Merge pull request #9832 from neondatabase/rc/release-proxy/2024-11-21

Proxy release 2024-11-21

.github/actions/allure-report-generate/action.yml

@@ -7,6 +7,10 @@ inputs:
     type: boolean
     required: false
     default: false
+  aws_oicd_role_arn:
+    description: 'the OIDC role arn to (re-)acquire for allure report upload - if not set call must acquire OIDC role'
+    required: false
+    default: ''
 
 outputs:
   base-url:
@@ -79,6 +83,14 @@ runs:
         ALLURE_VERSION: 2.27.0
         ALLURE_ZIP_SHA256: b071858fb2fa542c65d8f152c5c40d26267b2dfb74df1f1608a589ecca38e777
 
+    - name: (Re-)configure AWS credentials # necessary to upload reports to S3 after a long-running test
+      if: ${{ !cancelled() && (inputs.aws_oicd_role_arn != '') }}
+      uses: aws-actions/configure-aws-credentials@v4
+      with:
+        aws-region: eu-central-1
+        role-to-assume: ${{ inputs.aws_oicd_role_arn }}
+        role-duration-seconds: 3600 # 1 hour should be more than enough to upload report
+
     # Potentially we could have several running build for the same key (for example, for the main branch), so we use improvised lock for this
     - name: Acquire lock
       shell: bash -euxo pipefail {0}

.github/actions/allure-report-store/action.yml

@@ -8,6 +8,10 @@ inputs:
   unique-key:
     description: 'string to distinguish different results in the same run'
     required: true
+  aws_oicd_role_arn:
+    description: 'the OIDC role arn to (re-)acquire for allure report upload - if not set call must acquire OIDC role'
+    required: false
+    default: ''
 
 runs:
   using: "composite"
@@ -31,6 +35,14 @@ runs:
       env:
         REPORT_DIR: ${{ inputs.report-dir }}
 
+    - name: (Re-)configure AWS credentials # necessary to upload reports to S3 after a long-running test
+      if: ${{ !cancelled() && (inputs.aws_oicd_role_arn != '') }}
+      uses: aws-actions/configure-aws-credentials@v4
+      with:
+        aws-region: eu-central-1
+        role-to-assume: ${{ inputs.aws_oicd_role_arn }}
+        role-duration-seconds: 3600 # 1 hour should be more than enough to upload report
+
     - name: Upload test results
       shell: bash -euxo pipefail {0}
       run: |

.github/actions/run-python-test-set/action.yml

@@ -48,6 +48,10 @@ inputs:
     description: 'benchmark durations JSON'
     required: false
     default: '{}'
+  aws_oicd_role_arn:
+    description: 'the OIDC role arn to (re-)acquire for allure report upload - if not set call must acquire OIDC role'
+    required: false
+    default: ''
 
 runs:
   using: "composite"
@@ -222,6 +226,13 @@ runs:
         # (for example if we didn't run the test for non build-and-test workflow)
         skip-if-does-not-exist: true
 
+    - name: (Re-)configure AWS credentials # necessary to upload reports to S3 after a long-running test
+      if: ${{ !cancelled() && (inputs.aws_oicd_role_arn != '') }}
+      uses: aws-actions/configure-aws-credentials@v4
+      with:
+        aws-region: eu-central-1
+        role-to-assume: ${{ inputs.aws_oicd_role_arn }}
+        role-duration-seconds: 3600 # 1 hour should be more than enough to upload report
     - name: Upload test results
       if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report-store

.github/workflows/_create-release-pr.yml

@@ -0,0 +1,79 @@
+name: Create Release PR
+
+on:
+  workflow_call:
+    inputs:
+      component-name:
+        description: 'Component name'
+        required: true
+        type: string
+      release-branch:
+        description: 'Release branch'
+        required: true
+        type: string
+    secrets:
+      ci-access-token:
+        description: 'CI access token'
+        required: true
+
+defaults:
+  run:
+    shell: bash -euo pipefail {0}
+
+jobs:
+  create-storage-release-branch:
+    runs-on: ubuntu-22.04
+
+    permissions:
+      contents: write # for `git push`
+
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        ref: main
+
+    - name: Set variables
+      id: vars
+      env:
+        COMPONENT_NAME: ${{ inputs.component-name }}
+        RELEASE_BRANCH: ${{ inputs.release-branch }}
+      run: |
+        today=$(date +'%Y-%m-%d')
+        echo "title=${COMPONENT_NAME} release ${today}" | tee -a ${GITHUB_OUTPUT}
+        echo "rc-branch=rc/${RELEASE_BRANCH}/${today}"  | tee -a ${GITHUB_OUTPUT}
+
+    - name: Configure git
+      run: |
+        git config user.name "github-actions[bot]"
+        git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+
+    - name: Create RC branch
+      env:
+        RC_BRANCH: ${{ steps.vars.outputs.rc-branch }}
+        TITLE: ${{ steps.vars.outputs.title }}
+      run: |
+        git checkout -b "${RC_BRANCH}"
+
+        # create an empty commit to distinguish workflow runs
+        # from other possible releases from the same commit
+        git commit --allow-empty -m "${TITLE}"
+
+        git push origin "${RC_BRANCH}"
+
+    - name: Create a PR into ${{ inputs.release-branch }}
+      env:
+        GH_TOKEN: ${{ secrets.ci-access-token }}
+        RC_BRANCH: ${{ steps.vars.outputs.rc-branch }}
+        RELEASE_BRANCH: ${{ inputs.release-branch }}
+        TITLE: ${{ steps.vars.outputs.title }}
+      run: |
+        cat << EOF > body.md
+          ## ${TITLE}
+
+          **Please merge this Pull Request using 'Create a merge commit' button**
+        EOF
+
+        gh pr create --title "${TITLE}" \
+                     --body-file "body.md" \
+                     --head "${RC_BRANCH}" \
+                     --base "${RELEASE_BRANCH}"

.github/workflows/benchmarking.yml

@@ -122,6 +122,7 @@ jobs:
         run_in_parallel: false
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
         # Set --sparse-ordering option of pytest-order plugin
         # to ensure tests are running in order of appears in the file.
         # It's important for test_perf_pgbench.py::test_pgbench_remote_* tests
@@ -133,6 +134,7 @@ jobs:
           --ignore test_runner/performance/test_perf_pgvector_queries.py
           --ignore test_runner/performance/test_logical_replication.py
           --ignore test_runner/performance/test_physical_replication.py
+          --ignore test_runner/performance/test_perf_ingest_using_pgcopydb.py
       env:
         BENCHMARK_CONNSTR: ${{ steps.create-neon-project.outputs.dsn }}
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
@@ -149,12 +151,14 @@ jobs:
       id: create-allure-report
       if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report-generate
+      with:
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
 
     - name: Post to a Slack channel
       if: ${{ github.event.schedule && failure() }}
       uses: slackapi/slack-github-action@v1
       with:
-        channel-id: "C033QLM5P7D" # dev-staging-stream
+        channel-id: "C06KHQVQ7U3" # on-call-qa-staging-stream
         slack-message: |
           Periodic perf testing: ${{ job.status }}
           <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
@@ -210,6 +214,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 5400
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
         PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
@@ -226,6 +231,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 5400
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
         PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
@@ -237,11 +243,13 @@ jobs:
       uses: ./.github/actions/allure-report-generate
       with:
         store-test-results-into-db: true
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         REGRESS_TEST_RESULT_CONNSTR_NEW: ${{ secrets.REGRESS_TEST_RESULT_CONNSTR_NEW }}
 
+    # Post both success and failure to the Slack channel
     - name: Post to a Slack channel
-      if: ${{ github.event.schedule && failure() }}
+      if: ${{ github.event.schedule }}
       uses: slackapi/slack-github-action@v1
       with:
         channel-id: "C06T9AMNDQQ" # on-call-compute-staging-stream
@@ -444,6 +452,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 21600 -k test_pgbench_remote_init
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         BENCHMARK_CONNSTR: ${{ steps.set-up-connstr.outputs.connstr }}
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
@@ -458,6 +467,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 21600 -k test_pgbench_remote_simple_update
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         BENCHMARK_CONNSTR: ${{ steps.set-up-connstr.outputs.connstr }}
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
@@ -472,6 +482,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 21600 -k test_pgbench_remote_select_only
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         BENCHMARK_CONNSTR: ${{ steps.set-up-connstr.outputs.connstr }}
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
@@ -488,12 +499,14 @@ jobs:
       id: create-allure-report
       if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report-generate
+      with:
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
 
     - name: Post to a Slack channel
       if: ${{ github.event.schedule && failure() }}
       uses: slackapi/slack-github-action@v1
       with:
-        channel-id: "C033QLM5P7D" # dev-staging-stream
+        channel-id: "C06KHQVQ7U3" # on-call-qa-staging-stream
         slack-message: |
           Periodic perf testing on ${{ matrix.platform }}: ${{ job.status }}
           <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
@@ -545,12 +558,12 @@ jobs:
         arch=$(uname -m | sed 's/x86_64/amd64/g' | sed 's/aarch64/arm64/g')
 
         cd /home/nonroot
-        wget -q "https://apt.postgresql.org/pub/repos/apt/pool/main/p/postgresql-17/libpq5_17.0-1.pgdg110+1_${arch}.deb"
-        wget -q "https://apt.postgresql.org/pub/repos/apt/pool/main/p/postgresql-16/postgresql-client-16_16.4-1.pgdg110+2_${arch}.deb"
-        wget -q "https://apt.postgresql.org/pub/repos/apt/pool/main/p/postgresql-16/postgresql-16_16.4-1.pgdg110+2_${arch}.deb"
-        dpkg -x libpq5_17.0-1.pgdg110+1_${arch}.deb pg
-        dpkg -x postgresql-16_16.4-1.pgdg110+2_${arch}.deb pg
-        dpkg -x postgresql-client-16_16.4-1.pgdg110+2_${arch}.deb pg
+        wget -q "https://apt.postgresql.org/pub/repos/apt/pool/main/p/postgresql-17/libpq5_17.1-1.pgdg110+1_${arch}.deb"
+        wget -q "https://apt.postgresql.org/pub/repos/apt/pool/main/p/postgresql-16/postgresql-client-16_16.5-1.pgdg110+1_${arch}.deb"
+        wget -q "https://apt.postgresql.org/pub/repos/apt/pool/main/p/postgresql-16/postgresql-16_16.5-1.pgdg110+1_${arch}.deb"
+        dpkg -x libpq5_17.1-1.pgdg110+1_${arch}.deb pg
+        dpkg -x postgresql-16_16.5-1.pgdg110+1_${arch}.deb pg
+        dpkg -x postgresql-client-16_16.5-1.pgdg110+1_${arch}.deb pg
 
         mkdir -p /tmp/neon/pg_install/v16/bin
         ln -s /home/nonroot/pg/usr/lib/postgresql/16/bin/pgbench /tmp/neon/pg_install/v16/bin/pgbench
@@ -598,6 +611,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 21600 -k test_pgvector_indexing
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
         PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
@@ -612,6 +626,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 21600
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         BENCHMARK_CONNSTR: ${{ steps.set-up-connstr.outputs.connstr }}
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
@@ -621,12 +636,14 @@ jobs:
       id: create-allure-report
       if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report-generate
+      with:
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
 
     - name: Post to a Slack channel
       if: ${{ github.event.schedule && failure() }}
       uses: slackapi/slack-github-action@v1
       with:
-        channel-id: "C033QLM5P7D" # dev-staging-stream
+        channel-id: "C06KHQVQ7U3" # on-call-qa-staging-stream
         slack-message: |
           Periodic perf testing on ${{ env.PLATFORM }}: ${{ job.status }}
           <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
@@ -722,6 +739,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 43200 -k test_clickbench
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
         PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
@@ -734,12 +752,14 @@ jobs:
       id: create-allure-report
       if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report-generate
+      with:
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
 
     - name: Post to a Slack channel
       if: ${{ github.event.schedule && failure() }}
       uses: slackapi/slack-github-action@v1
       with:
-        channel-id: "C033QLM5P7D" # dev-staging-stream
+        channel-id: "C06KHQVQ7U3" # on-call-qa-staging-stream
         slack-message: |
           Periodic OLAP perf testing on ${{ matrix.platform }}: ${{ job.status }}
           <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
@@ -836,6 +856,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 21600 -k test_tpch
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
         PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
@@ -846,12 +867,14 @@ jobs:
       id: create-allure-report
       if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report-generate
+      with:
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
 
     - name: Post to a Slack channel
       if: ${{ github.event.schedule && failure() }}
       uses: slackapi/slack-github-action@v1
       with:
-        channel-id: "C033QLM5P7D" # dev-staging-stream
+        channel-id: "C06KHQVQ7U3" # on-call-qa-staging-stream
         slack-message: |
           Periodic TPC-H perf testing on ${{ matrix.platform }}: ${{ job.status }}
           <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
@@ -934,6 +957,7 @@ jobs:
         save_perf_report: ${{ env.SAVE_PERF_REPORT }}
         extra_params: -m remote_cluster --timeout 21600 -k test_user_examples
         pg_version: ${{ env.DEFAULT_PG_VERSION }}
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
         VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
         PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
@@ -943,12 +967,14 @@ jobs:
       id: create-allure-report
       if: ${{ !cancelled() }}
       uses: ./.github/actions/allure-report-generate
+      with:
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
 
     - name: Post to a Slack channel
       if: ${{ github.event.schedule && failure() }}
       uses: slackapi/slack-github-action@v1
       with:
-        channel-id: "C033QLM5P7D" # dev-staging-stream
+        channel-id: "C06KHQVQ7U3" # on-call-qa-staging-stream
         slack-message: |
           Periodic TPC-H perf testing on ${{ matrix.platform }}: ${{ job.status }}
           <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>

.github/workflows/build-build-tools-image.yml

@@ -2,18 +2,13 @@ name: Build build-tools image
 
 on:
   workflow_call:
-    inputs:
-      image-tag:
-        description: "build-tools image tag"
-        required: true
-        type: string
     outputs:
       image-tag:
         description: "build-tools tag"
-        value: ${{ inputs.image-tag }}
+        value: ${{ jobs.check-image.outputs.tag }}
       image:
         description: "build-tools image"
-        value: neondatabase/build-tools:${{ inputs.image-tag }}
+        value: neondatabase/build-tools:${{ jobs.check-image.outputs.tag }}
 
 defaults:
   run:
@@ -35,7 +30,36 @@ permissions: {}
 
 jobs:
   check-image:
-    uses: ./.github/workflows/check-build-tools-image.yml
+    runs-on: ubuntu-22.04
+    outputs:
+      tag: ${{ steps.get-build-tools-tag.outputs.image-tag }}
+      found: ${{ steps.check-image.outputs.found }}
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Get build-tools image tag for the current commit
+        id: get-build-tools-tag
+        env:
+          IMAGE_TAG: |
+            ${{ hashFiles('build-tools.Dockerfile',
+                          '.github/workflows/build-build-tools-image.yml') }}
+        run: |
+          echo "image-tag=${IMAGE_TAG}" | tee -a $GITHUB_OUTPUT
+
+      - name: Check if such tag found in the registry
+        id: check-image
+        env:
+          IMAGE_TAG: ${{ steps.get-build-tools-tag.outputs.image-tag }}
+        run: |
+          if docker manifest inspect neondatabase/build-tools:${IMAGE_TAG}; then
+            found=true
+          else
+            found=false
+          fi
+
+          echo "found=${found}" | tee -a $GITHUB_OUTPUT
+
 
   build-image:
     needs: [ check-image ]
@@ -48,20 +72,7 @@ jobs:
 
     runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}
 
-    env:
-      IMAGE_TAG: ${{ inputs.image-tag }}
-
     steps:
-      - name: Check `input.tag` is correct
-        env:
-          INPUTS_IMAGE_TAG: ${{ inputs.image-tag }}
-          CHECK_IMAGE_TAG : ${{ needs.check-image.outputs.image-tag }}
-        run: |
-          if [ "${INPUTS_IMAGE_TAG}" != "${CHECK_IMAGE_TAG}" ]; then
-            echo "'inputs.image-tag' (${INPUTS_IMAGE_TAG}) does not match the tag of the latest build-tools image 'inputs.image-tag' (${CHECK_IMAGE_TAG})"
-            exit 1
-          fi
-
       - uses: actions/checkout@v4
 
       - uses: neondatabase/dev-actions/set-docker-config-dir@6094485bf440001c94a94a3f9e221e81ff6b6193
@@ -92,10 +103,10 @@ jobs:
           cache-from: type=registry,ref=cache.neon.build/build-tools:cache-${{ matrix.debian-version }}-${{ matrix.arch }}
           cache-to: ${{ github.ref_name == 'main' && format('type=registry,ref=cache.neon.build/build-tools:cache-{0}-{1},mode=max', matrix.debian-version, matrix.arch) || '' }}
           tags: |
-            neondatabase/build-tools:${{ inputs.image-tag }}-${{ matrix.debian-version }}-${{ matrix.arch }}
+            neondatabase/build-tools:${{ needs.check-image.outputs.tag }}-${{ matrix.debian-version }}-${{ matrix.arch }}
 
   merge-images:
-    needs: [ build-image ]
+    needs: [ check-image, build-image ]
     runs-on: ubuntu-22.04
 
     steps:
@@ -107,7 +118,7 @@ jobs:
       - name: Create multi-arch image
         env:
           DEFAULT_DEBIAN_VERSION: bullseye
-          IMAGE_TAG: ${{ inputs.image-tag }}
+          IMAGE_TAG: ${{ needs.check-image.outputs.tag }}
         run: |
           for debian_version in bullseye bookworm; do
             tags=("-t" "neondatabase/build-tools:${IMAGE_TAG}-${debian_version}")

.github/workflows/build_and_test.yml

@@ -77,15 +77,9 @@ jobs:
         shell: bash
         id: build-tag
 
-  check-build-tools-image:
-    needs: [ check-permissions ]
-    uses: ./.github/workflows/check-build-tools-image.yml
-
   build-build-tools-image:
-    needs: [ check-build-tools-image ]
+    needs: [ check-permissions ]
     uses: ./.github/workflows/build-build-tools-image.yml
-    with:
-      image-tag: ${{ needs.check-build-tools-image.outputs.image-tag }}
     secrets: inherit
 
   check-codestyle-python:

.github/workflows/check-build-tools-image.yml

@@ -1,51 +0,0 @@
-name: Check build-tools image
-
-on:
-  workflow_call:
-    outputs:
-      image-tag:
-        description: "build-tools image tag"
-        value: ${{ jobs.check-image.outputs.tag }}
-      found:
-        description: "Whether the image is found in the registry"
-        value: ${{ jobs.check-image.outputs.found }}
-
-defaults:
-  run:
-    shell: bash -euo pipefail {0}
-
-# No permission for GITHUB_TOKEN by default; the **minimal required** set of permissions should be granted in each job.
-permissions: {}
-
-jobs:
-  check-image:
-    runs-on: ubuntu-22.04
-    outputs:
-      tag: ${{ steps.get-build-tools-tag.outputs.image-tag }}
-      found: ${{ steps.check-image.outputs.found }}
-
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Get build-tools image tag for the current commit
-        id: get-build-tools-tag
-        env:
-          IMAGE_TAG: |
-            ${{ hashFiles('build-tools.Dockerfile',
-                          '.github/workflows/check-build-tools-image.yml',
-                          '.github/workflows/build-build-tools-image.yml') }}
-        run: |
-          echo "image-tag=${IMAGE_TAG}" | tee -a $GITHUB_OUTPUT
-
-      - name: Check if such tag found in the registry
-        id: check-image
-        env:
-          IMAGE_TAG: ${{ steps.get-build-tools-tag.outputs.image-tag }}
-        run: |
-          if docker manifest inspect neondatabase/build-tools:${IMAGE_TAG}; then
-            found=true
-          else
-            found=false
-          fi
-
-          echo "found=${found}" | tee -a $GITHUB_OUTPUT

.github/workflows/ingest_benchmark.yml

@@ -1,4 +1,4 @@
-name: Benchmarking
+name: benchmarking ingest
 
 on:
   # uncomment to run on push for debugging your PR
@@ -74,18 +74,16 @@ jobs:
         compute_units: '[7, 7]' # we want to test large compute here to avoid compute-side bottleneck
         api_key: ${{ secrets.NEON_STAGING_API_KEY }}
 
-    - name: Initialize Neon project and retrieve current backpressure seconds
+    - name: Initialize Neon project
       if: ${{ matrix.target_project == 'new_empty_project' }}
       env:
-          NEW_PROJECT_CONNSTR: ${{ steps.create-neon-project-ingest-target.outputs.dsn }}
+          BENCHMARK_INGEST_TARGET_CONNSTR: ${{ steps.create-neon-project-ingest-target.outputs.dsn }}
           NEW_PROJECT_ID: ${{ steps.create-neon-project-ingest-target.outputs.project_id }}
       run: |
         echo "Initializing Neon project with project_id: ${NEW_PROJECT_ID}"
         export LD_LIBRARY_PATH=${PG_16_LIB_PATH}
-        ${PSQL} "${NEW_PROJECT_CONNSTR}" -c "CREATE EXTENSION IF NOT EXISTS neon; CREATE EXTENSION IF NOT EXISTS neon_utils;"
-        BACKPRESSURE_TIME_BEFORE_INGEST=$(${PSQL} "${NEW_PROJECT_CONNSTR}" -t -c "select backpressure_throttling_time()/1000000;")
-        echo "BACKPRESSURE_TIME_BEFORE_INGEST=${BACKPRESSURE_TIME_BEFORE_INGEST}" >> $GITHUB_ENV
-        echo "NEW_PROJECT_CONNSTR=${NEW_PROJECT_CONNSTR}" >> $GITHUB_ENV
+        ${PSQL} "${BENCHMARK_INGEST_TARGET_CONNSTR}" -c "CREATE EXTENSION IF NOT EXISTS neon; CREATE EXTENSION IF NOT EXISTS neon_utils;"
+        echo "BENCHMARK_INGEST_TARGET_CONNSTR=${BENCHMARK_INGEST_TARGET_CONNSTR}" >> $GITHUB_ENV
 
     - name: Create Neon Branch for large tenant
       if: ${{ matrix.target_project == 'large_existing_project' }}
@@ -95,266 +93,55 @@ jobs:
         project_id: ${{ vars.BENCHMARK_INGEST_TARGET_PROJECTID }}
         api_key: ${{ secrets.NEON_STAGING_API_KEY }}
 
-    - name: Initialize Neon project and retrieve current backpressure seconds
+    - name: Initialize Neon project 
       if: ${{ matrix.target_project == 'large_existing_project' }}
       env:
-          NEW_PROJECT_CONNSTR: ${{ steps.create-neon-branch-ingest-target.outputs.dsn }}
+          BENCHMARK_INGEST_TARGET_CONNSTR: ${{ steps.create-neon-branch-ingest-target.outputs.dsn }}
           NEW_BRANCH_ID: ${{ steps.create-neon-branch-ingest-target.outputs.branch_id }}
       run: |
         echo "Initializing Neon branch with branch_id: ${NEW_BRANCH_ID}"
         export LD_LIBRARY_PATH=${PG_16_LIB_PATH}
         # Extract the part before the database name
-        base_connstr="${NEW_PROJECT_CONNSTR%/*}"
+        base_connstr="${BENCHMARK_INGEST_TARGET_CONNSTR%/*}"
         # Extract the query parameters (if any) after the database name
-        query_params="${NEW_PROJECT_CONNSTR#*\?}"
+        query_params="${BENCHMARK_INGEST_TARGET_CONNSTR#*\?}"
         # Reconstruct the new connection string
-        if [ "$query_params" != "$NEW_PROJECT_CONNSTR" ]; then
+        if [ "$query_params" != "$BENCHMARK_INGEST_TARGET_CONNSTR" ]; then
           new_connstr="${base_connstr}/neondb?${query_params}"
         else
           new_connstr="${base_connstr}/neondb"
         fi
         ${PSQL} "${new_connstr}" -c "drop database ludicrous;"
         ${PSQL} "${new_connstr}" -c "CREATE DATABASE ludicrous;"
-        if [ "$query_params" != "$NEW_PROJECT_CONNSTR" ]; then
-          NEW_PROJECT_CONNSTR="${base_connstr}/ludicrous?${query_params}"
+        if [ "$query_params" != "$BENCHMARK_INGEST_TARGET_CONNSTR" ]; then
+          BENCHMARK_INGEST_TARGET_CONNSTR="${base_connstr}/ludicrous?${query_params}"
         else
-          NEW_PROJECT_CONNSTR="${base_connstr}/ludicrous"
+          BENCHMARK_INGEST_TARGET_CONNSTR="${base_connstr}/ludicrous"
         fi
-        ${PSQL} "${NEW_PROJECT_CONNSTR}" -c "CREATE EXTENSION IF NOT EXISTS neon; CREATE EXTENSION IF NOT EXISTS neon_utils;"
-        BACKPRESSURE_TIME_BEFORE_INGEST=$(${PSQL} "${NEW_PROJECT_CONNSTR}" -t -c "select backpressure_throttling_time()/1000000;")
-        echo "BACKPRESSURE_TIME_BEFORE_INGEST=${BACKPRESSURE_TIME_BEFORE_INGEST}" >> $GITHUB_ENV
-        echo "NEW_PROJECT_CONNSTR=${NEW_PROJECT_CONNSTR}" >> $GITHUB_ENV
-      
-        
-    - name: Create pgcopydb filter file
-      run: |
-        cat << EOF > /tmp/pgcopydb_filter.txt
-          [include-only-table]
-          public.events
-          public.emails
-          public.email_transmissions
-          public.payments
-          public.editions
-          public.edition_modules
-          public.sp_content
-          public.email_broadcasts
-          public.user_collections
-          public.devices
-          public.user_accounts
-          public.lessons
-          public.lesson_users
-          public.payment_methods
-          public.orders
-          public.course_emails
-          public.modules
-          public.users
-          public.module_users
-          public.courses
-          public.payment_gateway_keys
-          public.accounts
-          public.roles
-          public.payment_gateways
-          public.management
-          public.event_names
-        EOF
+        ${PSQL} "${BENCHMARK_INGEST_TARGET_CONNSTR}" -c "CREATE EXTENSION IF NOT EXISTS neon; CREATE EXTENSION IF NOT EXISTS neon_utils;"
+        echo "BENCHMARK_INGEST_TARGET_CONNSTR=${BENCHMARK_INGEST_TARGET_CONNSTR}" >> $GITHUB_ENV
 
-    - name: Invoke pgcopydb
+    - name: Invoke pgcopydb  
+      uses: ./.github/actions/run-python-test-set
+      with:
+        build_type: remote
+        test_selection: performance/test_perf_ingest_using_pgcopydb.py
+        run_in_parallel: false
+        extra_params: -s -m remote_cluster --timeout 86400 -k test_ingest_performance_using_pgcopydb
+        pg_version: v16
+        save_perf_report: true
+        aws_oicd_role_arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
       env:
-          BENCHMARK_INGEST_SOURCE_CONNSTR: ${{ secrets.BENCHMARK_INGEST_SOURCE_CONNSTR }}
-      run: |
-        export LD_LIBRARY_PATH=${PGCOPYDB_LIB_PATH}:${PG_16_LIB_PATH}
-        export PGCOPYDB_SOURCE_PGURI="${BENCHMARK_INGEST_SOURCE_CONNSTR}"
-        export PGCOPYDB_TARGET_PGURI="${NEW_PROJECT_CONNSTR}"
-        export PGOPTIONS="-c maintenance_work_mem=8388608 -c max_parallel_maintenance_workers=7"
-        ${PG_CONFIG} --bindir
-        ${PGCOPYDB} --version
-        ${PGCOPYDB} clone --skip-vacuum  --no-owner --no-acl --skip-db-properties --table-jobs 4 \
-          --index-jobs 4 --restore-jobs 4 --split-tables-larger-than 10GB --skip-extensions \
-          --use-copy-binary --filters /tmp/pgcopydb_filter.txt 2>&1 | tee /tmp/pgcopydb_${{ matrix.target_project }}.log
+        BENCHMARK_INGEST_SOURCE_CONNSTR: ${{ secrets.BENCHMARK_INGEST_SOURCE_CONNSTR }}
+        TARGET_PROJECT_TYPE: ${{ matrix.target_project }}
+        # we report PLATFORM in zenbenchmark NeonBenchmarker perf database and want to distinguish between new project and large tenant
+        PLATFORM: "${{ matrix.target_project }}-us-east-2-staging"
+        PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
 
