remove dynamic limiter internal Box<dyn>

## Problem
Sometimes the forced extension upgrade test fails (on schedule) due to a
timeout.
## Summary of changes
The timeout is increased to 60 mins.

.github/actionlint.yml

@@ -8,6 +8,7 @@ self-hosted-runner:
     - small-arm64
     - us-east-2
 config-variables:
+  - AWS_ECR_REGION
   - AZURE_DEV_CLIENT_ID
   - AZURE_DEV_REGISTRY_NAME
   - AZURE_DEV_SUBSCRIPTION_ID
@@ -15,23 +16,25 @@ config-variables:
   - AZURE_PROD_REGISTRY_NAME
   - AZURE_PROD_SUBSCRIPTION_ID
   - AZURE_TENANT_ID
+  - BENCHMARK_INGEST_TARGET_PROJECTID
+  - BENCHMARK_LARGE_OLTP_PROJECTID
   - BENCHMARK_PROJECT_ID_PUB
   - BENCHMARK_PROJECT_ID_SUB
-  - REMOTE_STORAGE_AZURE_CONTAINER
-  - REMOTE_STORAGE_AZURE_REGION
-  - SLACK_UPCOMING_RELEASE_CHANNEL_ID
   - DEV_AWS_OIDC_ROLE_ARN
-  - BENCHMARK_INGEST_TARGET_PROJECTID
-  - PGREGRESS_PG16_PROJECT_ID
-  - PGREGRESS_PG17_PROJECT_ID
-  - SLACK_ON_CALL_QA_STAGING_STREAM
   - DEV_AWS_OIDC_ROLE_MANAGE_BENCHMARK_EC2_VMS_ARN
-  - SLACK_ON_CALL_STORAGE_STAGING_STREAM
-  - SLACK_CICD_CHANNEL_ID
-  - SLACK_STORAGE_CHANNEL_ID
+  - HETZNER_CACHE_BUCKET
+  - HETZNER_CACHE_ENDPOINT
+  - HETZNER_CACHE_REGION
   - NEON_DEV_AWS_ACCOUNT_ID
   - NEON_PROD_AWS_ACCOUNT_ID
-  - AWS_ECR_REGION
-  - BENCHMARK_LARGE_OLTP_PROJECTID
+  - PGREGRESS_PG16_PROJECT_ID
+  - PGREGRESS_PG17_PROJECT_ID
+  - REMOTE_STORAGE_AZURE_CONTAINER
+  - REMOTE_STORAGE_AZURE_REGION
+  - SLACK_CICD_CHANNEL_ID
   - SLACK_ON_CALL_DEVPROD_STREAM
+  - SLACK_ON_CALL_QA_STAGING_STREAM
+  - SLACK_ON_CALL_STORAGE_STAGING_STREAM
   - SLACK_RUST_CHANNEL_ID
+  - SLACK_STORAGE_CHANNEL_ID
+  - SLACK_UPCOMING_RELEASE_CHANNEL_ID

.github/workflows/_build-and-test-locally.yml

@@ -128,29 +128,49 @@ jobs:
 
       - name: Cache postgres v14 build
         id: cache_pg_14
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/v14
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ inputs.build-type }}-pg-${{ steps.pg_v14_rev.outputs.pg_rev }}-bookworm-${{ hashFiles('Makefile', 'build-tools.Dockerfile') }}
 
       - name: Cache postgres v15 build
         id: cache_pg_15
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/v15
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ inputs.build-type }}-pg-${{ steps.pg_v15_rev.outputs.pg_rev }}-bookworm-${{ hashFiles('Makefile', 'build-tools.Dockerfile') }}
 
       - name: Cache postgres v16 build
         id: cache_pg_16
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/v16
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ inputs.build-type }}-pg-${{ steps.pg_v16_rev.outputs.pg_rev }}-bookworm-${{ hashFiles('Makefile', 'build-tools.Dockerfile') }}
 
       - name: Cache postgres v17 build
         id: cache_pg_17
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/v17
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ inputs.build-type }}-pg-${{ steps.pg_v17_rev.outputs.pg_rev }}-bookworm-${{ hashFiles('Makefile', 'build-tools.Dockerfile') }}
 

.github/workflows/_check-codestyle-python.yml

@@ -37,8 +37,14 @@ jobs:
 
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
 
-      - uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+      - name: Cache poetry deps
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: ~/.cache/pypoetry/virtualenvs
           key: v2-${{ runner.os }}-${{ runner.arch }}-python-deps-bookworm-${{ hashFiles('poetry.lock') }}
 

.github/workflows/_check-codestyle-rust.yml

@@ -48,8 +48,13 @@ jobs:
           submodules: true
 
       - name: Cache cargo deps
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: |
             ~/.cargo/registry
             !~/.cargo/registry/src

.github/workflows/_meta.yml

@@ -5,6 +5,9 @@ on:
       github-event-name:
         type: string
         required: true
+      github-event-json:
+        type: string
+        required: true
     outputs:
       build-tag:
         description: "Tag for the current workflow run"
@@ -27,6 +30,9 @@ on:
       release-pr-run-id:
         description: "Only available if `run-kind in [storage-release, proxy-release, compute-release]`. Contains the run ID of the `Build and Test` workflow, assuming one with the current commit can be found."
         value: ${{ jobs.tags.outputs.release-pr-run-id }}
+      sha:
+        description: "github.event.pull_request.head.sha on release PRs, github.sha otherwise"
+        value: ${{ jobs.tags.outputs.sha }}
 
 permissions: {}
 
@@ -45,6 +51,7 @@ jobs:
       storage: ${{ steps.previous-releases.outputs.storage }}
       run-kind: ${{ steps.run-kind.outputs.run-kind }}
       release-pr-run-id: ${{ steps.release-pr-run-id.outputs.release-pr-run-id }}
+      sha: ${{ steps.sha.outputs.sha }}
     permissions:
       contents: read
     steps:
@@ -54,10 +61,6 @@ jobs:
         with:
           egress-policy: audit
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-
       - name: Get run kind
         id: run-kind
         env:
@@ -78,6 +81,23 @@ jobs:
         run: |
           echo "run-kind=$RUN_KIND" | tee -a $GITHUB_OUTPUT
 
+      - name: Get the right SHA
+        id: sha
+        env:
+          SHA: >
+            ${{
+              contains(fromJSON('["storage-rc-pr", "proxy-rc-pr", "compute-rc-pr"]'), steps.run-kind.outputs.run-kind)
+              && fromJSON(inputs.github-event-json).pull_request.head.sha
+              || github.sha
+            }}
+        run: |
+          echo "sha=$SHA" | tee -a $GITHUB_OUTPUT
+
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+          ref: ${{ steps.sha.outputs.sha }}
+
       - name: Get build tag
         id: build-tag
         env:
@@ -143,7 +163,7 @@ jobs:
         if: ${{ contains(fromJSON('["storage-release", "compute-release", "proxy-release"]'), steps.run-kind.outputs.run-kind) }}
         env:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          CURRENT_SHA: ${{ github.event.pull_request.head.sha || github.sha }}
+          CURRENT_SHA: ${{ github.sha }}
         run: |
           RELEASE_PR_RUN_ID=$(gh api "/repos/${GITHUB_REPOSITORY}/actions/runs?head_sha=$CURRENT_SHA" | jq '[.workflow_runs[] | select(.name == "Build and Test") | select(.head_branch | test("^rc/release(-(proxy|compute))?/[0-9]{4}-[0-9]{2}-[0-9]{2}$"; "s"))] | first | .id // ("Failed to find Build and Test run from  RC PR!" | halt_error(1))')
           echo "release-pr-run-id=$RELEASE_PR_RUN_ID" | tee -a $GITHUB_OUTPUT

.github/workflows/build-macos.yml

@@ -63,8 +63,13 @@ jobs:
 
       - name: Cache postgres ${{ matrix.postgres-version }} build
         id: cache_pg
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/${{ matrix.postgres-version }}
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-pg-${{ matrix.postgres-version }}-${{ steps.pg_rev.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
 
@@ -129,15 +134,25 @@ jobs:
 
       - name: Cache postgres v17 build
         id: cache_pg
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/v17
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-pg-v17-${{ steps.pg_rev.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
 
       - name: Cache walproposer-lib
         id: cache_walproposer_lib
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/build/walproposer-lib
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-walproposer_lib-v17-${{ steps.pg_rev.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
 
@@ -203,32 +218,57 @@ jobs:
 
       - name: Cache postgres v14 build
         id: cache_pg
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/v14
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-pg-v14-${{ steps.pg_rev_v14.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
       - name: Cache postgres v15 build
         id: cache_pg_v15
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/v15
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-pg-v15-${{ steps.pg_rev_v15.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
       - name: Cache postgres v16 build
         id: cache_pg_v16
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/v16
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-pg-v16-${{ steps.pg_rev_v16.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
       - name: Cache postgres v17 build
         id: cache_pg_v17
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/v17
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-pg-v17-${{ steps.pg_rev_v17.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
 
       - name: Cache cargo deps (only for v17)
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: |
             ~/.cargo/registry
             !~/.cargo/registry/src
@@ -238,8 +278,13 @@ jobs:
 
       - name: Cache walproposer-lib
         id: cache_walproposer_lib
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: pg_install/build/walproposer-lib
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-walproposer_lib-v17-${{ steps.pg_rev_v17.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
 

.github/workflows/build_and_test.yml

@@ -80,6 +80,7 @@ jobs:
     uses: ./.github/workflows/_meta.yml
     with:
       github-event-name: ${{ github.event_name }}
+      github-event-json: ${{ toJSON(github.event) }}
 
   build-build-tools-image:
     needs: [ check-permissions ]
@@ -248,8 +249,13 @@ jobs:
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
 
       - name: Cache poetry deps
-        uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
+        uses: tespkg/actions-cache@b7bf5fcc2f98a52ac6080eb0fd282c2f752074b1  # v1.8.0
         with:
+          endpoint: ${{ vars.HETZNER_CACHE_REGION }}.${{ vars.HETZNER_CACHE_ENDPOINT }}
+          bucket: ${{ vars.HETZNER_CACHE_BUCKET }}
+          accessKey: ${{ secrets.HETZNER_CACHE_ACCESS_KEY }}
+          secretKey: ${{ secrets.HETZNER_CACHE_SECRET_KEY }}
+          use-fallback: false
           path: ~/.cache/pypoetry/virtualenvs
           key: v2-${{ runner.os }}-${{ runner.arch }}-python-deps-bookworm-${{ hashFiles('poetry.lock') }}
 
@@ -540,6 +546,7 @@ jobs:
     uses: ./.github/workflows/trigger-e2e-tests.yml
     with:
       github-event-name: ${{ github.event_name }}
+      github-event-json: ${{ toJSON(github.event) }}
     secrets: inherit
 
   neon-image-arch:
@@ -563,6 +570,7 @@ jobs:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           submodules: true
+          ref: ${{ needs.meta.outputs.sha }}
 
       - uses: neondatabase/dev-actions/set-docker-config-dir@6094485bf440001c94a94a3f9e221e81ff6b6193
       - uses: docker/setup-buildx-action@b5ca514318bd6ebac0fb2aedd5d36ec1b5c232a2 # v3.10.0
@@ -672,6 +680,7 @@ jobs:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           submodules: true
+          ref: ${{ needs.meta.outputs.sha }}
 
       - uses: neondatabase/dev-actions/set-docker-config-dir@6094485bf440001c94a94a3f9e221e81ff6b6193
       - uses: docker/setup-buildx-action@b5ca514318bd6ebac0fb2aedd5d36ec1b5c232a2 # v3.10.0

.github/workflows/force-test-extensions-upgrade.yml

@@ -55,7 +55,7 @@ jobs:
           echo tag=${tag} >> ${GITHUB_OUTPUT}
 
       - name: Test extension upgrade
-        timeout-minutes: 20
+        timeout-minutes: 60
         env:
           NEW_COMPUTE_TAG: latest
           OLD_COMPUTE_TAG: ${{ steps.get-last-compute-release-tag.outputs.tag }}

.github/workflows/report-workflow-stats-batch.yml

@@ -23,7 +23,7 @@ jobs:
         egress-policy: audit
 
     - name: Export Workflow Run for the past 2 hours
-      uses: neondatabase/gh-workflow-stats-action@4c998b25ab5cc6588b52a610b749531f6a566b6b # v0.2.1
+      uses: neondatabase/gh-workflow-stats-action@701b1f202666d0b82e67b4d387e909af2b920127 # v0.2.2
       with:
         db_uri: ${{ secrets.GH_REPORT_STATS_DB_RW_CONNSTR }}
         db_table: "gh_workflow_stats_neon"
@@ -43,7 +43,7 @@ jobs:
         egress-policy: audit
 
     - name: Export Workflow Run for the past 48 hours
-      uses: neondatabase/gh-workflow-stats-action@4c998b25ab5cc6588b52a610b749531f6a566b6b # v0.2.1
+      uses: neondatabase/gh-workflow-stats-action@701b1f202666d0b82e67b4d387e909af2b920127 # v0.2.2
       with:
         db_uri: ${{ secrets.GH_REPORT_STATS_DB_RW_CONNSTR }}
         db_table: "gh_workflow_stats_neon"
@@ -63,7 +63,7 @@ jobs:
         egress-policy: audit
 
     - name: Export Workflow Run for the past 30 days
-      uses: neondatabase/gh-workflow-stats-action@4c998b25ab5cc6588b52a610b749531f6a566b6b # v0.2.1
+      uses: neondatabase/gh-workflow-stats-action@701b1f202666d0b82e67b4d387e909af2b920127 # v0.2.2
       with:
         db_uri: ${{ secrets.GH_REPORT_STATS_DB_RW_CONNSTR }}
         db_table: "gh_workflow_stats_neon"

.github/workflows/trigger-e2e-tests.yml

@@ -9,6 +9,9 @@ on:
       github-event-name:
         type: string
         required: true
+      github-event-json:
+        type: string
+        required: true
 
 defaults:
   run:
@@ -48,6 +51,7 @@ jobs:
     uses: ./.github/workflows/_meta.yml
     with:
       github-event-name: ${{ inputs.github-event-name || github.event_name }}
+      github-event-json: ${{ inputs.github-event-json || toJSON(github.event) }}
 
   trigger-e2e-tests:
     needs: [ meta ]

Cargo.lock

@@ -148,9 +148,9 @@ dependencies = [
 
 [[package]]
 name = "arc-swap"
-version = "1.6.0"
+version = "1.7.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bddcadddf5e9015d310179a59bb28c4d4b9920ad0f11e8e14dbadf654890c9a6"
+checksum = "69f7f8c3906b62b754cd5326047894316021dcfe5a194c8ea52bdd94934a3457"
 
 [[package]]
 name = "archery"
@@ -3861,11 +3861,10 @@ dependencies = [
 
 [[package]]
 name = "num-bigint"
-version = "0.4.3"
+version = "0.4.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f93ab6289c7b344a8a9f60f88d80aa20032336fe78da341afc91c8a2341fc75f"
+checksum = "a5e44f723f1133c9deac646763579fdb3ac745e418f2a7af9cd0c431da1f20b9"
 dependencies = [
- "autocfg",
  "num-integer",
  "num-traits",
 ]
@@ -3914,11 +3913,10 @@ dependencies = [
 
 [[package]]
 name = "num-integer"
-version = "0.1.45"
+version = "0.1.46"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "225d3389fb3509a24c93f5c29eb6bde2586b98d9f016636dff58d7c6f7569cd9"
+checksum = "7969661fd2958a5cb096e56c8e1ad0444ac2bbcd0061bd28660485a44879858f"
 dependencies = [
- "autocfg",
  "num-traits",
 ]
 
@@ -3947,9 +3945,9 @@ dependencies = [
 
 [[package]]
 name = "num-traits"
-version = "0.2.15"
+version = "0.2.19"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "578ede34cf02f8924ab9447f50c28075b4d3e5b269972345e7e0372b38c6cdcd"
+checksum = "071dfc062690e90b734c0b2273ce72ad0ffa95f0c74596bc250dcfd960262841"
 dependencies = [
  "autocfg",
  "libm",
@@ -5362,26 +5360,25 @@ dependencies = [
 
 [[package]]
 name = "redis"
-version = "0.25.2"
+version = "0.29.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "71d64e978fd98a0e6b105d066ba4889a7301fca65aeac850a877d8797343feeb"
+checksum = "b110459d6e323b7cda23980c46c77157601199c9da6241552b284cd565a7a133"
 dependencies = [
- "async-trait",
+ "arc-swap",
  "bytes",
  "combine",
  "futures-util",
  "itoa",
+ "num-bigint",
  "percent-encoding",
  "pin-project-lite",
- "rustls 0.22.4",
- "rustls-native-certs 0.7.0",
- "rustls-pemfile 2.1.1",
- "rustls-pki-types",
+ "rustls 0.23.18",
+ "rustls-native-certs 0.8.0",
  "ryu",
  "sha1_smol",
  "socket2",
  "tokio",
- "tokio-rustls 0.25.0",
+ "tokio-rustls 0.26.0",
  "tokio-util",
  "url",
 ]
@@ -7217,15 +7214,14 @@ dependencies = [
  "bytes",
  "fallible-iterator",
  "futures-util",
- "log",
  "parking_lot 0.12.1",
- "phf",
  "pin-project-lite",
  "postgres-protocol2",
  "postgres-types2",
  "serde",
  "tokio",
  "tokio-util",
+ "tracing",
 ]
 
 [[package]]

Cargo.toml

@@ -50,7 +50,7 @@ license = "Apache-2.0"
 [workspace.dependencies]
 ahash = "0.8"
 anyhow = { version = "1.0", features = ["backtrace"] }
-arc-swap = "1.6"
+arc-swap = "1.7"
 async-compression = { version = "0.4.0", features = ["tokio", "gzip", "zstd"] }
 atomic-take = "1.1.0"
 flate2 = "1.0.26"
@@ -130,7 +130,7 @@ nix = { version = "0.27", features = ["dir", "fs", "process", "socket", "signal"
 # on compute startup metrics (start_postgres_ms), >= 25% degradation.
 notify = "6.0.0"
 num_cpus = "1.15"
-num-traits = "0.2.15"
+num-traits = "0.2.19"
 once_cell = "1.13"
 opentelemetry = "0.27"
 opentelemetry_sdk = "0.27"
@@ -146,7 +146,7 @@ procfs = "0.16"
 prometheus = {version = "0.13", default-features=false, features = ["process"]} # removes protobuf dependency
 prost = "0.13"
 rand = "0.8"
-redis = { version = "0.25.2", features = ["tokio-rustls-comp", "keep-alive"] }
+redis = { version = "0.29.2", features = ["tokio-rustls-comp", "keep-alive"] }
 regex = "1.10.2"
 reqwest = { version = "0.12", default-features = false, features = ["rustls-tls"] }
 reqwest-tracing = { version = "0.5", features = ["opentelemetry_0_27"] }

compute/compute-node.Dockerfile

@@ -1916,26 +1916,30 @@ RUN apt update && \
       ;; \
     esac && \
     apt install --no-install-recommends -y \
+        ca-certificates \
         gdb \
-        liblz4-1 \
-        libreadline8 \
+        iproute2 \
         libboost-iostreams1.74.0 \
         libboost-regex1.74.0 \
         libboost-serialization1.74.0 \
         libboost-system1.74.0 \
-        libossp-uuid16 \
+        libcurl4 \
+        libevent-2.1-7 \
         libgeos-c1v5 \
+        liblz4-1 \
+        libossp-uuid16 \
         libprotobuf-c1 \
+        libreadline8 \
         libsfcgal1 \
         libxml2 \
         libxslt1.1 \
         libzstd1 \
-        libcurl4 \
-        libevent-2.1-7 \
         locales \
+        lsof \
         procps \
-        ca-certificates \
         rsyslog \
+        screen \
+        tcpdump \
         $VERSION_INSTALLS && \
     apt clean && 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

compute_tools/src/bin/compute_ctl.rs

@@ -45,7 +45,9 @@ use anyhow::{Context, Result};
 use clap::Parser;
 use compute_api::responses::ComputeCtlConfig;
 use compute_api::spec::ComputeSpec;
-use compute_tools::compute::{ComputeNode, ComputeNodeParams, forward_termination_signal};
+use compute_tools::compute::{
+    BUILD_TAG, ComputeNode, ComputeNodeParams, forward_termination_signal,
+};
 use compute_tools::extension_server::get_pg_version_string;
 use compute_tools::logger::*;
 use compute_tools::params::*;
@@ -57,10 +59,6 @@ use tracing::{error, info};
 use url::Url;
 use utils::failpoint_support;
 
-// this is an arbitrary build tag. Fine as a default / for testing purposes
-// in-case of not-set environment var
-const BUILD_TAG_DEFAULT: &str = "latest";
-
 // Compatibility hack: if the control plane specified any remote-ext-config
 // use the default value for extension storage proxy gateway.
 // Remove this once the control plane is updated to pass the gateway URL
@@ -147,7 +145,7 @@ fn main() -> Result<()> {
         .build()?;
     let _rt_guard = runtime.enter();
 
-    let build_tag = runtime.block_on(init())?;
+    runtime.block_on(init())?;
 
     // enable core dumping for all child processes
     setrlimit(Resource::CORE, rlimit::INFINITY, rlimit::INFINITY)?;
@@ -174,8 +172,6 @@ fn main() -> Result<()> {
             cgroup: cli.cgroup,
             #[cfg(target_os = "linux")]
             vm_monitor_addr: cli.vm_monitor_addr,
-            build_tag,
-
             live_config_allowed: cli_spec.live_config_allowed,
         },
         cli_spec.spec,
@@ -189,7 +185,7 @@ fn main() -> Result<()> {
     deinit_and_exit(exit_code);
 }
 
-async fn init() -> Result<String> {
+async fn init() -> Result<()> {
     init_tracing_and_logging(DEFAULT_LOG_LEVEL).await?;
 
     let mut signals = Signals::new([SIGINT, SIGTERM, SIGQUIT])?;
@@ -199,12 +195,9 @@ async fn init() -> Result<String> {
         }
     });
 
-    let build_tag = option_env!("BUILD_TAG")
-        .unwrap_or(BUILD_TAG_DEFAULT)
-        .to_string();
-    info!("build_tag: {build_tag}");
+    info!("compute build_tag: {}", &BUILD_TAG.to_string());
 
-    Ok(build_tag)
+    Ok(())
 }
 
 fn try_spec_from_cli(cli: &Cli) -> Result<CliSpecParams> {

compute_tools/src/compute.rs

@@ -1,6 +1,6 @@
 use std::collections::HashMap;
 use std::os::unix::fs::{PermissionsExt, symlink};
-use std::path::{Path, PathBuf};
+use std::path::Path;
 use std::process::{Command, Stdio};
 use std::str::FromStr;
 use std::sync::atomic::{AtomicU32, Ordering};
@@ -20,6 +20,7 @@ use futures::future::join_all;
 use futures::stream::FuturesUnordered;
 use nix::sys::signal::{Signal, kill};
 use nix::unistd::Pid;
+use once_cell::sync::Lazy;
 use postgres;
 use postgres::NoTls;
 use postgres::error::SqlState;
@@ -35,6 +36,7 @@ use crate::disk_quota::set_disk_quota;
 use crate::installed_extensions::get_installed_extensions;
 use crate::logger::startup_context_from_env;
 use crate::lsn_lease::launch_lsn_lease_bg_task_for_static;
+use crate::metrics::COMPUTE_CTL_UP;
 use crate::monitor::launch_monitor;
 use crate::pg_helpers::*;
 use crate::rsyslog::{
@@ -49,6 +51,17 @@ use crate::{config, extension_server, local_proxy};
 
 pub static SYNC_SAFEKEEPERS_PID: AtomicU32 = AtomicU32::new(0);
 pub static PG_PID: AtomicU32 = AtomicU32::new(0);
+// This is an arbitrary build tag. Fine as a default / for testing purposes
+// in-case of not-set environment var
+const BUILD_TAG_DEFAULT: &str = "latest";
+/// Build tag/version of the compute node binaries/image. It's tricky and ugly
+/// to pass it everywhere as a part of `ComputeNodeParams`, so we use a
+/// global static variable.
+pub static BUILD_TAG: Lazy<String> = Lazy::new(|| {
+    option_env!("BUILD_TAG")
+        .unwrap_or(BUILD_TAG_DEFAULT)
+        .to_string()
+});
 
 /// Static configuration params that don't change after startup. These mostly
 /// come from the CLI args, or are derived from them.
@@ -72,7 +85,6 @@ pub struct ComputeNodeParams {
     pub pgdata: String,
     pub pgbin: String,
     pub pgversion: String,
-    pub build_tag: String,
 
     /// The port that the compute's external HTTP server listens on
     pub external_http_port: u16,
@@ -153,11 +165,6 @@ pub struct ComputeState {
     pub startup_span: Option<tracing::span::Span>,
 
     pub metrics: ComputeMetrics,
-
-    /// current audit log level
-    /// to know if it is already configured, or we need to set up audit
-    /// when compute receives a new spec
-    pub audit_log_level: ComputeAudit,
 }
 
 impl ComputeState {
@@ -170,7 +177,6 @@ impl ComputeState {
             pspec: None,
             startup_span: None,
             metrics: ComputeMetrics::default(),
-            audit_log_level: ComputeAudit::default(),
         }
     }
 
@@ -179,6 +185,11 @@ impl ComputeState {
         info!("Changing compute status from {} to {}", prev, status);
         self.status = status;
         state_changed.notify_all();
+
+        COMPUTE_CTL_UP.reset();
+        COMPUTE_CTL_UP
+            .with_label_values(&[&BUILD_TAG, format!("{}", status).as_str()])
+            .set(1);
     }
 
     pub fn set_failed_status(&mut self, err: anyhow::Error, state_changed: &Condvar) {
@@ -358,13 +369,19 @@ impl ComputeNode {
         }
         .launch(&this);
 
-        // The internal HTTP server could be launched later, but there isn't much
-        // sense in waiting.
+        // The internal HTTP server is needed for a further activation by control plane
+        // if compute was started for a pool, so we have to start server before hanging
+        // waiting for a spec.
         crate::http::server::Server::Internal {
             port: this.params.internal_http_port,
         }
         .launch(&this);
 
+        // HTTP server is running, so we can officially declare compute_ctl as 'up'
+        COMPUTE_CTL_UP
+            .with_label_values(&[&BUILD_TAG, ComputeStatus::Empty.to_string().as_str()])
+            .set(1);
+
         // If we got a spec from the CLI already, use that. Otherwise wait for the
         // control plane to pass it to us with a /configure HTTP request
         let pspec = if let Some(cli_spec) = cli_spec {
@@ -626,10 +643,16 @@ impl ComputeNode {
             });
         }
 
-        // If extended compute audit is enabled configure and start rsyslog
-        if pspec.spec.audit_log_level == ComputeAudit::Hipaa {
-            let log_directory_path = self.get_audit_log_dir().to_string_lossy().to_string();
-            configure_audit_rsyslog(&log_directory_path, pspec.spec.audit_log_level.as_str())?;
+        // Configure and start rsyslog for HIPAA if necessary
+        if let ComputeAudit::Hipaa = pspec.spec.audit_log_level {
+            let remote_endpoint = std::env::var("AUDIT_LOGGING_ENDPOINT").unwrap_or("".to_string());
+            if remote_endpoint.is_empty() {
+                anyhow::bail!("AUDIT_LOGGING_ENDPOINT is empty");
+            }
+
+            let log_directory_path = Path::new(&self.params.pgdata).join("log");
+            let log_directory_path = log_directory_path.to_string_lossy().to_string();
+            configure_audit_rsyslog(log_directory_path.clone(), "hipaa", &remote_endpoint)?;
 
             // Launch a background task to clean up the audit logs
             launch_pgaudit_gc(log_directory_path);
@@ -684,11 +707,6 @@ impl ComputeNode {
             });
         }
 
-        // after all the configuration is done
-        // preserve the information about the current audit log level
-        // so that we don't relaunch rsyslog on every spec change
-        self.set_audit_log_level(pspec.spec.audit_log_level);
-
         // All done!
         let startup_end_time = Utc::now();
         let metrics = {
@@ -843,19 +861,6 @@ impl ComputeNode {
         self.state.lock().unwrap().status
     }
 
