don't need the metric.
+    DontTrackSize { reason: &'static str },
+    /// Track the byte size of the call in applicable metric(s).
+    Bytes(u64),
+}
+
 impl RemoteTimelineClientMetrics {
-    /// Increment the metrics that track ongoing calls to the remote timeline client instance.
+    /// Update the metrics that change when a call to the remote timeline client instance starts.
     ///
-    /// Drop the returned guard object once the operation is finished to decrement the values.
+    /// Drop the returned guard object once the operation is finished to updates corresponding metrics that track completions.
     /// Or, use [`RemoteTimelineClientCallMetricGuard::will_decrement_manually`] and [`call_end`] if that
     /// is more suitable.
     /// Never do both.
@@ -849,24 +983,51 @@ impl RemoteTimelineClientMetrics {
         &self,
         file_kind: &RemoteOpFileKind,
         op_kind: &RemoteOpKind,
+        size: RemoteTimelineClientMetricsCallTrackSize,
     ) -> RemoteTimelineClientCallMetricGuard {
-        let unfinished_metric = self.calls_unfinished_gauge(file_kind, op_kind);
+        let calls_unfinished_metric = self.calls_unfinished_gauge(file_kind, op_kind);
         self.calls_started_hist(file_kind, op_kind)
-            .observe(unfinished_metric.get() as f64);
-        unfinished_metric.inc();
-        RemoteTimelineClientCallMetricGuard(Some(unfinished_metric))
+            .observe(calls_unfinished_metric.get() as f64);
+        calls_unfinished_metric.inc(); // NB: inc after the histogram, see comment on underlying metric
+
+        let bytes_finished = match size {
+            RemoteTimelineClientMetricsCallTrackSize::DontTrackSize { reason: _reason } => {
+                // nothing to do
+                None
+            }
+            RemoteTimelineClientMetricsCallTrackSize::Bytes(size) => {
+                self.bytes_started_counter(file_kind, op_kind).inc_by(size);
+                let finished_counter = self.bytes_finished_counter(file_kind, op_kind);
+                Some((finished_counter, size))
+            }
+        };
+        RemoteTimelineClientCallMetricGuard {
+            calls_unfinished_metric: Some(calls_unfinished_metric),
+            bytes_finished,
+        }
     }
 
-    /// Manually decrement the metric instead of using the guard object.
+    /// Manually udpate the metrics that track completions, instead of using the guard object.
     /// Using the guard object is generally preferable.
     /// See [`call_begin`] for more context.
-    pub(crate) fn call_end(&self, file_kind: &RemoteOpFileKind, op_kind: &RemoteOpKind) {
-        let unfinished_metric = self.calls_unfinished_gauge(file_kind, op_kind);
+    pub(crate) fn call_end(
+        &self,
+        file_kind: &RemoteOpFileKind,
+        op_kind: &RemoteOpKind,
+        size: RemoteTimelineClientMetricsCallTrackSize,
+    ) {
+        let calls_unfinished_metric = self.calls_unfinished_gauge(file_kind, op_kind);
         debug_assert!(
-            unfinished_metric.get() > 0,
+            calls_unfinished_metric.get() > 0,
             "begin and end should cancel out"
         );
-        unfinished_metric.dec();
+        calls_unfinished_metric.dec();
+        match size {
+            RemoteTimelineClientMetricsCallTrackSize::DontTrackSize { reason: _reason } => {}
+            RemoteTimelineClientMetricsCallTrackSize::Bytes(size) => {
+                self.bytes_finished_counter(file_kind, op_kind).inc_by(size);
+            }
+        }
     }
 }
 
@@ -879,6 +1040,8 @@ impl Drop for RemoteTimelineClientMetrics {
             remote_operation_time,
             calls_unfinished_gauge,
             calls_started_hist,
+            bytes_started_counter,
+            bytes_finished_counter,
         } = self;
         for ((a, b, c), _) in remote_operation_time.get_mut().unwrap().drain() {
             let _ = REMOTE_OPERATION_TIME.remove_label_values(&[tenant_id, timeline_id, a, b, c]);
@@ -899,6 +1062,22 @@ impl Drop for RemoteTimelineClientMetrics {
                 b,
             ]);
         }
+        for ((a, b), _) in bytes_started_counter.get_mut().unwrap().drain() {
+            let _ = REMOTE_TIMELINE_CLIENT_BYTES_STARTED_COUNTER.remove_label_values(&[
+                tenant_id,
+                timeline_id,
+                a,
+                b,
+            ]);
+        }
+        for ((a, b), _) in bytes_finished_counter.get_mut().unwrap().drain() {
+            let _ = REMOTE_TIMELINE_CLIENT_BYTES_FINISHED_COUNTER.remove_label_values(&[
+                tenant_id,
+                timeline_id,
+                a,
+                b,
+            ]);
+        }
         {
             let _ = remote_physical_size_gauge; // use to avoid 'unused' warning in desctructuring above
             let _ = REMOTE_PHYSICAL_SIZE.remove_label_values(&[tenant_id, timeline_id]);
@@ -962,3 +1141,10 @@ impl<F: Future<Output = Result<O, E>>, O, E> Future for MeasuredRemoteOp<F> {
         poll_result
     }
 }
+
+pub fn preinitialize_metrics() {
+    // We want to alert on this metric increasing.
+    // Initialize it eagerly, so that our alert rule can distinguish absence of the metric from metric value 0.
+    assert_eq!(UNEXPECTED_ONDEMAND_DOWNLOADS.get(), 0);
+    UNEXPECTED_ONDEMAND_DOWNLOADS.reset();
+}
diff --git a/pageserver/src/page_service.rs b/pageserver/src/page_service.rs
index c0e4a2a9cf..8b0795db3c 100644
--- a/pageserver/src/page_service.rs
+++ b/pageserver/src/page_service.rs
@@ -20,7 +20,6 @@ use pageserver_api::models::{
     PagestreamFeMessage, PagestreamGetPageRequest, PagestreamGetPageResponse,
     PagestreamNblocksRequest, PagestreamNblocksResponse,
 };
-use postgres_backend::PostgresBackendTCP;
 use postgres_backend::{self, is_expected_io_error, AuthType, PostgresBackend, QueryError};
 use pq_proto::framed::ConnectionError;
 use pq_proto::FeStartupPacket;
@@ -32,6 +31,7 @@ use std::str;
 use std::str::FromStr;
 use std::sync::Arc;
 use std::time::Duration;
+use tokio::io::{AsyncRead, AsyncWrite};
 use tokio_util::io::StreamReader;
 use tracing::*;
 use utils::id::ConnectionId;
@@ -57,7 +57,10 @@ use crate::trace::Tracer;
 use postgres_ffi::pg_constants::DEFAULTTABLESPACE_OID;
 use postgres_ffi::BLCKSZ;
 
-fn copyin_stream(pgb: &mut PostgresBackendTCP) -> impl Stream<Item = io::Result<Bytes>> + '_ {
+fn copyin_stream<IO>(pgb: &mut PostgresBackend<IO>) -> impl Stream<Item = io::Result<Bytes>> + '_
+where
+    IO: AsyncRead + AsyncWrite + Unpin,
+{
     async_stream::try_stream! {
         loop {
             let msg = tokio::select! {
@@ -65,8 +68,8 @@ fn copyin_stream(pgb: &mut PostgresBackendTCP) -> impl Stream<Item = io::Result<
 
                 _ = task_mgr::shutdown_watcher() => {
                     // We were requested to shut down.
-                    let msg = format!("pageserver is shutting down");
-                    let _ = pgb.write_message_noflush(&BeMessage::ErrorResponse(&msg, None));
+                    let msg = "pageserver is shutting down";
+                    let _ = pgb.write_message_noflush(&BeMessage::ErrorResponse(msg, None));
                     Err(QueryError::Other(anyhow::anyhow!(msg)))
                 }
 
@@ -125,7 +128,7 @@ fn copyin_stream(pgb: &mut PostgresBackendTCP) -> impl Stream<Item = io::Result<
 ///
 /// XXX: Currently, any trailing data after the EOF marker prints a warning.
 /// Perhaps it should be a hard error?
-async fn read_tar_eof(mut reader: (impl tokio::io::AsyncRead + Unpin)) -> anyhow::Result<()> {
+async fn read_tar_eof(mut reader: (impl AsyncRead + Unpin)) -> anyhow::Result<()> {
     use tokio::io::AsyncReadExt;
     let mut buf = [0u8; 512];
 
@@ -245,12 +248,23 @@ async fn page_service_conn_main(
         .set_nodelay(true)
         .context("could not set TCP_NODELAY")?;
 
+    let peer_addr = socket.peer_addr().context("get peer address")?;
+
+    // setup read timeout of 10 minutes. the timeout is rather arbitrary for requirements:
+    // - long enough for most valid compute connections
+    // - less than infinite to stop us from "leaking" connections to long-gone computes
+    //
+    // no write timeout is used, because the kernel is assumed to error writes after some time.
+    let mut socket = tokio_io_timeout::TimeoutReader::new(socket);
+    socket.set_timeout(Some(std::time::Duration::from_secs(60 * 10)));
+    let socket = std::pin::pin!(socket);
+
     // XXX: pgbackend.run() should take the connection_ctx,
     // and create a child per-query context when it invokes process_query.
     // But it's in a shared crate, so, we store connection_ctx inside PageServerHandler
     // and create the per-query context in process_query ourselves.
     let mut conn_handler = PageServerHandler::new(conf, auth, connection_ctx);
-    let pgbackend = PostgresBackend::new(socket, auth_type, None)?;
+    let pgbackend = PostgresBackend::new_from_io(socket, peer_addr, auth_type, None)?;
 
     match pgbackend
         .run(&mut conn_handler, task_mgr::shutdown_watcher)
@@ -332,13 +346,16 @@ impl PageServerHandler {
     }
 
     #[instrument(skip(self, pgb, ctx))]
-    async fn handle_pagerequests(
+    async fn handle_pagerequests<IO>(
         &self,
-        pgb: &mut PostgresBackendTCP,
+        pgb: &mut PostgresBackend<IO>,
         tenant_id: TenantId,
         timeline_id: TimelineId,
         ctx: RequestContext,
-    ) -> anyhow::Result<()> {
+    ) -> anyhow::Result<()>
+    where
+        IO: AsyncRead + AsyncWrite + Send + Sync + Unpin,
+    {
         // NOTE: pagerequests handler exits when connection is closed,
         //       so there is no need to reset the association
         task_mgr::associate_with(Some(tenant_id), Some(timeline_id));
@@ -436,16 +453,19 @@ impl PageServerHandler {
 
     #[allow(clippy::too_many_arguments)]
     #[instrument(skip(self, pgb, ctx))]
-    async fn handle_import_basebackup(
+    async fn handle_import_basebackup<IO>(
         &self,
-        pgb: &mut PostgresBackendTCP,
+        pgb: &mut PostgresBackend<IO>,
         tenant_id: TenantId,
         timeline_id: TimelineId,
         base_lsn: Lsn,
         _end_lsn: Lsn,
         pg_version: u32,
         ctx: RequestContext,
-    ) -> Result<(), QueryError> {
+    ) -> Result<(), QueryError>
+    where
+        IO: AsyncRead + AsyncWrite + Send + Sync + Unpin,
+    {
         task_mgr::associate_with(Some(tenant_id), Some(timeline_id));
         // Create empty timeline
         info!("creating new timeline");
@@ -486,15 +506,18 @@ impl PageServerHandler {
     }
 
     #[instrument(skip(self, pgb, ctx))]
-    async fn handle_import_wal(
+    async fn handle_import_wal<IO>(
         &self,
-        pgb: &mut PostgresBackendTCP,
+        pgb: &mut PostgresBackend<IO>,
         tenant_id: TenantId,
         timeline_id: TimelineId,
         start_lsn: Lsn,
         end_lsn: Lsn,
         ctx: RequestContext,
-    ) -> Result<(), QueryError> {
+    ) -> Result<(), QueryError>
+    where
+        IO: AsyncRead + AsyncWrite + Send + Sync + Unpin,
+    {
         task_mgr::associate_with(Some(tenant_id), Some(timeline_id));
 
         let timeline = get_active_tenant_timeline(tenant_id, timeline_id, &ctx).await?;
@@ -690,16 +713,21 @@ impl PageServerHandler {
 
     #[allow(clippy::too_many_arguments)]
     #[instrument(skip(self, pgb, ctx))]
-    async fn handle_basebackup_request(
+    async fn handle_basebackup_request<IO>(
         &mut self,
-        pgb: &mut PostgresBackendTCP,
+        pgb: &mut PostgresBackend<IO>,
         tenant_id: TenantId,
         timeline_id: TimelineId,
         lsn: Option<Lsn>,
         prev_lsn: Option<Lsn>,
         full_backup: bool,
         ctx: RequestContext,
-    ) -> anyhow::Result<()> {
+    ) -> anyhow::Result<()>
+    where
+        IO: AsyncRead + AsyncWrite + Send + Sync + Unpin,
+    {
+        let started = std::time::Instant::now();
+
         // check that the timeline exists
         let timeline = get_active_tenant_timeline(tenant_id, timeline_id, &ctx).await?;
         let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
@@ -712,6 +740,8 @@ impl PageServerHandler {
                 .context("invalid basebackup lsn")?;
         }
 
+        let lsn_awaited_after = started.elapsed();
+
         // switch client to COPYOUT
         pgb.write_message_noflush(&BeMessage::CopyOutResponse)?;
         pgb.flush().await?;
@@ -732,7 +762,17 @@ impl PageServerHandler {
 
         pgb.write_message_noflush(&BeMessage::CopyDone)?;
         pgb.flush().await?;
-        info!("basebackup complete");
+
+        let basebackup_after = started
+            .elapsed()
+            .checked_sub(lsn_awaited_after)
+            .unwrap_or(Duration::ZERO);
+
+        info!(
+            lsn_await_millis = lsn_awaited_after.as_millis(),
+            basebackup_millis = basebackup_after.as_millis(),
+            "basebackup complete"
+        );
 
         Ok(())
     }
@@ -756,10 +796,13 @@ impl PageServerHandler {
 }
 
 #[async_trait::async_trait]
-impl postgres_backend::Handler<tokio::net::TcpStream> for PageServerHandler {
+impl<IO> postgres_backend::Handler<IO> for PageServerHandler
+where
+    IO: AsyncRead + AsyncWrite + Send + Sync + Unpin,
+{
     fn check_auth_jwt(
         &mut self,
-        _pgb: &mut PostgresBackendTCP,
+        _pgb: &mut PostgresBackend<IO>,
         jwt_response: &[u8],
     ) -> Result<(), QueryError> {
         // this unwrap is never triggered, because check_auth_jwt only called when auth_type is NeonJWT
@@ -787,7 +830,7 @@ impl postgres_backend::Handler<tokio::net::TcpStream> for PageServerHandler {
 
     fn startup(
         &mut self,
-        _pgb: &mut PostgresBackendTCP,
+        _pgb: &mut PostgresBackend<IO>,
         _sm: &FeStartupPacket,
     ) -> Result<(), QueryError> {
         Ok(())
@@ -795,7 +838,7 @@ impl postgres_backend::Handler<tokio::net::TcpStream> for PageServerHandler {
 
     async fn process_query(
         &mut self,
-        pgb: &mut PostgresBackendTCP,
+        pgb: &mut PostgresBackend<IO>,
         query_string: &str,
     ) -> Result<(), QueryError> {
         let ctx = self.connection_ctx.attached_child();
diff --git a/pageserver/src/tenant.rs b/pageserver/src/tenant.rs
index 7fac7d2ac0..5cfc466111 100644
--- a/pageserver/src/tenant.rs
+++ b/pageserver/src/tenant.rs
@@ -118,6 +118,10 @@ pub struct Tenant {
     // Global pageserver config parameters
     pub conf: &'static PageServerConf,
 
+    /// The value creation timestamp, used to measure activation delay, see:
+    /// <https://github.com/neondatabase/neon/issues/4025>
+    loading_started_at: Instant,
+
     state: watch::Sender<TenantState>,
 
     // Overridden tenant-specific config parameters.
@@ -177,9 +181,9 @@ impl UninitializedTimeline<'_> {
     ///
     /// The new timeline is initialized in Active state, and its background jobs are
     /// started
-    pub fn initialize(self, _ctx: &RequestContext) -> anyhow::Result<Arc<Timeline>> {
+    pub fn initialize(self, ctx: &RequestContext) -> anyhow::Result<Arc<Timeline>> {
         let mut timelines = self.owning_tenant.timelines.lock().unwrap();
-        self.initialize_with_lock(&mut timelines, true, true)
+        self.initialize_with_lock(ctx, &mut timelines, true, true)
     }
 
     /// Like `initialize`, but the caller is already holding lock on Tenant::timelines.
@@ -189,6 +193,7 @@ impl UninitializedTimeline<'_> {
     /// been initialized.
     fn initialize_with_lock(
         mut self,
+        ctx: &RequestContext,
         timelines: &mut HashMap<TimelineId, Arc<Timeline>>,
         load_layer_map: bool,
         activate: bool,
@@ -229,7 +234,9 @@ impl UninitializedTimeline<'_> {
                 new_timeline.maybe_spawn_flush_loop();
 
                 if activate {
-                    new_timeline.activate();
+                    new_timeline
+                        .activate(ctx)
+                        .context("initializing timeline activation")?;
                 }
             }
         }
@@ -469,7 +476,7 @@ impl Tenant {
         local_metadata: Option<TimelineMetadata>,
         ancestor: Option<Arc<Timeline>>,
         first_save: bool,
-        _ctx: &RequestContext,
+        ctx: &RequestContext,
     ) -> anyhow::Result<()> {
         let tenant_id = self.tenant_id;
 
@@ -504,7 +511,7 @@ impl Tenant {
             // Do not start walreceiver here. We do need loaded layer map for reconcile_with_remote
             // But we shouldnt start walreceiver before we have all the data locally, because working walreceiver
             // will ingest data which may require looking at the layers which are not yet available locally
-            match timeline.initialize_with_lock(&mut timelines_accessor, true, false) {
+            match timeline.initialize_with_lock(ctx, &mut timelines_accessor, true, false) {
                 Ok(new_timeline) => new_timeline,
                 Err(e) => {
                     error!("Failed to initialize timeline {tenant_id}/{timeline_id}: {e:?}");
@@ -616,7 +623,7 @@ impl Tenant {
                 match tenant_clone.attach(ctx).await {
                     Ok(_) => {}
                     Err(e) => {
-                        tenant_clone.set_broken(&e.to_string());
+                        tenant_clone.set_broken(e.to_string());
                         error!("error attaching tenant: {:?}", e);
                     }
                 }
@@ -629,7 +636,7 @@ impl Tenant {
     ///
     /// Background task that downloads all data for a tenant and brings it to Active state.
     ///
-    #[instrument(skip(self, ctx), fields(tenant_id=%self.tenant_id))]
+    #[instrument(skip_all, fields(tenant_id=%self.tenant_id))]
     async fn attach(self: &Arc<Tenant>, ctx: RequestContext) -> anyhow::Result<()> {
         // Create directory with marker file to indicate attaching state.
         // The load_local_tenants() function in tenant::mgr relies on the marker file
@@ -750,7 +757,7 @@ impl Tenant {
 
         // Start background operations and open the tenant for business.
         // The loops will shut themselves down when they notice that the tenant is inactive.
-        self.activate()?;
+        self.activate(&ctx)?;
 
         info!("Done");
 
@@ -824,7 +831,10 @@ impl Tenant {
     pub fn create_broken_tenant(conf: &'static PageServerConf, tenant_id: TenantId) -> Arc<Tenant> {
         let wal_redo_manager = Arc::new(PostgresRedoManager::new(conf, tenant_id));
         Arc::new(Tenant::new(
-            TenantState::Broken,
+            TenantState::Broken {
+                reason: "create_broken_tenant".into(),
+                backtrace: String::new(),
+            },
             conf,
             TenantConfOpt::default(),
             wal_redo_manager,
@@ -885,7 +895,7 @@ impl Tenant {
                 match tenant_clone.load(&ctx).await {
                     Ok(()) => {}
                     Err(err) => {
-                        tenant_clone.set_broken(&err.to_string());
+                        tenant_clone.set_broken(err.to_string());
                         error!("could not load tenant {tenant_id}: {err:?}");
                     }
                 }
@@ -1022,7 +1032,7 @@ impl Tenant {
 
         // Start background operations and open the tenant for business.
         // The loops will shut themselves down when they notice that the tenant is inactive.
-        self.activate()?;
+        self.activate(ctx)?;
 
         info!("Done");
 
@@ -1358,12 +1368,7 @@ impl Tenant {
 
         // Stop the walreceiver first.
         debug!("waiting for wal receiver to shutdown");
-        task_mgr::shutdown_tasks(
-            Some(TaskKind::WalReceiverManager),
-            Some(self.tenant_id),
-            Some(timeline_id),
-        )
-        .await;
+        timeline.walreceiver.stop().await;
         debug!("wal receiver shutdown confirmed");
 
         info!("waiting for timeline tasks to shutdown");
@@ -1442,7 +1447,7 @@ impl Tenant {
     }
 
     pub fn current_state(&self) -> TenantState {
-        *self.state.borrow()
+        self.state.borrow().clone()
     }
 
     pub fn is_active(&self) -> bool {
@@ -1450,18 +1455,18 @@ impl Tenant {
     }
 
     /// Changes tenant status to active, unless shutdown was already requested.
-    fn activate(&self) -> anyhow::Result<()> {
+    fn activate(&self, ctx: &RequestContext) -> anyhow::Result<()> {
         let mut result = Ok(());
         self.state.send_modify(|current_state| {
-            match *current_state {
+            match &*current_state {
                 TenantState::Active => {
                     // activate() was called on an already Active tenant. Shouldn't happen.
                     result = Err(anyhow::anyhow!("Tenant is already active"));
                 }
-                TenantState::Broken => {
+                TenantState::Broken { reason, .. } => {
                     // This shouldn't happen either
                     result = Err(anyhow::anyhow!(
-                        "Could not activate tenant because it is in broken state"
+                        "Could not activate tenant because it is in broken state due to: {reason}",
                     ));
                 }
                 TenantState::Stopping => {
@@ -1472,7 +1477,7 @@ impl Tenant {
                 TenantState::Loading | TenantState::Attaching => {
                     *current_state = TenantState::Active;
 
-                    info!("Activating tenant {}", self.tenant_id);
+                    debug!(tenant_id = %self.tenant_id, "Activating tenant");
 
                     let timelines_accessor = self.timelines.lock().unwrap();
                     let not_broken_timelines = timelines_accessor
@@ -1483,9 +1488,47 @@ impl Tenant {
                     // down when they notice that the tenant is inactive.
                     tasks::start_background_loops(self.tenant_id);
 
+                    let mut activated_timelines = 0;
+                    let mut timelines_broken_during_activation = 0;
+
                     for timeline in not_broken_timelines {
-                        timeline.activate();
+                        match timeline
+                            .activate(ctx)
+                            .context("timeline activation for activating tenant")
+                        {
+                            Ok(()) => {
+                                activated_timelines += 1;
+                            }
+                            Err(e) => {
+                                error!(
+                                    "Failed to activate timeline {}: {:#}",
+                                    timeline.timeline_id, e
+                                );
+                                timeline.set_state(TimelineState::Broken);
+                                *current_state = TenantState::broken_from_reason(format!(
+                                    "failed to activate timeline {}: {}",
+                                    timeline.timeline_id, e
+                                ));
+
+                                timelines_broken_during_activation += 1;
+                            }
+                        }
                     }
+
+                    let elapsed = self.loading_started_at.elapsed();
+                    let total_timelines = timelines_accessor.len();
+
+                    // log a lot of stuff, because some tenants sometimes suffer from user-visible
+                    // times to activate. see https://github.com/neondatabase/neon/issues/4025
+                    info!(
+                        since_creation_millis = elapsed.as_millis(),
+                        tenant_id = %self.tenant_id,
+                        activated_timelines,
+                        timelines_broken_during_activation,
+                        total_timelines,
+                        post_state = <&'static str>::from(&*current_state),
+                        "activation attempt finished"
+                    );
                 }
             }
         });
@@ -1495,7 +1538,7 @@ impl Tenant {
     /// Change tenant status to Stopping, to mark that it is being shut down
     pub fn set_stopping(&self) {
         self.state.send_modify(|current_state| {
-            match *current_state {
+            match current_state {
                 TenantState::Active | TenantState::Loading | TenantState::Attaching => {
                     *current_state = TenantState::Stopping;
 
@@ -1511,8 +1554,8 @@ impl Tenant {
                         timeline.set_state(TimelineState::Stopping);
                     }
                 }
-                TenantState::Broken => {
-                    info!("Cannot set tenant to Stopping state, it is already in Broken state");
+                TenantState::Broken { reason, .. } => {
+                    info!("Cannot set tenant to Stopping state, it is in Broken state due to: {reason}");
                 }
                 TenantState::Stopping => {
                     // The tenant was detached, or system shutdown was requested, while we were
@@ -1523,7 +1566,7 @@ impl Tenant {
         });
     }
 
-    pub fn set_broken(&self, reason: &str) {
+    pub fn set_broken(&self, reason: String) {
         self.state.send_modify(|current_state| {
             match *current_state {
                 TenantState::Active => {
@@ -1531,24 +1574,24 @@ impl Tenant {
                     // while loading or attaching a tenant. A tenant that has already been
                     // activated should never be marked as broken. We cope with it the best
                     // we can, but it shouldn't happen.
-                    *current_state = TenantState::Broken;
                     warn!("Changing Active tenant to Broken state, reason: {}", reason);
+                    *current_state = TenantState::broken_from_reason(reason);
                 }
-                TenantState::Broken => {
+                TenantState::Broken { .. } => {
                     // This shouldn't happen either
                     warn!("Tenant is already in Broken state");
                 }
                 TenantState::Stopping => {
                     // This shouldn't happen either
-                    *current_state = TenantState::Broken;
                     warn!(
                         "Marking Stopping tenant as Broken state, reason: {}",
                         reason
                     );
+                    *current_state = TenantState::broken_from_reason(reason);
                 }
                 TenantState::Loading | TenantState::Attaching => {
                     info!("Setting tenant as Broken state, reason: {}", reason);
-                    *current_state = TenantState::Broken;
+                    *current_state = TenantState::broken_from_reason(reason);
                 }
             }
         });
@@ -1561,7 +1604,7 @@ impl Tenant {
     pub async fn wait_to_become_active(&self) -> anyhow::Result<()> {
         let mut receiver = self.state.subscribe();
         loop {
-            let current_state = *receiver.borrow_and_update();
+            let current_state = receiver.borrow_and_update().clone();
             match current_state {
                 TenantState::Loading | TenantState::Attaching => {
                     // in these states, there's a chance that we can reach ::Active
@@ -1570,12 +1613,12 @@ impl Tenant {
                 TenantState::Active { .. } => {
                     return Ok(());
                 }
-                TenantState::Broken | TenantState::Stopping => {
+                TenantState::Broken { .. } | TenantState::Stopping => {
                     // There's no chance the tenant can transition back into ::Active
                     anyhow::bail!(
                         "Tenant {} will not become active. Current state: {:?}",
                         self.tenant_id,
-                        current_state,
+                        &current_state,
                     );
                 }
             }
@@ -1715,6 +1758,13 @@ impl Tenant {
 
     pub fn set_new_tenant_config(&self, new_tenant_conf: TenantConfOpt) {
         *self.tenant_conf.write().unwrap() = new_tenant_conf;
+        // Don't hold self.timelines.lock() during the notifies.
+        // There's no risk of deadlock right now, but there could be if we consolidate
+        // mutexes in struct Timeline in the future.
+        let timelines = self.list_timelines();
+        for timeline in timelines {
+            timeline.tenant_conf_updated();
+        }
     }
 
     fn create_timeline_data(
@@ -1756,21 +1806,23 @@ impl Tenant {
         let (state, mut rx) = watch::channel(state);
 
         tokio::spawn(async move {
-            let current_state = *rx.borrow_and_update();
+            let mut current_state: &'static str = From::from(&*rx.borrow_and_update());
             let tid = tenant_id.to_string();
             TENANT_STATE_METRIC
-                .with_label_values(&[&tid, current_state.as_str()])
+                .with_label_values(&[&tid, current_state])
                 .inc();
             loop {
                 match rx.changed().await {
                     Ok(()) => {
-                        let new_state = *rx.borrow();
+                        let new_state: &'static str = From::from(&*rx.borrow_and_update());
                         TENANT_STATE_METRIC
-                            .with_label_values(&[&tid, current_state.as_str()])
+                            .with_label_values(&[&tid, current_state])
                             .dec();
                         TENANT_STATE_METRIC
-                            .with_label_values(&[&tid, new_state.as_str()])
+                            .with_label_values(&[&tid, new_state])
                             .inc();
+
+                        current_state = new_state;
                     }
                     Err(_sender_dropped_error) => {
                         info!("Tenant dropped the state updates sender, quitting waiting for tenant state change");
@@ -1783,6 +1835,9 @@ impl Tenant {
         Tenant {
             tenant_id,
             conf,
+            // using now here is good enough approximation to catch tenants with really long
+            // activation times.
+            loading_started_at: Instant::now(),
             tenant_conf: Arc::new(RwLock::new(tenant_conf)),
             timelines: Mutex::new(HashMap::new()),
             gc_cs: tokio::sync::Mutex::new(()),
@@ -1865,7 +1920,7 @@ impl Tenant {
             .to_string();
 
             // Convert the config to a toml file.
-            conf_content += &toml_edit::easy::to_string(&tenant_conf)?;
+            conf_content += &toml_edit::ser::to_string(&tenant_conf)?;
 
             let mut target_config_file = VirtualFile::open_with_options(
                 target_config_path,
@@ -2093,7 +2148,7 @@ impl Tenant {
         src_timeline: &Arc<Timeline>,
         dst_id: TimelineId,
         start_lsn: Option<Lsn>,
-        _ctx: &RequestContext,
+        ctx: &RequestContext,
     ) -> anyhow::Result<Arc<Timeline>> {
         let src_id = src_timeline.timeline_id;
 
@@ -2186,7 +2241,7 @@ impl Tenant {
                 false,
                 Some(Arc::clone(src_timeline)),
             )?
-            .initialize_with_lock(&mut timelines, true, true)?;
+            .initialize_with_lock(ctx, &mut timelines, true, true)?;
         drop(timelines);
 
         // Root timeline gets its layers during creation and uploads them along with the metadata.
@@ -2299,7 +2354,7 @@ impl Tenant {
 
         let timeline = {
             let mut timelines = self.timelines.lock().unwrap();
-            raw_timeline.initialize_with_lock(&mut timelines, false, true)?
+            raw_timeline.initialize_with_lock(ctx, &mut timelines, false, true)?
         };
 
         info!(
@@ -2791,6 +2846,9 @@ pub mod harness {
                 trace_read_requests: Some(tenant_conf.trace_read_requests),
                 eviction_policy: Some(tenant_conf.eviction_policy),
                 min_resident_size_override: tenant_conf.min_resident_size_override,
+                evictions_low_residence_duration_metric_threshold: Some(
+                    tenant_conf.evictions_low_residence_duration_metric_threshold,
+                ),
             }
         }
     }
@@ -2823,7 +2881,13 @@ pub mod harness {
             };
 
             LOG_HANDLE.get_or_init(|| {
-                logging::init(logging::LogFormat::Test).expect("Failed to init test logging")
+                logging::init(
+                    logging::LogFormat::Test,
+                    // enable it in case in case the tests exercise code paths that use
+                    // debug_assert_current_span_has_tenant_and_timeline_id
+                    logging::TracingErrorLayerEnablement::EnableWithRustLogFilter,
+                )
+                .expect("Failed to init test logging")
             });
 
             let repo_dir = PageServerConf::test_repo_dir(test_name);
diff --git a/pageserver/src/tenant/config.rs b/pageserver/src/tenant/config.rs
index cdabb23a7b..34f57840fb 100644
--- a/pageserver/src/tenant/config.rs
+++ b/pageserver/src/tenant/config.rs
@@ -39,6 +39,7 @@ pub mod defaults {
     pub const DEFAULT_WALRECEIVER_CONNECT_TIMEOUT: &str = "2 seconds";
     pub const DEFAULT_WALRECEIVER_LAGGING_WAL_TIMEOUT: &str = "3 seconds";
     pub const DEFAULT_MAX_WALRECEIVER_LSN_WAL_LAG: u64 = 10 * 1024 * 1024;
+    pub const DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD: &str = "24 hour";
 }
 
 /// Per-tenant configuration options
@@ -93,6 +94,9 @@ pub struct TenantConf {
     pub trace_read_requests: bool,
     pub eviction_policy: EvictionPolicy,
     pub min_resident_size_override: Option<u64>,
+    // See the corresponding metric's help string.
+    #[serde(with = "humantime_serde")]
+    pub evictions_low_residence_duration_metric_threshold: Duration,
 }
 
 /// Same as TenantConf, but this struct preserves the information about
@@ -164,6 +168,11 @@ pub struct TenantConfOpt {
     #[serde(skip_serializing_if = "Option::is_none")]
     #[serde(default)]
     pub min_resident_size_override: Option<u64>,
+
+    #[serde(skip_serializing_if = "Option::is_none")]
+    #[serde(with = "humantime_serde")]
+    #[serde(default)]
+    pub evictions_low_residence_duration_metric_threshold: Option<Duration>,
 }
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
@@ -228,6 +237,9 @@ impl TenantConfOpt {
             min_resident_size_override: self
                 .min_resident_size_override
                 .or(global_conf.min_resident_size_override),
+            evictions_low_residence_duration_metric_threshold: self
+                .evictions_low_residence_duration_metric_threshold
+                .unwrap_or(global_conf.evictions_low_residence_duration_metric_threshold),
         }
     }
 }
@@ -260,6 +272,10 @@ impl Default for TenantConf {
             trace_read_requests: false,
             eviction_policy: EvictionPolicy::NoEviction,
             min_resident_size_override: None,
+            evictions_low_residence_duration_metric_threshold: humantime::parse_duration(
+                DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD,
+            )
+            .expect("cannot parse default evictions_low_residence_duration_metric_threshold"),
         }
     }
 }
@@ -275,9 +291,9 @@ mod tests {
             ..TenantConfOpt::default()
         };
 
-        let toml_form = toml_edit::easy::to_string(&small_conf).unwrap();
+        let toml_form = toml_edit::ser::to_string(&small_conf).unwrap();
         assert_eq!(toml_form, "gc_horizon = 42\n");
-        assert_eq!(small_conf, toml_edit::easy::from_str(&toml_form).unwrap());
+        assert_eq!(small_conf, toml_edit::de::from_str(&toml_form).unwrap());
 
         let json_form = serde_json::to_string(&small_conf).unwrap();
         assert_eq!(json_form, "{\"gc_horizon\":42}");
diff --git a/pageserver/src/tenant/layer_map.rs b/pageserver/src/tenant/layer_map.rs
index 4c659be9aa..8d06ccd565 100644
--- a/pageserver/src/tenant/layer_map.rs
+++ b/pageserver/src/tenant/layer_map.rs
@@ -48,7 +48,6 @@ mod layer_coverage;
 
 use crate::context::RequestContext;
 use crate::keyspace::KeyPartitioning;
-use crate::metrics::NUM_ONDISK_LAYERS;
 use crate::repository::Key;
 use crate::tenant::storage_layer::InMemoryLayer;
 use crate::tenant::storage_layer::Layer;
@@ -275,6 +274,7 @@ where
     /// Helper function for BatchedUpdates::insert_historic
     ///
     pub(self) fn insert_historic_noflush(&mut self, layer: Arc<L>) {
+        // TODO: See #3869, resulting #4088, attempted fix and repro #4094
         self.historic.insert(
             historic_layer_coverage::LayerKey::from(&*layer),
             Arc::clone(&layer),
@@ -283,8 +283,6 @@ where
         if Self::is_l0(&layer) {
             self.l0_delta_layers.push(layer);
         }
-
-        NUM_ONDISK_LAYERS.inc();
     }
 
     ///
@@ -309,8 +307,6 @@ where
                 "failed to locate removed historic layer from l0_delta_layers"
             );
         }
-
-        NUM_ONDISK_LAYERS.dec();
     }
 
     pub(self) fn replace_historic_noflush(
diff --git a/pageserver/src/tenant/mgr.rs b/pageserver/src/tenant/mgr.rs
index 4971186206..754316b3cd 100644
--- a/pageserver/src/tenant/mgr.rs
+++ b/pageserver/src/tenant/mgr.rs
@@ -537,7 +537,7 @@ where
             Some(tenant) => match tenant.current_state() {
                 TenantState::Attaching
                 | TenantState::Loading
-                | TenantState::Broken
+                | TenantState::Broken { .. }
                 | TenantState::Active => tenant.set_stopping(),
                 TenantState::Stopping => return Err(TenantStateError::IsStopping(tenant_id)),
             },
@@ -565,7 +565,7 @@ where
             let tenants_accessor = TENANTS.read().await;
             match tenants_accessor.get(&tenant_id) {
                 Some(tenant) => {
-                    tenant.set_broken(&e.to_string());
+                    tenant.set_broken(e.to_string());
                 }
                 None => {
                     warn!("Tenant {tenant_id} got removed from memory");
diff --git a/pageserver/src/tenant/remote_timeline_client.rs b/pageserver/src/tenant/remote_timeline_client.rs
index 28c4943dbd..c42824a8b5 100644
--- a/pageserver/src/tenant/remote_timeline_client.rs
+++ b/pageserver/src/tenant/remote_timeline_client.rs
@@ -219,7 +219,8 @@ use utils::lsn::Lsn;
 
 use crate::metrics::{
     MeasureRemoteOp, RemoteOpFileKind, RemoteOpKind, RemoteTimelineClientMetrics,
-    REMOTE_ONDEMAND_DOWNLOADED_BYTES, REMOTE_ONDEMAND_DOWNLOADED_LAYERS,
+    RemoteTimelineClientMetricsCallTrackSize, REMOTE_ONDEMAND_DOWNLOADED_BYTES,
+    REMOTE_ONDEMAND_DOWNLOADED_LAYERS,
 };
 use crate::tenant::remote_timeline_client::index::LayerFileMetadata;
 use crate::{
@@ -367,9 +368,13 @@ impl RemoteTimelineClient {
 
     /// Download index file
     pub async fn download_index_file(&self) -> Result<IndexPart, DownloadError> {
-        let _unfinished_gauge_guard = self
-            .metrics
-            .call_begin(&RemoteOpFileKind::Index, &RemoteOpKind::Download);
+        let _unfinished_gauge_guard = self.metrics.call_begin(
+            &RemoteOpFileKind::Index,
+            &RemoteOpKind::Download,
+            crate::metrics::RemoteTimelineClientMetricsCallTrackSize::DontTrackSize {
+                reason: "no need for a downloads gauge",
+            },
+        );
 
         download::download_index_part(
             self.conf,
@@ -398,9 +403,13 @@ impl RemoteTimelineClient {
         layer_metadata: &LayerFileMetadata,
     ) -> anyhow::Result<u64> {
         let downloaded_size = {
-            let _unfinished_gauge_guard = self
-                .metrics
-                .call_begin(&RemoteOpFileKind::Layer, &RemoteOpKind::Download);
+            let _unfinished_gauge_guard = self.metrics.call_begin(
+                &RemoteOpFileKind::Layer,
+                &RemoteOpKind::Download,
+                crate::metrics::RemoteTimelineClientMetricsCallTrackSize::DontTrackSize {
+                    reason: "no need for a downloads gauge",
+                },
+            );
             download::download_layer_file(
                 self.conf,
                 &self.storage_impl,
@@ -886,11 +895,32 @@ impl RemoteTimelineClient {
     fn calls_unfinished_metric_impl(
         &self,
         op: &UploadOp,
-    ) -> Option<(RemoteOpFileKind, RemoteOpKind)> {
+    ) -> Option<(
+        RemoteOpFileKind,
+        RemoteOpKind,
+        RemoteTimelineClientMetricsCallTrackSize,
+    )> {
+        use RemoteTimelineClientMetricsCallTrackSize::DontTrackSize;
         let res = match op {
-            UploadOp::UploadLayer(_, _) => (RemoteOpFileKind::Layer, RemoteOpKind::Upload),
-            UploadOp::UploadMetadata(_, _) => (RemoteOpFileKind::Index, RemoteOpKind::Upload),
-            UploadOp::Delete(file_kind, _) => (*file_kind, RemoteOpKind::Delete),
+            UploadOp::UploadLayer(_, m) => (
+                RemoteOpFileKind::Layer,
+                RemoteOpKind::Upload,
+                RemoteTimelineClientMetricsCallTrackSize::Bytes(m.file_size()),
+            ),
+            UploadOp::UploadMetadata(_, _) => (
+                RemoteOpFileKind::Index,
+                RemoteOpKind::Upload,
+                DontTrackSize {
+                    reason: "metadata uploads are tiny",
+                },
+            ),
+            UploadOp::Delete(file_kind, _) => (
+                *file_kind,
+                RemoteOpKind::Delete,
+                DontTrackSize {
+                    reason: "should we track deletes? positive or negative sign?",
+                },
+            ),
             UploadOp::Barrier(_) => {
                 // we do not account these
                 return None;
@@ -900,20 +930,20 @@ impl RemoteTimelineClient {
     }
 
     fn calls_unfinished_metric_begin(&self, op: &UploadOp) {
-        let (file_kind, op_kind) = match self.calls_unfinished_metric_impl(op) {
+        let (file_kind, op_kind, track_bytes) = match self.calls_unfinished_metric_impl(op) {
             Some(x) => x,
             None => return,
         };
-        let guard = self.metrics.call_begin(&file_kind, &op_kind);
+        let guard = self.metrics.call_begin(&file_kind, &op_kind, track_bytes);
         guard.will_decrement_manually(); // in unfinished_ops_metric_end()
     }
 
     fn calls_unfinished_metric_end(&self, op: &UploadOp) {
-        let (file_kind, op_kind) = match self.calls_unfinished_metric_impl(op) {
+        let (file_kind, op_kind, track_bytes) = match self.calls_unfinished_metric_impl(op) {
             Some(x) => x,
             None => return,
         };
-        self.metrics.call_end(&file_kind, &op_kind);
+        self.metrics.call_end(&file_kind, &op_kind, track_bytes);
     }
 
     fn stop(&self) {
@@ -981,11 +1011,19 @@ impl RemoteTimelineClient {
 mod tests {
     use super::*;
     use crate::{
-        tenant::harness::{TenantHarness, TIMELINE_ID},
+        context::RequestContext,
+        tenant::{
+            harness::{TenantHarness, TIMELINE_ID},
+            Tenant,
+        },
         DEFAULT_PG_VERSION,
     };
     use remote_storage::{RemoteStorageConfig, RemoteStorageKind};
-    use std::{collections::HashSet, path::Path};
+    use std::{
+        collections::HashSet,
+        path::{Path, PathBuf},
+    };
+    use tokio::runtime::EnterGuard;
     use utils::lsn::Lsn;
 
     pub(super) fn dummy_contents(name: &str) -> Vec<u8> {
@@ -1034,39 +1072,80 @@ mod tests {
         assert_eq!(found, expected);
     }
 
+    struct TestSetup {
+        runtime: &'static tokio::runtime::Runtime,
+        entered_runtime: EnterGuard<'static>,
+        harness: TenantHarness<'static>,
+        tenant: Arc<Tenant>,
+        tenant_ctx: RequestContext,
+        remote_fs_dir: PathBuf,
+        client: Arc<RemoteTimelineClient>,
+    }
+
+    impl TestSetup {
+        fn new(test_name: &str) -> anyhow::Result<Self> {
+            // Use a current-thread runtime in the test
+            let runtime = Box::leak(Box::new(
+                tokio::runtime::Builder::new_current_thread()
+                    .enable_all()
+                    .build()?,
+            ));
+            let entered_runtime = runtime.enter();
+
+            let test_name = Box::leak(Box::new(format!("remote_timeline_client__{test_name}")));
+            let harness = TenantHarness::create(test_name)?;
+            let (tenant, ctx) = runtime.block_on(harness.load());
+            // create an empty timeline directory
+            let timeline =
+                tenant.create_empty_timeline(TIMELINE_ID, Lsn(0), DEFAULT_PG_VERSION, &ctx)?;
+            let _ = timeline.initialize(&ctx).unwrap();
+
+            let remote_fs_dir = harness.conf.workdir.join("remote_fs");
+            std::fs::create_dir_all(remote_fs_dir)?;
+            let remote_fs_dir = std::fs::canonicalize(harness.conf.workdir.join("remote_fs"))?;
+
+            let storage_config = RemoteStorageConfig {
+                max_concurrent_syncs: std::num::NonZeroUsize::new(
+                    remote_storage::DEFAULT_REMOTE_STORAGE_MAX_CONCURRENT_SYNCS,
+                )
+                .unwrap(),
+                max_sync_errors: std::num::NonZeroU32::new(
+                    remote_storage::DEFAULT_REMOTE_STORAGE_MAX_SYNC_ERRORS,
+                )
+                .unwrap(),
+                storage: RemoteStorageKind::LocalFs(remote_fs_dir.clone()),
+            };
+
+            let storage = GenericRemoteStorage::from_config(&storage_config).unwrap();
+
+            let client = Arc::new(RemoteTimelineClient {
+                conf: harness.conf,
+                runtime,
+                tenant_id: harness.tenant_id,
+                timeline_id: TIMELINE_ID,
+                storage_impl: storage,
+                upload_queue: Mutex::new(UploadQueue::Uninitialized),
+                metrics: Arc::new(RemoteTimelineClientMetrics::new(
+                    &harness.tenant_id,
+                    &TIMELINE_ID,
+                )),
+            });
+
+            Ok(Self {
+                runtime,
+                entered_runtime,
+                harness,
+                tenant,
+                tenant_ctx: ctx,
+                remote_fs_dir,
+                client,
+            })
+        }
+    }
+
     // Test scheduling
     #[test]
     fn upload_scheduling() -> anyhow::Result<()> {
-        // Use a current-thread runtime in the test
-        let runtime = Box::leak(Box::new(
-            tokio::runtime::Builder::new_current_thread()
-                .enable_all()
-                .build()?,
-        ));
-        let _entered = runtime.enter();
-
-        let harness = TenantHarness::create("upload_scheduling")?;
-        let (tenant, ctx) = runtime.block_on(harness.load());
-        let _timeline =
-            tenant.create_empty_timeline(TIMELINE_ID, Lsn(0), DEFAULT_PG_VERSION, &ctx)?;
-        let timeline_path = harness.timeline_path(&TIMELINE_ID);
-
-        let remote_fs_dir = harness.conf.workdir.join("remote_fs");
-        std::fs::create_dir_all(remote_fs_dir)?;
-        let remote_fs_dir = std::fs::canonicalize(harness.conf.workdir.join("remote_fs"))?;
-
-        let storage_config = RemoteStorageConfig {
-            max_concurrent_syncs: std::num::NonZeroUsize::new(
-                remote_storage::DEFAULT_REMOTE_STORAGE_MAX_CONCURRENT_SYNCS,
-            )
-            .unwrap(),
-            max_sync_errors: std::num::NonZeroU32::new(
-                remote_storage::DEFAULT_REMOTE_STORAGE_MAX_SYNC_ERRORS,
-            )
-            .unwrap(),
-            storage: RemoteStorageKind::LocalFs(remote_fs_dir.clone()),
-        };
-
         // Test outline:
         //
         // Schedule upload of a bunch of layers. Check that they are started immediately, not queued
@@ -1081,21 +1160,19 @@ mod tests {
         // Schedule another deletion. Check that it's launched immediately.
         // Schedule index upload. Check that it's queued
 
-        println!("workdir: {}", harness.conf.workdir.display());
-
-        let storage_impl = GenericRemoteStorage::from_config(&storage_config)?;
-        let client = Arc::new(RemoteTimelineClient {
-            conf: harness.conf,
+        let TestSetup {
             runtime,
-            tenant_id: harness.tenant_id,
-            timeline_id: TIMELINE_ID,
-            storage_impl,
-            upload_queue: Mutex::new(UploadQueue::Uninitialized),
-            metrics: Arc::new(RemoteTimelineClientMetrics::new(
-                &harness.tenant_id,
-                &TIMELINE_ID,
-            )),
-        });
+            entered_runtime: _entered_runtime,
+            harness,
+            tenant: _tenant,
+            tenant_ctx: _tenant_ctx,
+            remote_fs_dir,
+            client,
+        } = TestSetup::new("upload_scheduling").unwrap();
+
+        let timeline_path = harness.timeline_path(&TIMELINE_ID);
+
+        println!("workdir: {}", harness.conf.workdir.display());
 
         let remote_timeline_dir =
             remote_fs_dir.join(timeline_path.strip_prefix(&harness.conf.workdir)?);
@@ -1216,4 +1293,90 @@ mod tests {
 
         Ok(())
     }
+
+    #[test]
+    fn bytes_unfinished_gauge_for_layer_file_uploads() -> anyhow::Result<()> {
+        // Setup
+
+        let TestSetup {
+            runtime,
+            harness,
+            client,
+            ..
+        } = TestSetup::new("metrics")?;
+
+        let metadata = dummy_metadata(Lsn(0x10));
+        client.init_upload_queue_for_empty_remote(&metadata)?;
+
+        let timeline_path = harness.timeline_path(&TIMELINE_ID);
+
+        let layer_file_name_1: LayerFileName = "000000000000000000000000000000000000-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF__00000000016B59D8-00000000016B5A51".parse().unwrap();
+        let content_1 = dummy_contents("foo");
+        std::fs::write(
+            timeline_path.join(layer_file_name_1.file_name()),
+            &content_1,
+        )?;
+
+        #[derive(Debug, PartialEq)]
+        struct BytesStartedFinished {
+            started: Option<usize>,
+            finished: Option<usize>,
+        }
+        let get_bytes_started_stopped = || {
+            let started = client
+                .metrics
+                .get_bytes_started_counter_value(&RemoteOpFileKind::Layer, &RemoteOpKind::Upload)
+                .map(|v| v.try_into().unwrap());
+            let stopped = client
+                .metrics
+                .get_bytes_finished_counter_value(&RemoteOpFileKind::Layer, &RemoteOpKind::Upload)
+                .map(|v| v.try_into().unwrap());
+            BytesStartedFinished {
+                started,
+                finished: stopped,
+            }
+        };
+
+        // Test
+
+        let init = get_bytes_started_stopped();
+
+        client.schedule_layer_file_upload(
+            &layer_file_name_1,
+            &LayerFileMetadata::new(content_1.len() as u64),
+        )?;
+
+        let pre = get_bytes_started_stopped();
+
+        runtime.block_on(client.wait_completion())?;
+
+        let post = get_bytes_started_stopped();
+
+        // Validate
+
+        assert_eq!(
+            init,
+            BytesStartedFinished {
+                started: None,
+                finished: None
+            }
+        );
+        assert_eq!(
+            pre,
+            BytesStartedFinished {
+                started: Some(content_1.len()),
+                // assert that the _finished metric is created eagerly so that subtractions work on first sample
+                finished: Some(0),
+            }
+        );
+        assert_eq!(
+            post,
+            BytesStartedFinished {
+                started: Some(content_1.len()),
+                finished: Some(content_1.len())
+            }
+        );
+
+        Ok(())
+    }
 }
diff --git a/pageserver/src/tenant/remote_timeline_client/download.rs b/pageserver/src/tenant/remote_timeline_client/download.rs
index bda095d850..a0d8c0193a 100644
--- a/pageserver/src/tenant/remote_timeline_client/download.rs
+++ b/pageserver/src/tenant/remote_timeline_client/download.rs
@@ -16,6 +16,7 @@ use tracing::{info, warn};
 
 use crate::config::PageServerConf;
 use crate::tenant::storage_layer::LayerFileName;
+use crate::tenant::timeline::debug_assert_current_span_has_tenant_and_timeline_id;
 use crate::{exponential_backoff, DEFAULT_BASE_BACKOFF_SECONDS, DEFAULT_MAX_BACKOFF_SECONDS};
 use remote_storage::{DownloadError, GenericRemoteStorage};
 use utils::crashsafe::path_with_suffix_extension;
@@ -43,6 +44,8 @@ pub async fn download_layer_file<'a>(
     layer_file_name: &'a LayerFileName,
     layer_metadata: &'a LayerFileMetadata,
 ) -> Result<u64, DownloadError> {
+    debug_assert_current_span_has_tenant_and_timeline_id();
+
     let timeline_path = conf.timeline_path(&timeline_id, &tenant_id);
 
     let local_path = timeline_path.join(layer_file_name.file_name());
@@ -154,7 +157,7 @@ pub async fn download_layer_file<'a>(
         .with_context(|| format!("Could not fsync layer file {}", local_path.display(),))
         .map_err(DownloadError::Other)?;
 
-    tracing::info!("download complete: {}", local_path.display());
+    tracing::debug!("download complete: {}", local_path.display());
 
     Ok(bytes_amount)
 }
diff --git a/pageserver/src/tenant/remote_timeline_client/upload.rs b/pageserver/src/tenant/remote_timeline_client/upload.rs
index ce9f4d9bf8..699121ccd9 100644
--- a/pageserver/src/tenant/remote_timeline_client/upload.rs
+++ b/pageserver/src/tenant/remote_timeline_client/upload.rs
@@ -74,7 +74,7 @@ pub(super) async fn upload_timeline_layer<'a>(
     })?;
 
     storage
-        .upload(Box::new(source_file), fs_size, &storage_path, None)
+        .upload(source_file, fs_size, &storage_path, None)
         .await
         .with_context(|| {
             format!(
diff --git a/pageserver/src/tenant/tasks.rs b/pageserver/src/tenant/tasks.rs
index 8aeacc12f5..7e7dbd3c5c 100644
--- a/pageserver/src/tenant/tasks.rs
+++ b/pageserver/src/tenant/tasks.rs
@@ -209,7 +209,7 @@ async fn wait_for_active_tenant(
         loop {
             match tenant_state_updates.changed().await {
                 Ok(()) => {
-                    let new_state = *tenant_state_updates.borrow();
+                    let new_state = &*tenant_state_updates.borrow();
                     match new_state {
                         TenantState::Active => {
                             debug!("Tenant state changed to active, continuing the task loop");
diff --git a/pageserver/src/tenant/timeline.rs b/pageserver/src/tenant/timeline.rs
index e80e32644b..8768841d87 100644
--- a/pageserver/src/tenant/timeline.rs
+++ b/pageserver/src/tenant/timeline.rs
@@ -14,6 +14,7 @@ use pageserver_api::models::{
     DownloadRemoteLayersTaskState, LayerMapInfo, LayerResidenceStatus, TimelineState,
 };
 use remote_storage::GenericRemoteStorage;
+use storage_broker::BrokerClientChannel;
 use tokio::sync::{oneshot, watch, Semaphore, TryAcquireError};
 use tokio_util::sync::CancellationToken;
 use tracing::*;
@@ -30,7 +31,7 @@ use std::sync::atomic::{AtomicI64, Ordering as AtomicOrdering};
 use std::sync::{Arc, Mutex, MutexGuard, RwLock, Weak};
 use std::time::{Duration, Instant, SystemTime};
 
-use crate::broker_client::is_broker_client_initialized;
+use crate::broker_client::{get_broker_client, is_broker_client_initialized};
 use crate::context::{DownloadBehavior, RequestContext};
 use crate::tenant::remote_timeline_client::{self, index::LayerFileMetadata};
 use crate::tenant::storage_layer::{
@@ -47,7 +48,7 @@ use crate::tenant::{
 
 use crate::config::PageServerConf;
 use crate::keyspace::{KeyPartitioning, KeySpace};
-use crate::metrics::TimelineMetrics;
+use crate::metrics::{TimelineMetrics, UNEXPECTED_ONDEMAND_DOWNLOADS};
 use crate::pgdatadir_mapping::LsnForTimestamp;
 use crate::pgdatadir_mapping::{is_rel_fsm_block_key, is_rel_vm_block_key};
 use crate::pgdatadir_mapping::{BlockNumber, CalculateLogicalSizeError};
@@ -71,11 +72,12 @@ use crate::walredo::WalRedoManager;
 use crate::METADATA_FILE_NAME;
 use crate::ZERO_PAGE;
 use crate::{is_temporary, task_mgr};
-use walreceiver::spawn_connection_manager_task;
 
 pub(super) use self::eviction_task::EvictionTaskTenantState;
 use self::eviction_task::EvictionTaskTimelineState;
+use self::walreceiver::{WalReceiver, WalReceiverConf};
 
+use super::config::TenantConf;
 use super::layer_map::BatchedUpdates;
 use super::remote_timeline_client::index::IndexPart;
 use super::remote_timeline_client::RemoteTimelineClient;
@@ -160,7 +162,7 @@ pub struct Timeline {
     ancestor_timeline: Option<Arc<Timeline>>,
     ancestor_lsn: Lsn,
 
-    metrics: TimelineMetrics,
+    pub(super) metrics: TimelineMetrics,
 
     /// Ensures layers aren't frozen by checkpointer between
     /// [`Timeline::get_layer_for_write`] and layer reads.
@@ -214,6 +216,7 @@ pub struct Timeline {
     /// or None if WAL receiver has not received anything for this timeline
     /// yet.
     pub last_received_wal: Mutex<Option<WalReceiverInfo>>,
+    pub walreceiver: WalReceiver,
 
     /// Relation size cache
     pub rel_size_cache: RwLock<HashMap<RelTag, (Lsn, BlockNumber)>>,
@@ -866,10 +869,18 @@ impl Timeline {
         Ok(())
     }
 
-    pub fn activate(self: &Arc<Self>) {
+    pub fn activate(self: &Arc<Self>, ctx: &RequestContext) -> anyhow::Result<()> {
+        if is_broker_client_initialized() {
+            self.launch_wal_receiver(ctx, get_broker_client().clone())?;
+        } else if cfg!(test) {
+            info!("not launching WAL receiver because broker client hasn't been initialized");
+        } else {
+            anyhow::bail!("broker client not initialized");
+        }
+
         self.set_state(TimelineState::Active);
-        self.launch_wal_receiver();
         self.launch_eviction_task();
+        Ok(())
     }
 
     pub fn set_state(&self, new_state: TimelineState) {
@@ -925,6 +936,7 @@ impl Timeline {
         }
     }
 
+    #[instrument(skip_all, fields(tenant = %self.tenant_id, timeline = %self.timeline_id))]
     pub async fn download_layer(&self, layer_file_name: &str) -> anyhow::Result<Option<bool>> {
         let Some(layer) = self.find_layer(layer_file_name) else { return Ok(None) };
         let Some(remote_layer) = layer.downcast_remote_layer() else { return  Ok(Some(false)) };
@@ -1126,6 +1138,8 @@ impl Timeline {
                 if let Some(delta) = local_layer_residence_duration {
                     self.metrics
                         .evictions_with_low_residence_duration
+                        .read()
+                        .unwrap()
                         .observe(delta);
                     info!(layer=%local_layer.short_id(), residence_millis=delta.as_millis(), "evicted layer after known residence period");
                 } else {
@@ -1199,6 +1213,35 @@ impl Timeline {
             .unwrap_or(self.conf.default_tenant_conf.eviction_policy)
     }
 
+    fn get_evictions_low_residence_duration_metric_threshold(
+        tenant_conf: &TenantConfOpt,
+        default_tenant_conf: &TenantConf,
+    ) -> Duration {
+        tenant_conf
+            .evictions_low_residence_duration_metric_threshold
+            .unwrap_or(default_tenant_conf.evictions_low_residence_duration_metric_threshold)
+    }
+
+    pub(super) fn tenant_conf_updated(&self) {
+        // NB: Most tenant conf options are read by background loops, so,
+        // changes will automatically be picked up.
+
+        // The threshold is embedded in the metric. So, we need to update it.
+        {
+            let new_threshold = Self::get_evictions_low_residence_duration_metric_threshold(
+                &self.tenant_conf.read().unwrap(),
+                &self.conf.default_tenant_conf,
+            );
+            let tenant_id_str = self.tenant_id.to_string();
+            let timeline_id_str = self.timeline_id.to_string();
+            self.metrics
+                .evictions_with_low_residence_duration
+                .write()
+                .unwrap()
+                .change_threshold(&tenant_id_str, &timeline_id_str, new_threshold);
+        }
+    }
+
     /// Open a Timeline handle.
     ///
     /// Loads the metadata for the timeline into memory, but not the layer map.
@@ -1220,7 +1263,36 @@ impl Timeline {
         let (layer_flush_start_tx, _) = tokio::sync::watch::channel(0);
         let (layer_flush_done_tx, _) = tokio::sync::watch::channel((0, Ok(())));
 
+        let tenant_conf_guard = tenant_conf.read().unwrap();
+        let wal_connect_timeout = tenant_conf_guard
+            .walreceiver_connect_timeout
+            .unwrap_or(conf.default_tenant_conf.walreceiver_connect_timeout);
+        let lagging_wal_timeout = tenant_conf_guard
+            .lagging_wal_timeout
+            .unwrap_or(conf.default_tenant_conf.lagging_wal_timeout);
+        let max_lsn_wal_lag = tenant_conf_guard
+            .max_lsn_wal_lag
+            .unwrap_or(conf.default_tenant_conf.max_lsn_wal_lag);
+        let evictions_low_residence_duration_metric_threshold =
+            Self::get_evictions_low_residence_duration_metric_threshold(
+                &tenant_conf_guard,
+                &conf.default_tenant_conf,
+            );
+        drop(tenant_conf_guard);
+
         Arc::new_cyclic(|myself| {
+            let walreceiver = WalReceiver::new(
+                TenantTimelineId::new(tenant_id, timeline_id),
+                Weak::clone(myself),
+                WalReceiverConf {
+                    wal_connect_timeout,
+                    lagging_wal_timeout,
+                    max_lsn_wal_lag,
+                    auth_token: crate::config::SAFEKEEPER_AUTH_TOKEN.get().cloned(),
+                    availability_zone: conf.availability_zone.clone(),
+                },
+            );
+
             let mut result = Timeline {
                 conf,
                 tenant_conf,
@@ -1231,6 +1303,7 @@ impl Timeline {
                 layers: RwLock::new(LayerMap::default()),
 
                 walredo_mgr,
+                walreceiver,
 
                 remote_client: remote_client.map(Arc::new),
 
@@ -1252,7 +1325,7 @@ impl Timeline {
                     &timeline_id,
                     crate::metrics::EvictionsWithLowResidenceDurationBuilder::new(
                         "mtime",
-                        conf.evictions_low_residence_duration_metric_threshold,
+                        evictions_low_residence_duration_metric_threshold,
                     ),
                 ),
 