-    # create dummy pgcopydb log to test parsing
-    # - name: create dummy log for parser test
-    #   run: |
-    #     cat << EOF > /tmp/pgcopydb_${{ matrix.target_project }}.log
-    #     2024-11-04 18:00:53.433 500861 INFO   main.c:136                Running pgcopydb version 0.17.10.g8361a93 from "/usr/lib/postgresql/17/bin/pgcopydb"
-    #     2024-11-04 18:00:53.434 500861 INFO   cli_common.c:1225         [SOURCE] Copying database from "postgres://neondb_owner@ep-bitter-shape-w2c1ir0a.us-east-2.aws.neon.build/neondb?sslmode=require&keepalives=1&keepalives_idle=10&keepalives_interval=10&keepalives_count=60"
-    #     2024-11-04 18:00:53.434 500861 INFO   cli_common.c:1226         [TARGET] Copying database into "postgres://neondb_owner@ep-icy-union-w25qd5pj.us-east-2.aws.neon.build/ludicrous?sslmode=require&keepalives=1&keepalives_idle=10&keepalives_interval=10&keepalives_count=60"
-    #     2024-11-04 18:00:53.442 500861 INFO   copydb.c:105              Using work dir "/tmp/pgcopydb"
-    #     2024-11-04 18:00:53.541 500861 INFO   snapshot.c:107            Exported snapshot "00000008-00000033-1" from the source database
-    #     2024-11-04 18:00:53.556 500865 INFO   cli_clone_follow.c:543    STEP 1: fetch source database tables, indexes, and sequences
-    #     2024-11-04 18:00:54.570 500865 INFO   copydb_schema.c:716       Splitting source candidate tables larger than 10 GB
-    #     2024-11-04 18:00:54.570 500865 INFO   copydb_schema.c:829       Table public.events is 96 GB large which is larger than --split-tables-larger-than 10 GB, and does not have a unique column of type integer: splitting by CTID
-    #     2024-11-04 18:01:05.538 500865 INFO   copydb_schema.c:905       Table public.events is 96 GB large, 10 COPY processes will be used, partitioning on ctid.
-    #     2024-11-04 18:01:05.564 500865 INFO   copydb_schema.c:905       Table public.email_transmissions is 27 GB large, 4 COPY processes will be used, partitioning on id.
-    #     2024-11-04 18:01:05.584 500865 INFO   copydb_schema.c:905       Table public.lessons is 25 GB large, 4 COPY processes will be used, partitioning on id.
-    #     2024-11-04 18:01:05.605 500865 INFO   copydb_schema.c:905       Table public.lesson_users is 16 GB large, 3 COPY processes will be used, partitioning on id.
-    #     2024-11-04 18:01:05.605 500865 INFO   copydb_schema.c:761       Fetched information for 26 tables (including 4 tables split in 21 partitions total), with an estimated total of 907 million tuples and 175 GB on-disk
-    #     2024-11-04 18:01:05.687 500865 INFO   copydb_schema.c:968       Fetched information for 57 indexes (supporting 25 constraints)
-    #     2024-11-04 18:01:05.753 500865 INFO   sequences.c:78            Fetching information for 24 sequences
-    #     2024-11-04 18:01:05.903 500865 INFO   copydb_schema.c:1122      Fetched information for 4 extensions
-    #     2024-11-04 18:01:06.178 500865 INFO   copydb_schema.c:1538      Found 0 indexes (supporting 0 constraints) in the target database
-    #     2024-11-04 18:01:06.184 500865 INFO   cli_clone_follow.c:584    STEP 2: dump the source database schema (pre/post data)
-    #     2024-11-04 18:01:06.186 500865 INFO   pgcmd.c:468                /usr/lib/postgresql/16/bin/pg_dump -Fc --snapshot 00000008-00000033-1 --section=pre-data --section=post-data --file /tmp/pgcopydb/schema/schema.dump 'postgres://neondb_owner@ep-bitter-shape-w2c1ir0a.us-east-2.aws.neon.build/neondb?sslmode=require&keepalives=1&keepalives_idle=10&keepalives_interval=10&keepalives_count=60'
-    #     2024-11-04 18:01:06.952 500865 INFO   cli_clone_follow.c:592    STEP 3: restore the pre-data section to the target database
-    #     2024-11-04 18:01:07.004 500865 INFO   pgcmd.c:1001               /usr/lib/postgresql/16/bin/pg_restore --dbname 'postgres://neondb_owner@ep-icy-union-w25qd5pj.us-east-2.aws.neon.build/ludicrous?sslmode=require&keepalives=1&keepalives_idle=10&keepalives_interval=10&keepalives_count=60' --section pre-data --jobs 4 --no-owner --no-acl --use-list /tmp/pgcopydb/schema/pre-filtered.list /tmp/pgcopydb/schema/schema.dump
-    #     2024-11-04 18:01:07.438 500874 INFO   table-data.c:656          STEP 4: starting 4 table-data COPY processes
-    #     2024-11-04 18:01:07.451 500877 INFO   vacuum.c:139              STEP 8: skipping VACUUM jobs per --skip-vacuum
-    #     2024-11-04 18:01:07.457 500875 INFO   indexes.c:182             STEP 6: starting 4 CREATE INDEX processes
-    #     2024-11-04 18:01:07.457 500875 INFO   indexes.c:183             STEP 7: constraints are built by the CREATE INDEX processes
-    #     2024-11-04 18:01:07.507 500865 INFO   blobs.c:74                Skipping large objects: none found.
-    #     2024-11-04 18:01:07.509 500865 INFO   sequences.c:194           STEP 9: reset sequences values
-    #     2024-11-04 18:01:07.510 500886 INFO   sequences.c:290           Set sequences values on the target database
-    #     2024-11-04 20:49:00.587 500865 INFO   cli_clone_follow.c:608    STEP 10: restore the post-data section to the target database
-    #     2024-11-04 20:49:00.600 500865 INFO   pgcmd.c:1001               /usr/lib/postgresql/16/bin/pg_restore --dbname 'postgres://neondb_owner@ep-icy-union-w25qd5pj.us-east-2.aws.neon.build/ludicrous?sslmode=require&keepalives=1&keepalives_idle=10&keepalives_interval=10&keepalives_count=60' --section post-data --jobs 4 --no-owner --no-acl --use-list /tmp/pgcopydb/schema/post-filtered.list /tmp/pgcopydb/schema/schema.dump
-    #     2024-11-05 10:50:58.508 500865 INFO   cli_clone_follow.c:639    All step are now done, 16h49m elapsed
-    #     2024-11-05 10:50:58.508 500865 INFO   summary.c:3155            Printing summary for 26 tables and 57 indexes
-
-    #       OID | Schema |                 Name | Parts | copy duration | transmitted bytes | indexes | create index duration 
-    #     ------+--------+----------------------+-------+---------------+-------------------+---------+----------------------
-    #     24654 | public |               events |    10 |         1d11h |            878 GB |       1 |                 1h41m
-    #     24623 | public |  email_transmissions |     4 |         4h46m |             99 GB |       3 |                 2h04m
-    #     24665 | public |              lessons |     4 |         4h42m |            161 GB |       4 |                 1m11s
-    #     24661 | public |         lesson_users |     3 |         2h46m |             49 GB |       3 |                39m35s
-    #     24631 | public |               emails |     1 |        34m07s |             10 GB |       2 |                   17s
-    #     24739 | public |             payments |     1 |         5m47s |           1848 MB |       4 |                 4m40s
-    #     24681 | public |         module_users |     1 |         4m57s |           1610 MB |       3 |                 1m50s
-    #     24694 | public |               orders |     1 |         2m50s |            835 MB |       3 |                 1m05s
-    #     24597 | public |              devices |     1 |         1m45s |            498 MB |       2 |                   40s
-    #     24723 | public |      payment_methods |     1 |         1m24s |            548 MB |       2 |                   31s
-    #     24765 | public |     user_collections |     1 |         2m17s |           1005 MB |       2 |                 968ms
-    #     24774 | public |                users |     1 |           52s |            291 MB |       4 |                   27s
-    #     24760 | public |        user_accounts |     1 |           16s |            172 MB |       3 |                   16s
-    #     24606 | public |      edition_modules |     1 |         8s983 |             46 MB |       3 |                 4s749
-    #     24583 | public |        course_emails |     1 |         8s526 |             26 MB |       2 |                 996ms
-    #     24685 | public |              modules |     1 |         1s592 |             21 MB |       3 |                 1s696
-    #     24610 | public |             editions |     1 |         2s199 |           7483 kB |       2 |                 1s032
-    #     24755 | public |           sp_content |     1 |         1s555 |           4177 kB |       0 |                   0ms
-    #     24619 | public |     email_broadcasts |     1 |         744ms |           2645 kB |       2 |                 677ms
-    #     24590 | public |              courses |     1 |         387ms |           1540 kB |       2 |                 367ms
-    #     24704 | public | payment_gateway_keys |     1 |         1s972 |            164 kB |       2 |                  27ms
-    #     24576 | public |             accounts |     1 |          58ms |             24 kB |       1 |                  14ms
-    #     24647 | public |          event_names |     1 |          32ms |             397 B |       1 |                   8ms
-    #     24716 | public |     payment_gateways |     1 |         1s675 |             117 B |       1 |                  11ms
-    #     24748 | public |                roles |     1 |          71ms |             173 B |       1 |                   8ms
-    #     24676 | public |           management |     1 |          33ms |              40 B |       1 |                  19ms
-
-
-    #                                                   Step   Connection    Duration    Transfer   Concurrency
-    #     --------------------------------------------------   ----------  ----------  ----------  ------------
-    #       Catalog Queries (table ordering, filtering, etc)       source         12s                         1
-    #                                             Dump Schema       source       765ms                         1
-    #                                         Prepare Schema       target       466ms                         1
-    #           COPY, INDEX, CONSTRAINTS, VACUUM (wall clock)         both       2h47m                        12
-    #                                       COPY (cumulative)         both       7h46m     1225 GB             4
-    #                               CREATE INDEX (cumulative)       target       4h36m                         4
-    #                               CONSTRAINTS (cumulative)       target       8s493                         4
-    #                                     VACUUM (cumulative)       target         0ms                         4
-    #                                         Reset Sequences         both        60ms                         1
-    #                             Large Objects (cumulative)       (null)         0ms                         0
-    #                                         Finalize Schema         both      14h01m                         4
-    #     --------------------------------------------------   ----------  ----------  ----------  ------------
-    #                               Total Wall Clock Duration         both      16h49m                        20
-
-
-    #     EOF
-
-
-    - name: show tables sizes and retrieve current backpressure seconds
+    - name: show tables sizes after ingest
       run: |
         export LD_LIBRARY_PATH=${PG_16_LIB_PATH}
-        ${PSQL} "${NEW_PROJECT_CONNSTR}" -c "\dt+"
-        BACKPRESSURE_TIME_AFTER_INGEST=$(${PSQL} "${NEW_PROJECT_CONNSTR}" -t -c "select backpressure_throttling_time()/1000000;")
-        echo "BACKPRESSURE_TIME_AFTER_INGEST=${BACKPRESSURE_TIME_AFTER_INGEST}" >> $GITHUB_ENV
-
-    - name: Parse pgcopydb log and report performance metrics
-      env:
-        PERF_TEST_RESULT_CONNSTR: ${{ secrets.PERF_TEST_RESULT_CONNSTR }}
-      run: |
-        export LD_LIBRARY_PATH=${PG_16_LIB_PATH}
-
-        # Define the log file path
-        LOG_FILE="/tmp/pgcopydb_${{ matrix.target_project }}.log"
-        
-        # Get the current git commit hash
-        git config --global --add safe.directory /__w/neon/neon
-        COMMIT_HASH=$(git rev-parse --short HEAD)
-        
-        # Define the platform and test suite
-        PLATFORM="pg16-${{ matrix.target_project }}-us-east-2-staging"
-        SUIT="pgcopydb_ingest_bench"
-        
-        # Function to convert time (e.g., "2h47m", "4h36m", "118ms", "8s493") to seconds
-        convert_to_seconds() {
-          local duration=$1
-          local total_seconds=0
-    
-          # Check for hours (h)
-          if [[ "$duration" =~ ([0-9]+)h ]]; then
-            total_seconds=$((total_seconds + ${BASH_REMATCH[1]#0} * 3600))
-          fi
-    
-          # Check for seconds (s)
-          if [[ "$duration" =~ ([0-9]+)s ]]; then
-            total_seconds=$((total_seconds + ${BASH_REMATCH[1]#0}))
-          fi
-    
-          # Check for milliseconds (ms) (if applicable)
-          if [[ "$duration" =~ ([0-9]+)ms ]]; then
-            total_seconds=$((total_seconds + ${BASH_REMATCH[1]#0} / 1000))
-            duration=${duration/${BASH_REMATCH[0]}/} # need to remove it to avoid double counting with m 
-          fi
-
-          # Check for minutes (m) - must be checked after ms because m is contained in ms
-          if [[ "$duration" =~ ([0-9]+)m ]]; then
-            total_seconds=$((total_seconds + ${BASH_REMATCH[1]#0} * 60))
-          fi
-    
-          echo $total_seconds
-        }
-
-        # Calculate the backpressure difference in seconds
-        BACKPRESSURE_TIME_DIFF=$(awk "BEGIN {print $BACKPRESSURE_TIME_AFTER_INGEST - $BACKPRESSURE_TIME_BEFORE_INGEST}")
-
-        # Insert the backpressure time difference into the performance database
-        if [ -n "$BACKPRESSURE_TIME_DIFF" ]; then
-          PSQL_CMD="${PSQL} \"${PERF_TEST_RESULT_CONNSTR}\" -c \"
-          INSERT INTO public.perf_test_results (suit, revision, platform, metric_name, metric_value, metric_unit, metric_report_type, recorded_at_timestamp)
-          VALUES ('${SUIT}', '${COMMIT_HASH}', '${PLATFORM}', 'backpressure_time', ${BACKPRESSURE_TIME_DIFF}, 'seconds', 'lower_is_better', now());
-          \""
-          echo "Inserting backpressure time difference: ${BACKPRESSURE_TIME_DIFF} seconds"
-          eval $PSQL_CMD
-        fi
-
-        # Extract and process log lines
-        while IFS= read -r line; do
-          METRIC_NAME=""
-          # Match each desired line and extract the relevant information
-          if [[ "$line" =~ COPY,\ INDEX,\ CONSTRAINTS,\ VACUUM.* ]]; then
-            METRIC_NAME="COPY, INDEX, CONSTRAINTS, VACUUM (wall clock)"
-          elif [[ "$line" =~ COPY\ \(cumulative\).* ]]; then
-            METRIC_NAME="COPY (cumulative)"
-          elif [[ "$line" =~ CREATE\ INDEX\ \(cumulative\).* ]]; then
-            METRIC_NAME="CREATE INDEX (cumulative)"
-          elif [[ "$line" =~ CONSTRAINTS\ \(cumulative\).* ]]; then
-            METRIC_NAME="CONSTRAINTS (cumulative)"
-          elif [[ "$line" =~ Finalize\ Schema.* ]]; then
-            METRIC_NAME="Finalize Schema"
-          elif [[ "$line" =~ Total\ Wall\ Clock\ Duration.* ]]; then
-            METRIC_NAME="Total Wall Clock Duration"
-          fi
-          
-          # If a metric was matched, insert it into the performance database
-          if [ -n "$METRIC_NAME" ]; then
-            DURATION=$(echo "$line" | grep -oP '\d+h\d+m|\d+s|\d+ms|\d{1,2}h\d{1,2}m|\d+\.\d+s' | head -n 1)
-            METRIC_VALUE=$(convert_to_seconds "$DURATION")
-            PSQL_CMD="${PSQL} \"${PERF_TEST_RESULT_CONNSTR}\" -c \"
-            INSERT INTO public.perf_test_results (suit, revision, platform, metric_name, metric_value, metric_unit, metric_report_type, recorded_at_timestamp)
-            VALUES ('${SUIT}', '${COMMIT_HASH}', '${PLATFORM}', '${METRIC_NAME}', ${METRIC_VALUE}, 'seconds', 'lower_is_better', now());
-            \""
-            echo "Inserting ${METRIC_NAME} with value ${METRIC_VALUE} seconds"
-            eval $PSQL_CMD
-          fi
-        done < "$LOG_FILE"
+        ${PSQL} "${BENCHMARK_INGEST_TARGET_CONNSTR}" -c "\dt+"
       
     - name: Delete Neon Project
       if: ${{ always() && matrix.target_project == 'new_empty_project' }}

.github/workflows/neon_extra_builds.yml

@@ -26,15 +26,9 @@ jobs:
     with:
       github-event-name: ${{ github.event_name}}
 
-  check-build-tools-image:
-    needs: [ check-permissions ]
-    uses: ./.github/workflows/check-build-tools-image.yml
-
   build-build-tools-image:
-    needs: [ check-build-tools-image ]
+    needs: [ check-permissions ]
     uses: ./.github/workflows/build-build-tools-image.yml
-    with:
-      image-tag: ${{ needs.check-build-tools-image.outputs.image-tag }}
     secrets: inherit
 
   check-macos-build:
@@ -44,7 +38,7 @@ jobs:
       contains(github.event.pull_request.labels.*.name, 'run-extra-build-*') ||
       github.ref_name == 'main'
     timeout-minutes: 90
-    runs-on: macos-14
+    runs-on: macos-15
 
     env:
       # Use release build only, to have less debug info around
@@ -58,7 +52,7 @@ jobs:
           submodules: true
 
       - name: Install macOS postgres dependencies
-        run: brew install flex bison openssl protobuf icu4c pkg-config
+        run: brew install flex bison openssl protobuf icu4c
 
       - name: Set pg 14 revision for caching
         id: pg_v14_rev

.github/workflows/periodic_pagebench.yml

@@ -72,7 +72,7 @@ jobs:
           echo "COMMIT_HASH=$INPUT_COMMIT_HASH" >> $GITHUB_ENV
         fi
 
-    - name: Start Bench with run_id   
+    - name: Start Bench with run_id
       run: |
         curl -k -X 'POST' \
         "${EC2_MACHINE_URL_US}/start_test/${GITHUB_RUN_ID}" \
@@ -116,7 +116,7 @@ jobs:
         -H 'accept: application/gzip' \
         -H "Authorization: Bearer $API_KEY" \
         --output "test_log_${GITHUB_RUN_ID}.gz"
-    
+
     - name: Unzip Test Log and Print it into this job's log
       if: always() && steps.poll_step.outputs.too_many_runs != 'true'
       run: |
@@ -134,13 +134,13 @@ jobs:
       if: ${{ github.event.schedule && failure() }}
       uses: slackapi/slack-github-action@v1
       with:
-        channel-id: "C033QLM5P7D" # dev-staging-stream
+        channel-id: "C06KHQVQ7U3" # on-call-qa-staging-stream
         slack-message: "Periodic pagebench testing on dedicated hardware: ${{ job.status }}\n${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
       env:
         SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}
 
     - name: Cleanup Test Resources
-      if: always() 
+      if: always()
       run: |
         curl -k -X 'POST' \
         "${EC2_MACHINE_URL_US}/cleanup_test/${GITHUB_RUN_ID}" \

.github/workflows/pg-clients.yml

@@ -39,15 +39,9 @@ jobs:
     with:
       github-event-name: ${{ github.event_name }}
 
-  check-build-tools-image:
-    needs: [ check-permissions ]
-    uses: ./.github/workflows/check-build-tools-image.yml
-
   build-build-tools-image:
-    needs: [ check-build-tools-image ]
+    needs: [ check-permissions ]
     uses: ./.github/workflows/build-build-tools-image.yml
-    with:
-      image-tag: ${{ needs.check-build-tools-image.outputs.image-tag }}
     secrets: inherit
 
   test-logical-replication:

.github/workflows/pre-merge-checks.yml

@@ -34,16 +34,10 @@ jobs:
         run: |
           echo "${PYTHON_CHANGED_FILES}"
 
-  check-build-tools-image:
+  build-build-tools-image:
     if: needs.get-changed-files.outputs.python-changed == 'true'
     needs: [ get-changed-files ]
-    uses: ./.github/workflows/check-build-tools-image.yml
-
-  build-build-tools-image:
-    needs: [ check-build-tools-image ]
     uses: ./.github/workflows/build-build-tools-image.yml
-    with:
-      image-tag: ${{ needs.check-build-tools-image.outputs.image-tag }}
     secrets: inherit
 
   check-codestyle-python:

.github/workflows/release.yml

@@ -26,82 +26,26 @@ defaults:
 jobs:
   create-storage-release-branch:
     if: ${{ github.event.schedule == '0 6 * * MON' || format('{0}', inputs.create-storage-release-branch) == 'true' }}
-    runs-on: ubuntu-22.04
 
     permissions:
-      contents: write # for `git push`
+      contents: write
 
-    steps:
-    - name: Check out code
-      uses: actions/checkout@v4
-      with:
-        ref: main
-
-    - name: Set environment variables
-      run: |
-        echo "RELEASE_DATE=$(date +'%Y-%m-%d')" | tee -a $GITHUB_ENV
-        echo "RELEASE_BRANCH=rc/$(date +'%Y-%m-%d')" | tee -a $GITHUB_ENV
-
-    - name: Create release branch
-      run: git checkout -b $RELEASE_BRANCH
-
-    - name: Push new branch
-      run: git push origin $RELEASE_BRANCH
-
-    - name: Create pull request into release
-      env:
-        GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
-      run: |
-        TITLE="Storage & Compute release ${RELEASE_DATE}"
-
-        cat << EOF > body.md
-          ## ${TITLE}
-
-          **Please merge this Pull Request using 'Create a merge commit' button**
-        EOF
-
-        gh pr create --title "${TITLE}" \
-                     --body-file "body.md" \
-                     --head "${RELEASE_BRANCH}" \
-                     --base "release"
+    uses: ./.github/workflows/_create-release-pr.yml
+    with:
+      component-name: 'Storage & Compute'
+      release-branch: 'release'
+    secrets:
+      ci-access-token: ${{ secrets.CI_ACCESS_TOKEN }}
 
   create-proxy-release-branch:
     if: ${{ github.event.schedule == '0 6 * * THU' || format('{0}', inputs.create-proxy-release-branch) == 'true' }}
-    runs-on: ubuntu-22.04
 
     permissions:
-      contents: write # for `git push`
+      contents: write
 
-    steps:
-    - name: Check out code
-      uses: actions/checkout@v4
-      with:
-        ref: main
-
-    - name: Set environment variables
-      run: |
-        echo "RELEASE_DATE=$(date +'%Y-%m-%d')" | tee -a $GITHUB_ENV
-        echo "RELEASE_BRANCH=rc/proxy/$(date +'%Y-%m-%d')" | tee -a $GITHUB_ENV
-
-    - name: Create release branch
-      run: git checkout -b $RELEASE_BRANCH
-
-    - name: Push new branch
-      run: git push origin $RELEASE_BRANCH
-
-    - name: Create pull request into release
-      env:
-        GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
-      run: |
-        TITLE="Proxy release ${RELEASE_DATE}"
-
-        cat << EOF > body.md
-          ## ${TITLE}
-
-          **Please merge this Pull Request using 'Create a merge commit' button**
-        EOF
-
-        gh pr create --title "${TITLE}" \
-                     --body-file "body.md" \
-                     --head "${RELEASE_BRANCH}" \
-                     --base "release-proxy"
+    uses: ./.github/workflows/_create-release-pr.yml
+    with:
+      component-name: 'Proxy'
+      release-branch: 'release-proxy'
+    secrets:
+      ci-access-token: ${{ secrets.CI_ACCESS_TOKEN }}

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

@@ -9,7 +9,6 @@ on:
     - Build and Test Locally
     - Build build-tools image
     - Check Permissions
-    - Check build-tools image
     - Check neon with extra platform builds
     - Cloud Regression Test
     - Create Release Branch

CODEOWNERS

@@ -1,6 +1,5 @@
+/.github/ @neondatabase/developer-productivity
 /compute_tools/ @neondatabase/control-plane @neondatabase/compute
-/storage_controller @neondatabase/storage
-/storage_scrubber @neondatabase/storage
 /libs/pageserver_api/ @neondatabase/storage
 /libs/postgres_ffi/ @neondatabase/compute @neondatabase/storage
 /libs/remote_storage/ @neondatabase/storage
@@ -11,4 +10,6 @@
 /pgxn/neon/ @neondatabase/compute @neondatabase/storage
 /proxy/ @neondatabase/proxy
 /safekeeper/ @neondatabase/storage
+/storage_controller @neondatabase/storage
+/storage_scrubber @neondatabase/storage
 /vendor/ @neondatabase/compute

Cargo.lock

@@ -3578,7 +3578,6 @@ dependencies = [
  "thiserror",
  "tokio",
  "tokio-util",
- "toml_edit",
  "utils",
  "workspace_hack",
 ]
@@ -3642,6 +3641,7 @@ dependencies = [
  "serde_json",
  "serde_path_to_error",
  "serde_with",
+ "smallvec",
  "storage_broker",
  "strum",
  "strum_macros",
@@ -4009,7 +4009,7 @@ dependencies = [
 [[package]]
 name = "postgres"
 version = "0.19.4"
-source = "git+https://github.com/neondatabase/rust-postgres.git?rev=20031d7a9ee1addeae6e0968e3899ae6bf01cee2#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
+source = "git+https://github.com/neondatabase/rust-postgres.git?branch=neon#a130197713830a0ea0004b539b1f51a66b4c3e18"
 dependencies = [
  "bytes",
  "fallible-iterator",
@@ -4022,7 +4022,7 @@ dependencies = [
 [[package]]
 name = "postgres-protocol"
 version = "0.6.4"
-source = "git+https://github.com/neondatabase/rust-postgres.git?rev=20031d7a9ee1addeae6e0968e3899ae6bf01cee2#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
+source = "git+https://github.com/neondatabase/rust-postgres.git?branch=neon#a130197713830a0ea0004b539b1f51a66b4c3e18"
 dependencies = [
  "base64 0.20.0",
  "byteorder",
@@ -4041,7 +4041,7 @@ dependencies = [
 [[package]]
 name = "postgres-types"
 version = "0.2.4"
-source = "git+https://github.com/neondatabase/rust-postgres.git?rev=20031d7a9ee1addeae6e0968e3899ae6bf01cee2#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
+source = "git+https://github.com/neondatabase/rust-postgres.git?branch=neon#a130197713830a0ea0004b539b1f51a66b4c3e18"
 dependencies = [
  "bytes",
  "fallible-iterator",
@@ -5663,9 +5663,9 @@ dependencies = [
 
 [[package]]
 name = "smallvec"
-version = "1.13.1"
+version = "1.13.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e6ecd384b10a64542d77071bd64bd7b231f4ed5940fba55e98c3de13824cf3d7"
+checksum = "3c5e1a9a646d36c3599cd173a41282daf47c44583ad367b8e6837255952e5c67"
 
 [[package]]
 name = "smol_str"
@@ -6074,9 +6074,9 @@ dependencies = [
 
 [[package]]
 name = "tikv-jemalloc-ctl"
-version = "0.5.4"
+version = "0.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "619bfed27d807b54f7f776b9430d4f8060e66ee138a28632ca898584d462c31c"
+checksum = "f21f216790c8df74ce3ab25b534e0718da5a1916719771d3fec23315c99e468b"
 dependencies = [
  "libc",
  "paste",
@@ -6085,9 +6085,9 @@ dependencies = [
 
 [[package]]
 name = "tikv-jemalloc-sys"
-version = "0.5.4+5.3.0-patched"
+version = "0.6.0+5.3.0-1-ge13ca993e8ccb9ba9847cc330696e02839f328f7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9402443cb8fd499b6f327e40565234ff34dbda27460c5b47db0db77443dd85d1"
+checksum = "cd3c60906412afa9c2b5b5a48ca6a5abe5736aec9eb48ad05037a677e52e4e2d"
 dependencies = [
  "cc",
  "libc",
@@ -6095,9 +6095,9 @@ dependencies = [
 
 [[package]]
 name = "tikv-jemallocator"
-version = "0.5.4"
+version = "0.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "965fe0c26be5c56c94e38ba547249074803efd52adfb66de62107d95aab3eaca"
+checksum = "4cec5ff18518d81584f477e9bfdf957f5bb0979b0bac3af4ca30b5b3ae2d2865"
 dependencies = [
  "libc",
  "tikv-jemalloc-sys",
@@ -6227,7 +6227,7 @@ dependencies = [
 [[package]]
 name = "tokio-postgres"
 version = "0.7.7"
-source = "git+https://github.com/neondatabase/rust-postgres.git?rev=20031d7a9ee1addeae6e0968e3899ae6bf01cee2#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
+source = "git+https://github.com/neondatabase/rust-postgres.git?branch=neon#a130197713830a0ea0004b539b1f51a66b4c3e18"
 dependencies = [
  "async-trait",
  "byteorder",

Cargo.toml

@@ -168,8 +168,8 @@ sync_wrapper = "0.1.2"
 tar = "0.4"
 test-context = "0.3"
 thiserror = "1.0"
-tikv-jemallocator = "0.5"
-tikv-jemalloc-ctl = "0.5"
+tikv-jemallocator = { version = "0.6", features = ["stats"] }
+tikv-jemalloc-ctl = { version = "0.6", features = ["stats"] }
 tokio = { version = "1.17", features = ["macros"] }
 tokio-epoll-uring = { git = "https://github.com/neondatabase/tokio-epoll-uring.git" , branch = "main" }
 tokio-io-timeout = "1.2.0"
@@ -203,21 +203,10 @@ env_logger = "0.10"
 log = "0.4"
 
 ## Libraries from neondatabase/ git forks, ideally with changes to be upstreamed
-
-# We want to use the 'neon' branch for these, but there's currently one
-# incompatible change on the branch. See:
-#
-# - PR #8076 which contained changes that depended on the new changes in
-#   the rust-postgres crate, and
-# - PR #8654 which reverted those changes and made the code in proxy incompatible
-#   with the tip of the 'neon' branch again.
-#
-# When those proxy changes are re-applied (see PR #8747), we can switch using
-# the tip of the 'neon' branch again.
-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev = "20031d7a9ee1addeae6e0968e3899ae6bf01cee2" }
-postgres-protocol = { git = "https://github.com/neondatabase/rust-postgres.git", rev = "20031d7a9ee1addeae6e0968e3899ae6bf01cee2" }
-postgres-types = { git = "https://github.com/neondatabase/rust-postgres.git", rev = "20031d7a9ee1addeae6e0968e3899ae6bf01cee2" }
-tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev = "20031d7a9ee1addeae6e0968e3899ae6bf01cee2" }
+postgres = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
+postgres-protocol = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
+postgres-types = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
+tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
 
 ## Local libraries
 compute_api = { version = "0.1", path = "./libs/compute_api/" }
@@ -255,7 +244,7 @@ tonic-build = "0.12"
 [patch.crates-io]
 
 # Needed to get `tokio-postgres-rustls` to depend on our fork.
-tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev = "20031d7a9ee1addeae6e0968e3899ae6bf01cee2" }
+tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
 
 ################# Binary contents sections
 

Dockerfile

@@ -55,7 +55,6 @@ RUN set -e \
       --bin proxy  \
       --bin neon_local \
       --bin storage_scrubber \
-      --bin stream_events \
       --locked --release
 
 # Build final image
@@ -83,7 +82,6 @@ COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_control
 COPY --from=build --chown=neon:neon /home/nonroot/target/release/proxy               /usr/local/bin
 COPY --from=build --chown=neon:neon /home/nonroot/target/release/neon_local          /usr/local/bin
 COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_scrubber    /usr/local/bin
-COPY --from=build --chown=neon:neon /home/nonroot/target/release/stream_events       /usr/local/bin
 
 COPY --from=pg-build /home/nonroot/pg_install/v14 /usr/local/v14/
 COPY --from=pg-build /home/nonroot/pg_install/v15 /usr/local/v15/

compute/etc/neon_collector.jsonnet

@@ -6,6 +6,7 @@
     import 'sql_exporter/compute_backpressure_throttling_seconds.libsonnet',
     import 'sql_exporter/compute_current_lsn.libsonnet',
     import 'sql_exporter/compute_logical_snapshot_files.libsonnet',
+    import 'sql_exporter/compute_max_connections.libsonnet',
     import 'sql_exporter/compute_receive_lsn.libsonnet',
     import 'sql_exporter/compute_subscriptions_count.libsonnet',
     import 'sql_exporter/connection_counts.libsonnet',

compute/etc/sql_exporter/compute_max_connections.libsonnet

@@ -0,0 +1,10 @@
+{
+  metric_name: 'compute_max_connections',
+  type: 'gauge',
+  help: 'Max connections allowed for Postgres',
+  key_labels: null,
+  values: [
+    'max_connections',
+  ],
+  query: importstr 'sql_exporter/compute_max_connections.sql',
+}

compute/etc/sql_exporter/compute_max_connections.sql

@@ -0,0 +1 @@
+SELECT current_setting('max_connections') as max_connections;

compute/patches/cloud_regress_pg16.patch

@@ -147,7 +147,7 @@ index 542c2e098c..0062d3024f 100644
  ALTER TABLE ptnowner1 OWNER TO regress_ptnowner;
  ALTER TABLE ptnowner OWNER TO regress_ptnowner;
 diff --git a/src/test/regress/expected/collate.icu.utf8.out b/src/test/regress/expected/collate.icu.utf8.out
-index 97bbe53b64..eac3d42a79 100644
+index 3f9a8f539c..0a51b52940 100644
 --- a/src/test/regress/expected/collate.icu.utf8.out
 +++ b/src/test/regress/expected/collate.icu.utf8.out
 @@ -1016,7 +1016,7 @@ select * from collate_test1 where b ilike 'ABC';
@@ -309,7 +309,7 @@ index b48365ec98..a6ef910055 100644
  -- the wrong partition. This test is *not* guaranteed to trigger that bug, but
  -- does so when shared_buffers is small enough.  To test if we encountered the
 diff --git a/src/test/regress/expected/copy2.out b/src/test/regress/expected/copy2.out
-index faf1a4d1b0..a44c97db52 100644
+index 9a74820ee8..22400a5551 100644
 --- a/src/test/regress/expected/copy2.out
 +++ b/src/test/regress/expected/copy2.out
 @@ -553,8 +553,8 @@ select * from check_con_tbl;
@@ -573,7 +573,7 @@ index 93302a07ef..1a73f083ac 100644
  -- that does not match with what's expected.
  -- This checks all the object types that include schema qualifications.
 diff --git a/src/test/regress/expected/create_view.out b/src/test/regress/expected/create_view.out
-index f3f8c7b5a2..3e3e54ff4c 100644
+index f551624afb..57f1e432d4 100644
 --- a/src/test/regress/expected/create_view.out
 +++ b/src/test/regress/expected/create_view.out
 @@ -18,7 +18,8 @@ CREATE TABLE real_city (
@@ -700,12 +700,12 @@ index 6ed50fdcfa..caa00a345d 100644
  COMMENT ON FOREIGN DATA WRAPPER dummy IS 'useless';
  CREATE FOREIGN DATA WRAPPER postgresql VALIDATOR postgresql_fdw_validator;
 diff --git a/src/test/regress/expected/foreign_key.out b/src/test/regress/expected/foreign_key.out
-index 12e523c737..8872e23935 100644
+index 6b8c2f2414..8e13b7fa46 100644
 --- a/src/test/regress/expected/foreign_key.out
 +++ b/src/test/regress/expected/foreign_key.out
-@@ -1968,7 +1968,7 @@ ALTER TABLE fk_partitioned_fk ATTACH PARTITION fk_partitioned_fk_2
-   FOR VALUES IN (1600);
- -- leave these tables around intentionally
+@@ -1985,7 +1985,7 @@ ALTER TABLE fk_partitioned_fk_6 ATTACH PARTITION fk_partitioned_pk_6 FOR VALUES
+ ERROR:  cannot ALTER TABLE "fk_partitioned_pk_61" because it is being used by active queries in this session
+ DROP TABLE fk_partitioned_pk_6, fk_partitioned_fk_6;
  -- test the case when the referenced table is owned by a different user
 -create role regress_other_partitioned_fk_owner;
 +create role regress_other_partitioned_fk_owner PASSWORD NEON_PASSWORD_PLACEHOLDER;
@@ -713,7 +713,7 @@ index 12e523c737..8872e23935 100644
  set role regress_other_partitioned_fk_owner;
  create table other_partitioned_fk(a int, b int) partition by list (a);
 diff --git a/src/test/regress/expected/generated.out b/src/test/regress/expected/generated.out
-index 0f623f7119..b48588a54e 100644
+index 5881420388..4ae21aa43c 100644
 --- a/src/test/regress/expected/generated.out
 +++ b/src/test/regress/expected/generated.out
 @@ -534,7 +534,7 @@ CREATE TABLE gtest10a (a int PRIMARY KEY, b int GENERATED ALWAYS AS (a * 2) STOR
@@ -762,7 +762,7 @@ index a2036a1597..805d73b9d2 100644
  -- fields, leading to long bucket chains and lots of table expansion.
  -- this is therefore a stress test of the bucket overflow code (unlike
 diff --git a/src/test/regress/expected/identity.out b/src/test/regress/expected/identity.out
-index cc7772349f..98a08eb48d 100644
+index 1b74958de9..078187b542 100644
 --- a/src/test/regress/expected/identity.out
 +++ b/src/test/regress/expected/identity.out
 @@ -520,7 +520,7 @@ ALTER TABLE itest7 ALTER COLUMN a SET GENERATED BY DEFAULT;
@@ -775,10 +775,10 @@ index cc7772349f..98a08eb48d 100644
  GRANT SELECT, INSERT ON itest8 TO regress_identity_user1;
  SET ROLE regress_identity_user1;
 diff --git a/src/test/regress/expected/inherit.out b/src/test/regress/expected/inherit.out
-index 4943429e9b..0257f22b15 100644
+index 8f831c95c3..ec681b52af 100644
 --- a/src/test/regress/expected/inherit.out
 +++ b/src/test/regress/expected/inherit.out
-@@ -2606,7 +2606,7 @@ create index on permtest_parent (left(c, 3));
+@@ -2636,7 +2636,7 @@ create index on permtest_parent (left(c, 3));
  insert into permtest_parent
    select 1, 'a', left(fipshash(i::text), 5) from generate_series(0, 100) i;
  analyze permtest_parent;
@@ -1133,7 +1133,7 @@ index 8475231735..1afae5395f 100644
  SELECT rolname, rolpassword
      FROM pg_authid
 diff --git a/src/test/regress/expected/privileges.out b/src/test/regress/expected/privileges.out
-index fbb0489a4f..2905194e2c 100644
+index 5b9dba7b32..cc408dad42 100644
 --- a/src/test/regress/expected/privileges.out
 +++ b/src/test/regress/expected/privileges.out
 @@ -20,19 +20,19 @@ SELECT lo_unlink(oid) FROM pg_largeobject_metadata WHERE oid >= 1000 AND oid < 3
@@ -1185,7 +1185,7 @@ index fbb0489a4f..2905194e2c 100644
  GRANT pg_read_all_data TO regress_priv_user6;
  GRANT pg_write_all_data TO regress_priv_user7;
  GRANT pg_read_all_settings TO regress_priv_user8 WITH ADMIN OPTION;
-@@ -145,8 +145,8 @@ REVOKE pg_read_all_settings FROM regress_priv_user8;
+@@ -212,8 +212,8 @@ REVOKE pg_read_all_settings FROM regress_priv_user8;
  DROP USER regress_priv_user10;
  DROP USER regress_priv_user9;
  DROP USER regress_priv_user8;
@@ -1196,7 +1196,7 @@ index fbb0489a4f..2905194e2c 100644
  ALTER GROUP regress_priv_group1 ADD USER regress_priv_user4;
  GRANT regress_priv_group2 TO regress_priv_user2 GRANTED BY regress_priv_user1;
  SET SESSION AUTHORIZATION regress_priv_user1;
-@@ -172,12 +172,16 @@ GRANT regress_priv_role TO regress_priv_user1 WITH ADMIN OPTION GRANTED BY regre
+@@ -239,12 +239,16 @@ GRANT regress_priv_role TO regress_priv_user1 WITH ADMIN OPTION GRANTED BY regre
  ERROR:  permission denied to grant privileges as role "regress_priv_role"
  DETAIL:  The grantor must have the ADMIN option on role "regress_priv_role".
  GRANT regress_priv_role TO regress_priv_user1 WITH ADMIN OPTION GRANTED BY CURRENT_ROLE;
@@ -1213,7 +1213,7 @@ index fbb0489a4f..2905194e2c 100644
  DROP ROLE regress_priv_role;
  SET SESSION AUTHORIZATION regress_priv_user1;
  SELECT session_user, current_user;
-@@ -1709,7 +1713,7 @@ SELECT has_table_privilege('regress_priv_user1', 'atest4', 'SELECT WITH GRANT OP
+@@ -1776,7 +1780,7 @@ SELECT has_table_privilege('regress_priv_user1', 'atest4', 'SELECT WITH GRANT OP
  