-    pub fn set_audit_log_level(&self, audit_log_level: ComputeAudit) {
-        let mut state = self.state.lock().unwrap();
-        state.audit_log_level = audit_log_level;
-    }
-
-    pub fn get_audit_log_level(&self) -> ComputeAudit {
-        self.state.lock().unwrap().audit_log_level
-    }
-
-    pub fn get_audit_log_dir(&self) -> PathBuf {
-        Path::new(&self.params.pgdata).join("log")
-    }
-
     pub fn get_timeline_id(&self) -> Option<TimelineId> {
         self.state
             .lock()
@@ -1566,29 +1571,6 @@ impl ComputeNode {
             });
         }
 
-        // If extended compute audit is enabled configure and start rsyslog
-        // We check that the audit_log_level changed compared to the previous spec and skip this step if not.
-        let audit_log_level = self.get_audit_log_level();
-
-        if spec.audit_log_level == ComputeAudit::Hipaa && audit_log_level != spec.audit_log_level {
-            info!(
-                "Configuring audit logging because audit_log_level changed from {:?} to {:?}",
-                audit_log_level, spec.audit_log_level
-            );
-
-            let log_directory_path = self.get_audit_log_dir().to_string_lossy().to_string();
-            configure_audit_rsyslog(&log_directory_path, spec.audit_log_level.as_str())?;
-
-            // Launch a background task to clean up the audit logs
-            // If rsyslog was already configured, we don't need to start this process again.
-            match audit_log_level {
-                ComputeAudit::Disabled | ComputeAudit::Log => {
-                    launch_pgaudit_gc(log_directory_path);
-                }
-                _ => {}
-            }
-        }
-
         // Write new config
         let pgdata_path = Path::new(&self.params.pgdata);
         config::write_postgres_conf(
@@ -1598,14 +1580,7 @@ impl ComputeNode {
             &self.compute_ctl_config.tls,
         )?;
 
-        // Override the skip_catalog_updates flag
-        // if we need to install new extensions
-        //
-        // Check that audit_log_level changed compared to the previous spec and skip this step if not.
-        // All operations are idempotent, so this is just a performance optimization.
-        let force_catalog_updates = audit_log_level != spec.audit_log_level;
-
-        if !spec.skip_pg_catalog_updates || force_catalog_updates {
+        if !spec.skip_pg_catalog_updates {
             let max_concurrent_connections = spec.reconfigure_concurrency;
             // Temporarily reset max_cluster_size in config
             // to avoid the possibility of hitting the limit, while we are reconfiguring:
@@ -1630,11 +1605,6 @@ impl ComputeNode {
 
         self.pg_reload_conf()?;
 
-        // after all the configuration is done
-        // preserve the information about the current audit log level
-        // so that we don't relaunch rsyslog on every spec change
-        self.set_audit_log_level(spec.audit_log_level);
-
         let unknown_op = "unknown".to_string();
         let op_id = spec.operation_uuid.as_ref().unwrap_or(&unknown_op);
         info!(
@@ -2085,12 +2055,8 @@ LIMIT 100",
 
         let mut download_tasks = Vec::new();
         for library in &libs_vec {
-            let (ext_name, ext_path) = remote_extensions.get_ext(
-                library,
-                true,
-                &self.params.build_tag,
-                &self.params.pgversion,
-            )?;
+            let (ext_name, ext_path) =
+                remote_extensions.get_ext(library, true, &BUILD_TAG, &self.params.pgversion)?;
             download_tasks.push(self.download_extension(ext_name, ext_path));
         }
         let results = join_all(download_tasks).await;

compute_tools/src/config.rs

@@ -200,9 +200,8 @@ pub fn write_postgres_conf(
             )?;
             // This log level is very verbose
             // but this is necessary for HIPAA compliance.
-            // Exclude 'misc' category, because it doesn't contain anything relevant.
+            // Exclude 'misc' category, because it doesn't contain anythig relevant.
             writeln!(file, "pgaudit.log='all, -misc'")?;
-            // Log parameters for all queries
             writeln!(file, "pgaudit.log_parameter=on")?;
             // Disable logging of catalog queries
             // The catalog doesn't contain sensitive data, so we don't need to audit it.

compute_tools/src/http/routes/extension_server.rs

@@ -5,7 +5,7 @@ use axum::response::{IntoResponse, Response};
 use http::StatusCode;
 use serde::Deserialize;
 
-use crate::compute::ComputeNode;
+use crate::compute::{BUILD_TAG, ComputeNode};
 use crate::http::JsonResponse;
 use crate::http::extract::{Path, Query};
 
@@ -47,7 +47,7 @@ pub(in crate::http) async fn download_extension(
         remote_extensions.get_ext(
             &filename,
             ext_server_params.is_library,
-            &compute.params.build_tag,
+            &BUILD_TAG,
             &compute.params.pgversion,
         )
     };

compute_tools/src/metrics.rs

@@ -1,7 +1,8 @@
 use metrics::core::{AtomicF64, Collector, GenericGauge};
 use metrics::proto::MetricFamily;
 use metrics::{
-    IntCounterVec, UIntGaugeVec, register_gauge, register_int_counter_vec, register_uint_gauge_vec,
+    IntCounterVec, IntGaugeVec, UIntGaugeVec, register_gauge, register_int_counter_vec,
+    register_int_gauge_vec, register_uint_gauge_vec,
 };
 use once_cell::sync::Lazy;
 
@@ -70,8 +71,19 @@ pub(crate) static AUDIT_LOG_DIR_SIZE: Lazy<GenericGauge<AtomicF64>> = Lazy::new(
     .expect("failed to define a metric")
 });
 
+// Report that `compute_ctl` is up and what's the current compute status.
+pub(crate) static COMPUTE_CTL_UP: Lazy<IntGaugeVec> = Lazy::new(|| {
+    register_int_gauge_vec!(
+        "compute_ctl_up",
+        "Whether compute_ctl is running",
+        &["build_tag", "status"]
+    )
+    .expect("failed to define a metric")
+});
+
 pub fn collect() -> Vec<MetricFamily> {
-    let mut metrics = INSTALLED_EXTENSIONS.collect();
+    let mut metrics = COMPUTE_CTL_UP.collect();
+    metrics.extend(INSTALLED_EXTENSIONS.collect());
     metrics.extend(CPLANE_REQUESTS_TOTAL.collect());
     metrics.extend(REMOTE_EXT_REQUESTS_TOTAL.collect());
     metrics.extend(DB_MIGRATION_FAILED.collect());

compute_tools/src/rsyslog.rs

@@ -9,7 +9,6 @@ use anyhow::{Context, Result, anyhow};
 use tracing::{error, info, instrument, warn};
 
 const POSTGRES_LOGS_CONF_PATH: &str = "/etc/rsyslog.d/postgres_logs.conf";
-const AUDIT_LOGS_CONF_PATH: &str = "/etc/rsyslog.d/compute_audit_rsyslog.conf";
 
 fn get_rsyslog_pid() -> Option<String> {
     let output = Command::new("pgrep")
@@ -49,43 +48,32 @@ fn restart_rsyslog() -> Result<()> {
     Ok(())
 }
 
-pub fn configure_audit_rsyslog(log_directory: &str, audit_log_level: &str) -> Result<()> {
-    let remote_endpoint = std::env::var("AUDIT_LOGGING_ENDPOINT")?;
-    if remote_endpoint.is_empty() {
-        return Err(anyhow!("AUDIT_LOGGING_ENDPOINT is not set"));
-    }
-
-    let old_config_content = match std::fs::read_to_string(AUDIT_LOGS_CONF_PATH) {
-        Ok(c) => c,
-        Err(err) if err.kind() == ErrorKind::NotFound => String::new(),
-        Err(err) => return Err(err.into()),
-    };
-
+pub fn configure_audit_rsyslog(
+    log_directory: String,
+    tag: &str,
+    remote_endpoint: &str,
+) -> Result<()> {
     let config_content: String = format!(
         include_str!("config_template/compute_audit_rsyslog_template.conf"),
         log_directory = log_directory,
-        tag = audit_log_level,
+        tag = tag,
         remote_endpoint = remote_endpoint
     );
 
-    if old_config_content == config_content {
-        info!("rsyslog configuration is up-to-date");
-        return Ok(());
-    }
-
     info!("rsyslog config_content: {}", config_content);
 
+    let rsyslog_conf_path = "/etc/rsyslog.d/compute_audit_rsyslog.conf";
     let mut file = OpenOptions::new()
         .create(true)
         .write(true)
         .truncate(true)
-        .open(AUDIT_LOGS_CONF_PATH)?;
+        .open(rsyslog_conf_path)?;
 
     file.write_all(config_content.as_bytes())?;
 
     info!(
         "rsyslog configuration file {} added successfully. Starting rsyslogd",
-        AUDIT_LOGS_CONF_PATH
+        rsyslog_conf_path
     );
 
     // start the service, using the configuration

control_plane/src/local_env.rs

@@ -184,6 +184,8 @@ pub struct NeonStorageControllerConf {
     pub timelines_onto_safekeepers: bool,
 
     pub use_https_safekeeper_api: bool,
+
+    pub use_local_compute_notifications: bool,
 }
 
 impl NeonStorageControllerConf {
@@ -213,6 +215,7 @@ impl Default for NeonStorageControllerConf {
             use_https_pageserver_api: false,
             timelines_onto_safekeepers: false,
             use_https_safekeeper_api: false,
+            use_local_compute_notifications: true,
         }
     }
 }

control_plane/src/pageserver.rs

@@ -51,11 +51,19 @@ impl PageServerNode {
             parse_host_port(&conf.listen_pg_addr).expect("Unable to parse listen_pg_addr");
         let port = port.unwrap_or(5432);
 
-        let ssl_ca_cert = env.ssl_ca_cert_path().map(|ssl_ca_file| {
+        let ssl_ca_certs = env.ssl_ca_cert_path().map(|ssl_ca_file| {
             let buf = std::fs::read(ssl_ca_file).expect("SSL root CA file should exist");
-            Certificate::from_pem(&buf).expect("CA certificate should be valid")
+            Certificate::from_pem_bundle(&buf).expect("SSL CA file should be valid")
         });
 
+        let mut http_client = reqwest::Client::builder();
+        for ssl_ca_cert in ssl_ca_certs.unwrap_or_default() {
+            http_client = http_client.add_root_certificate(ssl_ca_cert);
+        }
+        let http_client = http_client
+            .build()
+            .expect("Client constructs with no errors");
+
         let endpoint = if env.storage_controller.use_https_pageserver_api {
             format!(
                 "https://{}",
@@ -72,6 +80,7 @@ impl PageServerNode {
             conf: conf.clone(),
             env: env.clone(),
             http_client: mgmt_api::Client::new(
+                http_client,
                 endpoint,
                 {
                     match conf.http_auth_type {
@@ -83,9 +92,7 @@ impl PageServerNode {
                     }
                 }
                 .as_deref(),
-                ssl_ca_cert,
-            )
-            .expect("Client constructs with no errors"),
+            ),
         }
     }
 
@@ -421,11 +428,6 @@ impl PageServerNode {
                 .map(|x| x.parse::<usize>())
                 .transpose()
                 .context("Failed to parse 'l0_flush_delay_threshold' as an integer")?,
-            l0_flush_wait_upload: settings
-                .remove("l0_flush_wait_upload")
-                .map(|x| x.parse::<bool>())
-                .transpose()
-                .context("Failed to parse 'l0_flush_wait_upload' as a boolean")?,
             l0_flush_stall_threshold: settings
                 .remove("l0_flush_stall_threshold")
                 .map(|x| x.parse::<usize>())

control_plane/src/storage_controller.rs

@@ -17,7 +17,7 @@ use pageserver_api::models::{TenantConfigRequest, TimelineCreateRequest, Timelin
 use pageserver_api::shard::TenantShardId;
 use pageserver_client::mgmt_api::ResponseErrorMessageExt;
 use postgres_backend::AuthType;
-use reqwest::Method;
+use reqwest::{Certificate, Method};
 use serde::de::DeserializeOwned;
 use serde::{Deserialize, Serialize};
 use tokio::process::Command;
@@ -143,11 +143,14 @@ impl StorageController {
             }
         };
 
-        let mut http_client = reqwest::Client::builder();
-        if let Some(ssl_ca_file) = env.ssl_ca_cert_path() {
+        let ssl_ca_certs = env.ssl_ca_cert_path().map(|ssl_ca_file| {
             let buf = std::fs::read(ssl_ca_file).expect("SSL CA file should exist");
-            let cert = reqwest::Certificate::from_pem(&buf).expect("SSL CA file should be valid");
-            http_client = http_client.add_root_certificate(cert);
+            Certificate::from_pem_bundle(&buf).expect("SSL CA file should be valid")
+        });
+
+        let mut http_client = reqwest::Client::builder();
+        for ssl_ca_cert in ssl_ca_certs.unwrap_or_default() {
+            http_client = http_client.add_root_certificate(ssl_ca_cert);
         }
         let http_client = http_client
             .build()
@@ -552,6 +555,10 @@ impl StorageController {
             args.push("--use-https-safekeeper-api".to_string());
         }
 
+        if self.config.use_local_compute_notifications {
+            args.push("--use-local-compute-notifications".to_string());
+        }
+
         if let Some(ssl_ca_file) = self.env.ssl_ca_cert_path() {
             args.push(format!("--ssl-ca-file={}", ssl_ca_file.to_str().unwrap()));
         }

control_plane/storcon_cli/src/main.rs

@@ -20,7 +20,7 @@ use pageserver_api::models::{
 };
 use pageserver_api::shard::{ShardStripeSize, TenantShardId};
 use pageserver_client::mgmt_api::{self};
-use reqwest::{Method, StatusCode, Url};
+use reqwest::{Certificate, Method, StatusCode, Url};
 use storage_controller_client::control_api::Client;
 use utils::id::{NodeId, TenantId, TimelineId};
 
@@ -274,7 +274,7 @@ struct Cli {
     jwt: Option<String>,
 
     #[arg(long)]
-    /// Trusted root CA certificate to use in https APIs.
+    /// Trusted root CA certificates to use in https APIs.
     ssl_ca_file: Option<PathBuf>,
 
     #[command(subcommand)]
@@ -387,17 +387,23 @@ async fn main() -> anyhow::Result<()> {
 
     let storcon_client = Client::new(cli.api.clone(), cli.jwt.clone());
 
-    let ssl_ca_cert = match &cli.ssl_ca_file {
+    let ssl_ca_certs = match &cli.ssl_ca_file {
         Some(ssl_ca_file) => {
             let buf = tokio::fs::read(ssl_ca_file).await?;
-            Some(reqwest::Certificate::from_pem(&buf)?)
+            Certificate::from_pem_bundle(&buf)?
         }
-        None => None,
+        None => Vec::new(),
     };
 
+    let mut http_client = reqwest::Client::builder();
+    for ssl_ca_cert in ssl_ca_certs {
+        http_client = http_client.add_root_certificate(ssl_ca_cert);
+    }
+    let http_client = http_client.build()?;
+
     let mut trimmed = cli.api.to_string();
     trimmed.pop();
-    let vps_client = mgmt_api::Client::new(trimmed, cli.jwt.as_deref(), ssl_ca_cert)?;
+    let vps_client = mgmt_api::Client::new(http_client, trimmed, cli.jwt.as_deref());
 
     match cli.command {
         Command::NodeRegister {

docs/rfcs/2025-02-14-storage-controller.md

@@ -0,0 +1,196 @@
+
+## Summary
+
+This is a retrospective RFC to document the design of the `storage-controller` service.
+
+This service manages the physical mapping of Tenants and Timelines to Pageservers and Safekeepers.  It
+acts as the API for "storage" as an abstract concept: enabling other parts of the system to reason
+about things like creating/deleting tenants and timelines without having to understand exactly which
+pageserver and safekeeper to communicate, or any subtle rules about how to orchestrate these things.
+
+The storage controller was implemented in the first half of 2024 as an essential part
+of storage sharding, especially [shard splitting](032-shard-splitting.md).
+
+It initially managed only pageservers, but has extended in 2025 to also manage safekeepers.  In
+some places you may seen unqualified references to 'nodes' -- those are pageservers.
+
+## Design Choices
+
+### Durability
+
+We rely on an external postgres for all durable state.  No local storage is used.
+
+We avoid any unnecessary I/O to durable storage.  For example:
+- most tracking of in-flight changes to the system is done in-memory rather than recording progress/steps in a database
+- When migrating tenant shards between pageservers we only touch the database to increment generation numbers,
+  we do not persist the total state of a tenant shard.
+
+Being frugal with database I/O has two benefits:
+- It avoids the database becoming a practical scaling bottleneck (we expect in-memory scale issues to be hit
+  before we hit e.g. transactions-per-second issues)
+- It reduces cost when using a cloud database service to run the controller's postgres database.
+
+The trade-off is that there is a "bootstrapping" problem: a controller can't be deployed in isolation, one
+must first have some existing database system.  In practice, we expect that Neon is deployed in one of the
+following ways:
+- into a cloud which has a postgres service that can be used to run the controller
+- into a mature on-prem environment that has existing facilities for running databases
+- into a test/dev environment where a simple one-node vanilla postgres installation is sufficient
+
+### Consensus
+
+The controller does _not_ implement any strong consensus mechanism of its own.  Instead:
+- Where strong consistency is required (for example, for pageserver generation numbers), this
+  responsibility is delegated to a transaction in our postgres database.
+- Highly available deploys are done using a simple in-database record of what controller instances
+  are available, distinguished by timestamps, rather than having controllers directly negotiate a leader.
+
+Avoiding strong consensus among controller processes is a cost saving (we avoid running three controllers
+all the time), and simplifies implementation (we do not have to phrase all configuration changes as e.g raft
+transactions).
+
+The trade-off is that under some circumstances a controller with partial network isolation can cause availability
+issues in the cluster, by making changes to pageserver state that might disagree with what the "true" active
+controller is trying to do.  The impact of this is bounded by our `controllers` database table, that enables
+a rogue node to eventually realise that it is not the leader and step down.  If a rogue node can't reach
+the database, then it implicitly stops making progress.  A rogue controller cannot durably damage the system
+because pageserver data and safekeeper configs are protected by generation numbers that are only updated
+via postgres transactions (i.e. no controller "trusts itself" to independently make decisions about generations).
+
+### Scale
+
+We design for high but not unlimited scale.  The memory footprint of each tenant shard is small (~8kB), so
+it is realistic to scale up to a million attached shards on a server with modest resources.  Tenants in
+a detached state (i.e. not active on pageservers) do not need to be managed by storage controller, and can
+be relegated from memory to the database.
+
+Typically, a tenant shard is updated about once a week, when we do a deploy.  During deploys, we relocate
+a few thousand tenants from each pageserver while it is restarted, so it is extremely rare for the controller
+to have to do O(N) work (on all shards at once).
+
+There are places where we do O(N) work:
+- On normal startup, when loading from the database into memory
+- On unclean startup (with no handover of observed state from a previous controller), where we will
+  scan all shards on all pageservers.
+
+It is important that these locations are written efficiently.  At high scale we should still expect runtimes
+of the order tens of seconds to complete a storage controller start.
+
+When the practical scale limit of a single storage controller is reached, just deploy another one with its
+own pageservers & safekeepers: each controller+its storage servers should be thought of as a logical cluster
+or "cell" of storage.
+
+# High Level Design
+
+The storage controller is an in-memory system (i.e. state for all attached
+tenants is held in memory _as well as_ being represented in durable postgres storage).
+
+## Infrastructure
+
+The storage controller is an async rust binary using tokio.
+
+The storage controller is built around the `Service` type.  This implements
+all the entry points for the outside world's interaction with the controller (HTTP handlers are mostly thin wrappers of service functions),
+and holds most in-memory state (e.g. the list of tenant shards).
+
+The state is held in a `ServiceInner` wrapped in a RwLock.  This monolithic
+lock is used to simplify reasoning about code that mutates state: each function that takes a write lock may be thought of as a serializable transaction on the in-memory state.  This lock is clearly a bottleneck, but
+nevertheless is scalable to managing millions of tenants.
+
+Persistent state is held in a postgres database, and we use the `diesel` crate to provide database client functionality.  All database access is wrapped in the `Persistence` type -- this makes it easy to understand which
+code is touching the database.  The database is only used when necessary, i.e. for state that cannot be recovered another way.  For example, we do not store the secondary pageserver locations of tenant shards in the database, rather we learn these at startup from running pageservers, and/or make scheduling decisions to fill in the gaps.  This adds some complexity, but massively reduces the load on the database, and enables running the storage controller with a very cheap postgres instance.
+
+## Pageserver tenant scheduling & reconciliation
+
+### Intent & observed state
+
+Each tenant shard is represented by type `TenantShard`, which has an 'intent' and 'observed' state.  Setting the
+intent state is called _scheduling_, and doing remote I/O to make observed
+state match intent state is called _reconciliation_.
+
+The `Scheduler` type is responsible for making choices about the intent
+state, such as choosing a pageserver for a new tenant shard, or assigning
+a replacement pageserver when the original one fails.
+
+The observed state is updated after tenant reconciliation (see below), and
+has the concept of a `None` state for a pageserver, indicating unknown state.  This is used to ensure that we can safely clean up after we start
+but do not finish a remote call to a pageserver, or if a pageserver restarts and we are uncertain of its state.
+
+### Tenant Reconciliation
+
+The `Reconciler` type is responsible for updating pageservers to achieve
+the intent state.  It is instantiated when `Service` determines that a shard requires reconciliation, and owned by a background tokio task that
+runs it to completion.  Reconciler does not have access to the `Service` state: it is populated with a snapshot of relevant information when constructed, and submits is results to a channel that `Service` consumes
+to update the tenant shard's observed state.
+
+The Reconciler does have access to the database, but only uses it for
+a single purpose: updating shards' generation numbers immediately before
+attaching them to a pageserver.
+
+Operations that change a tenant's scheduling will spawn a reconciler if
+necessary, and there is also a background loop which checks every shard
+for the need to reconcile -- this background loop ensures eventual progress
+if some earlier reconciliations failed for some reason.
+
+The reconciler has a general purpose code path which will attach/detach from pageservers as necessary, and a special case path for live migrations.  The live migration case is more common in practice, and is taken whenever the current observed state indicates that we have a healthy attached location to migrate from.  This implements live migration as described in the earlier [live migration RFC](028-pageserver-migration.md).
+
+### Scheduling optimisation
+
+During the periodic background reconciliation loop, the controller also
+performance _scheduling optimization_.  This is the process of looking for
+shards that are in sub-optimal locations, and moving them.
+
+Typically, this means:
+- Shards attached outside their preferred AZ (e.g. after a node failure), to migrate them back to their preferred AZ
+- Shards attached on the same pageserver as some other shards in the same
+  tenant, to migrate them elsewhere (e.g. after a shard split)
+
+Scheduling optimisation is a multi-step process to ensure graceful cutovers, e.g. by creating new secondary location, waiting for it to
+warm up, then cutting over.  This is not done as an explicit queue
+of operations, but rather by iteratively calling the optimisation
+function, which will recognise each intervening state as something
+that can generate the next optimisation.
+
+### Pageserver heartbeats and failure
+
+The `Heartbeater` type is responsible for detecting when a pageserver
+becomes unavailable.  This is fed back into `Service` for action: when
+a pageserver is marked unavailable, tenant shards on that pageserver are
+rescheduled and Reconcilers are spawned to cut them over to their new location.
+
+## Pageserver timeline CRUD operations
+
+By CRUD operations, we mean creating and deleting timelines.  The authoritative storage for which timelines exist on the pageserver
+is in S3, and is governed by the pageserver's system of generation
+numbers.  Because a shard can be attached to multiple pageservers
+concurrently, we need to handle this when doing timeline CRUD operations:
+- A timeline operation is only persistent if _after_ the ack from a pageserver, that pageserver's generation is still the latest.
+- For deletions in particular, they are only persistent if _all_ attached
+  locations have acked the deletion operation, since if only the latest one
+  has acked then the timeline could still return from the dead if some old-generation attachment writes an index for it.
+
+## Zero-downtime controller deployments
+
+When two storage controllers run at the same time, they coordinate via
+the database to establish one leader, and the other controller may proxy
+requests to this leader
+
+See  [Storage controller restarts RFC](037-storage-controller-restarts.md).
+
+Note that this is not a strong consensus mechanism: the controller must also survive split-brain situations.  This is respected by code that
+e.g. increments version numbers, which uses database transactions that
+check the expected value before modifying it.  A split-brain situation can
+impact availability (e.g. if two controllers are fighting over where to
+attach a shard), but it should never impact durability and data integrity.
+
+## Graceful drain & fill of pageservers during deploys
+
+The storage controller has functionality for draining + filling pageservers
+while deploying new pageserver binaries, so that clients are not actively
+using a pageserver while it restarts.
+
+See [Graceful restarts RFC](033-storage-controller-drain-and-fill.md)
+
+## Safekeeper timeline scheduling
+
+This is currently under development, see  [Safekeeper dynamic membership change RFC](035-safekeeper-dynamic-membership-change.md).

libs/compute_api/src/spec.rs

@@ -160,6 +160,12 @@ pub struct ComputeSpec {
     pub drop_subscriptions_before_start: bool,
 
     /// Log level for audit logging:
+    ///
+    /// Disabled - no audit logging. This is the default.
+    /// log - log masked statements to the postgres log using pgaudit extension
+    /// hipaa - log unmasked statements to the file using pgaudit and pgauditlogtofile extension
+    ///
+    /// Extensions should be present in shared_preload_libraries
     #[serde(default)]
     pub audit_log_level: ComputeAudit,
 }
@@ -282,27 +288,16 @@ impl ComputeMode {
 }
 
 /// Log level for audit logging
-#[derive(Clone, Copy, Debug, Default, Eq, PartialEq, Deserialize, Serialize)]
+/// Disabled, log, hipaa
+/// Default is Disabled
+#[derive(Clone, Debug, Default, Eq, PartialEq, Deserialize, Serialize)]
 pub enum ComputeAudit {
     #[default]
-    /// no audit logging. This is the default.
     Disabled,
-    /// write masked audit log statements to the postgres log using pgaudit extension
     Log,
-    /// log unmasked statements to the file using pgaudit and pgauditlogtofile extensions
     Hipaa,
 }
 
-impl ComputeAudit {
-    pub fn as_str(&self) -> &str {
-        match self {
-            ComputeAudit::Disabled => "disabled",
-            ComputeAudit::Log => "log",
-            ComputeAudit::Hipaa => "hipaa",
-        }
-    }
-}
-
 #[derive(Clone, Debug, Default, Deserialize, Serialize, PartialEq, Eq)]
 pub struct Cluster {
     pub cluster_id: Option<String>,

libs/pageserver_api/src/config.rs

@@ -285,12 +285,6 @@ pub struct TenantConfigToml {
     /// Level0 delta layer threshold at which to stall layer flushes. Must be >compaction_threshold
     /// to avoid deadlock. 0 to disable. Disabled by default.
     pub l0_flush_stall_threshold: Option<usize>,
-    /// If true, Level0 delta layer flushes will wait for S3 upload before flushing the next
-    /// layer. This is a temporary backpressure mechanism which should be removed once
-    /// l0_flush_{delay,stall}_threshold is fully enabled.
-    ///
-    /// TODO: this is no longer enabled, remove it when the config option is no longer set.
-    pub l0_flush_wait_upload: bool,
     // Determines how much history is retained, to allow
     // branching and read replicas at an older point in time.
     // The unit is #of bytes of WAL.
@@ -579,8 +573,6 @@ pub mod tenant_conf_defaults {
     pub const DEFAULT_COMPACTION_ALGORITHM: crate::models::CompactionAlgorithm =
         crate::models::CompactionAlgorithm::Legacy;
 
-    pub const DEFAULT_L0_FLUSH_WAIT_UPLOAD: bool = false;
-
     pub const DEFAULT_GC_HORIZON: u64 = 64 * 1024 * 1024;
 