@@ -1350,44 +1423,17 @@ impl Timeline {
         *flush_loop_state = FlushLoopState::Running;
     }
 
-    pub(super) fn launch_wal_receiver(self: &Arc<Self>) {
-        if !is_broker_client_initialized() {
-            if cfg!(test) {
-                info!("not launching WAL receiver because broker client hasn't been initialized");
-                return;
-            } else {
-                panic!("broker client not initialized");
-            }
-        }
-
+    pub(super) fn launch_wal_receiver(
+        &self,
+        ctx: &RequestContext,
+        broker_client: BrokerClientChannel,
+    ) -> anyhow::Result<()> {
         info!(
             "launching WAL receiver for timeline {} of tenant {}",
             self.timeline_id, self.tenant_id
         );
-        let tenant_conf_guard = self.tenant_conf.read().unwrap();
-        let lagging_wal_timeout = tenant_conf_guard
-            .lagging_wal_timeout
-            .unwrap_or(self.conf.default_tenant_conf.lagging_wal_timeout);
-        let walreceiver_connect_timeout = tenant_conf_guard
-            .walreceiver_connect_timeout
-            .unwrap_or(self.conf.default_tenant_conf.walreceiver_connect_timeout);
-        let max_lsn_wal_lag = tenant_conf_guard
-            .max_lsn_wal_lag
-            .unwrap_or(self.conf.default_tenant_conf.max_lsn_wal_lag);
-        drop(tenant_conf_guard);
-        let self_clone = Arc::clone(self);
-        let background_ctx =
-            // XXX: this is a detached_child. Plumb through the ctx from call sites.
-            RequestContext::todo_child(TaskKind::WalReceiverManager, DownloadBehavior::Error);
-        spawn_connection_manager_task(
-            self_clone,
-            walreceiver_connect_timeout,
-            lagging_wal_timeout,
-            max_lsn_wal_lag,
-            crate::config::SAFEKEEPER_AUTH_TOKEN.get().cloned(),
-            self.conf.availability_zone.clone(),
-            background_ctx,
-        );
+        self.walreceiver.start(ctx, broker_client)?;
+        Ok(())
     }
 
     ///
@@ -2303,6 +2349,7 @@ impl Timeline {
                             id,
                             ctx.task_kind()
                         );
+                        UNEXPECTED_ONDEMAND_DOWNLOADS.inc();
                         timeline.download_remote_layer(remote_layer).await?;
                         continue 'layer_map_search;
                     }
@@ -3766,11 +3813,13 @@ impl Timeline {
     /// If the caller has a deadline or needs a timeout, they can simply stop polling:
     /// we're **cancellation-safe** because the download happens in a separate task_mgr task.
     /// So, the current download attempt will run to completion even if we stop polling.
-    #[instrument(skip_all, fields(tenant_id=%self.tenant_id, timeline_id=%self.timeline_id, layer=%remote_layer.short_id()))]
+    #[instrument(skip_all, fields(layer=%remote_layer.short_id()))]
     pub async fn download_remote_layer(
         &self,
         remote_layer: Arc<RemoteLayer>,
     ) -> anyhow::Result<()> {
+        debug_assert_current_span_has_tenant_and_timeline_id();
+
         use std::sync::atomic::Ordering::Relaxed;
 
         let permit = match Arc::clone(&remote_layer.ongoing_download)
@@ -3814,6 +3863,8 @@ impl Timeline {
                     .await;
 
                 if let Ok(size) = &result {
+                    info!("layer file download finished");
+
                     // XXX the temp file is still around in Err() case
                     // and consumes space until we clean up upon pageserver restart.
                     self_clone.metrics.resident_physical_size_gauge.add(*size);
@@ -3885,6 +3936,8 @@ impl Timeline {
                     updates.flush();
                     drop(layers);
 
+                    info!("on-demand download successful");
+
                     // Now that we've inserted the download into the layer map,
                     // close the semaphore. This will make other waiters for
                     // this download return Ok(()).
@@ -3892,7 +3945,7 @@ impl Timeline {
                     remote_layer.ongoing_download.close();
                 } else {
                     // Keep semaphore open. We'll drop the permit at the end of the function.
-                    error!("on-demand download failed: {:?}", result.as_ref().unwrap_err());
+                    error!("layer file download failed: {:?}", result.as_ref().unwrap_err());
                 }
 
                 // Don't treat it as an error if the task that triggered the download
@@ -4203,3 +4256,36 @@ fn rename_to_backup(path: &Path) -> anyhow::Result<()> {
 
     bail!("couldn't find an unused backup number for {:?}", path)
 }
+
+#[cfg(not(debug_assertions))]
+#[inline]
+pub(crate) fn debug_assert_current_span_has_tenant_and_timeline_id() {}
+
+#[cfg(debug_assertions)]
+#[inline]
+pub(crate) fn debug_assert_current_span_has_tenant_and_timeline_id() {
+    use utils::tracing_span_assert;
+
+    pub static TENANT_ID_EXTRACTOR: once_cell::sync::Lazy<
+        tracing_span_assert::MultiNameExtractor<2>,
+    > = once_cell::sync::Lazy::new(|| {
+        tracing_span_assert::MultiNameExtractor::new("TenantId", ["tenant_id", "tenant"])
+    });
+
+    pub static TIMELINE_ID_EXTRACTOR: once_cell::sync::Lazy<
+        tracing_span_assert::MultiNameExtractor<2>,
+    > = once_cell::sync::Lazy::new(|| {
+        tracing_span_assert::MultiNameExtractor::new("TimelineId", ["timeline_id", "timeline"])
+    });
+
+    match tracing_span_assert::check_fields_present([
+        &*TENANT_ID_EXTRACTOR,
+        &*TIMELINE_ID_EXTRACTOR,
+    ]) {
+        Ok(()) => (),
+        Err(missing) => panic!(
+            "missing extractors: {:?}",
+            missing.into_iter().map(|e| e.name()).collect::<Vec<_>>()
+        ),
+    }
+}
diff --git a/pageserver/src/tenant/timeline/walreceiver.rs b/pageserver/src/tenant/timeline/walreceiver.rs
index f33a12c5cc..00f446af38 100644
--- a/pageserver/src/tenant/timeline/walreceiver.rs
+++ b/pageserver/src/tenant/timeline/walreceiver.rs
@@ -23,14 +23,133 @@
 mod connection_manager;
 mod walreceiver_connection;
 
-use crate::task_mgr::WALRECEIVER_RUNTIME;
+use crate::context::{DownloadBehavior, RequestContext};
+use crate::task_mgr::{self, TaskKind, WALRECEIVER_RUNTIME};
+use crate::tenant::timeline::walreceiver::connection_manager::{
+    connection_manager_loop_step, ConnectionManagerState,
+};
 
+use anyhow::Context;
 use std::future::Future;
+use std::num::NonZeroU64;
+use std::ops::ControlFlow;
+use std::sync::atomic::{self, AtomicBool};
+use std::sync::{Arc, Weak};
+use std::time::Duration;
+use storage_broker::BrokerClientChannel;
+use tokio::select;
 use tokio::sync::watch;
 use tokio_util::sync::CancellationToken;
 use tracing::*;
 
-pub use connection_manager::spawn_connection_manager_task;
+use utils::id::TenantTimelineId;
+
+use super::Timeline;
+
+#[derive(Clone)]
+pub struct WalReceiverConf {
+    /// The timeout on the connection to safekeeper for WAL streaming.
+    pub wal_connect_timeout: Duration,
+    /// The timeout to use to determine when the current connection is "stale" and reconnect to the other one.
+    pub lagging_wal_timeout: Duration,
+    /// The Lsn lag to use to determine when the current connection is lagging to much behind and reconnect to the other one.
+    pub max_lsn_wal_lag: NonZeroU64,
+    pub auth_token: Option<Arc<String>>,
+    pub availability_zone: Option<String>,
+}
+
+pub struct WalReceiver {
+    timeline: TenantTimelineId,
+    timeline_ref: Weak<Timeline>,
+    conf: WalReceiverConf,
+    started: AtomicBool,
+}
+
+impl WalReceiver {
+    pub fn new(
+        timeline: TenantTimelineId,
+        timeline_ref: Weak<Timeline>,
+        conf: WalReceiverConf,
+    ) -> Self {
+        Self {
+            timeline,
+            timeline_ref,
+            conf,
+            started: AtomicBool::new(false),
+        }
+    }
+
+    pub fn start(
+        &self,
+        ctx: &RequestContext,
+        mut broker_client: BrokerClientChannel,
+    ) -> anyhow::Result<()> {
+        if self.started.load(atomic::Ordering::Acquire) {
+            anyhow::bail!("Wal receiver is already started");
+        }
+
+        let timeline = self.timeline_ref.upgrade().with_context(|| {
+            format!("walreceiver start on a dropped timeline {}", self.timeline)
+        })?;
+
+        let tenant_id = timeline.tenant_id;
+        let timeline_id = timeline.timeline_id;
+        let walreceiver_ctx =
+            ctx.detached_child(TaskKind::WalReceiverManager, DownloadBehavior::Error);
+
+        let wal_receiver_conf = self.conf.clone();
+        task_mgr::spawn(
+            WALRECEIVER_RUNTIME.handle(),
+            TaskKind::WalReceiverManager,
+            Some(tenant_id),
+            Some(timeline_id),
+            &format!("walreceiver for timeline {tenant_id}/{timeline_id}"),
+            false,
+            async move {
+                info!("WAL receiver manager started, connecting to broker");
+                let mut connection_manager_state = ConnectionManagerState::new(
+                    timeline,
+                    wal_receiver_conf,
+                );
+                loop {
+                    select! {
+                        _ = task_mgr::shutdown_watcher() => {
+                            info!("WAL receiver shutdown requested, shutting down");
+                            connection_manager_state.shutdown().await;
+                            return Ok(());
+                        },
+                        loop_step_result = connection_manager_loop_step(
+                            &mut broker_client,
+                            &mut connection_manager_state,
+                            &walreceiver_ctx,
+                        ) => match loop_step_result {
+                            ControlFlow::Continue(()) => continue,
+                            ControlFlow::Break(()) => {
+                                info!("Connection manager loop ended, shutting down");
+                                connection_manager_state.shutdown().await;
+                                return Ok(());
+                            }
+                        },
+                    }
+                }
+            }.instrument(info_span!(parent: None, "wal_connection_manager", tenant = %tenant_id, timeline = %timeline_id))
+        );
+
+        self.started.store(true, atomic::Ordering::Release);
+
+        Ok(())
+    }
+
+    pub async fn stop(&self) {
+        task_mgr::shutdown_tasks(
+            Some(TaskKind::WalReceiverManager),
+            Some(self.timeline.tenant_id),
+            Some(self.timeline.timeline_id),
+        )
+        .await;
+        self.started.store(false, atomic::Ordering::Release);
+    }
+}
 
 /// A handle of an asynchronous task.
 /// The task has a channel that it can use to communicate its lifecycle events in a certain form, see [`TaskEvent`]
@@ -39,26 +158,26 @@ pub use connection_manager::spawn_connection_manager_task;
 /// Note that the communication happens via the `watch` channel, that does not accumulate the events, replacing the old one with the never one on submission.
 /// That may lead to certain events not being observed by the listener.
 #[derive(Debug)]
-pub struct TaskHandle<E> {
+struct TaskHandle<E> {
     join_handle: Option<tokio::task::JoinHandle<anyhow::Result<()>>>,
     events_receiver: watch::Receiver<TaskStateUpdate<E>>,
     cancellation: CancellationToken,
 }
 
-pub enum TaskEvent<E> {
+enum TaskEvent<E> {
     Update(TaskStateUpdate<E>),
     End(anyhow::Result<()>),
 }
 
 #[derive(Debug, Clone)]
-pub enum TaskStateUpdate<E> {
+enum TaskStateUpdate<E> {
     Started,
     Progress(E),
 }
 
 impl<E: Clone> TaskHandle<E> {
     /// Initializes the task, starting it immediately after the creation.
-    pub fn spawn<Fut>(
+    fn spawn<Fut>(
         task: impl FnOnce(watch::Sender<TaskStateUpdate<E>>, CancellationToken) -> Fut + Send + 'static,
     ) -> Self
     where
@@ -131,7 +250,7 @@ impl<E: Clone> TaskHandle<E> {
     }
 
     /// Aborts current task, waiting for it to finish.
-    pub async fn shutdown(self) {
+    async fn shutdown(self) {
         if let Some(jh) = self.join_handle {
             self.cancellation.cancel();
             match jh.await {
diff --git a/pageserver/src/tenant/timeline/walreceiver/connection_manager.rs b/pageserver/src/tenant/timeline/walreceiver/connection_manager.rs
index de07676ffe..731c5c4644 100644
--- a/pageserver/src/tenant/timeline/walreceiver/connection_manager.rs
+++ b/pageserver/src/tenant/timeline/walreceiver/connection_manager.rs
@@ -11,11 +11,9 @@
 
 use std::{collections::HashMap, num::NonZeroU64, ops::ControlFlow, sync::Arc, time::Duration};
 
-use super::TaskStateUpdate;
-use crate::broker_client::get_broker_client;
+use super::{TaskStateUpdate, WalReceiverConf};
 use crate::context::{DownloadBehavior, RequestContext};
-use crate::task_mgr::WALRECEIVER_RUNTIME;
-use crate::task_mgr::{self, TaskKind};
+use crate::task_mgr::TaskKind;
 use crate::tenant::Timeline;
 use anyhow::Context;
 use chrono::{NaiveDateTime, Utc};
@@ -38,75 +36,17 @@ use utils::{
 
 use super::{walreceiver_connection::WalConnectionStatus, TaskEvent, TaskHandle};
 
-/// Spawns the loop to take care of the timeline's WAL streaming connection.
-pub fn spawn_connection_manager_task(
-    timeline: Arc<Timeline>,
-    wal_connect_timeout: Duration,
-    lagging_wal_timeout: Duration,
-    max_lsn_wal_lag: NonZeroU64,
-    auth_token: Option<Arc<String>>,
-    availability_zone: Option<String>,
-    ctx: RequestContext,
-) {
-    let mut broker_client = get_broker_client().clone();
-
-    let tenant_id = timeline.tenant_id;
-    let timeline_id = timeline.timeline_id;
-
-    task_mgr::spawn(
-        WALRECEIVER_RUNTIME.handle(),
-        TaskKind::WalReceiverManager,
-        Some(tenant_id),
-        Some(timeline_id),
-        &format!("walreceiver for timeline {tenant_id}/{timeline_id}"),
-        false,
-        async move {
-            info!("WAL receiver manager started, connecting to broker");
-            let mut walreceiver_state = WalreceiverState::new(
-                timeline,
-                wal_connect_timeout,
-                lagging_wal_timeout,
-                max_lsn_wal_lag,
-                auth_token,
-                availability_zone,
-            );
-            loop {
-                select! {
-                    _ = task_mgr::shutdown_watcher() => {
-                        info!("WAL receiver shutdown requested, shutting down");
-                        walreceiver_state.shutdown().await;
-                        return Ok(());
-                    },
-                    loop_step_result = connection_manager_loop_step(
-                        &mut broker_client,
-                        &mut walreceiver_state,
-                        &ctx,
-                    ) => match loop_step_result {
-                        ControlFlow::Continue(()) => continue,
-                        ControlFlow::Break(()) => {
-                            info!("Connection manager loop ended, shutting down");
-                            walreceiver_state.shutdown().await;
-                            return Ok(());
-                        }
-                    },
-                }
-            }
-        }
-        .instrument(
-            info_span!(parent: None, "wal_connection_manager", tenant = %tenant_id, timeline = %timeline_id),
-        ),
-    );
-}
-
 /// Attempts to subscribe for timeline updates, pushed by safekeepers into the broker.
 /// Based on the updates, desides whether to start, keep or stop a WAL receiver task.
 /// If storage broker subscription is cancelled, exits.
-async fn connection_manager_loop_step(
+pub(super) async fn connection_manager_loop_step(
     broker_client: &mut BrokerClientChannel,
-    walreceiver_state: &mut WalreceiverState,
+    connection_manager_state: &mut ConnectionManagerState,
     ctx: &RequestContext,
 ) -> ControlFlow<(), ()> {
-    let mut timeline_state_updates = walreceiver_state.timeline.subscribe_for_state_updates();
+    let mut timeline_state_updates = connection_manager_state
+        .timeline
+        .subscribe_for_state_updates();
 
     match wait_for_active_timeline(&mut timeline_state_updates).await {
         ControlFlow::Continue(()) => {}
@@ -117,8 +57,8 @@ async fn connection_manager_loop_step(
     }
 
     let id = TenantTimelineId {
-        tenant_id: walreceiver_state.timeline.tenant_id,
-        timeline_id: walreceiver_state.timeline.timeline_id,
+        tenant_id: connection_manager_state.timeline.tenant_id,
+        timeline_id: connection_manager_state.timeline.timeline_id,
     };
 
     // Subscribe to the broker updates. Stream shares underlying TCP connection
@@ -128,7 +68,7 @@ async fn connection_manager_loop_step(
     info!("Subscribed for broker timeline updates");
 
     loop {
-        let time_until_next_retry = walreceiver_state.time_until_next_retry();
+        let time_until_next_retry = connection_manager_state.time_until_next_retry();
 
         // These things are happening concurrently:
         //
@@ -141,12 +81,12 @@ async fn connection_manager_loop_step(
         //  - timeline state changes to something that does not allow walreceiver to run concurrently
         select! {
             Some(wal_connection_update) = async {
-                match walreceiver_state.wal_connection.as_mut() {
+                match connection_manager_state.wal_connection.as_mut() {
                     Some(wal_connection) => Some(wal_connection.connection_task.next_task_event().await),
                     None => None,
                 }
             } => {
-                let wal_connection = walreceiver_state.wal_connection.as_mut()
+                let wal_connection = connection_manager_state.wal_connection.as_mut()
                     .expect("Should have a connection, as checked by the corresponding select! guard");
                 match wal_connection_update {
                     TaskEvent::Update(TaskStateUpdate::Started) => {},
@@ -156,7 +96,7 @@ async fn connection_manager_loop_step(
                             // from this safekeeper. This is good enough to clean unsuccessful
                             // retries history and allow reconnecting to this safekeeper without
                             // sleeping for a long time.
-                            walreceiver_state.wal_connection_retries.remove(&wal_connection.sk_id);
+                            connection_manager_state.wal_connection_retries.remove(&wal_connection.sk_id);
                         }
                         wal_connection.status = new_status;
                     }
@@ -165,7 +105,7 @@ async fn connection_manager_loop_step(
                             Ok(()) => debug!("WAL receiving task finished"),
                             Err(e) => error!("wal receiver task finished with an error: {e:?}"),
                         }
-                        walreceiver_state.drop_old_connection(false).await;
+                        connection_manager_state.drop_old_connection(false).await;
                     },
                 }
             },
@@ -173,7 +113,7 @@ async fn connection_manager_loop_step(
             // Got a new update from the broker
             broker_update = broker_subscription.message() => {
                 match broker_update {
-                    Ok(Some(broker_update)) => walreceiver_state.register_timeline_update(broker_update),
+                    Ok(Some(broker_update)) => connection_manager_state.register_timeline_update(broker_update),
                     Err(e) => {
                         error!("broker subscription failed: {e}");
                         return ControlFlow::Continue(());
@@ -187,12 +127,12 @@ async fn connection_manager_loop_step(
 
             new_event = async {
                 loop {
-                    if walreceiver_state.timeline.current_state() == TimelineState::Loading {
+                    if connection_manager_state.timeline.current_state() == TimelineState::Loading {
                         warn!("wal connection manager should only be launched after timeline has become active");
                     }
                     match timeline_state_updates.changed().await {
                         Ok(()) => {
-                            let new_state = walreceiver_state.timeline.current_state();
+                            let new_state = connection_manager_state.timeline.current_state();
                             match new_state {
                                 // we're already active as walreceiver, no need to reactivate
                                 TimelineState::Active => continue,
@@ -234,9 +174,9 @@ async fn connection_manager_loop_step(
             } => debug!("Waking up for the next retry after waiting for {time_until_next_retry:?}"),
         }
 
-        if let Some(new_candidate) = walreceiver_state.next_connection_candidate() {
+        if let Some(new_candidate) = connection_manager_state.next_connection_candidate() {
             info!("Switching to new connection candidate: {new_candidate:?}");
-            walreceiver_state
+            connection_manager_state
                 .change_connection(new_candidate, ctx)
                 .await
         }
@@ -314,25 +254,17 @@ const WALCONNECTION_RETRY_MAX_BACKOFF_SECONDS: f64 = 15.0;
 const WALCONNECTION_RETRY_BACKOFF_MULTIPLIER: f64 = 1.5;
 
 /// All data that's needed to run endless broker loop and keep the WAL streaming connection alive, if possible.
-struct WalreceiverState {
+pub(super) struct ConnectionManagerState {
     id: TenantTimelineId,
-
     /// Use pageserver data about the timeline to filter out some of the safekeepers.
     timeline: Arc<Timeline>,
-    /// The timeout on the connection to safekeeper for WAL streaming.
-    wal_connect_timeout: Duration,
-    /// The timeout to use to determine when the current connection is "stale" and reconnect to the other one.
-    lagging_wal_timeout: Duration,
-    /// The Lsn lag to use to determine when the current connection is lagging to much behind and reconnect to the other one.
-    max_lsn_wal_lag: NonZeroU64,
+    conf: WalReceiverConf,
     /// Current connection to safekeeper for WAL streaming.
     wal_connection: Option<WalConnection>,
     /// Info about retries and unsuccessful attempts to connect to safekeepers.
     wal_connection_retries: HashMap<NodeId, RetryInfo>,
     /// Data about all timelines, available for connection, fetched from storage broker, grouped by their corresponding safekeeper node id.
     wal_stream_candidates: HashMap<NodeId, BrokerSkTimeline>,
-    auth_token: Option<Arc<String>>,
-    availability_zone: Option<String>,
 }
 
 /// Current connection data.
@@ -375,15 +307,8 @@ struct BrokerSkTimeline {
     latest_update: NaiveDateTime,
 }
 
-impl WalreceiverState {
-    fn new(
-        timeline: Arc<Timeline>,
-        wal_connect_timeout: Duration,
-        lagging_wal_timeout: Duration,
-        max_lsn_wal_lag: NonZeroU64,
-        auth_token: Option<Arc<String>>,
-        availability_zone: Option<String>,
-    ) -> Self {
+impl ConnectionManagerState {
+    pub(super) fn new(timeline: Arc<Timeline>, conf: WalReceiverConf) -> Self {
         let id = TenantTimelineId {
             tenant_id: timeline.tenant_id,
             timeline_id: timeline.timeline_id,
@@ -391,14 +316,10 @@ impl WalreceiverState {
         Self {
             id,
             timeline,
-            wal_connect_timeout,
-            lagging_wal_timeout,
-            max_lsn_wal_lag,
+            conf,
             wal_connection: None,
             wal_stream_candidates: HashMap::new(),
             wal_connection_retries: HashMap::new(),
-            auth_token,
-            availability_zone,
         }
     }
 
@@ -407,7 +328,7 @@ impl WalreceiverState {
         self.drop_old_connection(true).await;
 
         let id = self.id;
-        let connect_timeout = self.wal_connect_timeout;
+        let connect_timeout = self.conf.wal_connect_timeout;
         let timeline = Arc::clone(&self.timeline);
         let ctx = ctx.detached_child(
             TaskKind::WalReceiverConnectionHandler,
@@ -427,7 +348,7 @@ impl WalreceiverState {
                 .context("walreceiver connection handling failure")
             }
             .instrument(
-                info_span!("walreceiver_connection", id = %id, node_id = %new_sk.safekeeper_id),
+                info_span!("walreceiver_connection", tenant_id = %id.tenant_id, timeline_id = %id.timeline_id, node_id = %new_sk.safekeeper_id),
             )
         });
 
@@ -563,7 +484,7 @@ impl WalreceiverState {
                     (now - existing_wal_connection.status.latest_connection_update).to_std()
                 {
                     // Drop connection if we haven't received keepalive message for a while.
-                    if latest_interaciton > self.wal_connect_timeout {
+                    if latest_interaciton > self.conf.wal_connect_timeout {
                         return Some(NewWalConnectionCandidate {
                             safekeeper_id: new_sk_id,
                             wal_source_connconf: new_wal_source_connconf,
@@ -573,7 +494,7 @@ impl WalreceiverState {
                                     existing_wal_connection.status.latest_connection_update,
                                 ),
                                 check_time: now,
-                                threshold: self.wal_connect_timeout,
+                                threshold: self.conf.wal_connect_timeout,
                             },
                         });
                     }
@@ -589,7 +510,7 @@ impl WalreceiverState {
                     // Check if the new candidate has much more WAL than the current one.
                     match new_commit_lsn.0.checked_sub(current_commit_lsn.0) {
                         Some(new_sk_lsn_advantage) => {
-                            if new_sk_lsn_advantage >= self.max_lsn_wal_lag.get() {
+                            if new_sk_lsn_advantage >= self.conf.max_lsn_wal_lag.get() {
                                 return Some(NewWalConnectionCandidate {
                                     safekeeper_id: new_sk_id,
                                     wal_source_connconf: new_wal_source_connconf,
@@ -597,16 +518,16 @@ impl WalreceiverState {
                                     reason: ReconnectReason::LaggingWal {
                                         current_commit_lsn,
                                         new_commit_lsn,
-                                        threshold: self.max_lsn_wal_lag,
+                                        threshold: self.conf.max_lsn_wal_lag,
                                     },
                                 });
                             }
                             // If we have a candidate with the same commit_lsn as the current one, which is in the same AZ as pageserver,
                             // and the current one is not, switch to the new one.
-                            if self.availability_zone.is_some()
+                            if self.conf.availability_zone.is_some()
                                 && existing_wal_connection.availability_zone
-                                    != self.availability_zone
-                                && self.availability_zone == new_availability_zone
+                                    != self.conf.availability_zone
+                                && self.conf.availability_zone == new_availability_zone
                             {
                                 return Some(NewWalConnectionCandidate {
                                     safekeeper_id: new_sk_id,
@@ -677,7 +598,7 @@ impl WalreceiverState {
                 if let Some(waiting_for_new_lsn_since) = waiting_for_new_lsn_since {
                     if let Ok(waiting_for_new_wal) = (now - waiting_for_new_lsn_since).to_std() {
                         if candidate_commit_lsn > current_commit_lsn
-                            && waiting_for_new_wal > self.lagging_wal_timeout
+                            && waiting_for_new_wal > self.conf.lagging_wal_timeout
                         {
                             return Some(NewWalConnectionCandidate {
                                 safekeeper_id: new_sk_id,
@@ -691,7 +612,7 @@ impl WalreceiverState {
                                         existing_wal_connection.status.latest_wal_update,
                                     ),
                                     check_time: now,
-                                    threshold: self.lagging_wal_timeout,
+                                    threshold: self.conf.lagging_wal_timeout,
                                 },
                             });
                         }
@@ -757,11 +678,11 @@ impl WalreceiverState {
                 match wal_stream_connection_config(
                     self.id,
                     info.safekeeper_connstr.as_ref(),
-                    match &self.auth_token {
+                    match &self.conf.auth_token {
                         None => None,
                         Some(x) => Some(x),
                     },
-                    self.availability_zone.as_deref(),
+                    self.conf.availability_zone.as_deref(),
                 ) {
                     Ok(connstr) => Some((*sk_id, info, connstr)),
                     Err(e) => {
@@ -775,7 +696,7 @@ impl WalreceiverState {
     /// Remove candidates which haven't sent broker updates for a while.
     fn cleanup_old_candidates(&mut self) {
         let mut node_ids_to_remove = Vec::with_capacity(self.wal_stream_candidates.len());
-        let lagging_wal_timeout = self.lagging_wal_timeout;
+        let lagging_wal_timeout = self.conf.lagging_wal_timeout;
 
         self.wal_stream_candidates.retain(|node_id, broker_info| {
             if let Ok(time_since_latest_broker_update) =
@@ -799,7 +720,7 @@ impl WalreceiverState {
         }
     }
 
-    async fn shutdown(mut self) {
+    pub(super) async fn shutdown(mut self) {
         if let Some(wal_connection) = self.wal_connection.take() {
             wal_connection.connection_task.shutdown().await;
         }
@@ -903,7 +824,7 @@ mod tests {
         let mut state = dummy_state(&harness).await;
         let now = Utc::now().naive_utc();
 
-        let lagging_wal_timeout = chrono::Duration::from_std(state.lagging_wal_timeout)?;
+        let lagging_wal_timeout = chrono::Duration::from_std(state.conf.lagging_wal_timeout)?;
         let delay_over_threshold = now - lagging_wal_timeout - lagging_wal_timeout;
 
         state.wal_connection = None;
@@ -914,7 +835,7 @@ mod tests {
             (
                 NodeId(3),
                 dummy_broker_sk_timeline(
-                    1 + state.max_lsn_wal_lag.get(),
+                    1 + state.conf.max_lsn_wal_lag.get(),
                     "delay_over_threshold",
                     delay_over_threshold,
                 ),
@@ -948,7 +869,7 @@ mod tests {
             streaming_lsn: Some(Lsn(current_lsn)),
         };
 
-        state.max_lsn_wal_lag = NonZeroU64::new(100).unwrap();
+        state.conf.max_lsn_wal_lag = NonZeroU64::new(100).unwrap();
         state.wal_connection = Some(WalConnection {
             started_at: now,
             sk_id: connected_sk_id,
@@ -966,7 +887,7 @@ mod tests {
             (
                 connected_sk_id,
                 dummy_broker_sk_timeline(
-                    current_lsn + state.max_lsn_wal_lag.get() * 2,
+                    current_lsn + state.conf.max_lsn_wal_lag.get() * 2,
                     DUMMY_SAFEKEEPER_HOST,
                     now,
                 ),
@@ -978,7 +899,7 @@ mod tests {
             (
                 NodeId(2),
                 dummy_broker_sk_timeline(
-                    current_lsn + state.max_lsn_wal_lag.get() / 2,
+                    current_lsn + state.conf.max_lsn_wal_lag.get() / 2,
                     "not_enough_advanced_lsn",
                     now,
                 ),
@@ -1003,7 +924,11 @@ mod tests {
         state.wal_connection = None;
         state.wal_stream_candidates = HashMap::from([(
             NodeId(0),
-            dummy_broker_sk_timeline(1 + state.max_lsn_wal_lag.get(), DUMMY_SAFEKEEPER_HOST, now),
+            dummy_broker_sk_timeline(
+                1 + state.conf.max_lsn_wal_lag.get(),
+                DUMMY_SAFEKEEPER_HOST,
+                now,
+            ),
         )]);
 
         let only_candidate = state
@@ -1101,7 +1026,7 @@ mod tests {
         let now = Utc::now().naive_utc();
 
         let connected_sk_id = NodeId(0);
-        let new_lsn = Lsn(current_lsn.0 + state.max_lsn_wal_lag.get() + 1);
+        let new_lsn = Lsn(current_lsn.0 + state.conf.max_lsn_wal_lag.get() + 1);
 
         let connection_status = WalConnectionStatus {
             is_connected: true,
@@ -1146,7 +1071,7 @@ mod tests {
             ReconnectReason::LaggingWal {
                 current_commit_lsn: current_lsn,
                 new_commit_lsn: new_lsn,
-                threshold: state.max_lsn_wal_lag
+                threshold: state.conf.max_lsn_wal_lag
             },
             "Should select bigger WAL safekeeper if it starts to lag enough"
         );
@@ -1165,7 +1090,7 @@ mod tests {
         let current_lsn = Lsn(100_000).align();
         let now = Utc::now().naive_utc();
 
-        let wal_connect_timeout = chrono::Duration::from_std(state.wal_connect_timeout)?;
+        let wal_connect_timeout = chrono::Duration::from_std(state.conf.wal_connect_timeout)?;
         let time_over_threshold =
             Utc::now().naive_utc() - wal_connect_timeout - wal_connect_timeout;
 
@@ -1208,7 +1133,7 @@ mod tests {
                 ..
             } => {
                 assert_eq!(last_keep_alive, Some(time_over_threshold));
-                assert_eq!(threshold, state.lagging_wal_timeout);
+                assert_eq!(threshold, state.conf.lagging_wal_timeout);
             }
             unexpected => panic!("Unexpected reason: {unexpected:?}"),
         }
@@ -1228,7 +1153,7 @@ mod tests {
         let new_lsn = Lsn(100_100).align();
         let now = Utc::now().naive_utc();
 
-        let lagging_wal_timeout = chrono::Duration::from_std(state.lagging_wal_timeout)?;
+        let lagging_wal_timeout = chrono::Duration::from_std(state.conf.lagging_wal_timeout)?;
         let time_over_threshold =
             Utc::now().naive_utc() - lagging_wal_timeout - lagging_wal_timeout;
 
@@ -1275,7 +1200,7 @@ mod tests {
                 assert_eq!(current_commit_lsn, current_lsn);
                 assert_eq!(candidate_commit_lsn, new_lsn);
                 assert_eq!(last_wal_interaction, Some(time_over_threshold));
-                assert_eq!(threshold, state.lagging_wal_timeout);
+                assert_eq!(threshold, state.conf.lagging_wal_timeout);
             }
             unexpected => panic!("Unexpected reason: {unexpected:?}"),
         }
@@ -1289,27 +1214,29 @@ mod tests {
 
     const DUMMY_SAFEKEEPER_HOST: &str = "safekeeper_connstr";
 
-    async fn dummy_state(harness: &TenantHarness<'_>) -> WalreceiverState {
+    async fn dummy_state(harness: &TenantHarness<'_>) -> ConnectionManagerState {
         let (tenant, ctx) = harness.load().await;
         let timeline = tenant
             .create_empty_timeline(TIMELINE_ID, Lsn(0), crate::DEFAULT_PG_VERSION, &ctx)
             .expect("Failed to create an empty timeline for dummy wal connection manager");
         let timeline = timeline.initialize(&ctx).unwrap();
 
-        WalreceiverState {
+        ConnectionManagerState {
             id: TenantTimelineId {
                 tenant_id: harness.tenant_id,
                 timeline_id: TIMELINE_ID,
             },
             timeline,
-            wal_connect_timeout: Duration::from_secs(1),
-            lagging_wal_timeout: Duration::from_secs(1),
-            max_lsn_wal_lag: NonZeroU64::new(1024 * 1024).unwrap(),
+            conf: WalReceiverConf {
+                wal_connect_timeout: Duration::from_secs(1),
+                lagging_wal_timeout: Duration::from_secs(1),
+                max_lsn_wal_lag: NonZeroU64::new(1024 * 1024).unwrap(),
+                auth_token: None,
+                availability_zone: None,
+            },
             wal_connection: None,
             wal_stream_candidates: HashMap::new(),
             wal_connection_retries: HashMap::new(),
-            auth_token: None,
-            availability_zone: None,
         }
     }
 
@@ -1321,7 +1248,7 @@ mod tests {
 
         let harness = TenantHarness::create("switch_to_same_availability_zone")?;
         let mut state = dummy_state(&harness).await;
-        state.availability_zone = test_az.clone();
+        state.conf.availability_zone = test_az.clone();
         let current_lsn = Lsn(100_000).align();
         let now = Utc::now().naive_utc();
 
diff --git a/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs b/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs
index ea2f2392ea..d80c7c5673 100644
--- a/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs
+++ b/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs
@@ -37,12 +37,12 @@ use crate::{
 use postgres_backend::is_expected_io_error;
 use postgres_connection::PgConnectionConfig;
 use postgres_ffi::waldecoder::WalStreamDecoder;
-use pq_proto::PageserverFeedback;
 use utils::lsn::Lsn;
+use utils::pageserver_feedback::PageserverFeedback;
 
 /// Status of the connection.
 #[derive(Debug, Clone, Copy)]
-pub struct WalConnectionStatus {
+pub(super) struct WalConnectionStatus {
     /// If we were able to initiate a postgres connection, this means that safekeeper process is at least running.
     pub is_connected: bool,
     /// Defines a healthy connection as one on which pageserver received WAL from safekeeper
@@ -60,7 +60,7 @@ pub struct WalConnectionStatus {
 
 /// Open a connection to the given safekeeper and receive WAL, sending back progress
 /// messages as we go.
-pub async fn handle_walreceiver_connection(
+pub(super) async fn handle_walreceiver_connection(
     timeline: Arc<Timeline>,
     wal_source_connconf: PgConnectionConfig,
     events_sender: watch::Sender<TaskStateUpdate<WalConnectionStatus>>,
@@ -319,12 +319,12 @@ pub async fn handle_walreceiver_connection(
                 timeline.get_remote_consistent_lsn().unwrap_or(Lsn(0));
 
             // The last LSN we processed. It is not guaranteed to survive pageserver crash.
-            let last_received_lsn = u64::from(last_lsn);
+            let last_received_lsn = last_lsn;
             // `disk_consistent_lsn` is the LSN at which page server guarantees local persistence of all received data
-            let disk_consistent_lsn = u64::from(timeline.get_disk_consistent_lsn());
+            let disk_consistent_lsn = timeline.get_disk_consistent_lsn();
             // The last LSN that is synced to remote storage and is guaranteed to survive pageserver crash
             // Used by safekeepers to remove WAL preceding `remote_consistent_lsn`.
-            let remote_consistent_lsn = u64::from(timeline_remote_consistent_lsn);
+            let remote_consistent_lsn = timeline_remote_consistent_lsn;
             let ts = SystemTime::now();
 
             // Update the status about what we just received. This is shown in the mgmt API.
diff --git a/pgxn/neon/file_cache.c b/pgxn/neon/file_cache.c
index 8dff259f02..cc46fb5a25 100644
--- a/pgxn/neon/file_cache.c
+++ b/pgxn/neon/file_cache.c
@@ -370,6 +370,74 @@ lfc_cache_contains(RelFileNode rnode, ForkNumber forkNum, BlockNumber blkno)
 	return found;
 }
 
+/*
+ * Evict a page (if present) from the local file cache
+ */
+void
+lfc_evict(RelFileNode rnode, ForkNumber forkNum, BlockNumber blkno)
+{
+	BufferTag tag;
+	FileCacheEntry* entry;
+	ssize_t rc;
+	bool found;
+	int chunk_offs = blkno & (BLOCKS_PER_CHUNK-1);
+	uint32 hash;
+
+	if (lfc_size_limit == 0) /* fast exit if file cache is disabled */
+		return;
+
+	INIT_BUFFERTAG(tag, rnode, forkNum, (blkno & ~(BLOCKS_PER_CHUNK-1)));
+
+	hash = get_hash_value(lfc_hash, &tag);
+
+	LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
+	entry = hash_search_with_hash_value(lfc_hash, &tag, hash, HASH_FIND, &found);
+
+	if (!found)
+	{
+		/* nothing to do */
+		LWLockRelease(lfc_lock);
+		return;
+	}
+
+	/* remove the page from the cache */
+	entry->bitmap[chunk_offs >> 5] &= ~(1 << (chunk_offs & (32 - 1)));
+
+	/*
+	 * If the chunk has no live entries, we can position the chunk to be
+	 * recycled first.
+	 */
+	if (entry->bitmap[chunk_offs >> 5] == 0)
+	{
+		bool has_remaining_pages;
+
+		for (int i = 0; i < (BLOCKS_PER_CHUNK / 32); i++) {
+			if (entry->bitmap[i] != 0)
+			{
+				has_remaining_pages = true;
+				break;
+			}
+		}
+
+		/*
+		 * Put the entry at the position that is first to be reclaimed when
+		 * we have no cached pages remaining in the chunk
+		 */
+		if (!has_remaining_pages)
+		{
+			dlist_delete(&entry->lru_node);
+			dlist_push_head(&lfc_ctl->lru, &entry->lru_node);
+		}
+	}
+
+	/*
+	 * Done: apart from empty chunks, we don't move chunks in the LRU when
+	 * they're empty because eviction isn't usage.
+	 */
+
+	LWLockRelease(lfc_lock);
+}
+
 /*
  * Try to read page from local cache.
  * Returns true if page is found in local cache.
@@ -528,7 +596,6 @@ lfc_write(RelFileNode rnode, ForkNumber forkNum, BlockNumber blkno,
 	LWLockRelease(lfc_lock);
 }
 
-
 /*
  * Record structure holding the to be exposed cache data.
  */
diff --git a/pgxn/neon/libpagestore.c b/pgxn/neon/libpagestore.c
index c44e8fcda5..21330c018f 100644
--- a/pgxn/neon/libpagestore.c
+++ b/pgxn/neon/libpagestore.c
@@ -17,6 +17,8 @@
 #include "pagestore_client.h"
 #include "fmgr.h"
 #include "access/xlog.h"
+#include "access/xlogutils.h"
+#include "storage/buf_internals.h"
 
 #include "libpq-fe.h"
 #include "libpq/pqformat.h"
@@ -57,6 +59,8 @@ int			n_unflushed_requests = 0;
 int			flush_every_n_requests = 8;
 int			readahead_buffer_size = 128;
 
+bool	(*old_redo_read_buffer_filter) (XLogReaderState *record, uint8 block_id) = NULL;
+
 static void pageserver_flush(void);
 
 static bool
@@ -467,6 +471,8 @@ pg_init_libpagestore(void)
 		smgr_hook = smgr_neon;
 		smgr_init_hook = smgr_init_neon;
 		dbsize_hook = neon_dbsize;
+		old_redo_read_buffer_filter = redo_read_buffer_filter;
+		redo_read_buffer_filter = neon_redo_read_buffer_filter;
 	}
 	lfc_init();
 }
diff --git a/pgxn/neon/neon.c b/pgxn/neon/neon.c
index 5c98902554..217c1974a0 100644
--- a/pgxn/neon/neon.c
+++ b/pgxn/neon/neon.c
@@ -24,6 +24,7 @@
 
 #include "neon.h"
 #include "walproposer.h"
+#include "pagestore_client.h"
 
 PG_MODULE_MAGIC;
 void		_PG_init(void);
diff --git a/pgxn/neon/neon.h b/pgxn/neon/neon.h
index 3eac8f4570..60d321a945 100644
--- a/pgxn/neon/neon.h
+++ b/pgxn/neon/neon.h
@@ -11,6 +11,7 @@
 
 #ifndef NEON_H
 #define NEON_H
+#include "access/xlogreader.h"
 
 /* GUCs */
 extern char *neon_auth_token;
@@ -20,4 +21,11 @@ extern char *neon_tenant;
 extern void pg_init_libpagestore(void);
 extern void pg_init_walproposer(void);
 
+/*
+ * Returns true if we shouldn't do REDO on that block in record indicated by
+ * block_id; false otherwise.
+ */
+extern bool neon_redo_read_buffer_filter(XLogReaderState *record, uint8 block_id);
+extern bool	(*old_redo_read_buffer_filter) (XLogReaderState *record, uint8 block_id);
+
 #endif							/* NEON_H */
diff --git a/pgxn/neon/pagestore_client.h b/pgxn/neon/pagestore_client.h
index a1f05ac685..22f5cdb73a 100644
--- a/pgxn/neon/pagestore_client.h
+++ b/pgxn/neon/pagestore_client.h
@@ -207,6 +207,7 @@ extern void forget_cached_relsize(RelFileNode rnode, ForkNumber forknum);
 extern void lfc_write(RelFileNode rnode, ForkNumber forkNum, BlockNumber blkno, char *buffer);
 extern bool lfc_read(RelFileNode rnode, ForkNumber forkNum, BlockNumber blkno, char *buffer);
 extern bool lfc_cache_contains(RelFileNode rnode, ForkNumber forkNum, BlockNumber blkno);
+extern void lfc_evict(RelFileNode rnode, ForkNumber forkNum, BlockNumber blkno);
 extern void lfc_init(void);
 
 
diff --git a/pgxn/neon/pagestore_smgr.c b/pgxn/neon/pagestore_smgr.c
index 5b30641856..528d4eb051 100644
--- a/pgxn/neon/pagestore_smgr.c
+++ b/pgxn/neon/pagestore_smgr.c
@@ -189,6 +189,7 @@ typedef struct PrfHashEntry {
 #define SH_DEFINE
 #define SH_DECLARE
 #include "lib/simplehash.h"
+#include "neon.h"
 
 /*
  * PrefetchState maintains the state of (prefetch) getPage@LSN requests.
@@ -1209,6 +1210,9 @@ neon_wallog_page(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum, ch
 
 	if (ShutdownRequestPending)
 		return;
+	/* Don't log any pages if we're not allowed to do so. */
+	if (!XLogInsertAllowed())
+		return;
 
 	/*
 	 * Whenever a VM or FSM page is evicted, WAL-log it. FSM and (some) VM
@@ -1375,8 +1379,18 @@ neon_get_request_lsn(bool *latest, RelFileNode rnode, ForkNumber forknum, BlockN
 
 	if (RecoveryInProgress())
 	{
+		/*
+		 * We don't know if WAL has been generated but not yet replayed, so
+		 * we're conservative in our estimates about latest pages.
+		 */
 		*latest = false;
-		lsn = GetXLogReplayRecPtr(NULL);
+
+		/*
+		 * Get the last written LSN of this page.
+		 */
+		lsn = GetLastWrittenLSN(rnode, forknum, blkno);
+		lsn = nm_adjust_lsn(lsn);
+
 		elog(DEBUG1, "neon_get_request_lsn GetXLogReplayRecPtr %X/%X request lsn 0 ",
 			 (uint32) ((lsn) >> 32), (uint32) (lsn));
 	}
@@ -1559,6 +1573,15 @@ neon_create(SMgrRelation reln, ForkNumber forkNum, bool isRedo)
 	/*
 	 * Newly created relation is empty, remember that in the relsize cache.
 	 *
+	 * Note that in REDO, this is called to make sure the relation fork exists,
+	 * but it does not truncate the relation. So, we can only update the
+	 * relsize if it didn't exist before.
+	 * 
+	 * Also, in redo, we must make sure to update the cached size of the
+	 * relation, as that is the primary source of truth for REDO's
+	 * file length considerations, and as file extension isn't (perfectly)
+	 * logged, we need to take care of that before we hit file size checks.
+	 *
 	 * FIXME: This is currently not just an optimization, but required for
 	 * correctness. Postgres can call smgrnblocks() on the newly-created
 	 * relation. Currently, we don't call SetLastWrittenLSN() when a new
@@ -1566,7 +1589,14 @@ neon_create(SMgrRelation reln, ForkNumber forkNum, bool isRedo)
 	 * cache, we might call smgrnblocks() on the newly-created relation before
 	 * the creation WAL record hass been received by the page server.
 	 */
-	set_cached_relsize(reln->smgr_rnode.node, forkNum, 0);
+	if (isRedo)
+	{
+		update_cached_relsize(reln->smgr_rnode.node, forkNum, 0);
+		get_cached_relsize(reln->smgr_rnode.node, forkNum,
+						   &reln->smgr_cached_nblocks[forkNum]);
+	}
+	else
+		set_cached_relsize(reln->smgr_rnode.node, forkNum, 0);
 
 #ifdef DEBUG_COMPARE_LOCAL
 	if (IS_LOCAL_REL(reln))
@@ -1831,6 +1861,26 @@ neon_read_at_lsn(RelFileNode rnode, ForkNumber forkNum, BlockNumber blkno,
 		.blockNum = blkno,
 	};
 
+	/*
+	 * The redo process does not lock pages that it needs to replay but are
+	 * not in the shared buffers, so a concurrent process may request the
+	 * page after redo has decided it won't redo that page and updated the
+	 * LwLSN for that page.
+	 * If we're in hot standby we need to take care that we don't return
+	 * until after REDO has finished replaying up to that LwLSN, as the page
+	 * should have been locked up to that point.
+	 *
+	 * See also the description on neon_redo_read_buffer_filter below.
+	 *
+	 * NOTE: It is possible that the WAL redo process will still do IO due to
+	 * concurrent failed read IOs. Those IOs should never have a request_lsn
+	 * that is as large as the WAL record we're currently replaying, if it
+	 * weren't for the behaviour of the LwLsn cache that uses the highest
+	 * value of the LwLsn cache when the entry is not found. 
+	 */
+	if (RecoveryInProgress() && !(MyBackendType == B_STARTUP))
+		XLogWaitForReplayOf(request_lsn);
+
 	/*
 	 * Try to find prefetched page in the list of received pages.
 	 */
@@ -2584,3 +2634,143 @@ smgr_init_neon(void)
 	smgr_init_standard();
 	neon_init();
 }
+
+
+/*
+ * Return whether we can skip the redo for this block.
+ * 
+ * The conditions for skipping the IO are:
+ *
+ * - The block is not in the shared buffers, and
+ * - The block is not in the local file cache
+ *
+ * ... because any subsequent read of the page requires us to read
+ * the new version of the page from the PageServer. We do not
+ * check the local file cache; we instead evict the page from LFC: it
+ * is cheaper than going through the FS calls to read the page, and
+ * limits the number of lock operations used in the REDO process.
+ *
+ * We have one exception to the rules for skipping IO: We always apply
+ * changes to shared catalogs' pages. Although this is mostly out of caution,
+ * catalog updates usually result in backends rebuilding their catalog snapshot,
+ * which means it's quite likely the modified page is going to be used soon.
+ *
+ * It is important to note that skipping WAL redo for a page also means
+ * the page isn't locked by the redo process, as there is no Buffer
+ * being returned, nor is there a buffer descriptor to lock.
+ * This means that any IO that wants to read this block needs to wait
+ * for the WAL REDO process to finish processing the WAL record before
+ * it allows the system to start reading the block, as releasing the
+ * block early could lead to phantom reads.
+ *
+ * For example, REDO for a WAL record that modifies 3 blocks could skip
+ * the first block, wait for a lock on the second, and then modify the
+ * third block. Without skipping, all blocks would be locked and phantom
+ * reads would not occur, but with skipping, a concurrent process could
+ * read block 1 with post-REDO contents and read block 3 with pre-REDO
+ * contents, where with REDO locking it would wait on block 1 and see
+ * block 3 with post-REDO contents only.
+ */
+bool
+neon_redo_read_buffer_filter(XLogReaderState *record, uint8 block_id)
+{
+	XLogRecPtr	end_recptr = record->EndRecPtr;
+	XLogRecPtr	prev_end_recptr = record->ReadRecPtr - 1;
+	RelFileNode	rnode;
+	ForkNumber	forknum;
+	BlockNumber	blkno;
+	BufferTag	tag;
+	uint32		hash;
+	LWLock	   *partitionLock;
+	Buffer		buffer;
+	bool		no_redo_needed;
+	BlockNumber relsize;
+
+	if (old_redo_read_buffer_filter && old_redo_read_buffer_filter(record, block_id))
+		return true;
+
+#if PG_VERSION_NUM < 150000
+	if (!XLogRecGetBlockTag(record, block_id, &rnode, &forknum, &blkno))
+		elog(PANIC, "failed to locate backup block with ID %d", block_id);
+#else
+	XLogRecGetBlockTag(record, block_id, &rnode, &forknum, &blkno);
+#endif
+
+	/*
+	 * Out of an abundance of caution, we always run redo on shared catalogs,
+	 * regardless of whether the block is stored in shared buffers.
+	 * See also this function's top comment.
+	 */
+	if (!OidIsValid(rnode.dbNode))
+		return false;
+
+	INIT_BUFFERTAG(tag, rnode, forknum, blkno);
+	hash = BufTableHashCode(&tag);
+	partitionLock = BufMappingPartitionLock(hash);
+
+	/*
+	 * Lock the partition of shared_buffers so that it can't be updated
+	 * concurrently.
+	 */
+	LWLockAcquire(partitionLock, LW_SHARED);
+
+	/* Try to find the relevant buffer */
+	buffer = BufTableLookup(&tag, hash);
+
+	no_redo_needed = buffer < 0;
+
+	/* we don't have the buffer in memory, update lwLsn past this record */
+	if (no_redo_needed)
+	{
+		SetLastWrittenLSNForBlock(end_recptr, rnode, forknum, blkno);
+		lfc_evict(rnode, forknum, blkno);
+	}
+	else
+	{
+		SetLastWrittenLSNForBlock(prev_end_recptr, rnode, forknum, blkno);
+	}
+
+	LWLockRelease(partitionLock);
+
+	/* Extend the relation if we know its size */
+	if (get_cached_relsize(rnode, forknum, &relsize))
+	{
+		if (relsize < blkno + 1)
+			update_cached_relsize(rnode, forknum, blkno + 1);
+	}
+	else
+	{
+		/*
+		 * Size was not cached. We populate the cache now, with the size of the
+		 * relation measured after this WAL record is applied.
+		 *
+		 * This length is later reused when we open the smgr to read the block,
+		 * which is fine and expected.
+		 */
+
+		NeonResponse *response;
+		NeonNblocksResponse *nbresponse;
+		NeonNblocksRequest request = {
+			.req = (NeonRequest) {
+				.lsn = end_recptr,
+				.latest = false,
+				.tag = T_NeonNblocksRequest,
+			},
+			.rnode = rnode,
+			.forknum = forknum,
+		};
+
+		response = page_server_request(&request);
+
+		Assert(response->tag == T_NeonNblocksResponse);
+		nbresponse = (NeonNblocksResponse *) response;
+
+		Assert(nbresponse->n_blocks > blkno);
+
+		set_cached_relsize(rnode, forknum, nbresponse->n_blocks);
+
+		elog(SmgrTrace, "Set length to %d", nbresponse->n_blocks);
+	}
+
+	return no_redo_needed;
+}
diff --git a/pgxn/neon/walproposer.c b/pgxn/neon/walproposer.c
index 45037a8c01..a99be40955 100644
--- a/pgxn/neon/walproposer.c
+++ b/pgxn/neon/walproposer.c
@@ -1964,18 +1964,26 @@ CombineHotStanbyFeedbacks(HotStandbyFeedback * hs)
 	{
 		if (safekeeper[i].appendResponse.hs.ts != 0)
 		{
-			if (FullTransactionIdPrecedes(safekeeper[i].appendResponse.hs.xmin, hs->xmin))
+			HotStandbyFeedback *skhs = &safekeeper[i].appendResponse.hs;
+			if (FullTransactionIdIsNormal(skhs->xmin)
+				&& FullTransactionIdPrecedes(skhs->xmin, hs->xmin))
 			{
-				hs->xmin = safekeeper[i].appendResponse.hs.xmin;
-				hs->ts = safekeeper[i].appendResponse.hs.ts;
+				hs->xmin = skhs->xmin;
+				hs->ts = skhs->ts;
 			}
-			if (FullTransactionIdPrecedes(safekeeper[i].appendResponse.hs.catalog_xmin, hs->catalog_xmin))
+			if (FullTransactionIdIsNormal(skhs->catalog_xmin)
+				&& FullTransactionIdPrecedes(skhs->catalog_xmin, hs->xmin))
 			{
-				hs->catalog_xmin = safekeeper[i].appendResponse.hs.catalog_xmin;
-				hs->ts = safekeeper[i].appendResponse.hs.ts;
+				hs->catalog_xmin = skhs->catalog_xmin;
+				hs->ts = skhs->ts;
 			}
 		}
 	}
+
+	if (hs->xmin.value == ~0)
+		hs->xmin = InvalidFullTransactionId;
+	if (hs->catalog_xmin.value == ~0)
+		hs->catalog_xmin = InvalidFullTransactionId;
 }
 
 /*
diff --git a/pgxn/neon_walredo/seccomp.c b/pgxn/neon_walredo/seccomp.c
index 5d5ba549ef..1e8f6682a2 100644
--- a/pgxn/neon_walredo/seccomp.c
+++ b/pgxn/neon_walredo/seccomp.c
@@ -9,6 +9,14 @@
  * To prevent this, it has been decided to limit possible interactions
  * with the outside world using the Secure Computing BPF mode.
  *
+ * This code is intended to support both x86_64 and aarch64. The latter
+ * doesn't implement some syscalls like open and select. We allow both
+ * select (absent on aarch64) and pselect6 (present on both architectures)
+ * We call select(2) through libc, and the libc wrapper calls select or pselect6
+ * depending on the architecture. You can check which syscalls are present on
+ * different architectures with the `scmp_sys_resolver` tool from the
+ * seccomp package.
+ *
  * We use this mode to disable all syscalls not in the allowlist. This
  * approach has its pros & cons:
  *
@@ -73,8 +81,6 @@
  *    I suspect that certain libc functions might involve slightly
  *    different syscalls, e.g. select/pselect6/pselect6_time64/whatever.
  *
- *  - Test on any arch other than amd64 to see if it works there.
- *
  *-------------------------------------------------------------------------
  */
 
@@ -122,9 +128,10 @@ seccomp_load_rules(PgSeccompRule *rules, int count)
 
 	/*
 	 * First, check that open of a well-known file works.
-	 * XXX: We use raw syscall() to call the very open().
+	 * XXX: We use raw syscall() to call the very openat() which is
+	 * present both on x86_64 and on aarch64.
 	 */
-	fd = syscall(SCMP_SYS(open), "/dev/null", O_RDONLY, 0);
+	fd = syscall(SCMP_SYS(openat), AT_FDCWD, "/dev/null", O_RDONLY, 0);
 	if (seccomp_test_sighandler_done)
 		ereport(FATAL,
 				(errcode(ERRCODE_SYSTEM_ERROR),
@@ -135,15 +142,15 @@ seccomp_load_rules(PgSeccompRule *rules, int count)
 				 errmsg("seccomp: could not open /dev/null for seccomp testing: %m")));
 	close((int) fd);
 
-	/* Set a trap on open() to test seccomp bpf */
-	rule = PG_SCMP(open, SCMP_ACT_TRAP);
+	/* Set a trap on openat() to test seccomp bpf */
+	rule = PG_SCMP(openat, SCMP_ACT_TRAP);
 	if (do_seccomp_load_rules(&rule, 1, SCMP_ACT_ALLOW) != 0)
 		ereport(FATAL,
 				(errcode(ERRCODE_SYSTEM_ERROR),
 				 errmsg("seccomp: could not load test trap")));
 
-	/* Finally, check that open() now raises SIGSYS */
-	(void) syscall(SCMP_SYS(open), "/dev/null", O_RDONLY, 0);
+	/* Finally, check that openat() now raises SIGSYS */
+	(void) syscall(SCMP_SYS(openat), AT_FDCWD, "/dev/null", O_RDONLY, 0);
 	if (!seccomp_test_sighandler_done)
 		ereport(FATAL,
 				(errcode(ERRCODE_SYSTEM_ERROR),
@@ -224,7 +231,7 @@ seccomp_test_sighandler(int signum, siginfo_t *info, void *cxt pg_attribute_unus
 		die(1, DIE_PREFIX "bad signal number\n");
 
 	/* TODO: maybe somehow extract the hardcoded syscall number */
-	if (info->si_syscall != SCMP_SYS(open))
+	if (info->si_syscall != SCMP_SYS(openat))
 		die(1, DIE_PREFIX "bad syscall number\n");
 
 #undef DIE_PREFIX
diff --git a/proxy/Cargo.toml b/proxy/Cargo.toml
index add8b14c95..9d702b29c3 100644
--- a/proxy/Cargo.toml
+++ b/proxy/Cargo.toml
@@ -64,6 +64,7 @@ webpki-roots.workspace = true
 x509-parser.workspace = true
 
 workspace_hack.workspace = true
+tokio-util.workspace = true
 
 [dev-dependencies]
 rcgen.workspace = true
diff --git a/proxy/src/compute.rs b/proxy/src/compute.rs
index b5efc72803..0465703ae6 100644
--- a/proxy/src/compute.rs
+++ b/proxy/src/compute.rs
@@ -1,8 +1,8 @@
 use crate::{cancellation::CancelClosure, error::UserFacingError};
-use futures::TryFutureExt;
+use futures::{FutureExt, TryFutureExt};
 use itertools::Itertools;
 use pq_proto::StartupMessageParams;
-use std::{io, net::SocketAddr};
+use std::{io, net::SocketAddr, time::Duration};
 use thiserror::Error;
 use tokio::net::TcpStream;
 use tokio_postgres::NoTls;
@@ -130,9 +130,23 @@ impl ConnCfg {
     async fn connect_raw(&self) -> io::Result<(SocketAddr, TcpStream)> {
         use tokio_postgres::config::Host;
 
+        // wrap TcpStream::connect with timeout
+        let connect_with_timeout = |host, port| {
+            let connection_timeout = Duration::from_millis(10000);
+            tokio::time::timeout(connection_timeout, TcpStream::connect((host, port))).map(
+                move |res| match res {
+                    Ok(tcpstream_connect_res) => tcpstream_connect_res,
+                    Err(_) => Err(io::Error::new(
+                        io::ErrorKind::TimedOut,
+                        format!("exceeded connection timeout {connection_timeout:?}"),
+                    )),
+                },
+            )
+        };
+
         let connect_once = |host, port| {
             info!("trying to connect to compute node at {host}:{port}");
-            TcpStream::connect((host, port)).and_then(|socket| async {
+            connect_with_timeout(host, port).and_then(|socket| async {
                 let socket_addr = socket.peer_addr()?;
                 // This prevents load balancer from severing the connection.
                 socket2::SockRef::from(&socket).set_keepalive(true)?;
@@ -165,7 +179,6 @@ impl ConnCfg {
                 Host::Unix(_) => continue, // unix sockets are not welcome here
             };
 
-            // TODO: maybe we should add a timeout.
             match connect_once(host, *port).await {
                 Ok(socket) => return Ok(socket),
                 Err(err) => {
diff --git a/proxy/src/config.rs b/proxy/src/config.rs
index ad51502b49..0ceb556ca1 100644
--- a/proxy/src/config.rs
+++ b/proxy/src/config.rs
@@ -40,7 +40,7 @@ pub fn configure_tls(
     let mut cert_resolver = CertResolver::new();
 