  -- security-restricted operations
  \c -
@@ -1222,7 +1222,7 @@ index fbb0489a4f..2905194e2c 100644
  -- Check that index expressions and predicates are run as the table's owner
  -- A dummy index function checking current_user
  CREATE FUNCTION sro_ifun(int) RETURNS int AS $$
-@@ -2601,8 +2605,8 @@ drop cascades to function testns.priv_testagg(integer)
+@@ -2668,8 +2672,8 @@ drop cascades to function testns.priv_testagg(integer)
  drop cascades to function testns.priv_testproc(integer)
  -- Change owner of the schema & and rename of new schema owner
  \c -
@@ -1233,7 +1233,7 @@ index fbb0489a4f..2905194e2c 100644
  SET SESSION ROLE regress_schemauser1;
  CREATE SCHEMA testns;
  SELECT nspname, rolname FROM pg_namespace, pg_roles WHERE pg_namespace.nspname = 'testns' AND pg_namespace.nspowner = pg_roles.oid;
-@@ -2725,7 +2729,7 @@ DROP USER regress_priv_user7;
+@@ -2792,7 +2796,7 @@ DROP USER regress_priv_user7;
  DROP USER regress_priv_user8; -- does not exist
  ERROR:  role "regress_priv_user8" does not exist
  -- permissions with LOCK TABLE
@@ -1242,7 +1242,7 @@ index fbb0489a4f..2905194e2c 100644
  CREATE TABLE lock_table (a int);
  -- LOCK TABLE and SELECT permission
  GRANT SELECT ON lock_table TO regress_locktable_user;
-@@ -2807,7 +2811,7 @@ DROP USER regress_locktable_user;
+@@ -2874,7 +2878,7 @@ DROP USER regress_locktable_user;
  -- pg_backend_memory_contexts.
  -- switch to superuser
  \c -
@@ -1251,7 +1251,7 @@ index fbb0489a4f..2905194e2c 100644
  SELECT has_table_privilege('regress_readallstats','pg_backend_memory_contexts','SELECT'); -- no
   has_table_privilege 
  ---------------------
-@@ -2851,10 +2855,10 @@ RESET ROLE;
+@@ -2918,10 +2922,10 @@ RESET ROLE;
  -- clean up
  DROP ROLE regress_readallstats;
  -- test role grantor machinery
@@ -1266,7 +1266,7 @@ index fbb0489a4f..2905194e2c 100644
  GRANT regress_group TO regress_group_direct_manager WITH INHERIT FALSE, ADMIN TRUE;
  GRANT regress_group_direct_manager TO regress_group_indirect_manager;
  SET SESSION AUTHORIZATION regress_group_direct_manager;
-@@ -2883,9 +2887,9 @@ DROP ROLE regress_group_direct_manager;
+@@ -2950,9 +2954,9 @@ DROP ROLE regress_group_direct_manager;
  DROP ROLE regress_group_indirect_manager;
  DROP ROLE regress_group_member;
  -- test SET and INHERIT options with object ownership changes
@@ -1813,7 +1813,7 @@ index 5e6969b173..2c4d52237f 100644
  
  -- clean up roles
 diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
-index 97ca9bf72c..b2a7a6f710 100644
+index 218c0c2863..f7af0cfb12 100644
 --- a/src/test/regress/expected/rowsecurity.out
 +++ b/src/test/regress/expected/rowsecurity.out
 @@ -14,13 +14,13 @@ DROP ROLE IF EXISTS regress_rls_group2;
@@ -1917,6 +1917,19 @@ index b79fe9a1c0..e29fab88ab 100644
  ALTER DEFAULT PRIVILEGES FOR ROLE regress_selinto_user
  	  REVOKE INSERT ON TABLES FROM regress_selinto_user;
  GRANT ALL ON SCHEMA selinto_schema TO public;
+diff --git a/src/test/regress/expected/select_parallel.out b/src/test/regress/expected/select_parallel.out
+index afc6ab08c2..dfcd891af3 100644
+--- a/src/test/regress/expected/select_parallel.out
++++ b/src/test/regress/expected/select_parallel.out
+@@ -1220,7 +1220,7 @@ SELECT 1 FROM tenk1_vw_sec
+ 
+ rollback;
+ -- test that function option SET ROLE works in parallel workers.
+-create role regress_parallel_worker;
++create role regress_parallel_worker PASSWORD NEON_PASSWORD_PLACEHOLDER;
+ create function set_and_report_role() returns text as
+   $$ select current_setting('role') $$ language sql parallel safe
+   set role = regress_parallel_worker;
 diff --git a/src/test/regress/expected/select_views.out b/src/test/regress/expected/select_views.out
 index 1aeed8452b..7d9427d070 100644
 --- a/src/test/regress/expected/select_views.out
@@ -2369,7 +2382,7 @@ index 6cb9c926c0..5e689e4062 100644
  ALTER TABLE ptnowner1 OWNER TO regress_ptnowner;
  ALTER TABLE ptnowner OWNER TO regress_ptnowner;
 diff --git a/src/test/regress/sql/collate.icu.utf8.sql b/src/test/regress/sql/collate.icu.utf8.sql
-index 3db9e25913..c66d5aa2c2 100644
+index 8aa902d5ab..24bb823b86 100644
 --- a/src/test/regress/sql/collate.icu.utf8.sql
 +++ b/src/test/regress/sql/collate.icu.utf8.sql
 @@ -353,7 +353,7 @@ reset enable_seqscan;
@@ -2532,7 +2545,7 @@ index 43d2e906dd..6c993d70f0 100644
  -- An earlier bug (see commit b1ecb9b3fcf) could end up using a buffer from
  -- the wrong partition. This test is *not* guaranteed to trigger that bug, but
 diff --git a/src/test/regress/sql/copy2.sql b/src/test/regress/sql/copy2.sql
-index d759635068..d58e50dcc5 100644
+index cf3828c16e..cf3ca38175 100644
 --- a/src/test/regress/sql/copy2.sql
 +++ b/src/test/regress/sql/copy2.sql
 @@ -365,8 +365,8 @@ copy check_con_tbl from stdin;
@@ -2774,7 +2787,7 @@ index 1b7064247a..be5b662ce1 100644
  -- Cases where schema creation fails as objects are qualified with a schema
  -- that does not match with what's expected.
 diff --git a/src/test/regress/sql/create_view.sql b/src/test/regress/sql/create_view.sql
-index 3a78be1b0c..617d2dc8d6 100644
+index ae6841308b..47bc792e30 100644
 --- a/src/test/regress/sql/create_view.sql
 +++ b/src/test/regress/sql/create_view.sql
 @@ -23,7 +23,8 @@ CREATE TABLE real_city (
@@ -2901,11 +2914,11 @@ index aa147b14a9..370e0dd570 100644
  CREATE FOREIGN DATA WRAPPER dummy;
  COMMENT ON FOREIGN DATA WRAPPER dummy IS 'useless';
 diff --git a/src/test/regress/sql/foreign_key.sql b/src/test/regress/sql/foreign_key.sql
-index 22e177f89b..7138d5e1d4 100644
+index 45c7a534cb..32dd26b8cd 100644
 --- a/src/test/regress/sql/foreign_key.sql
 +++ b/src/test/regress/sql/foreign_key.sql
-@@ -1418,7 +1418,7 @@ ALTER TABLE fk_partitioned_fk ATTACH PARTITION fk_partitioned_fk_2
- -- leave these tables around intentionally
+@@ -1435,7 +1435,7 @@ ALTER TABLE fk_partitioned_fk_6 ATTACH PARTITION fk_partitioned_pk_6 FOR VALUES
+ DROP TABLE fk_partitioned_pk_6, fk_partitioned_fk_6;
  
  -- test the case when the referenced table is owned by a different user
 -create role regress_other_partitioned_fk_owner;
@@ -2963,7 +2976,7 @@ index 527024f710..de49c0b85f 100644
  -- the data in this file has a lot of duplicates in the index key
  -- fields, leading to long bucket chains and lots of table expansion.
 diff --git a/src/test/regress/sql/identity.sql b/src/test/regress/sql/identity.sql
-index 91d2e443b4..241c93f373 100644
+index 7537258a75..9041e35e34 100644
 --- a/src/test/regress/sql/identity.sql
 +++ b/src/test/regress/sql/identity.sql
 @@ -287,7 +287,7 @@ ALTER TABLE itest7 ALTER COLUMN a RESTART;
@@ -2976,10 +2989,10 @@ index 91d2e443b4..241c93f373 100644
  GRANT SELECT, INSERT ON itest8 TO regress_identity_user1;
  SET ROLE regress_identity_user1;
 diff --git a/src/test/regress/sql/inherit.sql b/src/test/regress/sql/inherit.sql
-index fe699c54d5..bdd5993f45 100644
+index b5b554a125..109889ad24 100644
 --- a/src/test/regress/sql/inherit.sql
 +++ b/src/test/regress/sql/inherit.sql
-@@ -950,7 +950,7 @@ create index on permtest_parent (left(c, 3));
+@@ -958,7 +958,7 @@ create index on permtest_parent (left(c, 3));
  insert into permtest_parent
    select 1, 'a', left(fipshash(i::text), 5) from generate_series(0, 100) i;
  analyze permtest_parent;
@@ -3218,7 +3231,7 @@ index 53e86b0b6c..f07cf1ec54 100644
  CREATE ROLE regress_passwd5 PASSWORD 'md5e73a4b11df52a6068f8b39f90be36023';
  
 diff --git a/src/test/regress/sql/privileges.sql b/src/test/regress/sql/privileges.sql
-index 3f68cafcd1..004b26831d 100644
+index 249df17a58..b258e7f26a 100644
 --- a/src/test/regress/sql/privileges.sql
 +++ b/src/test/regress/sql/privileges.sql
 @@ -24,18 +24,18 @@ RESET client_min_messages;
@@ -3269,7 +3282,7 @@ index 3f68cafcd1..004b26831d 100644
  
  GRANT pg_read_all_data TO regress_priv_user6;
  GRANT pg_write_all_data TO regress_priv_user7;
-@@ -130,8 +130,8 @@ DROP USER regress_priv_user10;
+@@ -163,8 +163,8 @@ DROP USER regress_priv_user10;
  DROP USER regress_priv_user9;
  DROP USER regress_priv_user8;
  
@@ -3280,7 +3293,7 @@ index 3f68cafcd1..004b26831d 100644
  
  ALTER GROUP regress_priv_group1 ADD USER regress_priv_user4;
  
-@@ -1124,7 +1124,7 @@ SELECT has_table_privilege('regress_priv_user1', 'atest4', 'SELECT WITH GRANT OP
+@@ -1157,7 +1157,7 @@ SELECT has_table_privilege('regress_priv_user1', 'atest4', 'SELECT WITH GRANT OP
  
  -- security-restricted operations
  \c -
@@ -3289,7 +3302,7 @@ index 3f68cafcd1..004b26831d 100644
  
  -- Check that index expressions and predicates are run as the table's owner
  
-@@ -1620,8 +1620,8 @@ DROP SCHEMA testns CASCADE;
+@@ -1653,8 +1653,8 @@ DROP SCHEMA testns CASCADE;
  -- Change owner of the schema & and rename of new schema owner
  \c -
  
@@ -3300,7 +3313,7 @@ index 3f68cafcd1..004b26831d 100644
  
  SET SESSION ROLE regress_schemauser1;
  CREATE SCHEMA testns;
-@@ -1715,7 +1715,7 @@ DROP USER regress_priv_user8; -- does not exist
+@@ -1748,7 +1748,7 @@ DROP USER regress_priv_user8; -- does not exist
  
  
  -- permissions with LOCK TABLE
@@ -3309,7 +3322,7 @@ index 3f68cafcd1..004b26831d 100644
  CREATE TABLE lock_table (a int);
  
  -- LOCK TABLE and SELECT permission
-@@ -1803,7 +1803,7 @@ DROP USER regress_locktable_user;
+@@ -1836,7 +1836,7 @@ DROP USER regress_locktable_user;
  -- switch to superuser
  \c -
  
@@ -3318,7 +3331,7 @@ index 3f68cafcd1..004b26831d 100644
  
  SELECT has_table_privilege('regress_readallstats','pg_backend_memory_contexts','SELECT'); -- no
  SELECT has_table_privilege('regress_readallstats','pg_shmem_allocations','SELECT'); -- no
-@@ -1823,10 +1823,10 @@ RESET ROLE;
+@@ -1856,10 +1856,10 @@ RESET ROLE;
  DROP ROLE regress_readallstats;
  
  -- test role grantor machinery
@@ -3333,7 +3346,7 @@ index 3f68cafcd1..004b26831d 100644
  
  GRANT regress_group TO regress_group_direct_manager WITH INHERIT FALSE, ADMIN TRUE;
  GRANT regress_group_direct_manager TO regress_group_indirect_manager;
-@@ -1848,9 +1848,9 @@ DROP ROLE regress_group_indirect_manager;
+@@ -1881,9 +1881,9 @@ DROP ROLE regress_group_indirect_manager;
  DROP ROLE regress_group_member;
  
  -- test SET and INHERIT options with object ownership changes
@@ -3625,7 +3638,7 @@ index c961b2d730..0859b89c4f 100644
  -- clean up roles
  DROP ROLE regress_test_def_superuser;
 diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
-index dec7340538..cdbc03a5cc 100644
+index d3bfd53e23..919ce1d0c6 100644
 --- a/src/test/regress/sql/rowsecurity.sql
 +++ b/src/test/regress/sql/rowsecurity.sql
 @@ -20,13 +20,13 @@ DROP SCHEMA IF EXISTS regress_rls_schema CASCADE;
@@ -3701,6 +3714,19 @@ index 689c448cc2..223ceb1d75 100644
  ALTER DEFAULT PRIVILEGES FOR ROLE regress_selinto_user
  	  REVOKE INSERT ON TABLES FROM regress_selinto_user;
  GRANT ALL ON SCHEMA selinto_schema TO public;
+diff --git a/src/test/regress/sql/select_parallel.sql b/src/test/regress/sql/select_parallel.sql
+index 33d78e16dc..cb193c9b27 100644
+--- a/src/test/regress/sql/select_parallel.sql
++++ b/src/test/regress/sql/select_parallel.sql
+@@ -464,7 +464,7 @@ SELECT 1 FROM tenk1_vw_sec
+ rollback;
+ 
+ -- test that function option SET ROLE works in parallel workers.
+-create role regress_parallel_worker;
++create role regress_parallel_worker PASSWORD NEON_PASSWORD_PLACEHOLDER;
+ 
+ create function set_and_report_role() returns text as
+   $$ select current_setting('role') $$ language sql parallel safe
 diff --git a/src/test/regress/sql/select_views.sql b/src/test/regress/sql/select_views.sql
 index e742f13699..7bd0255df8 100644
 --- a/src/test/regress/sql/select_views.sql

compute_tools/src/bin/compute_ctl.rs

@@ -105,6 +105,11 @@ fn main() -> Result<()> {
 fn init() -> Result<(String, clap::ArgMatches)> {
     init_tracing_and_logging(DEFAULT_LOG_LEVEL)?;
 
+    opentelemetry::global::set_error_handler(|err| {
+        tracing::info!("OpenTelemetry error: {err}");
+    })
+    .expect("global error handler lock poisoned");
+
     let mut signals = Signals::new([SIGINT, SIGTERM, SIGQUIT])?;
     thread::spawn(move || {
         for sig in signals.forever() {

compute_tools/src/catalog.rs

@@ -1,38 +1,40 @@
-use compute_api::{
-    responses::CatalogObjects,
-    spec::{Database, Role},
-};
+use compute_api::responses::CatalogObjects;
 use futures::Stream;
-use postgres::{Client, NoTls};
+use postgres::NoTls;
 use std::{path::Path, process::Stdio, result::Result, sync::Arc};
 use tokio::{
     io::{AsyncBufReadExt, BufReader},
     process::Command,
-    task,
+    spawn,
 };
+use tokio_postgres::connect;
 use tokio_stream::{self as stream, StreamExt};
 use tokio_util::codec::{BytesCodec, FramedRead};
 use tracing::warn;
 
-use crate::{
-    compute::ComputeNode,
-    pg_helpers::{get_existing_dbs, get_existing_roles},
-};
+use crate::compute::ComputeNode;
+use crate::pg_helpers::{get_existing_dbs_async, get_existing_roles_async};
 
 pub async fn get_dbs_and_roles(compute: &Arc<ComputeNode>) -> anyhow::Result<CatalogObjects> {
     let connstr = compute.connstr.clone();
-    task::spawn_blocking(move || {
-        let mut client = Client::connect(connstr.as_str(), NoTls)?;
-        let roles: Vec<Role>;
-        {
-            let mut xact = client.transaction()?;
-            roles = get_existing_roles(&mut xact)?;
-        }
-        let databases: Vec<Database> = get_existing_dbs(&mut client)?.values().cloned().collect();
 
-        Ok(CatalogObjects { roles, databases })
-    })
-    .await?
+    let (client, connection): (tokio_postgres::Client, _) =
+        connect(connstr.as_str(), NoTls).await?;
+
+    spawn(async move {
+        if let Err(e) = connection.await {
+            eprintln!("connection error: {}", e);
+        }
+    });
+
+    let roles = get_existing_roles_async(&client).await?;
+
+    let databases = get_existing_dbs_async(&client)
+        .await?
+        .into_values()
+        .collect();
+
+    Ok(CatalogObjects { roles, databases })
 }
 
 #[derive(Debug, thiserror::Error)]

compute_tools/src/checker.rs

@@ -1,37 +1,9 @@
 use anyhow::{anyhow, Ok, Result};
-use postgres::Client;
 use tokio_postgres::NoTls;
 use tracing::{error, instrument, warn};
 
 use crate::compute::ComputeNode;
 
-/// Create a special service table for availability checks
-/// only if it does not exist already.
-pub fn create_availability_check_data(client: &mut Client) -> Result<()> {
-    let query = "
-        DO $$
-        BEGIN
-            IF NOT EXISTS(
-                SELECT 1
-                FROM pg_catalog.pg_tables
-                WHERE tablename = 'health_check'
-            )
-            THEN
-            CREATE TABLE health_check (
-                id serial primary key,
-                updated_at timestamptz default now()
-            );
-            INSERT INTO health_check VALUES (1, now())
-                ON CONFLICT (id) DO UPDATE
-                 SET updated_at = now();
-            END IF;
-        END
-        $$;";
-    client.execute(query, &[])?;
-
-    Ok(())
-}
-
 /// Update timestamp in a row in a special service table to check
 /// that we can actually write some data in this particular timeline.
 #[instrument(skip_all)]

compute_tools/src/compute.rs

@@ -1,20 +1,21 @@
-use std::collections::HashMap;
+use std::collections::{HashMap, HashSet};
 use std::env;
 use std::fs;
+use std::iter::once;
 use std::os::unix::fs::{symlink, PermissionsExt};
 use std::path::Path;
 use std::process::{Command, Stdio};
 use std::str::FromStr;
 use std::sync::atomic::AtomicU32;
 use std::sync::atomic::Ordering;
-use std::sync::{Condvar, Mutex, RwLock};
+use std::sync::{Arc, Condvar, Mutex, RwLock};
 use std::thread;
 use std::time::Duration;
 use std::time::Instant;
 
 use anyhow::{Context, Result};
 use chrono::{DateTime, Utc};
-use compute_api::spec::PgIdent;
+use compute_api::spec::{PgIdent, Role};
 use futures::future::join_all;
 use futures::stream::FuturesUnordered;
 use futures::StreamExt;
@@ -31,15 +32,23 @@ use compute_api::spec::{ComputeFeature, ComputeMode, ComputeSpec, ExtVersion};
 use utils::measured_stream::MeasuredReader;
 
 use nix::sys::signal::{kill, Signal};
-
 use remote_storage::{DownloadError, RemotePath};
+use tokio::spawn;
+use url::Url;
 
-use crate::checker::create_availability_check_data;
 use crate::installed_extensions::get_installed_extensions_sync;
 use crate::local_proxy;
-use crate::logger::inlinify;
 use crate::pg_helpers::*;
 use crate::spec::*;
+use crate::spec_apply::ApplySpecPhase::{
+    CreateAndAlterDatabases, CreateAndAlterRoles, CreateAvailabilityCheck, CreateSuperUser,
+    DropInvalidDatabases, DropRoles, HandleNeonExtension, HandleOtherExtensions,
+    RenameAndDeleteDatabases, RenameRoles, RunInEachDatabase,
+};
+use crate::spec_apply::PerDatabasePhase::{
+    ChangeSchemaPerms, DeleteDBRoleReferences, HandleAnonExtension,
+};
+use crate::spec_apply::{apply_operations, MutableApplyContext, DB};
 use crate::sync_sk::{check_if_synced, ping_safekeeper};
 use crate::{config, extension_server};
 
@@ -224,10 +233,7 @@ fn maybe_cgexec(cmd: &str) -> Command {
     }
 }
 
-/// Create special neon_superuser role, that's a slightly nerfed version of a real superuser
-/// that we give to customers
-#[instrument(skip_all)]
-fn create_neon_superuser(spec: &ComputeSpec, client: &mut Client) -> Result<()> {
+pub(crate) fn construct_superuser_query(spec: &ComputeSpec) -> String {
     let roles = spec
         .cluster
         .roles
@@ -296,11 +302,8 @@ fn create_neon_superuser(spec: &ComputeSpec, client: &mut Client) -> Result<()>
             $$;"#,
         roles_decl, database_decl,
     );
-    info!("Neon superuser created: {}", inlinify(&query));
-    client
-        .simple_query(&query)
-        .map_err(|e| anyhow::anyhow!(e).context(query))?;
-    Ok(())
+
+    query
 }
 
 impl ComputeNode {
@@ -813,21 +816,14 @@ impl ComputeNode {
         Ok(())
     }
 
-    /// Do initial configuration of the already started Postgres.
-    #[instrument(skip_all)]
-    pub fn apply_config(&self, compute_state: &ComputeState) -> Result<()> {
-        // If connection fails,
-        // it may be the old node with `zenith_admin` superuser.
-        //
-        // In this case we need to connect with old `zenith_admin` name
-        // and create new user. We cannot simply rename connected user,
-        // but we can create a new one and grant it all privileges.
-        let mut connstr = self.connstr.clone();
+    async fn get_maintenance_client(url: &Url) -> Result<tokio_postgres::Client> {
+        let mut connstr = url.clone();
+
         connstr
             .query_pairs_mut()
             .append_pair("application_name", "apply_config");
 
-        let mut client = match Client::connect(connstr.as_str(), NoTls) {
+        let (client, conn) = match tokio_postgres::connect(connstr.as_str(), NoTls).await {
             Err(e) => match e.code() {
                 Some(&SqlState::INVALID_PASSWORD)
                 | Some(&SqlState::INVALID_AUTHORIZATION_SPECIFICATION) => {
@@ -845,8 +841,8 @@ impl ComputeNode {
                     let mut client =
                         Client::connect(zenith_admin_connstr.as_str(), NoTls)
                             .context("broken cloud_admin credential: tried connecting with cloud_admin but could not authenticate, and zenith_admin does not work either")?;
-                    // Disable forwarding so that users don't get a cloud_admin role
 
+                    // Disable forwarding so that users don't get a cloud_admin role
                     let mut func = || {
                         client.simple_query("SET neon.forward_ddl = false")?;
                         client.simple_query("CREATE USER cloud_admin WITH SUPERUSER")?;
@@ -858,49 +854,309 @@ impl ComputeNode {
                     drop(client);
 
                     // reconnect with connstring with expected name
-                    Client::connect(connstr.as_str(), NoTls)?
+                    tokio_postgres::connect(connstr.as_str(), NoTls).await?
                 }
                 _ => return Err(e.into()),
             },
-            Ok(client) => client,
+            Ok((client, conn)) => (client, conn),
         };
 
-        // Disable DDL forwarding because control plane already knows about these roles/databases.
+        spawn(async move {
+            if let Err(e) = conn.await {
+                error!("maintenance client connection error: {}", e);
+            }
+        });
+
+        // Disable DDL forwarding because control plane already knows about the roles/databases
+        // we're about to modify.
         client
             .simple_query("SET neon.forward_ddl = false")
+            .await
             .context("apply_config SET neon.forward_ddl = false")?;
 
-        // Proceed with post-startup configuration. Note, that order of operations is important.
-        let spec = &compute_state.pspec.as_ref().expect("spec must be set").spec;
-        create_neon_superuser(spec, &mut client).context("apply_config create_neon_superuser")?;
-        cleanup_instance(&mut client).context("apply_config cleanup_instance")?;
-        handle_roles(spec, &mut client).context("apply_config handle_roles")?;
-        handle_databases(spec, &mut client).context("apply_config handle_databases")?;
-        handle_role_deletions(spec, connstr.as_str(), &mut client)
-            .context("apply_config handle_role_deletions")?;
-        handle_grants(
-            spec,
-            &mut client,
-            connstr.as_str(),
-            self.has_feature(ComputeFeature::AnonExtension),
-        )
-        .context("apply_config handle_grants")?;
-        handle_extensions(spec, &mut client).context("apply_config handle_extensions")?;
-        handle_extension_neon(&mut client).context("apply_config handle_extension_neon")?;
-        create_availability_check_data(&mut client)
-            .context("apply_config create_availability_check_data")?;
+        Ok(client)
+    }
 
-        // 'Close' connection
-        drop(client);
+    /// Apply the spec to the running PostgreSQL instance.
+    /// The caller can decide to run with multiple clients in parallel, or
+    /// single mode.  Either way, the commands executed will be the same, and
+    /// only commands run in different databases are parallelized.
+    #[instrument(skip_all)]
+    pub fn apply_spec_sql(
+        &self,
+        spec: Arc<ComputeSpec>,
+        url: Arc<Url>,
+        concurrency: usize,
+    ) -> Result<()> {
+        let rt = tokio::runtime::Builder::new_multi_thread()
+            .enable_all()
+            .build()?;
 
-        if let Some(ref local_proxy) = spec.local_proxy_config {
+        info!("Applying config with max {} concurrency", concurrency);
+        debug!("Config: {:?}", spec);
+
+        rt.block_on(async {
+            // Proceed with post-startup configuration. Note, that order of operations is important.
+            let client = Self::get_maintenance_client(&url).await?;
+            let spec = spec.clone();
+
+            let databases = get_existing_dbs_async(&client).await?;
+            let roles = get_existing_roles_async(&client)
+                .await?
+                .into_iter()
+                .map(|role| (role.name.clone(), role))
+                .collect::<HashMap<String, Role>>();
+
+            let jwks_roles = Arc::new(
+                spec.as_ref()
+                    .local_proxy_config
+                    .iter()
+                    .flat_map(|it| &it.jwks)
+                    .flatten()
+                    .flat_map(|setting| &setting.role_names)
+                    .cloned()
+                    .collect::<HashSet<_>>(),
+            );
+
+            let ctx = Arc::new(tokio::sync::RwLock::new(MutableApplyContext {
+                roles,
+                dbs: databases,
+            }));
+
+            for phase in [
+                CreateSuperUser,
+                DropInvalidDatabases,
+                RenameRoles,
+                CreateAndAlterRoles,
+                RenameAndDeleteDatabases,
+                CreateAndAlterDatabases,
+            ] {
+                debug!("Applying phase {:?}", &phase);
+                apply_operations(
+                    spec.clone(),
+                    ctx.clone(),
+                    jwks_roles.clone(),
+                    phase,
+                    || async { Ok(&client) },
+                )
+                .await?;
+            }
+
+            let concurrency_token = Arc::new(tokio::sync::Semaphore::new(concurrency));
+
+            let db_processes = spec
+                .cluster
+                .databases
+                .iter()
+                .map(|db| DB::new(db.clone()))
+                // include
+                .chain(once(DB::SystemDB))
+                .map(|db| {
+                    let spec = spec.clone();
+                    let ctx = ctx.clone();
+                    let jwks_roles = jwks_roles.clone();
+                    let mut url = url.as_ref().clone();
+                    let concurrency_token = concurrency_token.clone();
+                    let db = db.clone();
+
+                    debug!("Applying per-database phases for Database {:?}", &db);
+
+                    match &db {
+                        DB::SystemDB => {}
+                        DB::UserDB(db) => {
+                            url.set_path(db.name.as_str());
+                        }
+                    }
+
+                    let url = Arc::new(url);
+                    let fut = Self::apply_spec_sql_db(
+                        spec.clone(),
+                        url,
+                        ctx.clone(),
+                        jwks_roles.clone(),
+                        concurrency_token.clone(),
+                        db,
+                    );
+
+                    Ok(spawn(fut))
+                })
+                .collect::<Vec<Result<_, anyhow::Error>>>();
+
+            for process in db_processes.into_iter() {
+                let handle = process?;
+                handle.await??;
+            }
+
+            for phase in vec![
+                HandleOtherExtensions,
+                HandleNeonExtension,
+                CreateAvailabilityCheck,
+                DropRoles,
+            ] {
+                debug!("Applying phase {:?}", &phase);
+                apply_operations(
+                    spec.clone(),
+                    ctx.clone(),
+                    jwks_roles.clone(),
+                    phase,
+                    || async { Ok(&client) },
+                )
+                .await?;
+            }
+
+            Ok::<(), anyhow::Error>(())
+        })?;
+
+        Ok(())
+    }
+
+    /// Apply SQL migrations of the RunInEachDatabase phase.
+    ///
+    /// May opt to not connect to databases that don't have any scheduled
+    /// operations.  The function is concurrency-controlled with the provided
+    /// semaphore.  The caller has to make sure the semaphore isn't exhausted.
+    async fn apply_spec_sql_db(
+        spec: Arc<ComputeSpec>,
+        url: Arc<Url>,
+        ctx: Arc<tokio::sync::RwLock<MutableApplyContext>>,
+        jwks_roles: Arc<HashSet<String>>,
+        concurrency_token: Arc<tokio::sync::Semaphore>,
+        db: DB,
+    ) -> Result<()> {
+        let _permit = concurrency_token.acquire().await?;
+
+        let mut client_conn = None;
+
+        for subphase in [
+            DeleteDBRoleReferences,
+            ChangeSchemaPerms,
+            HandleAnonExtension,
+        ] {
+            apply_operations(
+                spec.clone(),
+                ctx.clone(),
+                jwks_roles.clone(),
+                RunInEachDatabase {
+                    db: db.clone(),
+                    subphase,
+                },
+                // Only connect if apply_operation actually wants a connection.
+                // It's quite possible this database doesn't need any queries,
+                // so by not connecting we save time and effort connecting to
+                // that database.
+                || async {
+                    if client_conn.is_none() {
+                        let db_client = Self::get_maintenance_client(&url).await?;
+                        client_conn.replace(db_client);
+                    }
+                    let client = client_conn.as_ref().unwrap();
+                    Ok(client)
+                },
+            )
+            .await?;
+        }
+
+        drop(client_conn);
+
+        Ok::<(), anyhow::Error>(())
+    }
+
+    /// Do initial configuration of the already started Postgres.
+    #[instrument(skip_all)]
+    pub fn apply_config(&self, compute_state: &ComputeState) -> Result<()> {
+        // If connection fails,
+        // it may be the old node with `zenith_admin` superuser.
+        //
+        // In this case we need to connect with old `zenith_admin` name
+        // and create new user. We cannot simply rename connected user,
+        // but we can create a new one and grant it all privileges.
+        let mut url = self.connstr.clone();
+        url.query_pairs_mut()
+            .append_pair("application_name", "apply_config");
+
+        let url = Arc::new(url);
+        let spec = Arc::new(
+            compute_state
+                .pspec
+                .as_ref()
+                .expect("spec must be set")
+                .spec
+                .clone(),
+        );
+
+        // Choose how many concurrent connections to use for applying the spec changes.
+        // If the cluster is not currently Running we don't have to deal with user connections,
+        // and can thus use all `max_connections` connection slots. However, that's generally not
+        // very efficient, so we generally still limit it to a smaller number.
+        let max_concurrent_connections = if compute_state.status != ComputeStatus::Running {
+            // If the settings contain 'max_connections', use that as template
+            if let Some(config) = spec.cluster.settings.find("max_connections") {
+                config.parse::<usize>().ok()
+            } else {
+                // Otherwise, try to find the setting in the postgresql_conf string
+                spec.cluster
+                    .postgresql_conf
+                    .iter()
+                    .flat_map(|conf| conf.split("\n"))
+                    .filter_map(|line| {
+                        if !line.contains("max_connections") {
+                            return None;
+                        }
+
+                        let (key, value) = line.split_once("=")?;
+                        let key = key
+                            .trim_start_matches(char::is_whitespace)
+                            .trim_end_matches(char::is_whitespace);
+
+                        let value = value
+                            .trim_start_matches(char::is_whitespace)
+                            .trim_end_matches(char::is_whitespace);
+
+                        if key != "max_connections" {
+                            return None;
+                        }
+
+                        value.parse::<usize>().ok()
+                    })
+                    .next()
+            }
+            // If max_connections is present, use at most 1/3rd of that.
+            // When max_connections is lower than 30, try to use at least 10 connections, but
+            // never more than max_connections.
+            .map(|limit| match limit {
+                0..10 => limit,
+                10..30 => 10,
+                30.. => limit / 3,
+            })
+            // If we didn't find max_connections, default to 10 concurrent connections.
+            .unwrap_or(10)
+        } else {
+            // state == Running
+            // Because the cluster is already in the Running state, we should assume users are
+            // already connected to the cluster, and high concurrency could negatively
+            // impact user connectivity. Therefore, we can limit concurrency to the number of
+            // reserved superuser connections, which users wouldn't be able to use anyway.
+            spec.cluster
+                .settings
+                .find("superuser_reserved_connections")
+                .iter()
+                .filter_map(|val| val.parse::<usize>().ok())
+                .map(|val| if val > 1 { val - 1 } else { 1 })
+                .last()
+                .unwrap_or(3)
+        };
+
+        // Merge-apply spec & changes to PostgreSQL state.
+        self.apply_spec_sql(spec.clone(), url.clone(), max_concurrent_connections)?;
+
+        if let Some(ref local_proxy) = &spec.clone().local_proxy_config {
             info!("configuring local_proxy");
             local_proxy::configure(local_proxy).context("apply_config local_proxy")?;
         }
 