     // Large DEFAULT_GC_PERIOD is fine as long as PITR_INTERVAL is larger.
@@ -627,7 +619,6 @@ impl Default for TenantConfigToml {
             compaction_l0_semaphore: DEFAULT_COMPACTION_L0_SEMAPHORE,
             l0_flush_delay_threshold: None,
             l0_flush_stall_threshold: None,
-            l0_flush_wait_upload: DEFAULT_L0_FLUSH_WAIT_UPLOAD,
             gc_horizon: DEFAULT_GC_HORIZON,
             gc_period: humantime::parse_duration(DEFAULT_GC_PERIOD)
                 .expect("cannot parse default gc period"),

libs/pageserver_api/src/models.rs

@@ -523,8 +523,6 @@ pub struct TenantConfigPatch {
     #[serde(skip_serializing_if = "FieldPatch::is_noop")]
     pub l0_flush_stall_threshold: FieldPatch<usize>,
     #[serde(skip_serializing_if = "FieldPatch::is_noop")]
-    pub l0_flush_wait_upload: FieldPatch<bool>,
-    #[serde(skip_serializing_if = "FieldPatch::is_noop")]
     pub gc_horizon: FieldPatch<u64>,
     #[serde(skip_serializing_if = "FieldPatch::is_noop")]
     pub gc_period: FieldPatch<String>,
@@ -614,9 +612,6 @@ pub struct TenantConfig {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub l0_flush_stall_threshold: Option<usize>,
 
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub l0_flush_wait_upload: Option<bool>,
-
     #[serde(skip_serializing_if = "Option::is_none")]
     pub gc_horizon: Option<u64>,
 
@@ -712,7 +707,6 @@ impl TenantConfig {
             mut compaction_l0_semaphore,
             mut l0_flush_delay_threshold,
             mut l0_flush_stall_threshold,
-            mut l0_flush_wait_upload,
             mut gc_horizon,
             mut gc_period,
             mut image_creation_threshold,
@@ -765,7 +759,6 @@ impl TenantConfig {
         patch
             .l0_flush_stall_threshold
             .apply(&mut l0_flush_stall_threshold);
-        patch.l0_flush_wait_upload.apply(&mut l0_flush_wait_upload);
         patch.gc_horizon.apply(&mut gc_horizon);
         patch
             .gc_period
@@ -844,7 +837,6 @@ impl TenantConfig {
             compaction_l0_semaphore,
             l0_flush_delay_threshold,
             l0_flush_stall_threshold,
-            l0_flush_wait_upload,
             gc_horizon,
             gc_period,
             image_creation_threshold,
@@ -911,9 +903,6 @@ impl TenantConfig {
             l0_flush_stall_threshold: self
                 .l0_flush_stall_threshold
                 .or(global_conf.l0_flush_stall_threshold),
-            l0_flush_wait_upload: self
-                .l0_flush_wait_upload
-                .unwrap_or(global_conf.l0_flush_wait_upload),
             gc_horizon: self.gc_horizon.unwrap_or(global_conf.gc_horizon),
             gc_period: self.gc_period.unwrap_or(global_conf.gc_period),
             image_creation_threshold: self

libs/proxy/tokio-postgres2/Cargo.toml

@@ -8,10 +8,9 @@ license = "MIT/Apache-2.0"
 bytes.workspace = true
 fallible-iterator.workspace = true
 futures-util = { workspace = true, features = ["sink"] }
-log = "0.4"
+tracing.workspace = true
 parking_lot.workspace = true
 pin-project-lite.workspace = true
-phf = "0.11"
 postgres-protocol2 = { path = "../postgres-protocol2" }
 postgres-types2 = { path = "../postgres-types2" }
 tokio = { workspace = true, features = ["io-util", "time", "net"] }

libs/proxy/tokio-postgres2/src/connection.rs

@@ -6,13 +6,13 @@ use std::task::{Context, Poll};
 use bytes::BytesMut;
 use fallible_iterator::FallibleIterator;
 use futures_util::{Sink, Stream, ready};
-use log::{info, trace};
 use postgres_protocol2::message::backend::Message;
 use postgres_protocol2::message::frontend;
 use tokio::io::{AsyncRead, AsyncWrite};
 use tokio::sync::mpsc;
 use tokio_util::codec::Framed;
 use tokio_util::sync::PollSender;
+use tracing::{info, trace};
 
 use crate::codec::{BackendMessage, BackendMessages, FrontendMessage, PostgresCodec};
 use crate::error::DbError;

libs/proxy/tokio-postgres2/src/prepare.rs

@@ -5,9 +5,9 @@ use std::sync::Arc;
 use bytes::Bytes;
 use fallible_iterator::FallibleIterator;
 use futures_util::{TryStreamExt, pin_mut};
-use log::debug;
 use postgres_protocol2::message::backend::Message;
 use postgres_protocol2::message::frontend;
+use tracing::debug;
 
 use crate::client::{CachedTypeInfo, InnerClient};
 use crate::codec::FrontendMessage;

libs/proxy/tokio-postgres2/src/query.rs

@@ -7,11 +7,11 @@ use std::task::{Context, Poll};
 use bytes::{BufMut, Bytes, BytesMut};
 use fallible_iterator::FallibleIterator;
 use futures_util::{Stream, ready};
-use log::{Level, debug, log_enabled};
 use pin_project_lite::pin_project;
 use postgres_protocol2::message::backend::Message;
 use postgres_protocol2::message::frontend;
 use postgres_types2::{Format, ToSql, Type};
+use tracing::debug;
 
 use crate::client::{InnerClient, Responses};
 use crate::codec::FrontendMessage;
@@ -36,7 +36,7 @@ where
     I: IntoIterator<Item = &'a (dyn ToSql + Sync)>,
     I::IntoIter: ExactSizeIterator,
 {
-    let buf = if log_enabled!(Level::Debug) {
+    let buf = if tracing::enabled!(tracing::Level::DEBUG) {
         let params = params.into_iter().collect::<Vec<_>>();
         debug!(
             "executing statement {} with parameters: {:?}",

libs/proxy/tokio-postgres2/src/simple_query.rs

@@ -6,10 +6,10 @@ use std::task::{Context, Poll};
 use bytes::Bytes;
 use fallible_iterator::FallibleIterator;
 use futures_util::{Stream, ready};
-use log::debug;
 use pin_project_lite::pin_project;
 use postgres_protocol2::message::backend::Message;
 use postgres_protocol2::message::frontend;
+use tracing::debug;
 
 use crate::client::{InnerClient, Responses};
 use crate::codec::FrontendMessage;

pageserver/client/src/mgmt_api.rs

@@ -7,7 +7,7 @@ use http_utils::error::HttpErrorBody;
 use pageserver_api::models::*;
 use pageserver_api::shard::TenantShardId;
 pub use reqwest::Body as ReqwestBody;
-use reqwest::{Certificate, IntoUrl, Method, StatusCode, Url};
+use reqwest::{IntoUrl, Method, StatusCode, Url};
 use utils::id::{TenantId, TimelineId};
 use utils::lsn::Lsn;
 
@@ -39,8 +39,8 @@ pub enum Error {
     #[error("Cancelled")]
     Cancelled,
 
-    #[error("create client: {0}{}", .0.source().map(|e| format!(": {e}")).unwrap_or_default())]
-    CreateClient(reqwest::Error),
+    #[error("request timed out: {0}")]
+    Timeout(String),
 }
 
 pub type Result<T> = std::result::Result<T, Error>;
@@ -72,24 +72,7 @@ pub enum ForceAwaitLogicalSize {
 }
 
 impl Client {
-    pub fn new(
-        mgmt_api_endpoint: String,
-        jwt: Option<&str>,
-        ssl_ca_cert: Option<Certificate>,
-    ) -> Result<Self> {
-        let mut http_client = reqwest::Client::builder();
-        if let Some(ssl_ca_cert) = ssl_ca_cert {
-            http_client = http_client.add_root_certificate(ssl_ca_cert);
-        }
-        let http_client = http_client.build().map_err(Error::CreateClient)?;
-        Ok(Self::from_client(http_client, mgmt_api_endpoint, jwt))
-    }
-
-    pub fn from_client(
-        client: reqwest::Client,
-        mgmt_api_endpoint: String,
-        jwt: Option<&str>,
-    ) -> Self {
+    pub fn new(client: reqwest::Client, mgmt_api_endpoint: String, jwt: Option<&str>) -> Self {
         Self {
             mgmt_api_endpoint,
             authorization_header: jwt.map(|jwt| format!("Bearer {jwt}")),

pageserver/pagebench/src/cmd/aux_files.rs

@@ -34,10 +34,10 @@ async fn main_impl(args: Args) -> anyhow::Result<()> {
     let args: &'static Args = Box::leak(Box::new(args));
 
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
+        reqwest::Client::new(), // TODO: support ssl_ca_file for https APIs in pagebench.
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-        None, // TODO: support ssl_ca_file for https APIs in pagebench.
-    )?);
+    ));
 
     // discover targets
     let timelines: Vec<TenantTimelineId> = crate::util::cli::targets::discover(

pageserver/pagebench/src/cmd/basebackup.rs

@@ -75,10 +75,10 @@ async fn main_impl(
     let args: &'static Args = Box::leak(Box::new(args));
 
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
+        reqwest::Client::new(), // TODO: support ssl_ca_file for https APIs in pagebench.
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-        None, // TODO: support ssl_ca_file for https APIs in pagebench.
-    )?);
+    ));
 
     // discover targets
     let timelines: Vec<TenantTimelineId> = crate::util::cli::targets::discover(

pageserver/pagebench/src/cmd/getpage_latest_lsn.rs

@@ -123,10 +123,10 @@ async fn main_impl(
     let args: &'static Args = Box::leak(Box::new(args));
 
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
+        reqwest::Client::new(), // TODO: support ssl_ca_file for https APIs in pagebench.
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-        None, // TODO: support ssl_ca_file for https APIs in pagebench.
-    )?);
+    ));
 
     if let Some(engine_str) = &args.set_io_engine {
         mgmt_api_client.put_io_engine(engine_str).await?;

pageserver/pagebench/src/cmd/ondemand_download_churn.rs

@@ -81,10 +81,10 @@ async fn main_impl(args: Args) -> anyhow::Result<()> {
     let args: &'static Args = Box::leak(Box::new(args));
 
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
+        reqwest::Client::new(), // TODO: support ssl_ca_file for https APIs in pagebench.
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-        None, // TODO: support ssl_ca_file for https APIs in pagebench.
-    )?);
+    ));
 
     if let Some(engine_str) = &args.set_io_engine {
         mgmt_api_client.put_io_engine(engine_str).await?;

pageserver/pagebench/src/cmd/trigger_initial_size_calculation.rs

@@ -38,10 +38,10 @@ async fn main_impl(args: Args) -> anyhow::Result<()> {
     let args: &'static Args = Box::leak(Box::new(args));
 
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
+        reqwest::Client::new(), // TODO: support ssl_ca_file for https APIs in pagebench.
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-        None, // TODO: support ssl_ca_file for https APIs in pagebench.
-    )?);
+    ));
 
     // discover targets
     let timelines: Vec<TenantTimelineId> = crate::util::cli::targets::discover(

pageserver/src/config.rs

@@ -65,8 +65,8 @@ pub struct PageServerConf {
     /// Period to reload certificate and private key from files.
     /// Default: 60s.
     pub ssl_cert_reload_period: Duration,
-    /// Trusted root CA certificate to use in https APIs.
-    pub ssl_ca_cert: Option<Certificate>,
+    /// Trusted root CA certificates to use in https APIs.
+    pub ssl_ca_certs: Vec<Certificate>,
 
     /// Current availability zone. Used for traffic metrics.
     pub availability_zone: Option<String>,
@@ -481,12 +481,12 @@ impl PageServerConf {
             validate_wal_contiguity: validate_wal_contiguity.unwrap_or(false),
             load_previous_heatmap: load_previous_heatmap.unwrap_or(true),
             generate_unarchival_heatmap: generate_unarchival_heatmap.unwrap_or(true),
-            ssl_ca_cert: match ssl_ca_file {
+            ssl_ca_certs: match ssl_ca_file {
                 Some(ssl_ca_file) => {
                     let buf = std::fs::read(ssl_ca_file)?;
-                    Some(Certificate::from_pem(&buf)?)
+                    Certificate::from_pem_bundle(&buf)?
                 }
-                None => None,
+                None => Vec::new(),
             },
         };
 

pageserver/src/controller_upcall_client.rs

@@ -76,7 +76,7 @@ impl StorageControllerUpcallClient {
             client = client.default_headers(headers);
         }
 
-        if let Some(ssl_ca_cert) = &conf.ssl_ca_cert {
+        for ssl_ca_cert in &conf.ssl_ca_certs {
             client = client.add_root_certificate(ssl_ca_cert.clone());
         }
 

pageserver/src/http/openapi_spec.yml

@@ -1133,6 +1133,40 @@ components:
         applied_gc_cutoff_lsn:
           type: string
           format: hex
+        safekeepers:
+          $ref: "#/components/schemas/TimelineSafekeepersInfo"
+
+    TimelineSafekeepersInfo:
+      type: object
+      required:
+        - tenant_id
+        - timeline_id
+        - generation
+        - safekeepers
+      properties:
+        tenant_id:
+          type: string
+          format: hex
+        timeline_id:
+          type: string
+          format: hex
+        generation:
+          type: integer
+        safekeepers:
+          type: array
+          items:
+            $ref: "#/components/schemas/TimelineSafekeeperInfo"
+
+    TimelineSafekeeperInfo:
+      type: object
+      required:
+        - id
+        - hostname
+      properties:
+        id:
+          type: integer
+        hostname:
+          type: string
 
     SyntheticSizeResponse:
       type: object

pageserver/src/http/routes.rs

@@ -2256,7 +2256,6 @@ async fn timeline_compact_handler(
     let state = get_state(&request);
 
     let mut flags = EnumSet::empty();
-    flags |= CompactFlags::NoYield; // run compaction to completion
 
     if Some(true) == parse_query_param::<_, bool>(&request, "force_l0_compaction")? {
         flags |= CompactFlags::ForceL0Compaction;
@@ -2417,7 +2416,6 @@ async fn timeline_checkpoint_handler(
     let state = get_state(&request);
 
     let mut flags = EnumSet::empty();
-    flags |= CompactFlags::NoYield; // run compaction to completion
     if Some(true) == parse_query_param::<_, bool>(&request, "force_l0_compaction")? {
         flags |= CompactFlags::ForceL0Compaction;
     }
@@ -3776,7 +3774,7 @@ pub fn make_router(
         )
         .put(
             "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/mark_invisible",
-            |r| testing_api_handler("mark timeline invisible", r, timeline_mark_invisible_handler),
+            |r| api_handler( r, timeline_mark_invisible_handler),
         )
         .put(
             "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/checkpoint",

pageserver/src/metrics.rs

@@ -10,7 +10,7 @@ use std::time::{Duration, Instant};
 use enum_map::{Enum as _, EnumMap};
 use futures::Future;
 use metrics::{
-    Counter, CounterVec, Gauge, GaugeVec, Histogram, HistogramVec, IntCounter, IntCounterPair,
+    Counter, CounterVec, GaugeVec, Histogram, HistogramVec, IntCounter, IntCounterPair,
     IntCounterPairVec, IntCounterVec, IntGauge, IntGaugeVec, UIntGauge, UIntGaugeVec,
     register_counter_vec, register_gauge_vec, register_histogram, register_histogram_vec,
     register_int_counter, register_int_counter_pair_vec, register_int_counter_vec,
@@ -499,15 +499,6 @@ pub(crate) static WAIT_LSN_IN_PROGRESS_GLOBAL_MICROS: Lazy<IntCounter> = Lazy::n
     .expect("failed to define a metric")
 });
 
-static FLUSH_WAIT_UPLOAD_TIME: Lazy<GaugeVec> = Lazy::new(|| {
-    register_gauge_vec!(
-        "pageserver_flush_wait_upload_seconds",
-        "Time spent waiting for preceding uploads during layer flush",
-        &["tenant_id", "shard_id", "timeline_id"]
-    )
-    .expect("failed to define a metric")
-});
-
 static LAST_RECORD_LSN: Lazy<IntGaugeVec> = Lazy::new(|| {
     register_int_gauge_vec!(
         "pageserver_last_record_lsn",
@@ -2864,7 +2855,6 @@ pub(crate) struct TimelineMetrics {
     timeline_id: String,
     pub flush_time_histo: StorageTimeMetrics,
     pub flush_delay_histo: StorageTimeMetrics,
-    pub flush_wait_upload_time_gauge: Gauge,
     pub compact_time_histo: StorageTimeMetrics,
     pub create_images_time_histo: StorageTimeMetrics,
     pub logical_size_histo: StorageTimeMetrics,
@@ -2916,9 +2906,6 @@ impl TimelineMetrics {
             &shard_id,
             &timeline_id,
         );
-        let flush_wait_upload_time_gauge = FLUSH_WAIT_UPLOAD_TIME
-            .get_metric_with_label_values(&[&tenant_id, &shard_id, &timeline_id])
-            .unwrap();
         let compact_time_histo = StorageTimeMetrics::new(
             StorageTimeOperation::Compact,
             &tenant_id,
@@ -3046,7 +3033,6 @@ impl TimelineMetrics {
             timeline_id,
             flush_time_histo,
             flush_delay_histo,
-            flush_wait_upload_time_gauge,
             compact_time_histo,
             create_images_time_histo,
             logical_size_histo,
@@ -3096,14 +3082,6 @@ impl TimelineMetrics {
         self.resident_physical_size_gauge.get()
     }
 
-    pub(crate) fn flush_wait_upload_time_gauge_add(&self, duration: f64) {
-        self.flush_wait_upload_time_gauge.add(duration);
-        crate::metrics::FLUSH_WAIT_UPLOAD_TIME
-            .get_metric_with_label_values(&[&self.tenant_id, &self.shard_id, &self.timeline_id])
-            .unwrap()
-            .add(duration);
-    }
-
     /// Generates TIMELINE_LAYER labels for a persistent layer.
     fn make_layer_labels(&self, layer_desc: &PersistentLayerDesc) -> [&str; 5] {
         let level = match LayerMap::is_l0(&layer_desc.key_range, layer_desc.is_delta()) {
@@ -3207,7 +3185,6 @@ impl TimelineMetrics {
         let shard_id = &self.shard_id;
         let _ = LAST_RECORD_LSN.remove_label_values(&[tenant_id, shard_id, timeline_id]);
         let _ = DISK_CONSISTENT_LSN.remove_label_values(&[tenant_id, shard_id, timeline_id]);
-        let _ = FLUSH_WAIT_UPLOAD_TIME.remove_label_values(&[tenant_id, shard_id, timeline_id]);
         let _ = STANDBY_HORIZON.remove_label_values(&[tenant_id, shard_id, timeline_id]);
         {
             RESIDENT_PHYSICAL_SIZE_GLOBAL.sub(self.resident_physical_size_get());

pageserver/src/task_mgr.rs

@@ -38,6 +38,7 @@ use std::panic::AssertUnwindSafe;
 use std::str::FromStr;
 use std::sync::atomic::{AtomicU64, Ordering};
 use std::sync::{Arc, Mutex};
+use std::time::Duration;
 
 use futures::FutureExt;
 use once_cell::sync::Lazy;
@@ -584,18 +585,25 @@ pub async fn shutdown_tasks(
                 // warn to catch these in tests; there shouldn't be any
                 warn!(name = task.name, tenant_shard_id = ?tenant_shard_id, timeline_id = ?timeline_id, kind = ?task_kind, "stopping left-over");
             }
-            if tokio::time::timeout(std::time::Duration::from_secs(1), &mut join_handle)
+            const INITIAL_COMPLAIN_TIMEOUT: Duration = Duration::from_secs(1);
+            const PERIODIC_COMPLAIN_TIMEOUT: Duration = Duration::from_secs(60);
+            if tokio::time::timeout(INITIAL_COMPLAIN_TIMEOUT, &mut join_handle)
                 .await
                 .is_err()
             {
                 // allow some time to elapse before logging to cut down the number of log
                 // lines.
                 info!("waiting for task {} to shut down", task.name);
-                // we never handled this return value, but:
-                // - we don't deschedule which would lead to is_cancelled
-                // - panics are already logged (is_panicked)
-                // - task errors are already logged in the wrapper
-                let _ = join_handle.await;
+                loop {
+                    tokio::select! {
+                        // we never handled this return value, but:
+                        // - we don't deschedule which would lead to is_cancelled
+                        // - panics are already logged (is_panicked)
+                        // - task errors are already logged in the wrapper
+                        _ = &mut join_handle => break,
+                        _ = tokio::time::sleep(PERIODIC_COMPLAIN_TIMEOUT) => info!("still waiting for task {} to shut down", task.name),
+                    }
+                }
                 info!("task {} completed", task.name);
             }
         } else {

pageserver/src/tenant.rs

@@ -3080,6 +3080,7 @@ impl Tenant {
             let mut has_pending_l0 = false;
             for timeline in compact_l0 {
                 let ctx = &ctx.with_scope_timeline(&timeline);
+                // NB: don't set CompactFlags::YieldForL0, since this is an L0-only compaction pass.
                 let outcome = timeline
                     .compact(cancel, CompactFlags::OnlyL0Compaction.into(), ctx)
                     .instrument(info_span!("compact_timeline", timeline_id = %timeline.timeline_id))
@@ -3097,14 +3098,9 @@ impl Tenant {
             }
         }
 
-        // Pass 2: image compaction and timeline offloading. If any timelines have accumulated
-        // more L0 layers, they may also be compacted here.
-        //
-        // NB: image compaction may yield if there is pending L0 compaction.
-        //
-        // TODO: it will only yield if there is pending L0 compaction on the same timeline. If a
-        // different timeline needs compaction, it won't. It should check `l0_compaction_trigger`.
-        // We leave this for a later PR.
+        // Pass 2: image compaction and timeline offloading. If any timelines have accumulated more
+        // L0 layers, they may also be compacted here. Image compaction will yield if there is
+        // pending L0 compaction on any tenant timeline.
         //
         // TODO: consider ordering timelines by some priority, e.g. time since last full compaction,
         // amount of L1 delta debt or garbage, offload-eligible timelines first, etc.
@@ -3115,8 +3111,14 @@ impl Tenant {
             }
             let ctx = &ctx.with_scope_timeline(&timeline);
 
+            // Yield for L0 if the separate L0 pass is enabled (otherwise there's no point).
+            let mut flags = EnumSet::default();
+            if self.get_compaction_l0_first() {
+                flags |= CompactFlags::YieldForL0;
+            }
+
             let mut outcome = timeline
-                .compact(cancel, EnumSet::default(), ctx)
+                .compact(cancel, flags, ctx)
                 .instrument(info_span!("compact_timeline", timeline_id = %timeline.timeline_id))
                 .await
                 .inspect_err(|err| self.maybe_trip_compaction_breaker(err))?;
@@ -6516,11 +6518,7 @@ mod tests {
 
         tline.freeze_and_flush().await?;
         tline
-            .compact(
-                &CancellationToken::new(),
-                CompactFlags::NoYield.into(),
-                &ctx,
-            )
+            .compact(&CancellationToken::new(), EnumSet::default(), &ctx)
             .await?;
 
         let mut writer = tline.writer().await;
@@ -6537,11 +6535,7 @@ mod tests {
 
         tline.freeze_and_flush().await?;
         tline
-            .compact(
-                &CancellationToken::new(),
-                CompactFlags::NoYield.into(),
-                &ctx,
-            )
+            .compact(&CancellationToken::new(), EnumSet::default(), &ctx)
             .await?;
 
         let mut writer = tline.writer().await;
@@ -6558,11 +6552,7 @@ mod tests {
 
         tline.freeze_and_flush().await?;
         tline
-            .compact(
-                &CancellationToken::new(),
-                CompactFlags::NoYield.into(),
-                &ctx,
-            )
+            .compact(&CancellationToken::new(), EnumSet::default(), &ctx)
             .await?;
 
         let mut writer = tline.writer().await;
@@ -6579,11 +6569,7 @@ mod tests {
 
         tline.freeze_and_flush().await?;
         tline
-            .compact(
-                &CancellationToken::new(),
-                CompactFlags::NoYield.into(),
-                &ctx,
-            )
+            .compact(&CancellationToken::new(), EnumSet::default(), &ctx)
             .await?;
 
         assert_eq!(
@@ -6666,9 +6652,7 @@ mod tests {
             timeline.freeze_and_flush().await?;
             if compact {
                 // this requires timeline to be &Arc<Timeline>
-                timeline
-                    .compact(&cancel, CompactFlags::NoYield.into(), ctx)
-                    .await?;
+                timeline.compact(&cancel, EnumSet::default(), ctx).await?;
             }
 
             // this doesn't really need to use the timeline_id target, but it is closer to what it
@@ -6995,7 +6979,6 @@ mod tests {
         child_timeline.freeze_and_flush().await?;
         let mut flags = EnumSet::new();
         flags.insert(CompactFlags::ForceRepartition);
-        flags.insert(CompactFlags::NoYield);
         child_timeline
             .compact(&CancellationToken::new(), flags, &ctx)
             .await?;
@@ -7374,9 +7357,7 @@ mod tests {
 
             // Perform a cycle of flush, compact, and GC
             tline.freeze_and_flush().await?;
-            tline
-                .compact(&cancel, CompactFlags::NoYield.into(), &ctx)
-                .await?;
+            tline.compact(&cancel, EnumSet::default(), &ctx).await?;
             tenant
                 .gc_iteration(Some(tline.timeline_id), 0, Duration::ZERO, &cancel, &ctx)
                 .await?;
@@ -7705,7 +7686,6 @@ mod tests {
                             let mut flags = EnumSet::new();
                             flags.insert(CompactFlags::ForceImageLayerCreation);
                             flags.insert(CompactFlags::ForceRepartition);
-                            flags.insert(CompactFlags::NoYield);
                             flags
                         } else {
                             EnumSet::empty()
@@ -7756,9 +7736,7 @@ mod tests {
         let before_num_l0_delta_files =
             tline.layers.read().await.layer_map()?.level0_deltas().len();
 
-        tline
-            .compact(&cancel, CompactFlags::NoYield.into(), &ctx)
-            .await?;
+        tline.compact(&cancel, EnumSet::default(), &ctx).await?;
 
         let after_num_l0_delta_files = tline.layers.read().await.layer_map()?.level0_deltas().len();
 
@@ -7923,7 +7901,6 @@ mod tests {
                             let mut flags = EnumSet::new();
                             flags.insert(CompactFlags::ForceImageLayerCreation);
                             flags.insert(CompactFlags::ForceRepartition);
-                            flags.insert(CompactFlags::NoYield);
                             flags
                         },
                         &ctx,
@@ -8386,7 +8363,6 @@ mod tests {
                     let mut flags = EnumSet::new();
                     flags.insert(CompactFlags::ForceImageLayerCreation);
                     flags.insert(CompactFlags::ForceRepartition);
-                    flags.insert(CompactFlags::NoYield);
                     flags
                 },
                 &ctx,
@@ -8454,7 +8430,6 @@ mod tests {
                     let mut flags = EnumSet::new();
                     flags.insert(CompactFlags::ForceImageLayerCreation);
                     flags.insert(CompactFlags::ForceRepartition);
-                    flags.insert(CompactFlags::NoYield);
                     flags
                 },
                 &ctx,
@@ -11551,4 +11526,255 @@ mod tests {
 