     // add default certificate
-    cert_resolver.add_cert(key_path, cert_path)?;
+    cert_resolver.add_cert(key_path, cert_path, true)?;
 
     // add extra certificates
     if let Some(certs_dir) = certs_dir {
@@ -52,8 +52,11 @@ pub fn configure_tls(
                 let key_path = path.join("tls.key");
                 let cert_path = path.join("tls.crt");
                 if key_path.exists() && cert_path.exists() {
-                    cert_resolver
-                        .add_cert(&key_path.to_string_lossy(), &cert_path.to_string_lossy())?;
+                    cert_resolver.add_cert(
+                        &key_path.to_string_lossy(),
+                        &cert_path.to_string_lossy(),
+                        false,
+                    )?;
                 }
             }
         }
@@ -78,16 +81,23 @@ pub fn configure_tls(
 
 struct CertResolver {
     certs: HashMap<String, Arc<rustls::sign::CertifiedKey>>,
+    default: Option<Arc<rustls::sign::CertifiedKey>>,
 }
 
 impl CertResolver {
     fn new() -> Self {
         Self {
             certs: HashMap::new(),
+            default: None,
         }
     }
 
-    fn add_cert(&mut self, key_path: &str, cert_path: &str) -> anyhow::Result<()> {
+    fn add_cert(
+        &mut self,
+        key_path: &str,
+        cert_path: &str,
+        is_default: bool,
+    ) -> anyhow::Result<()> {
         let priv_key = {
             let key_bytes = std::fs::read(key_path).context("TLS key file")?;
             let mut keys = rustls_pemfile::pkcs8_private_keys(&mut &key_bytes[..])
@@ -136,10 +146,13 @@ impl CertResolver {
             "Failed to parse common name from certificate at '{cert_path}'."
         ))?;
 
-        self.certs.insert(
-            common_name,
-            Arc::new(rustls::sign::CertifiedKey::new(cert_chain, key)),
-        );
+        let cert = Arc::new(rustls::sign::CertifiedKey::new(cert_chain, key));
+
+        if is_default {
+            self.default = Some(cert.clone());
+        }
+
+        self.certs.insert(common_name, cert);
 
         Ok(())
     }
@@ -172,7 +185,17 @@ impl rustls::server::ResolvesServerCert for CertResolver {
                 }
             }
         } else {
-            None
+            // No SNI, use the default certificate, otherwise we can't get to
+            // options parameter which can be used to set endpoint name too.
+            // That means that non-SNI flow will not work for CNAME domains in
+            // verify-full mode.
+            //
+            // If that will be a problem we can:
+            //
+            // a) Instead of multi-cert approach use single cert with extra
+            //    domains listed in Subject Alternative Name (SAN).
+            // b) Deploy separate proxy instances for extra domains.
+            self.default.as_ref().cloned()
         }
     }
 }
diff --git a/proxy/src/http/websocket.rs b/proxy/src/http/websocket.rs
index 1757652a90..c7676e8e14 100644
--- a/proxy/src/http/websocket.rs
+++ b/proxy/src/http/websocket.rs
@@ -22,6 +22,7 @@ use tokio::{
     io::{self, AsyncBufRead, AsyncRead, AsyncWrite, ReadBuf},
     net::TcpListener,
 };
+use tokio_util::sync::CancellationToken;
 use tracing::{error, info, info_span, warn, Instrument};
 use utils::http::{error::ApiError, json::json_response};
 
@@ -188,6 +189,7 @@ async fn ws_handler(
 pub async fn task_main(
     config: &'static ProxyConfig,
     ws_listener: TcpListener,
+    cancellation_token: CancellationToken,
 ) -> anyhow::Result<()> {
     scopeguard::defer! {
         info!("websocket server has shut down");
@@ -231,6 +233,7 @@ pub async fn task_main(
 
     hyper::Server::builder(accept::from_stream(tls_listener))
         .serve(make_svc)
+        .with_graceful_shutdown(cancellation_token.cancelled())
         .await?;
 
     Ok(())
diff --git a/proxy/src/main.rs b/proxy/src/main.rs
index c6526e9aff..1fd13c9f68 100644
--- a/proxy/src/main.rs
+++ b/proxy/src/main.rs
@@ -28,6 +28,7 @@ use config::ProxyConfig;
 use futures::FutureExt;
 use std::{borrow::Cow, future::Future, net::SocketAddr};
 use tokio::{net::TcpListener, task::JoinError};
+use tokio_util::sync::CancellationToken;
 use tracing::{info, warn};
 use utils::{project_git_version, sentry_init::init_sentry};
 
@@ -66,39 +67,48 @@ async fn main() -> anyhow::Result<()> {
     let proxy_address: SocketAddr = args.get_one::<String>("proxy").unwrap().parse()?;
     info!("Starting proxy on {proxy_address}");
     let proxy_listener = TcpListener::bind(proxy_address).await?;
+    let cancellation_token = CancellationToken::new();
 
-    let mut tasks = vec![
-        tokio::spawn(handle_signals()),
-        tokio::spawn(http::server::task_main(http_listener)),
-        tokio::spawn(proxy::task_main(config, proxy_listener)),
-        tokio::spawn(console::mgmt::task_main(mgmt_listener)),
-    ];
+    let mut client_tasks = vec![tokio::spawn(proxy::task_main(
+        config,
+        proxy_listener,
+        cancellation_token.clone(),
+    ))];
 
     if let Some(wss_address) = args.get_one::<String>("wss") {
         let wss_address: SocketAddr = wss_address.parse()?;
         info!("Starting wss on {wss_address}");
         let wss_listener = TcpListener::bind(wss_address).await?;
 
-        tasks.push(tokio::spawn(http::websocket::task_main(
+        client_tasks.push(tokio::spawn(http::websocket::task_main(
             config,
             wss_listener,
+            cancellation_token.clone(),
         )));
     }
 
+    let mut tasks = vec![
+        tokio::spawn(handle_signals(cancellation_token)),
+        tokio::spawn(http::server::task_main(http_listener)),
+        tokio::spawn(console::mgmt::task_main(mgmt_listener)),
+    ];
+
     if let Some(metrics_config) = &config.metric_collection {
         tasks.push(tokio::spawn(metrics::task_main(metrics_config)));
     }
 
-    // This combinator will block until either all tasks complete or
-    // one of them finishes with an error (others will be cancelled).
-    let tasks = tasks.into_iter().map(flatten_err);
-    let _: Vec<()> = futures::future::try_join_all(tasks).await?;
-
+    let tasks = futures::future::try_join_all(tasks.into_iter().map(flatten_err));
+    let client_tasks = futures::future::try_join_all(client_tasks.into_iter().map(flatten_err));
+    tokio::select! {
+        // We are only expecting an error from these forever tasks
+        res = tasks => { res?; },
+        res = client_tasks => { res?; },
+    }
     Ok(())
 }
 
 /// Handle unix signals appropriately.
-async fn handle_signals() -> anyhow::Result<()> {
+async fn handle_signals(token: CancellationToken) -> anyhow::Result<()> {
     use tokio::signal::unix::{signal, SignalKind};
 
     let mut hangup = signal(SignalKind::hangup())?;
@@ -116,11 +126,9 @@ async fn handle_signals() -> anyhow::Result<()> {
                 warn!("received SIGINT, exiting immediately");
                 bail!("interrupted");
             }
-            // TODO: Don't accept new proxy connections.
-            // TODO: Shut down once all exisiting connections have been closed.
             _ = terminate.recv() => {
-                warn!("received SIGTERM, exiting immediately");
-                bail!("terminated");
+                warn!("received SIGTERM, shutting down once all existing connections have closed");
+                token.cancel();
             }
         }
     }
diff --git a/proxy/src/metrics.rs b/proxy/src/metrics.rs
index 445c2e930c..6ae1e3a447 100644
--- a/proxy/src/metrics.rs
+++ b/proxy/src/metrics.rs
@@ -95,7 +95,7 @@ fn gather_proxy_io_bytes_per_client() -> Vec<(Ids, (u64, DateTime<Utc>))> {
                     current_metrics.push((
                         Ids {
                             endpoint_id: endpoint_id.to_string(),
-                            branch_id: "".to_string(),
+                            branch_id: branch_id.to_string(),
                         },
                         (value, Utc::now()),
                     ));
diff --git a/proxy/src/proxy.rs b/proxy/src/proxy.rs
index 70fb25474e..1169d76160 100644
--- a/proxy/src/proxy.rs
+++ b/proxy/src/proxy.rs
@@ -17,6 +17,7 @@ use once_cell::sync::Lazy;
 use pq_proto::{BeMessage as Be, FeStartupPacket, StartupMessageParams};
 use std::sync::Arc;
 use tokio::io::{AsyncRead, AsyncWrite, AsyncWriteExt};
+use tokio_util::sync::CancellationToken;
 use tracing::{error, info, warn};
 use utils::measured_stream::MeasuredStream;
 
@@ -63,6 +64,7 @@ static NUM_BYTES_PROXIED_COUNTER: Lazy<IntCounterVec> = Lazy::new(|| {
 pub async fn task_main(
     config: &'static ProxyConfig,
     listener: tokio::net::TcpListener,
+    cancellation_token: CancellationToken,
 ) -> anyhow::Result<()> {
     scopeguard::defer! {
         info!("proxy has shut down");
@@ -72,29 +74,48 @@ pub async fn task_main(
     // will be inherited by all accepted client sockets.
     socket2::SockRef::from(&listener).set_keepalive(true)?;
 
+    let mut connections = tokio::task::JoinSet::new();
     let cancel_map = Arc::new(CancelMap::default());
+
     loop {
-        let (socket, peer_addr) = listener.accept().await?;
-        info!("accepted postgres client connection from {peer_addr}");
+        tokio::select! {
+            accept_result = listener.accept() => {
+                let (socket, peer_addr) = accept_result?;
+                info!("accepted postgres client connection from {peer_addr}");
 
-        let session_id = uuid::Uuid::new_v4();
-        let cancel_map = Arc::clone(&cancel_map);
-        tokio::spawn(
-            async move {
-                info!("spawned a task for {peer_addr}");
+                let session_id = uuid::Uuid::new_v4();
+                let cancel_map = Arc::clone(&cancel_map);
+                connections.spawn(
+                    async move {
+                        info!("spawned a task for {peer_addr}");
 
-                socket
-                    .set_nodelay(true)
-                    .context("failed to set socket option")?;
+                        socket
+                            .set_nodelay(true)
+                            .context("failed to set socket option")?;
 
-                handle_client(config, &cancel_map, session_id, socket).await
+                        handle_client(config, &cancel_map, session_id, socket).await
+                    }
+                    .unwrap_or_else(move |e| {
+                        // Acknowledge that the task has finished with an error.
+                        error!(?session_id, "per-client task finished with an error: {e:#}");
+                    }),
+                );
             }
-            .unwrap_or_else(|e| {
-                // Acknowledge that the task has finished with an error.
-                error!("per-client task finished with an error: {e:#}");
-            }),
-        );
+            _ = cancellation_token.cancelled() => {
+                drop(listener);
+                break;
+            }
+        }
     }
+    // Drain connections
+    while let Some(res) = connections.join_next().await {
+        if let Err(e) = res {
+            if !e.is_panic() && !e.is_cancelled() {
+                warn!("unexpected error from joined connection task: {e:?}");
+            }
+        }
+    }
+    Ok(())
 }
 
 // TODO(tech debt): unite this with its twin below.
diff --git a/run_clippy.sh b/run_clippy.sh
index 9adfddedc2..ae2a17ec0c 100755
--- a/run_clippy.sh
+++ b/run_clippy.sh
@@ -1,4 +1,5 @@
-#!/bin/bash
+#!/usr/bin/env bash
+set -euo pipefail
 
 # If you save this in your path under the name "cargo-zclippy" (or whatever
 # name you like), then you can run it as "cargo zclippy" from the shell prompt.
@@ -8,7 +9,11 @@
 # warnings and errors right in the editor.
 # In vscode, this setting is Rust-analyzer>Check On Save:Command
 
-# * `-A unknown_lints` – do not warn about unknown lint suppressions
-#                        that people with newer toolchains might use
-# * `-D warnings`      - fail on any warnings (`cargo` returns non-zero exit status)
-cargo clippy --locked --all --all-targets --all-features -- -A unknown_lints -D warnings
+# NB: the CI runs the full feature powerset, so, it catches slightly more errors
+# at the expense of longer runtime. This script is used by developers, so, don't
+# do that here.
+
+thisscript="${BASH_SOURCE[0]}"
+thisscript_dir="$(dirname "$thisscript")"
+CLIPPY_COMMON_ARGS="$( source .neon_clippy_args; echo "$CLIPPY_COMMON_ARGS")"
+exec cargo clippy --all-features $CLIPPY_COMMON_ARGS
diff --git a/safekeeper/src/bin/safekeeper.rs b/safekeeper/src/bin/safekeeper.rs
index ace921a26d..3699a2a74c 100644
--- a/safekeeper/src/bin/safekeeper.rs
+++ b/safekeeper/src/bin/safekeeper.rs
@@ -134,7 +134,10 @@ fn main() -> anyhow::Result<()> {
     // 1. init logging
     // 2. tracing panic hook
     // 3. sentry
-    logging::init(LogFormat::from_config(&args.log_format)?)?;
+    logging::init(
+        LogFormat::from_config(&args.log_format)?,
+        logging::TracingErrorLayerEnablement::Disabled,
+    )?;
     logging::replace_panic_hook_with_tracing_panic_hook().forget();
     info!("version: {GIT_VERSION}");
 
diff --git a/safekeeper/src/broker.rs b/safekeeper/src/broker.rs
index 92f35bf51f..6a98d8fd84 100644
--- a/safekeeper/src/broker.rs
+++ b/safekeeper/src/broker.rs
@@ -91,7 +91,7 @@ async fn pull_loop(conf: SafeKeeperConf) -> Result<()> {
             // connection to the broker.
 
             // note: there are blocking operations below, but it's considered fine for now
-            tli.record_safekeeper_info(&msg).await?
+            tli.record_safekeeper_info(msg).await?
         }
     }
     bail!("end of stream");
diff --git a/safekeeper/src/debug_dump.rs b/safekeeper/src/debug_dump.rs
index 674cf9f6eb..f711c4429d 100644
--- a/safekeeper/src/debug_dump.rs
+++ b/safekeeper/src/debug_dump.rs
@@ -9,9 +9,10 @@ use std::path::PathBuf;
 use anyhow::Result;
 use chrono::{DateTime, Utc};
 use postgres_ffi::XLogSegNo;
+use serde::Deserialize;
 use serde::Serialize;
 
-use utils::http::json::display_serialize;
+use serde_with::{serde_as, DisplayFromStr};
 use utils::id::NodeId;
 use utils::id::TenantTimelineId;
 use utils::id::{TenantId, TimelineId};
@@ -22,11 +23,11 @@ use crate::safekeeper::SafekeeperMemState;
 use crate::safekeeper::TermHistory;
 use crate::SafeKeeperConf;
 
-use crate::timeline::ReplicaState;
+use crate::send_wal::WalSenderState;
 use crate::GlobalTimelines;
 
 /// Various filters that influence the resulting JSON output.
-#[derive(Debug, Serialize)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct Args {
     /// Dump all available safekeeper state. False by default.
     pub dump_all: bool,
@@ -51,7 +52,7 @@ pub struct Args {
 }
 
 /// Response for debug dump request.
-#[derive(Debug, Serialize)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct Response {
     pub start_time: DateTime<Utc>,
     pub finish_time: DateTime<Utc>,
@@ -61,7 +62,7 @@ pub struct Response {
 }
 
 /// Safekeeper configuration.
-#[derive(Debug, Serialize)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct Config {
     pub id: NodeId,
     pub workdir: PathBuf,
@@ -72,22 +73,23 @@ pub struct Config {
     pub wal_backup_enabled: bool,
 }
 
-#[derive(Debug, Serialize)]
+#[serde_as]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct Timeline {
-    #[serde(serialize_with = "display_serialize")]
+    #[serde_as(as = "DisplayFromStr")]
     pub tenant_id: TenantId,
-    #[serde(serialize_with = "display_serialize")]
+    #[serde_as(as = "DisplayFromStr")]
     pub timeline_id: TimelineId,
     pub control_file: Option<SafeKeeperState>,
     pub memory: Option<Memory>,
     pub disk_content: Option<DiskContent>,
 }
 
-#[derive(Debug, Serialize)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct Memory {
     pub is_cancelled: bool,
     pub peers_info_len: usize,
-    pub replicas: Vec<Option<ReplicaState>>,
+    pub walsenders: Vec<WalSenderState>,
     pub wal_backup_active: bool,
     pub active: bool,
     pub num_computes: u32,
@@ -102,12 +104,12 @@ pub struct Memory {
     pub file_open: bool,
 }
 
-#[derive(Debug, Serialize)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct DiskContent {
     pub files: Vec<FileInfo>,
 }
 
-#[derive(Debug, Serialize)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct FileInfo {
     pub name: String,
     pub size: u64,
diff --git a/safekeeper/src/handler.rs b/safekeeper/src/handler.rs
index a589fe1869..2c3d1cea0e 100644
--- a/safekeeper/src/handler.rs
+++ b/safekeeper/src/handler.rs
@@ -3,6 +3,7 @@
 
 use anyhow::Context;
 use std::str;
+use std::str::FromStr;
 use tokio::io::{AsyncRead, AsyncWrite};
 use tracing::{info, info_span, Instrument};
 
@@ -49,12 +50,14 @@ fn parse_cmd(cmd: &str) -> anyhow::Result<SafekeeperPostgresCommand> {
     if cmd.starts_with("START_WAL_PUSH") {
         Ok(SafekeeperPostgresCommand::StartWalPush)
     } else if cmd.starts_with("START_REPLICATION") {
-        let re =
-            Regex::new(r"START_REPLICATION(?: PHYSICAL)? ([[:xdigit:]]+/[[:xdigit:]]+)").unwrap();
+        let re = Regex::new(
+            r"START_REPLICATION(?: SLOT [^ ]+)?(?: PHYSICAL)? ([[:xdigit:]]+/[[:xdigit:]]+)",
+        )
+        .unwrap();
         let mut caps = re.captures_iter(cmd);
         let start_lsn = caps
             .next()
-            .map(|cap| cap[1].parse::<Lsn>())
+            .map(|cap| Lsn::from_str(&cap[1]))
             .context("parse start LSN from START_REPLICATION command")??;
         Ok(SafekeeperPostgresCommand::StartReplication { start_lsn })
     } else if cmd.starts_with("IDENTIFY_SYSTEM") {
diff --git a/safekeeper/src/http/routes.rs b/safekeeper/src/http/routes.rs
index cdec45c148..eeb08d2733 100644
--- a/safekeeper/src/http/routes.rs
+++ b/safekeeper/src/http/routes.rs
@@ -3,7 +3,8 @@ use hyper::{Body, Request, Response, StatusCode, Uri};
 use once_cell::sync::Lazy;
 use postgres_ffi::WAL_SEGMENT_SIZE;
 use safekeeper_api::models::SkTimelineInfo;
-use serde::Serialize;
+use serde::{Deserialize, Serialize};
+use serde_with::{serde_as, DisplayFromStr};
 use std::collections::{HashMap, HashSet};
 use std::fmt;
 use std::str::FromStr;
@@ -11,7 +12,6 @@ use std::sync::Arc;
 use storage_broker::proto::SafekeeperTimelineInfo;
 use storage_broker::proto::TenantTimelineId as ProtoTenantTimelineId;
 use tokio::task::JoinError;
-use utils::http::json::display_serialize;
 
 use crate::debug_dump;
 use crate::safekeeper::ServerInfo;
@@ -57,44 +57,46 @@ fn get_conf(request: &Request<Body>) -> &SafeKeeperConf {
 
 /// Same as TermSwitchEntry, but serializes LSN using display serializer
 /// in Postgres format, i.e. 0/FFFFFFFF. Used only for the API response.
-#[derive(Debug, Serialize)]
-struct TermSwitchApiEntry {
+#[serde_as]
+#[derive(Debug, Serialize, Deserialize)]
+pub struct TermSwitchApiEntry {
     pub term: Term,
-    #[serde(serialize_with = "display_serialize")]
+    #[serde_as(as = "DisplayFromStr")]
     pub lsn: Lsn,
 }
 
 /// Augment AcceptorState with epoch for convenience
-#[derive(Debug, Serialize)]
-struct AcceptorStateStatus {
-    term: Term,
-    epoch: Term,
-    term_history: Vec<TermSwitchApiEntry>,
+#[derive(Debug, Serialize, Deserialize)]
+pub struct AcceptorStateStatus {
+    pub term: Term,
+    pub epoch: Term,
+    pub term_history: Vec<TermSwitchApiEntry>,
 }
 
 /// Info about timeline on safekeeper ready for reporting.
-#[derive(Debug, Serialize)]
-struct TimelineStatus {
-    #[serde(serialize_with = "display_serialize")]
-    tenant_id: TenantId,
-    #[serde(serialize_with = "display_serialize")]
-    timeline_id: TimelineId,
-    acceptor_state: AcceptorStateStatus,
-    pg_info: ServerInfo,
-    #[serde(serialize_with = "display_serialize")]
-    flush_lsn: Lsn,
-    #[serde(serialize_with = "display_serialize")]
-    timeline_start_lsn: Lsn,
-    #[serde(serialize_with = "display_serialize")]
-    local_start_lsn: Lsn,
-    #[serde(serialize_with = "display_serialize")]
-    commit_lsn: Lsn,
-    #[serde(serialize_with = "display_serialize")]
-    backup_lsn: Lsn,
-    #[serde(serialize_with = "display_serialize")]
-    peer_horizon_lsn: Lsn,
-    #[serde(serialize_with = "display_serialize")]
-    remote_consistent_lsn: Lsn,
+#[serde_as]
+#[derive(Debug, Serialize, Deserialize)]
+pub struct TimelineStatus {
+    #[serde_as(as = "DisplayFromStr")]
+    pub tenant_id: TenantId,
+    #[serde_as(as = "DisplayFromStr")]
+    pub timeline_id: TimelineId,
+    pub acceptor_state: AcceptorStateStatus,
+    pub pg_info: ServerInfo,
+    #[serde_as(as = "DisplayFromStr")]
+    pub flush_lsn: Lsn,
+    #[serde_as(as = "DisplayFromStr")]
+    pub timeline_start_lsn: Lsn,
+    #[serde_as(as = "DisplayFromStr")]
+    pub local_start_lsn: Lsn,
+    #[serde_as(as = "DisplayFromStr")]
+    pub commit_lsn: Lsn,
+    #[serde_as(as = "DisplayFromStr")]
+    pub backup_lsn: Lsn,
+    #[serde_as(as = "DisplayFromStr")]
+    pub peer_horizon_lsn: Lsn,
+    #[serde_as(as = "DisplayFromStr")]
+    pub remote_consistent_lsn: Lsn,
 }
 
 fn check_permission(request: &Request<Body>, tenant_id: Option<TenantId>) -> Result<(), ApiError> {
@@ -144,7 +146,7 @@ async fn timeline_status_handler(request: Request<Body>) -> Result<Response<Body
         commit_lsn: inmem.commit_lsn,
         backup_lsn: inmem.backup_lsn,
         peer_horizon_lsn: inmem.peer_horizon_lsn,
-        remote_consistent_lsn: inmem.remote_consistent_lsn,
+        remote_consistent_lsn: tli.get_walsenders().get_remote_consistent_lsn(),
     };
     json_response(StatusCode::OK, status)
 }
@@ -246,7 +248,7 @@ async fn record_safekeeper_info(mut request: Request<Body>) -> Result<Response<B
     };
 
     let tli = GlobalTimelines::get(ttid).map_err(ApiError::from)?;
-    tli.record_safekeeper_info(&proto_sk_info)
+    tli.record_safekeeper_info(proto_sk_info)
         .await
         .map_err(ApiError::InternalServerError)?;
 
diff --git a/safekeeper/src/json_ctrl.rs b/safekeeper/src/json_ctrl.rs
index 2841cd195f..dc9188723e 100644
--- a/safekeeper/src/json_ctrl.rs
+++ b/safekeeper/src/json_ctrl.rs
@@ -50,7 +50,7 @@ pub struct AppendLogicalMessage {
     pub pg_version: u32,
 }
 
-#[derive(Debug, Serialize, Deserialize)]
+#[derive(Debug, Serialize)]
 struct AppendResult {
     // safekeeper state after append
     state: SafeKeeperState,
@@ -133,7 +133,7 @@ fn send_proposer_elected(tli: &Arc<Timeline>, term: Term, lsn: Lsn) -> anyhow::R
     Ok(())
 }
 
-#[derive(Debug, Serialize, Deserialize)]
+#[derive(Debug, Serialize)]
 pub struct InsertedWAL {
     begin_lsn: Lsn,
     pub end_lsn: Lsn,
diff --git a/safekeeper/src/metrics.rs b/safekeeper/src/metrics.rs
index 2aaa17bfc5..eafee557d7 100644
--- a/safekeeper/src/metrics.rs
+++ b/safekeeper/src/metrics.rs
@@ -15,11 +15,11 @@ use metrics::{
 use once_cell::sync::Lazy;
 
 use postgres_ffi::XLogSegNo;
+use utils::pageserver_feedback::PageserverFeedback;
 use utils::{id::TenantTimelineId, lsn::Lsn};
 
 use crate::{
     safekeeper::{SafeKeeperState, SafekeeperMemState},
-    timeline::ReplicaState,
     GlobalTimelines,
 };
 
@@ -231,7 +231,7 @@ pub fn time_io_closure(closure: impl FnOnce() -> Result<()>) -> Result<f64> {
 /// Metrics for a single timeline.
 pub struct FullTimelineInfo {
     pub ttid: TenantTimelineId,
-    pub replicas: Vec<ReplicaState>,
+    pub ps_feedback: PageserverFeedback,
     pub wal_backup_active: bool,
     pub timeline_is_active: bool,
     pub num_computes: u32,
@@ -242,6 +242,7 @@ pub struct FullTimelineInfo {
     pub persisted_state: SafeKeeperState,
 
     pub flush_lsn: Lsn,
+    pub remote_consistent_lsn: Lsn,
 
     pub wal_storage: WalStorageMetrics,
 }
@@ -514,19 +515,6 @@ impl Collector for TimelineCollector {
             let timeline_id = tli.ttid.timeline_id.to_string();
             let labels = &[tenant_id.as_str(), timeline_id.as_str()];
 
-            let mut most_advanced: Option<pq_proto::PageserverFeedback> = None;
-            for replica in tli.replicas.iter() {
-                if let Some(replica_feedback) = replica.pageserver_feedback {
-                    if let Some(current) = most_advanced {
-                        if current.last_received_lsn < replica_feedback.last_received_lsn {
-                            most_advanced = Some(replica_feedback);
-                        }
-                    } else {
-                        most_advanced = Some(replica_feedback);
-                    }
-                }
-            }
-
             self.commit_lsn
                 .with_label_values(labels)
                 .set(tli.mem_state.commit_lsn.into());
@@ -544,7 +532,7 @@ impl Collector for TimelineCollector {
                 .set(tli.mem_state.peer_horizon_lsn.into());
             self.remote_consistent_lsn
                 .with_label_values(labels)
-                .set(tli.mem_state.remote_consistent_lsn.into());
+                .set(tli.remote_consistent_lsn.into());
             self.timeline_active
                 .with_label_values(labels)
                 .set(tli.timeline_is_active as u64);
@@ -567,15 +555,17 @@ impl Collector for TimelineCollector {
                 .with_label_values(labels)
                 .set(tli.wal_storage.flush_wal_seconds);
 
-            if let Some(feedback) = most_advanced {
-                self.ps_last_received_lsn
+            self.ps_last_received_lsn
+                .with_label_values(labels)
+                .set(tli.ps_feedback.last_received_lsn.0);
+            if let Ok(unix_time) = tli
+                .ps_feedback
+                .replytime
+                .duration_since(SystemTime::UNIX_EPOCH)
+            {
+                self.feedback_last_time_seconds
                     .with_label_values(labels)
-                    .set(feedback.last_received_lsn);
-                if let Ok(unix_time) = feedback.replytime.duration_since(SystemTime::UNIX_EPOCH) {
-                    self.feedback_last_time_seconds
-                        .with_label_values(labels)
-                        .set(unix_time.as_secs());
-                }
+                    .set(unix_time.as_secs());
             }
 
             if tli.last_removed_segno != 0 {
diff --git a/safekeeper/src/receive_wal.rs b/safekeeper/src/receive_wal.rs
index 61e4c5f0fa..195470e3ca 100644
--- a/safekeeper/src/receive_wal.rs
+++ b/safekeeper/src/receive_wal.rs
@@ -27,6 +27,8 @@ use tokio::sync::mpsc::error::TryRecvError;
 use tokio::sync::mpsc::Receiver;
 use tokio::sync::mpsc::Sender;
 use tokio::task::spawn_blocking;
+use tokio::time::Duration;
+use tokio::time::Instant;
 use tracing::*;
 use utils::id::TenantTimelineId;
 use utils::lsn::Lsn;
@@ -206,6 +208,10 @@ async fn network_write<IO: AsyncRead + AsyncWrite + Unpin>(
     }
 }
 
+// Send keepalive messages to walproposer, to make sure it receives updates
+// even when it writes a steady stream of messages.
+const KEEPALIVE_INTERVAL: Duration = Duration::from_secs(1);
+
 /// Takes messages from msg_rx, processes and pushes replies to reply_tx.
 struct WalAcceptor {
     tli: Arc<Timeline>,
@@ -253,18 +259,25 @@ impl WalAcceptor {
             timeline: Arc::clone(&self.tli),
         };
 
-        let mut next_msg: ProposerAcceptorMessage;
+        // After this timestamp we will stop processing AppendRequests and send a response
+        // to the walproposer. walproposer sends at least one AppendRequest per second,
+        // we will send keepalives by replying to these requests once per second.
+        let mut next_keepalive = Instant::now();
 
         loop {
             let opt_msg = self.msg_rx.recv().await;
             if opt_msg.is_none() {
                 return Ok(()); // chan closed, streaming terminated
             }
-            next_msg = opt_msg.unwrap();
+            let mut next_msg = opt_msg.unwrap();
 
-            if matches!(next_msg, ProposerAcceptorMessage::AppendRequest(_)) {
+            let reply_msg = if matches!(next_msg, ProposerAcceptorMessage::AppendRequest(_)) {
                 // loop through AppendRequest's while it's readily available to
                 // write as many WAL as possible without fsyncing
+                //
+                // Note: this will need to be rewritten if we want to read non-AppendRequest messages here.
+                // Otherwise, we might end up in a situation where we read a message, but don't
+                // process it.
                 while let ProposerAcceptorMessage::AppendRequest(append_request) = next_msg {
                     let noflush_msg = ProposerAcceptorMessage::NoFlushAppendRequest(append_request);
 
@@ -274,6 +287,11 @@ impl WalAcceptor {
                         }
                     }
 
+                    // get out of this loop if keepalive time is reached
+                    if Instant::now() >= next_keepalive {
+                        break;
+                    }
+
                     match self.msg_rx.try_recv() {
                         Ok(msg) => next_msg = msg,
                         Err(TryRecvError::Empty) => break,
@@ -282,18 +300,18 @@ impl WalAcceptor {
                 }
 
                 // flush all written WAL to the disk
-                if let Some(reply) = self.tli.process_msg(&ProposerAcceptorMessage::FlushWAL)? {
-                    if self.reply_tx.send(reply).await.is_err() {
-                        return Ok(()); // chan closed, streaming terminated
-                    }
-                }
+                self.tli.process_msg(&ProposerAcceptorMessage::FlushWAL)?
             } else {
                 // process message other than AppendRequest
-                if let Some(reply) = self.tli.process_msg(&next_msg)? {
-                    if self.reply_tx.send(reply).await.is_err() {
-                        return Ok(()); // chan closed, streaming terminated
-                    }
+                self.tli.process_msg(&next_msg)?
+            };
+
+            if let Some(reply) = reply_msg {
+                if self.reply_tx.send(reply).await.is_err() {
+                    return Ok(()); // chan closed, streaming terminated
                 }
+                // reset keepalive time
+                next_keepalive = Instant::now() + KEEPALIVE_INTERVAL;
             }
         }
     }
diff --git a/safekeeper/src/safekeeper.rs b/safekeeper/src/safekeeper.rs
index 10b4842cbd..33da0c8e5a 100644
--- a/safekeeper/src/safekeeper.rs
+++ b/safekeeper/src/safekeeper.rs
@@ -18,7 +18,8 @@ use crate::control_file;
 use crate::send_wal::HotStandbyFeedback;
 
 use crate::wal_storage;
-use pq_proto::{PageserverFeedback, SystemId};
+use pq_proto::SystemId;
+use utils::pageserver_feedback::PageserverFeedback;
 use utils::{
     bin_ser::LeSer,
     id::{NodeId, TenantId, TenantTimelineId, TimelineId},
@@ -205,14 +206,13 @@ pub struct SafeKeeperState {
     pub peers: PersistedPeers,
 }
 
-#[derive(Debug, Clone, Serialize)]
+#[derive(Debug, Clone, Serialize, Deserialize)]
 // In memory safekeeper state. Fields mirror ones in `SafeKeeperState`; values
 // are not flushed yet.
 pub struct SafekeeperMemState {
     pub commit_lsn: Lsn,
     pub backup_lsn: Lsn,
     pub peer_horizon_lsn: Lsn,
-    pub remote_consistent_lsn: Lsn,
     #[serde(with = "hex")]
     pub proposer_uuid: PgUuid,
 }
@@ -347,7 +347,7 @@ pub struct AppendRequestHeader {
 }
 
 /// Report safekeeper state to proposer
-#[derive(Debug, Serialize, Deserialize)]
+#[derive(Debug, Serialize)]
 pub struct AppendResponse {
     // Current term of the safekeeper; if it is higher than proposer's, the
     // compute is out of date.
@@ -540,7 +540,6 @@ where
                 commit_lsn: state.commit_lsn,
                 backup_lsn: state.backup_lsn,
                 peer_horizon_lsn: state.peer_horizon_lsn,
-                remote_consistent_lsn: state.remote_consistent_lsn,
                 proposer_uuid: state.proposer_uuid,
             },
             state,
@@ -781,10 +780,6 @@ where
 
             // Initializing backup_lsn is useful to avoid making backup think it should upload 0 segment.
             self.inmem.backup_lsn = max(self.inmem.backup_lsn, state.timeline_start_lsn);
-            // Initializing remote_consistent_lsn sets that we have nothing to
-            // stream to pageserver(s) immediately after creation.
-            self.inmem.remote_consistent_lsn =
-                max(self.inmem.remote_consistent_lsn, state.timeline_start_lsn);
 
             state.acceptor_state.term_history = msg.term_history.clone();
             self.persist_control_file(state)?;
@@ -837,7 +832,6 @@ where
         state.commit_lsn = self.inmem.commit_lsn;
         state.backup_lsn = self.inmem.backup_lsn;
         state.peer_horizon_lsn = self.inmem.peer_horizon_lsn;
-        state.remote_consistent_lsn = self.inmem.remote_consistent_lsn;
         state.proposer_uuid = self.inmem.proposer_uuid;
         self.state.persist(&state)
     }
@@ -940,14 +934,12 @@ where
             self.state.backup_lsn + (self.state.server.wal_seg_size as u64) < new_backup_lsn;
         self.inmem.backup_lsn = new_backup_lsn;
 
-        let new_remote_consistent_lsn = max(
-            Lsn(sk_info.remote_consistent_lsn),
-            self.inmem.remote_consistent_lsn,
-        );
+        // value in sk_info should be maximized over our local in memory value.
+        let new_remote_consistent_lsn = Lsn(sk_info.remote_consistent_lsn);
+        assert!(self.state.remote_consistent_lsn <= new_remote_consistent_lsn);
         sync_control_file |= self.state.remote_consistent_lsn
             + (self.state.server.wal_seg_size as u64)
             < new_remote_consistent_lsn;
-        self.inmem.remote_consistent_lsn = new_remote_consistent_lsn;
 
         let new_peer_horizon_lsn = max(Lsn(sk_info.peer_horizon_lsn), self.inmem.peer_horizon_lsn);
         sync_control_file |= self.state.peer_horizon_lsn + (self.state.server.wal_seg_size as u64)
@@ -955,7 +947,12 @@ where
         self.inmem.peer_horizon_lsn = new_peer_horizon_lsn;
 
         if sync_control_file {
-            self.persist_control_file(self.state.clone())?;
+            let mut state = self.state.clone();
+            // Note: we do not persist remote_consistent_lsn in other paths of
+            // persisting cf -- that is not much needed currently. We could do
+            // that by storing Arc to walsenders in Safekeeper.
+            state.remote_consistent_lsn = new_remote_consistent_lsn;
+            self.persist_control_file(state)?;
         }
         Ok(())
     }
diff --git a/safekeeper/src/send_wal.rs b/safekeeper/src/send_wal.rs
index a6ca89efa4..6b303eb0fe 100644
--- a/safekeeper/src/send_wal.rs
+++ b/safekeeper/src/send_wal.rs
@@ -1,21 +1,28 @@
 //! This module implements the streaming side of replication protocol, starting
-//! with the "START_REPLICATION" message.
+//! with the "START_REPLICATION" message, and registry of walsenders.
 
 use crate::handler::SafekeeperPostgresHandler;
-use crate::timeline::{ReplicaState, Timeline};
+use crate::timeline::Timeline;
+use crate::wal_service::ConnectionId;
 use crate::wal_storage::WalReader;
 use crate::GlobalTimelines;
 use anyhow::Context as AnyhowContext;
 use bytes::Bytes;
+use parking_lot::Mutex;
 use postgres_backend::PostgresBackend;
 use postgres_backend::{CopyStreamHandlerEnd, PostgresBackendReader, QueryError};
 use postgres_ffi::get_current_timestamp;
 use postgres_ffi::{TimestampTz, MAX_SEND_SIZE};
-use pq_proto::{BeMessage, PageserverFeedback, WalSndKeepAlive, XLogDataBody};
+use pq_proto::{BeMessage, WalSndKeepAlive, XLogDataBody};
 use serde::{Deserialize, Serialize};
+use serde_with::{serde_as, DisplayFromStr};
 use tokio::io::{AsyncRead, AsyncWrite};
+use utils::id::TenantTimelineId;
+use utils::lsn::AtomicLsn;
+use utils::pageserver_feedback::PageserverFeedback;
 
-use std::cmp::min;
+use std::cmp::{max, min};
+use std::net::SocketAddr;
 use std::str;
 use std::sync::Arc;
 use std::time::Duration;
@@ -40,6 +47,8 @@ pub struct HotStandbyFeedback {
     pub catalog_xmin: FullTransactionId,
 }
 
+const INVALID_FULL_TRANSACTION_ID: FullTransactionId = 0;
+
 impl HotStandbyFeedback {
     pub fn empty() -> HotStandbyFeedback {
         HotStandbyFeedback {
@@ -51,24 +60,294 @@ impl HotStandbyFeedback {
 }
 
 /// Standby status update
-#[derive(Debug, Clone, Deserialize)]
+#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
 pub struct StandbyReply {
-    pub write_lsn: Lsn, // last lsn received by pageserver
-    pub flush_lsn: Lsn, // pageserver's disk consistent lSN
-    pub apply_lsn: Lsn, // pageserver's remote consistent lSN
-    pub reply_ts: TimestampTz,
+    pub write_lsn: Lsn, // The location of the last WAL byte + 1 received and written to disk in the standby.
+    pub flush_lsn: Lsn, // The location of the last WAL byte + 1 flushed to disk in the standby.
+    pub apply_lsn: Lsn, // The location of the last WAL byte + 1 applied in the standby.
+    pub reply_ts: TimestampTz, // The client's system clock at the time of transmission, as microseconds since midnight on 2000-01-01.
     pub reply_requested: bool,
 }
 
-/// Scope guard to unregister replication connection from timeline
-struct ReplicationConnGuard {
-    replica: usize, // replica internal ID assigned by timeline
-    timeline: Arc<Timeline>,
+impl StandbyReply {
+    fn empty() -> Self {
+        StandbyReply {
+            write_lsn: Lsn::INVALID,
+            flush_lsn: Lsn::INVALID,
+            apply_lsn: Lsn::INVALID,
+            reply_ts: 0,
+            reply_requested: false,
+        }
+    }
 }
 
-impl Drop for ReplicationConnGuard {
+#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
+pub struct StandbyFeedback {
+    reply: StandbyReply,
+    hs_feedback: HotStandbyFeedback,
+}
+
+/// WalSenders registry. Timeline holds it (wrapped in Arc).
+pub struct WalSenders {
+    /// Lsn maximized over all walsenders *and* peer data, so might be higher
+    /// than what we receive from replicas.
+    remote_consistent_lsn: AtomicLsn,
+    mutex: Mutex<WalSendersShared>,
+}
+
+impl WalSenders {
+    pub fn new(remote_consistent_lsn: Lsn) -> Arc<WalSenders> {
+        Arc::new(WalSenders {
+            remote_consistent_lsn: AtomicLsn::from(remote_consistent_lsn),
+            mutex: Mutex::new(WalSendersShared::new()),
+        })
+    }
+
+    /// Register new walsender. Returned guard provides access to the slot and
+    /// automatically deregisters in Drop.
+    fn register(
+        self: &Arc<WalSenders>,
+        ttid: TenantTimelineId,
+        addr: SocketAddr,
+        conn_id: ConnectionId,
+        appname: Option<String>,
+    ) -> WalSenderGuard {
+        let slots = &mut self.mutex.lock().slots;
+        let walsender_state = WalSenderState {
+            ttid,
+            addr,
+            conn_id,
+            appname,
+            feedback: ReplicationFeedback::Pageserver(PageserverFeedback::empty()),
+        };
+        // find empty slot or create new one
+        let pos = if let Some(pos) = slots.iter().position(|s| s.is_none()) {
+            slots[pos] = Some(walsender_state);
+            pos
+        } else {
+            let pos = slots.len();
+            slots.push(Some(walsender_state));
+            pos
+        };
+        WalSenderGuard {
+            id: pos,
+            walsenders: self.clone(),
+        }
+    }
+
+    /// Get state of all walsenders.
+    pub fn get_all(self: &Arc<WalSenders>) -> Vec<WalSenderState> {
+        self.mutex.lock().slots.iter().flatten().cloned().collect()
+    }
+
+    /// Get aggregated pageserver feedback.
+    pub fn get_ps_feedback(self: &Arc<WalSenders>) -> PageserverFeedback {
+        self.mutex.lock().agg_ps_feedback
+    }
+
+    /// Get aggregated pageserver and hot standby feedback (we send them to compute).
+    pub fn get_feedbacks(self: &Arc<WalSenders>) -> (PageserverFeedback, HotStandbyFeedback) {
+        let shared = self.mutex.lock();
+        (shared.agg_ps_feedback, shared.agg_hs_feedback)
+    }
+
+    /// Record new pageserver feedback, update aggregated values.
+    fn record_ps_feedback(self: &Arc<WalSenders>, id: WalSenderId, feedback: &PageserverFeedback) {
+        let mut shared = self.mutex.lock();
+        shared.get_slot_mut(id).feedback = ReplicationFeedback::Pageserver(*feedback);
+        shared.update_ps_feedback();
+        self.update_remote_consistent_lsn(shared.agg_ps_feedback.remote_consistent_lsn);
+    }
+
+    /// Record standby reply.
+    fn record_standby_reply(self: &Arc<WalSenders>, id: WalSenderId, reply: &StandbyReply) {
+        let mut shared = self.mutex.lock();
+        let slot = shared.get_slot_mut(id);
+        match &mut slot.feedback {
+            ReplicationFeedback::Standby(sf) => sf.reply = *reply,
+            ReplicationFeedback::Pageserver(_) => {
+                slot.feedback = ReplicationFeedback::Standby(StandbyFeedback {
+                    reply: *reply,
+                    hs_feedback: HotStandbyFeedback::empty(),
+                })
+            }
+        }
+    }
+
+    /// Record hot standby feedback, update aggregated value.
+    fn record_hs_feedback(self: &Arc<WalSenders>, id: WalSenderId, feedback: &HotStandbyFeedback) {
+        let mut shared = self.mutex.lock();
+        let slot = shared.get_slot_mut(id);
+        match &mut slot.feedback {
+            ReplicationFeedback::Standby(sf) => sf.hs_feedback = *feedback,
+            ReplicationFeedback::Pageserver(_) => {
+                slot.feedback = ReplicationFeedback::Standby(StandbyFeedback {
+                    reply: StandbyReply::empty(),
+                    hs_feedback: *feedback,
+                })
+            }
+        }
+        shared.update_hs_feedback();
+    }
+
+    /// Get remote_consistent_lsn reported by the pageserver. Returns None if
+    /// client is not pageserver.
+    fn get_ws_remote_consistent_lsn(self: &Arc<WalSenders>, id: WalSenderId) -> Option<Lsn> {
+        let shared = self.mutex.lock();
+        let slot = shared.get_slot(id);
+        match slot.feedback {
+            ReplicationFeedback::Pageserver(feedback) => Some(feedback.remote_consistent_lsn),
+            _ => None,
+        }
+    }
+
+    /// Get remote_consistent_lsn maximized across all walsenders and peers.
+    pub fn get_remote_consistent_lsn(self: &Arc<WalSenders>) -> Lsn {
+        self.remote_consistent_lsn.load()
+    }
+
+    /// Update maximized remote_consistent_lsn, return new (potentially) value.
+    pub fn update_remote_consistent_lsn(self: &Arc<WalSenders>, candidate: Lsn) -> Lsn {
+        self.remote_consistent_lsn
+            .fetch_max(candidate)
+            .max(candidate)
+    }
+
+    /// Unregister walsender.
+    fn unregister(self: &Arc<WalSenders>, id: WalSenderId) {
+        let mut shared = self.mutex.lock();
+        shared.slots[id] = None;
+        shared.update_hs_feedback();
+    }
+}
+
+struct WalSendersShared {
+    // aggregated over all walsenders value
+    agg_hs_feedback: HotStandbyFeedback,
+    // aggregated over all walsenders value
+    agg_ps_feedback: PageserverFeedback,
+    slots: Vec<Option<WalSenderState>>,
+}
+
+impl WalSendersShared {
+    fn new() -> Self {
+        WalSendersShared {
+            agg_hs_feedback: HotStandbyFeedback::empty(),
+            agg_ps_feedback: PageserverFeedback::empty(),
+            slots: Vec::new(),
+        }
+    }
+
+    /// Get content of provided id slot, it must exist.
+    fn get_slot(&self, id: WalSenderId) -> &WalSenderState {
+        self.slots[id].as_ref().expect("walsender doesn't exist")
+    }
+
+    /// Get mut content of provided id slot, it must exist.
+    fn get_slot_mut(&mut self, id: WalSenderId) -> &mut WalSenderState {
+        self.slots[id].as_mut().expect("walsender doesn't exist")
+    }
+
+    /// Update aggregated hot standy feedback. We just take min of valid xmins
+    /// and ts.
+    fn update_hs_feedback(&mut self) {
+        let mut agg = HotStandbyFeedback::empty();
+        for ws_state in self.slots.iter().flatten() {
+            if let ReplicationFeedback::Standby(standby_feedback) = ws_state.feedback {
+                let hs_feedback = standby_feedback.hs_feedback;
+                // doing Option math like op1.iter().chain(op2.iter()).min()
+                // would be nicer, but we serialize/deserialize this struct
+                // directly, so leave as is for now
+                if hs_feedback.xmin != INVALID_FULL_TRANSACTION_ID {
+                    if agg.xmin != INVALID_FULL_TRANSACTION_ID {
+                        agg.xmin = min(agg.xmin, hs_feedback.xmin);
+                    } else {
+                        agg.xmin = hs_feedback.xmin;
+                    }
+                    agg.ts = min(agg.ts, hs_feedback.ts);
+                }
+                if hs_feedback.catalog_xmin != INVALID_FULL_TRANSACTION_ID {
+                    if agg.catalog_xmin != INVALID_FULL_TRANSACTION_ID {
+                        agg.catalog_xmin = min(agg.catalog_xmin, hs_feedback.catalog_xmin);
+                    } else {
+                        agg.catalog_xmin = hs_feedback.catalog_xmin;
+                    }
+                    agg.ts = min(agg.ts, hs_feedback.ts);
+                }
+            }
+        }
+        self.agg_hs_feedback = agg;
+    }
+
+    /// Update aggregated pageserver feedback. LSNs (last_received,
+    /// disk_consistent, remote_consistent) and reply timestamp are just
+    /// maximized; timeline_size if taken from feedback with highest
+    /// last_received lsn. This is generally reasonable, but we might want to
+    /// implement other policies once multiple pageservers start to be actively
+    /// used.
+    fn update_ps_feedback(&mut self) {
+        let init = PageserverFeedback::empty();
+        let acc =
+            self.slots
+                .iter()
+                .flatten()
+                .fold(init, |mut acc, ws_state| match ws_state.feedback {
+                    ReplicationFeedback::Pageserver(feedback) => {
+                        if feedback.last_received_lsn > acc.last_received_lsn {
+                            acc.current_timeline_size = feedback.current_timeline_size;
+                        }
+                        acc.last_received_lsn =
+                            max(feedback.last_received_lsn, acc.last_received_lsn);
+                        acc.disk_consistent_lsn =
+                            max(feedback.disk_consistent_lsn, acc.disk_consistent_lsn);
+                        acc.remote_consistent_lsn =
+                            max(feedback.remote_consistent_lsn, acc.remote_consistent_lsn);
+                        acc.replytime = max(feedback.replytime, acc.replytime);
+                        acc
+                    }
+                    ReplicationFeedback::Standby(_) => acc,
+                });
+        self.agg_ps_feedback = acc;
+    }
+}
+
+// Serialized is used only for pretty printing in json.
+#[serde_as]
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct WalSenderState {
+    #[serde_as(as = "DisplayFromStr")]
+    ttid: TenantTimelineId,
+    addr: SocketAddr,
+    conn_id: ConnectionId,
+    // postgres application_name
+    appname: Option<String>,
+    feedback: ReplicationFeedback,
+}
+
+// Receiver is either pageserver or regular standby, which have different
+// feedbacks.
+#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
+enum ReplicationFeedback {
+    Pageserver(PageserverFeedback),
+    Standby(StandbyFeedback),
+}
+
+// id of the occupied slot in WalSenders to access it (and save in the
+// WalSenderGuard). We could give Arc directly to the slot, but there is not
+// much sense in that as values aggregation which is performed on each feedback
+// receival iterates over all walsenders.
+pub type WalSenderId = usize;
+
+/// Scope guard to access slot in WalSenders registry and unregister from it in
+/// Drop.
+pub struct WalSenderGuard {
+    id: WalSenderId,
+    walsenders: Arc<WalSenders>,
+}
+
+impl Drop for WalSenderGuard {
     fn drop(&mut self) {
-        self.timeline.remove_replica(self.replica);
+        self.walsenders.unregister(self.id);
     }
 }
 
@@ -97,16 +376,13 @@ impl SafekeeperPostgresHandler {
         let tli =
             GlobalTimelines::get(self.ttid).map_err(|e| CopyStreamHandlerEnd::Other(e.into()))?;
 
-        let state = ReplicaState::new();
-        // This replica_id is used below to check if it's time to stop replication.
-        let replica_id = tli.add_replica(state);
-
-        // Use a guard object to remove our entry from the timeline, when the background
-        // thread and us have both finished using it.
-        let _guard = Arc::new(ReplicationConnGuard {
-            replica: replica_id,
-            timeline: tli.clone(),
-        });
+        // Use a guard object to remove our entry from the timeline when we are done.
+        let ws_guard = Arc::new(tli.get_walsenders().register(
+            self.ttid,
+            *pgb.get_peer_addr(),
+            self.conn_id,
+            self.appname.clone(),
+        ));
 
         // Walproposer gets special handling: safekeeper must give proposer all
         // local WAL till the end, whether committed or not (walproposer will
@@ -154,16 +430,11 @@ impl SafekeeperPostgresHandler {
             end_pos,
             stop_pos,
             commit_lsn_watch_rx: tli.get_commit_lsn_watch_rx(),
-            replica_id,
+            ws_guard: ws_guard.clone(),
             wal_reader,
             send_buf: [0; MAX_SEND_SIZE],
         };
-        let mut reply_reader = ReplyReader {
-            reader,
-            tli,
-            replica_id,
-            feedback: ReplicaState::new(),
-        };
+        let mut reply_reader = ReplyReader { reader, ws_guard };
 
         let res = tokio::select! {
             // todo: add read|write .context to these errors
@@ -190,7 +461,7 @@ struct WalSender<'a, IO> {
     // in recovery.
     stop_pos: Option<Lsn>,
     commit_lsn_watch_rx: Receiver<Lsn>,
-    replica_id: usize,
+    ws_guard: Arc<WalSenderGuard>,
     wal_reader: WalReader,
     // buffer for readling WAL into to send it
     send_buf: [u8; MAX_SEND_SIZE],
@@ -264,14 +535,20 @@ impl<IO: AsyncRead + AsyncWrite + Unpin> WalSender<'_, IO> {
                 return Ok(());
             }
             // Timed out waiting for WAL, check for termination and send KA
-            if self.tli.should_walsender_stop(self.replica_id) {
-                // Terminate if there is nothing more to send.
-                // TODO close the stream properly
-                return Err(CopyStreamHandlerEnd::ServerInitiated(format!(
-                    "ending streaming to {:?} at {}, receiver is caughtup and there is no computes",
-                    self.appname, self.start_pos,
-                )));
+            if let Some(remote_consistent_lsn) = self
+                .ws_guard
+                .walsenders
+                .get_ws_remote_consistent_lsn(self.ws_guard.id)
+            {
+                if self.tli.should_walsender_stop(remote_consistent_lsn) {
+                    // Terminate if there is nothing more to send.
+                    return Err(CopyStreamHandlerEnd::ServerInitiated(format!(
+                        "ending streaming to {:?} at {}, receiver is caughtup and there is no computes",
+                        self.appname, self.start_pos,
+                    )));
+                }
             }
+
             self.pgb
                 .write_message(&BeMessage::KeepAlive(WalSndKeepAlive {
                     sent_ptr: self.end_pos.0,
@@ -286,9 +563,7 @@ impl<IO: AsyncRead + AsyncWrite + Unpin> WalSender<'_, IO> {
 /// A half driving receiving replies.
 struct ReplyReader<IO> {
     reader: PostgresBackendReader<IO>,
-    tli: Arc<Timeline>,
-    replica_id: usize,
-    feedback: ReplicaState,
+    ws_guard: Arc<WalSenderGuard>,
 }
 
 impl<IO: AsyncRead + AsyncWrite + Unpin> ReplyReader<IO> {
@@ -303,29 +578,32 @@ impl<IO: AsyncRead + AsyncWrite + Unpin> ReplyReader<IO> {
         match msg.first().cloned() {
             Some(HOT_STANDBY_FEEDBACK_TAG_BYTE) => {
                 // Note: deserializing is on m[1..] because we skip the tag byte.
-                self.feedback.hs_feedback = HotStandbyFeedback::des(&msg[1..])
+                let hs_feedback = HotStandbyFeedback::des(&msg[1..])
                     .context("failed to deserialize HotStandbyFeedback")?;
-                self.tli
-                    .update_replica_state(self.replica_id, self.feedback);
+                self.ws_guard
+                    .walsenders
+                    .record_hs_feedback(self.ws_guard.id, &hs_feedback);
             }
             Some(STANDBY_STATUS_UPDATE_TAG_BYTE) => {
-                let _reply =
+                let reply =
                     StandbyReply::des(&msg[1..]).context("failed to deserialize StandbyReply")?;
-                // This must be a regular postgres replica,
-                // because pageserver doesn't send this type of messages to safekeeper.
-                // Currently we just ignore this, tracking progress for them is not supported.
+                self.ws_guard
+                    .walsenders
+                    .record_standby_reply(self.ws_guard.id, &reply);
             }
             Some(NEON_STATUS_UPDATE_TAG_BYTE) => {
                 // pageserver sends this.
                 // Note: deserializing is on m[9..] because we skip the tag byte and len bytes.
                 let buf = Bytes::copy_from_slice(&msg[9..]);
-                let reply = PageserverFeedback::parse(buf);
+                let ps_feedback = PageserverFeedback::parse(buf);
 
-                trace!("PageserverFeedback is {:?}", reply);
-                self.feedback.pageserver_feedback = Some(reply);
-
-                self.tli
-                    .update_replica_state(self.replica_id, self.feedback);
+                trace!("PageserverFeedback is {:?}", ps_feedback);
+                self.ws_guard
+                    .walsenders
+                    .record_ps_feedback(self.ws_guard.id, &ps_feedback);
+                // in principle new remote_consistent_lsn could allow to
+                // deactivate the timeline, but we check that regularly through
+                // broker updated, not need to do it here
             }
             _ => warn!("unexpected message {:?}", msg),
         }
@@ -368,3 +646,89 @@ async fn wait_for_lsn(rx: &mut Receiver<Lsn>, lsn: Lsn) -> anyhow::Result<Option
         Err(_) => Ok(None),
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use postgres_protocol::PG_EPOCH;
+    use utils::id::{TenantId, TimelineId};
+
+    use super::*;
+
+    fn mock_ttid() -> TenantTimelineId {
+        TenantTimelineId {
+            tenant_id: TenantId::from_slice(&[0x00; 16]).unwrap(),
+            timeline_id: TimelineId::from_slice(&[0x00; 16]).unwrap(),
+        }
+    }
+
+    fn mock_addr() -> SocketAddr {
+        "127.0.0.1:8080".parse().unwrap()
+    }
+
+    // add to wss specified feedback setting other fields to dummy values
+    fn push_feedback(wss: &mut WalSendersShared, feedback: ReplicationFeedback) {
+        let walsender_state = WalSenderState {
+            ttid: mock_ttid(),
+            addr: mock_addr(),
+            conn_id: 1,
+            appname: None,
+            feedback,
+        };
+        wss.slots.push(Some(walsender_state))
+    }
+
+    // form standby feedback with given hot standby feedback ts/xmin and the
+    // rest set to dummy values.
+    fn hs_feedback(ts: TimestampTz, xmin: FullTransactionId) -> ReplicationFeedback {
+        ReplicationFeedback::Standby(StandbyFeedback {
+            reply: StandbyReply::empty(),
+            hs_feedback: HotStandbyFeedback {
+                ts,
+                xmin,
+                catalog_xmin: 0,
+            },
+        })
+    }
+
+    // test that hs aggregation works as expected
+    #[test]
+    fn test_hs_feedback_no_valid() {
+        let mut wss = WalSendersShared::new();
+        push_feedback(&mut wss, hs_feedback(1, INVALID_FULL_TRANSACTION_ID));
+        wss.update_hs_feedback();
+        assert_eq!(wss.agg_hs_feedback.xmin, INVALID_FULL_TRANSACTION_ID);
+    }
+
+    #[test]
+    fn test_hs_feedback() {
+        let mut wss = WalSendersShared::new();
+        push_feedback(&mut wss, hs_feedback(1, INVALID_FULL_TRANSACTION_ID));
+        push_feedback(&mut wss, hs_feedback(1, 42));
+        push_feedback(&mut wss, hs_feedback(1, 64));
+        wss.update_hs_feedback();
+        assert_eq!(wss.agg_hs_feedback.xmin, 42);
+    }
+
+    // form pageserver feedback with given last_record_lsn / tli size and the
+    // rest set to dummy values.
+    fn ps_feedback(current_timeline_size: u64, last_received_lsn: Lsn) -> ReplicationFeedback {
+        ReplicationFeedback::Pageserver(PageserverFeedback {
+            current_timeline_size,
+            last_received_lsn,
+            disk_consistent_lsn: Lsn::INVALID,
+            remote_consistent_lsn: Lsn::INVALID,
+            replytime: *PG_EPOCH,
+        })
+    }
+
+    // test that ps aggregation works as expected
+    #[test]
+    fn test_ps_feedback() {
+        let mut wss = WalSendersShared::new();
+        push_feedback(&mut wss, ps_feedback(8, Lsn(42)));
+        push_feedback(&mut wss, ps_feedback(4, Lsn(84)));
+        wss.update_ps_feedback();
+        assert_eq!(wss.agg_ps_feedback.current_timeline_size, 4);
+        assert_eq!(wss.agg_ps_feedback.last_received_lsn, Lsn(84));
+    }
+}
diff --git a/safekeeper/src/timeline.rs b/safekeeper/src/timeline.rs
index 8097c863fa..64ca6967df 100644
--- a/safekeeper/src/timeline.rs
+++ b/safekeeper/src/timeline.rs
@@ -4,10 +4,10 @@
 use anyhow::{anyhow, bail, Result};
 use parking_lot::{Mutex, MutexGuard};
 use postgres_ffi::XLogSegNo;
-use pq_proto::PageserverFeedback;
-use serde::Serialize;
-use std::cmp::{max, min};
+
+use std::cmp::max;
 use std::path::PathBuf;
+use std::sync::Arc;
 use tokio::{
     sync::{mpsc::Sender, watch},
     time::Instant,
@@ -26,7 +26,7 @@ use crate::safekeeper::{
     AcceptorProposerMessage, ProposerAcceptorMessage, SafeKeeper, SafeKeeperState,
     SafekeeperMemState, ServerInfo, Term,
 };
-use crate::send_wal::HotStandbyFeedback;
+use crate::send_wal::WalSenders;
 use crate::{control_file, safekeeper::UNKNOWN_SERVER_VERSION};
 
 use crate::metrics::FullTimelineInfo;
@@ -81,48 +81,12 @@ impl PeersInfo {
     }
 }
 