         // Run migrations separately to not hold up cold starts
         thread::spawn(move || {
-            let mut connstr = connstr.clone();
+            let mut connstr = url.as_ref().clone();
             connstr
                 .query_pairs_mut()
                 .append_pair("application_name", "migrations");
@@ -908,7 +1164,8 @@ impl ComputeNode {
             let mut client = Client::connect(connstr.as_str(), NoTls)?;
             handle_migrations(&mut client).context("apply_config handle_migrations")
         });
-        Ok(())
+
+        Ok::<(), anyhow::Error>(())
     }
 
     // Wrapped this around `pg_ctl reload`, but right now we don't use
@@ -971,32 +1228,16 @@ impl ComputeNode {
         config::with_compute_ctl_tmp_override(pgdata_path, "neon.max_cluster_size=-1", || {
             self.pg_reload_conf()?;
 
-            let mut client = Client::connect(self.connstr.as_str(), NoTls)?;
-
-            // Proceed with post-startup configuration. Note, that order of operations is important.
-            // Disable DDL forwarding because control plane already knows about these roles/databases.
             if spec.mode == ComputeMode::Primary {
-                client.simple_query("SET neon.forward_ddl = false")?;
-                cleanup_instance(&mut client)?;
-                handle_roles(&spec, &mut client)?;
-                handle_databases(&spec, &mut client)?;
-                handle_role_deletions(&spec, self.connstr.as_str(), &mut client)?;
-                handle_grants(
-                    &spec,
-                    &mut client,
-                    self.connstr.as_str(),
-                    self.has_feature(ComputeFeature::AnonExtension),
-                )?;
-                handle_extensions(&spec, &mut client)?;
-                handle_extension_neon(&mut client)?;
-                // We can skip handle_migrations here because a new migration can only appear
-                // if we have a new version of the compute_ctl binary, which can only happen
-                // if compute got restarted, in which case we'll end up inside of apply_config
-                // instead of reconfigure.
-            }
+                let mut url = self.connstr.clone();
+                url.query_pairs_mut()
+                    .append_pair("application_name", "apply_config");
+                let url = Arc::new(url);
 
-            // 'Close' connection
-            drop(client);
+                let spec = Arc::new(spec.clone());
+
+                self.apply_spec_sql(spec, url, 1)?;
+            }
 
             Ok(())
         })?;

compute_tools/src/lib.rs

@@ -23,5 +23,6 @@ pub mod monitor;
 pub mod params;
 pub mod pg_helpers;
 pub mod spec;
+mod spec_apply;
 pub mod swap;
 pub mod sync_sk;

compute_tools/src/pg_helpers.rs

@@ -10,9 +10,9 @@ use std::thread::JoinHandle;
 use std::time::{Duration, Instant};
 
 use anyhow::{bail, Result};
+use futures::StreamExt;
 use ini::Ini;
 use notify::{RecursiveMode, Watcher};
-use postgres::{Client, Transaction};
 use tokio::io::AsyncBufReadExt;
 use tokio::time::timeout;
 use tokio_postgres::NoTls;
@@ -197,27 +197,34 @@ impl Escaping for PgIdent {
 }
 
 /// Build a list of existing Postgres roles
-pub fn get_existing_roles(xact: &mut Transaction<'_>) -> Result<Vec<Role>> {
-    let postgres_roles = xact
-        .query("SELECT rolname, rolpassword FROM pg_catalog.pg_authid", &[])?
-        .iter()
+pub async fn get_existing_roles_async(client: &tokio_postgres::Client) -> Result<Vec<Role>> {
+    let postgres_roles = client
+        .query_raw::<str, &String, &[String; 0]>(
+            "SELECT rolname, rolpassword FROM pg_catalog.pg_authid",
+            &[],
+        )
+        .await?
+        .filter_map(|row| async { row.ok() })
         .map(|row| Role {
             name: row.get("rolname"),
             encrypted_password: row.get("rolpassword"),
             options: None,
         })
-        .collect();
+        .collect()
+        .await;
 
     Ok(postgres_roles)
 }
 
 /// Build a list of existing Postgres databases
-pub fn get_existing_dbs(client: &mut Client) -> Result<HashMap<String, Database>> {
+pub async fn get_existing_dbs_async(
+    client: &tokio_postgres::Client,
+) -> Result<HashMap<String, Database>> {
     // `pg_database.datconnlimit = -2` means that the database is in the
     // invalid state. See:
     //   https://github.com/postgres/postgres/commit/a4b4cc1d60f7e8ccfcc8ff8cb80c28ee411ad9a9
-    let postgres_dbs: Vec<Database> = client
-        .query(
+    let rowstream = client
+        .query_raw::<str, &String, &[String; 0]>(
             "SELECT
                 datname AS name,
                 datdba::regrole::text AS owner,
@@ -226,8 +233,11 @@ pub fn get_existing_dbs(client: &mut Client) -> Result<HashMap<String, Database>
             FROM
                 pg_catalog.pg_database;",
             &[],
-        )?
-        .iter()
+        )
+        .await?;
+
+    let dbs_map = rowstream
+        .filter_map(|r| async { r.ok() })
         .map(|row| Database {
             name: row.get("name"),
             owner: row.get("owner"),
@@ -235,12 +245,9 @@ pub fn get_existing_dbs(client: &mut Client) -> Result<HashMap<String, Database>
             invalid: row.get("invalid"),
             options: None,
         })
-        .collect();
-
-    let dbs_map = postgres_dbs
-        .iter()
         .map(|db| (db.name.clone(), db.clone()))
-        .collect::<HashMap<_, _>>();
+        .collect::<HashMap<_, _>>()
+        .await;
 
     Ok(dbs_map)
 }

compute_tools/src/spec.rs

@@ -1,22 +1,17 @@
-use std::collections::HashSet;
+use anyhow::{anyhow, bail, Result};
+use postgres::Client;
+use reqwest::StatusCode;
 use std::fs::File;
 use std::path::Path;
-use std::str::FromStr;
-
-use anyhow::{anyhow, bail, Context, Result};
-use postgres::config::Config;
-use postgres::{Client, NoTls};
-use reqwest::StatusCode;
-use tracing::{error, info, info_span, instrument, span_enabled, warn, Level};
+use tracing::{error, info, instrument, warn};
 
 use crate::config;
-use crate::logger::inlinify;
 use crate::migration::MigrationRunner;
 use crate::params::PG_HBA_ALL_MD5;
 use crate::pg_helpers::*;
 
 use compute_api::responses::{ControlPlaneComputeStatus, ControlPlaneSpecResponse};
-use compute_api::spec::{ComputeSpec, PgIdent, Role};
+use compute_api::spec::ComputeSpec;
 
 // Do control plane request and return response if any. In case of error it
 // returns a bool flag indicating whether it makes sense to retry the request
@@ -151,625 +146,6 @@ pub fn add_standby_signal(pgdata_path: &Path) -> Result<()> {
     Ok(())
 }
 
-/// Compute could be unexpectedly shut down, for example, during the
-/// database dropping. This leaves the database in the invalid state,
-/// which prevents new db creation with the same name. This function
-/// will clean it up before proceeding with catalog updates. All
-/// possible future cleanup operations may go here too.
-#[instrument(skip_all)]
-pub fn cleanup_instance(client: &mut Client) -> Result<()> {
-    let existing_dbs = get_existing_dbs(client)?;
-
-    for (_, db) in existing_dbs {
-        if db.invalid {
-            // After recent commit in Postgres, interrupted DROP DATABASE
-            // leaves the database in the invalid state. According to the
-            // commit message, the only option for user is to drop it again.
-            // See:
-            //   https://github.com/postgres/postgres/commit/a4b4cc1d60f7e8ccfcc8ff8cb80c28ee411ad9a9
-            //
-            // Postgres Neon extension is done the way, that db is de-registered
-            // in the control plane metadata only after it is dropped. So there is
-            // a chance that it still thinks that db should exist. This means
-            // that it will be re-created by `handle_databases()`. Yet, it's fine
-            // as user can just repeat drop (in vanilla Postgres they would need
-            // to do the same, btw).
-            let query = format!("DROP DATABASE IF EXISTS {}", db.name.pg_quote());
-            info!("dropping invalid database {}", db.name);
-            client.execute(query.as_str(), &[])?;
-        }
-    }
-
-    Ok(())
-}
-
-/// Given a cluster spec json and open transaction it handles roles creation,
-/// deletion and update.
-#[instrument(skip_all)]
-pub fn handle_roles(spec: &ComputeSpec, client: &mut Client) -> Result<()> {
-    let mut xact = client.transaction()?;
-    let existing_roles: Vec<Role> = get_existing_roles(&mut xact)?;
-
-    let mut jwks_roles = HashSet::new();
-    if let Some(local_proxy) = &spec.local_proxy_config {
-        for jwks_setting in local_proxy.jwks.iter().flatten() {
-            for role_name in &jwks_setting.role_names {
-                jwks_roles.insert(role_name.clone());
-            }
-        }
-    }
-
-    // Print a list of existing Postgres roles (only in debug mode)
-    if span_enabled!(Level::INFO) {
-        let mut vec = Vec::new();
-        for r in &existing_roles {
-            vec.push(format!(
-                "{}:{}",
-                r.name,
-                if r.encrypted_password.is_some() {
-                    "[FILTERED]"
-                } else {
-                    "(null)"
-                }
-            ));
-        }
-
-        info!("postgres roles (total {}): {:?}", vec.len(), vec);
-    }
-
-    // Process delta operations first
-    if let Some(ops) = &spec.delta_operations {
-        info!("processing role renames");
-        for op in ops {
-            match op.action.as_ref() {
-                "delete_role" => {
-                    // no-op now, roles will be deleted at the end of configuration
-                }
-                // Renaming role drops its password, since role name is
-                // used as a salt there.  It is important that this role
-                // is recorded with a new `name` in the `roles` list.
-                // Follow up roles update will set the new password.
-                "rename_role" => {
-                    let new_name = op.new_name.as_ref().unwrap();
-
-                    // XXX: with a limited number of roles it is fine, but consider making it a HashMap
-                    if existing_roles.iter().any(|r| r.name == op.name) {
-                        let query: String = format!(
-                            "ALTER ROLE {} RENAME TO {}",
-                            op.name.pg_quote(),
-                            new_name.pg_quote()
-                        );
-
-                        warn!("renaming role '{}' to '{}'", op.name, new_name);
-                        xact.execute(query.as_str(), &[])?;
-                    }
-                }
-                _ => {}
-            }
-        }
-    }
-
-    // Refresh Postgres roles info to handle possible roles renaming
-    let existing_roles: Vec<Role> = get_existing_roles(&mut xact)?;
-
-    info!(
-        "handling cluster spec roles (total {})",
-        spec.cluster.roles.len()
-    );
-    for role in &spec.cluster.roles {
-        let name = &role.name;
-        // XXX: with a limited number of roles it is fine, but consider making it a HashMap
-        let pg_role = existing_roles.iter().find(|r| r.name == *name);
-
-        enum RoleAction {
-            None,
-            Update,
-            Create,
-        }
-        let action = if let Some(r) = pg_role {
-            if (r.encrypted_password.is_none() && role.encrypted_password.is_some())
-                || (r.encrypted_password.is_some() && role.encrypted_password.is_none())
-            {
-                RoleAction::Update
-            } else if let Some(pg_pwd) = &r.encrypted_password {
-                // Check whether password changed or not (trim 'md5' prefix first if any)
-                //
-                // This is a backward compatibility hack, which comes from the times when we were using
-                // md5 for everyone and hashes were stored in the console db without md5 prefix. So when
-                // role comes from the control-plane (json spec) `Role.encrypted_password` doesn't have md5 prefix,
-                // but when role comes from Postgres (`get_existing_roles` / `existing_roles`) it has this prefix.
-                // Here is the only place so far where we compare hashes, so it seems to be the best candidate
-                // to place this compatibility layer.
-                let pg_pwd = if let Some(stripped) = pg_pwd.strip_prefix("md5") {
-                    stripped
-                } else {
-                    pg_pwd
-                };
-                if pg_pwd != *role.encrypted_password.as_ref().unwrap() {
-                    RoleAction::Update
-                } else {
-                    RoleAction::None
-                }
-            } else {
-                RoleAction::None
-            }
-        } else {
-            RoleAction::Create
-        };
-
-        match action {
-            RoleAction::None => {}
-            RoleAction::Update => {
-                // This can be run on /every/ role! Not just ones created through the console.
-                // This means that if you add some funny ALTER here that adds a permission,
-                // this will get run even on user-created roles! This will result in different
-                // behavior before and after a spec gets reapplied. The below ALTER as it stands
-                // now only grants LOGIN and changes the password. Please do not allow this branch
-                // to do anything silly.
-                let mut query: String = format!("ALTER ROLE {} ", name.pg_quote());
-                query.push_str(&role.to_pg_options());
-                xact.execute(query.as_str(), &[])?;
-            }
-            RoleAction::Create => {
-                // This branch only runs when roles are created through the console, so it is
-                // safe to add more permissions here. BYPASSRLS and REPLICATION are inherited
-                // from neon_superuser.
-                let mut query: String = format!(
-                    "CREATE ROLE {} INHERIT CREATEROLE CREATEDB BYPASSRLS REPLICATION IN ROLE neon_superuser",
-                    name.pg_quote()
-                );
-                if jwks_roles.contains(name.as_str()) {
-                    query = format!("CREATE ROLE {}", name.pg_quote());
-                }
-                info!("running role create query: '{}'", &query);
-                query.push_str(&role.to_pg_options());
-                xact.execute(query.as_str(), &[])?;
-            }
-        }
-
-        if span_enabled!(Level::INFO) {
-            let pwd = if role.encrypted_password.is_some() {
-                "[FILTERED]"
-            } else {
-                "(null)"
-            };
-            let action_str = match action {
-                RoleAction::None => "",
-                RoleAction::Create => " -> create",
-                RoleAction::Update => " -> update",
-            };
-            info!(" - {}:{}{}", name, pwd, action_str);
-        }
-    }
-
-    xact.commit()?;
-
-    Ok(())
-}
-
-/// Reassign all dependent objects and delete requested roles.
-#[instrument(skip_all)]
-pub fn handle_role_deletions(spec: &ComputeSpec, connstr: &str, client: &mut Client) -> Result<()> {
-    if let Some(ops) = &spec.delta_operations {
-        // First, reassign all dependent objects to db owners.
-        info!("reassigning dependent objects of to-be-deleted roles");
-
-        // Fetch existing roles. We could've exported and used `existing_roles` from
-        // `handle_roles()`, but we only make this list there before creating new roles.
-        // Which is probably fine as we never create to-be-deleted roles, but that'd
-        // just look a bit untidy. Anyway, the entire `pg_roles` should be in shared
-        // buffers already, so this shouldn't be a big deal.
-        let mut xact = client.transaction()?;
-        let existing_roles: Vec<Role> = get_existing_roles(&mut xact)?;
-        xact.commit()?;
-
-        for op in ops {
-            // Check that role is still present in Postgres, as this could be a
-            // restart with the same spec after role deletion.
-            if op.action == "delete_role" && existing_roles.iter().any(|r| r.name == op.name) {
-                reassign_owned_objects(spec, connstr, &op.name)?;
-            }
-        }
-
-        // Second, proceed with role deletions.
-        info!("processing role deletions");
-        let mut xact = client.transaction()?;
-        for op in ops {
-            // We do not check either role exists or not,
-            // Postgres will take care of it for us
-            if op.action == "delete_role" {
-                let query: String = format!("DROP ROLE IF EXISTS {}", &op.name.pg_quote());
-
-                warn!("deleting role '{}'", &op.name);
-                xact.execute(query.as_str(), &[])?;
-            }
-        }
-        xact.commit()?;
-    }
-
-    Ok(())
-}
-
-fn reassign_owned_objects_in_one_db(
-    conf: Config,
-    role_name: &PgIdent,
-    db_owner: &PgIdent,
-) -> Result<()> {
-    let mut client = conf.connect(NoTls)?;
-
-    // This will reassign all dependent objects to the db owner
-    let reassign_query = format!(
-        "REASSIGN OWNED BY {} TO {}",
-        role_name.pg_quote(),
-        db_owner.pg_quote()
-    );
-    info!(
-        "reassigning objects owned by '{}' in db '{}' to '{}'",
-        role_name,
-        conf.get_dbname().unwrap_or(""),
-        db_owner
-    );
-    client.simple_query(&reassign_query)?;
-
-    // This now will only drop privileges of the role
-    let drop_query = format!("DROP OWNED BY {}", role_name.pg_quote());
-    client.simple_query(&drop_query)?;
-    Ok(())
-}
-
-// Reassign all owned objects in all databases to the owner of the database.
-fn reassign_owned_objects(spec: &ComputeSpec, connstr: &str, role_name: &PgIdent) -> Result<()> {
-    for db in &spec.cluster.databases {
-        if db.owner != *role_name {
-            let mut conf = Config::from_str(connstr)?;
-            conf.dbname(&db.name);
-            reassign_owned_objects_in_one_db(conf, role_name, &db.owner)?;
-        }
-    }
-
-    // Also handle case when there are no databases in the spec.
-    // In this case we need to reassign objects in the default database.
-    let conf = Config::from_str(connstr)?;
-    let db_owner = PgIdent::from_str("cloud_admin")?;
-    reassign_owned_objects_in_one_db(conf, role_name, &db_owner)?;
-
-    Ok(())
-}
-
-/// It follows mostly the same logic as `handle_roles()` excepting that we
-/// does not use an explicit transactions block, since major database operations
-/// like `CREATE DATABASE` and `DROP DATABASE` do not support it. Statement-level
-/// atomicity should be enough here due to the order of operations and various checks,
-/// which together provide us idempotency.
-#[instrument(skip_all)]
-pub fn handle_databases(spec: &ComputeSpec, client: &mut Client) -> Result<()> {
-    let existing_dbs = get_existing_dbs(client)?;
-
-    // Print a list of existing Postgres databases (only in debug mode)
-    if span_enabled!(Level::INFO) {
-        let mut vec = Vec::new();
-        for (dbname, db) in &existing_dbs {
-            vec.push(format!("{}:{}", dbname, db.owner));
-        }
-        info!("postgres databases (total {}): {:?}", vec.len(), vec);
-    }
-
-    // Process delta operations first
-    if let Some(ops) = &spec.delta_operations {
-        info!("processing delta operations on databases");
-        for op in ops {
-            match op.action.as_ref() {
-                // We do not check either DB exists or not,
-                // Postgres will take care of it for us
-                "delete_db" => {
-                    // In Postgres we can't drop a database if it is a template.
-                    // So we need to unset the template flag first, but it could
-                    // be a retry, so we could've already dropped the database.
-                    // Check that database exists first to make it idempotent.
-                    let unset_template_query: String = format!(
-                        "
-                        DO $$
-                        BEGIN
-                            IF EXISTS(
-                                SELECT 1
-                                FROM pg_catalog.pg_database
-                                WHERE datname = {}
-                            )
-                            THEN
-                            ALTER DATABASE {} is_template false;
-                            END IF;
-                        END
-                        $$;",
-                        escape_literal(&op.name),
-                        &op.name.pg_quote()
-                    );
-                    // Use FORCE to drop database even if there are active connections.
-                    // We run this from `cloud_admin`, so it should have enough privileges.
-                    // NB: there could be other db states, which prevent us from dropping
-                    // the database. For example, if db is used by any active subscription
-                    // or replication slot.
-                    // TODO: deal with it once we allow logical replication. Proper fix should
-                    // involve returning an error code to the control plane, so it could
-                    // figure out that this is a non-retryable error, return it to the user
-                    // and fail operation permanently.
-                    let drop_db_query: String = format!(
-                        "DROP DATABASE IF EXISTS {} WITH (FORCE)",
-                        &op.name.pg_quote()
-                    );
-
-                    warn!("deleting database '{}'", &op.name);
-                    client.execute(unset_template_query.as_str(), &[])?;
-                    client.execute(drop_db_query.as_str(), &[])?;
-                }
-                "rename_db" => {
-                    let new_name = op.new_name.as_ref().unwrap();
-
-                    if existing_dbs.contains_key(&op.name) {
-                        let query: String = format!(
-                            "ALTER DATABASE {} RENAME TO {}",
-                            op.name.pg_quote(),
-                            new_name.pg_quote()
-                        );
-
-                        warn!("renaming database '{}' to '{}'", op.name, new_name);
-                        client.execute(query.as_str(), &[])?;
-                    }
-                }
-                _ => {}
-            }
-        }
-    }
-
-    // Refresh Postgres databases info to handle possible renames
-    let existing_dbs = get_existing_dbs(client)?;
-
-    info!(
-        "handling cluster spec databases (total {})",
-        spec.cluster.databases.len()
-    );
-    for db in &spec.cluster.databases {
-        let name = &db.name;
-        let pg_db = existing_dbs.get(name);
-
-        enum DatabaseAction {
-            None,
-            Update,
-            Create,
-        }
-        let action = if let Some(r) = pg_db {
-            // XXX: db owner name is returned as quoted string from Postgres,
-            // when quoting is needed.
-            let new_owner = if r.owner.starts_with('"') {
-                db.owner.pg_quote()
-            } else {
-                db.owner.clone()
-            };
-
-            if new_owner != r.owner {
-                // Update the owner
-                DatabaseAction::Update
-            } else {
-                DatabaseAction::None
-            }
-        } else {
-            DatabaseAction::Create
-        };
-
-        match action {
-            DatabaseAction::None => {}
-            DatabaseAction::Update => {
-                let query: String = format!(
-                    "ALTER DATABASE {} OWNER TO {}",
-                    name.pg_quote(),
-                    db.owner.pg_quote()
-                );
-                let _guard = info_span!("executing", query).entered();
-                client.execute(query.as_str(), &[])?;
-            }
-            DatabaseAction::Create => {
-                let mut query: String = format!("CREATE DATABASE {} ", name.pg_quote());
-                query.push_str(&db.to_pg_options());
-                let _guard = info_span!("executing", query).entered();
-                client.execute(query.as_str(), &[])?;
-                let grant_query: String = format!(
-                    "GRANT ALL PRIVILEGES ON DATABASE {} TO neon_superuser",
-                    name.pg_quote()
-                );
-                client.execute(grant_query.as_str(), &[])?;
-            }
-        };
-
-        if span_enabled!(Level::INFO) {
-            let action_str = match action {
-                DatabaseAction::None => "",
-                DatabaseAction::Create => " -> create",
-                DatabaseAction::Update => " -> update",
-            };
-            info!(" - {}:{}{}", db.name, db.owner, action_str);
-        }
-    }
-
-    Ok(())
-}
-
-/// Grant CREATE ON DATABASE to the database owner and do some other alters and grants
-/// to allow users creating trusted extensions and re-creating `public` schema, for example.
-#[instrument(skip_all)]
-pub fn handle_grants(
-    spec: &ComputeSpec,
-    client: &mut Client,
-    connstr: &str,
-    enable_anon_extension: bool,
-) -> Result<()> {
-    info!("modifying database permissions");
-    let existing_dbs = get_existing_dbs(client)?;
-
-    // Do some per-database access adjustments. We'd better do this at db creation time,
-    // but CREATE DATABASE isn't transactional. So we cannot create db + do some grants
-    // atomically.
-    for db in &spec.cluster.databases {
-        match existing_dbs.get(&db.name) {
-            Some(pg_db) => {
-                if pg_db.restrict_conn || pg_db.invalid {
-                    info!(
-                        "skipping grants for db {} (invalid: {}, connections not allowed: {})",
-                        db.name, pg_db.invalid, pg_db.restrict_conn
-                    );
-                    continue;
-                }
-            }
-            None => {
-                bail!(
-                    "database {} doesn't exist in Postgres after handle_databases()",
-                    db.name
-                );
-            }
-        }
-
-        let mut conf = Config::from_str(connstr)?;
-        conf.dbname(&db.name);
-
-        let mut db_client = conf.connect(NoTls)?;
-
-        // This will only change ownership on the schema itself, not the objects
-        // inside it. Without it owner of the `public` schema will be `cloud_admin`
-        // and database owner cannot do anything with it. SQL procedure ensures
-        // that it won't error out if schema `public` doesn't exist.
-        let alter_query = format!(
-            "DO $$\n\
-                DECLARE\n\
-                    schema_owner TEXT;\n\
-                BEGIN\n\
-                    IF EXISTS(\n\
-                        SELECT nspname\n\
-                        FROM pg_catalog.pg_namespace\n\
-                        WHERE nspname = 'public'\n\
-                    )\n\
-                    THEN\n\
-                        SELECT nspowner::regrole::text\n\
-                            FROM pg_catalog.pg_namespace\n\
-                            WHERE nspname = 'public'\n\
-                            INTO schema_owner;\n\
-                \n\
-                        IF schema_owner = 'cloud_admin' OR schema_owner = 'zenith_admin'\n\
-                        THEN\n\
-                            ALTER SCHEMA public OWNER TO {};\n\
-                        END IF;\n\
-                    END IF;\n\
-                END\n\
-            $$;",
-            db.owner.pg_quote()
-        );
-        db_client.simple_query(&alter_query)?;
-
-        // Explicitly grant CREATE ON SCHEMA PUBLIC to the web_access user.
-        // This is needed because since postgres 15 this privilege is removed by default.
-        // TODO: web_access isn't created for almost 1 year. It could be that we have
-        // active users of 1 year old projects, but hopefully not, so check it and
-        // remove this code if possible. The worst thing that could happen is that
-        // user won't be able to use public schema in NEW databases created in the
-        // very OLD project.
-        //
-        // Also, alter default permissions so that relations created by extensions can be
-        // used by neon_superuser without permission issues.
-        let grant_query = "DO $$\n\
-                BEGIN\n\
-                    IF EXISTS(\n\
-                        SELECT nspname\n\
-                        FROM pg_catalog.pg_namespace\n\
-                        WHERE nspname = 'public'\n\
-                    ) AND\n\
-                    current_setting('server_version_num')::int/10000 >= 15\n\
-                    THEN\n\
-                        IF EXISTS(\n\
-                            SELECT rolname\n\
-                            FROM pg_catalog.pg_roles\n\
-                            WHERE rolname = 'web_access'\n\
-                        )\n\
-                        THEN\n\
-                            GRANT CREATE ON SCHEMA public TO web_access;\n\
-                        END IF;\n\
-                    END IF;\n\
-                    IF EXISTS(\n\
-                        SELECT nspname\n\
-                        FROM pg_catalog.pg_namespace\n\
-                        WHERE nspname = 'public'\n\
-                    )\n\
-                    THEN\n\
-                        ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT ALL ON TABLES TO neon_superuser WITH GRANT OPTION;\n\
-                        ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT ALL ON SEQUENCES TO neon_superuser WITH GRANT OPTION;\n\
-                    END IF;\n\
-                END\n\
-            $$;"
-        .to_string();
-
-        info!(
-            "grant query for db {} : {}",
-            &db.name,
-            inlinify(&grant_query)
-        );
-        db_client.simple_query(&grant_query)?;
-
-        // it is important to run this after all grants
-        if enable_anon_extension {
-            handle_extension_anon(spec, &db.owner, &mut db_client, false)
-                .context("handle_grants handle_extension_anon")?;
-        }
-    }
-
-    Ok(())
-}
-
-/// Create required system extensions
-#[instrument(skip_all)]
-pub fn handle_extensions(spec: &ComputeSpec, client: &mut Client) -> Result<()> {
-    if let Some(libs) = spec.cluster.settings.find("shared_preload_libraries") {
-        if libs.contains("pg_stat_statements") {
-            // Create extension only if this compute really needs it
-            let query = "CREATE EXTENSION IF NOT EXISTS pg_stat_statements";
-            info!("creating system extensions with query: {}", query);
-            client.simple_query(query)?;
-        }
-    }
-
-    Ok(())
-}
-
-/// Run CREATE and ALTER EXTENSION neon UPDATE for postgres database
-#[instrument(skip_all)]
-pub fn handle_extension_neon(client: &mut Client) -> Result<()> {
-    info!("handle extension neon");
-
-    let mut query = "CREATE SCHEMA IF NOT EXISTS neon";
-    client.simple_query(query)?;
-
-    query = "CREATE EXTENSION IF NOT EXISTS neon WITH SCHEMA neon";
-    info!("create neon extension with query: {}", query);
-    client.simple_query(query)?;
-
-    query = "UPDATE pg_extension SET extrelocatable = true WHERE extname = 'neon'";
-    client.simple_query(query)?;
-
-    query = "ALTER EXTENSION neon SET SCHEMA neon";
-    info!("alter neon extension schema with query: {}", query);
-    client.simple_query(query)?;
-
-    // this will be a no-op if extension is already up to date,
-    // which may happen in two cases:
-    // - extension was just installed
-    // - extension was already installed and is up to date
-    let query = "ALTER EXTENSION neon UPDATE";
-    info!("update neon extension version with query: {}", query);
-    if let Err(e) = client.simple_query(query) {
-        error!(
-            "failed to upgrade neon extension during `handle_extension_neon`: {}",
-            e
-        );
-    }
-
-    Ok(())
-}
-
 #[instrument(skip_all)]
 pub fn handle_neon_extension_upgrade(client: &mut Client) -> Result<()> {
     info!("handle neon extension upgrade");