         Ok(())
     }
+
+    #[cfg(feature = "testing")]
+    #[tokio::test]
+    async fn test_synthetic_size_calculation_with_invisible_branches() -> anyhow::Result<()> {
+        use pageserver_api::models::TimelineVisibilityState;
+
+        use crate::tenant::size::gather_inputs;
+
+        let tenant_conf = pageserver_api::models::TenantConfig {
+            // Ensure that we don't compute gc_cutoffs (which needs reading the layer files)
+            pitr_interval: Some(Duration::ZERO),
+            ..Default::default()
+        };
+        let harness = TenantHarness::create_custom(
+            "test_synthetic_size_calculation_with_invisible_branches",
+            tenant_conf,
+            TenantId::generate(),
+            ShardIdentity::unsharded(),
+            Generation::new(0xdeadbeef),
+        )
+        .await?;
+        let (tenant, ctx) = harness.load().await;
+        let main_tline = tenant
+            .create_test_timeline_with_layers(
+                TIMELINE_ID,
+                Lsn(0x10),
+                DEFAULT_PG_VERSION,
+                &ctx,
+                vec![],
+                vec![],
+                vec![],
+                Lsn(0x100),
+            )
+            .await?;
+
+        let snapshot1 = TimelineId::from_array(hex!("11223344556677881122334455667790"));
+        tenant
+            .branch_timeline_test_with_layers(
+                &main_tline,
+                snapshot1,
+                Some(Lsn(0x20)),
+                &ctx,
+                vec![],
+                vec![],
+                Lsn(0x50),
+            )
+            .await?;
+        let snapshot2 = TimelineId::from_array(hex!("11223344556677881122334455667791"));
+        tenant
+            .branch_timeline_test_with_layers(
+                &main_tline,
+                snapshot2,
+                Some(Lsn(0x30)),
+                &ctx,
+                vec![],
+                vec![],
+                Lsn(0x50),
+            )
+            .await?;
+        let snapshot3 = TimelineId::from_array(hex!("11223344556677881122334455667792"));
+        tenant
+            .branch_timeline_test_with_layers(
+                &main_tline,
+                snapshot3,
+                Some(Lsn(0x40)),
+                &ctx,
+                vec![],
+                vec![],
+                Lsn(0x50),
+            )
+            .await?;
+        let limit = Arc::new(Semaphore::new(1));
+        let max_retention_period = None;
+        let mut logical_size_cache = HashMap::new();
+        let cause = LogicalSizeCalculationCause::EvictionTaskImitation;
+        let cancel = CancellationToken::new();
+
+        let inputs = gather_inputs(
+            &tenant,
+            &limit,
+            max_retention_period,
+            &mut logical_size_cache,
+            cause,
+            &cancel,
+            &ctx,
+        )
+        .instrument(info_span!(
+            "gather_inputs",
+            tenant_id = "unknown",
+            shard_id = "unknown",
+        ))
+        .await?;
+        use crate::tenant::size::{LsnKind, ModelInputs, SegmentMeta};
+        use LsnKind::*;
+        use tenant_size_model::Segment;
+        let ModelInputs { mut segments, .. } = inputs;
+        segments.retain(|s| s.timeline_id == TIMELINE_ID);
+        for segment in segments.iter_mut() {
+            segment.segment.parent = None; // We don't care about the parent for the test
+            segment.segment.size = None; // We don't care about the size for the test
+        }
+        assert_eq!(
+            segments,
+            [
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x10,
+                        size: None,
+                        needed: false,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchStart,
+                },
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x20,
+                        size: None,
+                        needed: false,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchPoint,
+                },
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x30,
+                        size: None,
+                        needed: false,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchPoint,
+                },
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x40,
+                        size: None,
+                        needed: false,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchPoint,
+                },
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x100,
+                        size: None,
+                        needed: false,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: GcCutOff,
+                }, // we need to retain everything above the last branch point
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x100,
+                        size: None,
+                        needed: true,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchEnd,
+                },
+            ]
+        );
+
+        main_tline
+            .remote_client
+            .schedule_index_upload_for_timeline_invisible_state(
+                TimelineVisibilityState::Invisible,
+            )?;
+        main_tline.remote_client.wait_completion().await?;
+        let inputs = gather_inputs(
+            &tenant,
+            &limit,
+            max_retention_period,
+            &mut logical_size_cache,
+            cause,
+            &cancel,
+            &ctx,
+        )
+        .instrument(info_span!(
+            "gather_inputs",
+            tenant_id = "unknown",
+            shard_id = "unknown",
+        ))
+        .await?;
+        let ModelInputs { mut segments, .. } = inputs;
+        segments.retain(|s| s.timeline_id == TIMELINE_ID);
+        for segment in segments.iter_mut() {
+            segment.segment.parent = None; // We don't care about the parent for the test
+            segment.segment.size = None; // We don't care about the size for the test
+        }
+        assert_eq!(
+            segments,
+            [
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x10,
+                        size: None,
+                        needed: false,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchStart,
+                },
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x20,
+                        size: None,
+                        needed: false,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchPoint,
+                },
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x30,
+                        size: None,
+                        needed: false,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchPoint,
+                },
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x40,
+                        size: None,
+                        needed: false,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchPoint,
+                },
+                SegmentMeta {
+                    segment: Segment {
+                        parent: None,
+                        lsn: 0x40, // Branch end LSN == last branch point LSN
+                        size: None,
+                        needed: true,
+                    },
+                    timeline_id: TIMELINE_ID,
+                    kind: BranchEnd,
+                },
+            ]
+        );
+        Ok(())
+    }
 }

pageserver/src/tenant/remote_timeline_client.rs

@@ -1968,9 +1968,7 @@ impl RemoteTimelineClient {
     /// Pick next tasks from the queue, and start as many of them as possible without violating
     /// the ordering constraints.
     ///
-    /// TODO: consider limiting the number of in-progress tasks, beyond what remote_storage does.
-    /// This can launch an unbounded number of queued tasks. `UploadQueue::next_ready()` also has
-    /// worst-case quadratic cost in the number of tasks, and may struggle beyond 10,000 tasks.
+    /// The number of inprogress tasks is limited by `Self::inprogress_tasks`, see `next_ready`.
     fn launch_queued_tasks(self: &Arc<Self>, upload_queue: &mut UploadQueueInitialized) {
         while let Some((mut next_op, coalesced_ops)) = upload_queue.next_ready() {
             debug!("starting op: {next_op}");
@@ -2218,6 +2216,11 @@ impl RemoteTimelineClient {
                     }
                     res
                 }
+                // TODO: this should wait for the deletion to be executed by the deletion queue.
+                // Otherwise, the deletion may race with an upload and wrongfully delete a newer
+                // file. Some of the above logic attempts to work around this, it should be replaced
+                // by the upload queue ordering guarantees (see `can_bypass`). See:
+                // <https://github.com/neondatabase/neon/issues/10283>.
                 UploadOp::Delete(delete) => {
                     if self.config.read().unwrap().block_deletions {
                         let mut queue_locked = self.upload_queue.lock().unwrap();

pageserver/src/tenant/size.rs

@@ -33,7 +33,7 @@ pub struct ModelInputs {
 }
 
 /// A [`Segment`], with some extra information for display purposes
-#[derive(Debug, serde::Serialize, serde::Deserialize)]
+#[derive(Debug, serde::Serialize, serde::Deserialize, PartialEq, Eq)]
 pub struct SegmentMeta {
     pub segment: Segment,
     pub timeline_id: TimelineId,
@@ -248,6 +248,8 @@ pub(super) async fn gather_inputs(
             None
         };
 
+        let branch_is_invisible = timeline.is_invisible() == Some(true);
+
         let lease_points = gc_info
             .leases
             .keys()
@@ -271,7 +273,10 @@ pub(super) async fn gather_inputs(
             .map(|(lsn, _child_id, _is_offloaded)| (lsn, LsnKind::BranchPoint))
             .collect::<Vec<_>>();
 
-        lsns.extend(lease_points.iter().map(|&lsn| (lsn, LsnKind::LeasePoint)));
+        if !branch_is_invisible {
+            // Do not count lease points for invisible branches.
+            lsns.extend(lease_points.iter().map(|&lsn| (lsn, LsnKind::LeasePoint)));
+        }
 
         drop(gc_info);
 
@@ -287,7 +292,9 @@ pub(super) async fn gather_inputs(
 
         // Add a point for the PITR cutoff
         let branch_start_needed = next_pitr_cutoff <= branch_start_lsn;
-        if !branch_start_needed {
+        if !branch_start_needed && !branch_is_invisible {
+            // Only add the GcCutOff point when the timeline is visible; otherwise, do not compute the size for the LSN
+            // range from the last branch point to the latest data.
             lsns.push((next_pitr_cutoff, LsnKind::GcCutOff));
         }
 
@@ -373,11 +380,19 @@ pub(super) async fn gather_inputs(
             }
         }
 
+        let branch_end_lsn = if branch_is_invisible {
+            // If the branch is invisible, the branch end is the last requested LSN (likely a branch cutoff point).
+            segments.last().unwrap().segment.lsn
+        } else {
+            // Otherwise, the branch end is the last record LSN.
+            last_record_lsn.0
+        };
+
         // Current end of the timeline
         segments.push(SegmentMeta {
             segment: Segment {
                 parent: Some(parent),
-                lsn: last_record_lsn.0,
+                lsn: branch_end_lsn,
                 size: None, // Filled in later, if necessary
                 needed: true,
             },
@@ -609,6 +624,7 @@ async fn calculate_logical_size(
     Ok(TimelineAtLsnSizeResult(timeline, lsn, size_res))
 }
 
+#[cfg(test)]
 #[test]
 fn verify_size_for_multiple_branches() {
     // this is generated from integration test test_tenant_size_with_multiple_branches, but this way
@@ -766,6 +782,7 @@ fn verify_size_for_multiple_branches() {
     assert_eq!(inputs.calculate(), 37_851_408);
 }
 
+#[cfg(test)]
 #[test]
 fn verify_size_for_one_branch() {
     let doc = r#"

pageserver/src/tenant/timeline.rs

@@ -84,8 +84,8 @@ use self::eviction_task::EvictionTaskTimelineState;
 use self::layer_manager::LayerManager;
 use self::logical_size::LogicalSize;
 use self::walreceiver::{WalReceiver, WalReceiverConf};
+use super::remote_timeline_client::RemoteTimelineClient;
 use super::remote_timeline_client::index::{GcCompactionState, IndexPart};
-use super::remote_timeline_client::{RemoteTimelineClient, WaitCompletionError};
 use super::secondary::heatmap::HeatMapLayer;
 use super::storage_layer::{LayerFringe, LayerVisibilityHint, ReadableLayer};
 use super::tasks::log_compaction_error;
@@ -870,9 +870,14 @@ pub(crate) enum CompactFlags {
     OnlyL0Compaction,
     EnhancedGcBottomMostCompaction,
     DryRun,
-    /// Disables compaction yielding e.g. due to high L0 count. This is set e.g. when requesting
-    /// compaction via HTTP API.
-    NoYield,
+    /// Makes image compaction yield if there's pending L0 compaction. This should always be used in
+    /// the background compaction task, since we want to aggressively compact down L0 to bound
+    /// read amplification.
+    ///
+    /// It only makes sense to use this when `compaction_l0_first` is enabled (such that we yield to
+    /// an L0 compaction pass), and without `OnlyL0Compaction` (L0 compaction shouldn't yield for L0
+    /// compaction).
+    YieldForL0,
 }
 
 #[serde_with::serde_as]
@@ -1891,18 +1896,19 @@ impl Timeline {
         // out by other background tasks (including image compaction). We request this via
         // `BackgroundLoopKind::L0Compaction`.
         //
-        // If this is a regular compaction pass, and L0-only compaction is enabled in the config,
-        // then we should yield for immediate L0 compaction if necessary while we're waiting for the
-        // background task semaphore. There's no point yielding otherwise, since we'd just end up
-        // right back here.
+        // Yield for pending L0 compaction while waiting for the semaphore.
         let is_l0_only = options.flags.contains(CompactFlags::OnlyL0Compaction);
         let semaphore_kind = match is_l0_only && self.get_compaction_l0_semaphore() {
             true => BackgroundLoopKind::L0Compaction,
             false => BackgroundLoopKind::Compaction,
         };
-        let yield_for_l0 = !is_l0_only
-            && self.get_compaction_l0_first()
-            && !options.flags.contains(CompactFlags::NoYield);
+        let yield_for_l0 = options.flags.contains(CompactFlags::YieldForL0);
+        if yield_for_l0 {
+            // If this is an L0 pass, it doesn't make sense to yield for L0.
+            debug_assert!(!is_l0_only, "YieldForL0 during L0 pass");
+            // If `compaction_l0_first` is disabled, there's no point yielding.
+            debug_assert!(self.get_compaction_l0_first(), "YieldForL0 without L0 pass");
+        }
 
         let acquire = async move {
             let guard = self.compaction_lock.lock().await;
@@ -2209,6 +2215,10 @@ impl Timeline {
         self.remote_client.is_archived()
     }
 
+    pub(crate) fn is_invisible(&self) -> Option<bool> {
+        self.remote_client.is_invisible()
+    }
+
     pub(crate) fn is_stopping(&self) -> bool {
         self.current_state() == TimelineState::Stopping
     }
@@ -2562,14 +2572,6 @@ impl Timeline {
         Some(max(l0_flush_stall_threshold, compaction_threshold))
     }
 
-    fn get_l0_flush_wait_upload(&self) -> bool {
-        let tenant_conf = self.tenant_conf.load();
-        tenant_conf
-            .tenant_conf
-            .l0_flush_wait_upload
-            .unwrap_or(self.conf.default_tenant_conf.l0_flush_wait_upload)
-    }
-
     fn get_image_creation_threshold(&self) -> usize {
         let tenant_conf = self.tenant_conf.load();
         tenant_conf
@@ -4591,27 +4593,6 @@ impl Timeline {
             // release lock on 'layers'
         };
 
-        // Backpressure mechanism: wait with continuation of the flush loop until we have uploaded all layer files.
-        // This makes us refuse ingest until the new layers have been persisted to the remote
-        // TODO: remove this, and rely on l0_flush_{delay,stall}_threshold instead.
-        if self.get_l0_flush_wait_upload() {
-            let start = Instant::now();
-            self.remote_client
-                .wait_completion()
-                .await
-                .map_err(|e| match e {
-                    WaitCompletionError::UploadQueueShutDownOrStopped
-                    | WaitCompletionError::NotInitialized(
-                        NotInitialized::ShuttingDown | NotInitialized::Stopped,
-                    ) => FlushLayerError::Cancelled,
-                    WaitCompletionError::NotInitialized(NotInitialized::Uninitialized) => {
-                        FlushLayerError::Other(anyhow!(e).into())
-                    }
-                })?;
-            let duration = start.elapsed().as_secs_f64();
-            self.metrics.flush_wait_upload_time_gauge_add(duration);
-        }
-
         // FIXME: between create_delta_layer and the scheduling of the upload in `update_metadata_file`,
         // a compaction can delete the file and then it won't be available for uploads any more.
         // We still schedule the upload, resulting in an error, but ideally we'd somehow avoid this

pageserver/src/tenant/timeline/compaction.rs

@@ -394,8 +394,8 @@ impl GcCompactionQueue {
                 if job.dry_run {
                     flags |= CompactFlags::DryRun;
                 }
-                if options.flags.contains(CompactFlags::NoYield) {
-                    flags |= CompactFlags::NoYield;
+                if options.flags.contains(CompactFlags::YieldForL0) {
+                    flags |= CompactFlags::YieldForL0;
                 }
                 let options = CompactOptions {
                     flags,
@@ -983,7 +983,7 @@ impl Timeline {
 
         // Yield if we have pending L0 compaction. The scheduler will do another pass.
         if (l0_outcome == CompactionOutcome::Pending || l0_outcome == CompactionOutcome::YieldForL0)
-            && !options.flags.contains(CompactFlags::NoYield)
+            && options.flags.contains(CompactFlags::YieldForL0)
         {
             info!("image/ancestor compaction yielding for L0 compaction");
             return Ok(CompactionOutcome::YieldForL0);
@@ -1028,7 +1028,7 @@ impl Timeline {
                             .load()
                             .as_ref()
                             .clone(),
-                        !options.flags.contains(CompactFlags::NoYield),
+                        options.flags.contains(CompactFlags::YieldForL0),
                     )
                     .await
                     .inspect_err(|err| {
@@ -2635,7 +2635,7 @@ impl Timeline {
     ) -> Result<CompactionOutcome, CompactionError> {
         let sub_compaction = options.sub_compaction;
         let job = GcCompactJob::from_compact_options(options.clone());
-        let no_yield = options.flags.contains(CompactFlags::NoYield);
+        let yield_for_l0 = options.flags.contains(CompactFlags::YieldForL0);
         if sub_compaction {
             info!(
                 "running enhanced gc bottom-most compaction with sub-compaction, splitting compaction jobs"
@@ -2650,7 +2650,7 @@ impl Timeline {
                     idx + 1,
                     jobs_len
                 );
-                self.compact_with_gc_inner(cancel, job, ctx, no_yield)
+                self.compact_with_gc_inner(cancel, job, ctx, yield_for_l0)
                     .await?;
             }
             if jobs_len == 0 {
@@ -2658,7 +2658,8 @@ impl Timeline {
             }
             return Ok(CompactionOutcome::Done);
         }
-        self.compact_with_gc_inner(cancel, job, ctx, no_yield).await
+        self.compact_with_gc_inner(cancel, job, ctx, yield_for_l0)
+            .await
     }
 
     async fn compact_with_gc_inner(
@@ -2666,7 +2667,7 @@ impl Timeline {
         cancel: &CancellationToken,
         job: GcCompactJob,
         ctx: &RequestContext,
-        no_yield: bool,
+        yield_for_l0: bool,
     ) -> Result<CompactionOutcome, CompactionError> {
         // Block other compaction/GC tasks from running for now. GC-compaction could run along
         // with legacy compaction tasks in the future. Always ensure the lock order is compaction -> gc.
@@ -2936,18 +2937,15 @@ impl Timeline {
             if cancel.is_cancelled() {
                 return Err(CompactionError::ShuttingDown);
             }
-            if !no_yield {
-                let should_yield = self
+            let should_yield = yield_for_l0
+                && self
                     .l0_compaction_trigger
                     .notified()
                     .now_or_never()
                     .is_some();
-                if should_yield {
-                    tracing::info!(
-                        "preempt gc-compaction when downloading layers: too many L0 layers"
-                    );
-                    return Ok(CompactionOutcome::YieldForL0);
-                }
+            if should_yield {
+                tracing::info!("preempt gc-compaction when downloading layers: too many L0 layers");
+                return Ok(CompactionOutcome::YieldForL0);
             }
             let resident_layer = layer
                 .download_and_keep_resident(ctx)
@@ -3081,21 +3079,17 @@ impl Timeline {
                 return Err(CompactionError::ShuttingDown);
             }
 
-            if !no_yield {
-                keys_processed += 1;
-                if keys_processed % 1000 == 0 {
-                    let should_yield = self
-                        .l0_compaction_trigger
-                        .notified()
-                        .now_or_never()
-                        .is_some();
-                    if should_yield {
-                        tracing::info!(
-                            "preempt gc-compaction in the main loop: too many L0 layers"
-                        );
-                        return Ok(CompactionOutcome::YieldForL0);
-                    }
-                }
+            keys_processed += 1;
+            let should_yield = yield_for_l0
+                && keys_processed % 1000 == 0
+                && self
+                    .l0_compaction_trigger
+                    .notified()
+                    .now_or_never()
+                    .is_some();
+            if should_yield {
+                tracing::info!("preempt gc-compaction in the main loop: too many L0 layers");
+                return Ok(CompactionOutcome::YieldForL0);
             }
             if self.shard_identity.is_key_disposable(&key) {
                 // If this shard does not need to store this key, simply skip it.

pageserver/src/tenant/timeline/detach_ancestor.rs

@@ -235,7 +235,7 @@ pub(super) async fn prepare(
         return Err(NoAncestor);
     }
 
-    check_no_archived_children_of_ancestor(tenant, detached, &ancestor, ancestor_lsn)?;
+    check_no_archived_children_of_ancestor(tenant, detached, &ancestor, ancestor_lsn, behavior)?;
 
     if let DetachBehavior::MultiLevelAndNoReparent = behavior {
         // If the ancestor has an ancestor, we might be able to fast-path detach it if the current ancestor does not have any data written/used by the detaching timeline.
@@ -249,7 +249,13 @@ pub(super) async fn prepare(
             ancestor_lsn = ancestor.ancestor_lsn; // Get the LSN first before resetting the `ancestor` variable
             ancestor = ancestor_of_ancestor;
             // TODO: do we still need to check if we don't want to reparent?
-            check_no_archived_children_of_ancestor(tenant, detached, &ancestor, ancestor_lsn)?;
+            check_no_archived_children_of_ancestor(
+                tenant,
+                detached,
+                &ancestor,
+                ancestor_lsn,
+                behavior,
+            )?;
         }
     } else if ancestor.ancestor_timeline.is_some() {
         // non-technical requirement; we could flatten N ancestors just as easily but we chose
@@ -1156,31 +1162,44 @@ fn check_no_archived_children_of_ancestor(
     detached: &Arc<Timeline>,
     ancestor: &Arc<Timeline>,
     ancestor_lsn: Lsn,
+    detach_behavior: DetachBehavior,
 ) -> Result<(), Error> {
-    let timelines = tenant.timelines.lock().unwrap();
-    let timelines_offloaded = tenant.timelines_offloaded.lock().unwrap();
-    for timeline in reparentable_timelines(timelines.values(), detached, ancestor, ancestor_lsn) {
-        if timeline.is_archived() == Some(true) {
-            return Err(Error::Archived(timeline.timeline_id));
-        }
-    }
-    for timeline_offloaded in timelines_offloaded.values() {
-        if timeline_offloaded.ancestor_timeline_id != Some(ancestor.timeline_id) {
-            continue;
-        }
-        // This forbids the detach ancestor feature if flattened timelines are present,
-        // even if the ancestor_lsn is from after the branchpoint of the detached timeline.
-        // But as per current design, we don't record the ancestor_lsn of flattened timelines.
-        // This is a bit unfortunate, but as of writing this we don't support flattening
-        // anyway. Maybe we can evolve the data model in the future.
-        if let Some(retain_lsn) = timeline_offloaded.ancestor_retain_lsn {
-            let is_earlier = retain_lsn <= ancestor_lsn;
-            if !is_earlier {
-                continue;
+    match detach_behavior {
+        DetachBehavior::NoAncestorAndReparent => {
+            let timelines = tenant.timelines.lock().unwrap();
+            let timelines_offloaded = tenant.timelines_offloaded.lock().unwrap();
+
+            for timeline in
+                reparentable_timelines(timelines.values(), detached, ancestor, ancestor_lsn)
+            {
+                if timeline.is_archived() == Some(true) {
+                    return Err(Error::Archived(timeline.timeline_id));
+                }
+            }
+
+            for timeline_offloaded in timelines_offloaded.values() {
+                if timeline_offloaded.ancestor_timeline_id != Some(ancestor.timeline_id) {
+                    continue;
+                }
+                // This forbids the detach ancestor feature if flattened timelines are present,
+                // even if the ancestor_lsn is from after the branchpoint of the detached timeline.
+                // But as per current design, we don't record the ancestor_lsn of flattened timelines.
+                // This is a bit unfortunate, but as of writing this we don't support flattening
+                // anyway. Maybe we can evolve the data model in the future.
+                if let Some(retain_lsn) = timeline_offloaded.ancestor_retain_lsn {
+                    let is_earlier = retain_lsn <= ancestor_lsn;
+                    if !is_earlier {
+                        continue;
+                    }
+                }
+                return Err(Error::Archived(timeline_offloaded.timeline_id));
             }
         }
-        return Err(Error::Archived(timeline_offloaded.timeline_id));
+        DetachBehavior::MultiLevelAndNoReparent => {
+            // We don't need to check anything if the user requested to not reparent.
+        }
     }
+
     Ok(())
 }
 

pgxn/neon/file_cache.c

@@ -647,18 +647,25 @@ lfc_cache_containsv(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 	return found;
 }
 
+#if PG_MAJORVERSION_NUM >= 16
+static PGIOAlignedBlock voidblock = {0};
+#else
+static PGAlignedBlock voidblock = {0};
+#endif
+#define SCRIBBLEPAGE (&voidblock.data)
+
 /*
  * Try to read pages from local cache.
  * Returns the number of pages read from the local cache, and sets bits in
- * 'read' for the pages which were read. This may scribble over buffers not
- * marked in 'read', so be careful with operation ordering.
+ * 'mask' for the pages which were read. This may scribble over buffers not
+ * marked in 'mask', so be careful with operation ordering.
  *
  * In case of error local file cache is disabled (lfc->limit is set to zero),
- * and -1 is returned. Note that 'read' and the buffers may be touched and in
- * an otherwise invalid state.
+ * and -1 is returned.
  *
- * If the mask argument is supplied, bits will be set at the offsets of pages
- * that were present and read from the LFC.
+ * If the mask argument is supplied, we'll only try to read those pages which
+ * don't have their bits set on entry. At exit, pages which were successfully
+ * read from LFC will have their bits set.
  */
 int
 lfc_readv_select(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
@@ -693,23 +700,43 @@ lfc_readv_select(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 	while (nblocks > 0)
 	{
 		struct iovec iov[PG_IOV_MAX];
-		int		chunk_offs = blkno & (BLOCKS_PER_CHUNK - 1);
+		int8	chunk_mask[BLOCKS_PER_CHUNK / 8] = {0};
+		int		chunk_offs = (blkno & (BLOCKS_PER_CHUNK - 1));
 		int		blocks_in_chunk = Min(nblocks, BLOCKS_PER_CHUNK - (blkno % BLOCKS_PER_CHUNK));
 		int		iteration_hits = 0;
 		int		iteration_misses = 0;
 		uint64	io_time_us = 0;
-		int     n_blocks_to_read = 0;
+		int		n_blocks_to_read = 0;
+		int		iov_last_used = 0;
+		int		first_block_in_chunk_read = -1;
 		ConditionVariable* cv;
 
 		Assert(blocks_in_chunk > 0);
 
 		for (int i = 0; i < blocks_in_chunk; i++)
 		{
-			n_blocks_to_read += (BITMAP_ISSET(mask, buf_offset + i) != 0);
-			iov[i].iov_base = buffers[buf_offset + i];
 			iov[i].iov_len = BLCKSZ;
-			BITMAP_CLR(mask,  buf_offset + i);
+			/* mask not set = we must do work */
+			if (!BITMAP_ISSET(mask, buf_offset + i))
+			{
+				iov[i].iov_base = buffers[buf_offset + i];
+				n_blocks_to_read++;
+				iov_last_used = i + 1;
+
+				if (first_block_in_chunk_read == -1)
+				{
+					first_block_in_chunk_read = i;
+				}
+			}
+			/* mask set = we must do no work */
+			else
+			{
+				/* don't scribble on pages we weren't requested to write to */
+				iov[i].iov_base = SCRIBBLEPAGE;
+			}
 		}
+
+		/* shortcut IO */
 		if (n_blocks_to_read == 0)
 		{
 			buf_offset += blocks_in_chunk;
@@ -718,6 +745,12 @@ lfc_readv_select(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 			continue;
 		}
 
+		/*
+		 * The effective iov size must be >= the number of blocks we're about
+		 * to read.
+		 */
+		Assert(iov_last_used - first_block_in_chunk_read >= n_blocks_to_read);
+
 		tag.blockNum = blkno - chunk_offs;
 		hash = get_hash_value(lfc_hash, &tag);
 		cv = &lfc_ctl->cv[hash % N_COND_VARS];
@@ -762,10 +795,15 @@ lfc_readv_select(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 		generation = lfc_ctl->generation;
 		entry_offset = entry->offset;
 
-		for (int i = 0; i < blocks_in_chunk; i++)
+		for (int i = first_block_in_chunk_read; i < iov_last_used; i++)
 		{
 			FileCacheBlockState state = UNAVAILABLE;
 			bool sleeping = false;
+
+			/* no need to work on something we're not interested in */
+			if (BITMAP_ISSET(mask, buf_offset + i))
+				continue;
+
 			while (lfc_ctl->generation == generation)
 			{
 				state = GET_STATE(entry, chunk_offs + i);
@@ -789,7 +827,7 @@ lfc_readv_select(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 			}
 			if (state == AVAILABLE)
 			{
-				BITMAP_SET(mask, buf_offset + i);
+				BITMAP_SET(chunk_mask, i);
 				iteration_hits++;
 			}
 			else
@@ -801,16 +839,34 @@ lfc_readv_select(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 
 		if (iteration_hits != 0)
 		{
+			/* chunk offset (# of pages) into the LFC file */
+			off_t	first_read_offset = (off_t) entry_offset * BLOCKS_PER_CHUNK;
+			int		nwrite = iov_last_used - first_block_in_chunk_read;
+			/* offset of first IOV */
+			first_read_offset += chunk_offs + first_block_in_chunk_read;
+
 			pgstat_report_wait_start(WAIT_EVENT_NEON_LFC_READ);
-			rc = preadv(lfc_desc, iov, blocks_in_chunk,
-						((off_t) entry_offset * BLOCKS_PER_CHUNK + chunk_offs) * BLCKSZ);
+
+			/* Read only the blocks we're interested in, limiting */
+			rc = preadv(lfc_desc, &iov[first_block_in_chunk_read],
+						nwrite, first_read_offset * BLCKSZ);
 			pgstat_report_wait_end();
 