-/// Replica status update + hot standby feedback
-#[derive(Debug, Clone, Copy, Serialize)]
-pub struct ReplicaState {
-    /// last known lsn received by replica
-    pub last_received_lsn: Lsn, // None means we don't know
-    /// combined remote consistent lsn of pageservers
-    pub remote_consistent_lsn: Lsn,
-    /// combined hot standby feedback from all replicas
-    pub hs_feedback: HotStandbyFeedback,
-    /// Replication specific feedback received from pageserver, if any
-    pub pageserver_feedback: Option<PageserverFeedback>,
-}
-
-impl Default for ReplicaState {
-    fn default() -> Self {
-        Self::new()
-    }
-}
-
-impl ReplicaState {
-    pub fn new() -> ReplicaState {
-        ReplicaState {
-            last_received_lsn: Lsn::MAX,
-            remote_consistent_lsn: Lsn(0),
-            hs_feedback: HotStandbyFeedback {
-                ts: 0,
-                xmin: u64::MAX,
-                catalog_xmin: u64::MAX,
-            },
-            pageserver_feedback: None,
-        }
-    }
-}
-
 /// Shared state associated with database instance
 pub struct SharedState {
     /// Safekeeper object
     sk: SafeKeeper<control_file::FileStorage, wal_storage::PhysicalStorage>,
     /// In memory list containing state of peers sent in latest messages from them.
     peers_info: PeersInfo,
-    /// State of replicas
-    replicas: Vec<Option<ReplicaState>>,
     /// True when WAL backup launcher oversees the timeline, making sure WAL is
     /// offloaded, allows to bother launcher less.
     wal_backup_active: bool,
@@ -171,7 +135,6 @@ impl SharedState {
         Ok(Self {
             sk,
             peers_info: PeersInfo(vec![]),
-            replicas: vec![],
             wal_backup_active: false,
             active: false,
             num_computes: 0,
@@ -191,7 +154,6 @@ impl SharedState {
         Ok(Self {
             sk: SafeKeeper::new(control_store, wal_store, conf.my_id)?,
             peers_info: PeersInfo(vec![]),
-            replicas: Vec::new(),
             wal_backup_active: false,
             active: false,
             num_computes: 0,
@@ -199,17 +161,17 @@ impl SharedState {
         })
     }
 
-    fn is_active(&self) -> bool {
+    fn is_active(&self, remote_consistent_lsn: Lsn) -> bool {
         self.is_wal_backup_required()
             // FIXME: add tracking of relevant pageservers and check them here individually,
             // otherwise migration won't work (we suspend too early).
-            || self.sk.inmem.remote_consistent_lsn < self.sk.inmem.commit_lsn
+            || remote_consistent_lsn < self.sk.inmem.commit_lsn
     }
 
     /// Mark timeline active/inactive and return whether s3 offloading requires
     /// start/stop action.
-    fn update_status(&mut self, ttid: TenantTimelineId) -> bool {
-        let is_active = self.is_active();
+    fn update_status(&mut self, remote_consistent_lsn: Lsn, ttid: TenantTimelineId) -> bool {
+        let is_active = self.is_active(remote_consistent_lsn);
         if self.active != is_active {
             info!("timeline {} active={} now", ttid, is_active);
         }
@@ -254,68 +216,11 @@ impl SharedState {
         self.sk.state.server.wal_seg_size as usize
     }
 
-    /// Get combined state of all alive replicas
-    pub fn get_replicas_state(&self) -> ReplicaState {
-        let mut acc = ReplicaState::new();
-        for state in self.replicas.iter().flatten() {
-            acc.hs_feedback.ts = max(acc.hs_feedback.ts, state.hs_feedback.ts);
-            acc.hs_feedback.xmin = min(acc.hs_feedback.xmin, state.hs_feedback.xmin);
-            acc.hs_feedback.catalog_xmin =
-                min(acc.hs_feedback.catalog_xmin, state.hs_feedback.catalog_xmin);
-
-            // FIXME
-            // If multiple pageservers are streaming WAL and send feedback for the same timeline simultaneously,
-            // this code is not correct.
-            // Now the most advanced feedback is used.
-            // If one pageserver lags when another doesn't, the backpressure won't be activated on compute and lagging
-            // pageserver is prone to timeout errors.
-            //
-            // To choose what feedback to use and resend to compute node,
-            // we need to know which pageserver compute node considers to be main.
-            // See https://github.com/neondatabase/neon/issues/1171
-            //
-            if let Some(pageserver_feedback) = state.pageserver_feedback {
-                if let Some(acc_feedback) = acc.pageserver_feedback {
-                    if acc_feedback.last_received_lsn < pageserver_feedback.last_received_lsn {
-                        warn!("More than one pageserver is streaming WAL for the timeline. Feedback resolving is not fully supported yet.");
-                        acc.pageserver_feedback = Some(pageserver_feedback);
-                    }
-                } else {
-                    acc.pageserver_feedback = Some(pageserver_feedback);
-                }
-
-                // last lsn received by pageserver
-                // FIXME if multiple pageservers are streaming WAL, last_received_lsn must be tracked per pageserver.
-                // See https://github.com/neondatabase/neon/issues/1171
-                acc.last_received_lsn = Lsn::from(pageserver_feedback.last_received_lsn);
-
-                // When at least one pageserver has preserved data up to remote_consistent_lsn,
-                // safekeeper is free to delete it, so choose max of all pageservers.
-                acc.remote_consistent_lsn = max(
-                    Lsn::from(pageserver_feedback.remote_consistent_lsn),
-                    acc.remote_consistent_lsn,
-                );
-            }
-        }
-        acc
-    }
-
-    /// Assign new replica ID. We choose first empty cell in the replicas vector
-    /// or extend the vector if there are no free slots.
-    pub fn add_replica(&mut self, state: ReplicaState) -> usize {
-        if let Some(pos) = self.replicas.iter().position(|r| r.is_none()) {
-            self.replicas[pos] = Some(state);
-            return pos;
-        }
-        let pos = self.replicas.len();
-        self.replicas.push(Some(state));
-        pos
-    }
-
     fn get_safekeeper_info(
         &self,
         ttid: &TenantTimelineId,
         conf: &SafeKeeperConf,
+        remote_consistent_lsn: Lsn,
     ) -> SafekeeperTimelineInfo {
         SafekeeperTimelineInfo {
             safekeeper_id: conf.my_id.0,
@@ -328,11 +233,7 @@ impl SharedState {
             // note: this value is not flushed to control file yet and can be lost
             commit_lsn: self.sk.inmem.commit_lsn.0,
             // TODO: rework feedbacks to avoid max here
-            remote_consistent_lsn: max(
-                self.get_replicas_state().remote_consistent_lsn,
-                self.sk.inmem.remote_consistent_lsn,
-            )
-            .0,
+            remote_consistent_lsn: remote_consistent_lsn.0,
             peer_horizon_lsn: self.sk.inmem.peer_horizon_lsn.0,
             safekeeper_connstr: conf.listen_pg_addr.clone(),
             backup_lsn: self.sk.inmem.backup_lsn.0,
@@ -387,6 +288,7 @@ pub struct Timeline {
     /// Safekeeper and other state, that should remain consistent and synchronized
     /// with the disk.
     mutex: Mutex<SharedState>,
+    walsenders: Arc<WalSenders>,
 
     /// Cancellation channel. Delete/cancel will send `true` here as a cancellation signal.
     cancellation_tx: watch::Sender<bool>,
@@ -409,6 +311,7 @@ impl Timeline {
         let _enter = info_span!("load_timeline", timeline = %ttid.timeline_id).entered();
 
         let shared_state = SharedState::restore(&conf, &ttid)?;
+        let rcl = shared_state.sk.state.remote_consistent_lsn;
         let (commit_lsn_watch_tx, commit_lsn_watch_rx) =
             watch::channel(shared_state.sk.state.commit_lsn);
         let (cancellation_tx, cancellation_rx) = watch::channel(false);
@@ -419,6 +322,7 @@ impl Timeline {
             commit_lsn_watch_tx,
             commit_lsn_watch_rx,
             mutex: Mutex::new(shared_state),
+            walsenders: WalSenders::new(rcl),
             cancellation_rx,
             cancellation_tx,
             timeline_dir: conf.timeline_dir(&ttid),
@@ -444,6 +348,7 @@ impl Timeline {
             commit_lsn_watch_tx,
             commit_lsn_watch_rx,
             mutex: Mutex::new(SharedState::create_new(&conf, &ttid, state)?),
+            walsenders: WalSenders::new(Lsn(0)),
             cancellation_rx,
             cancellation_tx,
             timeline_dir: conf.timeline_dir(&ttid),
@@ -475,7 +380,7 @@ impl Timeline {
         match || -> Result<()> {
             shared_state.sk.persist()?;
             // TODO: add more initialization steps here
-            shared_state.update_status(self.ttid);
+            self.update_status(shared_state);
             Ok(())
         }() {
             Ok(_) => Ok(()),
@@ -531,6 +436,10 @@ impl Timeline {
         self.mutex.lock()
     }
 
+    fn update_status(&self, shared_state: &mut SharedState) -> bool {
+        shared_state.update_status(self.get_walsenders().get_remote_consistent_lsn(), self.ttid)
+    }
+
     /// Register compute connection, starting timeline-related activity if it is
     /// not running yet.
     pub async fn on_compute_connect(&self) -> Result<()> {
@@ -542,7 +451,7 @@ impl Timeline {
         {
             let mut shared_state = self.write_shared_state();
             shared_state.num_computes += 1;
-            is_wal_backup_action_pending = shared_state.update_status(self.ttid);
+            is_wal_backup_action_pending = self.update_status(&mut shared_state);
         }
         // Wake up wal backup launcher, if offloading not started yet.
         if is_wal_backup_action_pending {
@@ -559,7 +468,7 @@ impl Timeline {
         {
             let mut shared_state = self.write_shared_state();
             shared_state.num_computes -= 1;
-            is_wal_backup_action_pending = shared_state.update_status(self.ttid);
+            is_wal_backup_action_pending = self.update_status(&mut shared_state);
         }
         // Wake up wal backup launcher, if it is time to stop the offloading.
         if is_wal_backup_action_pending {
@@ -574,26 +483,19 @@ impl Timeline {
         Ok(())
     }
 
-    /// Returns true if walsender should stop sending WAL to pageserver.
-    /// TODO: check this pageserver is actually interested in this timeline.
-    pub fn should_walsender_stop(&self, replica_id: usize) -> bool {
+    /// Returns true if walsender should stop sending WAL to pageserver. We
+    /// terminate it if remote_consistent_lsn reached commit_lsn and there is no
+    /// computes. While there might be nothing to stream already, we learn about
+    /// remote_consistent_lsn update through replication feedback, and we want
+    /// to stop pushing to the broker if pageserver is fully caughtup.
+    pub fn should_walsender_stop(&self, reported_remote_consistent_lsn: Lsn) -> bool {
         if self.is_cancelled() {
             return true;
         }
-        let mut shared_state = self.write_shared_state();
+        let shared_state = self.write_shared_state();
         if shared_state.num_computes == 0 {
-            let replica_state = shared_state.replicas[replica_id].unwrap();
-            let reported_remote_consistent_lsn = replica_state
-                .pageserver_feedback
-                .map(|f| Lsn(f.remote_consistent_lsn))
-                .unwrap_or(Lsn::INVALID);
-            let stop = shared_state.sk.inmem.commit_lsn == Lsn(0) || // no data at all yet
-            (reported_remote_consistent_lsn!= Lsn::MAX && // Lsn::MAX means that we don't know the latest LSN yet.
-            reported_remote_consistent_lsn >= shared_state.sk.inmem.commit_lsn);
-            if stop {
-                shared_state.update_status(self.ttid);
-                return true;
-            }
+            return shared_state.sk.inmem.commit_lsn == Lsn(0) || // no data at all yet
+            reported_remote_consistent_lsn >= shared_state.sk.inmem.commit_lsn;
         }
         false
     }
@@ -628,13 +530,12 @@ impl Timeline {
             let mut shared_state = self.write_shared_state();
             rmsg = shared_state.sk.process_msg(msg)?;
 
-            // if this is AppendResponse, fill in proper hot standby feedback and disk consistent lsn
+            // if this is AppendResponse, fill in proper pageserver and hot
+            // standby feedback.
             if let Some(AcceptorProposerMessage::AppendResponse(ref mut resp)) = rmsg {
-                let state = shared_state.get_replicas_state();
-                resp.hs_feedback = state.hs_feedback;
-                if let Some(pageserver_feedback) = state.pageserver_feedback {
-                    resp.pageserver_feedback = pageserver_feedback;
-                }
+                let (ps_feedback, hs_feedback) = self.walsenders.get_feedbacks();
+                resp.hs_feedback = hs_feedback;
+                resp.pageserver_feedback = ps_feedback;
             }
 
             commit_lsn = shared_state.sk.inmem.commit_lsn;
@@ -684,19 +585,29 @@ impl Timeline {
     /// Get safekeeper info for broadcasting to broker and other peers.
     pub fn get_safekeeper_info(&self, conf: &SafeKeeperConf) -> SafekeeperTimelineInfo {
         let shared_state = self.write_shared_state();
-        shared_state.get_safekeeper_info(&self.ttid, conf)
+        shared_state.get_safekeeper_info(
+            &self.ttid,
+            conf,
+            self.walsenders.get_remote_consistent_lsn(),
+        )
     }
 
     /// Update timeline state with peer safekeeper data.
-    pub async fn record_safekeeper_info(&self, sk_info: &SafekeeperTimelineInfo) -> Result<()> {
+    pub async fn record_safekeeper_info(&self, mut sk_info: SafekeeperTimelineInfo) -> Result<()> {
+        // Update local remote_consistent_lsn in memory (in .walsenders) and in
+        // sk_info to pass it down to control file.
+        sk_info.remote_consistent_lsn = self
+            .walsenders
+            .update_remote_consistent_lsn(Lsn(sk_info.remote_consistent_lsn))
+            .0;
         let is_wal_backup_action_pending: bool;
         let commit_lsn: Lsn;
         {
             let mut shared_state = self.write_shared_state();
-            shared_state.sk.record_safekeeper_info(sk_info)?;
-            let peer_info = PeerInfo::from_sk_info(sk_info, Instant::now());
+            shared_state.sk.record_safekeeper_info(&sk_info)?;
+            let peer_info = PeerInfo::from_sk_info(&sk_info, Instant::now());
             shared_state.peers_info.upsert(&peer_info);
-            is_wal_backup_action_pending = shared_state.update_status(self.ttid);
+            is_wal_backup_action_pending = self.update_status(&mut shared_state);
             commit_lsn = shared_state.sk.inmem.commit_lsn;
         }
         self.commit_lsn_watch_tx.send(commit_lsn)?;
@@ -723,22 +634,8 @@ impl Timeline {
             .collect()
     }
 
-    /// Add send_wal replica to the in-memory vector of replicas.
-    pub fn add_replica(&self, state: ReplicaState) -> usize {
-        self.write_shared_state().add_replica(state)
-    }
-
-    /// Update replication replica state.
-    pub fn update_replica_state(&self, id: usize, state: ReplicaState) {
-        let mut shared_state = self.write_shared_state();
-        shared_state.replicas[id] = Some(state);
-    }
-
-    /// Remove send_wal replica from the in-memory vector of replicas.
-    pub fn remove_replica(&self, id: usize) {
-        let mut shared_state = self.write_shared_state();
-        assert!(shared_state.replicas[id].is_some());
-        shared_state.replicas[id] = None;
+    pub fn get_walsenders(&self) -> &Arc<WalSenders> {
+        &self.walsenders
     }
 
     /// Returns flush_lsn.
@@ -781,16 +678,12 @@ impl Timeline {
             return None;
         }
 
+        let ps_feedback = self.walsenders.get_ps_feedback();
         let state = self.write_shared_state();
         if state.active {
             Some(FullTimelineInfo {
                 ttid: self.ttid,
-                replicas: state
-                    .replicas
-                    .iter()
-                    .filter_map(|r| r.as_ref())
-                    .copied()
-                    .collect(),
+                ps_feedback,
                 wal_backup_active: state.wal_backup_active,
                 timeline_is_active: state.active,
                 num_computes: state.num_computes,
@@ -799,6 +692,7 @@ impl Timeline {
                 mem_state: state.sk.inmem.clone(),
                 persisted_state: state.sk.state.clone(),
                 flush_lsn: state.sk.wal_store.flush_lsn(),
+                remote_consistent_lsn: self.get_walsenders().get_remote_consistent_lsn(),
                 wal_storage: state.sk.wal_store.get_metrics(),
             })
         } else {
@@ -816,7 +710,7 @@ impl Timeline {
         debug_dump::Memory {
             is_cancelled: self.is_cancelled(),
             peers_info_len: state.peers_info.0.len(),
-            replicas: state.replicas.clone(),
+            walsenders: self.walsenders.get_all(),
             wal_backup_active: state.wal_backup_active,
             active: state.active,
             num_computes: state.num_computes,
diff --git a/safekeeper/src/wal_storage.rs b/safekeeper/src/wal_storage.rs
index 9b385630c2..54e27714ea 100644
--- a/safekeeper/src/wal_storage.rs
+++ b/safekeeper/src/wal_storage.rs
@@ -18,6 +18,7 @@ use postgres_ffi::v14::xlog_utils::{IsPartialXLogFileName, IsXLogFileName, XLogF
 use postgres_ffi::{XLogSegNo, PG_TLI};
 use std::cmp::{max, min};
 
+use bytes::Bytes;
 use std::fs::{self, remove_file, File, OpenOptions};
 use std::io::Write;
 use std::path::{Path, PathBuf};
@@ -36,6 +37,7 @@ use postgres_ffi::XLOG_BLCKSZ;
 
 use postgres_ffi::waldecoder::WalStreamDecoder;
 
+use pq_proto::SystemId;
 use tokio::io::{AsyncReadExt, AsyncSeekExt};
 
 pub trait Storage {
@@ -478,6 +480,13 @@ pub struct WalReader {
 
     // We don't have WAL locally if LSN is less than local_start_lsn
     local_start_lsn: Lsn,
+    // We will respond with zero-ed bytes before this Lsn as long as
+    // pos is in the same segment as timeline_start_lsn.
+    timeline_start_lsn: Lsn,
+    // integer version number of PostgreSQL, e.g. 14; 15; 16
+    pg_version: u32,
+    system_id: SystemId,
+    timeline_start_segment: Option<Bytes>,
 }
 
 impl WalReader {
@@ -488,19 +497,27 @@ impl WalReader {
         start_pos: Lsn,
         enable_remote_read: bool,
     ) -> Result<Self> {
-        if start_pos < state.timeline_start_lsn {
+        if state.server.wal_seg_size == 0 || state.local_start_lsn == Lsn(0) {
+            bail!("state uninitialized, no data to read");
+        }
+
+        // TODO: Upgrade to bail!() once we know this couldn't possibly happen
+        if state.timeline_start_lsn == Lsn(0) {
+            warn!("timeline_start_lsn uninitialized before initializing wal reader");
+        }
+
+        if start_pos
+            < state
+                .timeline_start_lsn
+                .segment_lsn(state.server.wal_seg_size as usize)
+        {
             bail!(
-                "Requested streaming from {}, which is before the start of the timeline {}",
+                "Requested streaming from {}, which is before the start of the timeline {}, and also doesn't start at the first segment of that timeline",
                 start_pos,
                 state.timeline_start_lsn
             );
         }
 
-        // TODO: add state.timeline_start_lsn == Lsn(0) check
-        if state.server.wal_seg_size == 0 || state.local_start_lsn == Lsn(0) {
-            bail!("state uninitialized, no data to read");
-        }
-
         Ok(Self {
             workdir,
             timeline_dir,
@@ -509,10 +526,65 @@ impl WalReader {
             wal_segment: None,
             enable_remote_read,
             local_start_lsn: state.local_start_lsn,
+            timeline_start_lsn: state.timeline_start_lsn,
+            pg_version: state.server.pg_version / 10000,
+            system_id: state.server.system_id,
+            timeline_start_segment: None,
         })
     }
 
     pub async fn read(&mut self, buf: &mut [u8]) -> Result<usize> {
+        // If this timeline is new, we may not have a full segment yet, so
+        // we pad the first bytes of the timeline's first WAL segment with 0s
+        if self.pos < self.timeline_start_lsn {
+            debug_assert_eq!(
+                self.pos.segment_number(self.wal_seg_size),
+                self.timeline_start_lsn.segment_number(self.wal_seg_size)
+            );
+
+            // All bytes after timeline_start_lsn are in WAL, but those before
+            // are not, so we manually construct an empty segment for the bytes
+            // not available in this timeline.
+            if self.timeline_start_segment.is_none() {
+                let it = postgres_ffi::generate_wal_segment(
+                    self.timeline_start_lsn.segment_number(self.wal_seg_size),
+                    self.system_id,
+                    self.pg_version,
+                    self.timeline_start_lsn,
+                )?;
+                self.timeline_start_segment = Some(it);
+            }
+
+            assert!(self.timeline_start_segment.is_some());
+            let segment = self.timeline_start_segment.take().unwrap();
+
+            let seg_bytes = &segment[..];
+
+            // How much of the current segment have we already consumed?
+            let pos_seg_offset = self.pos.segment_offset(self.wal_seg_size);
+
+            // How many bytes may we consume in total?
+            let tl_start_seg_offset = self.timeline_start_lsn.segment_offset(self.wal_seg_size);
+
+            debug_assert!(seg_bytes.len() > pos_seg_offset);
+            debug_assert!(seg_bytes.len() > tl_start_seg_offset);
+
+            // Copy as many bytes as possible into the buffer
+            let len = (tl_start_seg_offset - pos_seg_offset).min(buf.len());
+            buf[0..len].copy_from_slice(&seg_bytes[pos_seg_offset..pos_seg_offset + len]);
+
+            self.pos += len as u64;
+
+            // If we're done with the segment, we can release it's memory.
+            // However, if we're not yet done, store it so that we don't have to
+            // construct the segment the next time this function is called.
+            if self.pos < self.timeline_start_lsn {
+                self.timeline_start_segment = Some(segment);
+            }
+
+            return Ok(len);
+        }
+
         let mut wal_segment = match self.wal_segment.take() {
             Some(reader) => reader,
             None => self.open_segment().await?,
diff --git a/scripts/flaky_tests.py b/scripts/flaky_tests.py
index 829cc814e8..262950b61d 100755
--- a/scripts/flaky_tests.py
+++ b/scripts/flaky_tests.py
@@ -42,12 +42,16 @@ def main(args: argparse.Namespace):
     res: DefaultDict[str, DefaultDict[str, Dict[str, bool]]]
     res = defaultdict(lambda: defaultdict(dict))
 
-    logging.info("connecting to the database...")
-    with psycopg2.connect(connstr, connect_timeout=10) as conn:
-        with conn.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
-            logging.info("fetching flaky tests...")
-            cur.execute(FLAKY_TESTS_QUERY, (interval_days,))
-            rows = cur.fetchall()
+    try:
+        logging.info("connecting to the database...")
+        with psycopg2.connect(connstr, connect_timeout=30) as conn:
+            with conn.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+                logging.info("fetching flaky tests...")
+                cur.execute(FLAKY_TESTS_QUERY, (interval_days,))
+                rows = cur.fetchall()
+    except psycopg2.OperationalError as exc:
+        logging.error("cannot fetch flaky tests from the DB due to an error", exc)
+        rows = []
 
     for row in rows:
         logging.info(f"\t{row['parent_suite'].replace('.', '/')}/{row['suite']}.py::{row['test']}")
diff --git a/storage_broker/src/bin/storage_broker.rs b/storage_broker/src/bin/storage_broker.rs
index d7ace28426..597d9860d8 100644
--- a/storage_broker/src/bin/storage_broker.rs
+++ b/storage_broker/src/bin/storage_broker.rs
@@ -23,7 +23,6 @@ use std::convert::Infallible;
 use std::net::SocketAddr;
 use std::pin::Pin;
 use std::sync::Arc;
-use std::task::Poll;
 use std::time::Duration;
 use tokio::sync::broadcast;
 use tokio::sync::broadcast::error::RecvError;
@@ -374,7 +373,7 @@ impl BrokerService for Broker {
                     Ok(info) => yield info,
                     Err(RecvError::Lagged(skipped_msg)) => {
                         missed_msgs += skipped_msg;
-                        if let Poll::Ready(_) = futures::poll!(Box::pin(warn_interval.tick())) {
+                        if (futures::poll!(Box::pin(warn_interval.tick()))).is_ready() {
                             warn!("subscription id={}, key={:?} addr={:?} dropped {} messages, channel is full",
                                 subscriber.id, subscriber.key, subscriber.remote_addr, missed_msgs);
                             missed_msgs = 0;
@@ -431,7 +430,10 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
     // 1. init logging
     // 2. tracing panic hook
     // 3. sentry
-    logging::init(LogFormat::from_config(&args.log_format)?)?;
+    logging::init(
+        LogFormat::from_config(&args.log_format)?,
+        logging::TracingErrorLayerEnablement::Disabled,
+    )?;
     logging::replace_panic_hook_with_tracing_panic_hook().forget();
     // initialize sentry if SENTRY_DSN is provided
     let _sentry_guard = init_sentry(Some(GIT_VERSION.into()), &[]);
diff --git a/test_runner/fixtures/compare_fixtures.py b/test_runner/fixtures/compare_fixtures.py
index b328cea5c6..f0d9ce4af2 100644
--- a/test_runner/fixtures/compare_fixtures.py
+++ b/test_runner/fixtures/compare_fixtures.py
@@ -114,7 +114,7 @@ class NeonCompare(PgCompare):
         self.timeline = self.env.neon_cli.create_timeline(branch_name, tenant_id=self.tenant)
 
         # Start pg
-        self._pg = self.env.postgres.create_start(branch_name, "main", self.tenant)
+        self._pg = self.env.endpoints.create_start(branch_name, "main", self.tenant)
 
     @property
     def pg(self) -> PgProtocol:
diff --git a/test_runner/fixtures/metrics.py b/test_runner/fixtures/metrics.py
index 2984f2c7d3..0e958ddd06 100644
--- a/test_runner/fixtures/metrics.py
+++ b/test_runner/fixtures/metrics.py
@@ -45,14 +45,18 @@ PAGESERVER_PER_TENANT_REMOTE_TIMELINE_CLIENT_METRICS: Tuple[str, ...] = (
     *[f"pageserver_remote_timeline_client_calls_started_{x}" for x in ["bucket", "count", "sum"]],
     *[f"pageserver_remote_operation_seconds_{x}" for x in ["bucket", "count", "sum"]],
     "pageserver_remote_physical_size",
+    "pageserver_remote_timeline_client_bytes_started_total",
+    "pageserver_remote_timeline_client_bytes_finished_total",
 )
 
 PAGESERVER_GLOBAL_METRICS: Tuple[str, ...] = (
     "pageserver_storage_operations_seconds_global_count",
     "pageserver_storage_operations_seconds_global_sum",
     "pageserver_storage_operations_seconds_global_bucket",
+    "pageserver_unexpected_ondemand_downloads_count_total",
     "libmetrics_launch_timestamp",
     "libmetrics_build_info",
+    "libmetrics_tracing_event_count_total",
 )
 
 PAGESERVER_PER_TENANT_METRICS: Tuple[str, ...] = (
diff --git a/test_runner/fixtures/neon_fixtures.py b/test_runner/fixtures/neon_fixtures.py
index 5b6f2e5c96..c71a063cb4 100644
--- a/test_runner/fixtures/neon_fixtures.py
+++ b/test_runner/fixtures/neon_fixtures.py
@@ -830,7 +830,7 @@ class NeonEnvBuilder:
         # Stop all the nodes.
         if self.env:
             log.info("Cleaning up all storage and compute nodes")
-            self.env.postgres.stop_all()
+            self.env.endpoints.stop_all()
             for sk in self.env.safekeepers:
                 sk.stop(immediate=True)
             self.env.pageserver.stop(immediate=True)
@@ -894,7 +894,7 @@ class NeonEnv:
         self.port_distributor = config.port_distributor
         self.s3_mock_server = config.mock_s3_server
         self.neon_cli = NeonCli(env=self)
-        self.postgres = PostgresFactory(self)
+        self.endpoints = EndpointFactory(self)
         self.safekeepers: List[Safekeeper] = []
         self.broker = config.broker
         self.remote_storage = config.remote_storage
@@ -902,6 +902,7 @@ class NeonEnv:
         self.pg_version = config.pg_version
         self.neon_binpath = config.neon_binpath
         self.pg_distrib_dir = config.pg_distrib_dir
+        self.endpoint_counter = 0
 
         # generate initial tenant ID here instead of letting 'neon init' generate it,
         # so that we don't need to dig it out of the config file afterwards.
@@ -1015,6 +1016,13 @@ class NeonEnv:
         priv = (Path(self.repo_dir) / "auth_private_key.pem").read_text()
         return AuthKeys(pub=pub, priv=priv)
 
+    def generate_endpoint_id(self) -> str:
+        """
+        Generate a unique endpoint ID
+        """
+        self.endpoint_counter += 1
+        return "ep-" + str(self.endpoint_counter)
+
 
 @pytest.fixture(scope=shareable_scope)
 def _shared_simple_env(
@@ -1073,7 +1081,7 @@ def neon_simple_env(_shared_simple_env: NeonEnv) -> Iterator[NeonEnv]:
     """
     yield _shared_simple_env
 
-    _shared_simple_env.postgres.stop_all()
+    _shared_simple_env.endpoints.stop_all()
 
 
 @pytest.fixture(scope="function")
@@ -1097,7 +1105,7 @@ def neon_env_builder(
     neon_env_builder.init_start().
 
     After the initialization, you can launch compute nodes by calling
-    the functions in the 'env.postgres' factory object, stop/start the
+    the functions in the 'env.endpoints' factory object, stop/start the
     nodes, etc.
     """
 
@@ -1438,16 +1446,17 @@ class NeonCli(AbstractNeonCli):
             args.extend(["-m", "immediate"])
         return self.raw_cli(args)
 
-    def pg_create(
+    def endpoint_create(
         self,
         branch_name: str,
-        node_name: Optional[str] = None,
+        endpoint_id: Optional[str] = None,
         tenant_id: Optional[TenantId] = None,
+        hot_standby: bool = False,
         lsn: Optional[Lsn] = None,
         port: Optional[int] = None,
     ) -> "subprocess.CompletedProcess[str]":
         args = [
-            "pg",
+            "endpoint",
             "create",
             "--tenant-id",
             str(tenant_id or self.env.initial_tenant),
@@ -1460,22 +1469,24 @@ class NeonCli(AbstractNeonCli):
             args.extend(["--lsn", str(lsn)])
         if port is not None:
             args.extend(["--port", str(port)])
-        if node_name is not None:
-            args.append(node_name)
+        if endpoint_id is not None:
+            args.append(endpoint_id)
+        if hot_standby:
+            args.extend(["--hot-standby", "true"])
 
         res = self.raw_cli(args)
         res.check_returncode()
         return res
 
-    def pg_start(
+    def endpoint_start(
         self,
-        node_name: str,
+        endpoint_id: str,
         tenant_id: Optional[TenantId] = None,
         lsn: Optional[Lsn] = None,
         port: Optional[int] = None,
     ) -> "subprocess.CompletedProcess[str]":
         args = [
-            "pg",
+            "endpoint",
             "start",
             "--tenant-id",
             str(tenant_id or self.env.initial_tenant),
@@ -1486,30 +1497,30 @@ class NeonCli(AbstractNeonCli):
             args.append(f"--lsn={lsn}")
         if port is not None:
             args.append(f"--port={port}")
-        if node_name is not None:
-            args.append(node_name)
+        if endpoint_id is not None:
+            args.append(endpoint_id)
 
         res = self.raw_cli(args)
         res.check_returncode()
         return res
 
-    def pg_stop(
+    def endpoint_stop(
         self,
-        node_name: str,
+        endpoint_id: str,
         tenant_id: Optional[TenantId] = None,
         destroy=False,
         check_return_code=True,
     ) -> "subprocess.CompletedProcess[str]":
         args = [
-            "pg",
+            "endpoint",
             "stop",
             "--tenant-id",
             str(tenant_id or self.env.initial_tenant),
         ]
         if destroy:
             args.append("--destroy")
-        if node_name is not None:
-            args.append(node_name)
+        if endpoint_id is not None:
+            args.append(endpoint_id)
 
         return self.raw_cli(args, check_return_code=check_return_code)
 
@@ -1905,15 +1916,26 @@ def remote_pg(
     connstr = os.getenv("BENCHMARK_CONNSTR")
     if connstr is None:
         raise ValueError("no connstr provided, use BENCHMARK_CONNSTR environment variable")
+
+    host = parse_dsn(connstr).get("host", "")
+    is_neon = host.endswith(".neon.build")
+
     start_ms = int(datetime.utcnow().timestamp() * 1000)
     with RemotePostgres(pg_bin, connstr) as remote_pg:
+        if is_neon:
+            timeline_id = TimelineId(remote_pg.safe_psql("SHOW neon.timeline_id")[0][0])
+
         yield remote_pg
 
     end_ms = int(datetime.utcnow().timestamp() * 1000)
-    host = parse_dsn(connstr).get("host", "")
-    if host.endswith(".neon.build"):
+    if is_neon:
         # Add 10s margin to the start and end times
-        allure_add_grafana_links(host, start_ms - 10_000, end_ms + 10_000)
+        allure_add_grafana_links(
+            host,
+            timeline_id,
+            start_ms - 10_000,
+            end_ms + 10_000,
+        )
 
 
 class PSQL:
@@ -2033,6 +2055,17 @@ class NeonProxy(PgProtocol):
         self._wait_until_ready()
         return self
 
+    # Sends SIGTERM to the proxy if it has been started
+    def terminate(self):
+        if self._popen:
+            self._popen.terminate()
+
+    # Waits for proxy to exit if it has been opened with a default timeout of
+    # two seconds. Raises subprocess.TimeoutExpired if the proxy does not exit in time.
+    def wait_for_exit(self, timeout=2):
+        if self._popen:
+            self._popen.wait(timeout=2)
+
     @backoff.on_exception(backoff.expo, requests.exceptions.RequestException, max_time=10)
     def _wait_until_ready(self):
         requests.get(f"http://{self.host}:{self.http_port}/v1/status")
@@ -2167,8 +2200,8 @@ def static_proxy(
         yield proxy
 
 
-class Postgres(PgProtocol):
-    """An object representing a running postgres daemon."""
+class Endpoint(PgProtocol):
+    """An object representing a Postgres compute endpoint managed by the control plane."""
 
     def __init__(
         self, env: NeonEnv, tenant_id: TenantId, port: int, check_stop_result: bool = True
@@ -2176,33 +2209,44 @@ class Postgres(PgProtocol):
         super().__init__(host="localhost", port=port, user="cloud_admin", dbname="postgres")
         self.env = env
         self.running = False
-        self.node_name: Optional[str] = None  # dubious, see asserts below
+        self.branch_name: Optional[str] = None  # dubious
+        self.endpoint_id: Optional[str] = None  # dubious, see asserts below
         self.pgdata_dir: Optional[str] = None  # Path to computenode PGDATA
         self.tenant_id = tenant_id
         self.port = port
         self.check_stop_result = check_stop_result
-        # path to conf is <repo_dir>/pgdatadirs/tenants/<tenant_id>/<node_name>/postgresql.conf
+        # path to conf is <repo_dir>/endpoints/<endpoint_id>/pgdata/postgresql.conf
 
     def create(
         self,
         branch_name: str,
-        node_name: Optional[str] = None,
+        endpoint_id: Optional[str] = None,
+        hot_standby: bool = False,
         lsn: Optional[Lsn] = None,
         config_lines: Optional[List[str]] = None,
-    ) -> "Postgres":
+    ) -> "Endpoint":
         """
-        Create the pg data directory.
+        Create a new Postgres endpoint.
         Returns self.
         """
 
         if not config_lines:
             config_lines = []
 
-        self.node_name = node_name or f"{branch_name}_pg_node"
-        self.env.neon_cli.pg_create(
-            branch_name, node_name=self.node_name, tenant_id=self.tenant_id, lsn=lsn, port=self.port
+        if endpoint_id is None:
+            endpoint_id = self.env.generate_endpoint_id()
+        self.endpoint_id = endpoint_id
+        self.branch_name = branch_name
+
+        self.env.neon_cli.endpoint_create(
+            branch_name,
+            endpoint_id=self.endpoint_id,
+            tenant_id=self.tenant_id,
+            lsn=lsn,
+            hot_standby=hot_standby,
+            port=self.port,
         )
-        path = Path("pgdatadirs") / "tenants" / str(self.tenant_id) / self.node_name
+        path = Path("endpoints") / self.endpoint_id / "pgdata"
         self.pgdata_dir = os.path.join(self.env.repo_dir, path)
 
         if config_lines is None:
@@ -2215,26 +2259,30 @@ class Postgres(PgProtocol):
 
         return self
 
-    def start(self) -> "Postgres":
+    def start(self) -> "Endpoint":
         """
         Start the Postgres instance.
         Returns self.
         """
 
-        assert self.node_name is not None
+        assert self.endpoint_id is not None
 
-        log.info(f"Starting postgres node {self.node_name}")
+        log.info(f"Starting postgres endpoint {self.endpoint_id}")
 
-        self.env.neon_cli.pg_start(self.node_name, tenant_id=self.tenant_id, port=self.port)
+        self.env.neon_cli.endpoint_start(self.endpoint_id, tenant_id=self.tenant_id, port=self.port)
         self.running = True
 
         return self
 
+    def endpoint_path(self) -> Path:
+        """Path to endpoint directory"""
+        assert self.endpoint_id
+        path = Path("endpoints") / self.endpoint_id
+        return self.env.repo_dir / path
+
     def pg_data_dir_path(self) -> str:
-        """Path to data directory"""
-        assert self.node_name
-        path = Path("pgdatadirs") / "tenants" / str(self.tenant_id) / self.node_name
-        return os.path.join(self.env.repo_dir, path)
+        """Path to Postgres data directory"""
+        return os.path.join(self.endpoint_path(), "pgdata")
 
     def pg_xact_dir_path(self) -> str:
         """Path to pg_xact dir"""
@@ -2248,7 +2296,7 @@ class Postgres(PgProtocol):
         """Path to postgresql.conf"""
         return os.path.join(self.pg_data_dir_path(), "postgresql.conf")
 
-    def adjust_for_safekeepers(self, safekeepers: str) -> "Postgres":
+    def adjust_for_safekeepers(self, safekeepers: str) -> "Endpoint":
         """
         Adjust instance config for working with wal acceptors instead of
         pageserver (pre-configured by CLI) directly.
@@ -2272,7 +2320,7 @@ class Postgres(PgProtocol):
             f.write("neon.safekeepers = '{}'\n".format(safekeepers))
         return self
 
-    def config(self, lines: List[str]) -> "Postgres":
+    def config(self, lines: List[str]) -> "Endpoint":
         """
         Add lines to postgresql.conf.
         Lines should be an array of valid postgresql.conf rows.
@@ -2286,32 +2334,32 @@ class Postgres(PgProtocol):
 
         return self
 
-    def stop(self) -> "Postgres":
+    def stop(self) -> "Endpoint":
         """
         Stop the Postgres instance if it's running.
         Returns self.
         """
 
         if self.running:
-            assert self.node_name is not None
-            self.env.neon_cli.pg_stop(
-                self.node_name, self.tenant_id, check_return_code=self.check_stop_result
+            assert self.endpoint_id is not None
+            self.env.neon_cli.endpoint_stop(
+                self.endpoint_id, self.tenant_id, check_return_code=self.check_stop_result
             )
             self.running = False
 
         return self
 
-    def stop_and_destroy(self) -> "Postgres":
+    def stop_and_destroy(self) -> "Endpoint":
         """
-        Stop the Postgres instance, then destroy it.
+        Stop the Postgres instance, then destroy the endpoint.
         Returns self.
         """
 
-        assert self.node_name is not None
-        self.env.neon_cli.pg_stop(
-            self.node_name, self.tenant_id, True, check_return_code=self.check_stop_result
+        assert self.endpoint_id is not None
+        self.env.neon_cli.endpoint_stop(
+            self.endpoint_id, self.tenant_id, True, check_return_code=self.check_stop_result
         )
-        self.node_name = None
+        self.endpoint_id = None
         self.running = False
 
         return self
@@ -2319,13 +2367,13 @@ class Postgres(PgProtocol):
     def create_start(
         self,
         branch_name: str,
-        node_name: Optional[str] = None,
+        endpoint_id: Optional[str] = None,
+        hot_standby: bool = False,
         lsn: Optional[Lsn] = None,
         config_lines: Optional[List[str]] = None,
-    ) -> "Postgres":
+    ) -> "Endpoint":
         """
-        Create a Postgres instance, apply config
-        and then start it.
+        Create an endpoint, apply config, and start Postgres.
         Returns self.
         """
 
@@ -2333,8 +2381,9 @@ class Postgres(PgProtocol):
 
         self.create(
             branch_name=branch_name,
-            node_name=node_name,
+            endpoint_id=endpoint_id,
             config_lines=config_lines,
+            hot_standby=hot_standby,
             lsn=lsn,
         ).start()
 
@@ -2342,7 +2391,7 @@ class Postgres(PgProtocol):
 
         return self
 
-    def __enter__(self) -> "Postgres":
+    def __enter__(self) -> "Endpoint":
         return self
 
     def __exit__(
@@ -2354,33 +2403,35 @@ class Postgres(PgProtocol):
         self.stop()
 
 
-class PostgresFactory:
-    """An object representing multiple running postgres daemons."""
+class EndpointFactory:
+    """An object representing multiple compute endpoints."""
 
     def __init__(self, env: NeonEnv):
         self.env = env
         self.num_instances: int = 0
-        self.instances: List[Postgres] = []
+        self.endpoints: List[Endpoint] = []
 
     def create_start(
         self,
         branch_name: str,
-        node_name: Optional[str] = None,
+        endpoint_id: Optional[str] = None,
         tenant_id: Optional[TenantId] = None,
         lsn: Optional[Lsn] = None,
+        hot_standby: bool = False,
         config_lines: Optional[List[str]] = None,
-    ) -> Postgres:
-        pg = Postgres(
+    ) -> Endpoint:
+        ep = Endpoint(
             self.env,
             tenant_id=tenant_id or self.env.initial_tenant,
             port=self.env.port_distributor.get_port(),
         )
         self.num_instances += 1
-        self.instances.append(pg)
+        self.endpoints.append(ep)
 
-        return pg.create_start(
+        return ep.create_start(
             branch_name=branch_name,
-            node_name=node_name,
+            endpoint_id=endpoint_id,
+            hot_standby=hot_standby,
             config_lines=config_lines,
             lsn=lsn,
         )
@@ -2388,33 +2439,68 @@ class PostgresFactory:
     def create(
         self,
         branch_name: str,
-        node_name: Optional[str] = None,
+        endpoint_id: Optional[str] = None,
         tenant_id: Optional[TenantId] = None,
         lsn: Optional[Lsn] = None,
+        hot_standby: bool = False,
         config_lines: Optional[List[str]] = None,
-    ) -> Postgres:
-        pg = Postgres(
+    ) -> Endpoint:
+        ep = Endpoint(
             self.env,
             tenant_id=tenant_id or self.env.initial_tenant,
             port=self.env.port_distributor.get_port(),
         )
 
-        self.num_instances += 1
-        self.instances.append(pg)
+        if endpoint_id is None:
+            endpoint_id = self.env.generate_endpoint_id()
 
-        return pg.create(
+        self.num_instances += 1
+        self.endpoints.append(ep)
+
+        return ep.create(
             branch_name=branch_name,
-            node_name=node_name,
+            endpoint_id=endpoint_id,
             lsn=lsn,
+            hot_standby=hot_standby,
             config_lines=config_lines,
         )
 
-    def stop_all(self) -> "PostgresFactory":
-        for pg in self.instances:
-            pg.stop()
+    def stop_all(self) -> "EndpointFactory":
+        for ep in self.endpoints:
+            ep.stop()
 
         return self
 
+    def new_replica(self, origin: Endpoint, endpoint_id: str, config_lines: Optional[List[str]]):
+        branch_name = origin.branch_name
+        assert origin in self.endpoints
+        assert branch_name is not None
+
+        return self.create(
+            branch_name=branch_name,
+            endpoint_id=endpoint_id,
+            tenant_id=origin.tenant_id,
+            lsn=None,
+            hot_standby=True,
+            config_lines=config_lines,
+        )
+
+    def new_replica_start(
+        self, origin: Endpoint, endpoint_id: str, config_lines: Optional[List[str]] = None
+    ):
+        branch_name = origin.branch_name
+        assert origin in self.endpoints
+        assert branch_name is not None
+
+        return self.create_start(
+            branch_name=branch_name,
+            endpoint_id=endpoint_id,
+            tenant_id=origin.tenant_id,
+            lsn=None,
+            hot_standby=True,
+            config_lines=config_lines,
+        )
+
 
 @dataclass
 class SafekeeperPort:
@@ -2500,6 +2586,7 @@ class SafekeeperTimelineStatus:
     commit_lsn: Lsn
     timeline_start_lsn: Lsn
     backup_lsn: Lsn
+    peer_horizon_lsn: Lsn
     remote_consistent_lsn: Lsn
 
 
@@ -2557,6 +2644,7 @@ class SafekeeperHttpClient(requests.Session):
             commit_lsn=Lsn(resj["commit_lsn"]),
             timeline_start_lsn=Lsn(resj["timeline_start_lsn"]),
             backup_lsn=Lsn(resj["backup_lsn"]),
+            peer_horizon_lsn=Lsn(resj["peer_horizon_lsn"]),
             remote_consistent_lsn=Lsn(resj["remote_consistent_lsn"]),
         )
 
@@ -2786,16 +2874,16 @@ def list_files_to_compare(pgdata_dir: Path) -> List[str]:
 def check_restored_datadir_content(
     test_output_dir: Path,
     env: NeonEnv,
-    pg: Postgres,
+    endpoint: Endpoint,
 ):
     # Get the timeline ID. We need it for the 'basebackup' command
-    timeline = TimelineId(pg.safe_psql("SHOW neon.timeline_id")[0][0])
+    timeline = TimelineId(endpoint.safe_psql("SHOW neon.timeline_id")[0][0])
 
     # stop postgres to ensure that files won't change
-    pg.stop()
+    endpoint.stop()
 
     # Take a basebackup from pageserver
-    restored_dir_path = env.repo_dir / f"{pg.node_name}_restored_datadir"
+    restored_dir_path = env.repo_dir / f"{endpoint.endpoint_id}_restored_datadir"
     restored_dir_path.mkdir(exist_ok=True)
 
     pg_bin = PgBin(test_output_dir, env.pg_distrib_dir, env.pg_version)
@@ -2805,7 +2893,7 @@ def check_restored_datadir_content(
         {psql_path}                                    \
             --no-psqlrc                                \
             postgres://localhost:{env.pageserver.service_port.pg}  \
-            -c 'basebackup {pg.tenant_id} {timeline}'  \
+            -c 'basebackup {endpoint.tenant_id} {timeline}'  \
          | tar -x -C {restored_dir_path}
     """
 
@@ -2822,8 +2910,8 @@ def check_restored_datadir_content(
     assert result.returncode == 0
 
     # list files we're going to compare
-    assert pg.pgdata_dir
-    pgdata_files = list_files_to_compare(Path(pg.pgdata_dir))
+    assert endpoint.pgdata_dir
+    pgdata_files = list_files_to_compare(Path(endpoint.pgdata_dir))
     restored_files = list_files_to_compare(restored_dir_path)
 
     # check that file sets are equal
@@ -2834,12 +2922,12 @@ def check_restored_datadir_content(
     # We've already filtered all mismatching files in list_files_to_compare(),
     # so here expect that the content is identical
     (match, mismatch, error) = filecmp.cmpfiles(
-        pg.pgdata_dir, restored_dir_path, pgdata_files, shallow=False
+        endpoint.pgdata_dir, restored_dir_path, pgdata_files, shallow=False
     )
     log.info(f"filecmp result mismatch and error lists:\n\t mismatch={mismatch}\n\t error={error}")
 
     for f in mismatch:
-        f1 = os.path.join(pg.pgdata_dir, f)
+        f1 = os.path.join(endpoint.pgdata_dir, f)
         f2 = os.path.join(restored_dir_path, f)
         stdout_filename = "{}.filediff".format(f2)
 
@@ -2854,24 +2942,24 @@ def check_restored_datadir_content(
 
 
 def wait_for_last_flush_lsn(
-    env: NeonEnv, pg: Postgres, tenant: TenantId, timeline: TimelineId
+    env: NeonEnv, endpoint: Endpoint, tenant: TenantId, timeline: TimelineId
 ) -> Lsn:
     """Wait for pageserver to catch up the latest flush LSN, returns the last observed lsn."""
-    last_flush_lsn = Lsn(pg.safe_psql("SELECT pg_current_wal_flush_lsn()")[0][0])
+    last_flush_lsn = Lsn(endpoint.safe_psql("SELECT pg_current_wal_flush_lsn()")[0][0])
     return wait_for_last_record_lsn(env.pageserver.http_client(), tenant, timeline, last_flush_lsn)
 
 
 def wait_for_wal_insert_lsn(
-    env: NeonEnv, pg: Postgres, tenant: TenantId, timeline: TimelineId
+    env: NeonEnv, endpoint: Endpoint, tenant: TenantId, timeline: TimelineId
 ) -> Lsn:
     """Wait for pageserver to catch up the latest flush LSN, returns the last observed lsn."""
-    last_flush_lsn = Lsn(pg.safe_psql("SELECT pg_current_wal_insert_lsn()")[0][0])
+    last_flush_lsn = Lsn(endpoint.safe_psql("SELECT pg_current_wal_insert_lsn()")[0][0])
     return wait_for_last_record_lsn(env.pageserver.http_client(), tenant, timeline, last_flush_lsn)
 
 
 def fork_at_current_lsn(
     env: NeonEnv,
-    pg: Postgres,
+    endpoint: Endpoint,
     new_branch_name: str,
     ancestor_branch_name: str,
     tenant_id: Optional[TenantId] = None,
@@ -2881,36 +2969,22 @@ def fork_at_current_lsn(
     The "last LSN" is taken from the given Postgres instance. The pageserver will wait for all the
     the WAL up to that LSN to arrive in the pageserver before creating the branch.
     """
-    current_lsn = pg.safe_psql("SELECT pg_current_wal_lsn()")[0][0]
+    current_lsn = endpoint.safe_psql("SELECT pg_current_wal_lsn()")[0][0]
     return env.neon_cli.create_branch(new_branch_name, ancestor_branch_name, tenant_id, current_lsn)
 
 
-def wait_for_sk_commit_lsn_to_arrive_at_pageserver_last_record_lsn(
-    tenant_id: TenantId,
-    timeline_id: TimelineId,
-    safekeepers: List[Safekeeper],
-    pageserver: NeonPageserver,
-):
-    sk_commit_lsns = [
-        sk.http_client().timeline_status(tenant_id, timeline_id).commit_lsn for sk in safekeepers
-    ]
-    lsn = max(sk_commit_lsns)
-    ps_http = pageserver.http_client()
-    wait_for_last_record_lsn(ps_http, tenant_id, timeline_id, lsn)
-    return lsn
-
-
-def wait_for_sk_commit_lsn_to_reach_remote_storage(
-    tenant_id: TenantId,
-    timeline_id: TimelineId,
-    safekeepers: List[Safekeeper],
-    pageserver: NeonPageserver,
-):
-    lsn = wait_for_sk_commit_lsn_to_arrive_at_pageserver_last_record_lsn(
-        tenant_id, timeline_id, safekeepers, pageserver
-    )
-    ps_http = pageserver.http_client()
+def last_flush_lsn_upload(
+    env: NeonEnv, endpoint: Endpoint, tenant_id: TenantId, timeline_id: TimelineId
+) -> Lsn:
+    """
+    Wait for pageserver to catch to the latest flush LSN of given endpoint,
+    checkpoint pageserver, and wait for it to be uploaded (remote_consistent_lsn
+    reaching flush LSN).
+    """
+    last_flush_lsn = wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
+    ps_http = env.pageserver.http_client()
+    wait_for_last_record_lsn(ps_http, tenant_id, timeline_id, last_flush_lsn)
     # force a checkpoint to trigger upload
     ps_http.timeline_checkpoint(tenant_id, timeline_id)
-    wait_for_upload(ps_http, tenant_id, timeline_id, lsn)
-    return lsn
+    wait_for_upload(ps_http, tenant_id, timeline_id, last_flush_lsn)
+    return last_flush_lsn
diff --git a/test_runner/fixtures/pageserver/http.py b/test_runner/fixtures/pageserver/http.py
index 1e1effe295..cf92aeb6c0 100644
--- a/test_runner/fixtures/pageserver/http.py
+++ b/test_runner/fixtures/pageserver/http.py
@@ -519,6 +519,13 @@ class PageserverHttpClient(requests.Session):
 
         assert res.status_code == 200
 
+    def download_all_layers(self, tenant_id: TenantId, timeline_id: TimelineId):
+        info = self.layer_map_info(tenant_id, timeline_id)
+        for layer in info.historic_layers:
+            if not layer.remote:
+                continue
+            self.download_layer(tenant_id, timeline_id, layer.layer_file_name)
+
     def evict_layer(self, tenant_id: TenantId, timeline_id: TimelineId, layer_name: str):
         res = self.delete(
             f"http://localhost:{self.port}/v1/tenant/{tenant_id}/timeline/{timeline_id}/layer/{layer_name}",
@@ -543,3 +550,13 @@ class PageserverHttpClient(requests.Session):
     def tenant_break(self, tenant_id: TenantId):
         res = self.put(f"http://localhost:{self.port}/v1/tenant/{tenant_id}/break")
         self.verbose_error(res)
+
+    def post_tracing_event(self, level: str, message: str):
+        res = self.post(
+            f"http://localhost:{self.port}/v1/tracing/event",
+            json={
+                "level": level,
+                "message": message,
+            },
+        )
+        self.verbose_error(res)
diff --git a/test_runner/fixtures/pageserver/utils.py b/test_runner/fixtures/pageserver/utils.py
index 65eda5b636..7f8bb40bda 100644
--- a/test_runner/fixtures/pageserver/utils.py
+++ b/test_runner/fixtures/pageserver/utils.py
@@ -1,16 +1,20 @@
 import time
+from typing import Optional
 
 from fixtures.log_helper import log
 from fixtures.pageserver.http import PageserverHttpClient
 from fixtures.types import Lsn, TenantId, TimelineId
 
 
-def assert_tenant_status(
-    pageserver_http: PageserverHttpClient, tenant: TenantId, expected_status: str
+def assert_tenant_state(
+    pageserver_http: PageserverHttpClient,
+    tenant: TenantId,
+    expected_state: str,
+    message: Optional[str] = None,
 ):
     tenant_status = pageserver_http.tenant_status(tenant)
     log.info(f"tenant_status: {tenant_status}")
-    assert tenant_status["state"] == expected_status, tenant_status
+    assert tenant_status["state"]["slug"] == expected_state, message or tenant_status
 
 
 def tenant_exists(pageserver_http: PageserverHttpClient, tenant_id: TenantId):
@@ -50,10 +54,9 @@ def wait_for_upload(
         if current_lsn >= lsn:
             log.info("wait finished")
             return
+        lr_lsn = last_record_lsn(pageserver_http, tenant, timeline)
         log.info(
-            "waiting for remote_consistent_lsn to reach {}, now {}, iteration {}".format(
-                lsn, current_lsn, i + 1
-            )
+            f"waiting for remote_consistent_lsn to reach {lsn}, now {current_lsn}, last_record_lsn={lr_lsn}, iteration {i + 1}"
         )
         time.sleep(1)
     raise Exception(
@@ -68,6 +71,7 @@ def wait_until_tenant_state(
     tenant_id: TenantId,
     expected_state: str,
     iterations: int,
+    period: float = 1.0,
 ) -> bool:
     """
     Does not use `wait_until` for debugging purposes
@@ -76,21 +80,28 @@ def wait_until_tenant_state(
         try:
             tenant = pageserver_http.tenant_status(tenant_id=tenant_id)
             log.debug(f"Tenant {tenant_id} data: {tenant}")
-            if tenant["state"] == expected_state:
+            if tenant["state"]["slug"] == expected_state:
                 return True
         except Exception as e:
             log.debug(f"Tenant {tenant_id} state retrieval failure: {e}")
 
-        time.sleep(1)
+        time.sleep(period)
 
     raise Exception(f"Tenant {tenant_id} did not become {expected_state} in {iterations} seconds")
 
 
 def wait_until_tenant_active(
-    pageserver_http: PageserverHttpClient, tenant_id: TenantId, iterations: int = 30
+    pageserver_http: PageserverHttpClient,
+    tenant_id: TenantId,
+    iterations: int = 30,
+    period: float = 1.0,
 ):
     wait_until_tenant_state(
-        pageserver_http, tenant_id, expected_state="Active", iterations=iterations
+        pageserver_http,
+        tenant_id,
+        expected_state="Active",
+        iterations=iterations,
+        period=period,
     )
 
 
diff --git a/test_runner/fixtures/utils.py b/test_runner/fixtures/utils.py
index 71df74dfba..30acd3f637 100644
--- a/test_runner/fixtures/utils.py
+++ b/test_runner/fixtures/utils.py
@@ -13,6 +13,7 @@ import allure
 from psycopg2.extensions import cursor
 
 from fixtures.log_helper import log
+from fixtures.types import TimelineId
 
 Fn = TypeVar("Fn", bound=Callable[..., Any])
 
@@ -186,11 +187,15 @@ def allure_attach_from_dir(dir: Path):
             allure.attach.file(source, name, attachment_type, extension)
 
 
-DATASOURCE_ID = "xHHYY0dVz"
+GRAFANA_URL = "https://neonprod.grafana.net"
+GRAFANA_EXPLORE_URL = f"{GRAFANA_URL}/explore"
+GRAFANA_TIMELINE_INSPECTOR_DASHBOARD_URL = f"{GRAFANA_URL}/d/8G011dlnk/timeline-inspector"
+LOGS_STAGING_DATASOURCE_ID = "xHHYY0dVz"
 
 
-def allure_add_grafana_links(host: str, start_ms: int, end_ms: int):
+def allure_add_grafana_links(host: str, timeline_id: TimelineId, start_ms: int, end_ms: int):
     """Add links to server logs in Grafana to Allure report"""
+    links = {}
     # We expect host to be in format like ep-divine-night-159320.us-east-2.aws.neon.build
     endpoint_id, region_id, _ = host.split(".", 2)
 
@@ -202,12 +207,12 @@ def allure_add_grafana_links(host: str, start_ms: int, end_ms: int):
     }
 
     params: Dict[str, Any] = {
-        "datasource": DATASOURCE_ID,
+        "datasource": LOGS_STAGING_DATASOURCE_ID,
         "queries": [
             {
                 "expr": "<PUT AN EXPRESSION HERE>",
                 "refId": "A",
-                "datasource": {"type": "loki", "uid": DATASOURCE_ID},
+                "datasource": {"type": "loki", "uid": LOGS_STAGING_DATASOURCE_ID},
                 "editorMode": "code",
                 "queryType": "range",
             }
@@ -220,8 +225,23 @@ def allure_add_grafana_links(host: str, start_ms: int, end_ms: int):
     for name, expr in expressions.items():
         params["queries"][0]["expr"] = expr
         query_string = urlencode({"orgId": 1, "left": json.dumps(params)})
-        link = f"https://neonprod.grafana.net/explore?{query_string}"
+        links[name] = f"{GRAFANA_EXPLORE_URL}?{query_string}"
 
+    timeline_qs = urlencode(
+        {
+            "orgId": 1,
+            "var-environment": "victoria-metrics-aws-dev",
+            "var-timeline_id": timeline_id,
+            "var-endpoint_id": endpoint_id,
+            "var-log_datasource": "grafanacloud-neonstaging-logs",
+            "from": start_ms,
+            "to": end_ms,
+        }
+    )
+    link = f"{GRAFANA_TIMELINE_INSPECTOR_DASHBOARD_URL}?{timeline_qs}"
+    links["Timeline Inspector"] = link
+
+    for name, link in links.items():
         allure.dynamic.link(link, name=name)
         log.info(f"{name}: {link}")
 
diff --git a/test_runner/performance/test_branch_creation.py b/test_runner/performance/test_branch_creation.py
index 16c5438b8f..6edcb8f1f2 100644
--- a/test_runner/performance/test_branch_creation.py
+++ b/test_runner/performance/test_branch_creation.py
@@ -52,13 +52,13 @@ def test_branch_creation_heavy_write(neon_compare: NeonCompare, n_branches: int)
     def run_pgbench(branch: str):
         log.info(f"Start a pgbench workload on branch {branch}")
 
-        pg = env.postgres.create_start(branch, tenant_id=tenant)
-        connstr = pg.connstr()
+        endpoint = env.endpoints.create_start(branch, tenant_id=tenant)
+        connstr = endpoint.connstr()
 
         pg_bin.run_capture(["pgbench", "-i", connstr])
         pg_bin.run_capture(["pgbench", "-c10", "-T10", connstr])
 
-        pg.stop()
+        endpoint.stop()
 
     env.neon_cli.create_branch("b0", tenant_id=tenant)
 
@@ -96,8 +96,8 @@ def test_branch_creation_many(neon_compare: NeonCompare, n_branches: int):
 
     env.neon_cli.create_branch("b0")
 
-    pg = env.postgres.create_start("b0")
-    neon_compare.pg_bin.run_capture(["pgbench", "-i", "-s10", pg.connstr()])
+    endpoint = env.endpoints.create_start("b0")
+    neon_compare.pg_bin.run_capture(["pgbench", "-i", "-s10", endpoint.connstr()])
 
     branch_creation_durations = []
 
@@ -124,15 +124,15 @@ def test_branch_creation_many_relations(neon_compare: NeonCompare):
 
     timeline_id = env.neon_cli.create_branch("root")
 
-    pg = env.postgres.create_start("root")
-    with closing(pg.connect()) as conn:
+    endpoint = env.endpoints.create_start("root")
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             for i in range(10000):
                 cur.execute(f"CREATE TABLE t{i} as SELECT g FROM generate_series(1, 1000) g")
 