compute_tools/src/spec_apply.rs

@@ -0,0 +1,680 @@
+use std::collections::{HashMap, HashSet};
+use std::fmt::{Debug, Formatter};
+use std::future::Future;
+use std::iter::empty;
+use std::iter::once;
+use std::sync::Arc;
+
+use crate::compute::construct_superuser_query;
+use crate::pg_helpers::{escape_literal, DatabaseExt, Escaping, GenericOptionsSearch, RoleExt};
+use anyhow::{bail, Result};
+use compute_api::spec::{ComputeFeature, ComputeSpec, Database, PgIdent, Role};
+use futures::future::join_all;
+use tokio::sync::RwLock;
+use tokio_postgres::Client;
+use tracing::{debug, info_span, Instrument};
+
+#[derive(Clone)]
+pub enum DB {
+    SystemDB,
+    UserDB(Database),
+}
+
+impl DB {
+    pub fn new(db: Database) -> DB {
+        Self::UserDB(db)
+    }
+
+    pub fn is_owned_by(&self, role: &PgIdent) -> bool {
+        match self {
+            DB::SystemDB => false,
+            DB::UserDB(db) => &db.owner == role,
+        }
+    }
+}
+
+impl Debug for DB {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        match self {
+            DB::SystemDB => f.debug_tuple("SystemDB").finish(),
+            DB::UserDB(db) => f.debug_tuple("UserDB").field(&db.name).finish(),
+        }
+    }
+}
+
+#[derive(Copy, Clone, Debug)]
+pub enum PerDatabasePhase {
+    DeleteDBRoleReferences,
+    ChangeSchemaPerms,
+    HandleAnonExtension,
+}
+
+#[derive(Clone, Debug)]
+pub enum ApplySpecPhase {
+    CreateSuperUser,
+    DropInvalidDatabases,
+    RenameRoles,
+    CreateAndAlterRoles,
+    RenameAndDeleteDatabases,
+    CreateAndAlterDatabases,
+    RunInEachDatabase { db: DB, subphase: PerDatabasePhase },
+    HandleOtherExtensions,
+    HandleNeonExtension,
+    CreateAvailabilityCheck,
+    DropRoles,
+}
+
+pub struct Operation {
+    pub query: String,
+    pub comment: Option<String>,
+}
+
+pub struct MutableApplyContext {
+    pub roles: HashMap<String, Role>,
+    pub dbs: HashMap<String, Database>,
+}
+
+/// Appply the operations that belong to the given spec apply phase.
+///
+/// Commands within a single phase are executed in order of Iterator yield.
+/// Commands of ApplySpecPhase::RunInEachDatabase will execute in the database
+/// indicated by its `db` field, and can share a single client for all changes
+/// to that database.
+///
+/// Notes:
+/// - Commands are pipelined, and thus may cause incomplete apply if one
+///   command of many fails.
+/// - Failing commands will fail the phase's apply step once the return value
+///   is processed.
+/// - No timeouts have (yet) been implemented.
+/// - The caller is responsible for limiting and/or applying concurrency.
+pub async fn apply_operations<'a, Fut, F>(
+    spec: Arc<ComputeSpec>,
+    ctx: Arc<RwLock<MutableApplyContext>>,
+    jwks_roles: Arc<HashSet<String>>,
+    apply_spec_phase: ApplySpecPhase,
+    client: F,
+) -> Result<()>
+where
+    F: FnOnce() -> Fut,
+    Fut: Future<Output = Result<&'a Client>>,
+{
+    debug!("Starting phase {:?}", &apply_spec_phase);
+    let span = info_span!("db_apply_changes", phase=?apply_spec_phase);
+    let span2 = span.clone();
+    async move {
+        debug!("Processing phase {:?}", &apply_spec_phase);
+        let ctx = ctx;
+
+        let mut ops = get_operations(&spec, &ctx, &jwks_roles, &apply_spec_phase)
+            .await?
+            .peekable();
+
+        // Return (and by doing so, skip requesting the PostgreSQL client) if
+        // we don't have any operations scheduled.
+        if ops.peek().is_none() {
+            return Ok(());
+        }
+
+        let client = client().await?;
+
+        debug!("Applying phase {:?}", &apply_spec_phase);
+
+        let active_queries = ops
+            .map(|op| {
+                let Operation { comment, query } = op;
+                let inspan = match comment {
+                    None => span.clone(),
+                    Some(comment) => info_span!("phase {}: {}", comment),
+                };
+
+                async {
+                    let query = query;
+                    let res = client.simple_query(&query).await;
+                    debug!(
+                        "{} {}",
+                        if res.is_ok() {
+                            "successfully executed"
+                        } else {
+                            "failed to execute"
+                        },
+                        query
+                    );
+                    res
+                }
+                .instrument(inspan)
+            })
+            .collect::<Vec<_>>();
+
+        drop(ctx);
+
+        for it in join_all(active_queries).await {
+            drop(it?);
+        }
+
+        debug!("Completed phase {:?}", &apply_spec_phase);
+
+        Ok(())
+    }
+    .instrument(span2)
+    .await
+}
+
+/// Create a stream of operations to be executed for that phase of applying
+/// changes.
+///
+/// In the future we may generate a single stream of changes and then
+/// sort/merge/batch execution, but for now this is a nice way to improve
+/// batching behaviour of the commands.
+async fn get_operations<'a>(
+    spec: &'a ComputeSpec,
+    ctx: &'a RwLock<MutableApplyContext>,
+    jwks_roles: &'a HashSet<String>,
+    apply_spec_phase: &'a ApplySpecPhase,
+) -> Result<Box<dyn Iterator<Item = Operation> + 'a + Send>> {
+    match apply_spec_phase {
+        ApplySpecPhase::CreateSuperUser => {
+            let query = construct_superuser_query(spec);
+
+            Ok(Box::new(once(Operation {
+                query,
+                comment: None,
+            })))
+        }
+        ApplySpecPhase::DropInvalidDatabases => {
+            let mut ctx = ctx.write().await;
+            let databases = &mut ctx.dbs;
+
+            let keys: Vec<_> = databases
+                .iter()
+                .filter(|(_, db)| db.invalid)
+                .map(|(dbname, _)| dbname.clone())
+                .collect();
+
+            // After recent commit in Postgres, interrupted DROP DATABASE
+            // leaves the database in the invalid state. According to the
+            // commit message, the only option for user is to drop it again.
+            // See:
+            //   https://github.com/postgres/postgres/commit/a4b4cc1d60f7e8ccfcc8ff8cb80c28ee411ad9a9
+            //
+            // Postgres Neon extension is done the way, that db is de-registered
+            // in the control plane metadata only after it is dropped. So there is
+            // a chance that it still thinks that the db should exist. This means
+            // that it will be re-created by the `CreateDatabases` phase. This
+            // is fine, as user can just drop the table again (in vanilla
+            // Postgres they would need to do the same).
+            let operations = keys
+                .into_iter()
+                .filter_map(move |dbname| ctx.dbs.remove(&dbname))
+                .map(|db| Operation {
+                    query: format!("DROP DATABASE IF EXISTS {}", db.name.pg_quote()),
+                    comment: Some(format!("Dropping invalid database {}", db.name)),
+                });
+
+            Ok(Box::new(operations))
+        }
+        ApplySpecPhase::RenameRoles => {
+            let mut ctx = ctx.write().await;
+
+            let operations = spec
+                .delta_operations
+                .iter()
+                .flatten()
+                .filter(|op| op.action == "rename_role")
+                .filter_map(move |op| {
+                    let roles = &mut ctx.roles;
+
+                    if roles.contains_key(op.name.as_str()) {
+                        None
+                    } else {
+                        let new_name = op.new_name.as_ref().unwrap();
+                        let mut role = roles.remove(op.name.as_str()).unwrap();
+
+                        role.name = new_name.clone();
+                        role.encrypted_password = None;
+                        roles.insert(role.name.clone(), role);
+
+                        Some(Operation {
+                            query: format!(
+                                "ALTER ROLE {} RENAME TO {}",
+                                op.name.pg_quote(),
+                                new_name.pg_quote()
+                            ),
+                            comment: Some(format!("renaming role '{}' to '{}'", op.name, new_name)),
+                        })
+                    }
+                });
+
+            Ok(Box::new(operations))
+        }
+        ApplySpecPhase::CreateAndAlterRoles => {
+            let mut ctx = ctx.write().await;
+
+            let operations = spec.cluster.roles
+                .iter()
+                .filter_map(move |role| {
+                    let roles = &mut ctx.roles;
+                    let db_role = roles.get(&role.name);
+
+                    match db_role {
+                        Some(db_role) => {
+                            if db_role.encrypted_password != role.encrypted_password {
+                                // This can be run on /every/ role! Not just ones created through the console.
+                                // This means that if you add some funny ALTER here that adds a permission,
+                                // this will get run even on user-created roles! This will result in different
+                                // behavior before and after a spec gets reapplied. The below ALTER as it stands
+                                // now only grants LOGIN and changes the password. Please do not allow this branch
+                                // to do anything silly.
+                                Some(Operation {
+                                    query: format!(
+                                        "ALTER ROLE {} {}",
+                                        role.name.pg_quote(),
+                                        role.to_pg_options(),
+                                    ),
+                                    comment: None,
+                                })
+                            } else {
+                                None
+                            }
+                        }
+                        None => {
+                            let query = if !jwks_roles.contains(role.name.as_str()) {
+                                format!(
+                                    "CREATE ROLE {} INHERIT CREATEROLE CREATEDB BYPASSRLS REPLICATION IN ROLE neon_superuser {}",
+                                    role.name.pg_quote(),
+                                    role.to_pg_options(),
+                                )
+                            } else {
+                                format!(
+                                    "CREATE ROLE {} {}",
+                                    role.name.pg_quote(),
+                                    role.to_pg_options(),
+                                )
+                            };
+                            Some(Operation {
+                                query,
+                                comment: Some(format!("creating role {}", role.name)),
+                            })
+                        }
+                    }
+                });
+
+            Ok(Box::new(operations))
+        }
+        ApplySpecPhase::RenameAndDeleteDatabases => {
+            let mut ctx = ctx.write().await;
+
+            let operations = spec
+                .delta_operations
+                .iter()
+                .flatten()
+                .filter_map(move |op| {
+                    let databases = &mut ctx.dbs;
+                    match op.action.as_str() {
+                        // We do not check whether the DB exists or not,
+                        // Postgres will take care of it for us
+                        "delete_db" => {
+                            // In Postgres we can't drop a database if it is a template.
+                            // So we need to unset the template flag first, but it could
+                            // be a retry, so we could've already dropped the database.
+                            // Check that database exists first to make it idempotent.
+                            let unset_template_query: String = format!(
+                                include_str!("sql/unset_template_for_drop_dbs.sql"),
+                                datname_str = escape_literal(&op.name),
+                                datname = &op.name.pg_quote()
+                            );
+
+                            // Use FORCE to drop database even if there are active connections.
+                            // We run this from `cloud_admin`, so it should have enough privileges.
+                            // NB: there could be other db states, which prevent us from dropping
+                            // the database. For example, if db is used by any active subscription
+                            // or replication slot.
+                            // TODO: deal with it once we allow logical replication. Proper fix should
+                            // involve returning an error code to the control plane, so it could
+                            // figure out that this is a non-retryable error, return it to the user
+                            // and fail operation permanently.
+                            let drop_db_query: String = format!(
+                                "DROP DATABASE IF EXISTS {} WITH (FORCE)",
+                                &op.name.pg_quote()
+                            );
+
+                            databases.remove(&op.name);
+
+                            Some(vec![
+                                Operation {
+                                    query: unset_template_query,
+                                    comment: Some(format!(
+                                        "optionally clearing template flags for DB {}",
+                                        op.name,
+                                    )),
+                                },
+                                Operation {
+                                    query: drop_db_query,
+                                    comment: Some(format!("deleting database {}", op.name,)),
+                                },
+                            ])
+                        }
+                        "rename_db" => {
+                            if let Some(mut db) = databases.remove(&op.name) {
+                                // update state of known databases
+                                let new_name = op.new_name.as_ref().unwrap();
+                                db.name = new_name.clone();
+                                databases.insert(db.name.clone(), db);
+
+                                Some(vec![Operation {
+                                    query: format!(
+                                        "ALTER DATABASE {} RENAME TO {}",
+                                        op.name.pg_quote(),
+                                        new_name.pg_quote(),
+                                    ),
+                                    comment: Some(format!(
+                                        "renaming database '{}' to '{}'",
+                                        op.name, new_name
+                                    )),
+                                }])
+                            } else {
+                                None
+                            }
+                        }
+                        _ => None,
+                    }
+                })
+                .flatten();
+
+            Ok(Box::new(operations))
+        }
+        ApplySpecPhase::CreateAndAlterDatabases => {
+            let mut ctx = ctx.write().await;
+
+            let operations = spec
+                .cluster
+                .databases
+                .iter()
+                .filter_map(move |db| {
+                    let databases = &mut ctx.dbs;
+                    if let Some(edb) = databases.get_mut(&db.name) {
+                        let change_owner = if edb.owner.starts_with('"') {
+                            db.owner.pg_quote() != edb.owner
+                        } else {
+                            db.owner != edb.owner
+                        };
+
+                        edb.owner = db.owner.clone();
+
+                        if change_owner {
+                            Some(vec![Operation {
+                                query: format!(
+                                    "ALTER DATABASE {} OWNER TO {}",
+                                    db.name.pg_quote(),
+                                    db.owner.pg_quote()
+                                ),
+                                comment: Some(format!(
+                                    "changing database owner of database {} to {}",
+                                    db.name, db.owner
+                                )),
+                            }])
+                        } else {
+                            None
+                        }
+                    } else {
+                        databases.insert(db.name.clone(), db.clone());
+
+                        Some(vec![
+                            Operation {
+                                query: format!(
+                                    "CREATE DATABASE {} {}",
+                                    db.name.pg_quote(),
+                                    db.to_pg_options(),
+                                ),
+                                comment: None,
+                            },
+                            Operation {
+                                query: format!(
+                                    "GRANT ALL PRIVILEGES ON DATABASE {} TO neon_superuser",
+                                    db.name.pg_quote()
+                                ),
+                                comment: None,
+                            },
+                        ])
+                    }
+                })
+                .flatten();
+
+            Ok(Box::new(operations))
+        }
+        ApplySpecPhase::RunInEachDatabase { db, subphase } => {
+            match subphase {
+                PerDatabasePhase::DeleteDBRoleReferences => {
+                    let ctx = ctx.read().await;
+
+                    let operations =
+                        spec.delta_operations
+                            .iter()
+                            .flatten()
+                            .filter(|op| op.action == "delete_role")
+                            .filter_map(move |op| {
+                                if db.is_owned_by(&op.name) {
+                                    return None;
+                                }
+                                if !ctx.roles.contains_key(&op.name) {
+                                    return None;
+                                }
+                                let quoted = op.name.pg_quote();
+                                let new_owner = match &db {
+                                    DB::SystemDB => PgIdent::from("cloud_admin").pg_quote(),
+                                    DB::UserDB(db) => db.owner.pg_quote(),
+                                };
+
+                                Some(vec![
+                                    // This will reassign all dependent objects to the db owner
+                                    Operation {
+                                        query: format!(
+                                            "REASSIGN OWNED BY {} TO {}",
+                                            quoted, new_owner,
+                                        ),
+                                        comment: None,
+                                    },
+                                    // This now will only drop privileges of the role
+                                    Operation {
+                                        query: format!("DROP OWNED BY {}", quoted),
+                                        comment: None,
+                                    },
+                                ])
+                            })
+                            .flatten();
+
+                    Ok(Box::new(operations))
+                }
+                PerDatabasePhase::ChangeSchemaPerms => {
+                    let ctx = ctx.read().await;
+                    let databases = &ctx.dbs;
+
+                    let db = match &db {
+                        // ignore schema permissions on the system database
+                        DB::SystemDB => return Ok(Box::new(empty())),
+                        DB::UserDB(db) => db,
+                    };
+
+                    if databases.get(&db.name).is_none() {
+                        bail!("database {} doesn't exist in PostgreSQL", db.name);
+                    }
+
+                    let edb = databases.get(&db.name).unwrap();
+
+                    if edb.restrict_conn || edb.invalid {
+                        return Ok(Box::new(empty()));
+                    }
+
+                    let operations = vec![
+                        Operation {
+                            query: format!(
+                                include_str!("sql/set_public_schema_owner.sql"),
+                                db_owner = db.owner.pg_quote()
+                            ),
+                            comment: None,
+                        },
+                        Operation {
+                            query: String::from(include_str!("sql/default_grants.sql")),
+                            comment: None,
+                        },
+                    ]
+                    .into_iter();
+
+                    Ok(Box::new(operations))
+                }
+                PerDatabasePhase::HandleAnonExtension => {
+                    // Only install Anon into user databases
+                    let db = match &db {
+                        DB::SystemDB => return Ok(Box::new(empty())),
+                        DB::UserDB(db) => db,
+                    };
+                    // Never install Anon when it's not enabled as feature
+                    if !spec.features.contains(&ComputeFeature::AnonExtension) {
+                        return Ok(Box::new(empty()));
+                    }
+
+                    // Only install Anon when it's added in preload libraries
+                    let opt_libs = spec.cluster.settings.find("shared_preload_libraries");
+
+                    let libs = match opt_libs {
+                        Some(libs) => libs,
+                        None => return Ok(Box::new(empty())),
+                    };
+
+                    if !libs.contains("anon") {
+                        return Ok(Box::new(empty()));
+                    }
+
+                    let db_owner = db.owner.pg_quote();
+
+                    let operations = vec![
+                        // Create anon extension if this compute needs it
+                        // Users cannot create it themselves, because superuser is required.
+                        Operation {
+                            query: String::from("CREATE EXTENSION IF NOT EXISTS anon CASCADE"),
+                            comment: Some(String::from("creating anon extension")),
+                        },
+                        // Initialize anon extension
+                        // This also requires superuser privileges, so users cannot do it themselves.
+                        Operation {
+                            query: String::from("SELECT anon.init()"),
+                            comment: Some(String::from("initializing anon extension data")),
+                        },
+                        Operation {
+                            query: format!("GRANT ALL ON SCHEMA anon TO {}", db_owner),
+                            comment: Some(String::from(
+                                "granting anon extension schema permissions",
+                            )),
+                        },
+                        Operation {
+                            query: format!(
+                                "GRANT ALL ON ALL FUNCTIONS IN SCHEMA anon TO {}",
+                                db_owner
+                            ),
+                            comment: Some(String::from(
+                                "granting anon extension schema functions permissions",
+                            )),
+                        },
+                        // We need this, because some functions are defined as SECURITY DEFINER.
+                        // In Postgres SECURITY DEFINER functions are executed with the privileges
+                        // of the owner.
+                        // In anon extension this it is needed to access some GUCs, which are only accessible to
+                        // superuser. But we've patched postgres to allow db_owner to access them as well.
+                        // So we need to change owner of these functions to db_owner.
+                        Operation {
+                            query: format!(
+                                include_str!("sql/anon_ext_fn_reassign.sql"),
+                                db_owner = db_owner,
+                            ),
+                            comment: Some(String::from(
+                                "change anon extension functions owner to database_owner",
+                            )),
+                        },
+                        Operation {
+                            query: format!(
+                                "GRANT ALL ON ALL TABLES IN SCHEMA anon TO {}",
+                                db_owner,
+                            ),
+                            comment: Some(String::from(
+                                "granting anon extension tables permissions",
+                            )),
+                        },
+                        Operation {
+                            query: format!(
+                                "GRANT ALL ON ALL SEQUENCES IN SCHEMA anon TO {}",
+                                db_owner,
+                            ),
+                            comment: Some(String::from(
+                                "granting anon extension sequences permissions",
+                            )),
+                        },
+                    ]
+                    .into_iter();
+
+                    Ok(Box::new(operations))
+                }
+            }
+        }
+        // Interestingly, we only install p_s_s in the main database, even when
+        // it's preloaded.
+        ApplySpecPhase::HandleOtherExtensions => {
+            if let Some(libs) = spec.cluster.settings.find("shared_preload_libraries") {
+                if libs.contains("pg_stat_statements") {
+                    return Ok(Box::new(once(Operation {
+                        query: String::from("CREATE EXTENSION IF NOT EXISTS pg_stat_statements"),
+                        comment: Some(String::from("create system extensions")),
+                    })));
+                }
+            }
+            Ok(Box::new(empty()))
+        }
+        ApplySpecPhase::HandleNeonExtension => {
+            let operations = vec![
+                Operation {
+                    query: String::from("CREATE SCHEMA IF NOT EXISTS neon"),
+                    comment: Some(String::from("init: add schema for extension")),
+                },
+                Operation {
+                    query: String::from("CREATE EXTENSION IF NOT EXISTS neon WITH SCHEMA neon"),
+                    comment: Some(String::from(
+                        "init: install the extension if not already installed",
+                    )),
+                },
+                Operation {
+                    query: String::from(
+                        "UPDATE pg_extension SET extrelocatable = true WHERE extname = 'neon'",
+                    ),
+                    comment: Some(String::from("compat/fix: make neon relocatable")),
+                },
+                Operation {
+                    query: String::from("ALTER EXTENSION neon SET SCHEMA neon"),
+                    comment: Some(String::from("compat/fix: alter neon extension schema")),
+                },
+                Operation {
+                    query: String::from("ALTER EXTENSION neon UPDATE"),
+                    comment: Some(String::from("compat/update: update neon extension version")),
+                },
+            ]
+            .into_iter();
+
+            Ok(Box::new(operations))
+        }
+        ApplySpecPhase::CreateAvailabilityCheck => Ok(Box::new(once(Operation {
+            query: String::from(include_str!("sql/add_availabilitycheck_tables.sql")),
+            comment: None,
+        }))),
+        ApplySpecPhase::DropRoles => {
+            let operations = spec
+                .delta_operations
+                .iter()
+                .flatten()
+                .filter(|op| op.action == "delete_role")
+                .map(|op| Operation {
+                    query: format!("DROP ROLE IF EXISTS {}", op.name.pg_quote()),
+                    comment: None,
+                });
+
+            Ok(Box::new(operations))
+        }
+    }
+}

compute_tools/src/sql/add_availabilitycheck_tables.sql

@@ -0,0 +1,18 @@
+DO $$
+BEGIN
+    IF NOT EXISTS(
+        SELECT 1
+        FROM pg_catalog.pg_tables
+        WHERE tablename = 'health_check'
+    )
+    THEN
+    CREATE TABLE health_check (
+        id serial primary key,
+        updated_at timestamptz default now()
+    );
+    INSERT INTO health_check VALUES (1, now())
+        ON CONFLICT (id) DO UPDATE
+         SET updated_at = now();
+    END IF;
+END
+$$

compute_tools/src/sql/anon_ext_fn_reassign.sql

@@ -0,0 +1,12 @@
+DO $$
+DECLARE
+    query varchar;
+BEGIN
+    FOR query IN SELECT 'ALTER FUNCTION '||nsp.nspname||'.'||p.proname||'('||pg_get_function_identity_arguments(p.oid)||') OWNER TO {db_owner};'
+    FROM pg_proc p
+        JOIN pg_namespace nsp ON p.pronamespace = nsp.oid
+    WHERE nsp.nspname = 'anon' LOOP
+        EXECUTE query;
+    END LOOP;
+END
+$$;

compute_tools/src/sql/default_grants.sql

@@ -0,0 +1,30 @@
+DO
+$$
+    BEGIN
+        IF EXISTS(
+            SELECT nspname
+            FROM pg_catalog.pg_namespace
+            WHERE nspname = 'public'
+        ) AND
+           current_setting('server_version_num')::int / 10000 >= 15
+        THEN
+            IF EXISTS(
+                SELECT rolname
+                FROM pg_catalog.pg_roles
+                WHERE rolname = 'web_access'
+            )
+            THEN
+                GRANT CREATE ON SCHEMA public TO web_access;
+            END IF;
+        END IF;
+        IF EXISTS(
+            SELECT nspname
+            FROM pg_catalog.pg_namespace
+            WHERE nspname = 'public'
+        )
+        THEN
+            ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT ALL ON TABLES TO neon_superuser WITH GRANT OPTION;
+            ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT ALL ON SEQUENCES TO neon_superuser WITH GRANT OPTION;
+        END IF;
+    END
+$$;

compute_tools/src/sql/set_public_schema_owner.sql

@@ -0,0 +1,23 @@
+DO
+$$
+    DECLARE
+        schema_owner TEXT;
+    BEGIN
+        IF EXISTS(
+            SELECT nspname
+            FROM pg_catalog.pg_namespace
+            WHERE nspname = 'public'
+        )
+        THEN
+            SELECT nspowner::regrole::text
+            FROM pg_catalog.pg_namespace
+            WHERE nspname = 'public'
+            INTO schema_owner;
+
+            IF schema_owner = 'cloud_admin' OR schema_owner = 'zenith_admin'
+            THEN
+                ALTER SCHEMA public OWNER TO {db_owner};
+            END IF;
+        END IF;
+    END
+$$;

compute_tools/src/sql/unset_template_for_drop_dbs.sql

@@ -0,0 +1,12 @@
+DO $$
+    BEGIN
+        IF EXISTS(
+            SELECT 1
+            FROM pg_catalog.pg_database
+            WHERE datname = {datname_str}
+        )
+        THEN
+            ALTER DATABASE {datname} is_template false;
+        END IF;
+    END
+$$;

deny.toml

@@ -37,6 +37,7 @@ allow = [
     "BSD-2-Clause",
     "BSD-3-Clause",
     "CC0-1.0",
+    "CDDL-1.0",
     "ISC",
     "MIT",
     "MPL-2.0",

libs/pageserver_api/src/config.rs

@@ -109,6 +109,8 @@ pub struct ConfigToml {
     pub virtual_file_io_mode: Option<crate::models::virtual_file::IoMode>,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub no_sync: Option<bool>,
+    #[serde(with = "humantime_serde")]
+    pub server_side_batch_timeout: Option<Duration>,
 }
 
 #[derive(Debug, Clone, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
@@ -317,6 +319,8 @@ pub mod defaults {
     pub const DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB: usize = 0;
 
     pub const DEFAULT_IO_BUFFER_ALIGNMENT: usize = 512;
+
+    pub const DEFAULT_SERVER_SIDE_BATCH_TIMEOUT: Option<&str> = None;
 }
 
 impl Default for ConfigToml {
@@ -397,6 +401,8 @@ impl Default for ConfigToml {
             ephemeral_bytes_per_memory_kb: (DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB),
             l0_flush: None,
             virtual_file_io_mode: None,
+            server_side_batch_timeout: DEFAULT_SERVER_SIDE_BATCH_TIMEOUT
+                .map(|duration| humantime::parse_duration(duration).unwrap()),
             tenant_config: TenantConfigToml::default(),
             no_sync: None,
         }

libs/pageserver_api/src/key.rs

@@ -24,7 +24,7 @@ pub struct Key {
 
 /// When working with large numbers of Keys in-memory, it is more efficient to handle them as i128 than as
 /// a struct of fields.
-#[derive(Clone, Copy, Hash, PartialEq, Eq, Ord, PartialOrd)]
+#[derive(Clone, Copy, Hash, PartialEq, Eq, Ord, PartialOrd, Serialize, Deserialize)]
 pub struct CompactKey(i128);
 
 /// The storage key size.

libs/pageserver_api/src/record.rs

@@ -41,6 +41,11 @@ pub enum NeonWalRecord {
         file_path: String,
         content: Option<Bytes>,
     },
+    // Truncate visibility map page
+    TruncateVisibilityMap {
+        trunc_byte: usize,
+        trunc_offs: usize,
+    },
 
     /// A testing record for unit testing purposes. It supports append data to an existing image, or clear it.
     #[cfg(feature = "testing")]

libs/pageserver_api/src/reltag.rs

@@ -24,7 +24,7 @@ use postgres_ffi::Oid;
 // FIXME: should move 'forknum' as last field to keep this consistent with Postgres.
 // Then we could replace the custom Ord and PartialOrd implementations below with
 // deriving them. This will require changes in walredoproc.c.
-#[derive(Debug, PartialEq, Eq, Hash, Clone, Copy, Serialize)]
+#[derive(Debug, PartialEq, Eq, Hash, Clone, Copy, Serialize, Deserialize)]
 pub struct RelTag {
     pub forknum: u8,
     pub spcnode: Oid,

libs/postgres_backend/src/lib.rs

@@ -716,6 +716,9 @@ impl<IO: AsyncRead + AsyncWrite + Unpin> PostgresBackend<IO> {
         Ok(())
     }
 
+    // Proto looks like this:
+    // FeMessage::Query("pagestream_v2{FeMessage::CopyData(PagesetreamFeMessage::GetPage(..))}")
+
     async fn process_message(
         &mut self,
         handler: &mut impl Handler<IO>,
@@ -831,7 +834,7 @@ impl<IO: AsyncRead + AsyncWrite + Unpin> PostgresBackend<IO> {
         use CopyStreamHandlerEnd::*;
 
         let expected_end = match &end {
-            ServerInitiated(_) | CopyDone | CopyFail | Terminate | EOF => true,
+            ServerInitiated(_) | CopyDone | CopyFail | Terminate | EOF | Cancelled => true,
             CopyStreamHandlerEnd::Disconnected(ConnectionError::Io(io_error))
                 if is_expected_io_error(io_error) =>
             {
@@ -871,6 +874,9 @@ impl<IO: AsyncRead + AsyncWrite + Unpin> PostgresBackend<IO> {
             // message from server' when it receives ErrorResponse (anything but
             // CopyData/CopyDone) back.
             CopyFail => Some((end.to_string(), SQLSTATE_SUCCESSFUL_COMPLETION)),
+
+            // When cancelled, send no response: we must not risk blocking on sending that response
+            Cancelled => None,
             _ => None,
         };
         if let Some((err, errcode)) = err_to_send_and_errcode {
@@ -1048,6 +1054,8 @@ pub enum CopyStreamHandlerEnd {
     /// The connection was lost
     #[error("connection error: {0}")]
     Disconnected(#[from] ConnectionError),
+    #[error("Shutdown")]
+    Cancelled,
     /// Some other error
     #[error(transparent)]
     Other(#[from] anyhow::Error),

libs/postgres_ffi/src/pg_constants.rs

@@ -243,8 +243,11 @@ const FSM_LEAF_NODES_PER_PAGE: usize = FSM_NODES_PER_PAGE - FSM_NON_LEAF_NODES_P
 pub const SLOTS_PER_FSM_PAGE: u32 = FSM_LEAF_NODES_PER_PAGE as u32;
 
 /* From visibilitymap.c */
-pub const VM_HEAPBLOCKS_PER_PAGE: u32 =
-    (BLCKSZ as usize - SIZEOF_PAGE_HEADER_DATA) as u32 * (8 / 2); // MAPSIZE * (BITS_PER_BYTE / BITS_PER_HEAPBLOCK)
+
+pub const VM_MAPSIZE: usize = BLCKSZ as usize - MAXALIGN_SIZE_OF_PAGE_HEADER_DATA;
+pub const VM_BITS_PER_HEAPBLOCK: usize = 2;
+pub const VM_HEAPBLOCKS_PER_BYTE: usize = 8 / VM_BITS_PER_HEAPBLOCK;
+pub const VM_HEAPBLOCKS_PER_PAGE: usize = VM_MAPSIZE * VM_HEAPBLOCKS_PER_BYTE;
 
 /* From origin.c */
 pub const REPLICATION_STATE_MAGIC: u32 = 0x1257DADE;

libs/postgres_ffi/src/walrecord.rs

@@ -16,7 +16,7 @@ use utils::bin_ser::DeserializeError;
 use utils::lsn::Lsn;
 
 #[repr(C)]
-#[derive(Debug)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct XlMultiXactCreate {
     pub mid: MultiXactId,
     /* new MultiXact's ID */
@@ -46,7 +46,7 @@ impl XlMultiXactCreate {
 }
 
 #[repr(C)]
-#[derive(Debug)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct XlMultiXactTruncate {
     pub oldest_multi_db: Oid,
     /* to-be-truncated range of multixact offsets */
@@ -72,7 +72,7 @@ impl XlMultiXactTruncate {
 }
 
 #[repr(C)]
-#[derive(Debug)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct XlRelmapUpdate {
     pub dbid: Oid,   /* database ID, or 0 for shared map */
     pub tsid: Oid,   /* database's tablespace, or pg_global */
@@ -90,7 +90,7 @@ impl XlRelmapUpdate {
 }
 
 #[repr(C)]
-#[derive(Debug)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct XlReploriginDrop {
     pub node_id: RepOriginId,
 }
@@ -104,7 +104,7 @@ impl XlReploriginDrop {
 }
 
 #[repr(C)]
-#[derive(Debug)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct XlReploriginSet {
     pub remote_lsn: Lsn,
     pub node_id: RepOriginId,
@@ -120,7 +120,7 @@ impl XlReploriginSet {
 }
 
 #[repr(C)]
-#[derive(Debug, Clone, Copy)]
+#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
 pub struct RelFileNode {
     pub spcnode: Oid, /* tablespace */
     pub dbnode: Oid,  /* database */
@@ -911,7 +911,7 @@ impl XlSmgrCreate {
 }
 
 #[repr(C)]
-#[derive(Debug)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct XlSmgrTruncate {
     pub blkno: BlockNumber,
     pub rnode: RelFileNode,
@@ -984,7 +984,7 @@ impl XlDropDatabase {
 /// xl_xact_parsed_abort structs in PostgreSQL, but we use the same
 /// struct for commits and aborts.
 ///
-#[derive(Debug)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct XlXactParsedRecord {
     pub xid: TransactionId,
     pub info: u8,

libs/pq_proto/src/lib.rs

@@ -184,6 +184,7 @@ pub struct CancelKeyData {
 
 impl fmt::Display for CancelKeyData {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        // TODO: this is producing strange results, with 0xffffffff........ always in the logs.
         let hi = (self.backend_pid as u64) << 32;
         let lo = self.cancel_key as u64;
         let id = hi | lo;

libs/remote_storage/src/azure_blob.rs

@@ -97,10 +97,7 @@ impl AzureBlobStorage {
 
     pub fn relative_path_to_name(&self, path: &RemotePath) -> String {
         assert_eq!(std::path::MAIN_SEPARATOR, REMOTE_STORAGE_PREFIX_SEPARATOR);
-        let path_string = path
-            .get_path()
-            .as_str()
-            .trim_end_matches(REMOTE_STORAGE_PREFIX_SEPARATOR);
+        let path_string = path.get_path().as_str();
         match &self.prefix_in_container {
             Some(prefix) => {
                 if prefix.ends_with(REMOTE_STORAGE_PREFIX_SEPARATOR) {
@@ -277,19 +274,14 @@ impl RemoteStorage for AzureBlobStorage {
         cancel: &CancellationToken,
     ) -> impl Stream<Item = Result<Listing, DownloadError>> {
         // get the passed prefix or if it is not set use prefix_in_bucket value
-        let list_prefix = prefix
-            .map(|p| self.relative_path_to_name(p))
-            .or_else(|| self.prefix_in_container.clone())
-            .map(|mut p| {
-                // required to end with a separator
-                // otherwise request will return only the entry of a prefix
-                if matches!(mode, ListingMode::WithDelimiter)
-                    && !p.ends_with(REMOTE_STORAGE_PREFIX_SEPARATOR)
-                {
-                    p.push(REMOTE_STORAGE_PREFIX_SEPARATOR);
+        let list_prefix = prefix.map(|p| self.relative_path_to_name(p)).or_else(|| {
+            self.prefix_in_container.clone().map(|mut s| {
+                if !s.ends_with(REMOTE_STORAGE_PREFIX_SEPARATOR) {
+                    s.push(REMOTE_STORAGE_PREFIX_SEPARATOR);
                 }
-                p
-            });
+                s
+            })
+        });
 
         async_stream::stream! {
             let _permit = self.permit(RequestKind::List, cancel).await?;

libs/remote_storage/src/config.rs

@@ -26,6 +26,16 @@ pub struct RemoteStorageConfig {
     pub timeout: Duration,
 }
 
+impl RemoteStorageKind {
+    pub fn bucket_name(&self) -> Option<&str> {
+        match self {
+            RemoteStorageKind::LocalFs { .. } => None,
+            RemoteStorageKind::AwsS3(config) => Some(&config.bucket_name),
+            RemoteStorageKind::AzureContainer(config) => Some(&config.container_name),
+        }
+    }
+}
+
 fn default_timeout() -> Duration {
     RemoteStorageConfig::DEFAULT_TIMEOUT
 }
@@ -178,6 +188,14 @@ impl RemoteStorageConfig {
     pub fn from_toml(toml: &toml_edit::Item) -> anyhow::Result<RemoteStorageConfig> {
         Ok(utils::toml_edit_ext::deserialize_item(toml)?)
     }
+
+    pub fn from_toml_str(input: &str) -> anyhow::Result<RemoteStorageConfig> {
+        let toml_document = toml_edit::DocumentMut::from_str(input)?;
+        if let Some(item) = toml_document.get("remote_storage") {
+            return Self::from_toml(item);
+        }
+        Self::from_toml(toml_document.as_item())
+    }
 }
 