-			if (rc != (BLCKSZ * blocks_in_chunk))
+			if (rc != (BLCKSZ * nwrite))
 			{
 				lfc_disable("read");
 				return -1;
 			}
+
+			/*
+			 * We successfully read the pages we know were valid when we
+			 * started reading; now mark those pages as read
+			 */
+			for (int i = first_block_in_chunk_read; i < iov_last_used; i++)
+			{
+				if (BITMAP_ISSET(chunk_mask, i))
+					BITMAP_SET(mask, buf_offset + i);
+			}
 		}
 
 		/* Place entry to the head of LRU list */

pgxn/neon/libpagestore.c

@@ -82,7 +82,8 @@ static int	max_reconnect_attempts = 60;
 static int	stripe_size;
 
 static int pageserver_response_log_timeout = 10000;
-static int pageserver_response_disconnect_timeout = 120000; /* 2 minutes */
+/* 2.5 minutes. A bit higher than highest default TCP retransmission timeout */
+static int pageserver_response_disconnect_timeout = 150000;
 
 typedef struct
 {
@@ -1141,37 +1142,23 @@ pageserver_try_receive(shardno_t shard_no)
 	NeonResponse *resp;
 	PageServer *shard = &page_servers[shard_no];
 	PGconn	   *pageserver_conn = shard->conn;
-	/* read response */
-	int			rc;
+	int	rc;
 
 	if (shard->state != PS_Connected)
 		return NULL;
 
 	Assert(pageserver_conn);
 
-	while (true)
+	rc = PQgetCopyData(shard->conn, &resp_buff.data, 1 /* async */);
+	if (rc == 0)
 	{
-		if (PQisBusy(shard->conn))
+		if (!PQconsumeInput(shard->conn))
 		{
-			WaitEvent	event;
-			if (WaitEventSetWait(shard->wes_read, 0, &event, 1,
-								 WAIT_EVENT_NEON_PS_READ) != 1
-				|| (event.events & WL_SOCKET_READABLE) == 0)
-			{
-				return NULL;
-			}
+			return NULL;
 		}
 		rc = PQgetCopyData(shard->conn, &resp_buff.data, 1 /* async */);
-		if (rc == 0)
-		{
-			if (!PQconsumeInput(shard->conn))
-			{
-				return NULL;
-			}
-		}
-		else
-			break;
 	}
+
 	if (rc == 0)
 		return NULL;
 	else if (rc > 0)
@@ -1450,7 +1437,7 @@ pg_init_libpagestore(void)
 							"If the pageserver doesn't respond to a request within this timeout, "
 							"disconnect and reconnect.",
 							&pageserver_response_disconnect_timeout,
-							120000, 100, INT_MAX,
+							150000, 100, INT_MAX,
 							PGC_SUSET,
 							GUC_UNIT_MS,
 							NULL, NULL, NULL);

pgxn/neon/pagestore_client.h

@@ -315,7 +315,7 @@ static inline bool
 lfc_read(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 		 void *buffer)
 {
-	bits8		rv = 1;
+	bits8		rv = 0;
 	return lfc_readv_select(rinfo, forkNum, blkno, &buffer, 1, &rv) == 1;
 }
 

pgxn/neon/pagestore_smgr.c

@@ -99,7 +99,7 @@ static char *hexdump_page(char *page);
 
 #define IS_LOCAL_REL(reln) (\
 	NInfoGetDbOid(InfoFromSMgrRel(reln)) != 0 && \
-		NInfoGetRelNumber(InfoFromSMgrRel(reln)) > FirstNormalObjectId \
+		NInfoGetRelNumber(InfoFromSMgrRel(reln)) >= FirstNormalObjectId \
 )
 
 const int	SmgrTrace = DEBUG5;
@@ -1081,6 +1081,9 @@ prefetch_lookup(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkn, neon_r
  * pageserver. If NULL, we utilize the lastWrittenLsn -infrastructure
  * to calculate the LSNs to send.
  *
+ * Bits set in *mask (if present) indicate pages already read; i.e. pages we
+ * can skip in this process.
+ *
  * When performing a prefetch rather than a synchronous request,
  * is_prefetch==true. Currently, it only affects how the request is accounted
  * in the perf counters.
@@ -1126,7 +1129,7 @@ Retry:
 		uint64		ring_index;
 		neon_request_lsns *lsns;
 
-		if (PointerIsValid(mask) && !BITMAP_ISSET(mask, i))
+		if (PointerIsValid(mask) && BITMAP_ISSET(mask, i))
 			continue;
 
 		if (frlsns)
@@ -2381,7 +2384,6 @@ neon_get_request_lsns(NRelFileInfo rinfo, ForkNumber forknum, BlockNumber blkno,
 						 LSN_FORMAT_ARGS(last_written_lsn),
 						 LSN_FORMAT_ARGS(flushlsn));
 				XLogFlush(last_written_lsn);
-				flushlsn = last_written_lsn;
 			}
 
 			/*
@@ -2397,18 +2399,35 @@ neon_get_request_lsns(NRelFileInfo rinfo, ForkNumber forknum, BlockNumber blkno,
 			 * requesting the latest page, by setting request LSN to
 			 * UINT64_MAX.
 			 *
-			 * Remember the current LSN, however, so that we can later
-			 * correctly determine if the response to the request is still
-			 * valid. The most up-to-date LSN we could use for that purpose
-			 * would be the current insert LSN, but to avoid the overhead of
-			 * looking it up, use 'flushlsn' instead. This relies on the
-			 * assumption that if the page was modified since the last WAL
-			 * flush, it should still be in the buffer cache, and we
-			 * wouldn't be requesting it.
+			 * effective_request_lsn is used to check that received response is still valid.
+			 * In case of primary node it is last written LSN. Originally we used flush_lsn here,
+			 * but it is not correct. Consider the following scenario:
+			 * 1. Backend A wants to prefetch block X
+			 * 2. Backend A checks that block X is not present in the shared buffer cache
+			 * 3. Backend A calls prefetch_do_request, which calls neon_get_request_lsns
+			 * 4. neon_get_request_lsns obtains LwLSN=11 for the block
+			 * 5. Backend B downloads block X, updates and wallogs it with LSN=13
+			 * 6. Block X is once again evicted from shared buffers, its LwLSN is set to LSN=13
+			 * 7. Backend A is still executing in neon_get_request_lsns(). It calls 'flushlsn = GetFlushRecPtr();'.
+			 *    Let's say that it is LSN=14
+			 * 8. Backend A uses LSN=14 as effective_lsn in the prefetch slot. The request stored in the slot is
+			 *    [not_modified_since=11, effective_request_lsn=14]
+			 * 9. Backend A sends the prefetch request, pageserver processes it, and sends response.
+			 *    The last LSN that the pageserver had processed was LSN=12, so the page image in the response is valid at LSN=12.
+			 * 10. Backend A calls smgrread() for page X with LwLSN=13
+			 * 11. Backend A finds in prefetch ring the response for the prefetch request with [not_modified_since=11, effective_lsn=Lsn14],
+			 * so it satisfies neon_prefetch_response_usable condition.
+			 *
+			 * Things go wrong in step 7-8, when [not_modified_since=11, effective_request_lsn=14] is determined for the request.
+			 * That is incorrect, because the page has in fact been modified at LSN=13. The invariant is that for any request,
+			 * there should not be any modifications to a page between its not_modified_since and (effective_)request_lsn values.
+			 *
+			 * The problem can be fixed by callingGetFlushRecPtr() before checking if the page is in the buffer cache.
+			 * But you can't do that within smgrprefetch(), would need to modify the caller.
 			 */
 			result->request_lsn = UINT64_MAX;
 			result->not_modified_since = last_written_lsn;
-			result->effective_request_lsn = flushlsn;
+			result->effective_request_lsn = last_written_lsn;
 		}
 	}
 }
@@ -2467,11 +2486,8 @@ neon_prefetch_response_usable(neon_request_lsns *request_lsns,
 	 * `not_modified_since` and `request_lsn` are sent to the pageserver, but
 	 * in the primary node, we always use UINT64_MAX as the `request_lsn`, so
 	 * we remember `effective_request_lsn` separately. In a primary,
-	 * `effective_request_lsn` is the last flush WAL position when the request
-	 * was sent to the pageserver. That's logically the LSN that we are
-	 * requesting the page at, but we send UINT64_MAX to the pageserver so
-	 * that if the GC horizon advances past that position, we still get a
-	 * valid response instead of an error.
+	 * `effective_request_lsn` is the same as  `not_modified_since`.
+	 * See comments in neon_get_request_lsns why we can not use last flush WAL position here.
 	 *
 	 * To determine whether a response to a GetPage request issued earlier is
 	 * still valid to satisfy a new page read, we look at the
@@ -3026,9 +3042,6 @@ neon_prefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
 
 		tag.blockNum = blocknum;
 
-		for (int i = 0; i < PG_IOV_MAX / 8; i++)
-			lfc_present[i] = ~(lfc_present[i]);
-
 		ring_index = prefetch_register_bufferv(tag, NULL, iterblocks,
 											   lfc_present, true);
 
@@ -3134,6 +3147,15 @@ neon_writeback(SMgrRelation reln, ForkNumber forknum,
 #endif
 }
 
+/*
+ * Read N pages at a specific LSN.
+ *
+ * *mask is set for pages read at a previous point in time, and which we
+ * should not touch, nor overwrite.
+ * New bits should be set in *mask for the pages we'successfully read.
+ *
+ * The offsets in request_lsns, buffers, and mask are linked.
+ */
 static void
 #if PG_MAJORVERSION_NUM < 16
 neon_read_at_lsnv(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber base_blockno, neon_request_lsns *request_lsns,
@@ -3186,7 +3208,7 @@ neon_read_at_lsnv(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber base_block
 		neon_request_lsns *reqlsns = &request_lsns[i];
 		TimestampTz		start_ts, end_ts;
 
-		if (PointerIsValid(mask) && !BITMAP_ISSET(mask, i))
+		if (PointerIsValid(mask) && BITMAP_ISSET(mask, i))
 			continue;
 
 		start_ts = GetCurrentTimestamp();
@@ -3485,9 +3507,7 @@ static void
 neon_readv(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
 		   void **buffers, BlockNumber nblocks)
 {
-	bits8		prefetch_hits[PG_IOV_MAX / 8] = {0};
-	bits8		lfc_hits[PG_IOV_MAX / 8];
-	bits8		read[PG_IOV_MAX / 8];
+	bits8		read_pages[PG_IOV_MAX / 8];
 	neon_request_lsns request_lsns[PG_IOV_MAX];
 	int			lfc_result;
 	int			prefetch_result;
@@ -3519,19 +3539,18 @@ neon_readv(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
 	neon_get_request_lsns(InfoFromSMgrRel(reln), forknum, blocknum,
 						  request_lsns, nblocks);
 
+	memset(read_pages, 0, sizeof(read_pages));
 
-	prefetch_result = prefetch_lookupv(InfoFromSMgrRel(reln), forknum, blocknum, request_lsns, nblocks, buffers, prefetch_hits);
+	prefetch_result = prefetch_lookupv(InfoFromSMgrRel(reln), forknum,
+									   blocknum, request_lsns, nblocks,
+									   buffers, read_pages);
 
 	if (prefetch_result == nblocks)
 		return;
 
-	/* invert the result: exclude prefetched blocks */
-	for (int i = 0; i < PG_IOV_MAX / 8; i++)
-		lfc_hits[i] = ~prefetch_hits[i];
-
 	/* Try to read from local file cache */
 	lfc_result = lfc_readv_select(InfoFromSMgrRel(reln), forknum, blocknum, buffers,
-								  nblocks, lfc_hits);
+								  nblocks, read_pages);
 
 	if (lfc_result > 0)
 		MyNeonCounters->file_cache_hits_total += lfc_result;
@@ -3540,21 +3559,8 @@ neon_readv(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
 	if (prefetch_result + lfc_result == nblocks)
 		return;
 
-	if (lfc_result <= 0)
-	{
-		/* can't use the LFC result, so read all blocks from PS */
-		for (int i = 0; i < PG_IOV_MAX / 8; i++)
-			read[i] = ~prefetch_hits[i];
-	}
-	else
-	{
-		/* invert the result: exclude blocks read from lfc */
-		for (int i = 0; i < PG_IOV_MAX / 8; i++)
-			read[i] = ~(prefetch_hits[i] | lfc_hits[i]);
-	}
-
 	neon_read_at_lsnv(InfoFromSMgrRel(reln), forknum, blocknum, request_lsns,
-					  buffers, nblocks, read);
+					  buffers, nblocks, read_pages);
 
 	/*
 	 * Try to receive prefetch results once again just to make sure we don't leave the smgr code while the OS might still have buffered bytes.

poetry.lock

@@ -3111,30 +3111,30 @@ six = "*"
 
 [[package]]
 name = "ruff"
-version = "0.7.0"
+version = "0.11.2"
 description = "An extremely fast Python linter and code formatter, written in Rust."
 optional = false
 python-versions = ">=3.7"
 groups = ["dev"]
 files = [
-    {file = "ruff-0.7.0-py3-none-linux_armv6l.whl", hash = "sha256:0cdf20c2b6ff98e37df47b2b0bd3a34aaa155f59a11182c1303cce79be715628"},
-    {file = "ruff-0.7.0-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:496494d350c7fdeb36ca4ef1c9f21d80d182423718782222c29b3e72b3512737"},
-    {file = "ruff-0.7.0-py3-none-macosx_11_0_arm64.whl", hash = "sha256:214b88498684e20b6b2b8852c01d50f0651f3cc6118dfa113b4def9f14faaf06"},
-    {file = "ruff-0.7.0-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:630fce3fefe9844e91ea5bbf7ceadab4f9981f42b704fae011bb8efcaf5d84be"},
-    {file = "ruff-0.7.0-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:211d877674e9373d4bb0f1c80f97a0201c61bcd1e9d045b6e9726adc42c156aa"},
-    {file = "ruff-0.7.0-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:194d6c46c98c73949a106425ed40a576f52291c12bc21399eb8f13a0f7073495"},
-    {file = "ruff-0.7.0-py3-none-manylinux_2_17_ppc64.manylinux2014_ppc64.whl", hash = "sha256:82c2579b82b9973a110fab281860403b397c08c403de92de19568f32f7178598"},
-    {file = "ruff-0.7.0-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:9af971fe85dcd5eaed8f585ddbc6bdbe8c217fb8fcf510ea6bca5bdfff56040e"},
-    {file = "ruff-0.7.0-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:b641c7f16939b7d24b7bfc0be4102c56562a18281f84f635604e8a6989948914"},
-    {file = "ruff-0.7.0-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:d71672336e46b34e0c90a790afeac8a31954fd42872c1f6adaea1dff76fd44f9"},
-    {file = "ruff-0.7.0-py3-none-musllinux_1_2_aarch64.whl", hash = "sha256:ab7d98c7eed355166f367597e513a6c82408df4181a937628dbec79abb2a1fe4"},
-    {file = "ruff-0.7.0-py3-none-musllinux_1_2_armv7l.whl", hash = "sha256:1eb54986f770f49edb14f71d33312d79e00e629a57387382200b1ef12d6a4ef9"},
-    {file = "ruff-0.7.0-py3-none-musllinux_1_2_i686.whl", hash = "sha256:dc452ba6f2bb9cf8726a84aa877061a2462afe9ae0ea1d411c53d226661c601d"},
-    {file = "ruff-0.7.0-py3-none-musllinux_1_2_x86_64.whl", hash = "sha256:4b406c2dce5be9bad59f2de26139a86017a517e6bcd2688da515481c05a2cb11"},
-    {file = "ruff-0.7.0-py3-none-win32.whl", hash = "sha256:f6c968509f767776f524a8430426539587d5ec5c662f6addb6aa25bc2e8195ec"},
-    {file = "ruff-0.7.0-py3-none-win_amd64.whl", hash = "sha256:ff4aabfbaaba880e85d394603b9e75d32b0693152e16fa659a3064a85df7fce2"},
-    {file = "ruff-0.7.0-py3-none-win_arm64.whl", hash = "sha256:10842f69c245e78d6adec7e1db0a7d9ddc2fff0621d730e61657b64fa36f207e"},
-    {file = "ruff-0.7.0.tar.gz", hash = "sha256:47a86360cf62d9cd53ebfb0b5eb0e882193fc191c6d717e8bef4462bc3b9ea2b"},
+    {file = "ruff-0.11.2-py3-none-linux_armv6l.whl", hash = "sha256:c69e20ea49e973f3afec2c06376eb56045709f0212615c1adb0eda35e8a4e477"},
+    {file = "ruff-0.11.2-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:2c5424cc1c4eb1d8ecabe6d4f1b70470b4f24a0c0171356290b1953ad8f0e272"},
+    {file = "ruff-0.11.2-py3-none-macosx_11_0_arm64.whl", hash = "sha256:ecf20854cc73f42171eedb66f006a43d0a21bfb98a2523a809931cda569552d9"},
+    {file = "ruff-0.11.2-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0c543bf65d5d27240321604cee0633a70c6c25c9a2f2492efa9f6d4b8e4199bb"},
+    {file = "ruff-0.11.2-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:20967168cc21195db5830b9224be0e964cc9c8ecf3b5a9e3ce19876e8d3a96e3"},
+    {file = "ruff-0.11.2-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:955a9ce63483999d9f0b8f0b4a3ad669e53484232853054cc8b9d51ab4c5de74"},
+    {file = "ruff-0.11.2-py3-none-manylinux_2_17_ppc64.manylinux2014_ppc64.whl", hash = "sha256:86b3a27c38b8fce73bcd262b0de32e9a6801b76d52cdb3ae4c914515f0cef608"},
+    {file = "ruff-0.11.2-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:a3b66a03b248c9fcd9d64d445bafdf1589326bee6fc5c8e92d7562e58883e30f"},
+    {file = "ruff-0.11.2-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:0397c2672db015be5aa3d4dac54c69aa012429097ff219392c018e21f5085147"},
+    {file = "ruff-0.11.2-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:869bcf3f9abf6457fbe39b5a37333aa4eecc52a3b99c98827ccc371a8e5b6f1b"},
+    {file = "ruff-0.11.2-py3-none-musllinux_1_2_aarch64.whl", hash = "sha256:2a2b50ca35457ba785cd8c93ebbe529467594087b527a08d487cf0ee7b3087e9"},
+    {file = "ruff-0.11.2-py3-none-musllinux_1_2_armv7l.whl", hash = "sha256:7c69c74bf53ddcfbc22e6eb2f31211df7f65054bfc1f72288fc71e5f82db3eab"},
+    {file = "ruff-0.11.2-py3-none-musllinux_1_2_i686.whl", hash = "sha256:6e8fb75e14560f7cf53b15bbc55baf5ecbe373dd5f3aab96ff7aa7777edd7630"},
+    {file = "ruff-0.11.2-py3-none-musllinux_1_2_x86_64.whl", hash = "sha256:842a472d7b4d6f5924e9297aa38149e5dcb1e628773b70e6387ae2c97a63c58f"},
+    {file = "ruff-0.11.2-py3-none-win32.whl", hash = "sha256:aca01ccd0eb5eb7156b324cfaa088586f06a86d9e5314b0eb330cb48415097cc"},
+    {file = "ruff-0.11.2-py3-none-win_amd64.whl", hash = "sha256:3170150172a8f994136c0c66f494edf199a0bbea7a409f649e4bc8f4d7084080"},
+    {file = "ruff-0.11.2-py3-none-win_arm64.whl", hash = "sha256:52933095158ff328f4c77af3d74f0379e34fd52f175144cefc1b192e7ccd32b4"},
+    {file = "ruff-0.11.2.tar.gz", hash = "sha256:ec47591497d5a1050175bdf4e1a4e6272cddff7da88a2ad595e1e326041d8d94"},
 ]
 
 [[package]]
@@ -3844,4 +3844,4 @@ cffi = ["cffi (>=1.11)"]
 [metadata]
 lock-version = "2.1"
 python-versions = "^3.11"
-content-hash = "715fc8c896dcfa1b15054deeddcdec557ef93af91b26e1c8e4688fe4dbef5296"
+content-hash = "fb50cb6b291169dce3188560cdb31a14af95647318f8f0f0d718131dbaf1817a"

proxy/src/binary/proxy.rs

@@ -314,9 +314,9 @@ pub async fn run() -> anyhow::Result<()> {
             None => {
                 bail!("plain auth requires redis_notifications to be set");
             }
-            Some(url) => Some(
-                ConnectionWithCredentialsProvider::new_with_static_credentials(url.to_string()),
-            ),
+            Some(url) => {
+                Some(ConnectionWithCredentialsProvider::new_with_static_credentials(url.clone()))
+            }
         },
         ("irsa", _) => match (&args.redis_host, args.redis_port) {
             (Some(host), Some(port)) => Some(

proxy/src/control_plane/client/mock.rs

@@ -1,5 +1,6 @@
 //! Mock console backend which relies on a user-provided postgres instance.
 
+use std::io;
 use std::net::{IpAddr, Ipv4Addr};
 use std::str::FromStr;
 use std::sync::Arc;
@@ -22,7 +23,6 @@ use crate::control_plane::errors::{
 };
 use crate::control_plane::messages::MetricsAuxInfo;
 use crate::control_plane::{AccessBlockerFlags, AuthInfo, AuthSecret, CachedNodeInfo, NodeInfo};
-use crate::error::io_error;
 use crate::intern::RoleNameInt;
 use crate::types::{BranchId, EndpointId, ProjectId, RoleName};
 use crate::url::ApiUrl;
@@ -36,13 +36,13 @@ enum MockApiError {
 
 impl From<MockApiError> for ControlPlaneError {
     fn from(e: MockApiError) -> Self {
-        io_error(e).into()
+        io::Error::other(e).into()
     }
 }
 
 impl From<tokio_postgres::Error> for ControlPlaneError {
     fn from(e: tokio_postgres::Error) -> Self {
-        io_error(e).into()
+        io::Error::other(e).into()
     }
 }
 

proxy/src/control_plane/errors.rs

@@ -1,8 +1,10 @@
+use std::io;
+
 use thiserror::Error;
 
 use crate::control_plane::client::ApiLockError;
 use crate::control_plane::messages::{self, ControlPlaneErrorMessage, Reason};
-use crate::error::{ErrorKind, ReportableError, UserFacingError, io_error};
+use crate::error::{ErrorKind, ReportableError, UserFacingError};
 use crate::proxy::retry::CouldRetry;
 
 /// A go-to error message which doesn't leak any detail.
@@ -79,13 +81,13 @@ impl CouldRetry for ControlPlaneError {
 
 impl From<reqwest::Error> for ControlPlaneError {
     fn from(e: reqwest::Error) -> Self {
-        io_error(e).into()
+        io::Error::other(e).into()
     }
 }
 
 impl From<reqwest_middleware::Error> for ControlPlaneError {
     fn from(e: reqwest_middleware::Error) -> Self {
-        io_error(e).into()
+        io::Error::other(e).into()
     }
 }
 

proxy/src/error.rs

@@ -1,15 +1,9 @@
-use std::error::Error as StdError;
-use std::{fmt, io};
+use std::fmt;
 
 use anyhow::Context;
 use measured::FixedCardinalityLabel;
 use tokio::task::JoinError;
 
-/// Upcast (almost) any error into an opaque [`io::Error`].
-pub(crate) fn io_error(e: impl Into<Box<dyn StdError + Send + Sync>>) -> io::Error {
-    io::Error::new(io::ErrorKind::Other, e)
-}
-
 /// Marks errors that may be safely shown to a client.
 /// This trait can be seen as a specialized version of [`ToString`].
 ///

proxy/src/protocol2.rs

@@ -163,8 +163,7 @@ fn process_proxy_payload(
         // other values are unassigned and must not be emitted by senders. Receivers
         // must drop connections presenting unexpected values here.
         #[rustfmt::skip] // https://github.com/rust-lang/rustfmt/issues/6384
-        _ => return Err(io::Error::new(
-            io::ErrorKind::Other,
+        _ => return Err(io::Error::other(
             format!(
                 "invalid proxy protocol command 0x{:02X}. expected local (0x20) or proxy (0x21)",
                 header.version_and_command
@@ -178,21 +177,20 @@ fn process_proxy_payload(
         TCP_OVER_IPV4 | UDP_OVER_IPV4 => {
             let addr = payload
                 .try_get::<ProxyProtocolV2HeaderV4>()
-                .ok_or_else(|| io::Error::new(io::ErrorKind::Other, size_err))?;
+                .ok_or_else(|| io::Error::other(size_err))?;
 
             SocketAddr::from((addr.src_addr.get(), addr.src_port.get()))
         }
         TCP_OVER_IPV6 | UDP_OVER_IPV6 => {
             let addr = payload
                 .try_get::<ProxyProtocolV2HeaderV6>()
-                .ok_or_else(|| io::Error::new(io::ErrorKind::Other, size_err))?;
+                .ok_or_else(|| io::Error::other(size_err))?;
 
             SocketAddr::from((addr.src_addr.get(), addr.src_port.get()))
         }
         // unspecified or unix stream. ignore the addresses
         _ => {
-            return Err(io::Error::new(
-                io::ErrorKind::Other,
+            return Err(io::Error::other(
                 "invalid proxy protocol address family/transport protocol.",
             ));
         }

proxy/src/rate_limiter/limit_algorithm.rs

@@ -65,23 +65,13 @@ pub struct RateLimiterConfig {
     pub(crate) initial_limit: usize,
 }
 
-impl RateLimiterConfig {
-    pub(crate) fn create_rate_limit_algorithm(self) -> Box<dyn LimitAlgorithm> {
-        match self.algorithm {
-            RateLimitAlgorithm::Fixed => Box::new(Fixed),
-            RateLimitAlgorithm::Aimd { conf } => Box::new(conf),
-        }
-    }
-}
-
-pub(crate) struct LimiterInner {
-    alg: Box<dyn LimitAlgorithm>,
-    available: usize,
+pub(crate) struct LimiterInner<L: ?Sized> {
     limit: usize,
     in_flight: usize,
+    alg: L,
 }
 
-impl LimiterInner {
+impl<L: ?Sized + LimitAlgorithm> LimiterInner<L> {
     fn update_limit(&mut self, latency: Duration, outcome: Option<Outcome>) {
         if let Some(outcome) = outcome {
             let sample = Sample {
@@ -94,14 +84,13 @@ impl LimiterInner {
     }
 
     fn take(&mut self, ready: &Notify) -> Option<()> {
-        if self.available >= 1 {
-            self.available -= 1;
+        if self.in_flight < self.limit {
             self.in_flight += 1;
 
-            // tell the next in the queue that there is a permit ready
-            if self.available >= 1 {
+            if self.in_flight < self.limit {
                 ready.notify_one();
             }
+
             Some(())
         } else {
             None
@@ -116,11 +105,11 @@ impl LimiterInner {
 ///
 /// The limit will be automatically adjusted based on observed latency (delay) and/or failures
 /// caused by overload (loss).
-pub(crate) struct DynamicLimiter {
-    config: RateLimiterConfig,
-    inner: Mutex<LimiterInner>,
+pub(crate) struct DynamicLimiter<L: ?Sized = dyn LimitAlgorithm> {
+    disabled: bool,
     // to notify when a token is available
     ready: Notify,
+    inner: Mutex<LimiterInner<L>>,
 }
 
 /// A concurrency token, required to run a job.
@@ -140,37 +129,46 @@ struct LimiterState {
     limit: usize,
 }
 
-impl DynamicLimiter {
-    /// Create a limiter with a given limit control algorithm.
-    pub(crate) fn new(config: RateLimiterConfig) -> Arc<Self> {
+impl<L> DynamicLimiter<L> {
+    pub(crate) fn new_inner(initial_limit: usize, alg: L) -> Arc<Self> {
         let ready = Notify::new();
         ready.notify_one();
 