     # Wait for the pageserver to finish processing all the pending WALs,
     # as we don't want the LSN wait time to be included during the branch creation
-    flush_lsn = Lsn(pg.safe_psql("SELECT pg_current_wal_flush_lsn()")[0][0])
+    flush_lsn = Lsn(endpoint.safe_psql("SELECT pg_current_wal_flush_lsn()")[0][0])
     wait_for_last_record_lsn(
         env.pageserver.http_client(), env.initial_tenant, timeline_id, flush_lsn
     )
@@ -142,7 +142,7 @@ def test_branch_creation_many_relations(neon_compare: NeonCompare):
 
     # run a concurrent insertion to make the ancestor "busy" during the branch creation
     thread = threading.Thread(
-        target=pg.safe_psql, args=("INSERT INTO t0 VALUES (generate_series(1, 100000))",)
+        target=endpoint.safe_psql, args=("INSERT INTO t0 VALUES (generate_series(1, 100000))",)
     )
     thread.start()
 
diff --git a/test_runner/performance/test_branching.py b/test_runner/performance/test_branching.py
index 4eaec40096..667d1a4c4a 100644
--- a/test_runner/performance/test_branching.py
+++ b/test_runner/performance/test_branching.py
@@ -42,41 +42,41 @@ def test_compare_child_and_root_pgbench_perf(neon_compare: NeonCompare):
         neon_compare.zenbenchmark.record_pg_bench_result(branch, res)
 
     env.neon_cli.create_branch("root")
-    pg_root = env.postgres.create_start("root")
-    pg_bin.run_capture(["pgbench", "-i", pg_root.connstr(), "-s10"])
+    endpoint_root = env.endpoints.create_start("root")
+    pg_bin.run_capture(["pgbench", "-i", endpoint_root.connstr(), "-s10"])
 
-    fork_at_current_lsn(env, pg_root, "child", "root")
+    fork_at_current_lsn(env, endpoint_root, "child", "root")
 
-    pg_child = env.postgres.create_start("child")
+    endpoint_child = env.endpoints.create_start("child")
 
-    run_pgbench_on_branch("root", ["pgbench", "-c10", "-T10", pg_root.connstr()])
-    run_pgbench_on_branch("child", ["pgbench", "-c10", "-T10", pg_child.connstr()])
+    run_pgbench_on_branch("root", ["pgbench", "-c10", "-T10", endpoint_root.connstr()])
+    run_pgbench_on_branch("child", ["pgbench", "-c10", "-T10", endpoint_child.connstr()])
 
 
 def test_compare_child_and_root_write_perf(neon_compare: NeonCompare):
     env = neon_compare.env
     env.neon_cli.create_branch("root")
-    pg_root = env.postgres.create_start("root")
+    endpoint_root = env.endpoints.create_start("root")
 
-    pg_root.safe_psql(
+    endpoint_root.safe_psql(
         "CREATE TABLE foo(key serial primary key, t text default 'foooooooooooooooooooooooooooooooooooooooooooooooooooo')",
     )
 
     env.neon_cli.create_branch("child", "root")
-    pg_child = env.postgres.create_start("child")
+    endpoint_child = env.endpoints.create_start("child")
 
     with neon_compare.record_duration("root_run_duration"):
-        pg_root.safe_psql("INSERT INTO foo SELECT FROM generate_series(1,1000000)")
+        endpoint_root.safe_psql("INSERT INTO foo SELECT FROM generate_series(1,1000000)")
     with neon_compare.record_duration("child_run_duration"):
-        pg_child.safe_psql("INSERT INTO foo SELECT FROM generate_series(1,1000000)")
+        endpoint_child.safe_psql("INSERT INTO foo SELECT FROM generate_series(1,1000000)")
 
 
 def test_compare_child_and_root_read_perf(neon_compare: NeonCompare):
     env = neon_compare.env
     env.neon_cli.create_branch("root")
-    pg_root = env.postgres.create_start("root")
+    endpoint_root = env.endpoints.create_start("root")
 
-    pg_root.safe_psql_many(
+    endpoint_root.safe_psql_many(
         [
             "CREATE TABLE foo(key serial primary key, t text default 'foooooooooooooooooooooooooooooooooooooooooooooooooooo')",
             "INSERT INTO foo SELECT FROM generate_series(1,1000000)",
@@ -84,12 +84,12 @@ def test_compare_child_and_root_read_perf(neon_compare: NeonCompare):
     )
 
     env.neon_cli.create_branch("child", "root")
-    pg_child = env.postgres.create_start("child")
+    endpoint_child = env.endpoints.create_start("child")
 
     with neon_compare.record_duration("root_run_duration"):
-        pg_root.safe_psql("SELECT count(*) from foo")
+        endpoint_root.safe_psql("SELECT count(*) from foo")
     with neon_compare.record_duration("child_run_duration"):
-        pg_child.safe_psql("SELECT count(*) from foo")
+        endpoint_child.safe_psql("SELECT count(*) from foo")
 
 
 # -----------------------------------------------------------------------
diff --git a/test_runner/performance/test_bulk_tenant_create.py b/test_runner/performance/test_bulk_tenant_create.py
index cef7ce0c6b..9b05903cfa 100644
--- a/test_runner/performance/test_bulk_tenant_create.py
+++ b/test_runner/performance/test_bulk_tenant_create.py
@@ -35,14 +35,14 @@ def test_bulk_tenant_create(
         # if use_safekeepers == 'with_sa':
         #    wa_factory.start_n_new(3)
 
-        pg_tenant = env.postgres.create_start(
+        endpoint_tenant = env.endpoints.create_start(
             f"test_bulk_tenant_create_{tenants_count}_{i}", tenant_id=tenant
         )
 
         end = timeit.default_timer()
         time_slices.append(end - start)
 
-        pg_tenant.stop()
+        endpoint_tenant.stop()
 
     zenbenchmark.record(
         "tenant_creation_time",
diff --git a/test_runner/performance/test_bulk_update.py b/test_runner/performance/test_bulk_update.py
index 7aa6f09a40..2ace31a2d7 100644
--- a/test_runner/performance/test_bulk_update.py
+++ b/test_runner/performance/test_bulk_update.py
@@ -18,8 +18,8 @@ def test_bulk_update(neon_env_builder: NeonEnvBuilder, zenbenchmark, fillfactor)
 
     timeline_id = env.neon_cli.create_branch("test_bulk_update")
     tenant_id = env.initial_tenant
-    pg = env.postgres.create_start("test_bulk_update")
-    cur = pg.connect().cursor()
+    endpoint = env.endpoints.create_start("test_bulk_update")
+    cur = endpoint.connect().cursor()
     cur.execute("set statement_timeout=0")
 
     cur.execute(f"create table t(x integer) WITH (fillfactor={fillfactor})")
@@ -28,13 +28,13 @@ def test_bulk_update(neon_env_builder: NeonEnvBuilder, zenbenchmark, fillfactor)
         cur.execute(f"insert into t values (generate_series(1,{n_records}))")
 
     cur.execute("vacuum t")
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
     with zenbenchmark.record_duration("update-no-prefetch"):
         cur.execute("update t set x=x+1")
 
     cur.execute("vacuum t")
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
     with zenbenchmark.record_duration("delete-no-prefetch"):
         cur.execute("delete from t")
@@ -50,13 +50,13 @@ def test_bulk_update(neon_env_builder: NeonEnvBuilder, zenbenchmark, fillfactor)
         cur.execute(f"insert into t2 values (generate_series(1,{n_records}))")
 
     cur.execute("vacuum t2")
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
     with zenbenchmark.record_duration("update-with-prefetch"):
         cur.execute("update t2 set x=x+1")
 
     cur.execute("vacuum t2")
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
     with zenbenchmark.record_duration("delete-with-prefetch"):
         cur.execute("delete from t2")
diff --git a/test_runner/performance/test_compaction.py b/test_runner/performance/test_compaction.py
index 89818ee8bd..326c4f5c6f 100644
--- a/test_runner/performance/test_compaction.py
+++ b/test_runner/performance/test_compaction.py
@@ -33,11 +33,11 @@ def test_compaction(neon_compare: NeonCompare):
 
     # Create some tables, and run a bunch of INSERTs and UPDATes on them,
     # to generate WAL and layers
-    pg = env.postgres.create_start(
+    endpoint = env.endpoints.create_start(
         "main", tenant_id=tenant_id, config_lines=["shared_buffers=512MB"]
     )
 
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             for i in range(100):
                 cur.execute(f"create table tbl{i} (i int, j int);")
@@ -45,7 +45,7 @@ def test_compaction(neon_compare: NeonCompare):
                 for j in range(100):
                     cur.execute(f"update tbl{i} set j = {j};")
 
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
     # First compaction generates L1 layers
     with neon_compare.zenbenchmark.record_duration("compaction"):
diff --git a/test_runner/performance/test_latency.py b/test_runner/performance/test_latency.py
index 257e0421af..6c94ecc482 100644
--- a/test_runner/performance/test_latency.py
+++ b/test_runner/performance/test_latency.py
@@ -2,13 +2,13 @@ import threading
 
 import pytest
 from fixtures.compare_fixtures import PgCompare
-from fixtures.neon_fixtures import Postgres
+from fixtures.neon_fixtures import PgProtocol
 
 from performance.test_perf_pgbench import get_scales_matrix
 from performance.test_wal_backpressure import record_read_latency
 
 
-def start_write_workload(pg: Postgres, scale: int = 10):
+def start_write_workload(pg: PgProtocol, scale: int = 10):
     with pg.connect().cursor() as cur:
         cur.execute(f"create table big as select generate_series(1,{scale*100_000})")
 
diff --git a/test_runner/performance/test_layer_map.py b/test_runner/performance/test_layer_map.py
index fb29c05273..18308e1077 100644
--- a/test_runner/performance/test_layer_map.py
+++ b/test_runner/performance/test_layer_map.py
@@ -25,8 +25,8 @@ def test_layer_map(neon_env_builder: NeonEnvBuilder, zenbenchmark):
     )
 
     env.neon_cli.create_timeline("test_layer_map", tenant_id=tenant)
-    pg = env.postgres.create_start("test_layer_map", tenant_id=tenant)
-    cur = pg.connect().cursor()
+    endpoint = env.endpoints.create_start("test_layer_map", tenant_id=tenant)
+    cur = endpoint.connect().cursor()
     cur.execute("create table t(x integer)")
     for i in range(n_iters):
         cur.execute(f"insert into t values (generate_series(1,{n_records}))")
diff --git a/test_runner/performance/test_startup.py b/test_runner/performance/test_startup.py
index e91b180154..fa2e058491 100644
--- a/test_runner/performance/test_startup.py
+++ b/test_runner/performance/test_startup.py
@@ -14,19 +14,19 @@ def test_startup(neon_env_builder: NeonEnvBuilder, zenbenchmark: NeonBenchmarker
     # Start
     env.neon_cli.create_branch("test_startup")
     with zenbenchmark.record_duration("startup_time"):
-        pg = env.postgres.create_start("test_startup")
-        pg.safe_psql("select 1;")
+        endpoint = env.endpoints.create_start("test_startup")
+        endpoint.safe_psql("select 1;")
 
     # Restart
-    pg.stop_and_destroy()
+    endpoint.stop_and_destroy()
     with zenbenchmark.record_duration("restart_time"):
-        pg.create_start("test_startup")
-        pg.safe_psql("select 1;")
+        endpoint.create_start("test_startup")
+        endpoint.safe_psql("select 1;")
 
     # Fill up
     num_rows = 1000000  # 30 MB
     num_tables = 100
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             for i in range(num_tables):
                 cur.execute(f"create table t_{i} (i integer);")
@@ -34,18 +34,18 @@ def test_startup(neon_env_builder: NeonEnvBuilder, zenbenchmark: NeonBenchmarker
 
     # Read
     with zenbenchmark.record_duration("read_time"):
-        pg.safe_psql("select * from t_0;")
+        endpoint.safe_psql("select * from t_0;")
 
     # Read again
     with zenbenchmark.record_duration("second_read_time"):
-        pg.safe_psql("select * from t_0;")
+        endpoint.safe_psql("select * from t_0;")
 
     # Restart
-    pg.stop_and_destroy()
+    endpoint.stop_and_destroy()
     with zenbenchmark.record_duration("restart_with_data"):
-        pg.create_start("test_startup")
-        pg.safe_psql("select 1;")
+        endpoint.create_start("test_startup")
+        endpoint.safe_psql("select 1;")
 
     # Read
     with zenbenchmark.record_duration("read_after_restart"):
-        pg.safe_psql("select * from t_0;")
+        endpoint.safe_psql("select * from t_0;")
diff --git a/test_runner/regress/test_ancestor_branch.py b/test_runner/regress/test_ancestor_branch.py
index 2406102756..e8c1a2f34c 100644
--- a/test_runner/regress/test_ancestor_branch.py
+++ b/test_runner/regress/test_ancestor_branch.py
@@ -22,8 +22,8 @@ def test_ancestor_branch(neon_env_builder: NeonEnvBuilder):
 
     pageserver_http.configure_failpoints(("flush-frozen-before-sync", "sleep(10000)"))
 
-    pg_branch0 = env.postgres.create_start("main", tenant_id=tenant)
-    branch0_cur = pg_branch0.connect().cursor()
+    endpoint_branch0 = env.endpoints.create_start("main", tenant_id=tenant)
+    branch0_cur = endpoint_branch0.connect().cursor()
     branch0_timeline = TimelineId(query_scalar(branch0_cur, "SHOW neon.timeline_id"))
     log.info(f"b0 timeline {branch0_timeline}")
 
@@ -44,10 +44,10 @@ def test_ancestor_branch(neon_env_builder: NeonEnvBuilder):
 
     # Create branch1.
     env.neon_cli.create_branch("branch1", "main", tenant_id=tenant, ancestor_start_lsn=lsn_100)
-    pg_branch1 = env.postgres.create_start("branch1", tenant_id=tenant)
+    endpoint_branch1 = env.endpoints.create_start("branch1", tenant_id=tenant)
     log.info("postgres is running on 'branch1' branch")
 
-    branch1_cur = pg_branch1.connect().cursor()
+    branch1_cur = endpoint_branch1.connect().cursor()
     branch1_timeline = TimelineId(query_scalar(branch1_cur, "SHOW neon.timeline_id"))
     log.info(f"b1 timeline {branch1_timeline}")
 
@@ -67,9 +67,9 @@ def test_ancestor_branch(neon_env_builder: NeonEnvBuilder):
 
     # Create branch2.
     env.neon_cli.create_branch("branch2", "branch1", tenant_id=tenant, ancestor_start_lsn=lsn_200)
-    pg_branch2 = env.postgres.create_start("branch2", tenant_id=tenant)
+    endpoint_branch2 = env.endpoints.create_start("branch2", tenant_id=tenant)
     log.info("postgres is running on 'branch2' branch")
-    branch2_cur = pg_branch2.connect().cursor()
+    branch2_cur = endpoint_branch2.connect().cursor()
 
     branch2_timeline = TimelineId(query_scalar(branch2_cur, "SHOW neon.timeline_id"))
     log.info(f"b2 timeline {branch2_timeline}")
diff --git a/test_runner/regress/test_auth.py b/test_runner/regress/test_auth.py
index f7c4736e04..3305869dce 100644
--- a/test_runner/regress/test_auth.py
+++ b/test_runner/regress/test_auth.py
@@ -64,9 +64,9 @@ def test_compute_auth_to_pageserver(neon_env_builder: NeonEnvBuilder):
 
     branch = "test_compute_auth_to_pageserver"
     env.neon_cli.create_branch(branch)
-    pg = env.postgres.create_start(branch)
+    endpoint = env.endpoints.create_start(branch)
 
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             # we rely upon autocommit after each statement
             # as waiting for acceptors happens there
@@ -83,7 +83,7 @@ def test_auth_failures(neon_env_builder: NeonEnvBuilder, auth_enabled: bool):
 
     branch = f"test_auth_failures_auth_enabled_{auth_enabled}"
     timeline_id = env.neon_cli.create_branch(branch)
-    env.postgres.create_start(branch)
+    env.endpoints.create_start(branch)
 
     tenant_token = env.auth_keys.generate_tenant_token(env.initial_tenant)
     invalid_tenant_token = env.auth_keys.generate_tenant_token(TenantId.generate())
diff --git a/test_runner/regress/test_backpressure.py b/test_runner/regress/test_backpressure.py
index a81fa380a9..352e149171 100644
--- a/test_runner/regress/test_backpressure.py
+++ b/test_runner/regress/test_backpressure.py
@@ -5,7 +5,7 @@ from contextlib import closing, contextmanager
 import psycopg2.extras
 import pytest
 from fixtures.log_helper import log
-from fixtures.neon_fixtures import NeonEnvBuilder, Postgres
+from fixtures.neon_fixtures import Endpoint, NeonEnvBuilder
 
 pytest_plugins = "fixtures.neon_fixtures"
 
@@ -20,10 +20,10 @@ def pg_cur(pg):
 # Periodically check that all backpressure lags are below the configured threshold,
 # assert if they are not.
 # If the check query fails, stop the thread. Main thread should notice that and stop the test.
-def check_backpressure(pg: Postgres, stop_event: threading.Event, polling_interval=5):
+def check_backpressure(endpoint: Endpoint, stop_event: threading.Event, polling_interval=5):
     log.info("checks started")
 
-    with pg_cur(pg) as cur:
+    with pg_cur(endpoint) as cur:
         cur.execute("CREATE EXTENSION neon")  # TODO move it to neon_fixtures?
 
         cur.execute("select pg_size_bytes(current_setting('max_replication_write_lag'))")
@@ -41,7 +41,7 @@ def check_backpressure(pg: Postgres, stop_event: threading.Event, polling_interv
         max_replication_apply_lag_bytes = res[0]
         log.info(f"max_replication_apply_lag: {max_replication_apply_lag_bytes} bytes")
 
-    with pg_cur(pg) as cur:
+    with pg_cur(endpoint) as cur:
         while not stop_event.is_set():
             try:
                 cur.execute(
@@ -102,14 +102,14 @@ def test_backpressure_received_lsn_lag(neon_env_builder: NeonEnvBuilder):
     # Create a branch for us
     env.neon_cli.create_branch("test_backpressure")
 
-    pg = env.postgres.create_start(
+    endpoint = env.endpoints.create_start(
         "test_backpressure", config_lines=["max_replication_write_lag=30MB"]
     )
     log.info("postgres is running on 'test_backpressure' branch")
 
     # setup check thread
     check_stop_event = threading.Event()
-    check_thread = threading.Thread(target=check_backpressure, args=(pg, check_stop_event))
+    check_thread = threading.Thread(target=check_backpressure, args=(endpoint, check_stop_event))
     check_thread.start()
 
     # Configure failpoint to slow down walreceiver ingest
@@ -125,7 +125,7 @@ def test_backpressure_received_lsn_lag(neon_env_builder: NeonEnvBuilder):
     # because of the lag and waiting for lsn to replay to arrive.
     time.sleep(2)
 
-    with pg_cur(pg) as cur:
+    with pg_cur(endpoint) as cur:
         # Create and initialize test table
         cur.execute("CREATE TABLE foo(x bigint)")
 
diff --git a/test_runner/regress/test_basebackup_error.py b/test_runner/regress/test_basebackup_error.py
index 94d3999d17..170b494884 100644
--- a/test_runner/regress/test_basebackup_error.py
+++ b/test_runner/regress/test_basebackup_error.py
@@ -15,4 +15,4 @@ def test_basebackup_error(neon_simple_env: NeonEnv):
     pageserver_http.configure_failpoints(("basebackup-before-control-file", "return"))
 
     with pytest.raises(Exception, match="basebackup-before-control-file"):
-        env.postgres.create_start("test_basebackup_error")
+        env.endpoints.create_start("test_basebackup_error")
diff --git a/test_runner/regress/test_branch_and_gc.py b/test_runner/regress/test_branch_and_gc.py
index cc807b7ff3..4a03421fcf 100644
--- a/test_runner/regress/test_branch_and_gc.py
+++ b/test_runner/regress/test_branch_and_gc.py
@@ -67,9 +67,9 @@ def test_branch_and_gc(neon_simple_env: NeonEnv):
     )
 
     timeline_main = env.neon_cli.create_timeline("test_main", tenant_id=tenant)
-    pg_main = env.postgres.create_start("test_main", tenant_id=tenant)
+    endpoint_main = env.endpoints.create_start("test_main", tenant_id=tenant)
 
-    main_cur = pg_main.connect().cursor()
+    main_cur = endpoint_main.connect().cursor()
 
     main_cur.execute(
         "CREATE TABLE foo(key serial primary key, t text default 'foooooooooooooooooooooooooooooooooooooooooooooooooooo')"
@@ -90,9 +90,9 @@ def test_branch_and_gc(neon_simple_env: NeonEnv):
     env.neon_cli.create_branch(
         "test_branch", "test_main", tenant_id=tenant, ancestor_start_lsn=lsn1
     )
-    pg_branch = env.postgres.create_start("test_branch", tenant_id=tenant)
+    endpoint_branch = env.endpoints.create_start("test_branch", tenant_id=tenant)
 
-    branch_cur = pg_branch.connect().cursor()
+    branch_cur = endpoint_branch.connect().cursor()
     branch_cur.execute("INSERT INTO foo SELECT FROM generate_series(1, 100000)")
 
     assert query_scalar(branch_cur, "SELECT count(*) FROM foo") == 200000
@@ -142,8 +142,8 @@ def test_branch_creation_before_gc(neon_simple_env: NeonEnv):
     )
 
     b0 = env.neon_cli.create_branch("b0", tenant_id=tenant)
-    pg0 = env.postgres.create_start("b0", tenant_id=tenant)
-    res = pg0.safe_psql_many(
+    endpoint0 = env.endpoints.create_start("b0", tenant_id=tenant)
+    res = endpoint0.safe_psql_many(
         queries=[
             "CREATE TABLE t(key serial primary key)",
             "INSERT INTO t SELECT FROM generate_series(1, 100000)",
diff --git a/test_runner/regress/test_branch_behind.py b/test_runner/regress/test_branch_behind.py
index d19f6a7d39..3f7d49ab03 100644
--- a/test_runner/regress/test_branch_behind.py
+++ b/test_runner/regress/test_branch_behind.py
@@ -18,10 +18,10 @@ def test_branch_behind(neon_env_builder: NeonEnvBuilder):
 
     # Branch at the point where only 100 rows were inserted
     env.neon_cli.create_branch("test_branch_behind")
-    pgmain = env.postgres.create_start("test_branch_behind")
+    endpoint_main = env.endpoints.create_start("test_branch_behind")
     log.info("postgres is running on 'test_branch_behind' branch")
 
-    main_cur = pgmain.connect().cursor()
+    main_cur = endpoint_main.connect().cursor()
 
     timeline = TimelineId(query_scalar(main_cur, "SHOW neon.timeline_id"))
 
@@ -74,15 +74,15 @@ def test_branch_behind(neon_env_builder: NeonEnvBuilder):
         "test_branch_behind_more", "test_branch_behind", ancestor_start_lsn=lsn_b
     )
 
-    pg_hundred = env.postgres.create_start("test_branch_behind_hundred")
-    pg_more = env.postgres.create_start("test_branch_behind_more")
+    endpoint_hundred = env.endpoints.create_start("test_branch_behind_hundred")
+    endpoint_more = env.endpoints.create_start("test_branch_behind_more")
 
     # On the 'hundred' branch, we should see only 100 rows
-    hundred_cur = pg_hundred.connect().cursor()
+    hundred_cur = endpoint_hundred.connect().cursor()
     assert query_scalar(hundred_cur, "SELECT count(*) FROM foo") == 100
 
     # On the 'more' branch, we should see 100200 rows
-    more_cur = pg_more.connect().cursor()
+    more_cur = endpoint_more.connect().cursor()
     assert query_scalar(more_cur, "SELECT count(*) FROM foo") == 200100
 
     # All the rows are visible on the main branch
@@ -94,8 +94,8 @@ def test_branch_behind(neon_env_builder: NeonEnvBuilder):
     env.neon_cli.create_branch(
         "test_branch_segment_boundary", "test_branch_behind", ancestor_start_lsn=Lsn("0/3000000")
     )
-    pg = env.postgres.create_start("test_branch_segment_boundary")
-    assert pg.safe_psql("SELECT 1")[0][0] == 1
+    endpoint = env.endpoints.create_start("test_branch_segment_boundary")
+    assert endpoint.safe_psql("SELECT 1")[0][0] == 1
 
     # branch at pre-initdb lsn
     with pytest.raises(Exception, match="invalid branch start lsn: .*"):
diff --git a/test_runner/regress/test_branching.py b/test_runner/regress/test_branching.py
index 3b78700e9f..31f9df6ebe 100644
--- a/test_runner/regress/test_branching.py
+++ b/test_runner/regress/test_branching.py
@@ -5,7 +5,7 @@ from typing import List
 
 import pytest
 from fixtures.log_helper import log
-from fixtures.neon_fixtures import NeonEnv, PgBin, Postgres
+from fixtures.neon_fixtures import Endpoint, NeonEnv, PgBin
 from fixtures.types import Lsn
 from fixtures.utils import query_scalar
 from performance.test_perf_pgbench import get_scales_matrix
@@ -40,20 +40,20 @@ def test_branching_with_pgbench(
         }
     )
 
-    def run_pgbench(pg: Postgres):
-        connstr = pg.connstr()
-
+    def run_pgbench(connstr: str):
         log.info(f"Start a pgbench workload on pg {connstr}")
 
         pg_bin.run_capture(["pgbench", "-i", f"-s{scale}", connstr])
         pg_bin.run_capture(["pgbench", "-T15", connstr])
 
     env.neon_cli.create_branch("b0", tenant_id=tenant)
-    pgs: List[Postgres] = []
-    pgs.append(env.postgres.create_start("b0", tenant_id=tenant))
+    endpoints: List[Endpoint] = []
+    endpoints.append(env.endpoints.create_start("b0", tenant_id=tenant))
 
     threads: List[threading.Thread] = []
-    threads.append(threading.Thread(target=run_pgbench, args=(pgs[0],), daemon=True))
+    threads.append(
+        threading.Thread(target=run_pgbench, args=(endpoints[0].connstr(),), daemon=True)
+    )
     threads[-1].start()
 
     thread_limit = 4
@@ -79,16 +79,18 @@ def test_branching_with_pgbench(
         else:
             env.neon_cli.create_branch("b{}".format(i + 1), "b0", tenant_id=tenant)
 
-        pgs.append(env.postgres.create_start("b{}".format(i + 1), tenant_id=tenant))
+        endpoints.append(env.endpoints.create_start("b{}".format(i + 1), tenant_id=tenant))
 
-        threads.append(threading.Thread(target=run_pgbench, args=(pgs[-1],), daemon=True))
+        threads.append(
+            threading.Thread(target=run_pgbench, args=(endpoints[-1].connstr(),), daemon=True)
+        )
         threads[-1].start()
 
     for thread in threads:
         thread.join()
 
-    for pg in pgs:
-        res = pg.safe_psql("SELECT count(*) from pgbench_accounts")
+    for ep in endpoints:
+        res = ep.safe_psql("SELECT count(*) from pgbench_accounts")
         assert res[0] == (100000 * scale,)
 
 
@@ -110,11 +112,11 @@ def test_branching_unnormalized_start_lsn(neon_simple_env: NeonEnv, pg_bin: PgBi
     env = neon_simple_env
 
     env.neon_cli.create_branch("b0")
-    pg0 = env.postgres.create_start("b0")
+    endpoint0 = env.endpoints.create_start("b0")
 
-    pg_bin.run_capture(["pgbench", "-i", pg0.connstr()])
+    pg_bin.run_capture(["pgbench", "-i", endpoint0.connstr()])
 
-    with pg0.cursor() as cur:
+    with endpoint0.cursor() as cur:
         curr_lsn = Lsn(query_scalar(cur, "SELECT pg_current_wal_flush_lsn()"))
 
     # Specify the `start_lsn` as a number that is divided by `XLOG_BLCKSZ`
@@ -123,6 +125,6 @@ def test_branching_unnormalized_start_lsn(neon_simple_env: NeonEnv, pg_bin: PgBi
 
     log.info(f"Branching b1 from b0 starting at lsn {start_lsn}...")
     env.neon_cli.create_branch("b1", "b0", ancestor_start_lsn=start_lsn)
-    pg1 = env.postgres.create_start("b1")
+    endpoint1 = env.endpoints.create_start("b1")
 
-    pg_bin.run_capture(["pgbench", "-i", pg1.connstr()])
+    pg_bin.run_capture(["pgbench", "-i", endpoint1.connstr()])
diff --git a/test_runner/regress/test_broken_timeline.py b/test_runner/regress/test_broken_timeline.py
index d12a0223a1..fb592bfbc3 100644
--- a/test_runner/regress/test_broken_timeline.py
+++ b/test_runner/regress/test_broken_timeline.py
@@ -4,7 +4,7 @@ from typing import List, Tuple
 
 import pytest
 from fixtures.log_helper import log
-from fixtures.neon_fixtures import NeonEnv, NeonEnvBuilder, Postgres
+from fixtures.neon_fixtures import Endpoint, NeonEnv, NeonEnvBuilder
 from fixtures.types import TenantId, TimelineId
 
 
@@ -24,17 +24,17 @@ def test_broken_timeline(neon_env_builder: NeonEnvBuilder):
         ]
     )
 
-    tenant_timelines: List[Tuple[TenantId, TimelineId, Postgres]] = []
+    tenant_timelines: List[Tuple[TenantId, TimelineId, Endpoint]] = []
 
     for n in range(4):
         tenant_id, timeline_id = env.neon_cli.create_tenant()
 
-        pg = env.postgres.create_start("main", tenant_id=tenant_id)
-        with pg.cursor() as cur:
+        endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
+        with endpoint.cursor() as cur:
             cur.execute("CREATE TABLE t(key int primary key, value text)")
             cur.execute("INSERT INTO t SELECT generate_series(1,100), 'payload'")
-        pg.stop()
-        tenant_timelines.append((tenant_id, timeline_id, pg))
+        endpoint.stop()
+        tenant_timelines.append((tenant_id, timeline_id, endpoint))
 
     # Stop the pageserver
     env.pageserver.stop()
diff --git a/test_runner/regress/test_clog_truncate.py b/test_runner/regress/test_clog_truncate.py
index f47e4a99bf..f22eca02cc 100644
--- a/test_runner/regress/test_clog_truncate.py
+++ b/test_runner/regress/test_clog_truncate.py
@@ -24,14 +24,14 @@ def test_clog_truncate(neon_simple_env: NeonEnv):
         "autovacuum_freeze_max_age=100000",
     ]
 
-    pg = env.postgres.create_start("test_clog_truncate", config_lines=config)
+    endpoint = env.endpoints.create_start("test_clog_truncate", config_lines=config)
     log.info("postgres is running on test_clog_truncate branch")
 
     # Install extension containing function needed for test
-    pg.safe_psql("CREATE EXTENSION neon_test_utils")
+    endpoint.safe_psql("CREATE EXTENSION neon_test_utils")
 
     # Consume many xids to advance clog
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("select test_consume_xids(1000*1000*10);")
         log.info("xids consumed")
 
@@ -44,7 +44,7 @@ def test_clog_truncate(neon_simple_env: NeonEnv):
 
     # wait for autovacuum to truncate the pg_xact
     # XXX Is it worth to add a timeout here?
-    pg_xact_0000_path = os.path.join(pg.pg_xact_dir_path(), "0000")
+    pg_xact_0000_path = os.path.join(endpoint.pg_xact_dir_path(), "0000")
     log.info(f"pg_xact_0000_path = {pg_xact_0000_path}")
 
     while os.path.isfile(pg_xact_0000_path):
@@ -52,7 +52,7 @@ def test_clog_truncate(neon_simple_env: NeonEnv):
         time.sleep(5)
 
     # checkpoint to advance latest lsn
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("CHECKPOINT;")
         lsn_after_truncation = query_scalar(cur, "select pg_current_wal_insert_lsn()")
 
@@ -61,10 +61,10 @@ def test_clog_truncate(neon_simple_env: NeonEnv):
     env.neon_cli.create_branch(
         "test_clog_truncate_new", "test_clog_truncate", ancestor_start_lsn=lsn_after_truncation
     )
-    pg2 = env.postgres.create_start("test_clog_truncate_new")
+    endpoint2 = env.endpoints.create_start("test_clog_truncate_new")
     log.info("postgres is running on test_clog_truncate_new branch")
 
     # check that new node doesn't contain truncated segment
-    pg_xact_0000_path_new = os.path.join(pg2.pg_xact_dir_path(), "0000")
+    pg_xact_0000_path_new = os.path.join(endpoint2.pg_xact_dir_path(), "0000")
     log.info(f"pg_xact_0000_path_new = {pg_xact_0000_path_new}")
     assert os.path.isfile(pg_xact_0000_path_new) is False
diff --git a/test_runner/regress/test_close_fds.py b/test_runner/regress/test_close_fds.py
index 22f245f79b..7059f3360e 100644
--- a/test_runner/regress/test_close_fds.py
+++ b/test_runner/regress/test_close_fds.py
@@ -24,8 +24,8 @@ def test_lsof_pageserver_pid(neon_simple_env: NeonEnv):
 
     def start_workload():
         env.neon_cli.create_branch("test_lsof_pageserver_pid")
-        pg = env.postgres.create_start("test_lsof_pageserver_pid")
-        with closing(pg.connect()) as conn:
+        endpoint = env.endpoints.create_start("test_lsof_pageserver_pid")
+        with closing(endpoint.connect()) as conn:
             with conn.cursor() as cur:
                 cur.execute("CREATE TABLE foo as SELECT x FROM generate_series(1,100000) x")
                 cur.execute("update foo set x=x+1")
diff --git a/test_runner/regress/test_compatibility.py b/test_runner/regress/test_compatibility.py
index 0cc111bd8c..e262202a73 100644
--- a/test_runner/regress/test_compatibility.py
+++ b/test_runner/regress/test_compatibility.py
@@ -1,3 +1,4 @@
+import copy
 import os
 import shutil
 import subprocess
@@ -55,29 +56,31 @@ def test_create_snapshot(neon_env_builder: NeonEnvBuilder, pg_bin: PgBin, test_o
     neon_env_builder.preserve_database_files = True
 
     env = neon_env_builder.init_start()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     # FIXME: Is this expected?
     env.pageserver.allowed_errors.append(
         ".*init_tenant_mgr: marking .* as locally complete, while it doesnt exist in remote index.*"
     )
 
-    pg_bin.run(["pgbench", "--initialize", "--scale=10", pg.connstr()])
-    pg_bin.run(["pgbench", "--time=60", "--progress=2", pg.connstr()])
-    pg_bin.run(["pg_dumpall", f"--dbname={pg.connstr()}", f"--file={test_output_dir / 'dump.sql'}"])
+    pg_bin.run(["pgbench", "--initialize", "--scale=10", endpoint.connstr()])
+    pg_bin.run(["pgbench", "--time=60", "--progress=2", endpoint.connstr()])
+    pg_bin.run(
+        ["pg_dumpall", f"--dbname={endpoint.connstr()}", f"--file={test_output_dir / 'dump.sql'}"]
+    )
 
     snapshot_config = toml.load(test_output_dir / "repo" / "config")
     tenant_id = snapshot_config["default_tenant_id"]
     timeline_id = dict(snapshot_config["branch_name_mappings"]["main"])[tenant_id]
 
     pageserver_http = env.pageserver.http_client()
-    lsn = Lsn(pg.safe_psql("SELECT pg_current_wal_flush_lsn()")[0][0])
+    lsn = Lsn(endpoint.safe_psql("SELECT pg_current_wal_flush_lsn()")[0][0])
 
     wait_for_last_record_lsn(pageserver_http, tenant_id, timeline_id, lsn)
     pageserver_http.timeline_checkpoint(tenant_id, timeline_id)
     wait_for_upload(pageserver_http, tenant_id, timeline_id, lsn)
 
-    env.postgres.stop_all()
+    env.endpoints.stop_all()
     for sk in env.safekeepers:
         sk.stop()
     env.pageserver.stop()
@@ -98,6 +101,9 @@ def test_backward_compatibility(
     pg_version: str,
     request: FixtureRequest,
 ):
+    """
+    Test that the new binaries can read old data
+    """
     compatibility_snapshot_dir_env = os.environ.get("COMPATIBILITY_SNAPSHOT_DIR")
     assert (
         compatibility_snapshot_dir_env is not None
@@ -120,6 +126,7 @@ def test_backward_compatibility(
         check_neon_works(
             test_output_dir / "compatibility_snapshot" / "repo",
             neon_binpath,
+            neon_binpath,
             pg_distrib_dir,
             pg_version,
             port_distributor,
@@ -148,7 +155,11 @@ def test_forward_compatibility(
     port_distributor: PortDistributor,
     pg_version: str,
     request: FixtureRequest,
+    neon_binpath: Path,
 ):
+    """
+    Test that the old binaries can read new data
+    """
     compatibility_neon_bin_env = os.environ.get("COMPATIBILITY_NEON_BIN")
     assert compatibility_neon_bin_env is not None, (
         "COMPATIBILITY_NEON_BIN is not set. It should be set to a path with Neon binaries "
@@ -183,6 +194,7 @@ def test_forward_compatibility(
         check_neon_works(
             test_output_dir / "compatibility_snapshot" / "repo",
             compatibility_neon_bin,
+            neon_binpath,
             compatibility_postgres_distrib_dir,
             pg_version,
             port_distributor,
@@ -223,9 +235,13 @@ def prepare_snapshot(
     for logfile in repo_dir.glob("**/*.log"):
         logfile.unlink()
 
-    # Remove tenants data for compute
-    for tenant in (repo_dir / "pgdatadirs" / "tenants").glob("*"):
-        shutil.rmtree(tenant)
+    # Remove old computes in 'endpoints'. Old versions of the control plane used a directory
+    # called "pgdatadirs". Delete it, too.
+    if (repo_dir / "endpoints").exists():
+        shutil.rmtree(repo_dir / "endpoints")
+    if (repo_dir / "pgdatadirs").exists():
+        shutil.rmtree(repo_dir / "pgdatadirs")
+    os.mkdir(repo_dir / "endpoints")
 
     # Remove wal-redo temp directory if it exists. Newer pageserver versions don't create
     # them anymore, but old versions did.
@@ -326,7 +342,8 @@ def get_neon_version(neon_binpath: Path):
 
 def check_neon_works(
     repo_dir: Path,
-    neon_binpath: Path,
+    neon_target_binpath: Path,
+    neon_current_binpath: Path,
     pg_distrib_dir: Path,
     pg_version: str,
     port_distributor: PortDistributor,
@@ -336,7 +353,7 @@ def check_neon_works(
 ):
     snapshot_config_toml = repo_dir / "config"
     snapshot_config = toml.load(snapshot_config_toml)
-    snapshot_config["neon_distrib_dir"] = str(neon_binpath)
+    snapshot_config["neon_distrib_dir"] = str(neon_target_binpath)
     snapshot_config["postgres_distrib_dir"] = str(pg_distrib_dir)
     with (snapshot_config_toml).open("w") as f:
         toml.dump(snapshot_config, f)
@@ -347,17 +364,25 @@ def check_neon_works(
     config.repo_dir = repo_dir
     config.pg_version = pg_version
     config.initial_tenant = snapshot_config["default_tenant_id"]
-    config.neon_binpath = neon_binpath
     config.pg_distrib_dir = pg_distrib_dir
     config.preserve_database_files = True
 
-    cli = NeonCli(config)
-    cli.raw_cli(["start"])
-    request.addfinalizer(lambda: cli.raw_cli(["stop"]))
+    # Use the "target" binaries to launch the storage nodes
+    config_target = config
+    config_target.neon_binpath = neon_target_binpath
+    cli_target = NeonCli(config_target)
+
+    # And the current binaries to launch computes
+    config_current = copy.copy(config)
+    config_current.neon_binpath = neon_current_binpath
+    cli_current = NeonCli(config_current)
+
+    cli_target.raw_cli(["start"])
+    request.addfinalizer(lambda: cli_target.raw_cli(["stop"]))
 
     pg_port = port_distributor.get_port()
-    cli.pg_start("main", port=pg_port)
-    request.addfinalizer(lambda: cli.pg_stop("main"))
+    cli_current.endpoint_start("main", port=pg_port)
+    request.addfinalizer(lambda: cli_current.endpoint_stop("main"))
 
     connstr = f"host=127.0.0.1 port={pg_port} user=cloud_admin dbname=postgres"
     pg_bin.run(["pg_dumpall", f"--dbname={connstr}", f"--file={test_output_dir / 'dump.sql'}"])
diff --git a/test_runner/regress/test_compute_ctl.py b/test_runner/regress/test_compute_ctl.py
index 05ac3841dc..d72ffe078d 100644
--- a/test_runner/regress/test_compute_ctl.py
+++ b/test_runner/regress/test_compute_ctl.py
@@ -13,10 +13,10 @@ def test_sync_safekeepers_logs(neon_env_builder: NeonEnvBuilder, pg_bin: PgBin):
     ctl = ComputeCtl(env)
 
     env.neon_cli.create_branch("test_compute_ctl", "main")
-    pg = env.postgres.create_start("test_compute_ctl")
-    pg.safe_psql("CREATE TABLE t(key int primary key, value text)")
+    endpoint = env.endpoints.create_start("test_compute_ctl")
+    endpoint.safe_psql("CREATE TABLE t(key int primary key, value text)")
 
-    with open(pg.config_file_path(), "r") as f:
+    with open(endpoint.config_file_path(), "r") as f:
         cfg_lines = f.readlines()
     cfg_map = {}
     for line in cfg_lines:
@@ -24,10 +24,13 @@ def test_sync_safekeepers_logs(neon_env_builder: NeonEnvBuilder, pg_bin: PgBin):
             k, v = line.split("=")
             cfg_map[k] = v.strip("\n '\"")
     log.info(f"postgres config: {cfg_map}")
-    pgdata = pg.pg_data_dir_path()
+    pgdata = endpoint.pg_data_dir_path()
     pg_bin_path = os.path.join(pg_bin.pg_bin_path, "postgres")
 
-    pg.stop_and_destroy()
+    endpoint.stop_and_destroy()
+
+    # stop_and_destroy removes the whole endpoint directory. Recreate it.
+    Path(pgdata).mkdir(parents=True)
 
     spec = (
         """
@@ -56,11 +59,6 @@ def test_sync_safekeepers_logs(neon_env_builder: NeonEnvBuilder, pg_bin: PgBin):
                 "value": "replica",
                 "vartype": "enum"
             },
-            {
-                "name": "hot_standby",
-                "value": "on",
-                "vartype": "bool"
-            },
             {
                 "name": "neon.safekeepers",
                 "value": """
diff --git a/test_runner/regress/test_config.py b/test_runner/regress/test_config.py
index 3477d96b89..0ea5784b67 100755
--- a/test_runner/regress/test_config.py
+++ b/test_runner/regress/test_config.py
@@ -12,10 +12,10 @@ def test_config(neon_simple_env: NeonEnv):
     env.neon_cli.create_branch("test_config", "empty")
 
     # change config
-    pg = env.postgres.create_start("test_config", config_lines=["log_min_messages=debug1"])
+    endpoint = env.endpoints.create_start("test_config", config_lines=["log_min_messages=debug1"])
     log.info("postgres is running on test_config branch")
 
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute(
                 """
diff --git a/test_runner/regress/test_crafted_wal_end.py b/test_runner/regress/test_crafted_wal_end.py
index 9899d424d1..7ec901af34 100644
--- a/test_runner/regress/test_crafted_wal_end.py
+++ b/test_runner/regress/test_crafted_wal_end.py
@@ -21,11 +21,11 @@ def test_crafted_wal_end(neon_env_builder: NeonEnvBuilder, wal_type: str):
     env = neon_env_builder.init_start()
     env.neon_cli.create_branch("test_crafted_wal_end")
 
-    pg = env.postgres.create("test_crafted_wal_end")
+    endpoint = env.endpoints.create("test_crafted_wal_end")
     wal_craft = WalCraft(env)
-    pg.config(wal_craft.postgres_config())
-    pg.start()
-    res = pg.safe_psql_many(
+    endpoint.config(wal_craft.postgres_config())
+    endpoint.start()
+    res = endpoint.safe_psql_many(
         queries=[
             "CREATE TABLE keys(key int primary key)",
             "INSERT INTO keys SELECT generate_series(1, 100)",
@@ -34,7 +34,7 @@ def test_crafted_wal_end(neon_env_builder: NeonEnvBuilder, wal_type: str):
     )
     assert res[-1][0] == (5050,)
 
-    wal_craft.in_existing(wal_type, pg.connstr())
+    wal_craft.in_existing(wal_type, endpoint.connstr())
 
     log.info("Restarting all safekeepers and pageservers")
     env.pageserver.stop()
@@ -43,7 +43,7 @@ def test_crafted_wal_end(neon_env_builder: NeonEnvBuilder, wal_type: str):
     env.pageserver.start()
 
     log.info("Trying more queries")
-    res = pg.safe_psql_many(
+    res = endpoint.safe_psql_many(
         queries=[
             "SELECT SUM(key) FROM keys",
             "INSERT INTO keys SELECT generate_series(101, 200)",
@@ -60,7 +60,7 @@ def test_crafted_wal_end(neon_env_builder: NeonEnvBuilder, wal_type: str):
     env.pageserver.start()
 
     log.info("Trying more queries (again)")
-    res = pg.safe_psql_many(
+    res = endpoint.safe_psql_many(
         queries=[
             "SELECT SUM(key) FROM keys",
             "INSERT INTO keys SELECT generate_series(201, 300)",
diff --git a/test_runner/regress/test_createdropdb.py b/test_runner/regress/test_createdropdb.py
index 036e50e6e8..68035b1b14 100644
--- a/test_runner/regress/test_createdropdb.py
+++ b/test_runner/regress/test_createdropdb.py
@@ -13,10 +13,10 @@ def test_createdb(neon_simple_env: NeonEnv):
     env = neon_simple_env
     env.neon_cli.create_branch("test_createdb", "empty")
 
-    pg = env.postgres.create_start("test_createdb")
+    endpoint = env.endpoints.create_start("test_createdb")
     log.info("postgres is running on 'test_createdb' branch")
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         # Cause a 'relmapper' change in the original branch
         cur.execute("VACUUM FULL pg_class")
 
@@ -26,10 +26,10 @@ def test_createdb(neon_simple_env: NeonEnv):
 
     # Create a branch
     env.neon_cli.create_branch("test_createdb2", "test_createdb", ancestor_start_lsn=lsn)
-    pg2 = env.postgres.create_start("test_createdb2")
+    endpoint2 = env.endpoints.create_start("test_createdb2")
 
     # Test that you can connect to the new database on both branches
-    for db in (pg, pg2):
+    for db in (endpoint, endpoint2):
         with db.cursor(dbname="foodb") as cur:
             # Check database size in both branches
             cur.execute(
@@ -55,17 +55,17 @@ def test_createdb(neon_simple_env: NeonEnv):
 def test_dropdb(neon_simple_env: NeonEnv, test_output_dir):
     env = neon_simple_env
     env.neon_cli.create_branch("test_dropdb", "empty")
-    pg = env.postgres.create_start("test_dropdb")
+    endpoint = env.endpoints.create_start("test_dropdb")
     log.info("postgres is running on 'test_dropdb' branch")
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("CREATE DATABASE foodb")
 
         lsn_before_drop = query_scalar(cur, "SELECT pg_current_wal_insert_lsn()")
 
         dboid = query_scalar(cur, "SELECT oid FROM pg_database WHERE datname='foodb';")
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("DROP DATABASE foodb")
 
         cur.execute("CHECKPOINT")
@@ -76,29 +76,29 @@ def test_dropdb(neon_simple_env: NeonEnv, test_output_dir):
     env.neon_cli.create_branch(
         "test_before_dropdb", "test_dropdb", ancestor_start_lsn=lsn_before_drop
     )
-    pg_before = env.postgres.create_start("test_before_dropdb")
+    endpoint_before = env.endpoints.create_start("test_before_dropdb")
 
     env.neon_cli.create_branch(
         "test_after_dropdb", "test_dropdb", ancestor_start_lsn=lsn_after_drop
     )
-    pg_after = env.postgres.create_start("test_after_dropdb")
+    endpoint_after = env.endpoints.create_start("test_after_dropdb")
 
     # Test that database exists on the branch before drop
-    pg_before.connect(dbname="foodb").close()
+    endpoint_before.connect(dbname="foodb").close()
 
     # Test that database subdir exists on the branch before drop
-    assert pg_before.pgdata_dir
-    dbpath = pathlib.Path(pg_before.pgdata_dir) / "base" / str(dboid)
+    assert endpoint_before.pgdata_dir
+    dbpath = pathlib.Path(endpoint_before.pgdata_dir) / "base" / str(dboid)
     log.info(dbpath)
 
     assert os.path.isdir(dbpath) is True
 
     # Test that database subdir doesn't exist on the branch after drop
-    assert pg_after.pgdata_dir
-    dbpath = pathlib.Path(pg_after.pgdata_dir) / "base" / str(dboid)
+    assert endpoint_after.pgdata_dir
+    dbpath = pathlib.Path(endpoint_after.pgdata_dir) / "base" / str(dboid)
     log.info(dbpath)
 
     assert os.path.isdir(dbpath) is False
 
     # Check that we restore the content of the datadir correctly
-    check_restored_datadir_content(test_output_dir, env, pg)
+    check_restored_datadir_content(test_output_dir, env, endpoint)
diff --git a/test_runner/regress/test_createuser.py b/test_runner/regress/test_createuser.py
index c5f8246f5b..f1bc405287 100644
--- a/test_runner/regress/test_createuser.py
+++ b/test_runner/regress/test_createuser.py
@@ -9,10 +9,10 @@ from fixtures.utils import query_scalar
 def test_createuser(neon_simple_env: NeonEnv):
     env = neon_simple_env
     env.neon_cli.create_branch("test_createuser", "empty")
-    pg = env.postgres.create_start("test_createuser")
+    endpoint = env.endpoints.create_start("test_createuser")
     log.info("postgres is running on 'test_createuser' branch")
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         # Cause a 'relmapper' change in the original branch
         cur.execute("CREATE USER testuser with password %s", ("testpwd",))
 
@@ -22,7 +22,7 @@ def test_createuser(neon_simple_env: NeonEnv):
 
     # Create a branch
     env.neon_cli.create_branch("test_createuser2", "test_createuser", ancestor_start_lsn=lsn)
-    pg2 = env.postgres.create_start("test_createuser2")
+    endpoint2 = env.endpoints.create_start("test_createuser2")
 
     # Test that you can connect to new branch as a new user
-    assert pg2.safe_psql("select current_user", user="testuser") == [("testuser",)]
+    assert endpoint2.safe_psql("select current_user", user="testuser") == [("testuser",)]
diff --git a/test_runner/regress/test_disk_usage_eviction.py b/test_runner/regress/test_disk_usage_eviction.py
index 413d6c9d5a..31c7ef2b17 100644
--- a/test_runner/regress/test_disk_usage_eviction.py
+++ b/test_runner/regress/test_disk_usage_eviction.py
@@ -91,8 +91,8 @@ class EvictionEnv:
         This assumes that the tenant is still at the state after pbench -i.
         """
         lsn = self.pgbench_init_lsns[tenant_id]
-        with self.neon_env.postgres.create_start("main", tenant_id=tenant_id, lsn=lsn) as pg:
-            self.pg_bin.run(["pgbench", "-S", pg.connstr()])
+        with self.neon_env.endpoints.create_start("main", tenant_id=tenant_id, lsn=lsn) as endpoint:
+            self.pg_bin.run(["pgbench", "-S", endpoint.connstr()])
 
     def pageserver_start_with_disk_usage_eviction(
         self, period, max_usage_pct, min_avail_bytes, mock_behavior
@@ -168,9 +168,9 @@ def eviction_env(request, neon_env_builder: NeonEnvBuilder, pg_bin: PgBin) -> Ev
             }
         )
 
-        with env.postgres.create_start("main", tenant_id=tenant_id) as pg:
-            pg_bin.run(["pgbench", "-i", f"-s{scale}", pg.connstr()])
-            wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+        with env.endpoints.create_start("main", tenant_id=tenant_id) as endpoint:
+            pg_bin.run(["pgbench", "-i", f"-s{scale}", endpoint.connstr()])
+            wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
         timelines.append((tenant_id, timeline_id))
 
diff --git a/test_runner/regress/test_fsm_truncate.py b/test_runner/regress/test_fsm_truncate.py
index 4551ff97e0..80e4da8380 100644
--- a/test_runner/regress/test_fsm_truncate.py
+++ b/test_runner/regress/test_fsm_truncate.py
@@ -4,7 +4,7 @@ from fixtures.neon_fixtures import NeonEnvBuilder
 def test_fsm_truncate(neon_env_builder: NeonEnvBuilder):
     env = neon_env_builder.init_start()
     env.neon_cli.create_branch("test_fsm_truncate")
-    pg = env.postgres.create_start("test_fsm_truncate")
-    pg.safe_psql(
+    endpoint = env.endpoints.create_start("test_fsm_truncate")
+    endpoint.safe_psql(
         "CREATE TABLE t1(key int); CREATE TABLE t2(key int); TRUNCATE TABLE t1; TRUNCATE TABLE t2;"
     )
diff --git a/test_runner/regress/test_fullbackup.py b/test_runner/regress/test_fullbackup.py
index b3d58edf6b..ece9dccf93 100644
--- a/test_runner/regress/test_fullbackup.py
+++ b/test_runner/regress/test_fullbackup.py
@@ -24,10 +24,10 @@ def test_fullbackup(
     env = neon_env_builder.init_start()
 
     env.neon_cli.create_branch("test_fullbackup")
-    pgmain = env.postgres.create_start("test_fullbackup")
+    endpoint_main = env.endpoints.create_start("test_fullbackup")
     log.info("postgres is running on 'test_fullbackup' branch")
 
-    with pgmain.cursor() as cur:
+    with endpoint_main.cursor() as cur:
         timeline = TimelineId(query_scalar(cur, "SHOW neon.timeline_id"))
 
         # data loading may take a while, so increase statement timeout
diff --git a/test_runner/regress/test_gc_aggressive.py b/test_runner/regress/test_gc_aggressive.py
index 702d94c691..d38be057d3 100644
--- a/test_runner/regress/test_gc_aggressive.py
+++ b/test_runner/regress/test_gc_aggressive.py
@@ -5,9 +5,9 @@ import random
 import pytest
 from fixtures.log_helper import log
 from fixtures.neon_fixtures import (
+    Endpoint,
     NeonEnv,
     NeonEnvBuilder,
-    Postgres,
     RemoteStorageKind,
     wait_for_last_flush_lsn,
 )
@@ -26,9 +26,9 @@ updates_performed = 0
 
 
 # Run random UPDATEs on test table
-async def update_table(pg: Postgres):
+async def update_table(endpoint: Endpoint):
     global updates_performed
-    pg_conn = await pg.connect_async()
+    pg_conn = await endpoint.connect_async()
 
     while updates_performed < updates_to_perform:
         updates_performed += 1
@@ -52,10 +52,10 @@ async def gc(env: NeonEnv, timeline: TimelineId):
 
 
 # At the same time, run UPDATEs and GC
-async def update_and_gc(env: NeonEnv, pg: Postgres, timeline: TimelineId):
+async def update_and_gc(env: NeonEnv, endpoint: Endpoint, timeline: TimelineId):
     workers = []
     for worker_id in range(num_connections):
-        workers.append(asyncio.create_task(update_table(pg)))
+        workers.append(asyncio.create_task(update_table(endpoint)))
     workers.append(asyncio.create_task(gc(env, timeline)))
 
     # await all workers
@@ -72,10 +72,10 @@ def test_gc_aggressive(neon_env_builder: NeonEnvBuilder):
     neon_env_builder.pageserver_config_override = "tenant_config={pitr_interval = '0 sec'}"
     env = neon_env_builder.init_start()
     env.neon_cli.create_branch("test_gc_aggressive", "main")
-    pg = env.postgres.create_start("test_gc_aggressive")
+    endpoint = env.endpoints.create_start("test_gc_aggressive")
     log.info("postgres is running on test_gc_aggressive branch")
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         timeline = TimelineId(query_scalar(cur, "SHOW neon.timeline_id"))
 
         # Create table, and insert the first 100 rows
@@ -89,7 +89,7 @@ def test_gc_aggressive(neon_env_builder: NeonEnvBuilder):
         )
         cur.execute("CREATE INDEX ON foo(id)")
 
-        asyncio.run(update_and_gc(env, pg, timeline))
+        asyncio.run(update_and_gc(env, endpoint, timeline))
 
         cur.execute("SELECT COUNT(*), SUM(counter) FROM foo")
         r = cur.fetchone()
@@ -110,11 +110,11 @@ def test_gc_index_upload(neon_env_builder: NeonEnvBuilder, remote_storage_kind:
 
     env = neon_env_builder.init_start()
     env.neon_cli.create_branch("test_gc_index_upload", "main")
-    pg = env.postgres.create_start("test_gc_index_upload")
+    endpoint = env.endpoints.create_start("test_gc_index_upload")
 
     pageserver_http = env.pageserver.http_client()
 
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     tenant_id = TenantId(query_scalar(cur, "SHOW neon.tenant_id"))
@@ -146,7 +146,7 @@ def test_gc_index_upload(neon_env_builder: NeonEnvBuilder, remote_storage_kind:
         return int(total)
 
     # Sanity check that the metric works
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
     pageserver_http.timeline_checkpoint(tenant_id, timeline_id)
     pageserver_http.timeline_gc(tenant_id, timeline_id, 10000)
     before = get_num_remote_ops("index", "upload")
diff --git a/test_runner/regress/test_gc_cutoff.py b/test_runner/regress/test_gc_cutoff.py
index 1b98a414da..79453c1bdc 100644
--- a/test_runner/regress/test_gc_cutoff.py
+++ b/test_runner/regress/test_gc_cutoff.py
@@ -31,8 +31,8 @@ def test_gc_cutoff(neon_env_builder: NeonEnvBuilder, pg_bin: PgBin):
             "image_creation_threshold": "2",
         }
     )
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
-    connstr = pg.connstr(options="-csynchronous_commit=off")
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
+    connstr = endpoint.connstr(options="-csynchronous_commit=off")
     pg_bin.run_capture(["pgbench", "-i", "-s10", connstr])
 
     pageserver_http.configure_failpoints(("after-timeline-gc-removed-layers", "exit"))
diff --git a/test_runner/regress/test_hot_standby.py b/test_runner/regress/test_hot_standby.py
new file mode 100644
index 0000000000..12e034cea2
--- /dev/null
+++ b/test_runner/regress/test_hot_standby.py
@@ -0,0 +1,79 @@
+import pytest
+from fixtures.neon_fixtures import NeonEnv
+
+
+@pytest.mark.timeout(1800)
+def test_hot_standby(neon_simple_env: NeonEnv):
+    env = neon_simple_env
+
+    with env.endpoints.create_start(
+        branch_name="main",
+        endpoint_id="primary",
+    ) as primary:
+        with env.endpoints.new_replica_start(origin=primary, endpoint_id="secondary") as secondary:
+            primary_lsn = None
+            cought_up = False
+            queries = [
+                "SHOW neon.timeline_id",
+                "SHOW neon.tenant_id",
+                "SELECT relname FROM pg_class WHERE relnamespace = current_schema()::regnamespace::oid",
+                "SELECT COUNT(*), SUM(i) FROM test",
+            ]
+            responses = dict()
+
+            with primary.connect() as p_con:
+                with p_con.cursor() as p_cur:
+                    p_cur.execute("CREATE TABLE test AS SELECT generate_series(1, 100) AS i")
+
+                # Explicit commit to make sure other connections (and replicas) can
+                # see the changes of this commit.
+                p_con.commit()
+
+                with p_con.cursor() as p_cur:
+                    p_cur.execute("SELECT pg_current_wal_insert_lsn()::text")
+                    res = p_cur.fetchone()
+                    assert res is not None
+                    (lsn,) = res
+                    primary_lsn = lsn
+
+                # Explicit commit to make sure other connections (and replicas) can
+                # see the changes of this commit.
+                # Note that this may generate more WAL if the transaction has changed
+                # things, but we don't care about that.
+                p_con.commit()
+
+                for query in queries:
+                    with p_con.cursor() as p_cur:
+                        p_cur.execute(query)
+                        res = p_cur.fetchone()
+                        assert res is not None
+                        response = res
+                        responses[query] = response
+
+            with secondary.connect() as s_con:
+                with s_con.cursor() as s_cur:
+                    s_cur.execute("SELECT 1 WHERE pg_is_in_recovery()")
+                    res = s_cur.fetchone()
+                    assert res is not None
+
+                while not cought_up:
+                    with s_con.cursor() as secondary_cursor:
+                        secondary_cursor.execute("SELECT pg_last_wal_replay_lsn()")
+                        res = secondary_cursor.fetchone()
+                        assert res is not None
+                        (secondary_lsn,) = res
+                        # There may be more changes on the primary after we got our LSN
+                        # due to e.g. autovacuum, but that shouldn't impact the content
+                        # of the tables, so we check whether we've replayed up to at
+                        # least after the commit of the `test` table.
+                        cought_up = secondary_lsn >= primary_lsn
+
+                # Explicit commit to flush any transient transaction-level state.
+                s_con.commit()
+
+                for query in queries:
+                    with s_con.cursor() as secondary_cursor:
+                        secondary_cursor.execute(query)
+                        response = secondary_cursor.fetchone()
+                        assert response is not None
+                        assert response == responses[query]
diff --git a/test_runner/regress/test_import.py b/test_runner/regress/test_import.py
index 774ed98563..137ce457bc 100644
--- a/test_runner/regress/test_import.py
+++ b/test_runner/regress/test_import.py
@@ -9,10 +9,10 @@ from pathlib import Path
 import pytest
 from fixtures.log_helper import log
 from fixtures.neon_fixtures import (
+    Endpoint,
     NeonEnv,
     NeonEnvBuilder,
     PgBin,
-    Postgres,
 )
 from fixtures.pageserver.utils import wait_for_last_record_lsn, wait_for_upload
 from fixtures.types import Lsn, TenantId, TimelineId
@@ -72,7 +72,7 @@ def test_import_from_vanilla(test_output_dir, pg_bin, vanilla_pg, neon_env_build
         start_lsn = manifest["WAL-Ranges"][0]["Start-LSN"]
         end_lsn = manifest["WAL-Ranges"][0]["End-LSN"]
 
-    node_name = "import_from_vanilla"
+    endpoint_id = "ep-import_from_vanilla"
     tenant = TenantId.generate()
     timeline = TimelineId.generate()
 
@@ -113,7 +113,7 @@ def test_import_from_vanilla(test_output_dir, pg_bin, vanilla_pg, neon_env_build
                 "--timeline-id",
                 str(timeline),
                 "--node-name",
-                node_name,
+                endpoint_id,
                 "--base-lsn",
                 start_lsn,
                 "--base-tarfile",
@@ -153,8 +153,8 @@ def test_import_from_vanilla(test_output_dir, pg_bin, vanilla_pg, neon_env_build
     wait_for_upload(client, tenant, timeline, Lsn(end_lsn))
 
     # Check it worked
-    pg = env.postgres.create_start(node_name, tenant_id=tenant)
-    assert pg.safe_psql("select count(*) from t") == [(300000,)]
+    endpoint = env.endpoints.create_start(endpoint_id, tenant_id=tenant)
+    assert endpoint.safe_psql("select count(*) from t") == [(300000,)]
 
 
 @pytest.mark.timeout(600)
@@ -168,10 +168,10 @@ def test_import_from_pageserver_small(pg_bin: PgBin, neon_env_builder: NeonEnvBu
     )
 
     timeline = env.neon_cli.create_branch("test_import_from_pageserver_small")
-    pg = env.postgres.create_start("test_import_from_pageserver_small")
+    endpoint = env.endpoints.create_start("test_import_from_pageserver_small")
 
     num_rows = 3000
-    lsn = _generate_data(num_rows, pg)
+    lsn = _generate_data(num_rows, endpoint)
     _import(num_rows, lsn, env, pg_bin, timeline, env.pg_distrib_dir)
 
 
@@ -185,14 +185,14 @@ def test_import_from_pageserver_multisegment(pg_bin: PgBin, neon_env_builder: Ne
     env = neon_env_builder.init_start()
 
     timeline = env.neon_cli.create_branch("test_import_from_pageserver_multisegment")
-    pg = env.postgres.create_start("test_import_from_pageserver_multisegment")
+    endpoint = env.endpoints.create_start("test_import_from_pageserver_multisegment")
 
     # For `test_import_from_pageserver_multisegment`, we want to make sure that the data
     # is large enough to create multi-segment files. Typically, a segment file's size is
     # at most 1GB. A large number of inserted rows (`30000000`) is used to increase the
     # DB size to above 1GB. Related: https://github.com/neondatabase/neon/issues/2097.
     num_rows = 30000000
-    lsn = _generate_data(num_rows, pg)
+    lsn = _generate_data(num_rows, endpoint)
 
     logical_size = env.pageserver.http_client().timeline_detail(env.initial_tenant, timeline)[
         "current_logical_size"
@@ -213,12 +213,12 @@ def test_import_from_pageserver_multisegment(pg_bin: PgBin, neon_env_builder: Ne
     assert cnt_seg_files > 0
 
 
-def _generate_data(num_rows: int, pg: Postgres) -> Lsn:
+def _generate_data(num_rows: int, endpoint: Endpoint) -> Lsn:
     """Generate a table with `num_rows` rows.
 