 #[cfg(test)]
@@ -185,8 +203,7 @@ mod tests {
     use super::*;
 
     fn parse(input: &str) -> anyhow::Result<RemoteStorageConfig> {
-        let toml = input.parse::<toml_edit::DocumentMut>().unwrap();
-        RemoteStorageConfig::from_toml(toml.as_item())
+        RemoteStorageConfig::from_toml_str(input)
     }
 
     #[test]

libs/utils/scripts/restore_from_wal.sh

@@ -50,8 +50,8 @@ REDO_POS=0x$("$PG_BIN"/pg_controldata -D "$DATA_DIR" | grep -F "REDO location"|
 declare -i WAL_SIZE=$REDO_POS+114
 "$PG_BIN"/pg_ctl -D "$DATA_DIR" -l "$DATA_DIR/logfile.log" start
 "$PG_BIN"/pg_ctl -D "$DATA_DIR" -l "$DATA_DIR/logfile.log" stop -m immediate
-cp "$DATA_DIR"/pg_wal/000000010000000000000001 .
+cp "$DATA_DIR"/pg_wal/000000010000000000000001 "$DATA_DIR"
 cp "$WAL_PATH"/* "$DATA_DIR"/pg_wal/
 for partial in "$DATA_DIR"/pg_wal/*.partial ; do mv "$partial" "${partial%.partial}" ; done
-dd if=000000010000000000000001 of="$DATA_DIR"/pg_wal/000000010000000000000001 bs=$WAL_SIZE count=1 conv=notrunc
-rm -f 000000010000000000000001
+dd if="$DATA_DIR"/000000010000000000000001 of="$DATA_DIR"/pg_wal/000000010000000000000001 bs=$WAL_SIZE count=1 conv=notrunc
+rm -f "$DATA_DIR"/000000010000000000000001

libs/utils/scripts/restore_from_wal_initdb.sh

@@ -14,8 +14,8 @@ REDO_POS=0x$("$PG_BIN"/pg_controldata -D "$DATA_DIR" | grep -F "REDO location"|
 declare -i WAL_SIZE=$REDO_POS+114
 "$PG_BIN"/pg_ctl -D "$DATA_DIR" -l "$DATA_DIR/logfile.log" start
 "$PG_BIN"/pg_ctl -D "$DATA_DIR" -l "$DATA_DIR/logfile.log" stop -m immediate
-cp "$DATA_DIR"/pg_wal/000000010000000000000001 .
+cp "$DATA_DIR"/pg_wal/000000010000000000000001 "$DATA_DIR"
 cp "$WAL_PATH"/* "$DATA_DIR"/pg_wal/
 for partial in "$DATA_DIR"/pg_wal/*.partial ; do mv "$partial" "${partial%.partial}" ; done
-dd if=000000010000000000000001 of="$DATA_DIR"/pg_wal/000000010000000000000001 bs=$WAL_SIZE count=1 conv=notrunc
-rm -f 000000010000000000000001
+dd if="$DATA_DIR"/000000010000000000000001 of="$DATA_DIR"/pg_wal/000000010000000000000001 bs=$WAL_SIZE count=1 conv=notrunc
+rm -f "$DATA_DIR"/000000010000000000000001

libs/utils/src/http/json.rs

@@ -5,6 +5,7 @@ use serde::{Deserialize, Serialize};
 
 use super::error::ApiError;
 
+/// Parse a json request body and deserialize it to the type `T`.
 pub async fn json_request<T: for<'de> Deserialize<'de>>(
     request: &mut Request<Body>,
 ) -> Result<T, ApiError> {
@@ -27,6 +28,27 @@ pub async fn json_request<T: for<'de> Deserialize<'de>>(
         .map_err(ApiError::BadRequest)
 }
 
+/// Parse a json request body and deserialize it to the type `T`. If the body is empty, return `T::default`.
+pub async fn json_request_maybe<T: for<'de> Deserialize<'de> + Default>(
+    request: &mut Request<Body>,
+) -> Result<T, ApiError> {
+    let body = hyper::body::aggregate(request.body_mut())
+        .await
+        .context("Failed to read request body")
+        .map_err(ApiError::BadRequest)?;
+
+    if body.remaining() == 0 {
+        return Ok(T::default());
+    }
+
+    let mut deser = serde_json::de::Deserializer::from_reader(body.reader());
+
+    serde_path_to_error::deserialize(&mut deser)
+        // intentionally stringify because the debug version is not helpful in python logs
+        .map_err(|e| anyhow::anyhow!("Failed to parse json request: {e}"))
+        .map_err(ApiError::BadRequest)
+}
+
 pub fn json_response<T: Serialize>(
     status: StatusCode,
     data: T,

libs/wal_decoder/src/decoder.rs

@@ -19,7 +19,7 @@ impl InterpretedWalRecord {
     pub fn from_bytes_filtered(
         buf: Bytes,
         shard: &ShardIdentity,
-        record_end_lsn: Lsn,
+        next_record_lsn: Lsn,
         pg_version: u32,
     ) -> anyhow::Result<InterpretedWalRecord> {
         let mut decoded = DecodedWALRecord::default();
@@ -32,18 +32,18 @@ impl InterpretedWalRecord {
             FlushUncommittedRecords::No
         };
 
-        let metadata_record = MetadataRecord::from_decoded(&decoded, record_end_lsn, pg_version)?;
+        let metadata_record = MetadataRecord::from_decoded(&decoded, next_record_lsn, pg_version)?;
         let batch = SerializedValueBatch::from_decoded_filtered(
             decoded,
             shard,
-            record_end_lsn,
+            next_record_lsn,
             pg_version,
         )?;
 
         Ok(InterpretedWalRecord {
             metadata_record,
             batch,
-            end_lsn: record_end_lsn,
+            next_record_lsn,
             flush_uncommitted,
             xid,
         })
@@ -53,7 +53,7 @@ impl InterpretedWalRecord {
 impl MetadataRecord {
     fn from_decoded(
         decoded: &DecodedWALRecord,
-        record_end_lsn: Lsn,
+        next_record_lsn: Lsn,
         pg_version: u32,
     ) -> anyhow::Result<Option<MetadataRecord>> {
         // Note: this doesn't actually copy the bytes since
@@ -74,7 +74,9 @@ impl MetadataRecord {
                 Ok(None)
             }
             pg_constants::RM_CLOG_ID => Self::decode_clog_record(&mut buf, decoded, pg_version),
-            pg_constants::RM_XACT_ID => Self::decode_xact_record(&mut buf, decoded, record_end_lsn),
+            pg_constants::RM_XACT_ID => {
+                Self::decode_xact_record(&mut buf, decoded, next_record_lsn)
+            }
             pg_constants::RM_MULTIXACT_ID => {
                 Self::decode_multixact_record(&mut buf, decoded, pg_version)
             }
@@ -86,7 +88,9 @@ impl MetadataRecord {
             //
             // Alternatively, one can make the checkpoint part of the subscription protocol
             // to the pageserver. This should work fine, but can be done at a later point.
-            pg_constants::RM_XLOG_ID => Self::decode_xlog_record(&mut buf, decoded, record_end_lsn),
+            pg_constants::RM_XLOG_ID => {
+                Self::decode_xlog_record(&mut buf, decoded, next_record_lsn)
+            }
             pg_constants::RM_LOGICALMSG_ID => {
                 Self::decode_logical_message_record(&mut buf, decoded)
             }

libs/wal_decoder/src/models.rs

@@ -32,16 +32,19 @@ use postgres_ffi::walrecord::{
     XlSmgrTruncate, XlXactParsedRecord,
 };
 use postgres_ffi::{Oid, TransactionId};
+use serde::{Deserialize, Serialize};
 use utils::lsn::Lsn;
 
 use crate::serialized_batch::SerializedValueBatch;
 
+#[derive(Serialize, Deserialize)]
 pub enum FlushUncommittedRecords {
     Yes,
     No,
 }
 
 /// An interpreted Postgres WAL record, ready to be handled by the pageserver
+#[derive(Serialize, Deserialize)]
 pub struct InterpretedWalRecord {
     /// Optional metadata record - may cause writes to metadata keys
     /// in the storage engine
@@ -49,8 +52,10 @@ pub struct InterpretedWalRecord {
     /// A pre-serialized batch along with the required metadata for ingestion
     /// by the pageserver
     pub batch: SerializedValueBatch,
-    /// Byte offset within WAL for the end of the original PG WAL record
-    pub end_lsn: Lsn,
+    /// Byte offset within WAL for the start of the next PG WAL record.
+    /// Usually this is the end LSN of the current record, but in case of
+    /// XLOG SWITCH records it will be within the next segment.
+    pub next_record_lsn: Lsn,
     /// Whether to flush all uncommitted modifications to the storage engine
     /// before ingesting this record. This is currently only used for legacy PG
     /// database creations which read pages from a template database. Such WAL
@@ -62,6 +67,7 @@ pub struct InterpretedWalRecord {
 
 /// The interpreted part of the Postgres WAL record which requires metadata
 /// writes to the underlying storage engine.
+#[derive(Serialize, Deserialize)]
 pub enum MetadataRecord {
     Heapam(HeapamRecord),
     Neonrmgr(NeonrmgrRecord),
@@ -77,10 +83,12 @@ pub enum MetadataRecord {
     Replorigin(ReploriginRecord),
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum HeapamRecord {
     ClearVmBits(ClearVmBits),
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct ClearVmBits {
     pub new_heap_blkno: Option<u32>,
     pub old_heap_blkno: Option<u32>,
@@ -88,24 +96,29 @@ pub struct ClearVmBits {
     pub flags: u8,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum NeonrmgrRecord {
     ClearVmBits(ClearVmBits),
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum SmgrRecord {
     Create(SmgrCreate),
     Truncate(XlSmgrTruncate),
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct SmgrCreate {
     pub rel: RelTag,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum DbaseRecord {
     Create(DbaseCreate),
     Drop(DbaseDrop),
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct DbaseCreate {
     pub db_id: Oid,
     pub tablespace_id: Oid,
@@ -113,27 +126,32 @@ pub struct DbaseCreate {
     pub src_tablespace_id: Oid,
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct DbaseDrop {
     pub db_id: Oid,
     pub tablespace_ids: Vec<Oid>,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum ClogRecord {
     ZeroPage(ClogZeroPage),
     Truncate(ClogTruncate),
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct ClogZeroPage {
     pub segno: u32,
     pub rpageno: u32,
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct ClogTruncate {
     pub pageno: u32,
     pub oldest_xid: TransactionId,
     pub oldest_xid_db: Oid,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum XactRecord {
     Commit(XactCommon),
     Abort(XactCommon),
@@ -142,6 +160,7 @@ pub enum XactRecord {
     Prepare(XactPrepare),
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct XactCommon {
     pub parsed: XlXactParsedRecord,
     pub origin_id: u16,
@@ -150,61 +169,73 @@ pub struct XactCommon {
     pub lsn: Lsn,
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct XactPrepare {
     pub xl_xid: TransactionId,
     pub data: Bytes,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum MultiXactRecord {
     ZeroPage(MultiXactZeroPage),
     Create(XlMultiXactCreate),
     Truncate(XlMultiXactTruncate),
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct MultiXactZeroPage {
     pub slru_kind: SlruKind,
     pub segno: u32,
     pub rpageno: u32,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum RelmapRecord {
     Update(RelmapUpdate),
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct RelmapUpdate {
     pub update: XlRelmapUpdate,
     pub buf: Bytes,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum XlogRecord {
     Raw(RawXlogRecord),
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct RawXlogRecord {
     pub info: u8,
     pub lsn: Lsn,
     pub buf: Bytes,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum LogicalMessageRecord {
     Put(PutLogicalMessage),
     #[cfg(feature = "testing")]
     Failpoint,
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct PutLogicalMessage {
     pub path: String,
     pub buf: Bytes,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum StandbyRecord {
     RunningXacts(StandbyRunningXacts),
 }
 
+#[derive(Serialize, Deserialize)]
 pub struct StandbyRunningXacts {
     pub oldest_running_xid: TransactionId,
 }
 
+#[derive(Serialize, Deserialize)]
 pub enum ReploriginRecord {
     Set(XlReploriginSet),
     Drop(XlReploriginDrop),

libs/wal_decoder/src/serialized_batch.rs

@@ -16,6 +16,7 @@ use pageserver_api::shard::ShardIdentity;
 use pageserver_api::{key::CompactKey, value::Value};
 use postgres_ffi::walrecord::{DecodedBkpBlock, DecodedWALRecord};
 use postgres_ffi::{page_is_new, page_set_lsn, pg_constants, BLCKSZ};
+use serde::{Deserialize, Serialize};
 use utils::bin_ser::BeSer;
 use utils::lsn::Lsn;
 
@@ -29,6 +30,7 @@ static ZERO_PAGE: Bytes = Bytes::from_static(&[0u8; BLCKSZ as usize]);
 /// relation sizes. In the case of "observed" values, we only need to know
 /// the key and LSN, so two types of metadata are supported to save on network
 /// bandwidth.
+#[derive(Serialize, Deserialize)]
 pub enum ValueMeta {
     Serialized(SerializedValueMeta),
     Observed(ObservedValueMeta),
@@ -75,6 +77,7 @@ impl PartialEq for OrderedValueMeta {
 impl Eq for OrderedValueMeta {}
 
 /// Metadata for a [`Value`] serialized into the batch.
+#[derive(Serialize, Deserialize)]
 pub struct SerializedValueMeta {
     pub key: CompactKey,
     pub lsn: Lsn,
@@ -86,12 +89,14 @@ pub struct SerializedValueMeta {
 }
 
 /// Metadata for a [`Value`] observed by the batch
+#[derive(Serialize, Deserialize)]
 pub struct ObservedValueMeta {
     pub key: CompactKey,
     pub lsn: Lsn,
 }
 
 /// Batch of serialized [`Value`]s.
+#[derive(Serialize, Deserialize)]
 pub struct SerializedValueBatch {
     /// [`Value`]s serialized in EphemeralFile's native format,
     /// ready for disk write by the pageserver
@@ -132,7 +137,7 @@ impl SerializedValueBatch {
     pub(crate) fn from_decoded_filtered(
         decoded: DecodedWALRecord,
         shard: &ShardIdentity,
-        record_end_lsn: Lsn,
+        next_record_lsn: Lsn,
         pg_version: u32,
     ) -> anyhow::Result<SerializedValueBatch> {
         // First determine how big the buffer needs to be and allocate it up-front.
@@ -156,13 +161,17 @@ impl SerializedValueBatch {
             let key = rel_block_to_key(rel, blk.blkno);
 
             if !key.is_valid_key_on_write_path() {
-                anyhow::bail!("Unsupported key decoded at LSN {}: {}", record_end_lsn, key);
+                anyhow::bail!(
+                    "Unsupported key decoded at LSN {}: {}",
+                    next_record_lsn,
+                    key
+                );
             }
 
             let key_is_local = shard.is_key_local(&key);
 
             tracing::debug!(
-                lsn=%record_end_lsn,
+                lsn=%next_record_lsn,
                 key=%key,
                 "ingest: shard decision {}",
                 if !key_is_local { "drop" } else { "keep" },
@@ -174,7 +183,7 @@ impl SerializedValueBatch {
                     // its blkno in case it implicitly extends a relation.
                     metadata.push(ValueMeta::Observed(ObservedValueMeta {
                         key: key.to_compact(),
-                        lsn: record_end_lsn,
+                        lsn: next_record_lsn,
                     }))
                 }
 
@@ -205,7 +214,7 @@ impl SerializedValueBatch {
                 // that would corrupt the page.
                 //
                 if !page_is_new(&image) {
-                    page_set_lsn(&mut image, record_end_lsn)
+                    page_set_lsn(&mut image, next_record_lsn)
                 }
                 assert_eq!(image.len(), BLCKSZ as usize);
 
@@ -224,12 +233,12 @@ impl SerializedValueBatch {
 
             metadata.push(ValueMeta::Serialized(SerializedValueMeta {
                 key: key.to_compact(),
-                lsn: record_end_lsn,
+                lsn: next_record_lsn,
                 batch_offset: relative_off,
                 len: val_ser_size,
                 will_init: val.will_init(),
             }));
-            max_lsn = std::cmp::max(max_lsn, record_end_lsn);
+            max_lsn = std::cmp::max(max_lsn, next_record_lsn);
             len += 1;
         }
 

pageserver/Cargo.toml

@@ -84,6 +84,7 @@ enumset = { workspace = true, features = ["serde"]}
 strum.workspace = true
 strum_macros.workspace = true
 wal_decoder.workspace = true
+smallvec.workspace = true
 
 [target.'cfg(target_os = "linux")'.dependencies]
 procfs.workspace = true

pageserver/benches/bench_ingest.rs

@@ -167,6 +167,7 @@ fn criterion_benchmark(c: &mut Criterion) {
         16384,
         virtual_file::io_engine_for_bench(),
         conf.virtual_file_io_mode,
+        virtual_file::SyncMode::Sync,
     );
     page_cache::init(conf.page_cache_size);
 

pageserver/ctl/Cargo.toml

@@ -18,7 +18,6 @@ postgres_ffi.workspace = true
 thiserror.workspace = true
 tokio.workspace = true
 tokio-util.workspace = true
-toml_edit.workspace = true
 utils.workspace = true
 svg_fmt.workspace = true
 workspace_hack.workspace = true

pageserver/ctl/src/layer_map_analyzer.rs

@@ -138,6 +138,7 @@ pub(crate) async fn main(cmd: &AnalyzeLayerMapCmd) -> Result<()> {
         10,
         virtual_file::api::IoEngineKind::StdFs,
         IoMode::preferred(),
+        virtual_file::SyncMode::Sync,
     );
     pageserver::page_cache::init(100);
 

pageserver/ctl/src/layers.rs

@@ -51,6 +51,7 @@ async fn read_delta_file(path: impl AsRef<Path>, ctx: &RequestContext) -> Result
         10,
         virtual_file::api::IoEngineKind::StdFs,
         IoMode::preferred(),
+        virtual_file::SyncMode::Sync,
     );
     page_cache::init(100);
     let path = Utf8Path::from_path(path.as_ref()).expect("non-Unicode path");
@@ -65,6 +66,7 @@ async fn read_image_file(path: impl AsRef<Path>, ctx: &RequestContext) -> Result
         10,
         virtual_file::api::IoEngineKind::StdFs,
         IoMode::preferred(),
+        virtual_file::SyncMode::Sync,
     );
     page_cache::init(100);
     let path = Utf8Path::from_path(path.as_ref()).expect("non-Unicode path");
@@ -171,6 +173,7 @@ pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
                 10,
                 virtual_file::api::IoEngineKind::StdFs,
                 IoMode::preferred(),
+                virtual_file::SyncMode::Sync,
             );
             pageserver::page_cache::init(100);
 

pageserver/ctl/src/main.rs

@@ -174,11 +174,7 @@ async fn main() -> anyhow::Result<()> {
                 println!("specified prefix '{}' failed validation", cmd.prefix);
                 return Ok(());
             };
-            let toml_document = toml_edit::DocumentMut::from_str(&cmd.config_toml_str)?;
-            let toml_item = toml_document
-                .get("remote_storage")
-                .expect("need remote_storage");
-            let config = RemoteStorageConfig::from_toml(toml_item)?;
+            let config = RemoteStorageConfig::from_toml_str(&cmd.config_toml_str)?;
             let storage = remote_storage::GenericRemoteStorage::from_config(&config).await;
             let cancel = CancellationToken::new();
             storage
@@ -209,6 +205,7 @@ async fn print_layerfile(path: &Utf8Path) -> anyhow::Result<()> {
         10,
         virtual_file::api::IoEngineKind::StdFs,
         IoMode::preferred(),
+        virtual_file::SyncMode::Sync,
     );
     page_cache::init(100);
     let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);

pageserver/src/bin/pageserver.rs

@@ -171,11 +171,18 @@ fn main() -> anyhow::Result<()> {
     let scenario = failpoint_support::init();
 
     // Basic initialization of things that don't change after startup
+    tracing::info!("Initializing virtual_file...");
     virtual_file::init(
         conf.max_file_descriptors,
         conf.virtual_file_io_engine,
         conf.virtual_file_io_mode,
+        if conf.no_sync {
+            virtual_file::SyncMode::UnsafeNoSync
+        } else {
+            virtual_file::SyncMode::Sync
+        },
     );
+    tracing::info!("Initializing page_cache...");
     page_cache::init(conf.page_cache_size);
 
     start_pageserver(launch_ts, conf).context("Failed to start pageserver")?;

pageserver/src/config.rs

@@ -182,6 +182,10 @@ pub struct PageServerConf {
 
     /// Optionally disable disk syncs (unsafe!)
     pub no_sync: bool,
+
+    /// Maximum amount of time for which a get page request request
+    /// might be held up for request merging.
+    pub server_side_batch_timeout: Option<Duration>,
 }
 
 /// Token for authentication to safekeepers
@@ -336,6 +340,7 @@ impl PageServerConf {
             concurrent_tenant_warmup,
             concurrent_tenant_size_logical_size_queries,
             virtual_file_io_engine,
+            server_side_batch_timeout,
             tenant_config,
             no_sync,
         } = config_toml;
@@ -377,6 +382,7 @@ impl PageServerConf {
             image_compression,
             timeline_offloading,
             ephemeral_bytes_per_memory_kb,
+            server_side_batch_timeout,
 
             // ------------------------------------------------------------
             // fields that require additional validation or custom handling

pageserver/src/http/routes.rs

@@ -83,6 +83,8 @@ use crate::tenant::storage_layer::LayerName;
 use crate::tenant::timeline::offload::offload_timeline;
 use crate::tenant::timeline::offload::OffloadError;
 use crate::tenant::timeline::CompactFlags;
+use crate::tenant::timeline::CompactOptions;
+use crate::tenant::timeline::CompactRange;
 use crate::tenant::timeline::CompactionError;
 use crate::tenant::timeline::Timeline;
 use crate::tenant::GetTimelineError;
@@ -100,7 +102,7 @@ use utils::{
     http::{
         endpoint::{self, attach_openapi_ui, auth_middleware, check_permission_with},
         error::{ApiError, HttpErrorBody},
-        json::{json_request, json_response},
+        json::{json_request, json_request_maybe, json_response},
         request::parse_request_param,
         RequestExt, RouterBuilder,
     },
@@ -324,6 +326,7 @@ impl From<crate::tenant::DeleteTimelineError> for ApiError {
                     .into_boxed_str(),
             ),
             a @ AlreadyInProgress(_) => ApiError::Conflict(a.to_string()),
+            Cancelled => ApiError::ResourceUnavailable("shutting down".into()),
             Other(e) => ApiError::InternalServerError(e),
         }
     }
@@ -1926,13 +1929,15 @@ async fn timeline_gc_handler(
 
 // Run compaction immediately on given timeline.
 async fn timeline_compact_handler(
-    request: Request<Body>,
+    mut request: Request<Body>,
     cancel: CancellationToken,
 ) -> Result<Response<Body>, ApiError> {
     let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
     let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
     check_permission(&request, Some(tenant_shard_id.tenant_id))?;
 
+    let compact_range = json_request_maybe::<Option<CompactRange>>(&mut request).await?;
+
     let state = get_state(&request);
 
     let mut flags = EnumSet::empty();
@@ -1956,11 +1961,16 @@ async fn timeline_compact_handler(
     let wait_until_uploaded =
         parse_query_param::<_, bool>(&request, "wait_until_uploaded")?.unwrap_or(false);
 
+    let options = CompactOptions {
+        compact_range,
+        flags,
+    };
+
     async {
         let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
         let timeline = active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id).await?;
         timeline
-            .compact(&cancel, flags, &ctx)
+            .compact_with_options(&cancel, options, &ctx)
             .await
             .map_err(|e| ApiError::InternalServerError(e.into()))?;
         if wait_until_uploaded {

pageserver/src/metrics.rs

@@ -1187,6 +1187,7 @@ struct GlobalAndPerTimelineHistogramTimer<'a, 'c> {
     ctx: &'c RequestContext,
     start: std::time::Instant,
     op: SmgrQueryType,
+    count: usize,
 }
 
 impl Drop for GlobalAndPerTimelineHistogramTimer<'_, '_> {
@@ -1214,10 +1215,13 @@ impl Drop for GlobalAndPerTimelineHistogramTimer<'_, '_> {
                 elapsed
             }
         };
-        self.global_latency_histo
-            .observe(ex_throttled.as_secs_f64());
-        if let Some(per_timeline_getpage_histo) = self.per_timeline_latency_histo {
-            per_timeline_getpage_histo.observe(ex_throttled.as_secs_f64());
+
+        for _ in 0..self.count {
+            self.global_latency_histo
+                .observe(ex_throttled.as_secs_f64());
+            if let Some(per_timeline_getpage_histo) = self.per_timeline_latency_histo {
+                per_timeline_getpage_histo.observe(ex_throttled.as_secs_f64());
+            }
         }
     }
 }
@@ -1385,6 +1389,14 @@ impl SmgrQueryTimePerTimeline {
         &'a self,
         op: SmgrQueryType,
         ctx: &'c RequestContext,
+    ) -> Option<impl Drop + 'a> {
+        self.start_timer_many(op, 1, ctx)
+    }
+    pub(crate) fn start_timer_many<'c: 'a, 'a>(
+        &'a self,
+        op: SmgrQueryType,
+        count: usize,
+        ctx: &'c RequestContext,
     ) -> Option<impl Drop + 'a> {
         let start = Instant::now();
 
@@ -1422,6 +1434,7 @@ impl SmgrQueryTimePerTimeline {
             ctx,
             start,
             op,
+            count,
         })
     }
 }

pageserver/src/page_service.rs

@@ -7,13 +7,13 @@ use bytes::Buf;
 use futures::FutureExt;
 use itertools::Itertools;
 use once_cell::sync::OnceCell;
-use pageserver_api::models::TenantState;
+use pageserver_api::models::{self, TenantState};
 use pageserver_api::models::{
     PagestreamBeMessage, PagestreamDbSizeRequest, PagestreamDbSizeResponse,
     PagestreamErrorResponse, PagestreamExistsRequest, PagestreamExistsResponse,
-    PagestreamFeMessage, PagestreamGetPageRequest, PagestreamGetPageResponse,
-    PagestreamGetSlruSegmentRequest, PagestreamGetSlruSegmentResponse, PagestreamNblocksRequest,
-    PagestreamNblocksResponse, PagestreamProtocolVersion,
+    PagestreamFeMessage, PagestreamGetPageRequest, PagestreamGetSlruSegmentRequest,
+    PagestreamGetSlruSegmentResponse, PagestreamNblocksRequest, PagestreamNblocksResponse,
+    PagestreamProtocolVersion,
 };
 use pageserver_api::shard::TenantShardId;
 use postgres_backend::{is_expected_io_error, AuthType, PostgresBackend, QueryError};
@@ -44,7 +44,7 @@ use crate::basebackup;
 use crate::basebackup::BasebackupError;
 use crate::config::PageServerConf;
 use crate::context::{DownloadBehavior, RequestContext};
-use crate::metrics;
+use crate::metrics::{self};
 use crate::metrics::{ComputeCommandKind, COMPUTE_COMMANDS_COUNTERS, LIVE_CONNECTIONS};
 use crate::pgdatadir_mapping::Version;
 use crate::span::debug_assert_current_span_has_tenant_and_timeline_id;
@@ -59,7 +59,7 @@ use crate::tenant::GetTimelineError;
 use crate::tenant::PageReconstructError;
 use crate::tenant::Timeline;
 use pageserver_api::key::rel_block_to_key;
-use pageserver_api::reltag::SlruKind;
+use pageserver_api::reltag::{BlockNumber, RelTag, SlruKind};
 use postgres_ffi::pg_constants::DEFAULTTABLESPACE_OID;
 use postgres_ffi::BLCKSZ;
 
@@ -105,6 +105,7 @@ pub fn spawn(
             pg_auth,
             tcp_listener,
             conf.pg_auth_type,
+            conf.server_side_batch_timeout,
             libpq_ctx,
             cancel.clone(),
         )
@@ -153,6 +154,7 @@ pub async fn libpq_listener_main(
     auth: Option<Arc<SwappableJwtAuth>>,
     listener: tokio::net::TcpListener,
     auth_type: AuthType,
+    server_side_batch_timeout: Option<Duration>,
     listener_ctx: RequestContext,
     listener_cancel: CancellationToken,
 ) -> Connections {
@@ -183,6 +185,7 @@ pub async fn libpq_listener_main(
                     local_auth,
                     socket,
                     auth_type,
+                    server_side_batch_timeout,
                     connection_ctx,
                     connections_cancel.child_token(),
                 ));
@@ -210,6 +213,7 @@ async fn page_service_conn_main(
     auth: Option<Arc<SwappableJwtAuth>>,
     socket: tokio::net::TcpStream,
     auth_type: AuthType,
+    server_side_batch_timeout: Option<Duration>,
     connection_ctx: RequestContext,
     cancel: CancellationToken,
 ) -> ConnectionHandlerResult {
@@ -260,8 +264,13 @@ async fn page_service_conn_main(
     // and create a child per-query context when it invokes process_query.
     // But it's in a shared crate, so, we store connection_ctx inside PageServerHandler
     // and create the per-query context in process_query ourselves.
-    let mut conn_handler =
-        PageServerHandler::new(tenant_manager, auth, connection_ctx, cancel.clone());
+    let mut conn_handler = PageServerHandler::new(
+        tenant_manager,
+        auth,
+        server_side_batch_timeout,
+        connection_ctx,
+        cancel.clone(),
+    );
     let pgbackend = PostgresBackend::new_from_io(socket, peer_addr, auth_type, None)?;
 
     match pgbackend.run(&mut conn_handler, &cancel).await {
@@ -304,6 +313,12 @@ struct PageServerHandler {
     cancel: CancellationToken,
 
     timeline_handles: TimelineHandles,
+
+    /// Messages queued up for the next processing batch
+    next_batch: Option<BatchedFeMessage>,
+
+    /// See [`PageServerConf::server_side_batch_timeout`]
+    server_side_batch_timeout: Option<Duration>,
 }
 
 struct TimelineHandles {
@@ -517,10 +532,47 @@ impl From<WaitLsnError> for QueryError {
     }
 }
 
+enum BatchedFeMessage {
+    Exists {
+        span: Span,
+        req: models::PagestreamExistsRequest,
+    },
+    Nblocks {
+        span: Span,
+        req: models::PagestreamNblocksRequest,
+    },
+    GetPage {
+        span: Span,
+        shard: timeline::handle::Handle<TenantManagerTypes>,
+        effective_request_lsn: Lsn,
+        pages: smallvec::SmallVec<[(RelTag, BlockNumber); 1]>,
+    },
+    DbSize {
+        span: Span,
+        req: models::PagestreamDbSizeRequest,
+    },
+    GetSlruSegment {
+        span: Span,
+        req: models::PagestreamGetSlruSegmentRequest,
+    },
+    RespondError {
+        span: Span,
+        error: PageStreamError,
+    },
+}
+
+enum BatchOrEof {
+    /// In the common case, this has one entry.
+    /// At most, it has two entries: the first is the leftover batch, the second is an error.
+    Batch(smallvec::SmallVec<[BatchedFeMessage; 1]>),
+    Eof,
+}
+
 impl PageServerHandler {
     pub fn new(
         tenant_manager: Arc<TenantManager>,
         auth: Option<Arc<SwappableJwtAuth>>,
+        server_side_batch_timeout: Option<Duration>,
         connection_ctx: RequestContext,
         cancel: CancellationToken,
     ) -> Self {
@@ -530,6 +582,8 @@ impl PageServerHandler {
             connection_ctx,
             timeline_handles: TimelineHandles::new(tenant_manager),
             cancel,
+            next_batch: None,
+            server_side_batch_timeout,
         }
     }
 