         Arc::new(Self {
+            disabled: initial_limit == 0,
             inner: Mutex::new(LimiterInner {
-                alg: config.create_rate_limit_algorithm(),
-                available: config.initial_limit,
-                limit: config.initial_limit,
+                limit: initial_limit,
                 in_flight: 0,
+                alg,
             }),
             ready,
-            config,
         })
     }
+}
+
+impl DynamicLimiter {
+    /// Create a limiter with a given limit control algorithm.
+    pub(crate) fn new(config: RateLimiterConfig) -> Arc<Self> {
+        let initial_limit = config.initial_limit;
+        match config.algorithm {
+            RateLimitAlgorithm::Fixed => DynamicLimiter::new_inner(initial_limit, Fixed),
+            RateLimitAlgorithm::Aimd { conf } => DynamicLimiter::new_inner(initial_limit, conf),
+        }
+    }
 
     /// Try to acquire a concurrency [Token], waiting for `duration` if there are none available.
     pub(crate) async fn acquire_timeout(
         self: &Arc<Self>,
         duration: Duration,
     ) -> Result<Token, Elapsed> {
-        tokio::time::timeout(duration, self.acquire()).await?
+        tokio::time::timeout(duration, self.acquire()).await
     }
 
     /// Try to acquire a concurrency [Token].
-    async fn acquire(self: &Arc<Self>) -> Result<Token, Elapsed> {
-        if self.config.initial_limit == 0 {
+    async fn acquire(self: &Arc<Self>) -> Token {
+        if self.disabled {
             // If the rate limiter is disabled, we can always acquire a token.
-            Ok(Token::disabled())
+            Token::disabled()
         } else {
             let mut notified = pin!(self.ready.notified());
             let mut ready = notified.as_mut().enable();
@@ -178,7 +176,7 @@ impl DynamicLimiter {
                 if ready {
                     let mut inner = self.inner.lock();
                     if inner.take(&self.ready).is_some() {
-                        break Ok(Token::new(self.clone()));
+                        break Token::new(self.clone());
                     }
                     notified.set(self.ready.notified());
                 }
@@ -200,7 +198,7 @@ impl DynamicLimiter {
         } else {
             tracing::debug!("outcome is {:?}", outcome);
         }
-        if self.config.initial_limit == 0 {
+        if self.disabled {
             return;
         }
 
@@ -210,7 +208,6 @@ impl DynamicLimiter {
 
         inner.in_flight -= 1;
         if inner.in_flight < inner.limit {
-            inner.available = inner.limit - inner.in_flight;
             // At least 1 permit is now available
             self.ready.notify_one();
         }

proxy/src/redis/connection_with_credentials_provider.rs

@@ -143,6 +143,8 @@ impl ConnectionWithCredentialsProvider {
                         db: 0,
                         username: Some(username),
                         password: Some(password.clone()),
+                        // TODO: switch to RESP3 after testing new client version.
+                        protocol: redis::ProtocolVersion::RESP2,
                     },
                 })
             }

proxy/src/serverless/json.rs

@@ -19,7 +19,7 @@ fn json_value_to_pg_text(value: &Value) -> Option<String> {
         v @ (Value::Bool(_) | Value::Number(_) | Value::Object(_)) => Some(v.to_string()),
 
         // avoid escaping here, as we pass this as a parameter
-        Value::String(s) => Some(s.to_string()),
+        Value::String(s) => Some(s.clone()),
 
         // special care for arrays
         Value::Array(_) => json_array_to_pg_array(value),

proxy/src/serverless/sql_over_http.rs

@@ -866,7 +866,7 @@ impl QueryData {
         let (inner, mut discard) = client.inner();
         let cancel_token = inner.cancel_token();
 
-        let res = match select(
+        match select(
             pin!(query_to_json(
                 config,
                 &mut *inner,
@@ -889,7 +889,7 @@ impl QueryData {
             // The query failed with an error
             Either::Left((Err(e), __not_yet_cancelled)) => {
                 discard.discard();
-                return Err(e);
+                Err(e)
             }
             // The query was cancelled.
             Either::Right((_cancelled, query)) => {
@@ -930,8 +930,7 @@ impl QueryData {
                     }
                 }
             }
-        };
-        res
+        }
     }
 }
 

proxy/src/serverless/websocket.rs

@@ -15,7 +15,7 @@ use tracing::warn;
 use crate::cancellation::CancellationHandler;
 use crate::config::ProxyConfig;
 use crate::context::RequestContext;
-use crate::error::{ReportableError, io_error};
+use crate::error::ReportableError;
 use crate::metrics::Metrics;
 use crate::proxy::{ClientMode, ErrorSource, handle_client};
 use crate::rate_limiter::EndpointRateLimiter;
@@ -50,23 +50,23 @@ impl<S: AsyncRead + AsyncWrite + Unpin> AsyncWrite for WebSocketRw<S> {
         let this = self.project();
         let mut stream = this.stream;
 
-        ready!(stream.as_mut().poll_ready(cx).map_err(io_error))?;
+        ready!(stream.as_mut().poll_ready(cx).map_err(io::Error::other))?;
 
         this.send.put(buf);
         match stream.as_mut().start_send(Frame::binary(this.send.split())) {
             Ok(()) => Poll::Ready(Ok(buf.len())),
-            Err(e) => Poll::Ready(Err(io_error(e))),
+            Err(e) => Poll::Ready(Err(io::Error::other(e))),
         }
     }
 
     fn poll_flush(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
         let stream = self.project().stream;
-        stream.poll_flush(cx).map_err(io_error)
+        stream.poll_flush(cx).map_err(io::Error::other)
     }
 
     fn poll_shutdown(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
         let stream = self.project().stream;
-        stream.poll_close(cx).map_err(io_error)
+        stream.poll_close(cx).map_err(io::Error::other)
     }
 }
 
@@ -97,7 +97,7 @@ impl<S: AsyncRead + AsyncWrite + Unpin> AsyncBufRead for WebSocketRw<S> {
             }
 
             let res = ready!(this.stream.as_mut().poll_next(cx));
-            match res.transpose().map_err(io_error)? {
+            match res.transpose().map_err(io::Error::other)? {
                 Some(message) => match message.opcode {
                     OpCode::Ping => {}
                     OpCode::Pong => {}
@@ -105,7 +105,7 @@ impl<S: AsyncRead + AsyncWrite + Unpin> AsyncBufRead for WebSocketRw<S> {
                         // We expect to see only binary messages.
                         let error = "unexpected text message in the websocket";
                         warn!(length = message.payload.len(), error);
-                        return Poll::Ready(Err(io_error(error)));
+                        return Poll::Ready(Err(io::Error::other(error)));
                     }
                     OpCode::Binary | OpCode::Continuation => {
                         debug_assert!(this.recv.is_empty());

proxy/src/tls/server_config.rs

@@ -173,7 +173,7 @@ impl CertResolver {
     }
 
     pub fn get_common_names(&self) -> HashSet<String> {
-        self.certs.keys().map(|s| s.to_string()).collect()
+        self.certs.keys().cloned().collect()
     }
 }
 

pyproject.toml

@@ -53,7 +53,7 @@ jsonnet = "^0.21.0-rc2"
 
 [tool.poetry.group.dev.dependencies]
 mypy = "==1.13.0"
-ruff = "^0.7.0"
+ruff = "^0.11.2"
 
 [build-system]
 requires = ["poetry-core>=1.0.0"]
@@ -109,4 +109,5 @@ select = [
     "W", # pycodestyle
     "B", # bugbear
     "UP", # pyupgrade
+    "TC", # flake8-type-checking
 ]

safekeeper/client/src/mgmt_api.rs

@@ -38,9 +38,8 @@ pub enum Error {
     #[error("Cancelled")]
     Cancelled,
 
-    /// Failed to create client.
-    #[error("create client: {0}{}", .0.source().map(|e| format!(": {e}")).unwrap_or_default())]
-    CreateClient(reqwest::Error),
+    #[error("request timed out: {0}")]
+    Timeout(String),
 }
 
 pub type Result<T> = std::result::Result<T, Error>;

safekeeper/src/bin/safekeeper.rs

@@ -217,7 +217,7 @@ struct Args {
     /// Period to reload certificate and private key from files.
     #[arg(long, value_parser = humantime::parse_duration, default_value = DEFAULT_SSL_CERT_RELOAD_PERIOD)]
     pub ssl_cert_reload_period: Duration,
-    /// Trusted root CA certificate to use in https APIs.
+    /// Trusted root CA certificates to use in https APIs.
     #[arg(long)]
     ssl_ca_file: Option<Utf8PathBuf>,
 }
@@ -353,13 +353,13 @@ async fn main() -> anyhow::Result<()> {
         }
     };
 
-    let ssl_ca_cert = match args.ssl_ca_file.as_ref() {
+    let ssl_ca_certs = match args.ssl_ca_file.as_ref() {
         Some(ssl_ca_file) => {
             tracing::info!("Using ssl root CA file: {ssl_ca_file:?}");
             let buf = tokio::fs::read(ssl_ca_file).await?;
-            Some(Certificate::from_pem(&buf)?)
+            Certificate::from_pem_bundle(&buf)?
         }
-        None => None,
+        None => Vec::new(),
     };
 
     let conf = Arc::new(SafeKeeperConf {
@@ -398,7 +398,7 @@ async fn main() -> anyhow::Result<()> {
         ssl_key_file: args.ssl_key_file,
         ssl_cert_file: args.ssl_cert_file,
         ssl_cert_reload_period: args.ssl_cert_reload_period,
-        ssl_ca_cert,
+        ssl_ca_certs,
     });
 
     // initialize sentry if SENTRY_DSN is provided

safekeeper/src/http/routes.rs

@@ -235,7 +235,7 @@ async fn timeline_pull_handler(mut request: Request<Body>) -> Result<Response<Bo
     let resp = pull_timeline::handle_request(
         data,
         conf.sk_auth_token.clone(),
-        conf.ssl_ca_cert.clone(),
+        conf.ssl_ca_certs.clone(),
         global_timelines,
     )
     .await

safekeeper/src/lib.rs

@@ -120,7 +120,7 @@ pub struct SafeKeeperConf {
     pub ssl_key_file: Utf8PathBuf,
     pub ssl_cert_file: Utf8PathBuf,
     pub ssl_cert_reload_period: Duration,
-    pub ssl_ca_cert: Option<Certificate>,
+    pub ssl_ca_certs: Vec<Certificate>,
 }
 
 impl SafeKeeperConf {
@@ -169,7 +169,7 @@ impl SafeKeeperConf {
             ssl_key_file: Utf8PathBuf::from(defaults::DEFAULT_SSL_KEY_FILE),
             ssl_cert_file: Utf8PathBuf::from(defaults::DEFAULT_SSL_CERT_FILE),
             ssl_cert_reload_period: Duration::from_secs(60),
-            ssl_ca_cert: None,
+            ssl_ca_certs: Vec::new(),
         }
     }
 }

safekeeper/src/pull_timeline.rs

@@ -393,7 +393,7 @@ pub struct DebugDumpResponse {
 pub async fn handle_request(
     request: PullTimelineRequest,
     sk_auth_token: Option<SecretString>,
-    ssl_ca_cert: Option<Certificate>,
+    ssl_ca_certs: Vec<Certificate>,
     global_timelines: Arc<GlobalTimelines>,
 ) -> Result<PullTimelineResponse> {
     let existing_tli = global_timelines.get(TenantTimelineId::new(
@@ -405,7 +405,7 @@ pub async fn handle_request(
     }
 
     let mut http_client = reqwest::Client::builder();
-    if let Some(ssl_ca_cert) = ssl_ca_cert {
+    for ssl_ca_cert in ssl_ca_certs {
         http_client = http_client.add_root_certificate(ssl_ca_cert);
     }
     let http_client = http_client.build()?;

safekeeper/tests/walproposer_sim/safekeeper.rs

@@ -183,7 +183,7 @@ pub fn run_server(os: NodeOs, disk: Arc<SafekeeperDisk>) -> Result<()> {
         ssl_key_file: Utf8PathBuf::from(""),
         ssl_cert_file: Utf8PathBuf::from(""),
         ssl_cert_reload_period: Duration::ZERO,
-        ssl_ca_cert: None,
+        ssl_ca_certs: Vec::new(),
     };
 
     let mut global = GlobalMap::new(disk, conf.clone())?;

scripts/download_basebackup.py

@@ -8,9 +8,12 @@
 from __future__ import annotations
 
 import argparse
+from typing import TYPE_CHECKING
 
 import psycopg2
-from psycopg2.extensions import connection as PgConnection
+
+if TYPE_CHECKING:
+    from psycopg2.extensions import connection as PgConnection
 
 
 def main(args: argparse.Namespace):

scripts/force_layer_download.py

@@ -7,13 +7,13 @@ import logging
 import signal
 import sys
 from collections import defaultdict
-from collections.abc import Awaitable
 from dataclasses import dataclass
 from typing import TYPE_CHECKING
 
 import aiohttp
 
 if TYPE_CHECKING:
+    from collections.abc import Awaitable
     from typing import Any
 
 

storage_controller/src/compute_hook.rs

@@ -624,16 +624,19 @@ impl ComputeHook {
             MaybeSendResult::Transmit((request, lock)) => (request, lock),
         };
 
-        let compute_hook_url = if let Some(control_plane_url) = &self.config.control_plane_url {
-            Some(if control_plane_url.ends_with('/') {
-                format!("{control_plane_url}notify-attach")
+        let result = if !self.config.use_local_compute_notifications {
+            let compute_hook_url = if let Some(control_plane_url) = &self.config.control_plane_url {
+                Some(if control_plane_url.ends_with('/') {
+                    format!("{control_plane_url}notify-attach")
+                } else {
+                    format!("{control_plane_url}/notify-attach")
+                })
             } else {
-                format!("{control_plane_url}/notify-attach")
-            })
-        } else {
-            self.config.compute_hook_url.clone()
-        };
-        let result = if let Some(notify_url) = &compute_hook_url {
+                self.config.compute_hook_url.clone()
+            };
+
+            // We validate this at startup
+            let notify_url = compute_hook_url.as_ref().unwrap();
             self.do_notify(notify_url, &request, cancel).await
         } else {
             self.do_notify_local(&request).await.map_err(|e| {

storage_controller/src/heartbeater.rs

@@ -8,7 +8,6 @@ use futures::StreamExt;
 use futures::stream::FuturesUnordered;
 use pageserver_api::controller_api::{NodeAvailability, SkSchedulingPolicy};
 use pageserver_api::models::PageserverUtilization;
-use reqwest::Certificate;
 use safekeeper_api::models::SafekeeperUtilization;
 use safekeeper_client::mgmt_api;
 use thiserror::Error;
@@ -27,8 +26,8 @@ struct HeartbeaterTask<Server, State> {
 
     max_offline_interval: Duration,
     max_warming_up_interval: Duration,
+    http_client: reqwest::Client,
     jwt_token: Option<String>,
-    ssl_ca_cert: Option<Certificate>,
 }
 
 #[derive(Debug, Clone)]
@@ -76,8 +75,8 @@ where
     HeartbeaterTask<Server, State>: HeartBeat<Server, State>,
 {
     pub(crate) fn new(
+        http_client: reqwest::Client,
         jwt_token: Option<String>,
-        ssl_ca_cert: Option<Certificate>,
         max_offline_interval: Duration,
         max_warming_up_interval: Duration,
         cancel: CancellationToken,
@@ -86,8 +85,8 @@ where
             tokio::sync::mpsc::unbounded_channel::<HeartbeatRequest<Server, State>>();
         let mut heartbeater = HeartbeaterTask::new(
             receiver,
+            http_client,
             jwt_token,
-            ssl_ca_cert,
             max_offline_interval,
             max_warming_up_interval,
             cancel,
@@ -122,8 +121,8 @@ where
 {
     fn new(
         receiver: tokio::sync::mpsc::UnboundedReceiver<HeartbeatRequest<Server, State>>,
+        http_client: reqwest::Client,
         jwt_token: Option<String>,
-        ssl_ca_cert: Option<Certificate>,
         max_offline_interval: Duration,
         max_warming_up_interval: Duration,
         cancel: CancellationToken,
@@ -134,8 +133,8 @@ where
             state: HashMap::new(),
             max_offline_interval,
             max_warming_up_interval,
+            http_client,
             jwt_token,
-            ssl_ca_cert,
         }
     }
     async fn run(&mut self) {
@@ -178,7 +177,7 @@ impl HeartBeat<Node, PageserverState> for HeartbeaterTask<Node, PageserverState>
         let mut heartbeat_futs = FuturesUnordered::new();
         for (node_id, node) in &*pageservers {
             heartbeat_futs.push({
-                let ssl_ca_cert = self.ssl_ca_cert.clone();
+                let http_client = self.http_client.clone();
                 let jwt_token = self.jwt_token.clone();
                 let cancel = self.cancel.clone();
 
@@ -193,8 +192,8 @@ impl HeartBeat<Node, PageserverState> for HeartbeaterTask<Node, PageserverState>
                     let response = node_clone
                         .with_client_retries(
                             |client| async move { client.get_utilization().await },
+                            &http_client,
                             &jwt_token,
-                            &ssl_ca_cert,
                             3,
                             3,
                             Duration::from_secs(1),
@@ -329,19 +328,19 @@ impl HeartBeat<Safekeeper, SafekeeperState> for HeartbeaterTask<Safekeeper, Safe
                 continue;
             }
             heartbeat_futs.push({
+                let http_client = self.http_client.clone();
                 let jwt_token = self
                     .jwt_token
                     .as_ref()
                     .map(|t| SecretString::from(t.to_owned()));
-                let ssl_ca_cert = self.ssl_ca_cert.clone();
                 let cancel = self.cancel.clone();
 
                 async move {
                     let response = sk
                         .with_client_retries(
                             |client| async move { client.get_utilization().await },
+                            &http_client,
                             &jwt_token,
-                            &ssl_ca_cert,
                             3,
                             3,
                             Duration::from_secs(1),

storage_controller/src/http.rs

@@ -24,9 +24,9 @@ use pageserver_api::controller_api::{
     ShardsPreferredAzsRequest, TenantCreateRequest, TenantPolicyRequest, TenantShardMigrateRequest,
 };
 use pageserver_api::models::{
-    DetachBehavior, TenantConfigPatchRequest, TenantConfigRequest, TenantLocationConfigRequest,
-    TenantShardSplitRequest, TenantTimeTravelRequest, TimelineArchivalConfigRequest,
-    TimelineCreateRequest,
+    DetachBehavior, LsnLeaseRequest, TenantConfigPatchRequest, TenantConfigRequest,
+    TenantLocationConfigRequest, TenantShardSplitRequest, TenantTimeTravelRequest,
+    TimelineArchivalConfigRequest, TimelineCreateRequest,
 };
 use pageserver_api::shard::TenantShardId;
 use pageserver_api::upcall_api::{ReAttachRequest, ValidateRequest};
@@ -582,6 +582,32 @@ async fn handle_tenant_timeline_download_heatmap_layers(
     json_response(StatusCode::OK, ())
 }
 
+async fn handle_tenant_timeline_lsn_lease(
+    service: Arc<Service>,
+    req: Request<Body>,
+) -> Result<Response<Body>, ApiError> {
+    let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
+    let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
+
+    check_permissions(&req, Scope::PageServerApi)?;
+    maybe_rate_limit(&req, tenant_id).await;
+
+    let mut req = match maybe_forward(req).await {
+        ForwardOutcome::Forwarded(res) => {
+            return res;
+        }
+        ForwardOutcome::NotForwarded(req) => req,
+    };
+
+    let lsn_lease_request = json_request::<LsnLeaseRequest>(&mut req).await?;
+
+    service
+        .tenant_timeline_lsn_lease(tenant_id, timeline_id, lsn_lease_request.lsn)
+        .await?;
+
+    json_response(StatusCode::OK, ())
+}
+
 // For metric labels where we would like to include the approximate path, but exclude high-cardinality fields like query parameters
 // and tenant/timeline IDs.  Since we are proxying to arbitrary paths, we don't have routing templates to
 // compare to, so we can just filter out our well known ID format with regexes.
@@ -656,11 +682,10 @@ async fn handle_tenant_timeline_passthrough(
     let _timer = latency.start_timer(labels.clone());
 
     let client = mgmt_api::Client::new(
+        service.get_http_client().clone(),
         node.base_url(),
         service.get_config().pageserver_jwt_token.as_deref(),
-        service.get_config().ssl_ca_cert.clone(),
-    )
-    .map_err(|e| ApiError::InternalServerError(anyhow::anyhow!(e)))?;
+    );
     let resp = client.get_raw(path).await.map_err(|e|
         // We return 503 here because if we can't successfully send a request to the pageserver,
         // either we aren't available or the pageserver is unavailable.
@@ -2193,6 +2218,17 @@ pub fn make_router(
                 )
             },
         )
+        // LSN lease passthrough to all shards
+        .post(
+            "/v1/tenant/:tenant_id/timeline/:timeline_id/lsn_lease",
+            |r| {
+                tenant_service_handler(
+                    r,
+                    handle_tenant_timeline_lsn_lease,
+                    RequestName("v1_tenant_timeline_lsn_lease"),
+                )
+            },
+        )
         // Tenant detail GET passthrough to shard zero:
         .get("/v1/tenant/:tenant_id", |r| {
             tenant_service_handler(

storage_controller/src/main.rs

@@ -200,9 +200,14 @@ struct Cli {
     /// Period to reload certificate and private key from files.
     #[arg(long, default_value = DEFAULT_SSL_CERT_RELOAD_PERIOD)]
     ssl_cert_reload_period: humantime::Duration,
-    /// Trusted root CA certificate to use in https APIs.
+    /// Trusted root CA certificates to use in https APIs.
     #[arg(long)]
     ssl_ca_file: Option<PathBuf>,
+
+    /// Neon local specific flag. When set, ignore [`Cli::control_plane_url`] and deliver
+    /// the compute notification directly (instead of via control plane).
+    #[arg(long, default_value = "false")]
+    use_local_compute_notifications: bool,
 }
 
 enum StrictMode {
@@ -368,6 +373,9 @@ async fn async_main() -> anyhow::Result<()> {
                 "neither `--compute-hook-url` nor `--control-plane-url` are set: this is only permitted in `--dev` mode"
             );
         }
+        StrictMode::Strict if args.use_local_compute_notifications => {
+            anyhow::bail!("`--use-local-compute-notifications` is only permitted in `--dev` mode");
+        }
         StrictMode::Strict => {
             tracing::info!("Starting in strict mode: configuration is OK.")
         }
@@ -376,13 +384,13 @@ async fn async_main() -> anyhow::Result<()> {
         }
     }
 
-    let ssl_ca_cert = match args.ssl_ca_file.as_ref() {
+    let ssl_ca_certs = match args.ssl_ca_file.as_ref() {
         Some(ssl_ca_file) => {
             tracing::info!("Using ssl root CA file: {ssl_ca_file:?}");
             let buf = tokio::fs::read(ssl_ca_file).await?;
-            Some(Certificate::from_pem(&buf)?)
+            Certificate::from_pem_bundle(&buf)?
         }
-        None => None,
+        None => Vec::new(),
     };
 
     let config = Config {
@@ -425,8 +433,9 @@ async fn async_main() -> anyhow::Result<()> {
         start_as_candidate: args.start_as_candidate,
         use_https_pageserver_api: args.use_https_pageserver_api,
         use_https_safekeeper_api: args.use_https_safekeeper_api,
-        ssl_ca_cert,
+        ssl_ca_certs,
         timelines_onto_safekeepers: args.timelines_onto_safekeepers,
+        use_local_compute_notifications: args.use_local_compute_notifications,
     };
 
     // Validate that we can connect to the database

storage_controller/src/node.rs

@@ -7,7 +7,7 @@ use pageserver_api::controller_api::{
 };
 use pageserver_api::shard::TenantShardId;
 use pageserver_client::mgmt_api;
-use reqwest::{Certificate, StatusCode};
+use reqwest::StatusCode;
 use serde::Serialize;
 use tokio_util::sync::CancellationToken;
 use utils::backoff;
@@ -280,8 +280,8 @@ impl Node {
     pub(crate) async fn with_client_retries<T, O, F>(
         &self,
         mut op: O,
+        http_client: &reqwest::Client,
         jwt: &Option<String>,
-        ssl_ca_cert: &Option<Certificate>,
         warn_threshold: u32,
         max_retries: u32,
         timeout: Duration,
@@ -300,24 +300,13 @@ impl Node {
                 | ApiError(StatusCode::REQUEST_TIMEOUT, _) => false,
                 ApiError(_, _) => true,
                 Cancelled => true,
-                CreateClient(_) => true,
+                Timeout(_) => false,
             }
         }
 
-        // TODO: refactor PageserverClient and with_client_retires (#11113).
-        let mut http_client = reqwest::ClientBuilder::new().timeout(timeout);
-        if let Some(ssl_ca_cert) = ssl_ca_cert.as_ref() {
-            http_client = http_client.add_root_certificate(ssl_ca_cert.clone())
-        }
-
-        let http_client = match http_client.build() {
-            Ok(http_client) => http_client,
-            Err(err) => return Some(Err(mgmt_api::Error::CreateClient(err))),
-        };
-
         backoff::retry(
             || {
-                let client = PageserverClient::from_client(
+                let client = PageserverClient::new(
                     self.get_id(),
                     http_client.clone(),
                     self.base_url(),
@@ -326,11 +315,14 @@ impl Node {
 
                 let node_cancel_fut = self.cancel.cancelled();
 
-                let op_fut = op(client);
+                let op_fut = tokio::time::timeout(timeout, op(client));
 
                 async {
                     tokio::select! {
-                        r = op_fut=> {r},
+                        r = op_fut => match r {
+                            Ok(r) => r,
+                            Err(e) => Err(mgmt_api::Error::Timeout(format!("{e}"))),
+                        },
                         _ = node_cancel_fut => {
                         Err(mgmt_api::Error::Cancelled)
                     }}

storage_controller/src/pageserver_client.rs

@@ -1,6 +1,6 @@
 use pageserver_api::models::detach_ancestor::AncestorDetached;
 use pageserver_api::models::{
-    DetachBehavior, LocationConfig, LocationConfigListResponse, PageserverUtilization,
+    DetachBehavior, LocationConfig, LocationConfigListResponse, LsnLease, PageserverUtilization,
     SecondaryProgress, TenantScanRemoteStorageResponse, TenantShardSplitRequest,
     TenantShardSplitResponse, TenantWaitLsnRequest, TimelineArchivalConfigRequest,
     TimelineCreateRequest, TimelineInfo, TopTenantShardsRequest, TopTenantShardsResponse,
@@ -8,8 +8,9 @@ use pageserver_api::models::{
 use pageserver_api::shard::TenantShardId;
 use pageserver_client::BlockUnblock;
 use pageserver_client::mgmt_api::{Client, Result};
-use reqwest::{Certificate, StatusCode};
+use reqwest::StatusCode;
 use utils::id::{NodeId, TenantId, TimelineId};
+use utils::lsn::Lsn;
 
 /// Thin wrapper around [`pageserver_client::mgmt_api::Client`]. It allows the storage
 /// controller to collect metrics in a non-intrusive manner.
@@ -47,25 +48,13 @@ macro_rules! measured_request {
 
 impl PageserverClient {
     pub(crate) fn new(
-        node_id: NodeId,
-        mgmt_api_endpoint: String,
-        jwt: Option<&str>,
-        ssl_ca_cert: Option<Certificate>,
-    ) -> Result<Self> {
-        Ok(Self {
-            inner: Client::new(mgmt_api_endpoint, jwt, ssl_ca_cert)?,
-            node_id_label: node_id.0.to_string(),
-        })
-    }
-
-    pub(crate) fn from_client(
         node_id: NodeId,
         raw_client: reqwest::Client,
         mgmt_api_endpoint: String,
         jwt: Option<&str>,
     ) -> Self {
         Self {
-            inner: Client::from_client(raw_client, mgmt_api_endpoint, jwt),
+            inner: Client::new(raw_client, mgmt_api_endpoint, jwt),
             node_id_label: node_id.0.to_string(),
         }
     }
@@ -207,6 +196,22 @@ impl PageserverClient {
         )
     }
 