     Returns:
     the latest insert WAL's LSN"""
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             # data loading may take a while, so increase statement timeout
             cur.execute("SET statement_timeout='300s'")
@@ -263,7 +263,7 @@ def _import(
     tar_output_file = result_basepath + ".stdout"
 
     # Stop the first pageserver instance, erase all its data
-    env.postgres.stop_all()
+    env.endpoints.stop_all()
     env.pageserver.stop()
 
     dir_to_clear = Path(env.repo_dir) / "tenants"
@@ -278,7 +278,7 @@ def _import(
     tenant = TenantId.generate()
 
     # Import to pageserver
-    node_name = "import_from_pageserver"
+    endpoint_id = "ep-import_from_pageserver"
     client = env.pageserver.http_client()
     client.tenant_create(tenant)
     env.neon_cli.raw_cli(
@@ -290,7 +290,7 @@ def _import(
             "--timeline-id",
             str(timeline),
             "--node-name",
-            node_name,
+            endpoint_id,
             "--base-lsn",
             str(lsn),
             "--base-tarfile",
@@ -305,8 +305,8 @@ def _import(
     wait_for_upload(client, tenant, timeline, lsn)
 
     # Check it worked
-    pg = env.postgres.create_start(node_name, tenant_id=tenant)
-    assert pg.safe_psql("select count(*) from tbl") == [(expected_num_rows,)]
+    endpoint = env.endpoints.create_start(endpoint_id, tenant_id=tenant)
+    assert endpoint.safe_psql("select count(*) from tbl") == [(expected_num_rows,)]
 
     # Take another fullbackup
     query = f"fullbackup { tenant} {timeline} {lsn}"
diff --git a/test_runner/regress/test_large_schema.py b/test_runner/regress/test_large_schema.py
index f14265f6fd..ac83131ba2 100644
--- a/test_runner/regress/test_large_schema.py
+++ b/test_runner/regress/test_large_schema.py
@@ -15,9 +15,9 @@ from fixtures.neon_fixtures import NeonEnvBuilder
 def test_large_schema(neon_env_builder: NeonEnvBuilder):
     env = neon_env_builder.init_start()
 
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
-    conn = pg.connect()
+    conn = endpoint.connect()
     cur = conn.cursor()
 
     tables = 2  # 10 is too much for debug build
@@ -27,18 +27,18 @@ def test_large_schema(neon_env_builder: NeonEnvBuilder):
 
         # Restart compute. Restart is actually not strictly needed.
         # It is done mostly because this test originally tries to model the problem reported by Ketteq.
-        pg.stop()
+        endpoint.stop()
         # Kill and restart the pageserver.
         # env.pageserver.stop(immediate=True)
         # env.pageserver.start()
-        pg.start()
+        endpoint.start()
 
         retry_sleep = 0.5
         max_retries = 200
         retries = 0
         while True:
             try:
-                conn = pg.connect()
+                conn = endpoint.connect()
                 cur = conn.cursor()
                 cur.execute(f"CREATE TABLE if not exists t_{i}(pk integer) partition by range (pk)")
                 for j in range(1, partitions + 1):
@@ -63,7 +63,7 @@ def test_large_schema(neon_env_builder: NeonEnvBuilder):
                     raise
             break
 
-    conn = pg.connect()
+    conn = endpoint.connect()
     cur = conn.cursor()
 
     for i in range(1, tables + 1):
@@ -74,8 +74,8 @@ def test_large_schema(neon_env_builder: NeonEnvBuilder):
     cur.execute("select * from pg_depend order by refclassid, refobjid, refobjsubid")
 
     # Check layer file sizes
-    tenant_id = pg.safe_psql("show neon.tenant_id")[0][0]
-    timeline_id = pg.safe_psql("show neon.timeline_id")[0][0]
+    tenant_id = endpoint.safe_psql("show neon.tenant_id")[0][0]
+    timeline_id = endpoint.safe_psql("show neon.timeline_id")[0][0]
     timeline_path = "{}/tenants/{}/timelines/{}/".format(env.repo_dir, tenant_id, timeline_id)
     for filename in os.listdir(timeline_path):
         if filename.startswith("00000"):
diff --git a/test_runner/regress/test_layer_eviction.py b/test_runner/regress/test_layer_eviction.py
index 2d07d02ce7..a96532c0d8 100644
--- a/test_runner/regress/test_layer_eviction.py
+++ b/test_runner/regress/test_layer_eviction.py
@@ -6,7 +6,6 @@ from fixtures.neon_fixtures import (
     NeonEnvBuilder,
     RemoteStorageKind,
     wait_for_last_flush_lsn,
-    wait_for_sk_commit_lsn_to_reach_remote_storage,
 )
 from fixtures.pageserver.utils import wait_for_last_record_lsn, wait_for_upload
 from fixtures.types import Lsn, TenantId, TimelineId
@@ -27,13 +26,13 @@ def test_basic_eviction(
 
     env = neon_env_builder.init_start()
     client = env.pageserver.http_client()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     # Create a number of layers in the tenant
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("CREATE TABLE foo (t text)")
         cur.execute(
             """
@@ -172,15 +171,15 @@ def test_gc_of_remote_layers(neon_env_builder: NeonEnvBuilder):
     env.initial_tenant = tenant_id  # update_and_gc relies on this
     ps_http = env.pageserver.http_client()
 
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     log.info("fill with data, creating delta & image layers, some of which are GC'able after")
     # no particular reason to create the layers like this, but we are sure
     # not to hit the image_creation_threshold here.
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("create table a (id bigserial primary key, some_value bigint not null)")
         cur.execute("insert into a(some_value) select i from generate_series(1, 10000) s(i)")
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
     ps_http.timeline_checkpoint(tenant_id, timeline_id)
 
     # Create delta layers, then turn them into image layers.
@@ -191,19 +190,19 @@ def test_gc_of_remote_layers(neon_env_builder: NeonEnvBuilder):
         for i in range(0, 2):
             for j in range(0, 3):
                 # create a minimal amount of "delta difficulty" for this table
-                with pg.cursor() as cur:
+                with endpoint.cursor() as cur:
                     cur.execute("update a set some_value = -some_value + %s", (j,))
 
-                with pg.cursor() as cur:
+                with endpoint.cursor() as cur:
                     # vacuuming should aid to reuse keys, though it's not really important
                     # with image_creation_threshold=1 which we will use on the last compaction
                     cur.execute("vacuum")
 
-                wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+                last_lsn = wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
                 if i == 1 and j == 2 and k == 1:
                     # last iteration; stop before checkpoint to avoid leaving an inmemory layer
-                    pg.stop_and_destroy()
+                    endpoint.stop_and_destroy()
 
                 ps_http.timeline_checkpoint(tenant_id, timeline_id)
 
@@ -222,10 +221,8 @@ def test_gc_of_remote_layers(neon_env_builder: NeonEnvBuilder):
         tenant_update_config({"image_creation_threshold": "1"})
         ps_http.timeline_compact(tenant_id, timeline_id)
 
-    # wait for all uploads to finish
-    wait_for_sk_commit_lsn_to_reach_remote_storage(
-        tenant_id, timeline_id, env.safekeepers, env.pageserver
-    )
+    # wait for all uploads to finish (checkpoint has been done above)
+    wait_for_upload(ps_http, tenant_id, timeline_id, last_lsn)
 
     # shutdown safekeepers to avoid on-demand downloads from walreceiver
     for sk in env.safekeepers:
diff --git a/test_runner/regress/test_layer_writers_fail.py b/test_runner/regress/test_layer_writers_fail.py
index e8ba0e7d91..d2d85a43e0 100644
--- a/test_runner/regress/test_layer_writers_fail.py
+++ b/test_runner/regress/test_layer_writers_fail.py
@@ -20,7 +20,7 @@ def test_image_layer_writer_fail_before_finish(neon_simple_env: NeonEnv):
         }
     )
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    pg = env.endpoints.create_start("main", tenant_id=tenant_id)
     pg.safe_psql_many(
         [
             "CREATE TABLE foo (t text) WITH (autovacuum_enabled = off)",
@@ -64,8 +64,8 @@ def test_delta_layer_writer_fail_before_finish(neon_simple_env: NeonEnv):
         }
     )
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
-    pg.safe_psql_many(
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
+    endpoint.safe_psql_many(
         [
             "CREATE TABLE foo (t text) WITH (autovacuum_enabled = off)",
             """INSERT INTO foo
diff --git a/test_runner/regress/test_logging.py b/test_runner/regress/test_logging.py
new file mode 100644
index 0000000000..d559be0a8f
--- /dev/null
+++ b/test_runner/regress/test_logging.py
@@ -0,0 +1,49 @@
+import uuid
+
+import pytest
+from fixtures.log_helper import log
+from fixtures.neon_fixtures import NeonEnvBuilder
+from fixtures.utils import wait_until
+
+
+@pytest.mark.parametrize("level", ["trace", "debug", "info", "warn", "error"])
+def test_logging_event_count(neon_env_builder: NeonEnvBuilder, level: str):
+    # self-test: make sure the event is logged (i.e., our testing endpoint works)
+    log_expected = {
+        "trace": False,
+        "debug": False,
+        "info": True,
+        "warn": True,
+        "error": True,
+    }[level]
+
+    env = neon_env_builder.init_start()
+    ps_http = env.pageserver.http_client()
+    msg_id = uuid.uuid4().hex
+
+    # NB: the _total suffix is added by our prometheus client
+    before = ps_http.get_metric_value("libmetrics_tracing_event_count_total", {"level": level})
+
+    # post the event
+    ps_http.post_tracing_event(level, msg_id)
+    if log_expected:
+        env.pageserver.allowed_errors.append(f".*{msg_id}.*")
+
+    def assert_logged():
+        if not log_expected:
+            return
+        assert env.pageserver.log_contains(f".*{msg_id}.*")
+
+    wait_until(10, 0.5, assert_logged)
+
+    # make sure it's counted
+    def assert_metric_value():
+        if not log_expected:
+            return
+        # NB: the _total suffix is added by our prometheus client
+        val = ps_http.get_metric_value("libmetrics_tracing_event_count_total", {"level": level})
+        val = val or 0.0
+        log.info("libmetrics_tracing_event_count: %s", val)
+        assert val > (before or 0.0)
+
+    wait_until(10, 1, assert_metric_value)
diff --git a/test_runner/regress/test_lsn_mapping.py b/test_runner/regress/test_lsn_mapping.py
index c5a49a6704..8ccfc21cf7 100644
--- a/test_runner/regress/test_lsn_mapping.py
+++ b/test_runner/regress/test_lsn_mapping.py
@@ -12,10 +12,10 @@ def test_lsn_mapping(neon_env_builder: NeonEnvBuilder):
     env = neon_env_builder.init_start()
 
     new_timeline_id = env.neon_cli.create_branch("test_lsn_mapping")
-    pgmain = env.postgres.create_start("test_lsn_mapping")
+    endpoint_main = env.endpoints.create_start("test_lsn_mapping")
     log.info("postgres is running on 'test_lsn_mapping' branch")
 
-    cur = pgmain.connect().cursor()
+    cur = endpoint_main.connect().cursor()
     # Create table, and insert rows, each in a separate transaction
     # Disable synchronous_commit to make this initialization go faster.
     #
@@ -35,7 +35,7 @@ def test_lsn_mapping(neon_env_builder: NeonEnvBuilder):
     cur.execute("INSERT INTO foo VALUES (-1)")
 
     # Wait until WAL is received by pageserver
-    wait_for_last_flush_lsn(env, pgmain, env.initial_tenant, new_timeline_id)
+    wait_for_last_flush_lsn(env, endpoint_main, env.initial_tenant, new_timeline_id)
 
     with env.pageserver.http_client() as client:
         # Check edge cases: timestamp in the future
@@ -61,9 +61,9 @@ def test_lsn_mapping(neon_env_builder: NeonEnvBuilder):
             # Call get_lsn_by_timestamp to get the LSN
             # Launch a new read-only node at that LSN, and check that only the rows
             # that were supposed to be committed at that point in time are visible.
-            pg_here = env.postgres.create_start(
-                branch_name="test_lsn_mapping", node_name="test_lsn_mapping_read", lsn=lsn
+            endpoint_here = env.endpoints.create_start(
+                branch_name="test_lsn_mapping", endpoint_id="ep-lsn_mapping_read", lsn=lsn
             )
-            assert pg_here.safe_psql("SELECT max(x) FROM foo")[0][0] == i
+            assert endpoint_here.safe_psql("SELECT max(x) FROM foo")[0][0] == i
 
-            pg_here.stop_and_destroy()
+            endpoint_here.stop_and_destroy()
diff --git a/test_runner/regress/test_metric_collection.py b/test_runner/regress/test_metric_collection.py
index a33af9a3b2..ecbce1f8f7 100644
--- a/test_runner/regress/test_metric_collection.py
+++ b/test_runner/regress/test_metric_collection.py
@@ -123,9 +123,9 @@ def test_metric_collection(
     # before pageserver, pageserver log might contain such errors in the end.
     env.pageserver.allowed_errors.append(".*metrics endpoint refused the sent metrics*")
     env.neon_cli.create_branch("test_metric_collection")
-    pg = env.postgres.create_start("test_metric_collection")
+    endpoint = env.endpoints.create_start("test_metric_collection")
 
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     tenant_id = TenantId(query_scalar(cur, "SHOW neon.tenant_id"))
@@ -158,7 +158,7 @@ def test_metric_collection(
 
     # upload some data to remote storage
     if remote_storage_kind == RemoteStorageKind.LOCAL_FS:
-        wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
         pageserver_http = env.pageserver.http_client()
         pageserver_http.timeline_checkpoint(tenant_id, timeline_id)
         pageserver_http.timeline_gc(tenant_id, timeline_id, 10000)
diff --git a/test_runner/regress/test_multixact.py b/test_runner/regress/test_multixact.py
index 635beb16b7..fe50969a0a 100644
--- a/test_runner/regress/test_multixact.py
+++ b/test_runner/regress/test_multixact.py
@@ -12,10 +12,10 @@ from fixtures.utils import query_scalar
 def test_multixact(neon_simple_env: NeonEnv, test_output_dir):
     env = neon_simple_env
     env.neon_cli.create_branch("test_multixact", "empty")
-    pg = env.postgres.create_start("test_multixact")
+    endpoint = env.endpoints.create_start("test_multixact")
 
     log.info("postgres is running on 'test_multixact' branch")
-    cur = pg.connect().cursor()
+    cur = endpoint.connect().cursor()
     cur.execute(
         """
         CREATE TABLE t1(i int primary key);
@@ -32,7 +32,7 @@ def test_multixact(neon_simple_env: NeonEnv, test_output_dir):
     connections = []
     for i in range(nclients):
         # Do not turn on autocommit. We want to hold the key-share locks.
-        conn = pg.connect(autocommit=False)
+        conn = endpoint.connect(autocommit=False)
         connections.append(conn)
 
     # On each iteration, we commit the previous transaction on a connection,
@@ -65,10 +65,10 @@ def test_multixact(neon_simple_env: NeonEnv, test_output_dir):
 
     # Branch at this point
     env.neon_cli.create_branch("test_multixact_new", "test_multixact", ancestor_start_lsn=lsn)
-    pg_new = env.postgres.create_start("test_multixact_new")
+    endpoint_new = env.endpoints.create_start("test_multixact_new")
 
     log.info("postgres is running on 'test_multixact_new' branch")
-    next_multixact_id_new = pg_new.safe_psql(
+    next_multixact_id_new = endpoint_new.safe_psql(
         "SELECT next_multixact_id FROM pg_control_checkpoint()"
     )[0][0]
 
@@ -76,4 +76,4 @@ def test_multixact(neon_simple_env: NeonEnv, test_output_dir):
     assert next_multixact_id_new == next_multixact_id
 
     # Check that we can restore the content of the datadir correctly
-    check_restored_datadir_content(test_output_dir, env, pg)
+    check_restored_datadir_content(test_output_dir, env, endpoint)
diff --git a/test_runner/regress/test_neon_local_cli.py b/test_runner/regress/test_neon_local_cli.py
index bd0f550ba5..f6629c54f9 100644
--- a/test_runner/regress/test_neon_local_cli.py
+++ b/test_runner/regress/test_neon_local_cli.py
@@ -9,9 +9,11 @@ def test_neon_cli_basics(neon_env_builder: NeonEnvBuilder, port_distributor: Por
     try:
         env.neon_cli.start()
         env.neon_cli.create_tenant(tenant_id=env.initial_tenant, set_default=True)
-        env.neon_cli.pg_start(node_name="main", port=port_distributor.get_port())
+        env.neon_cli.endpoint_start(endpoint_id="ep-main", port=port_distributor.get_port())
 
         env.neon_cli.create_branch(new_branch_name="migration_check")
-        env.neon_cli.pg_start(node_name="migration_check", port=port_distributor.get_port())
+        env.neon_cli.endpoint_start(
+            endpoint_id="ep-migration_check", port=port_distributor.get_port()
+        )
     finally:
         env.neon_cli.stop()
diff --git a/test_runner/regress/test_next_xid.py b/test_runner/regress/test_next_xid.py
index 698ea0e1d3..6e94e15227 100644
--- a/test_runner/regress/test_next_xid.py
+++ b/test_runner/regress/test_next_xid.py
@@ -8,9 +8,9 @@ from fixtures.neon_fixtures import NeonEnvBuilder
 def test_next_xid(neon_env_builder: NeonEnvBuilder):
     env = neon_env_builder.init_start()
 
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
-    conn = pg.connect()
+    conn = endpoint.connect()
     cur = conn.cursor()
     cur.execute("CREATE TABLE t(x integer)")
 
@@ -19,17 +19,17 @@ def test_next_xid(neon_env_builder: NeonEnvBuilder):
         print(f"iteration {i} / {iterations}")
 
         # Kill and restart the pageserver.
-        pg.stop()
+        endpoint.stop()
         env.pageserver.stop(immediate=True)
         env.pageserver.start()
-        pg.start()
+        endpoint.start()
 
         retry_sleep = 0.5
         max_retries = 200
         retries = 0
         while True:
             try:
-                conn = pg.connect()
+                conn = endpoint.connect()
                 cur = conn.cursor()
                 cur.execute(f"INSERT INTO t values({i})")
                 conn.close()
@@ -48,7 +48,7 @@ def test_next_xid(neon_env_builder: NeonEnvBuilder):
                     raise
             break
 
-    conn = pg.connect()
+    conn = endpoint.connect()
     cur = conn.cursor()
     cur.execute("SELECT count(*) FROM t")
     assert cur.fetchone() == (iterations,)
diff --git a/test_runner/regress/test_normal_work.py b/test_runner/regress/test_normal_work.py
index aa37a2411c..50de99adb5 100644
--- a/test_runner/regress/test_normal_work.py
+++ b/test_runner/regress/test_normal_work.py
@@ -6,9 +6,9 @@ from fixtures.pageserver.http import PageserverHttpClient
 
 def check_tenant(env: NeonEnv, pageserver_http: PageserverHttpClient):
     tenant_id, timeline_id = env.neon_cli.create_tenant()
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
     # we rely upon autocommit after each statement
-    res_1 = pg.safe_psql_many(
+    res_1 = endpoint.safe_psql_many(
         queries=[
             "CREATE TABLE t(key int primary key, value text)",
             "INSERT INTO t SELECT generate_series(1,100000), 'payload'",
@@ -19,14 +19,14 @@ def check_tenant(env: NeonEnv, pageserver_http: PageserverHttpClient):
     assert res_1[-1][0] == (5000050000,)
     # TODO check detach on live instance
     log.info("stopping compute")
-    pg.stop()
+    endpoint.stop()
     log.info("compute stopped")
 
-    pg.start()
-    res_2 = pg.safe_psql("SELECT sum(key) FROM t")
+    endpoint.start()
+    res_2 = endpoint.safe_psql("SELECT sum(key) FROM t")
     assert res_2[0] == (5000050000,)
 
-    pg.stop()
+    endpoint.stop()
     pageserver_http.tenant_detach(tenant_id)
 
 
diff --git a/test_runner/regress/test_old_request_lsn.py b/test_runner/regress/test_old_request_lsn.py
index 9885a811e1..814b9f3de0 100644
--- a/test_runner/regress/test_old_request_lsn.py
+++ b/test_runner/regress/test_old_request_lsn.py
@@ -19,10 +19,10 @@ def test_old_request_lsn(neon_env_builder: NeonEnvBuilder):
     neon_env_builder.pageserver_config_override = "tenant_config={pitr_interval = '0 sec'}"
     env = neon_env_builder.init_start()
     env.neon_cli.create_branch("test_old_request_lsn", "main")
-    pg = env.postgres.create_start("test_old_request_lsn")
+    endpoint = env.endpoints.create_start("test_old_request_lsn")
     log.info("postgres is running on test_old_request_lsn branch")
 
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     # Get the timeline ID of our branch. We need it for the 'do_gc' command
diff --git a/test_runner/regress/test_ondemand_download.py b/test_runner/regress/test_ondemand_download.py
index 90ab8e68d8..5c02708457 100644
--- a/test_runner/regress/test_ondemand_download.py
+++ b/test_runner/regress/test_ondemand_download.py
@@ -12,12 +12,12 @@ from fixtures.neon_fixtures import (
     NeonEnvBuilder,
     RemoteStorageKind,
     available_remote_storages,
+    last_flush_lsn_upload,
     wait_for_last_flush_lsn,
-    wait_for_sk_commit_lsn_to_reach_remote_storage,
 )
 from fixtures.pageserver.http import PageserverApiException, PageserverHttpClient
 from fixtures.pageserver.utils import (
-    assert_tenant_status,
+    assert_tenant_state,
     wait_for_last_record_lsn,
     wait_for_upload,
     wait_until_tenant_state,
@@ -73,17 +73,17 @@ def test_ondemand_download_large_rel(
     )
     env.initial_tenant = tenant
 
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     client = env.pageserver.http_client()
 
-    tenant_id = pg.safe_psql("show neon.tenant_id")[0][0]
-    timeline_id = pg.safe_psql("show neon.timeline_id")[0][0]
+    tenant_id = endpoint.safe_psql("show neon.tenant_id")[0][0]
+    timeline_id = endpoint.safe_psql("show neon.timeline_id")[0][0]
 
     # We want to make sure that the data is large enough that the keyspace is partitioned.
     num_rows = 1000000
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         # data loading may take a while, so increase statement timeout
         cur.execute("SET statement_timeout='300s'")
         cur.execute(
@@ -106,7 +106,7 @@ def test_ondemand_download_large_rel(
     log.info("uploads have finished")
 
     ##### Stop the first pageserver instance, erase all its data
-    pg.stop()
+    endpoint.stop()
     env.pageserver.stop()
 
     # remove all the layer files
@@ -117,7 +117,7 @@ def test_ondemand_download_large_rel(
     ##### Second start, restore the data and ensure it's the same
     env.pageserver.start()
 
-    pg.start()
+    endpoint.start()
     before_downloads = get_num_downloaded_layers(client, tenant_id, timeline_id)
 
     # Probe in the middle of the table. There's a high chance that the beginning
@@ -125,7 +125,7 @@ def test_ondemand_download_large_rel(
     # from other tables, and with the entry that stores the size of the
     # relation, so they are likely already downloaded. But the middle of the
     # table should not have been needed by anything yet.
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         assert query_scalar(cur, "select count(*) from tbl where id = 500000") == 1
 
     after_downloads = get_num_downloaded_layers(client, tenant_id, timeline_id)
@@ -167,17 +167,17 @@ def test_ondemand_download_timetravel(
     )
     env.initial_tenant = tenant
 
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     client = env.pageserver.http_client()
 
-    tenant_id = pg.safe_psql("show neon.tenant_id")[0][0]
-    timeline_id = pg.safe_psql("show neon.timeline_id")[0][0]
+    tenant_id = endpoint.safe_psql("show neon.tenant_id")[0][0]
+    timeline_id = endpoint.safe_psql("show neon.timeline_id")[0][0]
 
     lsns = []
 
     table_len = 10000
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute(
             f"""
         CREATE TABLE testtab(id serial primary key, checkpoint_number int, data text);
@@ -192,7 +192,7 @@ def test_ondemand_download_timetravel(
     lsns.append((0, current_lsn))
 
     for checkpoint_number in range(1, 20):
-        with pg.cursor() as cur:
+        with endpoint.cursor() as cur:
             cur.execute(f"UPDATE testtab SET checkpoint_number = {checkpoint_number}")
             current_lsn = Lsn(query_scalar(cur, "SELECT pg_current_wal_flush_lsn()"))
         lsns.append((checkpoint_number, current_lsn))
@@ -204,12 +204,10 @@ def test_ondemand_download_timetravel(
         client.timeline_checkpoint(tenant_id, timeline_id)
 
     ##### Stop the first pageserver instance, erase all its data
-    env.postgres.stop_all()
+    env.endpoints.stop_all()
 
     # wait until pageserver has successfully uploaded all the data to remote storage
-    wait_for_sk_commit_lsn_to_reach_remote_storage(
-        tenant_id, timeline_id, env.safekeepers, env.pageserver
-    )
+    wait_for_upload(client, tenant_id, timeline_id, current_lsn)
 
     def get_api_current_physical_size():
         d = client.timeline_detail(tenant_id, timeline_id)
@@ -239,7 +237,7 @@ def test_ondemand_download_timetravel(
     ##### Second start, restore the data and ensure it's the same
     env.pageserver.start()
 
-    wait_until(10, 0.2, lambda: assert_tenant_status(client, tenant_id, "Active"))
+    wait_until(10, 0.2, lambda: assert_tenant_state(client, tenant_id, "Active"))
 
     # The current_physical_size reports the sum of layers loaded in the layer
     # map, regardless of where the layer files are located. So even though we
@@ -251,10 +249,10 @@ def test_ondemand_download_timetravel(
     num_layers_downloaded = [0]
     resident_size = [get_resident_physical_size()]
     for checkpoint_number, lsn in lsns:
-        pg_old = env.postgres.create_start(
-            branch_name="main", node_name=f"test_old_lsn_{checkpoint_number}", lsn=lsn
+        endpoint_old = env.endpoints.create_start(
+            branch_name="main", endpoint_id=f"ep-old_lsn_{checkpoint_number}", lsn=lsn
         )
-        with pg_old.cursor() as cur:
+        with endpoint_old.cursor() as cur:
             # assert query_scalar(cur, f"select count(*) from testtab where checkpoint_number={checkpoint_number}") == 100000
             assert (
                 query_scalar(
@@ -331,15 +329,15 @@ def test_download_remote_layers_api(
     )
     env.initial_tenant = tenant
 
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     client = env.pageserver.http_client()
 
-    tenant_id = pg.safe_psql("show neon.tenant_id")[0][0]
-    timeline_id = pg.safe_psql("show neon.timeline_id")[0][0]
+    tenant_id = endpoint.safe_psql("show neon.tenant_id")[0][0]
+    timeline_id = endpoint.safe_psql("show neon.timeline_id")[0][0]
 
     table_len = 10000
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute(
             f"""
         CREATE TABLE testtab(id serial primary key, checkpoint_number int, data text);
@@ -347,11 +345,8 @@ def test_download_remote_layers_api(
         """
         )
 
-    env.postgres.stop_all()
-
-    wait_for_sk_commit_lsn_to_reach_remote_storage(
-        tenant_id, timeline_id, env.safekeepers, env.pageserver
-    )
+    last_flush_lsn_upload(env, endpoint, tenant_id, timeline_id)
+    env.endpoints.stop_all()
 
     def get_api_current_physical_size():
         d = client.timeline_detail(tenant_id, timeline_id)
@@ -392,7 +387,7 @@ def test_download_remote_layers_api(
         ]
     )
 
-    wait_until(10, 0.2, lambda: assert_tenant_status(client, tenant_id, "Active"))
+    wait_until(10, 0.2, lambda: assert_tenant_state(client, tenant_id, "Active"))
 
     ###### Phase 1: exercise download error code path
     assert (
@@ -463,8 +458,8 @@ def test_download_remote_layers_api(
         sk.start()
 
     # ensure that all the data is back
-    pg_old = env.postgres.create_start(branch_name="main")
-    with pg_old.cursor() as cur:
+    endpoint_old = env.endpoints.create_start(branch_name="main")
+    with endpoint_old.cursor() as cur:
         assert query_scalar(cur, "select count(*) from testtab") == table_len
 
 
@@ -513,17 +508,17 @@ def test_compaction_downloads_on_demand_without_image_creation(
     env.initial_tenant = tenant_id
     pageserver_http = env.pageserver.http_client()
 
-    with env.postgres.create_start("main") as pg:
+    with env.endpoints.create_start("main") as endpoint:
         # no particular reason to create the layers like this, but we are sure
         # not to hit the image_creation_threshold here.
-        with pg.cursor() as cur:
+        with endpoint.cursor() as cur:
             cur.execute("create table a as select id::bigint from generate_series(1, 204800) s(id)")
-        wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
         pageserver_http.timeline_checkpoint(tenant_id, timeline_id)
 
-        with pg.cursor() as cur:
+        with endpoint.cursor() as cur:
             cur.execute("update a set id = -id")
-        wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
         pageserver_http.timeline_checkpoint(tenant_id, timeline_id)
 
     layers = pageserver_http.layer_map_info(tenant_id, timeline_id)
@@ -589,32 +584,32 @@ def test_compaction_downloads_on_demand_with_image_creation(
     env.initial_tenant = tenant_id
     pageserver_http = env.pageserver.http_client()
 
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     # no particular reason to create the layers like this, but we are sure
     # not to hit the image_creation_threshold here.
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("create table a (id bigserial primary key, some_value bigint not null)")
         cur.execute("insert into a(some_value) select i from generate_series(1, 10000) s(i)")
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
     pageserver_http.timeline_checkpoint(tenant_id, timeline_id)
 
     for i in range(0, 2):
         for j in range(0, 3):
             # create a minimal amount of "delta difficulty" for this table
-            with pg.cursor() as cur:
+            with endpoint.cursor() as cur:
                 cur.execute("update a set some_value = -some_value + %s", (j,))
 
-            with pg.cursor() as cur:
+            with endpoint.cursor() as cur:
                 # vacuuming should aid to reuse keys, though it's not really important
                 # with image_creation_threshold=1 which we will use on the last compaction
                 cur.execute("vacuum")
 
-            wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+            wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
             if i == 1 and j == 2:
                 # last iteration; stop before checkpoint to avoid leaving an inmemory layer
-                pg.stop_and_destroy()
+                endpoint.stop_and_destroy()
 
             pageserver_http.timeline_checkpoint(tenant_id, timeline_id)
 
diff --git a/test_runner/regress/test_pageserver_api.py b/test_runner/regress/test_pageserver_api.py
index 5b05989ae4..e86cd18f58 100644
--- a/test_runner/regress/test_pageserver_api.py
+++ b/test_runner/regress/test_pageserver_api.py
@@ -150,7 +150,7 @@ def test_pageserver_http_get_wal_receiver_success(neon_simple_env: NeonEnv):
     env = neon_simple_env
     with env.pageserver.http_client() as client:
         tenant_id, timeline_id = env.neon_cli.create_tenant()
-        pg = env.postgres.create_start(DEFAULT_BRANCH_NAME, tenant_id=tenant_id)
+        endpoint = env.endpoints.create_start(DEFAULT_BRANCH_NAME, tenant_id=tenant_id)
 
         # Wait to make sure that we get a latest WAL receiver data.
         # We need to wait here because it's possible that we don't have access to
@@ -163,7 +163,7 @@ def test_pageserver_http_get_wal_receiver_success(neon_simple_env: NeonEnv):
         )
 
         # Make a DB modification then expect getting a new WAL receiver's data.
-        pg.safe_psql("CREATE TABLE t(key int primary key, value text)")
+        endpoint.safe_psql("CREATE TABLE t(key int primary key, value text)")
         wait_until(
             number_of_iterations=5,
             interval=1,
diff --git a/test_runner/regress/test_pageserver_catchup.py b/test_runner/regress/test_pageserver_catchup.py
index cba3203591..c16cbcb4ba 100644
--- a/test_runner/regress/test_pageserver_catchup.py
+++ b/test_runner/regress/test_pageserver_catchup.py
@@ -11,11 +11,11 @@ def test_pageserver_catchup_while_compute_down(neon_env_builder: NeonEnvBuilder)
 
     env.neon_cli.create_branch("test_pageserver_catchup_while_compute_down")
     # Make shared_buffers large to ensure we won't query pageserver while it is down.
-    pg = env.postgres.create_start(
+    endpoint = env.endpoints.create_start(
         "test_pageserver_catchup_while_compute_down", config_lines=["shared_buffers=512MB"]
     )
 
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     # Create table, and insert some rows.
@@ -59,10 +59,10 @@ def test_pageserver_catchup_while_compute_down(neon_env_builder: NeonEnvBuilder)
     env.safekeepers[2].start()
 
     # restart compute node
-    pg.stop_and_destroy().create_start("test_pageserver_catchup_while_compute_down")
+    endpoint.stop_and_destroy().create_start("test_pageserver_catchup_while_compute_down")
 
     # Ensure that basebackup went correct and pageserver returned all data
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     cur.execute("SELECT count(*) FROM foo")
diff --git a/test_runner/regress/test_pageserver_restart.py b/test_runner/regress/test_pageserver_restart.py
index 453ddec0d4..6da5503fb1 100644
--- a/test_runner/regress/test_pageserver_restart.py
+++ b/test_runner/regress/test_pageserver_restart.py
@@ -11,9 +11,9 @@ def test_pageserver_restart(neon_env_builder: NeonEnvBuilder):
     env = neon_env_builder.init_start()
 
     env.neon_cli.create_branch("test_pageserver_restart")
-    pg = env.postgres.create_start("test_pageserver_restart")
+    endpoint = env.endpoints.create_start("test_pageserver_restart")
 
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     # Create table, and insert some rows. Make it big enough that it doesn't fit in
@@ -59,7 +59,7 @@ def test_pageserver_restart(neon_env_builder: NeonEnvBuilder):
     client = env.pageserver.http_client()
     tenant_status = client.tenant_status(env.initial_tenant)
     log.info("Tenant status : %s", tenant_status)
-    assert tenant_status["state"] == "Loading"
+    assert tenant_status["state"]["slug"] == "Loading"
 
     # Try to read. This waits until the loading finishes, and then return normally.
     cur.execute("SELECT count(*) FROM foo")
@@ -84,13 +84,13 @@ def test_pageserver_chaos(neon_env_builder: NeonEnvBuilder):
         }
     )
     env.neon_cli.create_timeline("test_pageserver_chaos", tenant_id=tenant)
-    pg = env.postgres.create_start("test_pageserver_chaos", tenant_id=tenant)
+    endpoint = env.endpoints.create_start("test_pageserver_chaos", tenant_id=tenant)
 
     # Create table, and insert some rows. Make it big enough that it doesn't fit in
     # shared_buffers, otherwise the SELECT after restart will just return answer
     # from shared_buffers without hitting the page server, which defeats the point
     # of this test.
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute("CREATE TABLE foo (id int, t text, updates int)")
             cur.execute("CREATE INDEX ON foo (id)")
@@ -116,12 +116,12 @@ def test_pageserver_chaos(neon_env_builder: NeonEnvBuilder):
 
     # Update the whole table, then immediately kill and restart the pageserver
     for i in range(1, 15):
-        pg.safe_psql("UPDATE foo set updates = updates + 1")
+        endpoint.safe_psql("UPDATE foo set updates = updates + 1")
 
         # This kills the pageserver immediately, to simulate a crash
         env.pageserver.stop(immediate=True)
         env.pageserver.start()
 
         # Check that all the updates are visible
-        num_updates = pg.safe_psql("SELECT sum(updates) FROM foo")[0][0]
+        num_updates = endpoint.safe_psql("SELECT sum(updates) FROM foo")[0][0]
         assert num_updates == i * 100000
diff --git a/test_runner/regress/test_pageserver_restarts_under_workload.py b/test_runner/regress/test_pageserver_restarts_under_workload.py
index eab8b112f0..bc3f3f2be4 100644
--- a/test_runner/regress/test_pageserver_restarts_under_workload.py
+++ b/test_runner/regress/test_pageserver_restarts_under_workload.py
@@ -5,7 +5,7 @@ import threading
 import time
 
 from fixtures.log_helper import log
-from fixtures.neon_fixtures import NeonEnv, PgBin, Postgres
+from fixtures.neon_fixtures import NeonEnv, PgBin
 
 
 # Test restarting page server, while safekeeper and compute node keep
@@ -13,7 +13,7 @@ from fixtures.neon_fixtures import NeonEnv, PgBin, Postgres
 def test_pageserver_restarts_under_worload(neon_simple_env: NeonEnv, pg_bin: PgBin):
     env = neon_simple_env
     env.neon_cli.create_branch("test_pageserver_restarts")
-    pg = env.postgres.create_start("test_pageserver_restarts")
+    endpoint = env.endpoints.create_start("test_pageserver_restarts")
     n_restarts = 10
     scale = 10
 
@@ -23,13 +23,12 @@ def test_pageserver_restarts_under_worload(neon_simple_env: NeonEnv, pg_bin: PgB
         r".*Gc failed, retrying in \S+: Cannot run GC iteration on inactive tenant"
     )
 
-    def run_pgbench(pg: Postgres):
-        connstr = pg.connstr()
+    def run_pgbench(connstr: str):
         log.info(f"Start a pgbench workload on pg {connstr}")
         pg_bin.run_capture(["pgbench", "-i", f"-s{scale}", connstr])
         pg_bin.run_capture(["pgbench", f"-T{n_restarts}", connstr])
 
-    thread = threading.Thread(target=run_pgbench, args=(pg,), daemon=True)
+    thread = threading.Thread(target=run_pgbench, args=(endpoint.connstr(),), daemon=True)
     thread.start()
 
     for i in range(n_restarts):
diff --git a/test_runner/regress/test_parallel_copy.py b/test_runner/regress/test_parallel_copy.py
index 59f19026cc..577bbc21bf 100644
--- a/test_runner/regress/test_parallel_copy.py
+++ b/test_runner/regress/test_parallel_copy.py
@@ -2,7 +2,7 @@ import asyncio
 from io import BytesIO
 
 from fixtures.log_helper import log
-from fixtures.neon_fixtures import NeonEnv, Postgres
+from fixtures.neon_fixtures import Endpoint, NeonEnv
 
 
 async def repeat_bytes(buf, repetitions: int):
@@ -10,7 +10,7 @@ async def repeat_bytes(buf, repetitions: int):
         yield buf
 
 
-async def copy_test_data_to_table(pg: Postgres, worker_id: int, table_name: str):
+async def copy_test_data_to_table(endpoint: Endpoint, worker_id: int, table_name: str):
     buf = BytesIO()
     for i in range(1000):
         buf.write(
@@ -20,7 +20,7 @@ async def copy_test_data_to_table(pg: Postgres, worker_id: int, table_name: str)
 
     copy_input = repeat_bytes(buf.read(), 5000)
 
-    pg_conn = await pg.connect_async()
+    pg_conn = await endpoint.connect_async()
 
     # PgProtocol.connect_async sets statement_timeout to 2 minutes.
     # That's not enough for this test, on a slow system in debug mode.
@@ -29,10 +29,10 @@ async def copy_test_data_to_table(pg: Postgres, worker_id: int, table_name: str)
     await pg_conn.copy_to_table(table_name, source=copy_input)
 
 
-async def parallel_load_same_table(pg: Postgres, n_parallel: int):
+async def parallel_load_same_table(endpoint: Endpoint, n_parallel: int):
     workers = []
     for worker_id in range(n_parallel):
-        worker = copy_test_data_to_table(pg, worker_id, "copytest")
+        worker = copy_test_data_to_table(endpoint, worker_id, "copytest")
         workers.append(asyncio.create_task(worker))
 
     # await all workers
@@ -43,13 +43,13 @@ async def parallel_load_same_table(pg: Postgres, n_parallel: int):
 def test_parallel_copy(neon_simple_env: NeonEnv, n_parallel=5):
     env = neon_simple_env
     env.neon_cli.create_branch("test_parallel_copy", "empty")
-    pg = env.postgres.create_start("test_parallel_copy")
+    endpoint = env.endpoints.create_start("test_parallel_copy")
     log.info("postgres is running on 'test_parallel_copy' branch")
 
     # Create test table
-    conn = pg.connect()
+    conn = endpoint.connect()
     cur = conn.cursor()
     cur.execute("CREATE TABLE copytest (i int, t text)")
 
     # Run COPY TO to load the table with parallel connections.
-    asyncio.run(parallel_load_same_table(pg, n_parallel))
+    asyncio.run(parallel_load_same_table(endpoint, n_parallel))
diff --git a/test_runner/regress/test_pg_regress.py b/test_runner/regress/test_pg_regress.py
index 5eb1ebb3de..64625ea4ee 100644
--- a/test_runner/regress/test_pg_regress.py
+++ b/test_runner/regress/test_pg_regress.py
@@ -24,8 +24,8 @@ def test_pg_regress(
 
     env.neon_cli.create_branch("test_pg_regress", "empty")
     # Connect to postgres and create a database called "regression".
-    pg = env.postgres.create_start("test_pg_regress")
-    pg.safe_psql("CREATE DATABASE regression")
+    endpoint = env.endpoints.create_start("test_pg_regress")
+    endpoint.safe_psql("CREATE DATABASE regression")
 
     # Create some local directories for pg_regress to run in.
     runpath = test_output_dir / "regress"
@@ -49,9 +49,9 @@ def test_pg_regress(
     ]
 
     env_vars = {
-        "PGPORT": str(pg.default_options["port"]),
-        "PGUSER": pg.default_options["user"],
-        "PGHOST": pg.default_options["host"],
+        "PGPORT": str(endpoint.default_options["port"]),
+        "PGUSER": endpoint.default_options["user"],
+        "PGHOST": endpoint.default_options["host"],
     }
 
     # Run the command.
@@ -61,10 +61,10 @@ def test_pg_regress(
         pg_bin.run(pg_regress_command, env=env_vars, cwd=runpath)
 
         # checkpoint one more time to ensure that the lsn we get is the latest one
-        pg.safe_psql("CHECKPOINT")
+        endpoint.safe_psql("CHECKPOINT")
 
         # Check that we restore the content of the datadir correctly
-        check_restored_datadir_content(test_output_dir, env, pg)
+        check_restored_datadir_content(test_output_dir, env, endpoint)
 
 
 # Run the PostgreSQL "isolation" tests, in src/test/isolation.
@@ -85,8 +85,10 @@ def test_isolation(
     env.neon_cli.create_branch("test_isolation", "empty")
     # Connect to postgres and create a database called "regression".
     # isolation tests use prepared transactions, so enable them
-    pg = env.postgres.create_start("test_isolation", config_lines=["max_prepared_transactions=100"])
-    pg.safe_psql("CREATE DATABASE isolation_regression")
+    endpoint = env.endpoints.create_start(
+        "test_isolation", config_lines=["max_prepared_transactions=100"]
+    )
+    endpoint.safe_psql("CREATE DATABASE isolation_regression")
 
     # Create some local directories for pg_isolation_regress to run in.
     runpath = test_output_dir / "regress"
@@ -109,9 +111,9 @@ def test_isolation(
     ]
 
     env_vars = {
-        "PGPORT": str(pg.default_options["port"]),
-        "PGUSER": pg.default_options["user"],
-        "PGHOST": pg.default_options["host"],
+        "PGPORT": str(endpoint.default_options["port"]),
+        "PGUSER": endpoint.default_options["user"],
+        "PGHOST": endpoint.default_options["host"],
     }
 
     # Run the command.
@@ -135,8 +137,8 @@ def test_sql_regress(
 
     env.neon_cli.create_branch("test_sql_regress", "empty")
     # Connect to postgres and create a database called "regression".
-    pg = env.postgres.create_start("test_sql_regress")
-    pg.safe_psql("CREATE DATABASE regression")
+    endpoint = env.endpoints.create_start("test_sql_regress")
+    endpoint.safe_psql("CREATE DATABASE regression")
 
     # Create some local directories for pg_regress to run in.
     runpath = test_output_dir / "regress"
@@ -160,9 +162,9 @@ def test_sql_regress(
     ]
 
     env_vars = {
-        "PGPORT": str(pg.default_options["port"]),
-        "PGUSER": pg.default_options["user"],
-        "PGHOST": pg.default_options["host"],
+        "PGPORT": str(endpoint.default_options["port"]),
+        "PGUSER": endpoint.default_options["user"],
+        "PGHOST": endpoint.default_options["host"],
     }
 
     # Run the command.
@@ -172,8 +174,8 @@ def test_sql_regress(
         pg_bin.run(pg_regress_command, env=env_vars, cwd=runpath)
 
         # checkpoint one more time to ensure that the lsn we get is the latest one
-        pg.safe_psql("CHECKPOINT")
-        pg.safe_psql("select pg_current_wal_insert_lsn()")[0][0]
+        endpoint.safe_psql("CHECKPOINT")
+        endpoint.safe_psql("select pg_current_wal_insert_lsn()")[0][0]
 
         # Check that we restore the content of the datadir correctly
-        check_restored_datadir_content(test_output_dir, env, pg)
+        check_restored_datadir_content(test_output_dir, env, endpoint)
diff --git a/test_runner/regress/test_pitr_gc.py b/test_runner/regress/test_pitr_gc.py
index fe4fbc0927..c2ea5b332a 100644
--- a/test_runner/regress/test_pitr_gc.py
+++ b/test_runner/regress/test_pitr_gc.py
@@ -15,10 +15,10 @@ def test_pitr_gc(neon_env_builder: NeonEnvBuilder):
     )
 
     env = neon_env_builder.init_start()
-    pgmain = env.postgres.create_start("main")
+    endpoint_main = env.endpoints.create_start("main")
     log.info("postgres is running on 'main' branch")
 
-    main_pg_conn = pgmain.connect()
+    main_pg_conn = endpoint_main.connect()
     main_cur = main_pg_conn.cursor()
     timeline = TimelineId(query_scalar(main_cur, "SHOW neon.timeline_id"))
 
@@ -62,10 +62,10 @@ def test_pitr_gc(neon_env_builder: NeonEnvBuilder):
     # It must have been preserved by PITR setting
     env.neon_cli.create_branch("test_pitr_gc_hundred", "main", ancestor_start_lsn=lsn_a)
 
-    pg_hundred = env.postgres.create_start("test_pitr_gc_hundred")
+    endpoint_hundred = env.endpoints.create_start("test_pitr_gc_hundred")
 
     # On the 'hundred' branch, we should see only 100 rows
-    hundred_pg_conn = pg_hundred.connect()
+    hundred_pg_conn = endpoint_hundred.connect()
     hundred_cur = hundred_pg_conn.cursor()
     hundred_cur.execute("SELECT count(*) FROM foo")
     assert hundred_cur.fetchone() == (100,)
diff --git a/test_runner/regress/test_proxy.py b/test_runner/regress/test_proxy.py
index 51fabdd2a1..ee6349436b 100644
--- a/test_runner/regress/test_proxy.py
+++ b/test_runner/regress/test_proxy.py
@@ -1,3 +1,5 @@
+import subprocess
+
 import psycopg2
 import pytest
 from fixtures.neon_fixtures import PSQL, NeonProxy, VanillaPostgres
@@ -134,3 +136,19 @@ def test_forward_params_to_client(static_proxy: NeonProxy):
             for name, value in cur.fetchall():
                 # Check that proxy has forwarded this parameter.
                 assert conn.get_parameter_status(name) == value
+
+
+@pytest.mark.timeout(5)
+def test_close_on_connections_exit(static_proxy: NeonProxy):
+    # Open two connections, send SIGTERM, then ensure that proxy doesn't exit
+    # until after connections close.
+    with static_proxy.connect(options="project=irrelevant"), static_proxy.connect(
+        options="project=irrelevant"
+    ):
+        static_proxy.terminate()
+        with pytest.raises(subprocess.TimeoutExpired):
+            static_proxy.wait_for_exit(timeout=2)
+        # Ensure we don't accept any more connections
+        with pytest.raises(psycopg2.OperationalError):
+            static_proxy.connect(options="project=irrelevant")
+    static_proxy.wait_for_exit()
diff --git a/test_runner/regress/test_read_trace.py b/test_runner/regress/test_read_trace.py
index be0eb76ccd..9ebe53fc17 100644
--- a/test_runner/regress/test_read_trace.py
+++ b/test_runner/regress/test_read_trace.py
@@ -21,22 +21,22 @@ def test_read_request_tracing(neon_env_builder: NeonEnvBuilder):
     )
 
     timeline = env.neon_cli.create_timeline("test_trace_replay", tenant_id=tenant)
-    pg = env.postgres.create_start("test_trace_replay", "main", tenant)
+    endpoint = env.endpoints.create_start("test_trace_replay", "main", tenant)
 
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute("create table t (i integer);")
             cur.execute(f"insert into t values (generate_series(1,{10000}));")
             cur.execute("select count(*) from t;")
-            tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-            timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+            tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+            timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
             current_lsn = Lsn(query_scalar(cur, "SELECT pg_current_wal_flush_lsn()"))
     # wait until pageserver receives that data
     pageserver_http = env.pageserver.http_client()
     wait_for_last_record_lsn(pageserver_http, tenant_id, timeline_id, current_lsn)
 
-    # Stop pg so we drop the connection and flush the traces
-    pg.stop()
+    # Stop postgres so we drop the connection and flush the traces
+    endpoint.stop()
 
     trace_path = env.repo_dir / "traces" / str(tenant) / str(timeline)
     assert trace_path.exists()
diff --git a/test_runner/regress/test_read_validation.py b/test_runner/regress/test_read_validation.py
index 47135dc56c..47a06359bb 100644
--- a/test_runner/regress/test_read_validation.py
+++ b/test_runner/regress/test_read_validation.py
@@ -17,10 +17,10 @@ def test_read_validation(neon_simple_env: NeonEnv):
     env = neon_simple_env
     env.neon_cli.create_branch("test_read_validation", "empty")
 
-    pg = env.postgres.create_start("test_read_validation")
+    endpoint = env.endpoints.create_start("test_read_validation")
     log.info("postgres is running on 'test_read_validation' branch")
 
-    with closing(pg.connect()) as con:
+    with closing(endpoint.connect()) as con:
         with con.cursor() as c:
             for e in extensions:
                 c.execute("create extension if not exists {};".format(e))
@@ -144,10 +144,10 @@ def test_read_validation_neg(neon_simple_env: NeonEnv):
 
     env.pageserver.allowed_errors.append(".*invalid LSN\\(0\\) in request.*")
 
-    pg = env.postgres.create_start("test_read_validation_neg")
+    endpoint = env.endpoints.create_start("test_read_validation_neg")
     log.info("postgres is running on 'test_read_validation_neg' branch")
 
-    with closing(pg.connect()) as con:
+    with closing(endpoint.connect()) as con:
         with con.cursor() as c:
             for e in extensions:
                 c.execute("create extension if not exists {};".format(e))
diff --git a/test_runner/regress/test_readonly_node.py b/test_runner/regress/test_readonly_node.py
index 69d6e427ce..2d641e36a7 100644
--- a/test_runner/regress/test_readonly_node.py
+++ b/test_runner/regress/test_readonly_node.py
@@ -15,12 +15,12 @@ from fixtures.utils import query_scalar
 def test_readonly_node(neon_simple_env: NeonEnv):
     env = neon_simple_env
     env.neon_cli.create_branch("test_readonly_node", "empty")
-    pgmain = env.postgres.create_start("test_readonly_node")
+    endpoint_main = env.endpoints.create_start("test_readonly_node")
     log.info("postgres is running on 'test_readonly_node' branch")
 
     env.pageserver.allowed_errors.append(".*basebackup .* failed: invalid basebackup lsn.*")
 
-    main_pg_conn = pgmain.connect()
+    main_pg_conn = endpoint_main.connect()
     main_cur = main_pg_conn.cursor()
 
     # Create table, and insert the first 100 rows
@@ -61,23 +61,23 @@ def test_readonly_node(neon_simple_env: NeonEnv):
     log.info("LSN after 400100 rows: " + lsn_c)
 
     # Create first read-only node at the point where only 100 rows were inserted
-    pg_hundred = env.postgres.create_start(
-        branch_name="test_readonly_node", node_name="test_readonly_node_hundred", lsn=lsn_a
+    endpoint_hundred = env.endpoints.create_start(
+        branch_name="test_readonly_node", endpoint_id="ep-readonly_node_hundred", lsn=lsn_a
     )
 
     # And another at the point where 200100 rows were inserted
-    pg_more = env.postgres.create_start(
-        branch_name="test_readonly_node", node_name="test_readonly_node_more", lsn=lsn_b
+    endpoint_more = env.endpoints.create_start(
+        branch_name="test_readonly_node", endpoint_id="ep-readonly_node_more", lsn=lsn_b
     )
 
     # On the 'hundred' node, we should see only 100 rows
-    hundred_pg_conn = pg_hundred.connect()
+    hundred_pg_conn = endpoint_hundred.connect()
     hundred_cur = hundred_pg_conn.cursor()
     hundred_cur.execute("SELECT count(*) FROM foo")
     assert hundred_cur.fetchone() == (100,)
 
     # On the 'more' node, we should see 100200 rows
-    more_pg_conn = pg_more.connect()
+    more_pg_conn = endpoint_more.connect()
     more_cur = more_pg_conn.cursor()
     more_cur.execute("SELECT count(*) FROM foo")
     assert more_cur.fetchone() == (200100,)
@@ -87,21 +87,21 @@ def test_readonly_node(neon_simple_env: NeonEnv):
     assert main_cur.fetchone() == (400100,)
 
     # Check creating a node at segment boundary
-    pg = env.postgres.create_start(
+    endpoint = env.endpoints.create_start(
         branch_name="test_readonly_node",
-        node_name="test_branch_segment_boundary",
+        endpoint_id="ep-branch_segment_boundary",
         lsn=Lsn("0/3000000"),
     )
-    cur = pg.connect().cursor()
+    cur = endpoint.connect().cursor()
     cur.execute("SELECT 1")
     assert cur.fetchone() == (1,)
 
     # Create node at pre-initdb lsn
     with pytest.raises(Exception, match="invalid basebackup lsn"):
         # compute node startup with invalid LSN should fail
-        env.postgres.create_start(
+        env.endpoints.create_start(
             branch_name="test_readonly_node",
-            node_name="test_readonly_node_preinitdb",
+            endpoint_id="ep-readonly_node_preinitdb",
             lsn=Lsn("0/42"),
         )
 
@@ -111,16 +111,16 @@ def test_timetravel(neon_simple_env: NeonEnv):
     env = neon_simple_env
     pageserver_http_client = env.pageserver.http_client()
     env.neon_cli.create_branch("test_timetravel", "empty")
-    pg = env.postgres.create_start("test_timetravel")
+    endpoint = env.endpoints.create_start("test_timetravel")
 
     client = env.pageserver.http_client()
 
-    tenant_id = pg.safe_psql("show neon.tenant_id")[0][0]
-    timeline_id = pg.safe_psql("show neon.timeline_id")[0][0]
+    tenant_id = endpoint.safe_psql("show neon.tenant_id")[0][0]
+    timeline_id = endpoint.safe_psql("show neon.timeline_id")[0][0]
 
     lsns = []
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute(
             """
         CREATE TABLE testtab(id serial primary key, iteration int, data text);
@@ -131,7 +131,7 @@ def test_timetravel(neon_simple_env: NeonEnv):
     lsns.append((0, current_lsn))
 
     for i in range(1, 5):
-        with pg.cursor() as cur:
+        with endpoint.cursor() as cur:
             cur.execute(f"UPDATE testtab SET iteration = {i}")
             current_lsn = Lsn(query_scalar(cur, "SELECT pg_current_wal_flush_lsn()"))
         lsns.append((i, current_lsn))
@@ -143,14 +143,14 @@ def test_timetravel(neon_simple_env: NeonEnv):
         pageserver_http_client.timeline_checkpoint(tenant_id, timeline_id)
 
     ##### Restart pageserver
-    env.postgres.stop_all()
+    env.endpoints.stop_all()
     env.pageserver.stop()
     env.pageserver.start()
 
     for i, lsn in lsns:
-        pg_old = env.postgres.create_start(
-            branch_name="test_timetravel", node_name=f"test_old_lsn_{i}", lsn=lsn
+        endpoint_old = env.endpoints.create_start(
+            branch_name="test_timetravel", endpoint_id=f"ep-old_lsn_{i}", lsn=lsn
         )
-        with pg_old.cursor() as cur:
+        with endpoint_old.cursor() as cur:
             assert query_scalar(cur, f"select count(*) from testtab where iteration={i}") == 100000
             assert query_scalar(cur, f"select count(*) from testtab where iteration<>{i}") == 0
diff --git a/test_runner/regress/test_recovery.py b/test_runner/regress/test_recovery.py
index 09644eaaa1..76e97a35a4 100644
--- a/test_runner/regress/test_recovery.py
+++ b/test_runner/regress/test_recovery.py
@@ -22,10 +22,10 @@ def test_pageserver_recovery(neon_env_builder: NeonEnvBuilder):
     # Create a branch for us
     env.neon_cli.create_branch("test_pageserver_recovery", "main")
 
-    pg = env.postgres.create_start("test_pageserver_recovery")
+    endpoint = env.endpoints.create_start("test_pageserver_recovery")
     log.info("postgres is running on 'test_pageserver_recovery' branch")
 
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             with env.pageserver.http_client() as pageserver_http:
                 # Create and initialize test table
@@ -54,7 +54,7 @@ def test_pageserver_recovery(neon_env_builder: NeonEnvBuilder):
     env.pageserver.stop()
     env.pageserver.start()
 
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute("select count(*) from foo")
             assert cur.fetchone() == (100000,)
diff --git a/test_runner/regress/test_remote_storage.py b/test_runner/regress/test_remote_storage.py
index 222305f006..6de5f7db04 100644
--- a/test_runner/regress/test_remote_storage.py
+++ b/test_runner/regress/test_remote_storage.py
@@ -87,17 +87,17 @@ def test_remote_storage_backup_and_restore(
     env.pageserver.allowed_errors.append(".*simulated failure of remote operation.*")
 
     pageserver_http = env.pageserver.http_client()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     client = env.pageserver.http_client()
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     checkpoint_numbers = range(1, 3)
 
     for checkpoint_number in checkpoint_numbers:
-        with pg.cursor() as cur:
+        with endpoint.cursor() as cur:
             cur.execute(
                 f"""
                 CREATE TABLE t{checkpoint_number}(id int primary key, data text);
@@ -126,7 +126,7 @@ def test_remote_storage_backup_and_restore(
     )
 