@@ -557,6 +611,221 @@ impl PageServerHandler {
         )
     }
 
+    async fn read_batch_from_connection<IO>(
+        &mut self,
+        pgb: &mut PostgresBackend<IO>,
+        tenant_id: &TenantId,
+        timeline_id: &TimelineId,
+        ctx: &RequestContext,
+    ) -> Result<Option<BatchOrEof>, QueryError>
+    where
+        IO: AsyncRead + AsyncWrite + Send + Sync + Unpin,
+    {
+        let mut batch = self.next_batch.take();
+        let mut batch_started_at: Option<std::time::Instant> = None;
+
+        let next_batch: Option<BatchedFeMessage> = loop {
+            let sleep_fut = match (self.server_side_batch_timeout, batch_started_at) {
+                (Some(batch_timeout), Some(started_at)) => futures::future::Either::Left(
+                    tokio::time::sleep_until((started_at + batch_timeout).into()),
+                ),
+                _ => futures::future::Either::Right(futures::future::pending()),
+            };
+
+            let msg = tokio::select! {
+                biased;
+                _ = self.cancel.cancelled() => {
+                    return Err(QueryError::Shutdown)
+                }
+                msg = pgb.read_message() => {
+                    msg
+                }
+                _ = sleep_fut => {
+                    assert!(batch.is_some());
+                    break None;
+                }
+            };
+            let copy_data_bytes = match msg? {
+                Some(FeMessage::CopyData(bytes)) => bytes,
+                Some(FeMessage::Terminate) => {
+                    return Ok(Some(BatchOrEof::Eof));
+                }
+                Some(m) => {
+                    return Err(QueryError::Other(anyhow::anyhow!(
+                        "unexpected message: {m:?} during COPY"
+                    )));
+                }
+                None => {
+                    return Ok(Some(BatchOrEof::Eof));
+                } // client disconnected
+            };
+            trace!("query: {copy_data_bytes:?}");
+            fail::fail_point!("ps::handle-pagerequest-message");
+
+            // parse request
+            let neon_fe_msg = PagestreamFeMessage::parse(&mut copy_data_bytes.reader())?;
+
+            let this_msg = match neon_fe_msg {
+                PagestreamFeMessage::Exists(req) => BatchedFeMessage::Exists {
+                    span: tracing::info_span!("handle_get_rel_exists_request", rel = %req.rel, req_lsn = %req.request_lsn),
+                    req,
+                },
+                PagestreamFeMessage::Nblocks(req) => BatchedFeMessage::Nblocks {
+                    span: tracing::info_span!("handle_get_nblocks_request", rel = %req.rel, req_lsn = %req.request_lsn),
+                    req,
+                },
+                PagestreamFeMessage::DbSize(req) => BatchedFeMessage::DbSize {
+                    span: tracing::info_span!("handle_db_size_request", dbnode = %req.dbnode, req_lsn = %req.request_lsn),
+                    req,
+                },
+                PagestreamFeMessage::GetSlruSegment(req) => BatchedFeMessage::GetSlruSegment {
+                    span: tracing::info_span!("handle_get_slru_segment_request", kind = %req.kind, segno = %req.segno, req_lsn = %req.request_lsn),
+                    req,
+                },
+                PagestreamFeMessage::GetPage(PagestreamGetPageRequest {
+                    request_lsn,
+                    not_modified_since,
+                    rel,
+                    blkno,
+                }) => {
+                    // shard_id is filled in by the handler
+                    let span = tracing::info_span!(
+                        "handle_get_page_at_lsn_request_batched",
+                        %tenant_id, %timeline_id, shard_id = tracing::field::Empty, req_lsn = %request_lsn,
+                        batch_size = tracing::field::Empty, batch_id = tracing::field::Empty
+                    );
+
+                    macro_rules! current_batch_and_error {
+                        ($error:expr) => {{
+                            let error = BatchedFeMessage::RespondError {
+                                span,
+                                error: $error,
+                            };
+                            let batch_and_error = match batch {
+                                Some(b) => smallvec::smallvec![b, error],
+                                None => smallvec::smallvec![error],
+                            };
+                            Ok(Some(BatchOrEof::Batch(batch_and_error)))
+                        }};
+                    }
+
+                    let key = rel_block_to_key(rel, blkno);
+                    let shard = match self
+                        .timeline_handles
+                        .get(*tenant_id, *timeline_id, ShardSelector::Page(key))
+                        .instrument(span.clone())
+                        .await
+                    {
+                        Ok(tl) => tl,
+                        Err(GetActiveTimelineError::Tenant(GetActiveTenantError::NotFound(_))) => {
+                            // We already know this tenant exists in general, because we resolved it at
+                            // start of connection.  Getting a NotFound here indicates that the shard containing
+                            // the requested page is not present on this node: the client's knowledge of shard->pageserver
+                            // mapping is out of date.
+                            //
+                            // Closing the connection by returning ``::Reconnect` has the side effect of rate-limiting above message, via
+                            // client's reconnect backoff, as well as hopefully prompting the client to load its updated configuration
+                            // and talk to a different pageserver.
+                            return current_batch_and_error!(PageStreamError::Reconnect(
+                                "getpage@lsn request routed to wrong shard".into()
+                            ));
+                        }
+                        Err(e) => {
+                            return current_batch_and_error!(e.into());
+                        }
+                    };
+                    let effective_request_lsn = match Self::wait_or_get_last_lsn(
+                        &shard,
+                        request_lsn,
+                        not_modified_since,
+                        &shard.get_latest_gc_cutoff_lsn(),
+                        ctx,
+                    )
+                    // TODO: if we actually need to wait for lsn here, it delays the entire batch which doesn't need to wait
+                    .await
+                    {
+                        Ok(lsn) => lsn,
+                        Err(e) => {
+                            return current_batch_and_error!(e);
+                        }
+                    };
+                    BatchedFeMessage::GetPage {
+                        span,
+                        shard,
+                        effective_request_lsn,
+                        pages: smallvec::smallvec![(rel, blkno)],
+                    }
+                }
+            };
+
+            let batch_timeout = match self.server_side_batch_timeout {
+                Some(value) => value,
+                None => {
+                    // Batching is not enabled - stop on the first message.
+                    return Ok(Some(BatchOrEof::Batch(smallvec::smallvec![this_msg])));
+                }
+            };
+
+            // check if we can batch
+            match (&mut batch, this_msg) {
+                (None, this_msg) => {
+                    batch = Some(this_msg);
+                }
+                (
+                    Some(BatchedFeMessage::GetPage {
+                        span: _,
+                        shard: accum_shard,
+                        pages: accum_pages,
+                        effective_request_lsn: accum_lsn,
+                    }),
+                    BatchedFeMessage::GetPage {
+                        span: _,
+                        shard: this_shard,
+                        pages: this_pages,
+                        effective_request_lsn: this_lsn,
+                    },
+                ) if async {
+                    assert_eq!(this_pages.len(), 1);
+                    if accum_pages.len() >= Timeline::MAX_GET_VECTORED_KEYS as usize {
+                        assert_eq!(accum_pages.len(), Timeline::MAX_GET_VECTORED_KEYS as usize);
+                        return false;
+                    }
+                    if (accum_shard.tenant_shard_id, accum_shard.timeline_id)
+                        != (this_shard.tenant_shard_id, this_shard.timeline_id)
+                    {
+                        // TODO: we _could_ batch & execute each shard seperately (and in parallel).
+                        // But the current logic for keeping responses in order does not support that.
+                        return false;
+                    }
+                    // the vectored get currently only supports a single LSN, so, bounce as soon
+                    // as the effective request_lsn changes
+                    if *accum_lsn != this_lsn {
+                        return false;
+                    }
+                    true
+                }
+                .await =>
+                {
+                    // ok to batch
+                    accum_pages.extend(this_pages);
+                }
+                (Some(_), this_msg) => {
+                    // by default, don't continue batching
+                    break Some(this_msg);
+                }
+            }
+
+            // batching impl piece
+            let started_at = batch_started_at.get_or_insert_with(Instant::now);
+            if started_at.elapsed() > batch_timeout {
+                break None;
+            }
+        };
+
+        self.next_batch = next_batch;
+        Ok(batch.map(|b| BatchOrEof::Batch(smallvec::smallvec![b])))
+    }
+
     /// Pagestream sub-protocol handler.
     ///
     /// It is a simple request-response protocol inside a COPYBOTH session.
@@ -592,133 +861,165 @@ impl PageServerHandler {
             }
         }
 
+        // If [`PageServerHandler`] is reused for multiple pagestreams,
+        // then make sure to not process requests from the previous ones.
+        self.next_batch = None;
+
         loop {
-            // read request bytes (it's exactly 1 PagestreamFeMessage per CopyData)
-            let msg = tokio::select! {
-                biased;
-                _ = self.cancel.cancelled() => {
-                    return Err(QueryError::Shutdown)
+            let maybe_batched = self
+                .read_batch_from_connection(pgb, &tenant_id, &timeline_id, &ctx)
+                .await?;
+            let batched = match maybe_batched {
+                Some(BatchOrEof::Batch(b)) => b,
+                Some(BatchOrEof::Eof) => {
+                    break;
                 }
-                msg = pgb.read_message() => { msg }
-            };
-            let copy_data_bytes = match msg? {
-                Some(FeMessage::CopyData(bytes)) => bytes,
-                Some(FeMessage::Terminate) => break,
-                Some(m) => {
-                    return Err(QueryError::Other(anyhow::anyhow!(
-                        "unexpected message: {m:?} during COPY"
-                    )));
-                }
-                None => break, // client disconnected
-            };
-
-            trace!("query: {copy_data_bytes:?}");
-            fail::fail_point!("ps::handle-pagerequest-message");
-
-            // parse request
-            let neon_fe_msg = PagestreamFeMessage::parse(&mut copy_data_bytes.reader())?;
-
-            // invoke handler function
-            let (handler_result, span) = match neon_fe_msg {
-                PagestreamFeMessage::Exists(req) => {
-                    fail::fail_point!("ps::handle-pagerequest-message::exists");
-                    let span = tracing::info_span!("handle_get_rel_exists_request", rel = %req.rel, req_lsn = %req.request_lsn);
-                    (
-                        self.handle_get_rel_exists_request(tenant_id, timeline_id, &req, &ctx)
-                            .instrument(span.clone())
-                            .await,
-                        span,
-                    )
-                }
-                PagestreamFeMessage::Nblocks(req) => {
-                    fail::fail_point!("ps::handle-pagerequest-message::nblocks");
-                    let span = tracing::info_span!("handle_get_nblocks_request", rel = %req.rel, req_lsn = %req.request_lsn);
-                    (
-                        self.handle_get_nblocks_request(tenant_id, timeline_id, &req, &ctx)
-                            .instrument(span.clone())
-                            .await,
-                        span,
-                    )
-                }
-                PagestreamFeMessage::GetPage(req) => {
-                    fail::fail_point!("ps::handle-pagerequest-message::getpage");
-                    // shard_id is filled in by the handler
-                    let span = tracing::info_span!("handle_get_page_at_lsn_request", rel = %req.rel, blkno = %req.blkno, req_lsn = %req.request_lsn);
-                    (
-                        self.handle_get_page_at_lsn_request(tenant_id, timeline_id, &req, &ctx)
-                            .instrument(span.clone())
-                            .await,
-                        span,
-                    )
-                }
-                PagestreamFeMessage::DbSize(req) => {
-                    fail::fail_point!("ps::handle-pagerequest-message::dbsize");
-                    let span = tracing::info_span!("handle_db_size_request", dbnode = %req.dbnode, req_lsn = %req.request_lsn);
-                    (
-                        self.handle_db_size_request(tenant_id, timeline_id, &req, &ctx)
-                            .instrument(span.clone())
-                            .await,
-                        span,
-                    )
-                }
-                PagestreamFeMessage::GetSlruSegment(req) => {
-                    fail::fail_point!("ps::handle-pagerequest-message::slrusegment");
-                    let span = tracing::info_span!("handle_get_slru_segment_request", kind = %req.kind, segno = %req.segno, req_lsn = %req.request_lsn);
-                    (
-                        self.handle_get_slru_segment_request(tenant_id, timeline_id, &req, &ctx)
-                            .instrument(span.clone())
-                            .await,
-                        span,
-                    )
+                None => {
+                    continue;
                 }
             };
 
-            // Map handler result to protocol behavior.
-            // Some handler errors cause exit from pagestream protocol.
-            // Other handler errors are sent back as an error message and we stay in pagestream protocol.
-            let response_msg = match handler_result {
-                Err(e) => match &e {
-                    PageStreamError::Shutdown => {
-                        // If we fail to fulfil a request during shutdown, which may be _because_ of
-                        // shutdown, then do not send the error to the client.  Instead just drop the
-                        // connection.
-                        span.in_scope(|| info!("dropping connection due to shutdown"));
-                        return Err(QueryError::Shutdown);
+            for batch in batched {
+                // invoke handler function
+                let (handler_results, span): (
+                    Vec<Result<PagestreamBeMessage, PageStreamError>>,
+                    _,
+                ) = match batch {
+                    BatchedFeMessage::Exists { span, req } => {
+                        fail::fail_point!("ps::handle-pagerequest-message::exists");
+                        (
+                            vec![
+                                self.handle_get_rel_exists_request(
+                                    tenant_id,
+                                    timeline_id,
+                                    &req,
+                                    &ctx,
+                                )
+                                .instrument(span.clone())
+                                .await,
+                            ],
+                            span,
+                        )
                     }
-                    PageStreamError::Reconnect(reason) => {
-                        span.in_scope(|| info!("handler requested reconnect: {reason}"));
-                        return Err(QueryError::Reconnect);
+                    BatchedFeMessage::Nblocks { span, req } => {
+                        fail::fail_point!("ps::handle-pagerequest-message::nblocks");
+                        (
+                            vec![
+                                self.handle_get_nblocks_request(tenant_id, timeline_id, &req, &ctx)
+                                    .instrument(span.clone())
+                                    .await,
+                            ],
+                            span,
+                        )
                     }
-                    PageStreamError::Read(_)
-                    | PageStreamError::LsnTimeout(_)
-                    | PageStreamError::NotFound(_)
-                    | PageStreamError::BadRequest(_) => {
-                        // print the all details to the log with {:#}, but for the client the
-                        // error message is enough.  Do not log if shutting down, as the anyhow::Error
-                        // here includes cancellation which is not an error.
-                        let full = utils::error::report_compact_sources(&e);
-                        span.in_scope(|| {
-                            error!("error reading relation or page version: {full:#}")
-                        });
-                        PagestreamBeMessage::Error(PagestreamErrorResponse {
-                            message: e.to_string(),
-                        })
+                    BatchedFeMessage::GetPage {
+                        span,
+                        shard,
+                        effective_request_lsn,
+                        pages,
+                    } => {
+                        fail::fail_point!("ps::handle-pagerequest-message::getpage");
+                        (
+                            {
+                                let npages = pages.len();
+                                let res = self
+                                    .handle_get_page_at_lsn_request_batched(
+                                        &shard,
+                                        effective_request_lsn,
+                                        pages,
+                                        &ctx,
+                                    )
+                                    .instrument(span.clone())
+                                    .await;
+                                assert_eq!(res.len(), npages);
+                                res
+                            },
+                            span,
+                        )
                     }
-                },
-                Ok(response_msg) => response_msg,
-            };
+                    BatchedFeMessage::DbSize { span, req } => {
+                        fail::fail_point!("ps::handle-pagerequest-message::dbsize");
+                        (
+                            vec![
+                                self.handle_db_size_request(tenant_id, timeline_id, &req, &ctx)
+                                    .instrument(span.clone())
+                                    .await,
+                            ],
+                            span,
+                        )
+                    }
+                    BatchedFeMessage::GetSlruSegment { span, req } => {
+                        fail::fail_point!("ps::handle-pagerequest-message::slrusegment");
+                        (
+                            vec![
+                                self.handle_get_slru_segment_request(
+                                    tenant_id,
+                                    timeline_id,
+                                    &req,
+                                    &ctx,
+                                )
+                                .instrument(span.clone())
+                                .await,
+                            ],
+                            span,
+                        )
+                    }
+                    BatchedFeMessage::RespondError { span, error } => {
+                        // We've already decided to respond with an error, so we don't need to
+                        // call the handler.
+                        (vec![Err(error)], span)
+                    }
+                };
 
-            // marshal & transmit response message
-            pgb.write_message_noflush(&BeMessage::CopyData(&response_msg.serialize()))?;
-            tokio::select! {
-                biased;
-                _ = self.cancel.cancelled() => {
-                    // We were requested to shut down.
-                    info!("shutdown request received in page handler");
-                    return Err(QueryError::Shutdown)
+                // Map handler result to protocol behavior.
+                // Some handler errors cause exit from pagestream protocol.
+                // Other handler errors are sent back as an error message and we stay in pagestream protocol.
+                for handler_result in handler_results {
+                    let response_msg = match handler_result {
+                        Err(e) => match &e {
+                            PageStreamError::Shutdown => {
+                                // If we fail to fulfil a request during shutdown, which may be _because_ of
+                                // shutdown, then do not send the error to the client.  Instead just drop the
+                                // connection.
+                                span.in_scope(|| info!("dropping connection due to shutdown"));
+                                return Err(QueryError::Shutdown);
+                            }
+                            PageStreamError::Reconnect(reason) => {
+                                span.in_scope(|| info!("handler requested reconnect: {reason}"));
+                                return Err(QueryError::Reconnect);
+                            }
+                            PageStreamError::Read(_)
+                            | PageStreamError::LsnTimeout(_)
+                            | PageStreamError::NotFound(_)
+                            | PageStreamError::BadRequest(_) => {
+                                // print the all details to the log with {:#}, but for the client the
+                                // error message is enough.  Do not log if shutting down, as the anyhow::Error
+                                // here includes cancellation which is not an error.
+                                let full = utils::error::report_compact_sources(&e);
+                                span.in_scope(|| {
+                                    error!("error reading relation or page version: {full:#}")
+                                });
+                                PagestreamBeMessage::Error(PagestreamErrorResponse {
+                                    message: e.to_string(),
+                                })
+                            }
+                        },
+                        Ok(response_msg) => response_msg,
+                    };
+
+                    // marshal & transmit response message
+                    pgb.write_message_noflush(&BeMessage::CopyData(&response_msg.serialize()))?;
                 }
-                res = pgb.flush() => {
-                    res?;
+                tokio::select! {
+                    biased;
+                    _ = self.cancel.cancelled() => {
+                        // We were requested to shut down.
+                        info!("shutdown request received in page handler");
+                        return Err(QueryError::Shutdown)
+                    }
+                    res = pgb.flush() => {
+                        res?;
+                    }
                 }
             }
         }
@@ -964,60 +1265,30 @@ impl PageServerHandler {
         }))
     }
 
-    #[instrument(skip_all, fields(shard_id))]
-    async fn handle_get_page_at_lsn_request(
+    #[instrument(skip_all)]
+    async fn handle_get_page_at_lsn_request_batched(
         &mut self,
-        tenant_id: TenantId,
-        timeline_id: TimelineId,
-        req: &PagestreamGetPageRequest,
+        timeline: &Timeline,
+        effective_lsn: Lsn,
+        pages: smallvec::SmallVec<[(RelTag, BlockNumber); 1]>,
         ctx: &RequestContext,
-    ) -> Result<PagestreamBeMessage, PageStreamError> {
-        let timeline = match self
-            .timeline_handles
-            .get(
-                tenant_id,
-                timeline_id,
-                ShardSelector::Page(rel_block_to_key(req.rel, req.blkno)),
-            )
-            .await
-        {
-            Ok(tl) => tl,
-            Err(GetActiveTimelineError::Tenant(GetActiveTenantError::NotFound(_))) => {
-                // We already know this tenant exists in general, because we resolved it at
-                // start of connection.  Getting a NotFound here indicates that the shard containing
-                // the requested page is not present on this node: the client's knowledge of shard->pageserver
-                // mapping is out of date.
-                //
-                // Closing the connection by returning ``::Reconnect` has the side effect of rate-limiting above message, via
-                // client's reconnect backoff, as well as hopefully prompting the client to load its updated configuration
-                // and talk to a different pageserver.
-                return Err(PageStreamError::Reconnect(
-                    "getpage@lsn request routed to wrong shard".into(),
-                ));
-            }
-            Err(e) => return Err(e.into()),
-        };
-
-        let _timer = timeline
-            .query_metrics
-            .start_timer(metrics::SmgrQueryType::GetPageAtLsn, ctx);
-
-        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
-        let lsn = Self::wait_or_get_last_lsn(
-            &timeline,
-            req.request_lsn,
-            req.not_modified_since,
-            &latest_gc_cutoff_lsn,
+    ) -> Vec<Result<PagestreamBeMessage, PageStreamError>> {
+        debug_assert_current_span_has_tenant_and_timeline_id();
+        let _timer = timeline.query_metrics.start_timer_many(
+            metrics::SmgrQueryType::GetPageAtLsn,
+            pages.len(),
             ctx,
-        )
-        .await?;
+        );
 
-        let page = timeline
-            .get_rel_page_at_lsn(req.rel, req.blkno, Version::Lsn(lsn), ctx)
-            .await?;
+        let pages = timeline
+            .get_rel_page_at_lsn_batched(pages, effective_lsn, ctx)
+            .await;
 
-        Ok(PagestreamBeMessage::GetPage(PagestreamGetPageResponse {
-            page,
+        Vec::from_iter(pages.into_iter().map(|page| {
+            page.map(|page| {
+                PagestreamBeMessage::GetPage(models::PagestreamGetPageResponse { page })
+            })
+            .map_err(PageStreamError::from)
         }))
     }
 
@@ -1674,6 +1945,13 @@ fn set_tracing_field_shard_id(timeline: &Timeline) {
     debug_assert_current_span_has_tenant_and_timeline_id();
 }
 
+struct WaitedForLsn(Lsn);
+impl From<WaitedForLsn> for Lsn {
+    fn from(WaitedForLsn(lsn): WaitedForLsn) -> Self {
+        lsn
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use utils::shard::ShardCount;

pageserver/src/pgdatadir_mapping.rs

@@ -10,10 +10,15 @@ use super::tenant::{PageReconstructError, Timeline};
 use crate::aux_file;
 use crate::context::RequestContext;
 use crate::keyspace::{KeySpace, KeySpaceAccum};
-use crate::span::debug_assert_current_span_has_tenant_and_timeline_id_no_shard_id;
+use crate::span::{
+    debug_assert_current_span_has_tenant_and_timeline_id,
+    debug_assert_current_span_has_tenant_and_timeline_id_no_shard_id,
+};
+use crate::tenant::timeline::GetVectoredError;
 use anyhow::{ensure, Context};
 use bytes::{Buf, Bytes, BytesMut};
 use enum_map::Enum;
+use itertools::Itertools;
 use pageserver_api::key::Key;
 use pageserver_api::key::{
     dbdir_key_range, rel_block_to_key, rel_dir_to_key, rel_key_range, rel_size_to_key,
@@ -30,7 +35,7 @@ use postgres_ffi::relfile_utils::{FSM_FORKNUM, VISIBILITYMAP_FORKNUM};
 use postgres_ffi::BLCKSZ;
 use postgres_ffi::{Oid, RepOriginId, TimestampTz, TransactionId};
 use serde::{Deserialize, Serialize};
-use std::collections::{hash_map, HashMap, HashSet};
+use std::collections::{hash_map, BTreeMap, HashMap, HashSet};
 use std::ops::ControlFlow;
 use std::ops::Range;
 use strum::IntoEnumIterator;
@@ -193,26 +198,195 @@ impl Timeline {
         version: Version<'_>,
         ctx: &RequestContext,
     ) -> Result<Bytes, PageReconstructError> {
-        if tag.relnode == 0 {
-            return Err(PageReconstructError::Other(
-                RelationError::InvalidRelnode.into(),
-            ));
+        match version {
+            Version::Lsn(effective_lsn) => {
+                let pages = smallvec::smallvec![(tag, blknum)];
+                let res = self
+                    .get_rel_page_at_lsn_batched(pages, effective_lsn, ctx)
+                    .await;
+                assert_eq!(res.len(), 1);
+                res.into_iter().next().unwrap()
+            }
+            Version::Modified(modification) => {
+                if tag.relnode == 0 {
+                    return Err(PageReconstructError::Other(
+                        RelationError::InvalidRelnode.into(),
+                    ));
+                }
+
+                let nblocks = self.get_rel_size(tag, version, ctx).await?;
+                if blknum >= nblocks {
+                    debug!(
+                        "read beyond EOF at {} blk {} at {}, size is {}: returning all-zeros page",
+                        tag,
+                        blknum,
+                        version.get_lsn(),
+                        nblocks
+                    );
+                    return Ok(ZERO_PAGE.clone());
+                }
+
+                let key = rel_block_to_key(tag, blknum);
+                modification.get(key, ctx).await
+            }
+        }
+    }
+
+    /// Like [`Self::get_rel_page_at_lsn`], but returns a batch of pages.
+    ///
+    /// The ordering of the returned vec corresponds to the ordering of `pages`.
+    pub(crate) async fn get_rel_page_at_lsn_batched(
+        &self,
+        pages: smallvec::SmallVec<[(RelTag, BlockNumber); 1]>,
+        effective_lsn: Lsn,
+        ctx: &RequestContext,
+    ) -> Vec<Result<Bytes, PageReconstructError>> {
+        debug_assert_current_span_has_tenant_and_timeline_id();
+
+        let mut slots_filled = 0;
+        let page_count = pages.len();
+
+        // Would be nice to use smallvec here but it doesn't provide the spare_capacity_mut() API.
+        let mut result = Vec::with_capacity(pages.len());
+        let result_slots = result.spare_capacity_mut();
+
+        let mut keys_slots: BTreeMap<Key, smallvec::SmallVec<[usize; 1]>> = BTreeMap::default();
+        for (response_slot_idx, (tag, blknum)) in pages.into_iter().enumerate() {
+            if tag.relnode == 0 {
+                result_slots[response_slot_idx].write(Err(PageReconstructError::Other(
+                    RelationError::InvalidRelnode.into(),
+                )));
+
+                slots_filled += 1;
+                continue;
+            }
+
+            let nblocks = match self
+                .get_rel_size(tag, Version::Lsn(effective_lsn), ctx)
+                .await
+            {
+                Ok(nblocks) => nblocks,
+                Err(err) => {
+                    result_slots[response_slot_idx].write(Err(err));
+                    slots_filled += 1;
+                    continue;
+                }
+            };
+
+            if blknum >= nblocks {
+                debug!(
+                    "read beyond EOF at {} blk {} at {}, size is {}: returning all-zeros page",
+                    tag, blknum, effective_lsn, nblocks
+                );
+                result_slots[response_slot_idx].write(Ok(ZERO_PAGE.clone()));
+                slots_filled += 1;
+                continue;
+            }
+
+            let key = rel_block_to_key(tag, blknum);
+
+            let key_slots = keys_slots.entry(key).or_default();
+            key_slots.push(response_slot_idx);
         }
 
-        let nblocks = self.get_rel_size(tag, version, ctx).await?;
-        if blknum >= nblocks {
-            debug!(
-                "read beyond EOF at {} blk {} at {}, size is {}: returning all-zeros page",
-                tag,
-                blknum,
-                version.get_lsn(),
-                nblocks
-            );
-            return Ok(ZERO_PAGE.clone());
+        let keyspace = {
+            // add_key requires monotonicity
+            let mut acc = KeySpaceAccum::new();
+            for key in keys_slots
+                .keys()
+                // in fact it requires strong monotonicity
+                .dedup()
+            {
+                acc.add_key(*key);
+            }
+            acc.to_keyspace()
+        };
+
+        match self.get_vectored(keyspace, effective_lsn, ctx).await {
+            Ok(results) => {
+                for (key, res) in results {
+                    let mut key_slots = keys_slots.remove(&key).unwrap().into_iter();
+                    let first_slot = key_slots.next().unwrap();
+
+                    for slot in key_slots {
+                        let clone = match &res {
+                            Ok(buf) => Ok(buf.clone()),
+                            Err(err) => Err(match err {
+                                PageReconstructError::Cancelled => {
+                                    PageReconstructError::Cancelled
+                                }
+
+                                x @ PageReconstructError::Other(_) |
+                                x @ PageReconstructError::AncestorLsnTimeout(_) |
+                                x @ PageReconstructError::WalRedo(_) |
+                                x @ PageReconstructError::MissingKey(_) => {
+                                    PageReconstructError::Other(anyhow::anyhow!("there was more than one request for this key in the batch, error logged once: {x:?}"))
+                                },
+                            }),
+                        };
+
+                        result_slots[slot].write(clone);
+                        slots_filled += 1;
+                    }
+
+                    result_slots[first_slot].write(res);
+                    slots_filled += 1;
+                }
+            }
+            Err(err) => {
+                // this cannot really happen because get_vectored only errors globally on invalid LSN or too large batch size
+                // (We enforce the max batch size outside of this function, in the code that constructs the batch request.)
+                for slot in keys_slots.values().flatten() {
+                    // this whole `match` is a lot like `From<GetVectoredError> for PageReconstructError`
+                    // but without taking ownership of the GetVectoredError
+                    let err = match &err {
+                        GetVectoredError::Cancelled => {
+                            Err(PageReconstructError::Cancelled)
+                        }
+                        // TODO: restructure get_vectored API to make this error per-key
+                        GetVectoredError::MissingKey(err) => {
+                            Err(PageReconstructError::Other(anyhow::anyhow!("whole vectored get request failed because one or more of the requested keys were missing: {err:?}")))
+                        }
+                        // TODO: restructure get_vectored API to make this error per-key
+                        GetVectoredError::GetReadyAncestorError(err) => {
+                            Err(PageReconstructError::Other(anyhow::anyhow!("whole vectored get request failed because one or more key required ancestor that wasn't ready: {err:?}")))
+                        }
+                        // TODO: restructure get_vectored API to make this error per-key
+                        GetVectoredError::Other(err) => {
+                            Err(PageReconstructError::Other(
+                                anyhow::anyhow!("whole vectored get request failed: {err:?}"),
+                            ))
+                        }
+                        // TODO: we can prevent this error class by moving this check into the type system
+                        GetVectoredError::InvalidLsn(e) => {
+                            Err(anyhow::anyhow!("invalid LSN: {e:?}").into())
+                        }
+                        // NB: this should never happen in practice because we limit MAX_GET_VECTORED_KEYS
+                        // TODO: we can prevent this error class by moving this check into the type system
+                        GetVectoredError::Oversized(err) => {
+                            Err(anyhow::anyhow!(
+                                "batching oversized: {err:?}"
+                            )
+                            .into())
+                        }
+                    };
+
+                    result_slots[*slot].write(err);
+                }
+
+                slots_filled += keys_slots.values().map(|slots| slots.len()).sum::<usize>();
+            }
+        };
+
+        assert_eq!(slots_filled, page_count);
+        // SAFETY:
+        // 1. `result` and any of its uninint members are not read from until this point
+        // 2. The length below is tracked at run-time and matches the number of requested pages.
+        unsafe {
+            result.set_len(page_count);
         }
 
-        let key = rel_block_to_key(tag, blknum);
-        version.get(self, key, ctx).await
+        result
     }
 
     // Get size of a database in blocks

pageserver/src/tenant.rs

@@ -39,6 +39,7 @@ use remote_timeline_client::UploadQueueNotReadyError;
 use std::collections::BTreeMap;
 use std::fmt;
 use std::future::Future;
+use std::sync::atomic::AtomicBool;
 use std::sync::Weak;
 use std::time::SystemTime;
 use storage_broker::BrokerClientChannel;
@@ -188,6 +189,7 @@ pub struct TenantSharedResources {
 /// A [`Tenant`] is really an _attached_ tenant.  The configuration
 /// for an attached tenant is a subset of the [`LocationConf`], represented
 /// in this struct.
+#[derive(Clone)]
 pub(super) struct AttachedTenantConf {
     tenant_conf: TenantConfOpt,
     location: AttachedLocationConfig,
@@ -248,7 +250,8 @@ struct TimelinePreload {
 
 pub(crate) struct TenantPreload {
     tenant_manifest: TenantManifest,
-    timelines: HashMap<TimelineId, TimelinePreload>,
+    /// Map from timeline ID to a possible timeline preload. It is None iff the timeline is offloaded according to the manifest.
+    timelines: HashMap<TimelineId, Option<TimelinePreload>>,
 }
 
 /// When we spawn a tenant, there is a special mode for tenant creation that
@@ -524,6 +527,9 @@ pub struct OffloadedTimeline {
     /// Prevent two tasks from deleting the timeline at the same time. If held, the
     /// timeline is being deleted. If 'true', the timeline has already been deleted.
     pub delete_progress: TimelineDeleteProgress,
+
+    /// Part of the `OffloadedTimeline` object's lifecycle: this needs to be set before we drop it
+    pub deleted_from_ancestor: AtomicBool,
 }
 
 impl OffloadedTimeline {
@@ -533,9 +539,16 @@ impl OffloadedTimeline {
     /// the timeline is not in a stopped state.
     /// Panics if the timeline is not archived.
     fn from_timeline(timeline: &Timeline) -> Result<Self, UploadQueueNotReadyError> {
-        let ancestor_retain_lsn = timeline
-            .get_ancestor_timeline_id()
-            .map(|_timeline_id| timeline.get_ancestor_lsn());
+        let (ancestor_retain_lsn, ancestor_timeline_id) =
+            if let Some(ancestor_timeline) = timeline.ancestor_timeline() {
+                let ancestor_lsn = timeline.get_ancestor_lsn();
+                let ancestor_timeline_id = ancestor_timeline.timeline_id;
+                let mut gc_info = ancestor_timeline.gc_info.write().unwrap();
+                gc_info.insert_child(timeline.timeline_id, ancestor_lsn, MaybeOffloaded::Yes);
+                (Some(ancestor_lsn), Some(ancestor_timeline_id))
+            } else {
+                (None, None)
+            };
         let archived_at = timeline
             .remote_client
             .archived_at_stopped_queue()?
@@ -543,14 +556,17 @@ impl OffloadedTimeline {
         Ok(Self {
             tenant_shard_id: timeline.tenant_shard_id,
             timeline_id: timeline.timeline_id,
-            ancestor_timeline_id: timeline.get_ancestor_timeline_id(),
+            ancestor_timeline_id,
             ancestor_retain_lsn,
             archived_at,
 
             delete_progress: timeline.delete_progress.clone(),
+            deleted_from_ancestor: AtomicBool::new(false),
         })
     }
     fn from_manifest(tenant_shard_id: TenantShardId, manifest: &OffloadedTimelineManifest) -> Self {
+        // We expect to reach this case in tenant loading, where the `retain_lsn` is populated in the parent's `gc_info`
+        // by the `initialize_gc_info` function.
         let OffloadedTimelineManifest {
             timeline_id,
             ancestor_timeline_id,
@@ -564,6 +580,7 @@ impl OffloadedTimeline {
             ancestor_retain_lsn,
             archived_at,
             delete_progress: TimelineDeleteProgress::default(),
+            deleted_from_ancestor: AtomicBool::new(false),
         }
     }
     fn manifest(&self) -> OffloadedTimelineManifest {
@@ -581,6 +598,37 @@ impl OffloadedTimeline {
             archived_at: *archived_at,
         }
     }
+    /// Delete this timeline's retain_lsn from its ancestor, if present in the given tenant
+    fn delete_from_ancestor_with_timelines(
+        &self,
+        timelines: &std::sync::MutexGuard<'_, HashMap<TimelineId, Arc<Timeline>>>,
+    ) {
+        if let (Some(_retain_lsn), Some(ancestor_timeline_id)) =
+            (self.ancestor_retain_lsn, self.ancestor_timeline_id)
+        {
+            if let Some((_, ancestor_timeline)) = timelines
+                .iter()
+                .find(|(tid, _tl)| **tid == ancestor_timeline_id)
+            {
+                let removal_happened = ancestor_timeline
+                    .gc_info
+                    .write()
+                    .unwrap()
+                    .remove_child_offloaded(self.timeline_id);
+                if !removal_happened {
+                    tracing::error!(tenant_id = %self.tenant_shard_id.tenant_id, shard_id = %self.tenant_shard_id.shard_slug(), timeline_id = %self.timeline_id,
+                        "Couldn't remove retain_lsn entry from offloaded timeline's parent: already removed");
+                }
+            }
+        }
+        self.deleted_from_ancestor.store(true, Ordering::Release);
+    }
+    /// Call [`Self::delete_from_ancestor_with_timelines`] instead if possible.
+    ///
+    /// As the entire tenant is being dropped, don't bother deregistering the `retain_lsn` from the ancestor.
+    fn defuse_for_tenant_drop(&self) {
+        self.deleted_from_ancestor.store(true, Ordering::Release);
+    }
 }
 
 impl fmt::Debug for OffloadedTimeline {
@@ -589,6 +637,17 @@ impl fmt::Debug for OffloadedTimeline {
     }
 }
 