+    pub(crate) async fn timeline_lease_lsn(
+        &self,
+        tenant_shard_id: TenantShardId,
+        timeline_id: TimelineId,
+        lsn: Lsn,
+    ) -> Result<LsnLease> {
+        measured_request!(
+            "timeline_lease_lsn",
+            crate::metrics::Method::Post,
+            &self.node_id_label,
+            self.inner
+                .timeline_init_lsn_lease(tenant_shard_id, timeline_id, lsn)
+                .await
+        )
+    }
+
     pub(crate) async fn tenant_shard_split(
         &self,
         tenant_shard_id: TenantShardId,

storage_controller/src/persistence.rs

@@ -997,10 +997,11 @@ impl Persistence {
                 // Clear sharding flag
                 let updated = diesel::update(tenant_shards)
                     .filter(tenant_id.eq(split_tenant_id.to_string()))
+                    .filter(shard_count.eq(new_shard_count.literal() as i32))
                     .set((splitting.eq(0),))
                     .execute(conn)
                     .await?;
-                debug_assert!(updated > 0);
+                assert!(updated == new_shard_count.count() as usize);
 
                 Ok(())
             })
@@ -1368,6 +1369,65 @@ impl Persistence {
         Ok(timeline_from_db)
     }
 
+    /// Set `delete_at` for the given timeline
+    pub(crate) async fn timeline_set_deleted_at(
+        &self,
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+    ) -> DatabaseResult<()> {
+        use crate::schema::timelines;
+
+        let deletion_time = chrono::Local::now().to_utc();
+        self.with_measured_conn(DatabaseOperation::InsertTimeline, move |conn| {
+            Box::pin(async move {
+                let updated = diesel::update(timelines::table)
+                    .filter(timelines::tenant_id.eq(tenant_id.to_string()))
+                    .filter(timelines::timeline_id.eq(timeline_id.to_string()))
+                    .set(timelines::deleted_at.eq(Some(deletion_time)))
+                    .execute(conn)
+                    .await?;
+
+                match updated {
+                    0 => Ok(()),
+                    1 => Ok(()),
+                    _ => Err(DatabaseError::Logical(format!(
+                        "unexpected number of rows ({})",
+                        updated
+                    ))),
+                }
+            })
+        })
+        .await
+    }
+
+    /// Load timeline from db. Returns `None` if not present.
+    ///
+    /// Only works if `deleted_at` is set, so you should call [`Self::timeline_set_deleted_at`] before.
+    pub(crate) async fn delete_timeline(
+        &self,
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+    ) -> DatabaseResult<()> {
+        use crate::schema::timelines::dsl;
+
+        let tenant_id = &tenant_id;
+        let timeline_id = &timeline_id;
+        self.with_measured_conn(DatabaseOperation::GetTimeline, move |conn| {
+            Box::pin(async move {
+                diesel::delete(dsl::timelines)
+                    .filter(dsl::tenant_id.eq(&tenant_id.to_string()))
+                    .filter(dsl::timeline_id.eq(&timeline_id.to_string()))
+                    .filter(dsl::deleted_at.is_not_null())
+                    .execute(conn)
+                    .await?;
+                Ok(())
+            })
+        })
+        .await?;
+
+        Ok(())
+    }
+
     /// Loads a list of all timelines from database.
     pub(crate) async fn list_timelines_for_tenant(
         &self,
@@ -1490,6 +1550,34 @@ impl Persistence {
 
         Ok(timeline_from_db)
     }
+    /// List pending operations for a given timeline (including tenant-global ones)
+    pub(crate) async fn list_pending_ops_for_timeline(
+        &self,
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+    ) -> DatabaseResult<Vec<TimelinePendingOpPersistence>> {
+        use crate::schema::safekeeper_timeline_pending_ops::dsl;
+
+        let timelines_from_db = self
+            .with_measured_conn(DatabaseOperation::ListTimelineReconcile, move |conn| {
+                Box::pin(async move {
+                    let from_db: Vec<TimelinePendingOpPersistence> =
+                        dsl::safekeeper_timeline_pending_ops
+                            .filter(dsl::tenant_id.eq(tenant_id.to_string()))
+                            .filter(
+                                dsl::timeline_id
+                                    .eq(timeline_id.to_string())
+                                    .or(dsl::timeline_id.eq("")),
+                            )
+                            .load(conn)
+                            .await?;
+                    Ok(from_db)
+                })
+            })
+            .await?;
+
+        Ok(timelines_from_db)
+    }
 
     /// Delete all pending ops for the given timeline.
     ///
@@ -1973,7 +2061,7 @@ impl ToSql<crate::schema::sql_types::PgLsn, Pg> for LsnWrapper {
     }
 }
 
-#[derive(Insertable, AsChangeset, Queryable, Selectable, Clone)]
+#[derive(Insertable, AsChangeset, Clone)]
 #[diesel(table_name = crate::schema::timelines)]
 pub(crate) struct TimelinePersistence {
     pub(crate) tenant_id: String,

storage_controller/src/reconciler.rs

@@ -86,6 +86,9 @@ pub(super) struct Reconciler {
 
     /// Access to persistent storage for updating generation numbers
     pub(crate) persistence: Arc<Persistence>,
+
+    /// HTTP client with proper CA certs.
+    pub(crate) http_client: reqwest::Client,
 }
 
 pub(crate) struct ReconcilerConfigBuilder {
@@ -298,8 +301,8 @@ impl Reconciler {
                         .location_config(tenant_shard_id, config.clone(), flush_ms, lazy)
                         .await
                 },
+                &self.http_client,
                 &self.service_config.pageserver_jwt_token,
-                &self.service_config.ssl_ca_cert,
                 1,
                 3,
                 timeout,
@@ -419,10 +422,10 @@ impl Reconciler {
 
         let client = PageserverClient::new(
             node.get_id(),
+            self.http_client.clone(),
             node.base_url(),
             self.service_config.pageserver_jwt_token.as_deref(),
-            self.service_config.ssl_ca_cert.clone(),
-        )?;
+        );
 
         client
             .wait_lsn(
@@ -443,10 +446,10 @@ impl Reconciler {
     ) -> anyhow::Result<HashMap<TimelineId, Lsn>> {
         let client = PageserverClient::new(
             node.get_id(),
+            self.http_client.clone(),
             node.base_url(),
             self.service_config.pageserver_jwt_token.as_deref(),
-            self.service_config.ssl_ca_cert.clone(),
-        )?;
+        );
 
         let timelines = client.timeline_list(&tenant_shard_id).await?;
         Ok(timelines
@@ -483,8 +486,8 @@ impl Reconciler {
                             )
                             .await
                     },
+                    &self.http_client,
                     &self.service_config.pageserver_jwt_token,
-                    &self.service_config.ssl_ca_cert,
                     1,
                     3,
                     request_download_timeout * 2,
@@ -778,8 +781,8 @@ impl Reconciler {
             let observed_conf = match attached_node
                 .with_client_retries(
                     |client| async move { client.get_location_config(tenant_shard_id).await },
+                    &self.http_client,
                     &self.service_config.pageserver_jwt_token,
-                    &self.service_config.ssl_ca_cert,
                     1,
                     1,
                     Duration::from_secs(5),
@@ -1127,8 +1130,8 @@ impl Reconciler {
             match origin
                 .with_client_retries(
                     |client| async move { client.get_location_config(tenant_shard_id).await },
+                    &self.http_client,
                     &self.service_config.pageserver_jwt_token,
-                    &self.service_config.ssl_ca_cert,
                     1,
                     3,
                     Duration::from_secs(5),

storage_controller/src/safekeeper.rs

@@ -1,7 +1,7 @@
 use std::time::Duration;
 
 use pageserver_api::controller_api::{SafekeeperDescribeResponse, SkSchedulingPolicy};
-use reqwest::{Certificate, StatusCode};
+use reqwest::StatusCode;
 use safekeeper_client::mgmt_api;
 use tokio_util::sync::CancellationToken;
 use utils::backoff;
@@ -94,8 +94,8 @@ impl Safekeeper {
     pub(crate) async fn with_client_retries<T, O, F>(
         &self,
         mut op: O,
+        http_client: &reqwest::Client,
         jwt: &Option<SecretString>,
-        ssl_ca_cert: &Option<Certificate>,
         warn_threshold: u32,
         max_retries: u32,
         timeout: Duration,
@@ -114,17 +114,10 @@ impl Safekeeper {
                 | ApiError(StatusCode::REQUEST_TIMEOUT, _) => false,
                 ApiError(_, _) => true,
                 Cancelled => true,
-                CreateClient(_) => true,
+                Timeout(_) => false,
             }
         }
 
-        // TODO: refactor SafekeeperClient and with_client_retires (#11113).
-        let mut http_client = reqwest::Client::builder().timeout(timeout);
-        if let Some(ssl_ca_cert) = ssl_ca_cert.as_ref() {
-            http_client = http_client.add_root_certificate(ssl_ca_cert.clone());
-        }
-        let http_client = http_client.build().map_err(mgmt_api::Error::CreateClient)?;
-
         backoff::retry(
             || {
                 let client = SafekeeperClient::new(
@@ -136,11 +129,14 @@ impl Safekeeper {
 
                 let node_cancel_fut = self.cancel.cancelled();
 
-                let op_fut = op(client);
+                let op_fut = tokio::time::timeout(timeout, op(client));
 
                 async {
                     tokio::select! {
-                        r = op_fut=> {r},
+                        r = op_fut => match r {
+                            Ok(r) => r,
+                            Err(e) => Err(mgmt_api::Error::Timeout(format!("{e}"))),
+                        },
                         _ = node_cancel_fut => {
                         Err(mgmt_api::Error::Cancelled)
                     }}

storage_controller/src/service.rs

@@ -12,7 +12,7 @@ use std::ops::{Deref, DerefMut};
 use std::path::PathBuf;
 use std::str::FromStr;
 use std::sync::Arc;
-use std::time::{Duration, Instant};
+use std::time::{Duration, Instant, SystemTime};
 
 use anyhow::Context;
 use context_iterator::TenantShardContextIterator;
@@ -34,7 +34,7 @@ use pageserver_api::controller_api::{
     TenantShardMigrateRequest, TenantShardMigrateResponse,
 };
 use pageserver_api::models::{
-    self, DetachBehavior, LocationConfig, LocationConfigListResponse, LocationConfigMode,
+    self, DetachBehavior, LocationConfig, LocationConfigListResponse, LocationConfigMode, LsnLease,
     PageserverUtilization, SecondaryProgress, ShardParameters, TenantConfig,
     TenantConfigPatchRequest, TenantConfigRequest, TenantLocationConfigRequest,
     TenantLocationConfigResponse, TenantShardLocation, TenantShardSplitRequest,
@@ -60,6 +60,7 @@ use tracing::{Instrument, debug, error, info, info_span, instrument, warn};
 use utils::completion::Barrier;
 use utils::generation::Generation;
 use utils::id::{NodeId, TenantId, TimelineId};
+use utils::lsn::Lsn;
 use utils::sync::gate::Gate;
 use utils::{failpoint_support, pausable_failpoint};
 
@@ -152,6 +153,7 @@ enum TenantOperations {
     TimelineGcBlockUnblock,
     DropDetached,
     DownloadHeatmapLayers,
+    TimelineLsnLease,
 }
 
 #[derive(Clone, strum_macros::Display)]
@@ -267,7 +269,7 @@ fn passthrough_api_error(node: &Node, e: mgmt_api::Error) -> ApiError {
             ApiError::Conflict(format!("{node} {status}: {status} {msg}"))
         }
         mgmt_api::Error::Cancelled => ApiError::ShuttingDown,
-        mgmt_api::Error::CreateClient(e) => ApiError::InternalServerError(anyhow::anyhow!(e)),
+        mgmt_api::Error::Timeout(e) => ApiError::Timeout(e.into()),
     }
 }
 
@@ -445,9 +447,11 @@ pub struct Config {
 
     pub use_https_safekeeper_api: bool,
 
-    pub ssl_ca_cert: Option<Certificate>,
+    pub ssl_ca_certs: Vec<Certificate>,
 
     pub timelines_onto_safekeepers: bool,
+
+    pub use_local_compute_notifications: bool,
 }
 
 impl From<DatabaseError> for ApiError {
@@ -524,6 +528,9 @@ pub struct Service {
     /// This waits for initial reconciliation with pageservers to complete.  Until this barrier
     /// passes, it isn't safe to do any actions that mutate tenants.
     pub(crate) startup_complete: Barrier,
+
+    /// HTTP client with proper CA certs.
+    http_client: reqwest::Client,
 }
 
 impl From<ReconcileWaitError> for ApiError {
@@ -667,6 +674,10 @@ impl Service {
         &self.config
     }
 
+    pub fn get_http_client(&self) -> &reqwest::Client {
+        &self.http_client
+    }
+
     /// Called once on startup, this function attempts to contact all pageservers to build an up-to-date
     /// view of the world, and determine which pageservers are responsive.
     #[instrument(skip_all)]
@@ -965,8 +976,8 @@ impl Service {
                     let response = node
                         .with_client_retries(
                             |client| async move { client.list_location_config().await },
+                            &self.http_client,
                             &self.config.pageserver_jwt_token,
-                            &self.config.ssl_ca_cert,
                             1,
                             5,
                             timeout,
@@ -1064,20 +1075,12 @@ impl Service {
                 break;
             }
 
-            let client = match PageserverClient::new(
+            let client = PageserverClient::new(
                 node.get_id(),
+                self.http_client.clone(),
                 node.base_url(),
                 self.config.pageserver_jwt_token.as_deref(),
-                self.config.ssl_ca_cert.clone(),
-            ) {
-                Ok(client) => client,
-                Err(e) => {
-                    tracing::error!(
-                        "Failed to create client to detach unknown shard {tenant_shard_id} on pageserver {node_id}: {e}"
-                    );
-                    continue;
-                }
-            };
+            );
             match client
                 .location_config(
                     tenant_shard_id,
@@ -1655,17 +1658,36 @@ impl Service {
         let cancel = CancellationToken::new();
         let reconcilers_cancel = cancel.child_token();
 
+        let mut http_client = reqwest::Client::builder();
+        // We intentionally disable the connection pool, so every request will create its own TCP connection.
+        // It's especially important for heartbeaters to notice more network problems.
+        //
+        // TODO: It makes sense to use this client only in heartbeaters and create a second one with
+        // connection pooling for everything else. But reqwest::Client may create a connection without
+        // ever using it (it uses hyper's Client under the hood):
+        // https://github.com/hyperium/hyper-util/blob/d51318df3461d40e5f5e5ca163cb3905ac960209/src/client/legacy/client.rs#L415
+        //
+        // Because of a bug in hyper0::Connection::graceful_shutdown such connections hang during
+        // graceful server shutdown: https://github.com/hyperium/hyper/issues/2730
+        //
+        // The bug has been fixed in hyper v1, so keep alive may be enabled only after we migrate to hyper1.
+        http_client = http_client.pool_max_idle_per_host(0);
+        for ssl_ca_cert in &config.ssl_ca_certs {
+            http_client = http_client.add_root_certificate(ssl_ca_cert.clone());
+        }
+        let http_client = http_client.build()?;
+
         let heartbeater_ps = Heartbeater::new(
+            http_client.clone(),
             config.pageserver_jwt_token.clone(),
-            config.ssl_ca_cert.clone(),
             config.max_offline_interval,
             config.max_warming_up_interval,
             cancel.clone(),
         );
 
         let heartbeater_sk = Heartbeater::new(
+            http_client.clone(),
             config.safekeeper_jwt_token.clone(),
-            config.ssl_ca_cert.clone(),
             config.max_offline_interval,
             config.max_warming_up_interval,
             cancel.clone(),
@@ -1708,6 +1730,7 @@ impl Service {
             reconcilers_gate: Gate::default(),
             tenant_op_locks: Default::default(),
             node_op_locks: Default::default(),
+            http_client,
         });
 
         let result_task_this = this.clone();
@@ -2013,8 +2036,8 @@ impl Service {
         let configs = match node
             .with_client_retries(
                 |client| async move { client.list_location_config().await },
+                &self.http_client,
                 &self.config.pageserver_jwt_token,
-                &self.config.ssl_ca_cert,
                 1,
                 5,
                 SHORT_RECONCILE_TIMEOUT,
@@ -2092,8 +2115,8 @@ impl Service {
                             .location_config(tenant_shard_id, config, None, false)
                             .await
                     },
+                    &self.http_client,
                     &self.config.pageserver_jwt_token,
-                    &self.config.ssl_ca_cert,
                     1,
                     5,
                     SHORT_RECONCILE_TIMEOUT,
@@ -3235,11 +3258,10 @@ impl Service {
             for tenant_shard_id in shard_ids {
                 let client = PageserverClient::new(
                     node.get_id(),
+                    self.http_client.clone(),
                     node.base_url(),
                     self.config.pageserver_jwt_token.as_deref(),
-                    self.config.ssl_ca_cert.clone(),
-                )
-                .map_err(|e| passthrough_api_error(&node, e))?;
+                );
 
                 tracing::info!("Doing time travel recovery for shard {tenant_shard_id}",);
 
@@ -3298,11 +3320,10 @@ impl Service {
         for (tenant_shard_id, node) in targets {
             let client = PageserverClient::new(
                 node.get_id(),
+                self.http_client.clone(),
                 node.base_url(),
                 self.config.pageserver_jwt_token.as_deref(),
-                self.config.ssl_ca_cert.clone(),
-            )
-            .map_err(|e| passthrough_api_error(&node, e))?;
+            );
             futs.push(async move {
                 let result = client
                     .tenant_secondary_download(tenant_shard_id, wait)
@@ -3427,8 +3448,8 @@ impl Service {
                         .tenant_delete(TenantShardId::unsharded(tenant_id))
                         .await
                 },
+                &self.http_client,
                 &self.config.pageserver_jwt_token,
-                &self.config.ssl_ca_cert,
                 1,
                 3,
                 RECONCILE_TIMEOUT,
@@ -3580,8 +3601,8 @@ impl Service {
             async fn create_one(
                 tenant_shard_id: TenantShardId,
                 locations: ShardMutationLocations,
+                http_client: reqwest::Client,
                 jwt: Option<String>,
-                ssl_ca_cert: Option<Certificate>,
                 create_req: TimelineCreateRequest,
             ) -> Result<TimelineInfo, ApiError> {
                 let latest = locations.latest.node;
@@ -3594,8 +3615,7 @@ impl Service {
                 );
 
                 let client =
-                    PageserverClient::new(latest.get_id(), latest.base_url(), jwt.as_deref(), ssl_ca_cert.clone())
-                    .map_err(|e| passthrough_api_error(&latest, e))?;
+                    PageserverClient::new(latest.get_id(), http_client.clone(), latest.base_url(), jwt.as_deref());
 
                 let timeline_info = client
                     .timeline_create(tenant_shard_id, &create_req)
@@ -3616,11 +3636,10 @@ impl Service {
 
                     let client = PageserverClient::new(
                         location.node.get_id(),
+                        http_client.clone(),
                         location.node.base_url(),
                         jwt.as_deref(),
-                        ssl_ca_cert.clone(),
-                    )
-                    .map_err(|e| passthrough_api_error(&location.node, e))?;
+                    );
 
                     let res = client
                         .timeline_create(tenant_shard_id, &create_req)
@@ -3648,8 +3667,8 @@ impl Service {
             let timeline_info = create_one(
                 shard_zero_tid,
                 shard_zero_locations,
+                self.http_client.clone(),
                 self.config.pageserver_jwt_token.clone(),
-                self.config.ssl_ca_cert.clone(),
                 create_req.clone(),
             )
             .await?;
@@ -3678,8 +3697,8 @@ impl Service {
                         Box::pin(create_one(
                             tenant_shard_id,
                             mutation_locations,
+                            self.http_client.clone(),
                             jwt.clone(),
-                            self.config.ssl_ca_cert.clone(),
                             create_req,
                         ))
                     },
@@ -3762,16 +3781,15 @@ impl Service {
                 tenant_shard_id: TenantShardId,
                 timeline_id: TimelineId,
                 node: Node,
+                http_client: reqwest::Client,
                 jwt: Option<String>,
-                ssl_ca_cert: Option<Certificate>,
                 req: TimelineArchivalConfigRequest,
             ) -> Result<(), ApiError> {
                 tracing::info!(
                     "Setting archival config of timeline on shard {tenant_shard_id}/{timeline_id}, attached to node {node}",
                 );
 
-                let client = PageserverClient::new(node.get_id(), node.base_url(), jwt.as_deref(), ssl_ca_cert)
-                    .map_err(|e| passthrough_api_error(&node, e))?;
+                let client = PageserverClient::new(node.get_id(),  http_client, node.base_url(), jwt.as_deref());
 
                 client
                     .timeline_archival_config(tenant_shard_id, timeline_id, &req)
@@ -3793,8 +3811,8 @@ impl Service {
                         tenant_shard_id,
                         timeline_id,
                         node,
+                        self.http_client.clone(),
                         self.config.pageserver_jwt_token.clone(),
-                        self.config.ssl_ca_cert.clone(),
                         req.clone(),
                     ))
                 })
@@ -3831,16 +3849,15 @@ impl Service {
                 tenant_shard_id: TenantShardId,
                 timeline_id: TimelineId,
                 node: Node,
+                http_client: reqwest::Client,
                 jwt: Option<String>,
-                ssl_ca_cert: Option<Certificate>,
                 behavior: Option<DetachBehavior>,
             ) -> Result<(ShardNumber, models::detach_ancestor::AncestorDetached), ApiError> {
                 tracing::info!(
                     "Detaching timeline on shard {tenant_shard_id}/{timeline_id}, attached to node {node}",
                 );
 
-                let client = PageserverClient::new(node.get_id(), node.base_url(), jwt.as_deref(), ssl_ca_cert)
-                    .map_err(|e| passthrough_api_error(&node, e))?;
+                let client = PageserverClient::new(node.get_id(), http_client, node.base_url(), jwt.as_deref());
 
                 client
                     .timeline_detach_ancestor(tenant_shard_id, timeline_id, behavior)
@@ -3879,8 +3896,8 @@ impl Service {
                         tenant_shard_id,
                         timeline_id,
                         node,
+                        self.http_client.clone(),
                         self.config.pageserver_jwt_token.clone(),
-                        self.config.ssl_ca_cert.clone(),
                         behavior,
                     ))
                 })
@@ -3933,17 +3950,16 @@ impl Service {
                 tenant_shard_id: TenantShardId,
                 timeline_id: TimelineId,
                 node: Node,
+                http_client: reqwest::Client,
                 jwt: Option<String>,
-                ssl_ca_cert: Option<Certificate>,
                 dir: BlockUnblock,
             ) -> Result<(), ApiError> {
                 let client = PageserverClient::new(
                     node.get_id(),
+                    http_client,
                     node.base_url(),
                     jwt.as_deref(),
-                    ssl_ca_cert,
-                )
-                .map_err(|e| passthrough_api_error(&node, e))?;
+                );
 
                 client
                     .timeline_block_unblock_gc(tenant_shard_id, timeline_id, dir)
@@ -3962,8 +3978,8 @@ impl Service {
                     tenant_shard_id,
                     timeline_id,
                     node,
+                    self.http_client.clone(),
                     self.config.pageserver_jwt_token.clone(),
-                    self.config.ssl_ca_cert.clone(),
                     dir,
                 ))
             })
@@ -3973,6 +3989,75 @@ impl Service {
         Ok(())
     }
 
+    pub(crate) async fn tenant_timeline_lsn_lease(
+        &self,
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+        lsn: Lsn,
+    ) -> Result<LsnLease, ApiError> {
+        let _tenant_lock = trace_shared_lock(
+            &self.tenant_op_locks,
+            tenant_id,
+            TenantOperations::TimelineLsnLease,
+        )
+        .await;
+
+        let targets = {
+            let locked = self.inner.read().unwrap();
+            let mut targets = Vec::new();
+
+            // If the request got an unsharded tenant id, then apply
+            // the operation to all shards. Otherwise, apply it to a specific shard.
+            let shards_range = TenantShardId::tenant_range(tenant_id);
+
+            for (tenant_shard_id, shard) in locked.tenants.range(shards_range) {
+                if let Some(node_id) = shard.intent.get_attached() {
+                    let node = locked
+                        .nodes
+                        .get(node_id)
+                        .expect("Pageservers may not be deleted while referenced");
+
+                    targets.push((*tenant_shard_id, node.clone()));
+                }
+            }
+            targets
+        };
+
+        let res = self
+            .tenant_for_shards_api(
+                targets,
+                |tenant_shard_id, client| async move {
+                    client
+                        .timeline_lease_lsn(tenant_shard_id, timeline_id, lsn)
+                        .await
+                },
+                1,
+                1,
+                SHORT_RECONCILE_TIMEOUT,
+                &self.cancel,
+            )
+            .await;
+
+        let mut valid_until = None;
+        for r in res {
+            match r {
+                Ok(lease) => {
+                    if let Some(ref mut valid_until) = valid_until {
+                        *valid_until = std::cmp::min(*valid_until, lease.valid_until);
+                    } else {
+                        valid_until = Some(lease.valid_until);
+                    }
+                }
+                Err(e) => {
+                    return Err(ApiError::InternalServerError(anyhow::anyhow!(e)));
+                }
+            }
+        }
+        Ok(LsnLease {
+            valid_until: valid_until.unwrap_or_else(SystemTime::now),
+        })
+    }
+
     pub(crate) async fn tenant_timeline_download_heatmap_layers(
         &self,
         tenant_shard_id: TenantShardId,
@@ -4091,8 +4176,8 @@ impl Service {
                 let r = node
                     .with_client_retries(
                         |client| op(tenant_shard_id, client),
+                        &self.http_client,
                         &self.config.pageserver_jwt_token,
-                        &self.config.ssl_ca_cert,
                         warn_threshold,
                         max_retries,
                         timeout,
@@ -4316,15 +4401,14 @@ impl Service {
                 tenant_shard_id: TenantShardId,
                 timeline_id: TimelineId,
                 node: Node,
+                http_client: reqwest::Client,
                 jwt: Option<String>,
-                ssl_ca_cert: Option<Certificate>,
             ) -> Result<StatusCode, ApiError> {
                 tracing::info!(
                     "Deleting timeline on shard {tenant_shard_id}/{timeline_id}, attached to node {node}",
                 );
 
-                let client = PageserverClient::new(node.get_id(), node.base_url(), jwt.as_deref(), ssl_ca_cert)
-                    .map_err(|e| passthrough_api_error(&node, e))?;
+                let client = PageserverClient::new(node.get_id(), http_client, node.base_url(), jwt.as_deref());
                 let res = client
                     .timeline_delete(tenant_shard_id, timeline_id)
                     .await;
@@ -4350,8 +4434,8 @@ impl Service {
                         tenant_shard_id,
                         timeline_id,
                         node,
+                        self.http_client.clone(),
                         self.config.pageserver_jwt_token.clone(),
-                        self.config.ssl_ca_cert.clone(),
                     ))
                 })
                 .await?;
@@ -4373,8 +4457,8 @@ impl Service {
                 shard_zero_tid,
                 timeline_id,
                 shard_zero_locations.latest.node,
+                self.http_client.clone(),
                 self.config.pageserver_jwt_token.clone(),
-                self.config.ssl_ca_cert.clone(),
             )
             .await?;
             Ok(shard_zero_status)
@@ -4809,8 +4893,8 @@ impl Service {
 
                         client.location_config(child_id, config, None, false).await
                     },
+                    &self.http_client,
                     &self.config.pageserver_jwt_token,
-                    &self.config.ssl_ca_cert,
                     1,
                     10,
                     Duration::from_secs(5),
@@ -5412,11 +5496,10 @@ impl Service {
             } = target;
             let client = PageserverClient::new(
                 node.get_id(),
+                self.http_client.clone(),
                 node.base_url(),
                 self.config.pageserver_jwt_token.as_deref(),
-                self.config.ssl_ca_cert.clone(),
-            )
-            .map_err(|e| passthrough_api_error(node, e))?;
+            );
             let response = client
                 .tenant_shard_split(
                     *parent_id,
@@ -5456,6 +5539,8 @@ impl Service {
             }
         }
 