     ##### Stop the first pageserver instance, erase all its data
-    env.postgres.stop_all()
+    env.endpoints.stop_all()
     env.pageserver.stop()
 
     dir_to_clear = Path(env.repo_dir) / "tenants"
@@ -187,8 +187,8 @@ def test_remote_storage_backup_and_restore(
     ), "current db Lsn should should not be less than the one stored on remote storage"
 
     log.info("select some data, this will cause layers to be downloaded")
-    pg = env.postgres.create_start("main")
-    with pg.cursor() as cur:
+    endpoint = env.endpoints.create_start("main")
+    with endpoint.cursor() as cur:
         for checkpoint_number in checkpoint_numbers:
             assert (
                 query_scalar(cur, f"SELECT data FROM t{checkpoint_number} WHERE id = {data_id};")
@@ -238,9 +238,9 @@ def test_remote_storage_upload_queue_retries(
 
     client = env.pageserver.http_client()
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
 
-    pg.safe_psql("CREATE TABLE foo (id INTEGER PRIMARY KEY, val text)")
+    endpoint.safe_psql("CREATE TABLE foo (id INTEGER PRIMARY KEY, val text)")
 
     def configure_storage_sync_failpoints(action):
         client.configure_failpoints(
@@ -253,7 +253,7 @@ def test_remote_storage_upload_queue_retries(
 
     def overwrite_data_and_wait_for_it_to_arrive_at_pageserver(data):
         # create initial set of layers & upload them with failpoints configured
-        pg.safe_psql_many(
+        endpoint.safe_psql_many(
             [
                 f"""
                INSERT INTO foo (id, val)
@@ -266,7 +266,7 @@ def test_remote_storage_upload_queue_retries(
                 "VACUUM foo",
             ]
         )
-        wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
     def get_queued_count(file_kind, op_kind):
         val = client.get_remote_timeline_client_metric(
@@ -343,7 +343,7 @@ def test_remote_storage_upload_queue_retries(
     #      but how do we validate the result after restore?
 
     env.pageserver.stop(immediate=True)
-    env.postgres.stop_all()
+    env.endpoints.stop_all()
 
     dir_to_clear = Path(env.repo_dir) / "tenants"
     shutil.rmtree(dir_to_clear)
@@ -357,8 +357,8 @@ def test_remote_storage_upload_queue_retries(
     wait_until_tenant_active(client, tenant_id)
 
     log.info("restarting postgres to validate")
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
-    with pg.cursor() as cur:
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
+    with endpoint.cursor() as cur:
         assert query_scalar(cur, "SELECT COUNT(*) FROM foo WHERE val = 'd'") == 10000
 
 
@@ -394,13 +394,13 @@ def test_remote_timeline_client_calls_started_metric(
 
     client = env.pageserver.http_client()
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
 
-    pg.safe_psql("CREATE TABLE foo (id INTEGER PRIMARY KEY, val text)")
+    endpoint.safe_psql("CREATE TABLE foo (id INTEGER PRIMARY KEY, val text)")
 
     def overwrite_data_and_wait_for_it_to_arrive_at_pageserver(data):
         # create initial set of layers & upload them with failpoints configured
-        pg.safe_psql_many(
+        endpoint.safe_psql_many(
             [
                 f"""
                INSERT INTO foo (id, val)
@@ -413,7 +413,7 @@ def test_remote_timeline_client_calls_started_metric(
                 "VACUUM foo",
             ]
         )
-        wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
     calls_started: Dict[Tuple[str, str], List[int]] = {
         ("layer", "upload"): [0],
@@ -478,7 +478,7 @@ def test_remote_timeline_client_calls_started_metric(
     )
 
     env.pageserver.stop(immediate=True)
-    env.postgres.stop_all()
+    env.endpoints.stop_all()
 
     dir_to_clear = Path(env.repo_dir) / "tenants"
     shutil.rmtree(dir_to_clear)
@@ -492,8 +492,8 @@ def test_remote_timeline_client_calls_started_metric(
     wait_until_tenant_active(client, tenant_id)
 
     log.info("restarting postgres to validate")
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
-    with pg.cursor() as cur:
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
+    with endpoint.cursor() as cur:
         assert query_scalar(cur, "SELECT COUNT(*) FROM foo WHERE val = 'd'") == 10000
 
     # ensure that we updated the calls_started download metric
@@ -543,17 +543,17 @@ def test_timeline_deletion_with_files_stuck_in_upload_queue(
         )
         return int(val) if val is not None else val
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
 
     client.configure_failpoints(("before-upload-layer", "return"))
 
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         [
             "CREATE TABLE foo (x INTEGER)",
             "INSERT INTO foo SELECT g FROM generate_series(1, 10000) g",
         ]
     )
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
     # Kick off a checkpoint operation.
     # It will get stuck in remote_client.wait_completion(), since the select query will have
@@ -627,8 +627,8 @@ def test_empty_branch_remote_storage_upload(
     new_branch_name = "new_branch"
     new_branch_timeline_id = env.neon_cli.create_branch(new_branch_name, "main", env.initial_tenant)
 
-    with env.postgres.create_start(new_branch_name, tenant_id=env.initial_tenant) as pg:
-        wait_for_last_flush_lsn(env, pg, env.initial_tenant, new_branch_timeline_id)
+    with env.endpoints.create_start(new_branch_name, tenant_id=env.initial_tenant) as endpoint:
+        wait_for_last_flush_lsn(env, endpoint, env.initial_tenant, new_branch_timeline_id)
     wait_upload_queue_empty(client, env.initial_tenant, new_branch_timeline_id)
 
     timelines_before_detach = set(
@@ -676,8 +676,8 @@ def test_empty_branch_remote_storage_upload_on_restart(
     new_branch_name = "new_branch"
     new_branch_timeline_id = env.neon_cli.create_branch(new_branch_name, "main", env.initial_tenant)
 
-    with env.postgres.create_start(new_branch_name, tenant_id=env.initial_tenant) as pg:
-        wait_for_last_flush_lsn(env, pg, env.initial_tenant, new_branch_timeline_id)
+    with env.endpoints.create_start(new_branch_name, tenant_id=env.initial_tenant) as endpoint:
+        wait_for_last_flush_lsn(env, endpoint, env.initial_tenant, new_branch_timeline_id)
     wait_upload_queue_empty(client, env.initial_tenant, new_branch_timeline_id)
 
     env.pageserver.stop()
diff --git a/test_runner/regress/test_subxacts.py b/test_runner/regress/test_subxacts.py
index 42234bf535..494820ef8e 100644
--- a/test_runner/regress/test_subxacts.py
+++ b/test_runner/regress/test_subxacts.py
@@ -11,10 +11,10 @@ from fixtures.neon_fixtures import NeonEnv, check_restored_datadir_content
 def test_subxacts(neon_simple_env: NeonEnv, test_output_dir):
     env = neon_simple_env
     env.neon_cli.create_branch("test_subxacts", "empty")
-    pg = env.postgres.create_start("test_subxacts")
+    endpoint = env.endpoints.create_start("test_subxacts")
 
     log.info("postgres is running on 'test_subxacts' branch")
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     cur.execute(
@@ -37,4 +37,4 @@ def test_subxacts(neon_simple_env: NeonEnv, test_output_dir):
     cur.execute("checkpoint")
 
     # Check that we can restore the content of the datadir correctly
-    check_restored_datadir_content(test_output_dir, env, pg)
+    check_restored_datadir_content(test_output_dir, env, endpoint)
diff --git a/test_runner/regress/test_tenant_conf.py b/test_runner/regress/test_tenant_conf.py
index 67aba227e5..8677a554f7 100644
--- a/test_runner/regress/test_tenant_conf.py
+++ b/test_runner/regress/test_tenant_conf.py
@@ -1,3 +1,4 @@
+import json
 from contextlib import closing
 
 import psycopg2.extras
@@ -7,7 +8,7 @@ from fixtures.neon_fixtures import (
     NeonEnvBuilder,
     RemoteStorageKind,
 )
-from fixtures.pageserver.utils import assert_tenant_status, wait_for_upload
+from fixtures.pageserver.utils import assert_tenant_state, wait_for_upload
 from fixtures.types import Lsn
 from fixtures.utils import wait_until
 
@@ -18,9 +19,16 @@ def test_tenant_config(neon_env_builder: NeonEnvBuilder):
     neon_env_builder.pageserver_config_override = """
 page_cache_size=444;
 wait_lsn_timeout='111 s';
-tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
+[tenant_config]
+checkpoint_distance = 10000
+compaction_target_size = 1048576
+evictions_low_residence_duration_metric_threshold = "2 days"
+eviction_policy = { "kind" = "LayerAccessThreshold", period = "20s", threshold = "23 hours" }
+"""
 
     env = neon_env_builder.init_start()
+    # we configure eviction but no remote storage, there might be error lines
+    env.pageserver.allowed_errors.append(".* no remote storage configured, cannot evict layers .*")
     http_client = env.pageserver.http_client()
 
     # Check that we raise on misspelled configs
@@ -39,15 +47,13 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     new_conf = {
         "checkpoint_distance": "20000",
         "gc_period": "30sec",
+        "evictions_low_residence_duration_metric_threshold": "42s",
+        "eviction_policy": json.dumps({"kind": "NoEviction"}),
     }
     tenant, _ = env.neon_cli.create_tenant(conf=new_conf)
 
     env.neon_cli.create_timeline("test_tenant_conf", tenant_id=tenant)
-    env.postgres.create_start(
-        "test_tenant_conf",
-        "main",
-        tenant,
-    )
+    env.endpoints.create_start("test_tenant_conf", "main", tenant)
 
     # check the configuration of the default tenant
     # it should match global configuration
@@ -82,6 +88,12 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     assert effective_config["gc_period"] == "1h"
     assert effective_config["image_creation_threshold"] == 3
     assert effective_config["pitr_interval"] == "7days"
+    assert effective_config["evictions_low_residence_duration_metric_threshold"] == "2days"
+    assert effective_config["eviction_policy"] == {
+        "kind": "LayerAccessThreshold",
+        "period": "20s",
+        "threshold": "23h",
+    }
 
     # check the configuration of the new tenant
     with closing(env.pageserver.connect()) as psconn:
@@ -116,6 +128,12 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     assert (
         new_effective_config["gc_period"] == "30s"
     ), "Specific 'gc_period' config should override the default value"
+    assert (
+        new_effective_config["evictions_low_residence_duration_metric_threshold"] == "42s"
+    ), "Should override default value"
+    assert new_effective_config["eviction_policy"] == {
+        "kind": "NoEviction"
+    }, "Specific 'eviction_policy' config should override the default value"
     assert new_effective_config["compaction_target_size"] == 1048576
     assert new_effective_config["compaction_period"] == "20s"
     assert new_effective_config["compaction_threshold"] == 10
@@ -129,6 +147,10 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
         "gc_period": "80sec",
         "compaction_period": "80sec",
         "image_creation_threshold": "2",
+        "evictions_low_residence_duration_metric_threshold": "23h",
+        "eviction_policy": json.dumps(
+            {"kind": "LayerAccessThreshold", "period": "80s", "threshold": "42h"}
+        ),
     }
     env.neon_cli.config_tenant(
         tenant_id=tenant,
@@ -171,6 +193,14 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     assert (
         updated_effective_config["compaction_period"] == "1m 20s"
     ), "Specific 'compaction_period' config should override the default value"
+    assert (
+        updated_effective_config["evictions_low_residence_duration_metric_threshold"] == "23h"
+    ), "Should override default value"
+    assert updated_effective_config["eviction_policy"] == {
+        "kind": "LayerAccessThreshold",
+        "period": "1m 20s",
+        "threshold": "1day 18h",
+    }, "Specific 'eviction_policy' config should override the default value"
     assert updated_effective_config["compaction_target_size"] == 1048576
     assert updated_effective_config["compaction_threshold"] == 10
     assert updated_effective_config["gc_horizon"] == 67108864
@@ -229,6 +259,12 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     assert final_effective_config["gc_horizon"] == 67108864
     assert final_effective_config["gc_period"] == "1h"
     assert final_effective_config["image_creation_threshold"] == 3
+    assert final_effective_config["evictions_low_residence_duration_metric_threshold"] == "2days"
+    assert final_effective_config["eviction_policy"] == {
+        "kind": "LayerAccessThreshold",
+        "period": "20s",
+        "threshold": "23h",
+    }
 
     # restart the pageserver and ensure that the config is still correct
     env.pageserver.stop()
@@ -278,7 +314,7 @@ def test_creating_tenant_conf_after_attach(neon_env_builder: NeonEnvBuilder):
     wait_until(
         number_of_iterations=5,
         interval=1,
-        func=lambda: assert_tenant_status(http_client, tenant_id, "Active"),
+        func=lambda: assert_tenant_state(http_client, tenant_id, "Active"),
     )
 
     env.neon_cli.config_tenant(tenant_id, {"gc_horizon": "1000000"})
@@ -289,3 +325,81 @@ def test_creating_tenant_conf_after_attach(neon_env_builder: NeonEnvBuilder):
     # dont test applying the setting here, we have that another test case to show it
     # we just care about being able to create the file
     assert len(contents_first) > len(contents_later)
+
+
+def test_live_reconfig_get_evictions_low_residence_duration_metric_threshold(
+    neon_env_builder: NeonEnvBuilder,
+):
+    neon_env_builder.enable_remote_storage(
+        remote_storage_kind=RemoteStorageKind.LOCAL_FS,
+        test_name="test_live_reconfig_get_evictions_low_residence_duration_metric_threshold",
+    )
+
+    env = neon_env_builder.init_start()
+    assert isinstance(env.remote_storage, LocalFsStorage)
+
+    (tenant_id, timeline_id) = env.neon_cli.create_tenant()
+    ps_http = env.pageserver.http_client()
+
+    def get_metric():
+        metrics = ps_http.get_metrics()
+        metric = metrics.query_one(
+            "pageserver_evictions_with_low_residence_duration_total",
+            {
+                "tenant_id": str(tenant_id),
+                "timeline_id": str(timeline_id),
+            },
+        )
+        return metric
+
+    default_value = ps_http.tenant_config(tenant_id).effective_config[
+        "evictions_low_residence_duration_metric_threshold"
+    ]
+    metric = get_metric()
+    assert int(metric.value) == 0, "metric is present with default value"
+
+    assert default_value == "1day"
+
+    ps_http.download_all_layers(tenant_id, timeline_id)
+    ps_http.evict_all_layers(tenant_id, timeline_id)
+    metric = get_metric()
+    assert int(metric.value) > 0, "metric is updated"
+
+    env.neon_cli.config_tenant(
+        tenant_id, {"evictions_low_residence_duration_metric_threshold": default_value}
+    )
+    updated_metric = get_metric()
+    assert int(updated_metric.value) == int(
+        metric.value
+    ), "metric is unchanged when setting same value"
+
+    env.neon_cli.config_tenant(
+        tenant_id, {"evictions_low_residence_duration_metric_threshold": "2day"}
+    )
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 2 * 24 * 60 * 60
+    assert int(metric.value) == 0
+
+    ps_http.download_all_layers(tenant_id, timeline_id)
+    ps_http.evict_all_layers(tenant_id, timeline_id)
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 2 * 24 * 60 * 60
+    assert int(metric.value) > 0
+
+    env.neon_cli.config_tenant(
+        tenant_id, {"evictions_low_residence_duration_metric_threshold": "2h"}
+    )
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 2 * 60 * 60
+    assert int(metric.value) == 0, "value resets if label changes"
+
+    ps_http.download_all_layers(tenant_id, timeline_id)
+    ps_http.evict_all_layers(tenant_id, timeline_id)
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 2 * 60 * 60
+    assert int(metric.value) > 0, "set a non-zero value for next step"
+
+    env.neon_cli.config_tenant(tenant_id, {})
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 24 * 60 * 60, "label resets to default"
+    assert int(metric.value) == 0, "value resets to default"
diff --git a/test_runner/regress/test_tenant_detach.py b/test_runner/regress/test_tenant_detach.py
index 58a010951e..847ae4b2b8 100644
--- a/test_runner/regress/test_tenant_detach.py
+++ b/test_runner/regress/test_tenant_detach.py
@@ -7,9 +7,9 @@ import asyncpg
 import pytest
 from fixtures.log_helper import log
 from fixtures.neon_fixtures import (
+    Endpoint,
     NeonEnv,
     NeonEnvBuilder,
-    Postgres,
     RemoteStorageKind,
     available_remote_storages,
 )
@@ -59,8 +59,8 @@ def test_tenant_reattach(
     # create new nenant
     tenant_id, timeline_id = env.neon_cli.create_tenant()
 
-    with env.postgres.create_start("main", tenant_id=tenant_id) as pg:
-        with pg.cursor() as cur:
+    with env.endpoints.create_start("main", tenant_id=tenant_id) as endpoint:
+        with endpoint.cursor() as cur:
             cur.execute("CREATE TABLE t(key int primary key, value text)")
             cur.execute("INSERT INTO t SELECT generate_series(1,100000), 'payload'")
             current_lsn = Lsn(query_scalar(cur, "SELECT pg_current_wal_flush_lsn()"))
@@ -99,8 +99,8 @@ def test_tenant_reattach(
 
     assert pageserver_last_record_lsn_before_detach == pageserver_last_record_lsn
 
-    with env.postgres.create_start("main", tenant_id=tenant_id) as pg:
-        with pg.cursor() as cur:
+    with env.endpoints.create_start("main", tenant_id=tenant_id) as endpoint:
+        with endpoint.cursor() as cur:
             assert query_scalar(cur, "SELECT count(*) FROM t") == 100000
 
         # Check that we had to retry the downloads
@@ -157,11 +157,11 @@ async def sleep_and_reattach(pageserver_http: PageserverHttpClient, tenant_id: T
 
 # async guts of test_tenant_reattach_while_bysy test
 async def reattach_while_busy(
-    env: NeonEnv, pg: Postgres, pageserver_http: PageserverHttpClient, tenant_id: TenantId
+    env: NeonEnv, endpoint: Endpoint, pageserver_http: PageserverHttpClient, tenant_id: TenantId
 ):
     workers = []
     for worker_id in range(num_connections):
-        pg_conn = await pg.connect_async()
+        pg_conn = await endpoint.connect_async()
         workers.append(asyncio.create_task(update_table(pg_conn)))
 
     workers.append(asyncio.create_task(sleep_and_reattach(pageserver_http, tenant_id)))
@@ -238,15 +238,15 @@ def test_tenant_reattach_while_busy(
         conf={"checkpoint_distance": "100000"}
     )
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
 
-    cur = pg.connect().cursor()
+    cur = endpoint.connect().cursor()
 
     cur.execute("CREATE TABLE t(id int primary key, counter int)")
     cur.execute(f"INSERT INTO t SELECT generate_series(1,{num_rows}), 0")
 
     # Run the test
-    asyncio.run(reattach_while_busy(env, pg, pageserver_http, tenant_id))
+    asyncio.run(reattach_while_busy(env, endpoint, pageserver_http, tenant_id))
 
     # Verify table contents
     assert query_scalar(cur, "SELECT count(*) FROM t") == num_rows
@@ -278,9 +278,9 @@ def test_tenant_detach_smoke(neon_env_builder: NeonEnvBuilder):
     # assert tenant exists on disk
     assert (env.repo_dir / "tenants" / str(tenant_id)).exists()
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
     # we rely upon autocommit after each statement
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         queries=[
             "CREATE TABLE t(key int primary key, value text)",
             "INSERT INTO t SELECT generate_series(1,100000), 'payload'",
@@ -339,9 +339,9 @@ def test_tenant_detach_ignored_tenant(neon_simple_env: NeonEnv):
     # assert tenant exists on disk
     assert (env.repo_dir / "tenants" / str(tenant_id)).exists()
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
     # we rely upon autocommit after each statement
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         queries=[
             "CREATE TABLE t(key int primary key, value text)",
             "INSERT INTO t SELECT generate_series(1,100000), 'payload'",
@@ -388,9 +388,9 @@ def test_tenant_detach_regular_tenant(neon_simple_env: NeonEnv):
     # assert tenant exists on disk
     assert (env.repo_dir / "tenants" / str(tenant_id)).exists()
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
     # we rely upon autocommit after each statement
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         queries=[
             "CREATE TABLE t(key int primary key, value text)",
             "INSERT INTO t SELECT generate_series(1,100000), 'payload'",
@@ -425,18 +425,18 @@ def test_detach_while_attaching(
     ##### First start, insert secret data and upload it to the remote storage
     env = neon_env_builder.init_start()
     pageserver_http = env.pageserver.http_client()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     client = env.pageserver.http_client()
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     # Create table, and insert some rows. Make it big enough that it doesn't fit in
     # shared_buffers, otherwise the SELECT after restart will just return answer
     # from shared_buffers without hitting the page server, which defeats the point
     # of this test.
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("CREATE TABLE foo (t text)")
         cur.execute(
             """
@@ -477,7 +477,7 @@ def test_detach_while_attaching(
     # cycle are still running, things could get really confusing..
     pageserver_http.tenant_attach(tenant_id)
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("SELECT COUNT(*) FROM foo")
 
 
@@ -572,14 +572,14 @@ def test_ignored_tenant_download_missing_layers(
     )
     env = neon_env_builder.init_start()
     pageserver_http = env.pageserver.http_client()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     data_id = 1
     data_secret = "very secret secret"
-    insert_test_data(pageserver_http, tenant_id, timeline_id, data_id, data_secret, pg)
+    insert_test_data(pageserver_http, tenant_id, timeline_id, data_id, data_secret, endpoint)
 
     tenants_before_ignore = [tenant["id"] for tenant in pageserver_http.tenant_list()]
     tenants_before_ignore.sort()
@@ -611,9 +611,9 @@ def test_ignored_tenant_download_missing_layers(
     ]
     assert timelines_before_ignore == timelines_after_ignore, "Should have all timelines back"
 
-    pg.stop()
-    pg.start()
-    ensure_test_data(data_id, data_secret, pg)
+    endpoint.stop()
+    endpoint.start()
+    ensure_test_data(data_id, data_secret, endpoint)
 
 
 # Tests that it's possible to `load` broken tenants:
@@ -631,10 +631,10 @@ def test_ignored_tenant_stays_broken_without_metadata(
     )
     env = neon_env_builder.init_start()
     pageserver_http = env.pageserver.http_client()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     # ignore the tenant and remove its metadata
     pageserver_http.tenant_ignore(tenant_id)
@@ -666,9 +666,9 @@ def test_load_attach_negatives(
     )
     env = neon_env_builder.init_start()
     pageserver_http = env.pageserver.http_client()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
 
     env.pageserver.allowed_errors.append(".*tenant .*? already exists, state:.*")
     with pytest.raises(
@@ -707,16 +707,16 @@ def test_ignore_while_attaching(
 
     env = neon_env_builder.init_start()
     pageserver_http = env.pageserver.http_client()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     pageserver_http = env.pageserver.http_client()
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     data_id = 1
     data_secret = "very secret secret"
-    insert_test_data(pageserver_http, tenant_id, timeline_id, data_id, data_secret, pg)
+    insert_test_data(pageserver_http, tenant_id, timeline_id, data_id, data_secret, endpoint)
 
     tenants_before_ignore = [tenant["id"] for tenant in pageserver_http.tenant_list()]
 
@@ -754,9 +754,9 @@ def test_ignore_while_attaching(
 
     wait_until_tenant_state(pageserver_http, tenant_id, "Active", 5)
 
-    pg.stop()
-    pg.start()
-    ensure_test_data(data_id, data_secret, pg)
+    endpoint.stop()
+    endpoint.start()
+    ensure_test_data(data_id, data_secret, endpoint)
 
 
 def insert_test_data(
@@ -765,9 +765,9 @@ def insert_test_data(
     timeline_id: TimelineId,
     data_id: int,
     data: str,
-    pg: Postgres,
+    endpoint: Endpoint,
 ):
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute(
             f"""
             CREATE TABLE test(id int primary key, secret text);
@@ -787,8 +787,8 @@ def insert_test_data(
     wait_for_upload(pageserver_http, tenant_id, timeline_id, current_lsn)
 
 
-def ensure_test_data(data_id: int, data: str, pg: Postgres):
-    with pg.cursor() as cur:
+def ensure_test_data(data_id: int, data: str, endpoint: Endpoint):
+    with endpoint.cursor() as cur:
         assert (
             query_scalar(cur, f"SELECT secret FROM test WHERE id = {data_id};") == data
         ), "Should have timeline data back"
diff --git a/test_runner/regress/test_tenant_relocation.py b/test_runner/regress/test_tenant_relocation.py
index 8ad4bd1c11..180afd88cd 100644
--- a/test_runner/regress/test_tenant_relocation.py
+++ b/test_runner/regress/test_tenant_relocation.py
@@ -7,15 +7,15 @@ from typing import Any, Dict, Optional, Tuple
 import pytest
 from fixtures.log_helper import log
 from fixtures.neon_fixtures import (
+    Endpoint,
     NeonBroker,
     NeonEnv,
     NeonEnvBuilder,
     PortDistributor,
-    Postgres,
 )
 from fixtures.pageserver.http import PageserverHttpClient
 from fixtures.pageserver.utils import (
-    assert_tenant_status,
+    assert_tenant_state,
     tenant_exists,
     wait_for_last_record_lsn,
     wait_for_upload,
@@ -87,20 +87,20 @@ def new_pageserver_service(
 
 
 @contextmanager
-def pg_cur(pg):
-    with closing(pg.connect()) as conn:
+def pg_cur(endpoint):
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             yield cur
 
 
-def load(pg: Postgres, stop_event: threading.Event, load_ok_event: threading.Event):
+def load(endpoint: Endpoint, stop_event: threading.Event, load_ok_event: threading.Event):
     log.info("load started")
 
     inserted_ctr = 0
     failed = False
     while not stop_event.is_set():
         try:
-            with pg_cur(pg) as cur:
+            with pg_cur(endpoint) as cur:
                 cur.execute("INSERT INTO load VALUES ('some payload')")
                 inserted_ctr += 1
         except:  # noqa: E722
@@ -110,7 +110,7 @@ def load(pg: Postgres, stop_event: threading.Event, load_ok_event: threading.Eve
             load_ok_event.clear()
         else:
             if failed:
-                with pg_cur(pg) as cur:
+                with pg_cur(endpoint) as cur:
                     # if we recovered after failure verify that we have correct number of rows
                     log.info("recovering at %s", inserted_ctr)
                     cur.execute("SELECT count(*) FROM load")
@@ -124,14 +124,14 @@ def load(pg: Postgres, stop_event: threading.Event, load_ok_event: threading.Eve
 
 
 def populate_branch(
-    pg: Postgres,
+    endpoint: Endpoint,
     tenant_id: TenantId,
     ps_http: PageserverHttpClient,
     create_table: bool,
     expected_sum: Optional[int],
 ) -> Tuple[TimelineId, Lsn]:
     # insert some data
-    with pg_cur(pg) as cur:
+    with pg_cur(endpoint) as cur:
         cur.execute("SHOW neon.timeline_id")
         timeline_id = TimelineId(cur.fetchone()[0])
         log.info("timeline to relocate %s", timeline_id)
@@ -196,19 +196,19 @@ def check_timeline_attached(
 
 def switch_pg_to_new_pageserver(
     env: NeonEnv,
-    pg: Postgres,
+    endpoint: Endpoint,
     new_pageserver_port: int,
     tenant_id: TenantId,
     timeline_id: TimelineId,
 ) -> Path:
-    pg.stop()
+    endpoint.stop()
 
-    pg_config_file_path = Path(pg.config_file_path())
+    pg_config_file_path = Path(endpoint.config_file_path())
     pg_config_file_path.open("a").write(
         f"\nneon.pageserver_connstring = 'postgresql://no_user:@localhost:{new_pageserver_port}'"
     )
 
-    pg.start()
+    endpoint.start()
 
     timeline_to_detach_local_path = (
         env.repo_dir / "tenants" / str(tenant_id) / "timelines" / str(timeline_id)
@@ -226,8 +226,8 @@ def switch_pg_to_new_pageserver(
     return timeline_to_detach_local_path
 
 
-def post_migration_check(pg: Postgres, sum_before_migration: int, old_local_path: Path):
-    with pg_cur(pg) as cur:
+def post_migration_check(endpoint: Endpoint, sum_before_migration: int, old_local_path: Path):
+    with pg_cur(endpoint) as cur:
         # check that data is still there
         cur.execute("SELECT sum(key) FROM t")
         assert cur.fetchone() == (sum_before_migration,)
@@ -288,12 +288,12 @@ def test_tenant_relocation(
     log.info("tenant to relocate %s initial_timeline_id %s", tenant_id, initial_timeline_id)
 
     env.neon_cli.create_branch("test_tenant_relocation_main", tenant_id=tenant_id)
-    pg_main = env.postgres.create_start(
+    ep_main = env.endpoints.create_start(
         branch_name="test_tenant_relocation_main", tenant_id=tenant_id
     )
 
     timeline_id_main, current_lsn_main = populate_branch(
-        pg_main,
+        ep_main,
         tenant_id=tenant_id,
         ps_http=pageserver_http,
         create_table=True,
@@ -306,12 +306,12 @@ def test_tenant_relocation(
         ancestor_start_lsn=current_lsn_main,
         tenant_id=tenant_id,
     )
-    pg_second = env.postgres.create_start(
+    ep_second = env.endpoints.create_start(
         branch_name="test_tenant_relocation_second", tenant_id=tenant_id
     )
 
     timeline_id_second, current_lsn_second = populate_branch(
-        pg_second,
+        ep_second,
         tenant_id=tenant_id,
         ps_http=pageserver_http,
         create_table=False,
@@ -327,14 +327,14 @@ def test_tenant_relocation(
 
     if with_load == "with_load":
         # create load table
-        with pg_cur(pg_main) as cur:
+        with pg_cur(ep_main) as cur:
             cur.execute("CREATE TABLE load(value text)")
 
         load_stop_event = threading.Event()
         load_ok_event = threading.Event()
         load_thread = threading.Thread(
             target=load,
-            args=(pg_main, load_stop_event, load_ok_event),
+            args=(ep_main, load_stop_event, load_ok_event),
             daemon=True,  # To make sure the child dies when the parent errors
         )
         load_thread.start()
@@ -416,11 +416,11 @@ def test_tenant_relocation(
 
             # wait for tenant to finish attaching
             tenant_status = new_pageserver_http.tenant_status(tenant_id=tenant_id)
-            assert tenant_status["state"] in ["Attaching", "Active"]
+            assert tenant_status["state"]["slug"] in ["Attaching", "Active"]
             wait_until(
                 number_of_iterations=10,
                 interval=1,
-                func=lambda: assert_tenant_status(new_pageserver_http, tenant_id, "Active"),
+                func=lambda: assert_tenant_state(new_pageserver_http, tenant_id, "Active"),
             )
 
             check_timeline_attached(
@@ -450,7 +450,7 @@ def test_tenant_relocation(
 
         old_local_path_main = switch_pg_to_new_pageserver(
             env,
-            pg_main,
+            ep_main,
             new_pageserver_pg_port,
             tenant_id,
             timeline_id_main,
@@ -458,7 +458,7 @@ def test_tenant_relocation(
 
         old_local_path_second = switch_pg_to_new_pageserver(
             env,
-            pg_second,
+            ep_second,
             new_pageserver_pg_port,
             tenant_id,
             timeline_id_second,
@@ -475,11 +475,11 @@ def test_tenant_relocation(
             interval=1,
             func=lambda: tenant_exists(pageserver_http, tenant_id),
         )
-        post_migration_check(pg_main, 500500, old_local_path_main)
-        post_migration_check(pg_second, 1001000, old_local_path_second)
+        post_migration_check(ep_main, 500500, old_local_path_main)
+        post_migration_check(ep_second, 1001000, old_local_path_second)
 
         # ensure that we can successfully read all relations on the new pageserver
-        with pg_cur(pg_second) as cur:
+        with pg_cur(ep_second) as cur:
             cur.execute(
                 """
                 DO $$
diff --git a/test_runner/regress/test_tenant_size.py b/test_runner/regress/test_tenant_size.py
index 9037fe0045..2d905910f8 100644
--- a/test_runner/regress/test_tenant_size.py
+++ b/test_runner/regress/test_tenant_size.py
@@ -4,9 +4,9 @@ from typing import List, Tuple
 import pytest
 from fixtures.log_helper import log
 from fixtures.neon_fixtures import (
+    Endpoint,
     NeonEnv,
     NeonEnvBuilder,
-    Postgres,
     wait_for_last_flush_lsn,
     wait_for_wal_insert_lsn,
 )
@@ -28,12 +28,12 @@ def test_empty_tenant_size(neon_simple_env: NeonEnv, test_output_dir: Path):
     branch_name, main_timeline_id = env.neon_cli.list_timelines(tenant_id)[0]
     assert branch_name == main_branch_name
 
-    with env.postgres.create_start(
+    with env.endpoints.create_start(
         main_branch_name,
         tenant_id=tenant_id,
         config_lines=["autovacuum=off", "checkpoint_timeout=10min"],
-    ) as pg:
-        with pg.cursor() as cur:
+    ) as endpoint:
+        with endpoint.cursor() as cur:
             cur.execute("SELECT 1")
             row = cur.fetchone()
             assert row is not None
@@ -105,12 +105,12 @@ def test_branched_empty_timeline_size(neon_simple_env: NeonEnv, test_output_dir:
 
     first_branch_timeline_id = env.neon_cli.create_branch("first-branch", tenant_id=tenant_id)
 
-    with env.postgres.create_start("first-branch", tenant_id=tenant_id) as pg:
-        with pg.cursor() as cur:
+    with env.endpoints.create_start("first-branch", tenant_id=tenant_id) as endpoint:
+        with endpoint.cursor() as cur:
             cur.execute(
                 "CREATE TABLE t0 AS SELECT i::bigint n FROM generate_series(0, 1000000) s(i)"
             )
-        wait_for_last_flush_lsn(env, pg, tenant_id, first_branch_timeline_id)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, first_branch_timeline_id)
 
     size_after_branching = http_client.tenant_size(tenant_id)
     log.info(f"size_after_branching: {size_after_branching}")
@@ -164,12 +164,12 @@ def test_branched_from_many_empty_parents_size(neon_simple_env: NeonEnv, test_ou
 
     assert last_branch is not None
 
-    with env.postgres.create_start(last_branch_name, tenant_id=tenant_id) as pg:
-        with pg.cursor() as cur:
+    with env.endpoints.create_start(last_branch_name, tenant_id=tenant_id) as endpoint:
+        with endpoint.cursor() as cur:
             cur.execute(
                 "CREATE TABLE t0 AS SELECT i::bigint n FROM generate_series(0, 1000000) s(i)"
             )
-        wait_for_last_flush_lsn(env, pg, tenant_id, last_branch)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, last_branch)
 
     size_after_writes = http_client.tenant_size(tenant_id)
     assert size_after_writes > initial_size
@@ -194,11 +194,11 @@ def test_branch_point_within_horizon(neon_simple_env: NeonEnv, test_output_dir:
     (tenant_id, main_id) = env.neon_cli.create_tenant(conf={"gc_horizon": str(gc_horizon)})
     http_client = env.pageserver.http_client()
 
-    with env.postgres.create_start("main", tenant_id=tenant_id) as pg:
-        initdb_lsn = wait_for_last_flush_lsn(env, pg, tenant_id, main_id)
-        with pg.cursor() as cur:
+    with env.endpoints.create_start("main", tenant_id=tenant_id) as endpoint:
+        initdb_lsn = wait_for_last_flush_lsn(env, endpoint, tenant_id, main_id)
+        with endpoint.cursor() as cur:
             cur.execute("CREATE TABLE t0 AS SELECT i::bigint n FROM generate_series(0, 1000) s(i)")
-        flushed_lsn = wait_for_last_flush_lsn(env, pg, tenant_id, main_id)
+        flushed_lsn = wait_for_last_flush_lsn(env, endpoint, tenant_id, main_id)
 
     size_before_branching = http_client.tenant_size(tenant_id)
 
@@ -208,10 +208,10 @@ def test_branch_point_within_horizon(neon_simple_env: NeonEnv, test_output_dir:
         "branch", tenant_id=tenant_id, ancestor_start_lsn=flushed_lsn
     )
 
-    with env.postgres.create_start("branch", tenant_id=tenant_id) as pg:
-        with pg.cursor() as cur:
+    with env.endpoints.create_start("branch", tenant_id=tenant_id) as endpoint:
+        with endpoint.cursor() as cur:
             cur.execute("CREATE TABLE t1 AS SELECT i::bigint n FROM generate_series(0, 1000) s(i)")
-        wait_for_last_flush_lsn(env, pg, tenant_id, branch_id)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, branch_id)
 
     size_after = http_client.tenant_size(tenant_id)
 
@@ -237,17 +237,17 @@ def test_parent_within_horizon(neon_simple_env: NeonEnv, test_output_dir: Path):
     (tenant_id, main_id) = env.neon_cli.create_tenant(conf={"gc_horizon": str(gc_horizon)})
     http_client = env.pageserver.http_client()
 
-    with env.postgres.create_start("main", tenant_id=tenant_id) as pg:
-        initdb_lsn = wait_for_last_flush_lsn(env, pg, tenant_id, main_id)
-        with pg.cursor() as cur:
+    with env.endpoints.create_start("main", tenant_id=tenant_id) as endpoint:
+        initdb_lsn = wait_for_last_flush_lsn(env, endpoint, tenant_id, main_id)
+        with endpoint.cursor() as cur:
             cur.execute("CREATE TABLE t0 AS SELECT i::bigint n FROM generate_series(0, 1000) s(i)")
 
-        flushed_lsn = wait_for_last_flush_lsn(env, pg, tenant_id, main_id)
+        flushed_lsn = wait_for_last_flush_lsn(env, endpoint, tenant_id, main_id)
 
-        with pg.cursor() as cur:
+        with endpoint.cursor() as cur:
             cur.execute("CREATE TABLE t00 AS SELECT i::bigint n FROM generate_series(0, 2000) s(i)")
 
-        wait_for_last_flush_lsn(env, pg, tenant_id, main_id)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, main_id)
 
     size_before_branching = http_client.tenant_size(tenant_id)
 
@@ -257,10 +257,10 @@ def test_parent_within_horizon(neon_simple_env: NeonEnv, test_output_dir: Path):
         "branch", tenant_id=tenant_id, ancestor_start_lsn=flushed_lsn
     )
 
-    with env.postgres.create_start("branch", tenant_id=tenant_id) as pg:
-        with pg.cursor() as cur:
+    with env.endpoints.create_start("branch", tenant_id=tenant_id) as endpoint:
+        with endpoint.cursor() as cur:
             cur.execute("CREATE TABLE t1 AS SELECT i::bigint n FROM generate_series(0, 10000) s(i)")
-        wait_for_last_flush_lsn(env, pg, tenant_id, branch_id)
+        wait_for_last_flush_lsn(env, endpoint, tenant_id, branch_id)
 
     size_after = http_client.tenant_size(tenant_id)
 
@@ -297,12 +297,12 @@ def test_only_heads_within_horizon(neon_simple_env: NeonEnv, test_output_dir: Pa
     # gc is not expected to change the results
 
     for branch_name, amount in [("main", 2000), ("first", 15000), ("second", 3000)]:
-        with env.postgres.create_start(branch_name, tenant_id=tenant_id) as pg:
-            with pg.cursor() as cur:
+        with env.endpoints.create_start(branch_name, tenant_id=tenant_id) as endpoint:
+            with endpoint.cursor() as cur:
                 cur.execute(
                     f"CREATE TABLE t0 AS SELECT i::bigint n FROM generate_series(0, {amount}) s(i)"
                 )
-            wait_for_last_flush_lsn(env, pg, tenant_id, ids[branch_name])
+            wait_for_last_flush_lsn(env, endpoint, tenant_id, ids[branch_name])
             size_now = http_client.tenant_size(tenant_id)
             if latest_size is not None:
                 assert size_now > latest_size
@@ -359,7 +359,7 @@ def test_single_branch_get_tenant_size_grows(
 
     def get_current_consistent_size(
         env: NeonEnv,
-        pg: Postgres,
+        endpoint: Endpoint,
         size_debug_file,  # apparently there is no public signature for open()...
         http_client: PageserverHttpClient,
         tenant_id: TenantId,
@@ -368,7 +368,7 @@ def test_single_branch_get_tenant_size_grows(
         consistent = False
         size_debug = None
 
-        current_lsn = wait_for_wal_insert_lsn(env, pg, tenant_id, timeline_id)
+        current_lsn = wait_for_wal_insert_lsn(env, endpoint, tenant_id, timeline_id)
         # We want to make sure we have a self-consistent set of values.
         # Size changes with WAL, so only if both before and after getting
         # the size of the tenant reports the same WAL insert LSN, we're OK
@@ -382,35 +382,36 @@ def test_single_branch_get_tenant_size_grows(
             size, sizes = http_client.tenant_size_and_modelinputs(tenant_id)
             size_debug = http_client.tenant_size_debug(tenant_id)
 
-            after_lsn = wait_for_wal_insert_lsn(env, pg, tenant_id, timeline_id)
+            after_lsn = wait_for_wal_insert_lsn(env, endpoint, tenant_id, timeline_id)
             consistent = current_lsn == after_lsn
             current_lsn = after_lsn
         size_debug_file.write(size_debug)
+        assert size > 0
         return (current_lsn, size)
 
-    with env.postgres.create_start(
+    with env.endpoints.create_start(
         branch_name,
         tenant_id=tenant_id,
         ### autovacuum is disabled to limit WAL logging.
         config_lines=["autovacuum=off"],
-    ) as pg:
+    ) as endpoint:
         (initdb_lsn, size) = get_current_consistent_size(
-            env, pg, size_debug_file, http_client, tenant_id, timeline_id
+            env, endpoint, size_debug_file, http_client, tenant_id, timeline_id
         )
         collected_responses.append(("INITDB", initdb_lsn, size))
 
-        with pg.cursor() as cur:
+        with endpoint.cursor() as cur:
             cur.execute("CREATE TABLE t0 (i BIGINT NOT NULL) WITH (fillfactor = 40)")
 
         (current_lsn, size) = get_current_consistent_size(
-            env, pg, size_debug_file, http_client, tenant_id, timeline_id
+            env, endpoint, size_debug_file, http_client, tenant_id, timeline_id
         )
         collected_responses.append(("CREATE", current_lsn, size))
 
         batch_size = 100
 
         for i in range(3):
-            with pg.cursor() as cur:
+            with endpoint.cursor() as cur:
                 cur.execute(
                     f"INSERT INTO t0(i) SELECT i FROM generate_series({batch_size} * %s, ({batch_size} * (%s + 1)) - 1) s(i)",
                     (i, i),
@@ -419,26 +420,24 @@ def test_single_branch_get_tenant_size_grows(
             i += 1
 
             (current_lsn, size) = get_current_consistent_size(
-                env, pg, size_debug_file, http_client, tenant_id, timeline_id
+                env, endpoint, size_debug_file, http_client, tenant_id, timeline_id
             )
 
             prev_size = collected_responses[-1][2]
-            if size == 0:
-                assert prev_size == 0
-            else:
-                # branch start shouldn't be past gc_horizon yet
-                # thus the size should grow as we insert more data
-                # "gc_horizon" is tuned so that it kicks in _after_ the
-                # insert phase, but before the update phase ends.
-                assert (
-                    current_lsn - initdb_lsn <= gc_horizon
-                ), "Tuning of GC window is likely out-of-date"
-                assert size > prev_size
+
+            # branch start shouldn't be past gc_horizon yet
+            # thus the size should grow as we insert more data
+            # "gc_horizon" is tuned so that it kicks in _after_ the
+            # insert phase, but before the update phase ends.
+            assert (
+                current_lsn - initdb_lsn <= gc_horizon
+            ), "Tuning of GC window is likely out-of-date"
+            assert size > prev_size
 
             collected_responses.append(("INSERT", current_lsn, size))
 
         while True:
-            with pg.cursor() as cur:
+            with endpoint.cursor() as cur:
                 cur.execute(
                     f"UPDATE t0 SET i = -i WHERE i IN (SELECT i FROM t0 WHERE i > 0 LIMIT {batch_size})"
                 )
@@ -448,7 +447,7 @@ def test_single_branch_get_tenant_size_grows(
                 break
 
             (current_lsn, size) = get_current_consistent_size(
-                env, pg, size_debug_file, http_client, tenant_id, timeline_id
+                env, endpoint, size_debug_file, http_client, tenant_id, timeline_id
             )
 
             prev_size = collected_responses[-1][2]
@@ -458,7 +457,7 @@ def test_single_branch_get_tenant_size_grows(
             collected_responses.append(("UPDATE", current_lsn, size))
 
         while True:
-            with pg.cursor() as cur:
+            with endpoint.cursor() as cur:
                 cur.execute(f"DELETE FROM t0 WHERE i IN (SELECT i FROM t0 LIMIT {batch_size})")
                 deleted = cur.rowcount
 
@@ -466,7 +465,7 @@ def test_single_branch_get_tenant_size_grows(
                 break
 
             (current_lsn, size) = get_current_consistent_size(
-                env, pg, size_debug_file, http_client, tenant_id, timeline_id
+                env, endpoint, size_debug_file, http_client, tenant_id, timeline_id
             )
 
             prev_size = collected_responses[-1][2]
@@ -475,14 +474,14 @@ def test_single_branch_get_tenant_size_grows(
 
             collected_responses.append(("DELETE", current_lsn, size))
 
-        with pg.cursor() as cur:
+        with endpoint.cursor() as cur:
             cur.execute("DROP TABLE t0")
 
         # The size of the tenant should still be as large as before we dropped
         # the table, because the drop operation can still be undone in the PITR
         # defined by gc_horizon.
         (current_lsn, size) = get_current_consistent_size(
-            env, pg, size_debug_file, http_client, tenant_id, timeline_id
+            env, endpoint, size_debug_file, http_client, tenant_id, timeline_id
         )
 
         prev_size = collected_responses[-1][2]
@@ -491,6 +490,9 @@ def test_single_branch_get_tenant_size_grows(
 
         collected_responses.append(("DROP", current_lsn, size))
 
+    # Should have gone past gc_horizon, otherwise gc_horizon is too large
+    assert current_lsn - initdb_lsn > gc_horizon
+
     # this isn't too many lines to forget for a while. observed while
     # developing these tests that locally the value is a bit more than what we
     # get in the ci.
@@ -532,16 +534,16 @@ def test_get_tenant_size_with_multiple_branches(
 
     http_client = env.pageserver.http_client()
 
-    main_pg = env.postgres.create_start(main_branch_name, tenant_id=tenant_id)
+    main_endpoint = env.endpoints.create_start(main_branch_name, tenant_id=tenant_id)
 
     batch_size = 10000
 
-    with main_pg.cursor() as cur:
+    with main_endpoint.cursor() as cur:
         cur.execute(
             f"CREATE TABLE t0 AS SELECT i::bigint n FROM generate_series(0, {batch_size}) s(i)"
         )
 
-    wait_for_last_flush_lsn(env, main_pg, tenant_id, main_timeline_id)
+    wait_for_last_flush_lsn(env, main_endpoint, tenant_id, main_timeline_id)
     size_at_branch = http_client.tenant_size(tenant_id)
     assert size_at_branch > 0
 
@@ -552,23 +554,23 @@ def test_get_tenant_size_with_multiple_branches(
     size_after_first_branch = http_client.tenant_size(tenant_id)
     assert size_after_first_branch == size_at_branch
 
-    first_branch_pg = env.postgres.create_start("first-branch", tenant_id=tenant_id)
+    first_branch_endpoint = env.endpoints.create_start("first-branch", tenant_id=tenant_id)
 
-    with first_branch_pg.cursor() as cur:
+    with first_branch_endpoint.cursor() as cur:
         cur.execute(
             f"CREATE TABLE t1 AS SELECT i::bigint n FROM generate_series(0, {batch_size}) s(i)"
         )
 
-    wait_for_last_flush_lsn(env, first_branch_pg, tenant_id, first_branch_timeline_id)
+    wait_for_last_flush_lsn(env, first_branch_endpoint, tenant_id, first_branch_timeline_id)
     size_after_growing_first_branch = http_client.tenant_size(tenant_id)
     assert size_after_growing_first_branch > size_after_first_branch
 
-    with main_pg.cursor() as cur:
+    with main_endpoint.cursor() as cur:
         cur.execute(
             f"CREATE TABLE t1 AS SELECT i::bigint n FROM generate_series(0, 2*{batch_size}) s(i)"
         )
 
-    wait_for_last_flush_lsn(env, main_pg, tenant_id, main_timeline_id)
+    wait_for_last_flush_lsn(env, main_endpoint, tenant_id, main_timeline_id)
     size_after_continuing_on_main = http_client.tenant_size(tenant_id)
     assert size_after_continuing_on_main > size_after_growing_first_branch
 
@@ -578,31 +580,31 @@ def test_get_tenant_size_with_multiple_branches(
     size_after_second_branch = http_client.tenant_size(tenant_id)
     assert size_after_second_branch == size_after_continuing_on_main
 
-    second_branch_pg = env.postgres.create_start("second-branch", tenant_id=tenant_id)
+    second_branch_endpoint = env.endpoints.create_start("second-branch", tenant_id=tenant_id)
 
-    with second_branch_pg.cursor() as cur:
+    with second_branch_endpoint.cursor() as cur:
         cur.execute(
             f"CREATE TABLE t2 AS SELECT i::bigint n FROM generate_series(0, 3*{batch_size}) s(i)"
         )
 
-    wait_for_last_flush_lsn(env, second_branch_pg, tenant_id, second_branch_timeline_id)
+    wait_for_last_flush_lsn(env, second_branch_endpoint, tenant_id, second_branch_timeline_id)
     size_after_growing_second_branch = http_client.tenant_size(tenant_id)
     assert size_after_growing_second_branch > size_after_second_branch
 
-    with second_branch_pg.cursor() as cur:
+    with second_branch_endpoint.cursor() as cur:
         cur.execute("DROP TABLE t0")
         cur.execute("DROP TABLE t1")
         cur.execute("VACUUM FULL")
 
-    wait_for_last_flush_lsn(env, second_branch_pg, tenant_id, second_branch_timeline_id)
+    wait_for_last_flush_lsn(env, second_branch_endpoint, tenant_id, second_branch_timeline_id)
     size_after_thinning_branch = http_client.tenant_size(tenant_id)
     assert (
         size_after_thinning_branch > size_after_growing_second_branch
     ), "tenant_size should grow with dropped tables and full vacuum"
 
-    first_branch_pg.stop_and_destroy()
-    second_branch_pg.stop_and_destroy()
-    main_pg.stop()
+    first_branch_endpoint.stop_and_destroy()
+    second_branch_endpoint.stop_and_destroy()
+    main_endpoint.stop()
     env.pageserver.stop()
     env.pageserver.start()
 
diff --git a/test_runner/regress/test_tenant_tasks.py b/test_runner/regress/test_tenant_tasks.py
index 24b211e368..21e4af4127 100644
--- a/test_runner/regress/test_tenant_tasks.py
+++ b/test_runner/regress/test_tenant_tasks.py
@@ -1,5 +1,6 @@
 from fixtures.log_helper import log
 from fixtures.neon_fixtures import NeonEnvBuilder
+from fixtures.pageserver.utils import assert_tenant_state, wait_until_tenant_active
 from fixtures.types import TenantId, TimelineId
 from fixtures.utils import wait_until
 
@@ -25,19 +26,19 @@ def test_tenant_tasks(neon_env_builder: NeonEnvBuilder):
         for t in timelines:
             client.timeline_delete(tenant, t)
 
-    def assert_active(tenant):
-        assert get_state(tenant) == "Active"
-
     # Create tenant, start compute
     tenant, _ = env.neon_cli.create_tenant()
     env.neon_cli.create_timeline(name, tenant_id=tenant)
-    pg = env.postgres.create_start(name, tenant_id=tenant)
-    assert (
-        get_state(tenant) == "Active"
-    ), "Pageserver should activate a tenant and start background jobs if timelines are loaded"
+    endpoint = env.endpoints.create_start(name, tenant_id=tenant)
+    assert_tenant_state(
+        client,
+        tenant,
+        expected_state="Active",
+        message="Pageserver should activate a tenant and start background jobs if timelines are loaded",
+    )
 
     # Stop compute
-    pg.stop()
+    endpoint.stop()
 
     # Delete all timelines on all tenants.
     #
@@ -47,7 +48,7 @@ def test_tenant_tasks(neon_env_builder: NeonEnvBuilder):
     for tenant_info in client.tenant_list():
         tenant_id = TenantId(tenant_info["id"])
         delete_all_timelines(tenant_id)
-        wait_until(10, 0.2, lambda: assert_active(tenant_id))
+        wait_until_tenant_active(client, tenant_id, iterations=10, period=0.2)
 
     # Assert that all tasks finish quickly after tenant is detached
     task_starts = client.get_metric_value("pageserver_tenant_task_events_total", {"event": "start"})
diff --git a/test_runner/regress/test_tenants.py b/test_runner/regress/test_tenants.py
index 8021bf9914..8026d7f5c6 100644
--- a/test_runner/regress/test_tenants.py
+++ b/test_runner/regress/test_tenants.py
@@ -66,17 +66,17 @@ def test_tenants_normal_work(neon_env_builder: NeonEnvBuilder):
     env.neon_cli.create_timeline("test_tenants_normal_work", tenant_id=tenant_1)
     env.neon_cli.create_timeline("test_tenants_normal_work", tenant_id=tenant_2)
 
-    pg_tenant1 = env.postgres.create_start(
+    endpoint_tenant1 = env.endpoints.create_start(
         "test_tenants_normal_work",
         tenant_id=tenant_1,
     )
-    pg_tenant2 = env.postgres.create_start(
+    endpoint_tenant2 = env.endpoints.create_start(
         "test_tenants_normal_work",
         tenant_id=tenant_2,
     )
 
-    for pg in [pg_tenant1, pg_tenant2]:
-        with closing(pg.connect()) as conn:
+    for endpoint in [endpoint_tenant1, endpoint_tenant2]:
+        with closing(endpoint.connect()) as conn:
             with conn.cursor() as cur:
                 # we rely upon autocommit after each statement
                 # as waiting for acceptors happens there
@@ -97,11 +97,11 @@ def test_metrics_normal_work(neon_env_builder: NeonEnvBuilder):
     timeline_1 = env.neon_cli.create_timeline("test_metrics_normal_work", tenant_id=tenant_1)
     timeline_2 = env.neon_cli.create_timeline("test_metrics_normal_work", tenant_id=tenant_2)
 
-    pg_tenant1 = env.postgres.create_start("test_metrics_normal_work", tenant_id=tenant_1)
-    pg_tenant2 = env.postgres.create_start("test_metrics_normal_work", tenant_id=tenant_2)
+    endpoint_tenant1 = env.endpoints.create_start("test_metrics_normal_work", tenant_id=tenant_1)
+    endpoint_tenant2 = env.endpoints.create_start("test_metrics_normal_work", tenant_id=tenant_2)
 
-    for pg in [pg_tenant1, pg_tenant2]:
-        with closing(pg.connect()) as conn:
+    for endpoint in [endpoint_tenant1, endpoint_tenant2]:
+        with closing(endpoint.connect()) as conn:
             with conn.cursor() as cur:
                 cur.execute("CREATE TABLE t(key int primary key, value text)")
                 cur.execute("INSERT INTO t SELECT generate_series(1,100000), 'payload'")
@@ -242,11 +242,15 @@ def test_pageserver_metrics_removed_after_detach(
     env.neon_cli.create_timeline("test_metrics_removed_after_detach", tenant_id=tenant_1)
     env.neon_cli.create_timeline("test_metrics_removed_after_detach", tenant_id=tenant_2)
 
-    pg_tenant1 = env.postgres.create_start("test_metrics_removed_after_detach", tenant_id=tenant_1)
-    pg_tenant2 = env.postgres.create_start("test_metrics_removed_after_detach", tenant_id=tenant_2)
+    endpoint_tenant1 = env.endpoints.create_start(
+        "test_metrics_removed_after_detach", tenant_id=tenant_1
+    )
+    endpoint_tenant2 = env.endpoints.create_start(
+        "test_metrics_removed_after_detach", tenant_id=tenant_2
+    )
 
-    for pg in [pg_tenant1, pg_tenant2]:
-        with closing(pg.connect()) as conn:
+    for endpoint in [endpoint_tenant1, endpoint_tenant2]:
+        with closing(endpoint.connect()) as conn:
             with conn.cursor() as cur:
                 cur.execute("CREATE TABLE t(key int primary key, value text)")
                 cur.execute("INSERT INTO t SELECT generate_series(1,100000), 'payload'")
@@ -317,7 +321,7 @@ def test_pageserver_with_empty_tenants(
     ), f"Tenant {tenant_with_empty_timelines_dir} should have an empty timelines/ directory"
 
     # Trigger timeline re-initialization after pageserver restart
-    env.postgres.stop_all()
+    env.endpoints.stop_all()
     env.pageserver.stop()
 
     tenant_without_timelines_dir = env.initial_tenant
@@ -332,24 +336,24 @@ def test_pageserver_with_empty_tenants(
 
     [broken_tenant] = [t for t in tenants if t["id"] == str(tenant_without_timelines_dir)]
     assert (
-        broken_tenant["state"] == "Broken"
+        broken_tenant["state"]["slug"] == "Broken"
     ), f"Tenant {tenant_without_timelines_dir} without timelines dir should be broken"
 
     broken_tenant_status = client.tenant_status(tenant_without_timelines_dir)
     assert (
-        broken_tenant_status["state"] == "Broken"
+        broken_tenant_status["state"]["slug"] == "Broken"
     ), f"Tenant {tenant_without_timelines_dir} without timelines dir should be broken"
 
     assert env.pageserver.log_contains(".*Setting tenant as Broken state, reason:.*")
 
     [loaded_tenant] = [t for t in tenants if t["id"] == str(tenant_with_empty_timelines_dir)]
     assert (
-        loaded_tenant["state"] == "Active"
+        loaded_tenant["state"]["slug"] == "Active"
     ), "Tenant {tenant_with_empty_timelines_dir} with empty timelines dir should be active and ready for timeline creation"
 
     loaded_tenant_status = client.tenant_status(tenant_with_empty_timelines_dir)
     assert (
-        loaded_tenant_status["state"] == "Active"
+        loaded_tenant_status["state"]["slug"] == "Active"
     ), f"Tenant {tenant_with_empty_timelines_dir} without timelines dir should be active"
 
     time.sleep(1)  # to allow metrics propagation
@@ -357,11 +361,11 @@ def test_pageserver_with_empty_tenants(
     ps_metrics = client.get_metrics()
     broken_tenants_metric_filter = {
         "tenant_id": str(tenant_without_timelines_dir),
-        "state": "broken",
+        "state": "Broken",
     }
     active_tenants_metric_filter = {
         "tenant_id": str(tenant_with_empty_timelines_dir),
-        "state": "active",
+        "state": "Active",
     }
 
     tenant_active_count = int(
diff --git a/test_runner/regress/test_tenants_with_remote_storage.py b/test_runner/regress/test_tenants_with_remote_storage.py
index ec1c12a0d8..dca2cd3d28 100644
--- a/test_runner/regress/test_tenants_with_remote_storage.py
+++ b/test_runner/regress/test_tenants_with_remote_storage.py
@@ -15,16 +15,16 @@ from typing import List, Tuple
 import pytest
 from fixtures.log_helper import log
 from fixtures.neon_fixtures import (
+    Endpoint,
     LocalFsStorage,
     NeonEnv,
     NeonEnvBuilder,
-    Postgres,
     RemoteStorageKind,
     available_remote_storages,
-    wait_for_sk_commit_lsn_to_reach_remote_storage,
+    last_flush_lsn_upload,
 )
 from fixtures.pageserver.utils import (
-    assert_tenant_status,
+    assert_tenant_state,
     wait_for_last_record_lsn,
     wait_for_upload,
 )
@@ -32,10 +32,10 @@ from fixtures.types import Lsn, TenantId, TimelineId
 from fixtures.utils import query_scalar, wait_until
 
 
-async def tenant_workload(env: NeonEnv, pg: Postgres):
+async def tenant_workload(env: NeonEnv, endpoint: Endpoint):
     await env.pageserver.connect_async()
 
-    pg_conn = await pg.connect_async()
+    pg_conn = await endpoint.connect_async()
 
     await pg_conn.execute("CREATE TABLE t(key int primary key, value text)")
     for i in range(1, 100):
@@ -49,10 +49,10 @@ async def tenant_workload(env: NeonEnv, pg: Postgres):
         assert res == i * 1000
 
 
-async def all_tenants_workload(env: NeonEnv, tenants_pgs):
+async def all_tenants_workload(env: NeonEnv, tenants_endpoints):
     workers = []
-    for _, pg in tenants_pgs:
-        worker = tenant_workload(env, pg)
+    for _, endpoint in tenants_endpoints:
+        worker = tenant_workload(env, endpoint)
         workers.append(asyncio.create_task(worker))
 
     # await all workers
@@ -73,7 +73,7 @@ def test_tenants_many(neon_env_builder: NeonEnvBuilder, remote_storage_kind: Rem
         ".*init_tenant_mgr: marking .* as locally complete, while it doesnt exist in remote index.*"
     )
 
-    tenants_pgs: List[Tuple[TenantId, Postgres]] = []
+    tenants_endpoints: List[Tuple[TenantId, Endpoint]] = []
 
     for _ in range(1, 5):
         # Use a tiny checkpoint distance, to create a lot of layers quickly
@@ -84,18 +84,18 @@ def test_tenants_many(neon_env_builder: NeonEnvBuilder, remote_storage_kind: Rem
         )
         env.neon_cli.create_timeline("test_tenants_many", tenant_id=tenant)
 
-        pg = env.postgres.create_start(
+        endpoint = env.endpoints.create_start(
             "test_tenants_many",
             tenant_id=tenant,
         )
-        tenants_pgs.append((tenant, pg))
+        tenants_endpoints.append((tenant, endpoint))
 
-    asyncio.run(all_tenants_workload(env, tenants_pgs))
+    asyncio.run(all_tenants_workload(env, tenants_endpoints))
 