+        pausable_failpoint!("shard-split-pre-complete");
+
         // TODO: if the pageserver restarted concurrently with our split API call,
         // the actual generation of the child shard might differ from the generation
         // we expect it to have.  In order for our in-database generation to end up
@@ -5898,11 +5983,10 @@ impl Service {
 
         let client = PageserverClient::new(
             node.get_id(),
+            self.http_client.clone(),
             node.base_url(),
             self.config.pageserver_jwt_token.as_deref(),
-            self.config.ssl_ca_cert.clone(),
-        )
-        .map_err(|e| passthrough_api_error(&node, e))?;
+        );
 
         let scan_result = client
             .tenant_scan_remote_storage(tenant_id)
@@ -7136,6 +7220,7 @@ impl Service {
             units,
             gate_guard,
             &self.reconcilers_cancel,
+            self.http_client.clone(),
         )
     }
 
@@ -7543,8 +7628,8 @@ impl Service {
         match attached_node
             .with_client_retries(
                 |client| async move { client.tenant_heatmap_upload(tenant_shard_id).await },
+                &self.http_client,
                 &self.config.pageserver_jwt_token,
-                &self.config.ssl_ca_cert,
                 3,
                 10,
                 SHORT_RECONCILE_TIMEOUT,
@@ -7580,8 +7665,8 @@ impl Service {
                             )
                             .await
                     },
+                    &self.http_client,
                     &self.config.pageserver_jwt_token,
-                    &self.config.ssl_ca_cert,
                     3,
                     10,
                     SHORT_RECONCILE_TIMEOUT,
@@ -7854,8 +7939,8 @@ impl Service {
             futures.push(async move {
                 node.with_client_retries(
                     |client| async move { client.top_tenant_shards(request.clone()).await },
+                    &self.http_client,
                     &self.config.pageserver_jwt_token,
-                    &self.config.ssl_ca_cert,
                     3,
                     3,
                     Duration::from_secs(5),
@@ -7974,8 +8059,8 @@ impl Service {
         match node
             .with_client_retries(
                 |client| async move { client.tenant_secondary_status(tenant_shard_id).await },
+                &self.http_client,
                 &self.config.pageserver_jwt_token,
-                &self.config.ssl_ca_cert,
                 1,
                 3,
                 Duration::from_millis(250),

storage_controller/src/service/chaos_injector.rs

@@ -4,7 +4,7 @@ use std::time::Duration;
 
 use pageserver_api::controller_api::ShardSchedulingPolicy;
 use rand::seq::SliceRandom;
-use rand::thread_rng;
+use rand::{Rng, thread_rng};
 use tokio_util::sync::CancellationToken;
 use utils::id::NodeId;
 use utils::shard::TenantShardId;
@@ -64,17 +64,22 @@ impl ChaosInjector {
         let mut interval = tokio::time::interval(self.interval);
         #[derive(Debug)]
         enum ChaosEvent {
-            ShuffleTenant,
-            ForceKill,
+            MigrationsToSecondary,
+            ForceKillController,
+            GracefulMigrationsAnywhere,
         }
         loop {
             let cron_interval = self.get_cron_interval_sleep_future();
             let chaos_type = tokio::select! {
                 _ = interval.tick() => {
-                    ChaosEvent::ShuffleTenant
+                    if thread_rng().gen_bool(0.5) {
+                        ChaosEvent::MigrationsToSecondary
+                    } else {
+                        ChaosEvent::GracefulMigrationsAnywhere
+                    }
                 }
                 Some(_) = maybe_sleep(cron_interval) => {
-                    ChaosEvent::ForceKill
+                    ChaosEvent::ForceKillController
                 }
                 _ = cancel.cancelled() => {
                     tracing::info!("Shutting down");
@@ -83,16 +88,29 @@ impl ChaosInjector {
             };
             tracing::info!("Chaos iteration: {chaos_type:?}...");
             match chaos_type {
-                ChaosEvent::ShuffleTenant => {
-                    self.inject_chaos().await;
+                ChaosEvent::MigrationsToSecondary => {
+                    self.inject_migrations_to_secondary();
                 }
-                ChaosEvent::ForceKill => {
+                ChaosEvent::GracefulMigrationsAnywhere => {
+                    self.inject_graceful_migrations_anywhere();
+                }
+                ChaosEvent::ForceKillController => {
                     self.force_kill().await;
                 }
             }
         }
     }
 
+    fn is_shard_eligible_for_chaos(&self, shard: &TenantShard) -> bool {
+        // - Skip non-active scheduling policies, so that a shard with a policy like Pause can
+        //   be pinned without being disrupted by us.
+        // - Skip shards doing a graceful migration already, so that we allow these to run to
+        //   completion rather than only exercising the first part and then cancelling with
+        //   some other chaos.
+        !matches!(shard.get_scheduling_policy(), ShardSchedulingPolicy::Active)
+            && shard.get_preferred_node().is_none()
+    }
+
     /// If a shard has a secondary and attached location, then re-assign the secondary to be
     /// attached and the attached to be secondary.
     ///
@@ -108,13 +126,7 @@ impl ChaosInjector {
             .get_mut(&tenant_shard_id)
             .expect("Held lock between choosing ID and this get");
 
-        if !matches!(shard.get_scheduling_policy(), ShardSchedulingPolicy::Active) {
-            // Skip non-active scheduling policies, so that a shard with a policy like Pause can
-            // be pinned without being disrupted by us.
-            tracing::info!(
-                "Skipping shard {tenant_shard_id}: scheduling policy is {:?}",
-                shard.get_scheduling_policy()
-            );
+        if !self.is_shard_eligible_for_chaos(shard) {
             return;
         }
 
@@ -152,7 +164,77 @@ impl ChaosInjector {
         std::process::exit(1);
     }
 
-    async fn inject_chaos(&mut self) {
+    // Unlike [`Self::inject_migrations_to_secondary`], this function will not only cut over to secondary, it
+    // will migrate a tenant to a random node in its home AZ using a graceful migration of the same type
+    // that my be initiated by an API caller using prewarm=true.
+    //
+    // This is a much more expensive operation in terms of I/O and time, as we will fully warm up
+    // some new location in order to migrate the tenant there.  For that reason we do far fewer of these.
+    fn inject_graceful_migrations_anywhere(&mut self) {
+        let batch_size = 1;
+        let mut inner = self.service.inner.write().unwrap();
+        let (nodes, tenants, _scheduler) = inner.parts_mut();
+
+        let mut candidates = tenants
+            .values_mut()
+            .filter(|shard| self.is_shard_eligible_for_chaos(shard))
+            .collect::<Vec<_>>();
+
+        tracing::info!(
+            "Injecting chaos: found {} candidates for graceful migrations anywhere",
+            candidates.len()
+        );
+
+        let mut victims: Vec<&mut TenantShard> = Vec::new();
+
+        // Pick our victims: use a hand-rolled loop rather than choose_multiple() because we want
+        // to take the mutable refs from our candidates rather than ref'ing them.
+        while !candidates.is_empty() && victims.len() < batch_size {
+            let i = thread_rng().gen_range(0..candidates.len());
+            victims.push(candidates.swap_remove(i));
+        }
+
+        for victim in victims.into_iter() {
+            // Find a node in the same AZ as the shard, or if the shard has no AZ preference, which
+            // is not where they are currently attached.
+            let candidate_nodes = nodes
+                .values()
+                .filter(|node| {
+                    if let Some(preferred_az) = victim.preferred_az() {
+                        node.get_availability_zone_id() == preferred_az
+                    } else if let Some(attached) = *victim.intent.get_attached() {
+                        node.get_id() != attached
+                    } else {
+                        true
+                    }
+                })
+                .collect::<Vec<_>>();
+
+            let Some(victim_node) = candidate_nodes.choose(&mut thread_rng()) else {
+                // This can happen if e.g. we are in a small region with only one pageserver per AZ.
+                tracing::info!(
+                    "no candidate nodes found for migrating shard {tenant_shard_id} within its home AZ",
+                    tenant_shard_id = victim.tenant_shard_id
+                );
+                continue;
+            };
+
+            // This doesn't change intent immediately: next iteration of Service::optimize_all should do that.  We avoid
+            // doing it here because applying optimizations requires dropping lock to do some async work to check the optimisation
+            // is valid given remote state, and it would be a shame to duplicate that dance here.
+            tracing::info!(
+                "Injecting chaos: migrate {} to {}",
+                victim.tenant_shard_id,
+                victim_node
+            );
+            victim.set_preferred_node(Some(victim_node.get_id()));
+        }
+    }
+
+    /// Migrations of attached locations to their secondary location.  This exercises reconciliation in general,
+    /// live migration in particular, and the pageserver code for cleanly shutting down and starting up tenants
+    /// during such migrations.
+    fn inject_migrations_to_secondary(&mut self) {
         // Pick some shards to interfere with
         let batch_size = 128;
         let mut inner = self.service.inner.write().unwrap();

storage_controller/src/service/safekeeper_reconciler.rs

@@ -160,9 +160,8 @@ pub(crate) struct ScheduleRequest {
 }
 
 struct ReconcilerHandle {
-    tx: UnboundedSender<(ScheduleRequest, Arc<CancellationToken>)>,
-    #[allow(clippy::type_complexity)]
-    ongoing_tokens: Arc<ClashMap<(TenantId, Option<TimelineId>), Arc<CancellationToken>>>,
+    tx: UnboundedSender<(ScheduleRequest, CancellationToken)>,
+    ongoing_tokens: Arc<ClashMap<(TenantId, Option<TimelineId>), CancellationToken>>,
     cancel: CancellationToken,
 }
 
@@ -172,13 +171,13 @@ impl ReconcilerHandle {
         &self,
         tenant_id: TenantId,
         timeline_id: Option<TimelineId>,
-    ) -> Arc<CancellationToken> {
+    ) -> CancellationToken {
         let entry = self.ongoing_tokens.entry((tenant_id, timeline_id));
         if let Entry::Occupied(entry) = &entry {
             let cancel: &CancellationToken = entry.get();
             cancel.cancel();
         }
-        entry.insert(Arc::new(self.cancel.child_token())).clone()
+        entry.insert(self.cancel.child_token()).clone()
     }
     /// Cancel an ongoing reconciliation
     fn cancel_reconciliation(&self, tenant_id: TenantId, timeline_id: Option<TimelineId>) {
@@ -197,7 +196,7 @@ impl ReconcilerHandle {
 
 pub(crate) struct SafekeeperReconciler {
     service: Arc<Service>,
-    rx: UnboundedReceiver<(ScheduleRequest, Arc<CancellationToken>)>,
+    rx: UnboundedReceiver<(ScheduleRequest, CancellationToken)>,
     cancel: CancellationToken,
 }
 
@@ -243,7 +242,7 @@ impl SafekeeperReconciler {
                 .await;
         }
     }
-    async fn reconcile_one(&self, req: ScheduleRequest, req_cancel: Arc<CancellationToken>) {
+    async fn reconcile_one(&self, req: ScheduleRequest, req_cancel: CancellationToken) {
         let req_host = req.safekeeper.skp.host.clone();
         match req.kind {
             SafekeeperTimelineOpKind::Pull => {
@@ -300,36 +299,96 @@ impl SafekeeperReconciler {
             SafekeeperTimelineOpKind::Delete => {
                 let tenant_id = req.tenant_id;
                 if let Some(timeline_id) = req.timeline_id {
-                    self.reconcile_inner(
+                    let deleted = self.reconcile_inner(
                         req,
                         async |client| client.delete_timeline(tenant_id, timeline_id).await,
                         |_resp| {
-                            tracing::info!("deleted timeline from {req_host}");
+                            tracing::info!(%tenant_id, %timeline_id, "deleted timeline from {req_host}");
                         },
                         req_cancel,
                     )
                     .await;
+                    if deleted {
+                        self.delete_timeline_from_db(tenant_id, timeline_id).await;
+                    }
                 } else {
-                    self.reconcile_inner(
-                        req,
-                        async |client| client.delete_tenant(tenant_id).await,
-                        |_resp| {
-                            tracing::info!("deleted tenant from {req_host}");
-                        },
-                        req_cancel,
-                    )
-                    .await;
+                    let deleted = self
+                        .reconcile_inner(
+                            req,
+                            async |client| client.delete_tenant(tenant_id).await,
+                            |_resp| {
+                                tracing::info!(%tenant_id, "deleted tenant from {req_host}");
+                            },
+                            req_cancel,
+                        )
+                        .await;
+                    if deleted {
+                        self.delete_tenant_timelines_from_db(tenant_id).await;
+                    }
                 }
             }
         }
     }
+    async fn delete_timeline_from_db(&self, tenant_id: TenantId, timeline_id: TimelineId) {
+        match self
+            .service
+            .persistence
+            .list_pending_ops_for_timeline(tenant_id, timeline_id)
+            .await
+        {
+            Ok(list) => {
+                if !list.is_empty() {
+                    tracing::info!(%tenant_id, %timeline_id, "not deleting timeline from db as there is {} open reconciles", list.len());
+                    return;
+                }
+            }
+            Err(e) => {
+                tracing::warn!(%tenant_id, %timeline_id, "couldn't query pending ops: {e}");
+                return;
+            }
+        }
+        tracing::info!(%tenant_id, %timeline_id, "deleting timeline from db after all reconciles succeeded");
+        // In theory we could crash right after deleting the op from the db and right before reaching this,
+        // but then we'll boot up with a timeline that has deleted_at set, so hopefully we'll issue deletion ops for it again.
+        if let Err(err) = self
+            .service
+            .persistence
+            .delete_timeline(tenant_id, timeline_id)
+            .await
+        {
+            tracing::warn!(%tenant_id, %timeline_id, "couldn't delete timeline from db: {err}");
+        }
+    }
+    async fn delete_tenant_timelines_from_db(&self, tenant_id: TenantId) {
+        let timeline_list = match self
+            .service
+            .persistence
+            .list_timelines_for_tenant(tenant_id)
+            .await
+        {
+            Ok(timeline_list) => timeline_list,
+            Err(e) => {
+                tracing::warn!(%tenant_id, "couldn't query timelines: {e}");
+                return;
+            }
+        };
+        for timeline in timeline_list {
+            let Ok(timeline_id) = TimelineId::from_str(&timeline.timeline_id) else {
+                tracing::warn!("Invalid timeline ID in database {}", timeline.timeline_id);
+                continue;
+            };
+            self.delete_timeline_from_db(tenant_id, timeline_id).await;
+        }
+    }
+    /// Returns whether the reconciliation happened successfully
     async fn reconcile_inner<T, F, U>(
         &self,
         req: ScheduleRequest,
         closure: impl Fn(SafekeeperClient) -> F,
         log_success: impl FnOnce(T) -> U,
-        req_cancel: Arc<CancellationToken>,
-    ) where
+        req_cancel: CancellationToken,
+    ) -> bool
+    where
         F: Future<Output = Result<T, safekeeper_client::mgmt_api::Error>>,
     {
         let jwt = self
@@ -338,7 +397,6 @@ impl SafekeeperReconciler {
             .safekeeper_jwt_token
             .clone()
             .map(SecretString::from);
-        let ssl_ca_cert = self.service.config.ssl_ca_cert.clone();
         loop {
             let res = req
                 .safekeeper
@@ -347,8 +405,8 @@ impl SafekeeperReconciler {
                         let closure = &closure;
                         async move { closure(client).await }
                     },
+                    self.service.get_http_client(),
                     &jwt,
-                    &ssl_ca_cert,
                     3,
                     10,
                     Duration::from_secs(10),
@@ -374,11 +432,11 @@ impl SafekeeperReconciler {
                             req.safekeeper.skp.host
                         );
                     }
-                    return;
+                    return true;
                 }
                 Err(mgmt_api::Error::Cancelled) => {
                     // On cancellation, the code that issued it will take care of removing db entries (if needed)
-                    return;
+                    return false;
                 }
                 Err(e) => {
                     tracing::info!(

storage_controller/src/service/safekeeper_service.rs

@@ -1,4 +1,4 @@
-use std::collections::{HashMap, HashSet};
+use std::collections::HashSet;
 use std::str::FromStr;
 use std::sync::Arc;
 use std::time::Duration;
@@ -78,8 +78,8 @@ impl Service {
         for sk in timeline_persistence.sk_set.iter() {
             let sk_id = NodeId(*sk as u64);
             let safekeepers = safekeepers.clone();
+            let http_client = self.http_client.clone();
             let jwt = jwt.clone();
-            let ssl_ca_cert = self.config.ssl_ca_cert.clone();
             let req = req.clone();
             joinset.spawn(async move {
                 // Unwrap is fine as we already would have returned error above
@@ -90,8 +90,8 @@ impl Service {
                             let req = req.clone();
                             async move { client.create_timeline(&req).await }
                         },
+                        &http_client,
                         &jwt,
-                        &ssl_ca_cert,
                         3,
                         3,
                         SK_CREATE_TIMELINE_RECONCILE_TIMEOUT,
@@ -313,25 +313,32 @@ impl Service {
             );
             return Ok(());
         };
+        self.persistence
+            .timeline_set_deleted_at(tenant_id, timeline_id)
+            .await?;
         let all_sks = tl
             .new_sk_set
             .iter()
-            .flat_map(|sks| {
-                sks.iter()
-                    .map(|sk| (*sk, SafekeeperTimelineOpKind::Exclude))
-            })
-            .chain(
-                tl.sk_set
-                    .iter()
-                    .map(|v| (*v, SafekeeperTimelineOpKind::Delete)),
-            )
-            .collect::<HashMap<_, _>>();
+            .flatten()
+            .chain(tl.sk_set.iter())
+            .collect::<HashSet<_>>();
 
         // Schedule reconciliations
+        for &sk_id in all_sks.iter() {
+            let pending_op = TimelinePendingOpPersistence {
+                tenant_id: tenant_id.to_string(),
+                timeline_id: timeline_id.to_string(),
+                generation: tl.generation,
+                op_kind: SafekeeperTimelineOpKind::Delete,
+                sk_id: *sk_id,
+            };
+            tracing::info!("writing pending op for sk id {sk_id}");
+            self.persistence.insert_pending_op(pending_op).await?;
+        }
         {
             let mut locked = self.inner.write().unwrap();
-            for (sk_id, kind) in all_sks {
-                let sk_id = NodeId(sk_id as u64);
+            for sk_id in all_sks {
+                let sk_id = NodeId(*sk_id as u64);
                 let Some(sk) = locked.safekeepers.get(&sk_id) else {
                     return Err(ApiError::InternalServerError(anyhow::anyhow!(
                         "Couldn't find safekeeper with id {sk_id}"
@@ -345,7 +352,7 @@ impl Service {
                     tenant_id,
                     timeline_id: Some(timeline_id),
                     generation: tl.generation as u32,
-                    kind,
+                    kind: SafekeeperTimelineOpKind::Delete,
                 };
                 locked.safekeeper_reconcilers.schedule_request(self, req);
             }
@@ -379,32 +386,50 @@ impl Service {
             })
             .collect::<Result<Vec<_>, ApiError>>()?;
 
-        // Remove pending ops from db.
+        // Remove pending ops from db, and set `deleted_at`.
         // We cancel them in a later iteration once we hold the state lock.
         for (timeline_id, _timeline) in timeline_list.iter() {
             self.persistence
                 .remove_pending_ops_for_timeline(tenant_id, Some(*timeline_id))
                 .await?;
+            self.persistence
+                .timeline_set_deleted_at(tenant_id, *timeline_id)
+                .await?;
         }
 
-        let mut locked = self.inner.write().unwrap();
-
         // The list of safekeepers that have any of the timelines
         let mut sk_list = HashSet::new();
 
         // List all pending ops for all timelines, cancel them
-        for (timeline_id, timeline) in timeline_list.iter() {
+        for (_timeline_id, timeline) in timeline_list.iter() {
             let sk_iter = timeline
                 .sk_set
                 .iter()
                 .chain(timeline.new_sk_set.iter().flatten())
                 .map(|id| NodeId(*id as u64));
-            for sk_id in sk_iter.clone() {
+            sk_list.extend(sk_iter);
+        }
+
+        for &sk_id in sk_list.iter() {
+            let pending_op = TimelinePendingOpPersistence {
+                tenant_id: tenant_id.to_string(),
+                timeline_id: String::new(),
+                generation: i32::MAX,
+                op_kind: SafekeeperTimelineOpKind::Delete,
+                sk_id: sk_id.0 as i64,
+            };
+            tracing::info!("writing pending op for sk id {sk_id}");
+            self.persistence.insert_pending_op(pending_op).await?;
+        }
+
+        let mut locked = self.inner.write().unwrap();
+
+        for (timeline_id, _timeline) in timeline_list.iter() {
+            for sk_id in sk_list.iter() {
                 locked
                     .safekeeper_reconcilers
-                    .cancel_reconciles_for_timeline(sk_id, tenant_id, Some(*timeline_id));
+                    .cancel_reconciles_for_timeline(*sk_id, tenant_id, Some(*timeline_id));
             }
-            sk_list.extend(sk_iter);
         }
 
         // unwrap is safe: we return above for an empty timeline list

storage_controller/src/tenant_shard.rs

@@ -1588,6 +1588,7 @@ impl TenantShard {
         units: ReconcileUnits,
         gate_guard: GateGuard,
         cancel: &CancellationToken,
+        http_client: reqwest::Client,
     ) -> Option<ReconcilerWaiter> {
         // Reconcile in flight for a stale sequence?  Our sequence's task will wait for it before
         // doing our sequence's work.
@@ -1633,6 +1634,7 @@ impl TenantShard {
             cancel: reconciler_cancel.clone(),
             persistence: persistence.clone(),
             compute_notify_failure: false,
+            http_client,
         };
 
         let reconcile_seq = self.sequence;

test_runner/cloud_regress/test_cloud_regress.py

@@ -4,11 +4,15 @@ Run the regression tests on the cloud instance of Neon
 
 from __future__ import annotations
 
-from pathlib import Path
+from typing import TYPE_CHECKING
 
 import pytest
-from fixtures.neon_fixtures import RemotePostgres
-from fixtures.pg_version import PgVersion
+
+if TYPE_CHECKING:
+    from pathlib import Path
+
+    from fixtures.neon_fixtures import RemotePostgres
+    from fixtures.pg_version import PgVersion
 
 
 @pytest.mark.timeout(7200)

test_runner/fixtures/auth_tokens.py

@@ -2,11 +2,12 @@ from __future__ import annotations
 
 from dataclasses import dataclass
 from enum import StrEnum
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import jwt
 
-from fixtures.common_types import TenantId
+if TYPE_CHECKING:
+    from fixtures.common_types import TenantId
 
 
 @dataclass

test_runner/fixtures/benchmark_fixture.py

@@ -15,18 +15,20 @@ from typing import TYPE_CHECKING
 
 import allure
 import pytest
-from _pytest.config import Config
-from _pytest.config.argparsing import Parser
-from _pytest.fixtures import FixtureRequest
-from _pytest.terminal import TerminalReporter
 
-from fixtures.common_types import TenantId, TimelineId
 from fixtures.log_helper import log
-from fixtures.neon_fixtures import NeonPageserver
 
 if TYPE_CHECKING:
     from collections.abc import Callable, Iterator, Mapping
 
+    from _pytest.config import Config
+    from _pytest.config.argparsing import Parser
+    from _pytest.fixtures import FixtureRequest
+    from _pytest.terminal import TerminalReporter
+
+    from fixtures.common_types import TenantId, TimelineId
+    from fixtures.neon_fixtures import NeonPageserver
+
 
 """
 This file contains fixtures for micro-benchmarks.

test_runner/fixtures/compare_fixtures.py

@@ -11,7 +11,6 @@ from pathlib import Path
 from typing import TYPE_CHECKING, final
 
 import pytest
-from _pytest.fixtures import FixtureRequest
 from typing_extensions import override
 
 from fixtures.benchmark_fixture import MetricReport, NeonBenchmarker
@@ -24,11 +23,14 @@ from fixtures.neon_fixtures import (
     VanillaPostgres,
     wait_for_last_flush_lsn,
 )
-from fixtures.pg_stats import PgStatTable
 
 if TYPE_CHECKING:
     from collections.abc import Iterator
 
+    from _pytest.fixtures import FixtureRequest
+
+    from fixtures.pg_stats import PgStatTable
+
 
 class PgCompare(ABC):
     """Common interface of all postgres implementations, useful for benchmarks.

test_runner/fixtures/compute_reconfigure.py

@@ -4,8 +4,6 @@ import concurrent.futures
 from typing import TYPE_CHECKING
 
 import pytest
-from pytest_httpserver import HTTPServer
-from werkzeug.wrappers.request import Request
 from werkzeug.wrappers.response import Response
 
 from fixtures.common_types import TenantId
@@ -15,6 +13,9 @@ if TYPE_CHECKING:
     from collections.abc import Callable
     from typing import Any
 
+    from pytest_httpserver import HTTPServer
+    from werkzeug.wrappers.request import Request
+
 
 class ComputeReconfigure:
     def __init__(self, server: HTTPServer):

test_runner/fixtures/endpoint/http.py

@@ -5,6 +5,8 @@ import urllib.parse
 import requests
 from requests.adapters import HTTPAdapter
 
+from fixtures.log_helper import log
+
 
 class EndpointHttpClient(requests.Session):
     def __init__(
@@ -51,6 +53,7 @@ class EndpointHttpClient(requests.Session):
     def metrics(self) -> str:
         res = self.get(f"http://localhost:{self.external_port}/metrics")
         res.raise_for_status()
+        log.debug("raw compute metrics: %s", res.text)
         return res.text
 
     # Current compute status.

test_runner/fixtures/fast_import.py

@@ -147,7 +147,7 @@ def fast_import(
         pg_distrib_dir,
         pg_version,
         workdir,
-        cleanup=not cast(bool, pytestconfig.getoption("--preserve-database-files")),
+        cleanup=not cast("bool", pytestconfig.getoption("--preserve-database-files")),
     ) as fi:
         yield fi
 

test_runner/fixtures/h2server.py

@@ -10,7 +10,6 @@ import asyncio
 import collections
 import io
 import json
-from collections.abc import AsyncIterable
 from typing import TYPE_CHECKING, final
 
 import pytest_asyncio
@@ -31,6 +30,7 @@ from h2.settings import SettingCodes
 from typing_extensions import override
 
 if TYPE_CHECKING:
+    from collections.abc import AsyncIterable
     from typing import Any
 
 

test_runner/fixtures/metrics.py

@@ -1,12 +1,15 @@
 from __future__ import annotations
 
 from collections import defaultdict
+from typing import TYPE_CHECKING
 
 from prometheus_client.parser import text_string_to_metric_families
-from prometheus_client.samples import Sample
 
 from fixtures.log_helper import log
 
+if TYPE_CHECKING:
+    from prometheus_client.samples import Sample
+
 
 class Metrics:
     metrics: dict[str, list[Sample]]
@@ -168,7 +171,6 @@ PAGESERVER_PER_TENANT_METRICS: tuple[str, ...] = (
     "pageserver_evictions_with_low_residence_duration_total",
     "pageserver_aux_file_estimated_size",
     "pageserver_valid_lsn_lease_count",
-    "pageserver_flush_wait_upload_seconds",
     counter("pageserver_tenant_throttling_count_accounted_start"),
     counter("pageserver_tenant_throttling_count_accounted_finish"),
     counter("pageserver_tenant_throttling_wait_usecs_sum"),

test_runner/fixtures/neon_cli.py

@@ -7,7 +7,6 @@ import subprocess
 import tempfile
 import textwrap
 from itertools import chain, product
-from pathlib import Path
 from typing import TYPE_CHECKING, cast
 
 import toml
@@ -15,14 +14,15 @@ import toml
 from fixtures.common_types import Lsn, TenantId, TimelineId
 from fixtures.log_helper import log
 from fixtures.pageserver.common_types import IndexPartDump
-from fixtures.pg_version import PgVersion
 
 if TYPE_CHECKING:
+    from pathlib import Path
     from typing import (
         Any,
-        cast,
     )
 
+    from fixtures.pg_version import PgVersion
+
 
 # Used to be an ABC. abc.ABC removed due to linter without name change.
 class AbstractNeonCli:
@@ -36,7 +36,7 @@ class AbstractNeonCli:
         self.extra_env = extra_env
         self.binpath = binpath
 
-    COMMAND: str = cast(str, None)  # To be overwritten by the derived class.
+    COMMAND: str = cast("str", None)  # To be overwritten by the derived class.
 
     def raw_cli(
         self,