     # Wait for the remote storage uploads to finish
     pageserver_http = env.pageserver.http_client()
-    for tenant, pg in tenants_pgs:
-        res = pg.safe_psql_many(
+    for tenant, endpoint in tenants_endpoints:
+        res = endpoint.safe_psql_many(
             ["SHOW neon.tenant_id", "SHOW neon.timeline_id", "SELECT pg_current_wal_flush_lsn()"]
         )
         tenant_id = TenantId(res[0][0][0])
@@ -137,15 +137,15 @@ def test_tenants_attached_after_download(
     )
 
     pageserver_http = env.pageserver.http_client()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
     client = env.pageserver.http_client()
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     for checkpoint_number in range(1, 3):
-        with pg.cursor() as cur:
+        with endpoint.cursor() as cur:
             cur.execute(
                 f"""
                 CREATE TABLE t{checkpoint_number}(id int primary key, secret text);
@@ -174,11 +174,8 @@ def test_tenants_attached_after_download(
     )
 
     ##### Stop the pageserver, erase its layer file to force it being downloaded from S3
-    env.postgres.stop_all()
-
-    wait_for_sk_commit_lsn_to_reach_remote_storage(
-        tenant_id, timeline_id, env.safekeepers, env.pageserver
-    )
+    last_flush_lsn_upload(env, endpoint, tenant_id, timeline_id)
+    env.endpoints.stop_all()
 
     env.pageserver.stop()
 
@@ -202,7 +199,7 @@ def test_tenants_attached_after_download(
     wait_until(
         number_of_iterations=5,
         interval=1,
-        func=lambda: assert_tenant_status(client, tenant_id, "Active"),
+        func=lambda: assert_tenant_state(client, tenant_id, "Active"),
     )
 
     restored_timelines = client.timeline_list(tenant_id)
@@ -244,12 +241,12 @@ def test_tenant_redownloads_truncated_file_on_startup(
     env.pageserver.allowed_errors.append(".*No timelines to attach received.*")
 
     pageserver_http = env.pageserver.http_client()
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("CREATE TABLE t1 AS VALUES (123, 'foobar');")
         current_lsn = Lsn(query_scalar(cur, "SELECT pg_current_wal_flush_lsn()"))
 
@@ -257,7 +254,7 @@ def test_tenant_redownloads_truncated_file_on_startup(
     pageserver_http.timeline_checkpoint(tenant_id, timeline_id)
     wait_for_upload(pageserver_http, tenant_id, timeline_id, current_lsn)
 
-    env.postgres.stop_all()
+    env.endpoints.stop_all()
     env.pageserver.stop()
 
     timeline_dir = Path(env.repo_dir) / "tenants" / str(tenant_id) / "timelines" / str(timeline_id)
@@ -286,7 +283,7 @@ def test_tenant_redownloads_truncated_file_on_startup(
     wait_until(
         number_of_iterations=5,
         interval=1,
-        func=lambda: assert_tenant_status(client, tenant_id, "Active"),
+        func=lambda: assert_tenant_state(client, tenant_id, "Active"),
     )
 
     restored_timelines = client.timeline_list(tenant_id)
@@ -313,9 +310,9 @@ def test_tenant_redownloads_truncated_file_on_startup(
         os.stat(remote_layer_path).st_size == expected_size
     ), "truncated file should not had been uploaded around re-download"
 
-    pg = env.postgres.create_start("main")
+    endpoint = env.endpoints.create_start("main")
 
-    with pg.cursor() as cur:
+    with endpoint.cursor() as cur:
         cur.execute("INSERT INTO t1 VALUES (234, 'test data');")
         current_lsn = Lsn(query_scalar(cur, "SELECT pg_current_wal_flush_lsn()"))
 
diff --git a/test_runner/regress/test_timeline_size.py b/test_runner/regress/test_timeline_size.py
index 7c77e1fe59..db278d5646 100644
--- a/test_runner/regress/test_timeline_size.py
+++ b/test_runner/regress/test_timeline_size.py
@@ -12,18 +12,18 @@ import psycopg2.extras
 import pytest
 from fixtures.log_helper import log
 from fixtures.neon_fixtures import (
+    Endpoint,
     NeonEnv,
     NeonEnvBuilder,
     PgBin,
     PortDistributor,
-    Postgres,
     RemoteStorageKind,
     VanillaPostgres,
     wait_for_last_flush_lsn,
 )
 from fixtures.pageserver.http import PageserverApiException, PageserverHttpClient
 from fixtures.pageserver.utils import (
-    assert_tenant_status,
+    assert_tenant_state,
     wait_for_upload_queue_empty,
     wait_until_tenant_active,
 )
@@ -38,10 +38,10 @@ def test_timeline_size(neon_simple_env: NeonEnv):
     client = env.pageserver.http_client()
     wait_for_timeline_size_init(client, tenant=env.initial_tenant, timeline=new_timeline_id)
 
-    pgmain = env.postgres.create_start("test_timeline_size")
+    endpoint_main = env.endpoints.create_start("test_timeline_size")
     log.info("postgres is running on 'test_timeline_size' branch")
 
-    with closing(pgmain.connect()) as conn:
+    with closing(endpoint_main.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute("CREATE TABLE foo (t text)")
             cur.execute(
@@ -74,10 +74,10 @@ def test_timeline_size_createdropdb(neon_simple_env: NeonEnv):
         env.initial_tenant, new_timeline_id, include_non_incremental_logical_size=True
     )
 
-    pgmain = env.postgres.create_start("test_timeline_size_createdropdb")
+    endpoint_main = env.endpoints.create_start("test_timeline_size_createdropdb")
     log.info("postgres is running on 'test_timeline_size_createdropdb' branch")
 
-    with closing(pgmain.connect()) as conn:
+    with closing(endpoint_main.connect()) as conn:
         with conn.cursor() as cur:
             res = client.timeline_detail(
                 env.initial_tenant, new_timeline_id, include_non_incremental_logical_size=True
@@ -89,7 +89,7 @@ def test_timeline_size_createdropdb(neon_simple_env: NeonEnv):
             ), "no writes should not change the incremental logical size"
 
             cur.execute("CREATE DATABASE foodb")
-            with closing(pgmain.connect(dbname="foodb")) as conn:
+            with closing(endpoint_main.connect(dbname="foodb")) as conn:
                 with conn.cursor() as cur2:
                     cur2.execute("CREATE TABLE foo (t text)")
                     cur2.execute(
@@ -118,7 +118,7 @@ def test_timeline_size_createdropdb(neon_simple_env: NeonEnv):
 
 
 # wait until received_lsn_lag is 0
-def wait_for_pageserver_catchup(pgmain: Postgres, polling_interval=1, timeout=60):
+def wait_for_pageserver_catchup(endpoint_main: Endpoint, polling_interval=1, timeout=60):
     started_at = time.time()
 
     received_lsn_lag = 1
@@ -129,7 +129,7 @@ def wait_for_pageserver_catchup(pgmain: Postgres, polling_interval=1, timeout=60
                 "timed out waiting for pageserver to reach pg_current_wal_flush_lsn()"
             )
 
-        res = pgmain.safe_psql(
+        res = endpoint_main.safe_psql(
             """
             SELECT
                 pg_size_pretty(pg_cluster_size()),
@@ -150,20 +150,20 @@ def test_timeline_size_quota(neon_env_builder: NeonEnvBuilder):
 
     wait_for_timeline_size_init(client, tenant=env.initial_tenant, timeline=new_timeline_id)
 
-    pgmain = env.postgres.create_start(
+    endpoint_main = env.endpoints.create_start(
         "test_timeline_size_quota",
         # Set small limit for the test
         config_lines=["neon.max_cluster_size=30MB"],
     )
     log.info("postgres is running on 'test_timeline_size_quota' branch")
 
-    with closing(pgmain.connect()) as conn:
+    with closing(endpoint_main.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute("CREATE EXTENSION neon")  # TODO move it to neon_fixtures?
 
             cur.execute("CREATE TABLE foo (t text)")
 
-            wait_for_pageserver_catchup(pgmain)
+            wait_for_pageserver_catchup(endpoint_main)
 
             # Insert many rows. This query must fail because of space limit
             try:
@@ -175,7 +175,7 @@ def test_timeline_size_quota(neon_env_builder: NeonEnvBuilder):
                 """
                 )
 
-                wait_for_pageserver_catchup(pgmain)
+                wait_for_pageserver_catchup(endpoint_main)
 
                 cur.execute(
                     """
@@ -195,7 +195,7 @@ def test_timeline_size_quota(neon_env_builder: NeonEnvBuilder):
             # drop table to free space
             cur.execute("DROP TABLE foo")
 
-            wait_for_pageserver_catchup(pgmain)
+            wait_for_pageserver_catchup(endpoint_main)
 
             # create it again and insert some rows. This query must succeed
             cur.execute("CREATE TABLE foo (t text)")
@@ -207,7 +207,7 @@ def test_timeline_size_quota(neon_env_builder: NeonEnvBuilder):
             """
             )
 
-            wait_for_pageserver_catchup(pgmain)
+            wait_for_pageserver_catchup(endpoint_main)
 
             cur.execute("SELECT * from pg_size_pretty(pg_cluster_size())")
             pg_cluster_size = cur.fetchone()
@@ -231,15 +231,15 @@ def test_timeline_initial_logical_size_calculation_cancellation(
     tenant_id, timeline_id = env.neon_cli.create_tenant()
 
     # load in some data
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
-    pg.safe_psql_many(
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
+    endpoint.safe_psql_many(
         [
             "CREATE TABLE foo (x INTEGER)",
             "INSERT INTO foo SELECT g FROM generate_series(1, 10000) g",
         ]
     )
-    wait_for_last_flush_lsn(env, pg, tenant_id, timeline_id)
-    pg.stop()
+    wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
+    endpoint.stop()
 
     # restart with failpoint inside initial size calculation task
     env.pageserver.stop()
@@ -311,9 +311,9 @@ def test_timeline_physical_size_init(
     env = neon_env_builder.init_start()
 
     new_timeline_id = env.neon_cli.create_branch("test_timeline_physical_size_init")
-    pg = env.postgres.create_start("test_timeline_physical_size_init")
+    endpoint = env.endpoints.create_start("test_timeline_physical_size_init")
 
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         [
             "CREATE TABLE foo (t text)",
             """INSERT INTO foo
@@ -322,7 +322,7 @@ def test_timeline_physical_size_init(
         ]
     )
 
-    wait_for_last_flush_lsn(env, pg, env.initial_tenant, new_timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, env.initial_tenant, new_timeline_id)
 
     # restart the pageserer to force calculating timeline's initial physical size
     env.pageserver.stop()
@@ -333,7 +333,7 @@ def test_timeline_physical_size_init(
     wait_until(
         number_of_iterations=5,
         interval=1,
-        func=lambda: assert_tenant_status(client, env.initial_tenant, "Active"),
+        func=lambda: assert_tenant_state(client, env.initial_tenant, "Active"),
     )
 
     assert_physical_size_invariants(
@@ -355,9 +355,9 @@ def test_timeline_physical_size_post_checkpoint(
 
     pageserver_http = env.pageserver.http_client()
     new_timeline_id = env.neon_cli.create_branch("test_timeline_physical_size_post_checkpoint")
-    pg = env.postgres.create_start("test_timeline_physical_size_post_checkpoint")
+    endpoint = env.endpoints.create_start("test_timeline_physical_size_post_checkpoint")
 
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         [
             "CREATE TABLE foo (t text)",
             """INSERT INTO foo
@@ -366,7 +366,7 @@ def test_timeline_physical_size_post_checkpoint(
         ]
     )
 
-    wait_for_last_flush_lsn(env, pg, env.initial_tenant, new_timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, env.initial_tenant, new_timeline_id)
     pageserver_http.timeline_checkpoint(env.initial_tenant, new_timeline_id)
 
     assert_physical_size_invariants(
@@ -394,7 +394,7 @@ def test_timeline_physical_size_post_compaction(
     pageserver_http = env.pageserver.http_client()
 
     new_timeline_id = env.neon_cli.create_branch("test_timeline_physical_size_post_compaction")
-    pg = env.postgres.create_start("test_timeline_physical_size_post_compaction")
+    endpoint = env.endpoints.create_start("test_timeline_physical_size_post_compaction")
 
     # We don't want autovacuum to run on the table, while we are calculating the
     # physical size, because that could cause a new layer to be created and a
@@ -402,7 +402,7 @@ def test_timeline_physical_size_post_compaction(
     # happens, because of some other background activity or autovacuum on other
     # tables, we could simply retry the size calculations. It's unlikely that
     # that would happen more than once.)
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         [
             "CREATE TABLE foo (t text) WITH (autovacuum_enabled = off)",
             """INSERT INTO foo
@@ -411,7 +411,7 @@ def test_timeline_physical_size_post_compaction(
         ]
     )
 
-    wait_for_last_flush_lsn(env, pg, env.initial_tenant, new_timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, env.initial_tenant, new_timeline_id)
 
     # shutdown safekeepers to prevent new data from coming in
     for sk in env.safekeepers:
@@ -446,10 +446,10 @@ def test_timeline_physical_size_post_gc(
     pageserver_http = env.pageserver.http_client()
 
     new_timeline_id = env.neon_cli.create_branch("test_timeline_physical_size_post_gc")
-    pg = env.postgres.create_start("test_timeline_physical_size_post_gc")
+    endpoint = env.endpoints.create_start("test_timeline_physical_size_post_gc")
 
     # Like in test_timeline_physical_size_post_compaction, disable autovacuum
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         [
             "CREATE TABLE foo (t text) WITH (autovacuum_enabled = off)",
             """INSERT INTO foo
@@ -458,10 +458,10 @@ def test_timeline_physical_size_post_gc(
         ]
     )
 
-    wait_for_last_flush_lsn(env, pg, env.initial_tenant, new_timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, env.initial_tenant, new_timeline_id)
     pageserver_http.timeline_checkpoint(env.initial_tenant, new_timeline_id)
 
-    pg.safe_psql(
+    endpoint.safe_psql(
         """
         INSERT INTO foo
             SELECT 'long string to consume some space' || g
@@ -469,7 +469,7 @@ def test_timeline_physical_size_post_gc(
     """
     )
 
-    wait_for_last_flush_lsn(env, pg, env.initial_tenant, new_timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, env.initial_tenant, new_timeline_id)
     pageserver_http.timeline_checkpoint(env.initial_tenant, new_timeline_id)
     pageserver_http.timeline_gc(env.initial_tenant, new_timeline_id, gc_horizon=None)
 
@@ -495,9 +495,9 @@ def test_timeline_size_metrics(
     pageserver_http = env.pageserver.http_client()
 
     new_timeline_id = env.neon_cli.create_branch("test_timeline_size_metrics")
-    pg = env.postgres.create_start("test_timeline_size_metrics")
+    endpoint = env.endpoints.create_start("test_timeline_size_metrics")
 
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         [
             "CREATE TABLE foo (t text)",
             """INSERT INTO foo
@@ -506,7 +506,7 @@ def test_timeline_size_metrics(
         ]
     )
 
-    wait_for_last_flush_lsn(env, pg, env.initial_tenant, new_timeline_id)
+    wait_for_last_flush_lsn(env, endpoint, env.initial_tenant, new_timeline_id)
     pageserver_http.timeline_checkpoint(env.initial_tenant, new_timeline_id)
 
     # get the metrics and parse the metric for the current timeline's physical size
@@ -558,7 +558,7 @@ def test_timeline_size_metrics(
     # The sum of the sizes of all databases, as seen by pg_database_size(), should also
     # be close. Again allow some slack, the logical size metric includes some things like
     # the SLRUs that are not included in pg_database_size().
-    dbsize_sum = pg.safe_psql("select sum(pg_database_size(oid)) from pg_database")[0][0]
+    dbsize_sum = endpoint.safe_psql("select sum(pg_database_size(oid)) from pg_database")[0][0]
     assert math.isclose(dbsize_sum, tl_logical_size_metric, abs_tol=2 * 1024 * 1024)
 
 
@@ -592,16 +592,16 @@ def test_tenant_physical_size(
         n_rows = random.randint(100, 1000)
 
         timeline = env.neon_cli.create_branch(f"test_tenant_physical_size_{i}", tenant_id=tenant)
-        pg = env.postgres.create_start(f"test_tenant_physical_size_{i}", tenant_id=tenant)
+        endpoint = env.endpoints.create_start(f"test_tenant_physical_size_{i}", tenant_id=tenant)
 
-        pg.safe_psql_many(
+        endpoint.safe_psql_many(
             [
                 "CREATE TABLE foo (t text)",
                 f"INSERT INTO foo SELECT 'long string to consume some space' || g FROM generate_series(1, {n_rows}) g",
             ]
         )
 
-        wait_for_last_flush_lsn(env, pg, tenant, timeline)
+        wait_for_last_flush_lsn(env, endpoint, tenant, timeline)
         pageserver_http.timeline_checkpoint(tenant, timeline)
 
         if remote_storage_kind is not None:
@@ -609,7 +609,7 @@ def test_tenant_physical_size(
 
         timeline_total_resident_physical_size += get_timeline_resident_physical_size(timeline)
 
-        pg.stop()
+        endpoint.stop()
 
     # ensure that tenant_status current_physical size reports sum of timeline current_physical_size
     tenant_current_physical_size = int(
diff --git a/test_runner/regress/test_truncate.py b/test_runner/regress/test_truncate.py
index cfe8a7f067..b1ddd93a40 100644
--- a/test_runner/regress/test_truncate.py
+++ b/test_runner/regress/test_truncate.py
@@ -27,8 +27,8 @@ def test_truncate(neon_env_builder: NeonEnvBuilder, zenbenchmark):
     )
 
     env.neon_cli.create_timeline("test_truncate", tenant_id=tenant)
-    pg = env.postgres.create_start("test_truncate", tenant_id=tenant)
-    cur = pg.connect().cursor()
+    endpoint = env.endpoints.create_start("test_truncate", tenant_id=tenant)
+    cur = endpoint.connect().cursor()
     cur.execute("create table t1(x integer)")
     cur.execute(f"insert into t1 values (generate_series(1,{n_records}))")
     cur.execute("vacuum t1")
diff --git a/test_runner/regress/test_twophase.py b/test_runner/regress/test_twophase.py
index f3b0f9ca06..305271c715 100644
--- a/test_runner/regress/test_twophase.py
+++ b/test_runner/regress/test_twophase.py
@@ -10,10 +10,12 @@ from fixtures.neon_fixtures import NeonEnv, fork_at_current_lsn
 def test_twophase(neon_simple_env: NeonEnv):
     env = neon_simple_env
     env.neon_cli.create_branch("test_twophase", "empty")
-    pg = env.postgres.create_start("test_twophase", config_lines=["max_prepared_transactions=5"])
+    endpoint = env.endpoints.create_start(
+        "test_twophase", config_lines=["max_prepared_transactions=5"]
+    )
     log.info("postgres is running on 'test_twophase' branch")
 
-    conn = pg.connect()
+    conn = endpoint.connect()
     cur = conn.cursor()
 
     cur.execute("CREATE TABLE foo (t text)")
@@ -42,7 +44,7 @@ def test_twophase(neon_simple_env: NeonEnv):
     # pg_twophase directory and fsynced
     cur.execute("CHECKPOINT")
 
-    twophase_files = os.listdir(pg.pg_twophase_dir_path())
+    twophase_files = os.listdir(endpoint.pg_twophase_dir_path())
     log.info(twophase_files)
     assert len(twophase_files) == 4
 
@@ -50,25 +52,25 @@ def test_twophase(neon_simple_env: NeonEnv):
     cur.execute("ROLLBACK PREPARED 'insert_four'")
     cur.execute("CHECKPOINT")
 
-    twophase_files = os.listdir(pg.pg_twophase_dir_path())
+    twophase_files = os.listdir(endpoint.pg_twophase_dir_path())
     log.info(twophase_files)
     assert len(twophase_files) == 2
 
     # Create a branch with the transaction in prepared state
-    fork_at_current_lsn(env, pg, "test_twophase_prepared", "test_twophase")
+    fork_at_current_lsn(env, endpoint, "test_twophase_prepared", "test_twophase")
 
     # Start compute on the new branch
-    pg2 = env.postgres.create_start(
+    endpoint2 = env.endpoints.create_start(
         "test_twophase_prepared",
         config_lines=["max_prepared_transactions=5"],
     )
 
     # Check that we restored only needed twophase files
-    twophase_files2 = os.listdir(pg2.pg_twophase_dir_path())
+    twophase_files2 = os.listdir(endpoint2.pg_twophase_dir_path())
     log.info(twophase_files2)
     assert twophase_files2.sort() == twophase_files.sort()
 
-    conn2 = pg2.connect()
+    conn2 = endpoint2.connect()
     cur2 = conn2.cursor()
 
     # On the new branch, commit one of the prepared transactions,
diff --git a/test_runner/regress/test_unlogged.py b/test_runner/regress/test_unlogged.py
index b6b20f1230..708bf0dfeb 100644
--- a/test_runner/regress/test_unlogged.py
+++ b/test_runner/regress/test_unlogged.py
@@ -9,9 +9,9 @@ from fixtures.neon_fixtures import NeonEnv, fork_at_current_lsn
 def test_unlogged(neon_simple_env: NeonEnv):
     env = neon_simple_env
     env.neon_cli.create_branch("test_unlogged", "empty")
-    pg = env.postgres.create_start("test_unlogged")
+    endpoint = env.endpoints.create_start("test_unlogged")
 
-    conn = pg.connect()
+    conn = endpoint.connect()
     cur = conn.cursor()
 
     cur.execute("CREATE UNLOGGED TABLE iut (id int);")
@@ -20,12 +20,10 @@ def test_unlogged(neon_simple_env: NeonEnv):
     cur.execute("INSERT INTO iut values (42);")
 
     # create another compute to fetch inital empty contents from pageserver
-    fork_at_current_lsn(env, pg, "test_unlogged_basebackup", "test_unlogged")
-    pg2 = env.postgres.create_start(
-        "test_unlogged_basebackup",
-    )
+    fork_at_current_lsn(env, endpoint, "test_unlogged_basebackup", "test_unlogged")
+    endpoint2 = env.endpoints.create_start("test_unlogged_basebackup")
 
-    conn2 = pg2.connect()
+    conn2 = endpoint2.connect()
     cur2 = conn2.cursor()
     # after restart table should be empty but valid
     cur2.execute("PREPARE iut_plan (int) AS INSERT INTO iut VALUES ($1)")
diff --git a/test_runner/regress/test_vm_bits.py b/test_runner/regress/test_vm_bits.py
index 16a870471b..d8034b31b0 100644
--- a/test_runner/regress/test_vm_bits.py
+++ b/test_runner/regress/test_vm_bits.py
@@ -10,10 +10,10 @@ def test_vm_bit_clear(neon_simple_env: NeonEnv):
     env = neon_simple_env
 
     env.neon_cli.create_branch("test_vm_bit_clear", "empty")
-    pg = env.postgres.create_start("test_vm_bit_clear")
+    endpoint = env.endpoints.create_start("test_vm_bit_clear")
 
     log.info("postgres is running on 'test_vm_bit_clear' branch")
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     # Install extension containing function needed for test
@@ -33,7 +33,7 @@ def test_vm_bit_clear(neon_simple_env: NeonEnv):
     cur.execute("UPDATE vmtest_update SET id = 5000 WHERE id = 1")
 
     # Branch at this point, to test that later
-    fork_at_current_lsn(env, pg, "test_vm_bit_clear_new", "test_vm_bit_clear")
+    fork_at_current_lsn(env, endpoint, "test_vm_bit_clear_new", "test_vm_bit_clear")
 
     # Clear the buffer cache, to force the VM page to be re-fetched from
     # the page server
@@ -63,10 +63,10 @@ def test_vm_bit_clear(neon_simple_env: NeonEnv):
     # a dirty VM page is evicted. If the VM bit was not correctly cleared by the
     # earlier WAL record, the full-page image hides the problem. Starting a new
     # server at the right point-in-time avoids that full-page image.
-    pg_new = env.postgres.create_start("test_vm_bit_clear_new")
+    endpoint_new = env.endpoints.create_start("test_vm_bit_clear_new")
 
     log.info("postgres is running on 'test_vm_bit_clear_new' branch")
-    pg_new_conn = pg_new.connect()
+    pg_new_conn = endpoint_new.connect()
     cur_new = pg_new_conn.cursor()
 
     cur_new.execute(
diff --git a/test_runner/regress/test_wal_acceptor.py b/test_runner/regress/test_wal_acceptor.py
index 306c492e8f..e8cfa4f318 100644
--- a/test_runner/regress/test_wal_acceptor.py
+++ b/test_runner/regress/test_wal_acceptor.py
@@ -16,6 +16,7 @@ from typing import Any, List, Optional
 import pytest
 from fixtures.log_helper import log
 from fixtures.neon_fixtures import (
+    Endpoint,
     NeonBroker,
     NeonEnv,
     NeonEnvBuilder,
@@ -23,7 +24,6 @@ from fixtures.neon_fixtures import (
     PgBin,
     PgProtocol,
     PortDistributor,
-    Postgres,
     RemoteStorageKind,
     RemoteStorageUsers,
     Safekeeper,
@@ -39,11 +39,11 @@ from fixtures.utils import get_dir_size, query_scalar, start_in_background
 def wait_lsn_force_checkpoint(
     tenant_id: TenantId,
     timeline_id: TimelineId,
-    pg: Postgres,
+    endpoint: Endpoint,
     ps: NeonPageserver,
     pageserver_conn_options={},
 ):
-    lsn = Lsn(pg.safe_psql("SELECT pg_current_wal_flush_lsn()")[0][0])
+    lsn = Lsn(endpoint.safe_psql("SELECT pg_current_wal_flush_lsn()")[0][0])
     log.info(f"pg_current_wal_flush_lsn is {lsn}, waiting for it on pageserver")
 
     auth_token = None
@@ -97,10 +97,10 @@ def test_many_timelines(neon_env_builder: NeonEnvBuilder):
     branch_names_to_timeline_ids = {}
 
     # start postgres on each timeline
-    pgs = []
+    endpoints = []
     for branch_name in branch_names:
         new_timeline_id = env.neon_cli.create_branch(branch_name)
-        pgs.append(env.postgres.create_start(branch_name))
+        endpoints.append(env.endpoints.create_start(branch_name))
         branch_names_to_timeline_ids[branch_name] = new_timeline_id
 
     tenant_id = env.initial_tenant
@@ -160,8 +160,8 @@ def test_many_timelines(neon_env_builder: NeonEnvBuilder):
     # Do everything in different loops to have actions on different timelines
     # interleaved.
     # create schema
-    for pg in pgs:
-        pg.safe_psql("CREATE TABLE t(key int primary key, value text)")
+    for endpoint in endpoints:
+        endpoint.safe_psql("CREATE TABLE t(key int primary key, value text)")
     init_m = collect_metrics("after CREATE TABLE")
 
     # Populate data for 2/3 timelines
@@ -197,16 +197,16 @@ def test_many_timelines(neon_env_builder: NeonEnvBuilder):
     metrics_checker = MetricsChecker()
     metrics_checker.start()
 
-    for pg in pgs[:-1]:
-        pg.safe_psql("INSERT INTO t SELECT generate_series(1,100000), 'payload'")
+    for endpoint in endpoints[:-1]:
+        endpoint.safe_psql("INSERT INTO t SELECT generate_series(1,100000), 'payload'")
 
     metrics_checker.stop()
 
     collect_metrics("after INSERT INTO")
 
     # Check data for 2/3 timelines
-    for pg in pgs[:-1]:
-        res = pg.safe_psql("SELECT sum(key) FROM t")
+    for endpoint in endpoints[:-1]:
+        res = endpoint.safe_psql("SELECT sum(key) FROM t")
         assert res[0] == (5000050000,)
 
     final_m = collect_metrics("after SELECT")
@@ -233,11 +233,11 @@ def test_restarts(neon_env_builder: NeonEnvBuilder):
     env = neon_env_builder.init_start()
 
     env.neon_cli.create_branch("test_safekeepers_restarts")
-    pg = env.postgres.create_start("test_safekeepers_restarts")
+    endpoint = env.endpoints.create_start("test_safekeepers_restarts")
 
     # we rely upon autocommit after each statement
     # as waiting for acceptors happens there
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     failed_node = None
@@ -268,22 +268,22 @@ def test_broker(neon_env_builder: NeonEnvBuilder):
         ".*init_tenant_mgr: marking .* as locally complete, while it doesnt exist in remote index.*"
     )
 
-    pg = env.postgres.create_start("test_broker")
-    pg.safe_psql("CREATE TABLE t(key int primary key, value text)")
+    endpoint = env.endpoints.create_start("test_broker")
+    endpoint.safe_psql("CREATE TABLE t(key int primary key, value text)")
 
     # learn neon timeline from compute
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     # wait until remote_consistent_lsn gets advanced on all safekeepers
     clients = [sk.http_client() for sk in env.safekeepers]
     stat_before = [cli.timeline_status(tenant_id, timeline_id) for cli in clients]
     log.info(f"statuses is {stat_before}")
 
-    pg.safe_psql("INSERT INTO t SELECT generate_series(1,100), 'payload'")
+    endpoint.safe_psql("INSERT INTO t SELECT generate_series(1,100), 'payload'")
 
     # force checkpoint in pageserver to advance remote_consistent_lsn
-    wait_lsn_force_checkpoint(tenant_id, timeline_id, pg, env.pageserver)
+    wait_lsn_force_checkpoint(tenant_id, timeline_id, endpoint, env.pageserver)
 
     # and wait till remote_consistent_lsn propagates to all safekeepers
     started_at = time.time()
@@ -299,7 +299,7 @@ def test_broker(neon_env_builder: NeonEnvBuilder):
             raise RuntimeError(
                 f"timed out waiting {elapsed:.0f}s for remote_consistent_lsn propagation: status before {stat_before}, status current {stat_after}"
             )
-        time.sleep(0.5)
+        time.sleep(1)
 
 
 # Test that old WAL consumed by peers and pageserver is removed from safekeepers.
@@ -317,26 +317,28 @@ def test_wal_removal(neon_env_builder: NeonEnvBuilder, auth_enabled: bool):
     )
 
     env.neon_cli.create_branch("test_safekeepers_wal_removal")
-    pg = env.postgres.create_start("test_safekeepers_wal_removal")
+    endpoint = env.endpoints.create_start("test_safekeepers_wal_removal")
 
     # Note: it is important to insert at least two segments, as currently
     # control file is synced roughly once in segment range and WAL is not
     # removed until all horizons are persisted.
-    pg.safe_psql_many(
+    endpoint.safe_psql_many(
         [
             "CREATE TABLE t(key int primary key, value text)",
             "INSERT INTO t SELECT generate_series(1,200000), 'payload'",
         ]
     )
 
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     # force checkpoint to advance remote_consistent_lsn
     pageserver_conn_options = {}
     if auth_enabled:
         pageserver_conn_options["password"] = env.auth_keys.generate_tenant_token(tenant_id)
-    wait_lsn_force_checkpoint(tenant_id, timeline_id, pg, env.pageserver, pageserver_conn_options)
+    wait_lsn_force_checkpoint(
+        tenant_id, timeline_id, endpoint, env.pageserver, pageserver_conn_options
+    )
 
     # We will wait for first segment removal. Make sure they exist for starter.
     first_segments = [
@@ -381,12 +383,15 @@ def test_wal_removal(neon_env_builder: NeonEnvBuilder, auth_enabled: bool):
     wait(
         lambda first_segments=first_segments: all(not os.path.exists(p) for p in first_segments),
         "first segment get removed",
+        wait_f=lambda http_cli=http_cli, tenant_id=tenant_id, timeline_id=timeline_id: log.info(
+            f"waiting for segments removal, sk info: {http_cli.timeline_status(tenant_id=tenant_id, timeline_id=timeline_id)}"
+        ),
     )
 
 
 # Wait for something, defined as f() returning True, raising error if this
-# doesn't happen without timeout seconds.
-def wait(f, desc, timeout=30):
+# doesn't happen without timeout seconds, and calling wait_f while waiting.
+def wait(f, desc, timeout=30, wait_f=None):
     started_at = time.time()
     while True:
         if f():
@@ -395,6 +400,8 @@ def wait(f, desc, timeout=30):
         if elapsed > timeout:
             raise RuntimeError(f"timed out waiting {elapsed:.0f}s for {desc}")
         time.sleep(0.5)
+        if wait_f is not None:
+            wait_f()
 
 
 def is_segment_offloaded(
@@ -436,13 +443,13 @@ def test_wal_backup(neon_env_builder: NeonEnvBuilder, remote_storage_kind: Remot
     env = neon_env_builder.init_start()
 
     env.neon_cli.create_branch("test_safekeepers_wal_backup")
-    pg = env.postgres.create_start("test_safekeepers_wal_backup")
+    endpoint = env.endpoints.create_start("test_safekeepers_wal_backup")
 
     # learn neon timeline from compute
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
     cur.execute("create table t(key int, value text)")
 
@@ -465,9 +472,9 @@ def test_wal_backup(neon_env_builder: NeonEnvBuilder, remote_storage_kind: Remot
     # put one of safekeepers down again
     env.safekeepers[0].stop()
     # restart postgres
-    pg.stop_and_destroy().create_start("test_safekeepers_wal_backup")
+    endpoint.stop_and_destroy().create_start("test_safekeepers_wal_backup")
     # and ensure offloading still works
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute("insert into t select generate_series(1,250000), 'payload'")
     seg_end = Lsn("0/5000000")
@@ -491,15 +498,15 @@ def test_s3_wal_replay(neon_env_builder: NeonEnvBuilder, remote_storage_kind: Re
     env = neon_env_builder.init_start()
     env.neon_cli.create_branch("test_s3_wal_replay")
 
-    pg = env.postgres.create_start("test_s3_wal_replay")
+    endpoint = env.endpoints.create_start("test_s3_wal_replay")
 
     # learn neon timeline from compute
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     expected_sum = 0
 
-    with closing(pg.connect()) as conn:
+    with closing(endpoint.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute("create table t(key int, value text)")
             cur.execute("insert into t values (1, 'payload')")
@@ -547,7 +554,7 @@ def test_s3_wal_replay(neon_env_builder: NeonEnvBuilder, remote_storage_kind: Re
         f"Pageserver last_record_lsn={pageserver_lsn}; flush_lsn={last_lsn}; lag before replay is {lag / 1024}kb"
     )
 
-    pg.stop_and_destroy()
+    endpoint.stop_and_destroy()
     ps_cli.timeline_delete(tenant_id, timeline_id)
 
     # Also delete and manually create timeline on safekeepers -- this tests
@@ -587,7 +594,7 @@ def test_s3_wal_replay(neon_env_builder: NeonEnvBuilder, remote_storage_kind: Re
             raise RuntimeError("Timed out waiting for WAL redo")
 
         tenant_status = ps_cli.tenant_status(tenant_id)
-        if tenant_status["state"] == "Loading":
+        if tenant_status["state"]["slug"] == "Loading":
             log.debug(f"Tenant {tenant_id} is still loading, retrying")
         else:
             pageserver_lsn = Lsn(
@@ -609,9 +616,9 @@ def test_s3_wal_replay(neon_env_builder: NeonEnvBuilder, remote_storage_kind: Re
     log.info(f"WAL redo took {elapsed} s")
 
     # verify data
-    pg.create_start("test_s3_wal_replay")
+    endpoint.create_start("test_s3_wal_replay")
 
-    assert pg.safe_psql("select sum(key) from t")[0][0] == expected_sum
+    assert endpoint.safe_psql("select sum(key) from t")[0][0] == expected_sum
 
 
 class ProposerPostgres(PgProtocol):
@@ -762,13 +769,13 @@ def test_timeline_status(neon_env_builder: NeonEnvBuilder, auth_enabled: bool):
     env = neon_env_builder.init_start()
 
     env.neon_cli.create_branch("test_timeline_status")
-    pg = env.postgres.create_start("test_timeline_status")
+    endpoint = env.endpoints.create_start("test_timeline_status")
 
     wa = env.safekeepers[0]
 
     # learn neon timeline from compute
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
     if not auth_enabled:
         wa_http_cli = wa.http_client()
@@ -806,11 +813,11 @@ def test_timeline_status(neon_env_builder: NeonEnvBuilder, auth_enabled: bool):
     assert debug_dump_0["timelines_count"] == 1
     assert debug_dump_0["timelines"][0]["timeline_id"] == str(timeline_id)
 
-    pg.safe_psql("create table t(i int)")
+    endpoint.safe_psql("create table t(i int)")
 
     # ensure epoch goes up after reboot
-    pg.stop().start()
-    pg.safe_psql("insert into t values(10)")
+    endpoint.stop().start()
+    endpoint.safe_psql("insert into t values(10)")
 
     tli_status = wa_http_cli.timeline_status(tenant_id, timeline_id)
     epoch_after_reboot = tli_status.acceptor_epoch
@@ -992,8 +999,8 @@ def test_replace_safekeeper(neon_env_builder: NeonEnvBuilder):
     def safekeepers_guc(env: NeonEnv, sk_names: List[int]) -> str:
         return ",".join([f"localhost:{sk.port.pg}" for sk in env.safekeepers if sk.id in sk_names])
 
-    def execute_payload(pg: Postgres):
-        with closing(pg.connect()) as conn:
+    def execute_payload(endpoint: Endpoint):
+        with closing(endpoint.connect()) as conn:
             with conn.cursor() as cur:
                 # we rely upon autocommit after each statement
                 # as waiting for acceptors happens there
@@ -1021,26 +1028,26 @@ def test_replace_safekeeper(neon_env_builder: NeonEnvBuilder):
     log.info("Use only first 3 safekeepers")
     env.safekeepers[3].stop()
     active_safekeepers = [1, 2, 3]
-    pg = env.postgres.create("test_replace_safekeeper")
-    pg.adjust_for_safekeepers(safekeepers_guc(env, active_safekeepers))
-    pg.start()
+    endpoint = env.endpoints.create("test_replace_safekeeper")
+    endpoint.adjust_for_safekeepers(safekeepers_guc(env, active_safekeepers))
+    endpoint.start()
 
     # learn neon timeline from compute
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
-    timeline_id = TimelineId(pg.safe_psql("show neon.timeline_id")[0][0])
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
+    timeline_id = TimelineId(endpoint.safe_psql("show neon.timeline_id")[0][0])
 
-    execute_payload(pg)
+    execute_payload(endpoint)
     show_statuses(env.safekeepers, tenant_id, timeline_id)
 
     log.info("Restart all safekeepers to flush everything")
     env.safekeepers[0].stop(immediate=True)
-    execute_payload(pg)
+    execute_payload(endpoint)
     env.safekeepers[0].start()
     env.safekeepers[1].stop(immediate=True)
-    execute_payload(pg)
+    execute_payload(endpoint)
     env.safekeepers[1].start()
     env.safekeepers[2].stop(immediate=True)
-    execute_payload(pg)
+    execute_payload(endpoint)
     env.safekeepers[2].start()
 
     env.safekeepers[0].stop(immediate=True)
@@ -1050,27 +1057,27 @@ def test_replace_safekeeper(neon_env_builder: NeonEnvBuilder):
     env.safekeepers[1].start()
     env.safekeepers[2].start()
 
-    execute_payload(pg)
+    execute_payload(endpoint)
     show_statuses(env.safekeepers, tenant_id, timeline_id)
 
     log.info("Stop sk1 (simulate failure) and use only quorum of sk2 and sk3")
     env.safekeepers[0].stop(immediate=True)
-    execute_payload(pg)
+    execute_payload(endpoint)
     show_statuses(env.safekeepers, tenant_id, timeline_id)
 
     log.info("Recreate postgres to replace failed sk1 with new sk4")
-    pg.stop_and_destroy().create("test_replace_safekeeper")
+    endpoint.stop_and_destroy().create("test_replace_safekeeper")
     active_safekeepers = [2, 3, 4]
     env.safekeepers[3].start()
-    pg.adjust_for_safekeepers(safekeepers_guc(env, active_safekeepers))
-    pg.start()
+    endpoint.adjust_for_safekeepers(safekeepers_guc(env, active_safekeepers))
+    endpoint.start()
 
-    execute_payload(pg)
+    execute_payload(endpoint)
     show_statuses(env.safekeepers, tenant_id, timeline_id)
 
     log.info("Stop sk2 to require quorum of sk3 and sk4 for normal work")
     env.safekeepers[1].stop(immediate=True)
-    execute_payload(pg)
+    execute_payload(endpoint)
     show_statuses(env.safekeepers, tenant_id, timeline_id)
 
 
@@ -1082,13 +1089,13 @@ def test_wal_deleted_after_broadcast(neon_env_builder: NeonEnvBuilder):
     last_lsn = Lsn(0)
 
     # returns pg_wal size in MB
-    def collect_stats(pg: Postgres, cur, enable_logs=True):
+    def collect_stats(endpoint: Endpoint, cur, enable_logs=True):
         nonlocal last_lsn
-        assert pg.pgdata_dir is not None
+        assert endpoint.pgdata_dir is not None
 
         log.info("executing INSERT to generate WAL")
         current_lsn = Lsn(query_scalar(cur, "select pg_current_wal_lsn()"))
-        pg_wal_size_mb = get_dir_size(os.path.join(pg.pgdata_dir, "pg_wal")) / 1024 / 1024
+        pg_wal_size_mb = get_dir_size(os.path.join(endpoint.pgdata_dir, "pg_wal")) / 1024 / 1024
         if enable_logs:
             lsn_delta_mb = (current_lsn - last_lsn) / 1024 / 1024
             log.info(f"LSN delta: {lsn_delta_mb} MB, current WAL size: {pg_wal_size_mb} MB")
@@ -1104,25 +1111,25 @@ def test_wal_deleted_after_broadcast(neon_env_builder: NeonEnvBuilder):
 
     env.neon_cli.create_branch("test_wal_deleted_after_broadcast")
     # Adjust checkpoint config to prevent keeping old WAL segments
-    pg = env.postgres.create_start(
+    endpoint = env.endpoints.create_start(
         "test_wal_deleted_after_broadcast",
         config_lines=["min_wal_size=32MB", "max_wal_size=32MB", "log_checkpoints=on"],
     )
 
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
     cur.execute("CREATE TABLE t(key int, value text)")
 
-    collect_stats(pg, cur)
+    collect_stats(endpoint, cur)
 
     # generate WAL to simulate normal workload
     for i in range(5):
         generate_wal(cur)
-        collect_stats(pg, cur)
+        collect_stats(endpoint, cur)
 
     log.info("executing checkpoint")
     cur.execute("CHECKPOINT")
-    wal_size_after_checkpoint = collect_stats(pg, cur)
+    wal_size_after_checkpoint = collect_stats(endpoint, cur)
 
     # there shouldn't be more than 2 WAL segments (but dir may have archive_status files)
     assert wal_size_after_checkpoint < 16 * 2.5
@@ -1151,13 +1158,13 @@ def test_delete_force(neon_env_builder: NeonEnvBuilder, auth_enabled: bool):
     tenant_id_other, timeline_id_other = env.neon_cli.create_tenant()
 
     # Populate branches
-    pg_1 = env.postgres.create_start("br1")
-    pg_2 = env.postgres.create_start("br2")
-    pg_3 = env.postgres.create_start("br3")
-    pg_4 = env.postgres.create_start("br4")
-    pg_other = env.postgres.create_start("main", tenant_id=tenant_id_other)
-    for pg in [pg_1, pg_2, pg_3, pg_4, pg_other]:
-        with closing(pg.connect()) as conn:
+    endpoint_1 = env.endpoints.create_start("br1")
+    endpoint_2 = env.endpoints.create_start("br2")
+    endpoint_3 = env.endpoints.create_start("br3")
+    endpoint_4 = env.endpoints.create_start("br4")
+    endpoint_other = env.endpoints.create_start("main", tenant_id=tenant_id_other)
+    for endpoint in [endpoint_1, endpoint_2, endpoint_3, endpoint_4, endpoint_other]:
+        with closing(endpoint.connect()) as conn:
             with conn.cursor() as cur:
                 cur.execute("CREATE TABLE t(key int primary key)")
     sk = env.safekeepers[0]
@@ -1178,14 +1185,14 @@ def test_delete_force(neon_env_builder: NeonEnvBuilder, auth_enabled: bool):
     assert (sk_data_dir / str(tenant_id_other) / str(timeline_id_other)).is_dir()
 
     # Stop branches which should be inactive and restart Safekeeper to drop its in-memory state.
-    pg_2.stop_and_destroy()
-    pg_4.stop_and_destroy()
+    endpoint_2.stop_and_destroy()
+    endpoint_4.stop_and_destroy()
     sk.stop()
     sk.start()
 
     # Ensure connections to Safekeeper are established
-    for pg in [pg_1, pg_3, pg_other]:
-        with closing(pg.connect()) as conn:
+    for endpoint in [endpoint_1, endpoint_3, endpoint_other]:
+        with closing(endpoint.connect()) as conn:
             with conn.cursor() as cur:
                 cur.execute("INSERT INTO t (key) VALUES (1)")
 
@@ -1244,6 +1251,6 @@ def test_delete_force(neon_env_builder: NeonEnvBuilder, auth_enabled: bool):
 
     # Ensure the other tenant still works
     sk_http_other.timeline_status(tenant_id_other, timeline_id_other)
-    with closing(pg_other.connect()) as conn:
+    with closing(endpoint_other.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute("INSERT INTO t (key) VALUES (123)")
diff --git a/test_runner/regress/test_wal_acceptor_async.py b/test_runner/regress/test_wal_acceptor_async.py
index f10a40690e..7debeed140 100644
--- a/test_runner/regress/test_wal_acceptor_async.py
+++ b/test_runner/regress/test_wal_acceptor_async.py
@@ -6,7 +6,7 @@ from typing import List, Optional
 
 import asyncpg
 from fixtures.log_helper import getLogger
-from fixtures.neon_fixtures import NeonEnv, NeonEnvBuilder, Postgres, Safekeeper
+from fixtures.neon_fixtures import Endpoint, NeonEnv, NeonEnvBuilder, Safekeeper
 from fixtures.types import Lsn, TenantId, TimelineId
 
 log = getLogger("root.safekeeper_async")
@@ -82,8 +82,10 @@ class WorkerStats(object):
         log.info("All workers made {} transactions".format(progress))
 
 
-async def run_random_worker(stats: WorkerStats, pg: Postgres, worker_id, n_accounts, max_transfer):
-    pg_conn = await pg.connect_async()
+async def run_random_worker(
+    stats: WorkerStats, endpoint: Endpoint, worker_id, n_accounts, max_transfer
+):
+    pg_conn = await endpoint.connect_async()
     log.debug("Started worker {}".format(worker_id))
 
     while stats.running:
@@ -141,7 +143,7 @@ async def wait_for_lsn(
 # consistent.
 async def run_restarts_under_load(
     env: NeonEnv,
-    pg: Postgres,
+    endpoint: Endpoint,
     acceptors: List[Safekeeper],
     n_workers=10,
     n_accounts=100,
@@ -154,7 +156,7 @@ async def run_restarts_under_load(
     # taking into account that this timeout is checked only at the beginning of every iteration.
     test_timeout_at = time.monotonic() + 5 * 60
 
-    pg_conn = await pg.connect_async()
+    pg_conn = await endpoint.connect_async()
     tenant_id = TenantId(await pg_conn.fetchval("show neon.tenant_id"))
     timeline_id = TimelineId(await pg_conn.fetchval("show neon.timeline_id"))
 
@@ -165,7 +167,7 @@ async def run_restarts_under_load(
     stats = WorkerStats(n_workers)
     workers = []
     for worker_id in range(n_workers):
-        worker = run_random_worker(stats, pg, worker_id, bank.n_accounts, max_transfer)
+        worker = run_random_worker(stats, endpoint, worker_id, bank.n_accounts, max_transfer)
         workers.append(asyncio.create_task(worker))
 
     for it in range(iterations):
@@ -212,11 +214,11 @@ def test_restarts_under_load(neon_env_builder: NeonEnvBuilder):
 
     env.neon_cli.create_branch("test_safekeepers_restarts_under_load")
     # Enable backpressure with 1MB maximal lag, because we don't want to block on `wait_for_lsn()` for too long
-    pg = env.postgres.create_start(
+    endpoint = env.endpoints.create_start(
         "test_safekeepers_restarts_under_load", config_lines=["max_replication_write_lag=1MB"]
     )
 
-    asyncio.run(run_restarts_under_load(env, pg, env.safekeepers))
+    asyncio.run(run_restarts_under_load(env, endpoint, env.safekeepers))
 
 
 # Restart acceptors one by one and test that everything is working as expected
@@ -228,7 +230,7 @@ def test_restarts_frequent_checkpoints(neon_env_builder: NeonEnvBuilder):
 
     env.neon_cli.create_branch("test_restarts_frequent_checkpoints")
     # Enable backpressure with 1MB maximal lag, because we don't want to block on `wait_for_lsn()` for too long
-    pg = env.postgres.create_start(
+    endpoint = env.endpoints.create_start(
         "test_restarts_frequent_checkpoints",
         config_lines=[
             "max_replication_write_lag=1MB",
@@ -240,11 +242,13 @@ def test_restarts_frequent_checkpoints(neon_env_builder: NeonEnvBuilder):
 
     # we try to simulate large (flush_lsn - truncate_lsn) lag, to test that WAL segments
     # are not removed before broadcasted to all safekeepers, with the help of replication slot
-    asyncio.run(run_restarts_under_load(env, pg, env.safekeepers, period_time=15, iterations=5))
+    asyncio.run(
+        run_restarts_under_load(env, endpoint, env.safekeepers, period_time=15, iterations=5)
+    )
 
 
-def postgres_create_start(env: NeonEnv, branch: str, pgdir_name: Optional[str]):
-    pg = Postgres(
+def endpoint_create_start(env: NeonEnv, branch: str, pgdir_name: Optional[str]):
+    endpoint = Endpoint(
         env,
         tenant_id=env.initial_tenant,
         port=env.port_distributor.get_port(),
@@ -253,19 +257,19 @@ def postgres_create_start(env: NeonEnv, branch: str, pgdir_name: Optional[str]):
         check_stop_result=False,
     )
 
-    # embed current time in node name
-    node_name = pgdir_name or f"pg_node_{time.time()}"
-    return pg.create_start(
-        branch_name=branch, node_name=node_name, config_lines=["log_statement=all"]
+    # embed current time in endpoint ID
+    endpoint_id = pgdir_name or f"ep-{time.time()}"
+    return endpoint.create_start(
+        branch_name=branch, endpoint_id=endpoint_id, config_lines=["log_statement=all"]
     )
 
 
 async def exec_compute_query(
     env: NeonEnv, branch: str, query: str, pgdir_name: Optional[str] = None
 ):
-    with postgres_create_start(env, branch=branch, pgdir_name=pgdir_name) as pg:
+    with endpoint_create_start(env, branch=branch, pgdir_name=pgdir_name) as endpoint:
         before_conn = time.time()
-        conn = await pg.connect_async()
+        conn = await endpoint.connect_async()
         res = await conn.fetch(query)
         await conn.close()
         after_conn = time.time()
@@ -436,8 +440,8 @@ async def check_unavailability(
     assert bg_query.done()
 
 
-async def run_unavailability(env: NeonEnv, pg: Postgres):
-    conn = await pg.connect_async()
+async def run_unavailability(env: NeonEnv, endpoint: Endpoint):
+    conn = await endpoint.connect_async()
 
     # check basic work with table
     await conn.execute("CREATE TABLE t(key int primary key, value text)")
@@ -462,9 +466,9 @@ def test_unavailability(neon_env_builder: NeonEnvBuilder):
     env = neon_env_builder.init_start()
 
     env.neon_cli.create_branch("test_safekeepers_unavailability")
-    pg = env.postgres.create_start("test_safekeepers_unavailability")
+    endpoint = env.endpoints.create_start("test_safekeepers_unavailability")
 
-    asyncio.run(run_unavailability(env, pg))
+    asyncio.run(run_unavailability(env, endpoint))
 
 
 @dataclass
@@ -493,8 +497,8 @@ async def xmas_garland(safekeepers: List[Safekeeper], data: RaceConditionTest):
         await asyncio.sleep(1)
 
 
-async def run_race_conditions(env: NeonEnv, pg: Postgres):
-    conn = await pg.connect_async()
+async def run_race_conditions(env: NeonEnv, endpoint: Endpoint):
+    conn = await endpoint.connect_async()
     await conn.execute("CREATE TABLE t(key int primary key, value text)")
 
     data = RaceConditionTest(0, False)
@@ -525,14 +529,14 @@ def test_race_conditions(neon_env_builder: NeonEnvBuilder):
     env = neon_env_builder.init_start()
 
     env.neon_cli.create_branch("test_safekeepers_race_conditions")
-    pg = env.postgres.create_start("test_safekeepers_race_conditions")
+    endpoint = env.endpoints.create_start("test_safekeepers_race_conditions")
 
-    asyncio.run(run_race_conditions(env, pg))
+    asyncio.run(run_race_conditions(env, endpoint))
 
 
 # Check that pageserver can select safekeeper with largest commit_lsn
 # and switch if LSN is not updated for some time (NoWalTimeout).
-async def run_wal_lagging(env: NeonEnv, pg: Postgres):
+async def run_wal_lagging(env: NeonEnv, endpoint: Endpoint):
     def safekeepers_guc(env: NeonEnv, active_sk: List[bool]) -> str:
         # use ports 10, 11 and 12 to simulate unavailable safekeepers
         return ",".join(
@@ -542,10 +546,10 @@ async def run_wal_lagging(env: NeonEnv, pg: Postgres):
             ]
         )
 
-    conn = await pg.connect_async()
+    conn = await endpoint.connect_async()
     await conn.execute("CREATE TABLE t(key int primary key, value text)")
     await conn.close()
-    pg.stop()
+    endpoint.stop()
 
     n_iterations = 20
     n_txes = 10000
@@ -561,11 +565,11 @@ async def run_wal_lagging(env: NeonEnv, pg: Postgres):
             it -= 1
             continue
 
-        pg.adjust_for_safekeepers(safekeepers_guc(env, active_sk))
+        endpoint.adjust_for_safekeepers(safekeepers_guc(env, active_sk))
         log.info(f"Iteration {it}: {active_sk}")
 
-        pg.start()
-        conn = await pg.connect_async()
+        endpoint.start()
+        conn = await endpoint.connect_async()
 
         for _ in range(n_txes):
             await conn.execute(f"INSERT INTO t values ({i}, 'payload')")
@@ -573,11 +577,11 @@ async def run_wal_lagging(env: NeonEnv, pg: Postgres):
             i += 1
 
         await conn.close()
-        pg.stop()
+        endpoint.stop()
 
-    pg.adjust_for_safekeepers(safekeepers_guc(env, [True] * len(env.safekeepers)))
-    pg.start()
-    conn = await pg.connect_async()
+    endpoint.adjust_for_safekeepers(safekeepers_guc(env, [True] * len(env.safekeepers)))
+    endpoint.start()
+    conn = await endpoint.connect_async()
 
     log.info(f"Executed {i-1} queries")
 
@@ -591,6 +595,6 @@ def test_wal_lagging(neon_env_builder: NeonEnvBuilder):
     env = neon_env_builder.init_start()
 
     env.neon_cli.create_branch("test_wal_lagging")
-    pg = env.postgres.create_start("test_wal_lagging")
+    endpoint = env.endpoints.create_start("test_wal_lagging")
 
-    asyncio.run(run_wal_lagging(env, pg))
+    asyncio.run(run_wal_lagging(env, endpoint))
diff --git a/test_runner/regress/test_wal_restore.py b/test_runner/regress/test_wal_restore.py
index 63d0b46f63..dd944af7eb 100644
--- a/test_runner/regress/test_wal_restore.py
+++ b/test_runner/regress/test_wal_restore.py
@@ -19,9 +19,9 @@ def test_wal_restore(
 ):
     env = neon_env_builder.init_start()
     env.neon_cli.create_branch("test_wal_restore")
-    pg = env.postgres.create_start("test_wal_restore")
-    pg.safe_psql("create table t as select generate_series(1,300000)")
-    tenant_id = TenantId(pg.safe_psql("show neon.tenant_id")[0][0])
+    endpoint = env.endpoints.create_start("test_wal_restore")
+    endpoint.safe_psql("create table t as select generate_series(1,300000)")
+    tenant_id = TenantId(endpoint.safe_psql("show neon.tenant_id")[0][0])
     env.neon_cli.pageserver_stop()
     port = port_distributor.get_port()
     data_dir = test_output_dir / "pgsql.restored"
diff --git a/test_runner/regress/test_walredo_not_left_behind_on_detach.py b/test_runner/regress/test_walredo_not_left_behind_on_detach.py
index d6302f8632..7d944bebb3 100644
--- a/test_runner/regress/test_walredo_not_left_behind_on_detach.py
+++ b/test_runner/regress/test_walredo_not_left_behind_on_detach.py
@@ -45,9 +45,9 @@ def test_walredo_not_left_behind_on_detach(neon_env_builder: NeonEnvBuilder):
     # assert tenant exists on disk
     assert (env.repo_dir / "tenants" / str(tenant_id)).exists()
 
-    pg = env.postgres.create_start("main", tenant_id=tenant_id)
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
 
-    pg_conn = pg.connect()
+    pg_conn = endpoint.connect()
     cur = pg_conn.cursor()
 
     # Create table, and insert some rows. Make it big enough that it doesn't fit in
diff --git a/test_runner/test_broken.py b/test_runner/test_broken.py
index 0281f4f48b..7e8aef5a5f 100644
--- a/test_runner/test_broken.py
+++ b/test_runner/test_broken.py
@@ -24,7 +24,7 @@ def test_broken(neon_simple_env: NeonEnv, pg_bin):
     env = neon_simple_env
 
     env.neon_cli.create_branch("test_broken", "empty")
-    env.postgres.create_start("test_broken")
+    env.endpoints.create_start("test_broken")
     log.info("postgres is running")
 
     log.info("THIS NEXT COMMAND WILL FAIL:")
diff --git a/trace/Cargo.toml b/trace/Cargo.toml
index 6ced992d4c..d6eed3f49c 100644
--- a/trace/Cargo.toml
+++ b/trace/Cargo.toml
@@ -4,8 +4,6 @@ version = "0.1.0"
 edition.workspace = true
 license.workspace = true
 
-# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
-
 [dependencies]
 clap.workspace = true
 anyhow.workspace = true
diff --git a/vendor/postgres-v14 b/vendor/postgres-v14
index 3e70693c91..a2daebc6b4 160000
--- a/vendor/postgres-v14
+++ b/vendor/postgres-v14
@@ -1 +1 @@
-Subproject commit 3e70693c9178878404d14a61c96b15b74eb02688
+Subproject commit a2daebc6b445dcbcca9c18e1711f47c1db7ffb04
diff --git a/vendor/postgres-v15 b/vendor/postgres-v15
index 4ad87b0f36..aee72b7be9 160000
--- a/vendor/postgres-v15
+++ b/vendor/postgres-v15
@@ -1 +1 @@
-Subproject commit 4ad87b0f364a2313600c1d9774ca33df00e606f4
+Subproject commit aee72b7be903e52d9bdc6449aa4c17fb852d8708
diff --git a/workspace_hack/Cargo.toml b/workspace_hack/Cargo.toml
index f885f4a94d..3d40f5dede 100644
--- a/workspace_hack/Cargo.toml
+++ b/workspace_hack/Cargo.toml
@@ -14,12 +14,11 @@ publish = false
 ### BEGIN HAKARI SECTION
 [dependencies]
 anyhow = { version = "1", features = ["backtrace"] }
-byteorder = { version = "1" }
 bytes = { version = "1", features = ["serde"] }
 chrono = { version = "0.4", default-features = false, features = ["clock", "serde"] }
 clap = { version = "4", features = ["derive", "string"] }
+clap_builder = { version = "4", default-features = false, features = ["color", "help", "std", "string", "suggestions", "usage"] }
 crossbeam-utils = { version = "0.8" }
-digest = { version = "0.10", features = ["mac", "std"] }
 either = { version = "1" }
 fail = { version = "0.5", default-features = false, features = ["failpoints"] }
 futures = { version = "0.3" }
@@ -29,7 +28,6 @@ futures-executor = { version = "0.3" }
 futures-sink = { version = "0.3" }
 futures-util = { version = "0.3", features = ["channel", "io", "sink"] }
 hashbrown = { version = "0.12", features = ["raw"] }
-indexmap = { version = "1", default-features = false, features = ["std"] }
 itertools = { version = "0.10" }
 libc = { version = "0.2", features = ["extra_traits"] }
 log = { version = "0.4", default-features = false, features = ["std"] }
@@ -52,7 +50,8 @@ socket2 = { version = "0.4", default-features = false, features = ["all"] }
 tokio = { version = "1", features = ["fs", "io-std", "io-util", "macros", "net", "process", "rt-multi-thread", "signal", "sync", "time"] }
 tokio-rustls = { version = "0.23" }
 tokio-util = { version = "0.7", features = ["codec", "io"] }
-tonic = { version = "0.8", features = ["tls-roots"] }
+toml_datetime = { version = "0.6", default-features = false, features = ["serde"] }
+toml_edit = { version = "0.19", features = ["serde"] }
 tower = { version = "0.4", features = ["balance", "buffer", "limit", "retry", "timeout", "util"] }
 tracing = { version = "0.1", features = ["log"] }
 tracing-core = { version = "0.1" }
@@ -64,7 +63,6 @@ anyhow = { version = "1", features = ["backtrace"] }
 bytes = { version = "1", features = ["serde"] }
 either = { version = "1" }
 hashbrown = { version = "0.12", features = ["raw"] }
-indexmap = { version = "1", default-features = false, features = ["std"] }
 itertools = { version = "0.10" }
 libc = { version = "0.2", features = ["extra_traits"] }
 log = { version = "0.4", default-features = false, features = ["std"] }
@@ -74,6 +72,7 @@ prost = { version = "0.11" }
 regex = { version = "1" }
 regex-syntax = { version = "0.6" }
 serde = { version = "1", features = ["alloc", "derive"] }
-syn = { version = "1", features = ["extra-traits", "full", "visit", "visit-mut"] }
+syn-dff4ba8e3ae991db = { package = "syn", version = "1", features = ["extra-traits", "full", "visit", "visit-mut"] }
+syn-f595c2ba2a3f28df = { package = "syn", version = "2", features = ["extra-traits", "full", "visit-mut"] }
 
 ### END HAKARI SECTION