+impl Drop for OffloadedTimeline {
+    fn drop(&mut self) {
+        if !self.deleted_from_ancestor.load(Ordering::Acquire) {
+            tracing::warn!(
+                "offloaded timeline {} was dropped without having cleaned it up at the ancestor",
+                self.timeline_id
+            );
+        }
+    }
+}
+
 #[derive(Copy, Clone, PartialEq, Eq, Hash, Debug)]
 pub enum MaybeOffloaded {
     Yes,
@@ -700,6 +759,9 @@ pub enum DeleteTimelineError {
     #[error("Timeline deletion is already in progress")]
     AlreadyInProgress(Arc<tokio::sync::Mutex<DeleteTimelineFlow>>),
 
+    #[error("Cancelled")]
+    Cancelled,
+
     #[error(transparent)]
     Other(#[from] anyhow::Error),
 }
@@ -710,6 +772,7 @@ impl Debug for DeleteTimelineError {
             Self::NotFound => write!(f, "NotFound"),
             Self::HasChildren(c) => f.debug_tuple("HasChildren").field(c).finish(),
             Self::AlreadyInProgress(_) => f.debug_tuple("AlreadyInProgress").finish(),
+            Self::Cancelled => f.debug_tuple("Cancelled").finish(),
             Self::Other(e) => f.debug_tuple("Other").field(e).finish(),
         }
     }
@@ -1336,7 +1399,7 @@ impl Tenant {
         // Get list of remote timelines
         // download index files for every tenant timeline
         info!("listing remote timelines");
-        let (remote_timeline_ids, other_keys) = remote_timeline_client::list_remote_timelines(
+        let (mut remote_timeline_ids, other_keys) = remote_timeline_client::list_remote_timelines(
             remote_storage,
             self.tenant_shard_id,
             cancel.clone(),
@@ -1370,11 +1433,27 @@ impl Tenant {
             warn!("Unexpected non timeline key {k}");
         }
 
+        // Avoid downloading IndexPart of offloaded timelines.
+        let mut offloaded_with_prefix = HashSet::new();
+        for offloaded in tenant_manifest.offloaded_timelines.iter() {
+            if remote_timeline_ids.remove(&offloaded.timeline_id) {
+                offloaded_with_prefix.insert(offloaded.timeline_id);
+            } else {
+                // We'll take care later of timelines in the manifest without a prefix
+            }
+        }
+
+        let timelines = self
+            .load_timelines_metadata(remote_timeline_ids, remote_storage, cancel)
+            .await?;
+
         Ok(TenantPreload {
             tenant_manifest,
-            timelines: self
-                .load_timelines_metadata(remote_timeline_ids, remote_storage, cancel)
-                .await?,
+            timelines: timelines
+                .into_iter()
+                .map(|(id, tl)| (id, Some(tl)))
+                .chain(offloaded_with_prefix.into_iter().map(|id| (id, None)))
+                .collect(),
         })
     }
 
@@ -1405,6 +1484,19 @@ impl Tenant {
             offloaded_timelines_list.push((timeline_id, Arc::new(offloaded_timeline)));
             offloaded_timeline_ids.insert(timeline_id);
         }
+        // Complete deletions for offloaded timeline id's from manifest.
+        // The manifest will be uploaded later in this function.
+        offloaded_timelines_list
+            .retain(|(offloaded_id, offloaded)| {
+                // Existence of a timeline is finally determined by the existence of an index-part.json in remote storage.
+                // If there is dangling references in another location, they need to be cleaned up.
+                let delete = !preload.timelines.contains_key(offloaded_id);
+                if delete {
+                    tracing::info!("Removing offloaded timeline {offloaded_id} from manifest as no remote prefix was found");
+                    offloaded.defuse_for_tenant_drop();
+                }
+                !delete
+        });
 
         let mut timelines_to_resume_deletions = vec![];
 
@@ -1412,10 +1504,9 @@ impl Tenant {
         let mut timeline_ancestors = HashMap::new();
         let mut existent_timelines = HashSet::new();
         for (timeline_id, preload) in preload.timelines {
-            if offloaded_timeline_ids.remove(&timeline_id) {
-                // The timeline is offloaded, skip loading it.
-                continue;
-            }
+            let Some(preload) = preload else { continue };
+            // This is an invariant of the `preload` function's API
+            assert!(!offloaded_timeline_ids.contains(&timeline_id));
             let index_part = match preload.index_part {
                 Ok(i) => {
                     debug!("remote index part exists for timeline {timeline_id}");
@@ -1525,30 +1616,13 @@ impl Tenant {
             .context("resume_deletion")
             .map_err(LoadLocalTimelineError::ResumeDeletion)?;
         }
-        // Complete deletions for offloaded timeline id's.
-        offloaded_timelines_list
-            .retain(|(offloaded_id, _offloaded)| {
-                // At this point, offloaded_timeline_ids has the list of all offloaded timelines
-                // without a prefix in S3, so they are inexistent.
-                // In the end, existence of a timeline is finally determined by the existence of an index-part.json in remote storage.
-                // If there is a dangling reference in another location, they need to be cleaned up.
-                let delete = offloaded_timeline_ids.contains(offloaded_id);
-                if delete {
-                    tracing::info!("Removing offloaded timeline {offloaded_id} from manifest as no remote prefix was found");
-                }
-                !delete
-        });
-        if !offloaded_timelines_list.is_empty() {
-            tracing::info!(
-                "Tenant has {} offloaded timelines",
-                offloaded_timelines_list.len()
-            );
-        }
+        let needs_manifest_upload =
+            offloaded_timelines_list.len() != preload.tenant_manifest.offloaded_timelines.len();
         {
             let mut offloaded_timelines_accessor = self.timelines_offloaded.lock().unwrap();
             offloaded_timelines_accessor.extend(offloaded_timelines_list.into_iter());
         }
-        if !offloaded_timeline_ids.is_empty() {
+        if needs_manifest_upload {
             self.store_tenant_manifest().await?;
         }
 
@@ -1734,6 +1808,7 @@ impl Tenant {
             self.tenant_shard_id,
             timeline_id,
             self.generation,
+            &self.tenant_conf.load().location,
         )
     }
 
@@ -1923,9 +1998,15 @@ impl Tenant {
                 )));
             };
             let mut offloaded_timelines = self.timelines_offloaded.lock().unwrap();
-            if offloaded_timelines.remove(&timeline_id).is_none() {
-                warn!("timeline already removed from offloaded timelines");
+            match offloaded_timelines.remove(&timeline_id) {
+                Some(offloaded) => {
+                    offloaded.delete_from_ancestor_with_timelines(&timelines);
+                }
+                None => warn!("timeline already removed from offloaded timelines"),
             }
+
+            self.initialize_gc_info(&timelines, &offloaded_timelines, Some(timeline_id));
+
             Arc::clone(timeline)
         };
 
@@ -2378,6 +2459,12 @@ impl Tenant {
             .remote_client
             .wait_completion()
             .await
+            .map_err(|e| match e {
+                WaitCompletionError::NotInitialized(
+                    e, // If the queue is already stopped, it's a shutdown error.
+                ) if e.is_stopping() => CreateTimelineError::ShuttingDown,
+                e => CreateTimelineError::Other(e.into()),
+            })
             .context("wait for timeline initial uploads to complete")?;
 
         // The creating task is responsible for activating the timeline.
@@ -2442,6 +2529,10 @@ impl Tenant {
         {
             let conf = self.tenant_conf.load();
 
+            // If we may not delete layers, then simply skip GC.  Even though a tenant
+            // in AttachedMulti state could do GC and just enqueue the blocked deletions,
+            // the only advantage to doing it is to perhaps shrink the LayerMap metadata
+            // a bit sooner than we would achieve by waiting for AttachedSingle status.
             if !conf.location.may_delete_layers_hint() {
                 info!("Skipping GC in location state {:?}", conf.location);
                 return Ok(GcResult::default());
@@ -2483,7 +2574,14 @@ impl Tenant {
 
         {
             let conf = self.tenant_conf.load();
-            if !conf.location.may_delete_layers_hint() || !conf.location.may_upload_layers_hint() {
+
+            // Note that compaction usually requires deletions, but we don't respect
+            // may_delete_layers_hint here: that is because tenants in AttachedMulti
+            // should proceed with compaction even if they can't do deletion, to avoid
+            // accumulating dangerously deep stacks of L0 layers.  Deletions will be
+            // enqueued inside RemoteTimelineClient, and executed layer if/when we transition
+            // to AttachedSingle state.
+            if !conf.location.may_upload_layers_hint() {
                 info!("Skipping compaction in location state {:?}", conf.location);
                 return Ok(false);
             }
@@ -2663,7 +2761,7 @@ impl Tenant {
                 .filter(|timeline| !(timeline.is_broken() || timeline.is_stopping()));
 
             // Before activation, populate each Timeline's GcInfo with information about its children
-            self.initialize_gc_info(&timelines_accessor, &timelines_offloaded_accessor);
+            self.initialize_gc_info(&timelines_accessor, &timelines_offloaded_accessor, None);
 
             // Spawn gc and compaction loops. The loops will shut themselves
             // down when they notice that the tenant is inactive.
@@ -2778,8 +2876,14 @@ impl Tenant {
                 let timeline_id = timeline.timeline_id;
                 let span = tracing::info_span!("timeline_shutdown", %timeline_id, ?shutdown_mode);
                 js.spawn(async move { timeline.shutdown(shutdown_mode).instrument(span).await });
-            })
-        };
+            });
+        }
+        {
+            let timelines_offloaded = self.timelines_offloaded.lock().unwrap();
+            timelines_offloaded.values().for_each(|timeline| {
+                timeline.defuse_for_tenant_drop();
+            });
+        }
         // test_long_timeline_create_then_tenant_delete is leaning on this message
         tracing::info!("Waiting for timelines...");
         while let Some(res) = js.join_next().await {
@@ -3355,6 +3459,7 @@ impl Tenant {
         // this race is not possible if both request types come from the storage
         // controller (as they should!) because an exclusive op lock is required
         // on the storage controller side.
+
         self.tenant_conf.rcu(|inner| {
             Arc::new(AttachedTenantConf {
                 tenant_conf: new_tenant_conf.clone(),
@@ -3364,20 +3469,22 @@ impl Tenant {
             })
         });
 
+        let updated = self.tenant_conf.load().clone();
+
         self.tenant_conf_updated(&new_tenant_conf);
         // Don't hold self.timelines.lock() during the notifies.
         // There's no risk of deadlock right now, but there could be if we consolidate
         // mutexes in struct Timeline in the future.
         let timelines = self.list_timelines();
         for timeline in timelines {
-            timeline.tenant_conf_updated(&new_tenant_conf);
+            timeline.tenant_conf_updated(&updated);
         }
     }
 
     pub(crate) fn set_new_location_config(&self, new_conf: AttachedTenantConf) {
         let new_tenant_conf = new_conf.tenant_conf.clone();
 
-        self.tenant_conf.store(Arc::new(new_conf));
+        self.tenant_conf.store(Arc::new(new_conf.clone()));
 
         self.tenant_conf_updated(&new_tenant_conf);
         // Don't hold self.timelines.lock() during the notifies.
@@ -3385,7 +3492,7 @@ impl Tenant {
         // mutexes in struct Timeline in the future.
         let timelines = self.list_timelines();
         for timeline in timelines {
-            timeline.tenant_conf_updated(&new_tenant_conf);
+            timeline.tenant_conf_updated(&new_conf);
         }
     }
 
@@ -3763,10 +3870,13 @@ impl Tenant {
         &self,
         timelines: &std::sync::MutexGuard<HashMap<TimelineId, Arc<Timeline>>>,
         timelines_offloaded: &std::sync::MutexGuard<HashMap<TimelineId, Arc<OffloadedTimeline>>>,
+        restrict_to_timeline: Option<TimelineId>,
     ) {
-        // This function must be called before activation: after activation timeline create/delete operations
-        // might happen, and this function is not safe to run concurrently with those.
-        assert!(!self.is_active());
+        if restrict_to_timeline.is_none() {
+            // This function must be called before activation: after activation timeline create/delete operations
+            // might happen, and this function is not safe to run concurrently with those.
+            assert!(!self.is_active());
+        }
 
         // Scan all timelines. For each timeline, remember the timeline ID and
         // the branch point where it was created.
@@ -3799,7 +3909,12 @@ impl Tenant {
         let horizon = self.get_gc_horizon();
 
         // Populate each timeline's GcInfo with information about its child branches
-        for timeline in timelines.values() {
+        let timelines_to_write = if let Some(timeline_id) = restrict_to_timeline {
+            itertools::Either::Left(timelines.get(&timeline_id).into_iter())
+        } else {
+            itertools::Either::Right(timelines.values())
+        };
+        for timeline in timelines_to_write {
             let mut branchpoints: Vec<(Lsn, TimelineId, MaybeOffloaded)> = all_branchpoints
                 .remove(&timeline.timeline_id)
                 .unwrap_or_default();
@@ -4445,6 +4560,7 @@ impl Tenant {
             self.tenant_shard_id,
             timeline_id,
             self.generation,
+            &self.tenant_conf.load().location,
         )
     }
 
@@ -5166,7 +5282,7 @@ mod tests {
     use storage_layer::PersistentLayerKey;
     use tests::storage_layer::ValuesReconstructState;
     use tests::timeline::{GetVectoredError, ShutdownMode};
-    use timeline::DeltaLayerTestDesc;
+    use timeline::{CompactOptions, DeltaLayerTestDesc};
     use utils::id::TenantId;
 
     #[cfg(feature = "testing")]
@@ -7640,7 +7756,7 @@ mod tests {
 
         let cancel = CancellationToken::new();
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
 
@@ -7717,7 +7833,7 @@ mod tests {
             guard.cutoffs.space = Lsn(0x40);
         }
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
 
@@ -8149,7 +8265,7 @@ mod tests {
 
         let cancel = CancellationToken::new();
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
 
@@ -8178,7 +8294,7 @@ mod tests {
             guard.cutoffs.space = Lsn(0x40);
         }
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
 
@@ -8731,7 +8847,14 @@ mod tests {
         dryrun_flags.insert(CompactFlags::DryRun);
 
         tline
-            .compact_with_gc(&cancel, dryrun_flags, &ctx)
+            .compact_with_gc(
+                &cancel,
+                CompactOptions {
+                    flags: dryrun_flags,
+                    compact_range: None,
+                },
+                &ctx,
+            )
             .await
             .unwrap();
         // We expect layer map to be the same b/c the dry run flag, but we don't know whether there will be other background jobs
@@ -8739,14 +8862,14 @@ mod tests {
         verify_result().await;
 
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
         verify_result().await;
 
         // compact again
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
         verify_result().await;
@@ -8759,14 +8882,14 @@ mod tests {
             guard.cutoffs.space = Lsn(0x38);
         }
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
         verify_result().await; // no wals between 0x30 and 0x38, so we should obtain the same result
 
         // not increasing the GC horizon and compact again
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
         verify_result().await;
@@ -8960,7 +9083,14 @@ mod tests {
         dryrun_flags.insert(CompactFlags::DryRun);
 
         tline
-            .compact_with_gc(&cancel, dryrun_flags, &ctx)
+            .compact_with_gc(
+                &cancel,
+                CompactOptions {
+                    flags: dryrun_flags,
+                    compact_range: None,
+                },
+                &ctx,
+            )
             .await
             .unwrap();
         // We expect layer map to be the same b/c the dry run flag, but we don't know whether there will be other background jobs
@@ -8968,14 +9098,14 @@ mod tests {
         verify_result().await;
 
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
         verify_result().await;
 
         // compact again
         tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
         verify_result().await;
@@ -9160,7 +9290,7 @@ mod tests {
 
         let cancel = CancellationToken::new();
         branch_tline
-            .compact_with_gc(&cancel, EnumSet::new(), &ctx)
+            .compact_with_gc(&cancel, CompactOptions::default(), &ctx)
             .await
             .unwrap();
 
@@ -9646,4 +9776,54 @@ mod tests {
 
         Ok(())
     }
+
+    #[cfg(feature = "testing")]
+    #[tokio::test]
+    async fn test_timeline_offload_retain_lsn() -> anyhow::Result<()> {
+        let harness = TenantHarness::create("test_timeline_offload_retain_lsn")
+            .await
+            .unwrap();
+        let (tenant, ctx) = harness.load().await;
+        let tline_parent = tenant
+            .create_test_timeline(TIMELINE_ID, Lsn(0x10), DEFAULT_PG_VERSION, &ctx)
+            .await
+            .unwrap();
+        let tline_child = tenant
+            .branch_timeline_test(&tline_parent, NEW_TIMELINE_ID, Some(Lsn(0x20)), &ctx)
+            .await
+            .unwrap();
+        {
+            let gc_info_parent = tline_parent.gc_info.read().unwrap();
+            assert_eq!(
+                gc_info_parent.retain_lsns,
+                vec![(Lsn(0x20), tline_child.timeline_id, MaybeOffloaded::No)]
+            );
+        }
+        // We have to directly call the remote_client instead of using the archive function to avoid constructing broker client...
+        tline_child
+            .remote_client
+            .schedule_index_upload_for_timeline_archival_state(TimelineArchivalState::Archived)
+            .unwrap();
+        tline_child.remote_client.wait_completion().await.unwrap();
+        offload_timeline(&tenant, &tline_child)
+            .instrument(tracing::info_span!(parent: None, "offload_test", tenant_id=%"test", shard_id=%"test", timeline_id=%"test"))
+            .await.unwrap();
+        let child_timeline_id = tline_child.timeline_id;
+        Arc::try_unwrap(tline_child).unwrap();
+
+        {
+            let gc_info_parent = tline_parent.gc_info.read().unwrap();
+            assert_eq!(
+                gc_info_parent.retain_lsns,
+                vec![(Lsn(0x20), child_timeline_id, MaybeOffloaded::Yes)]
+            );
+        }
+
+        tenant
+            .get_offloaded_timeline(child_timeline_id)
+            .unwrap()
+            .defuse_for_tenant_drop();
+
+        Ok(())
+    }
 }

pageserver/src/tenant/mgr.rs

@@ -1719,10 +1719,11 @@ impl TenantManager {
                     parent_layers.push(relative_path.to_owned());
                 }
             }
-            debug_assert!(
-                !parent_layers.is_empty(),
-                "shutdown cannot empty the layermap"
-            );
+
+            if parent_layers.is_empty() {
+                tracing::info!("Ancestor shard has no resident layer to hard link");
+            }
+
             (parent_timelines, parent_layers)
         };
 

pageserver/src/tenant/remote_timeline_client.rs

@@ -197,6 +197,7 @@ use utils::backoff::{
     self, exponential_backoff, DEFAULT_BASE_BACKOFF_SECONDS, DEFAULT_MAX_BACKOFF_SECONDS,
 };
 use utils::pausable_failpoint;
+use utils::shard::ShardNumber;
 
 use std::collections::{HashMap, VecDeque};
 use std::sync::atomic::{AtomicU32, Ordering};
@@ -240,10 +241,11 @@ use utils::id::{TenantId, TimelineId};
 
 use self::index::IndexPart;
 
+use super::config::AttachedLocationConfig;
 use super::metadata::MetadataUpdate;
 use super::storage_layer::{Layer, LayerName, ResidentLayer};
 use super::upload_queue::{NotInitialized, SetDeletedFlagProgress};
-use super::Generation;
+use super::{DeleteTimelineError, Generation};
 
 pub(crate) use download::{
     download_index_part, download_tenant_manifest, is_temp_download_file,
@@ -301,6 +303,36 @@ pub enum WaitCompletionError {
 #[derive(Debug, thiserror::Error)]
 #[error("Upload queue either in unexpected state or hasn't downloaded manifest yet")]
 pub struct UploadQueueNotReadyError;
+/// Behavioral modes that enable seamless live migration.
+///
+/// See docs/rfcs/028-pageserver-migration.md to understand how these fit in.
+struct RemoteTimelineClientConfig {
+    /// If this is false, then update to remote_consistent_lsn are dropped rather
+    /// than being submitted to DeletionQueue for validation.  This behavior is
+    /// used when a tenant attachment is known to have a stale generation number,
+    /// such that validation attempts will always fail.  This is not necessary
+    /// for correctness, but avoids spamming error statistics with failed validations
+    /// when doing migrations of tenants.
+    process_remote_consistent_lsn_updates: bool,
+
+    /// If this is true, then object deletions are held in a buffer in RemoteTimelineClient
+    /// rather than being submitted to the DeletionQueue.  This behavior is used when a tenant
+    /// is known to be multi-attached, in order to avoid disrupting other attached tenants
+    /// whose generations' metadata refers to the deleted objects.
+    block_deletions: bool,
+}
+
+/// RemoteTimelineClientConfig's state is entirely driven by LocationConf, but we do
+/// not carry the entire LocationConf structure: it's much more than we need.  The From
+/// impl extracts the subset of the LocationConf that is interesting to RemoteTimelineClient.
+impl From<&AttachedLocationConfig> for RemoteTimelineClientConfig {
+    fn from(lc: &AttachedLocationConfig) -> Self {
+        Self {
+            block_deletions: !lc.may_delete_layers_hint(),
+            process_remote_consistent_lsn_updates: lc.may_upload_layers_hint(),
+        }
+    }
+}
 
 /// A client for accessing a timeline's data in remote storage.
 ///
@@ -321,7 +353,7 @@ pub struct UploadQueueNotReadyError;
 /// in the index part file, whenever timeline metadata is uploaded.
 ///
 /// Downloads are not queued, they are performed immediately.
-pub struct RemoteTimelineClient {
+pub(crate) struct RemoteTimelineClient {
     conf: &'static PageServerConf,
 
     runtime: tokio::runtime::Handle,
@@ -338,6 +370,9 @@ pub struct RemoteTimelineClient {
 
     deletion_queue_client: DeletionQueueClient,
 
+    /// Subset of tenant configuration used to control upload behaviors during migrations
+    config: std::sync::RwLock<RemoteTimelineClientConfig>,
+
     cancel: CancellationToken,
 }
 
@@ -348,13 +383,14 @@ impl RemoteTimelineClient {
     /// Note: the caller must initialize the upload queue before any uploads can be scheduled,
     /// by calling init_upload_queue.
     ///
-    pub fn new(
+    pub(crate) fn new(
         remote_storage: GenericRemoteStorage,
         deletion_queue_client: DeletionQueueClient,
         conf: &'static PageServerConf,
         tenant_shard_id: TenantShardId,
         timeline_id: TimelineId,
         generation: Generation,
+        location_conf: &AttachedLocationConfig,
     ) -> RemoteTimelineClient {
         RemoteTimelineClient {
             conf,
@@ -374,6 +410,7 @@ impl RemoteTimelineClient {
                 &tenant_shard_id,
                 &timeline_id,
             )),
+            config: std::sync::RwLock::new(RemoteTimelineClientConfig::from(location_conf)),
             cancel: CancellationToken::new(),
         }
     }
@@ -429,6 +466,43 @@ impl RemoteTimelineClient {
         Ok(())
     }
 
+    /// Notify this client of a change to its parent tenant's config, as this may cause us to
+    /// take action (unblocking deletions when transitioning from AttachedMulti to AttachedSingle)
+    pub(super) fn update_config(&self, location_conf: &AttachedLocationConfig) {
+        let new_conf = RemoteTimelineClientConfig::from(location_conf);
+        let unblocked = !new_conf.block_deletions;
+
+        // Update config before draining deletions, so that we don't race with more being
+        // inserted.  This can result in deletions happening our of order, but that does not
+        // violate any invariants: deletions only need to be ordered relative to upload of the index
+        // that dereferences the deleted objects, and we are not changing that order.
+        *self.config.write().unwrap() = new_conf;
+
+        if unblocked {
+            // If we may now delete layers, drain any that were blocked in our old
+            // configuration state
+            let mut queue_locked = self.upload_queue.lock().unwrap();
+
+            if let Ok(queue) = queue_locked.initialized_mut() {
+                let blocked_deletions = std::mem::take(&mut queue.blocked_deletions);
+                for d in blocked_deletions {
+                    if let Err(e) = self.deletion_queue_client.push_layers_sync(
+                        self.tenant_shard_id,
+                        self.timeline_id,
+                        self.generation,
+                        d.layers,
+                    ) {
+                        // This could happen if the pageserver is shut down while a tenant
+                        // is transitioning from a deletion-blocked state: we will leak some
+                        // S3 objects in this case.
+                        warn!("Failed to drain blocked deletions: {}", e);
+                        break;
+                    }
+                }
+            }
+        }
+    }
+
     /// Returns `None` if nothing is yet uplodaded, `Some(disk_consistent_lsn)` otherwise.
     pub fn remote_consistent_lsn_projected(&self) -> Option<Lsn> {
         match &mut *self.upload_queue.lock().unwrap() {
@@ -1550,15 +1624,17 @@ impl RemoteTimelineClient {
     /// Prerequisites: UploadQueue should be in stopped state and deleted_at should be successfuly set.
     /// The function deletes layer files one by one, then lists the prefix to see if we leaked something
     /// deletes leaked files if any and proceeds with deletion of index file at the end.
-    pub(crate) async fn delete_all(self: &Arc<Self>) -> anyhow::Result<()> {
+    pub(crate) async fn delete_all(self: &Arc<Self>) -> Result<(), DeleteTimelineError> {
         debug_assert_current_span_has_tenant_and_timeline_id();
 
         let layers: Vec<RemotePath> = {
             let mut locked = self.upload_queue.lock().unwrap();
-            let stopped = locked.stopped_mut()?;
+            let stopped = locked.stopped_mut().map_err(DeleteTimelineError::Other)?;
 
             if !matches!(stopped.deleted_at, SetDeletedFlagProgress::Successful(_)) {
-                anyhow::bail!("deleted_at is not set")
+                return Err(DeleteTimelineError::Other(anyhow::anyhow!(
+                    "deleted_at is not set"
+                )));
             }
 
             debug_assert!(stopped.upload_queue_for_deletion.no_pending_work());
@@ -1593,7 +1669,10 @@ impl RemoteTimelineClient {
         };
 
         let layer_deletion_count = layers.len();
-        self.deletion_queue_client.push_immediate(layers).await?;
+        self.deletion_queue_client
+            .push_immediate(layers)
+            .await
+            .map_err(|_| DeleteTimelineError::Cancelled)?;
 
         // Delete the initdb.tar.zst, which is not always present, but deletion attempts of
         // inexistant objects are not considered errors.
@@ -1601,7 +1680,8 @@ impl RemoteTimelineClient {
             remote_initdb_archive_path(&self.tenant_shard_id.tenant_id, &self.timeline_id);
         self.deletion_queue_client
             .push_immediate(vec![initdb_path])
-            .await?;
+            .await
+            .map_err(|_| DeleteTimelineError::Cancelled)?;
 
         // Do not delete index part yet, it is needed for possible retry. If we remove it first
         // and retry will arrive to different pageserver there wont be any traces of it on remote storage
@@ -1609,7 +1689,9 @@ impl RemoteTimelineClient {
 
         // Execute all pending deletions, so that when we proceed to do a listing below, we aren't
         // taking the burden of listing all the layers that we already know we should delete.
-        self.flush_deletion_queue().await?;
+        self.flush_deletion_queue()
+            .await
+            .map_err(|_| DeleteTimelineError::Cancelled)?;
 
         let cancel = shutdown_token();
 
@@ -1672,28 +1754,32 @@ impl RemoteTimelineClient {
         if !remaining_layers.is_empty() {
             self.deletion_queue_client
                 .push_immediate(remaining_layers)
-                .await?;
+                .await
+                .map_err(|_| DeleteTimelineError::Cancelled)?;
         }
 
         fail::fail_point!("timeline-delete-before-index-delete", |_| {
-            Err(anyhow::anyhow!(
+            Err(DeleteTimelineError::Other(anyhow::anyhow!(
                 "failpoint: timeline-delete-before-index-delete"
-            ))?
+            )))?
         });
 
         debug!("enqueuing index part deletion");
         self.deletion_queue_client
             .push_immediate([latest_index].to_vec())
-            .await?;
+            .await
+            .map_err(|_| DeleteTimelineError::Cancelled)?;
 
         // Timeline deletion is rare and we have probably emitted a reasonably number of objects: wait
         // for a flush to a persistent deletion list so that we may be sure deletion will occur.
-        self.flush_deletion_queue().await?;
+        self.flush_deletion_queue()
+            .await
+            .map_err(|_| DeleteTimelineError::Cancelled)?;
 
         fail::fail_point!("timeline-delete-after-index-delete", |_| {
-            Err(anyhow::anyhow!(
+            Err(DeleteTimelineError::Other(anyhow::anyhow!(
                 "failpoint: timeline-delete-after-index-delete"
-            ))?
+            )))?
         });
 
         info!(prefix=%timeline_storage_path, referenced=layer_deletion_count, not_referenced=%not_referenced_count, "done deleting in timeline prefix, including index_part.json");
@@ -1900,16 +1986,24 @@ impl RemoteTimelineClient {
                     res
                 }
                 UploadOp::Delete(delete) => {
-                    pausable_failpoint!("before-delete-layer-pausable");
-                    self.deletion_queue_client
-                        .push_layers(
-                            self.tenant_shard_id,
-                            self.timeline_id,
-                            self.generation,
-                            delete.layers.clone(),
-                        )
-                        .await
-                        .map_err(|e| anyhow::anyhow!(e))
+                    if self.config.read().unwrap().block_deletions {
+                        let mut queue_locked = self.upload_queue.lock().unwrap();
+                        if let Ok(queue) = queue_locked.initialized_mut() {
+                            queue.blocked_deletions.push(delete.clone());
+                        }
+                        Ok(())
+                    } else {
+                        pausable_failpoint!("before-delete-layer-pausable");
+                        self.deletion_queue_client
+                            .push_layers(
+                                self.tenant_shard_id,
+                                self.timeline_id,
+                                self.generation,
+                                delete.layers.clone(),
+                            )
+                            .await
+                            .map_err(|e| anyhow::anyhow!(e))
+                    }
                 }
                 unexpected @ UploadOp::Barrier(_) | unexpected @ UploadOp::Shutdown => {
                     // unreachable. Barrier operations are handled synchronously in
@@ -2016,8 +2110,16 @@ impl RemoteTimelineClient {
                         // Legacy mode: skip validating generation
                         upload_queue.visible_remote_consistent_lsn.store(lsn);
                         None
-                    } else {
+                    } else if self
+                        .config
+                        .read()
+                        .unwrap()
+                        .process_remote_consistent_lsn_updates
+                    {
                         Some((lsn, upload_queue.visible_remote_consistent_lsn.clone()))
+                    } else {
+                        // Our config disables remote_consistent_lsn updates: drop it.
+                        None
                     }
                 }
                 UploadOp::Delete(_) => {
@@ -2154,6 +2256,7 @@ impl RemoteTimelineClient {
                         queued_operations: VecDeque::default(),
                         #[cfg(feature = "testing")]
                         dangling_files: HashMap::default(),
+                        blocked_deletions: Vec::new(),
                         shutting_down: false,
                         shutdown_ready: Arc::new(tokio::sync::Semaphore::new(0)),
                     };
@@ -2219,6 +2322,28 @@ impl RemoteTimelineClient {
             UploadQueue::Initialized(x) => x.no_pending_work(),
         }
     }
+
+    /// 'foreign' in the sense that it does not belong to this tenant shard.  This method
+    /// is used during GC for other shards to get the index of shard zero.
+    pub(crate) async fn download_foreign_index(
+        &self,
+        shard_number: ShardNumber,
+        cancel: &CancellationToken,
+    ) -> Result<(IndexPart, Generation, std::time::SystemTime), DownloadError> {
+        let foreign_shard_id = TenantShardId {
+            shard_number,
+            shard_count: self.tenant_shard_id.shard_count,
+            tenant_id: self.tenant_shard_id.tenant_id,
+        };
+        download_index_part(
+            &self.storage_impl,
+            &foreign_shard_id,
+            &self.timeline_id,
+            Generation::MAX,
+            cancel,
+        )
+        .await
+    }
 }
 
 pub(crate) struct UploadQueueAccessor<'a> {
@@ -2367,6 +2492,7 @@ mod tests {
     use crate::{
         context::RequestContext,
         tenant::{
+            config::AttachmentMode,
             harness::{TenantHarness, TIMELINE_ID},
             storage_layer::layer::local_layer_path,
             Tenant, Timeline,
@@ -2452,6 +2578,10 @@ mod tests {
 
         /// Construct a RemoteTimelineClient in an arbitrary generation
         fn build_client(&self, generation: Generation) -> Arc<RemoteTimelineClient> {
+            let location_conf = AttachedLocationConfig {
+                generation,
+                attach_mode: AttachmentMode::Single,
+            };
             Arc::new(RemoteTimelineClient {
                 conf: self.harness.conf,
                 runtime: tokio::runtime::Handle::current(),
@@ -2465,6 +2595,7 @@ mod tests {
                     &self.harness.tenant_shard_id,
                     &TIMELINE_ID,
                 )),
+                config: std::sync::RwLock::new(RemoteTimelineClientConfig::from(&location_conf)),
                 cancel: CancellationToken::new(),
             })
         }

pageserver/src/tenant/timeline.rs

@@ -38,6 +38,7 @@ use pageserver_api::{
     shard::{ShardIdentity, ShardNumber, TenantShardId},
 };
 use rand::Rng;
+use remote_storage::DownloadError;
 use serde_with::serde_as;
 use storage_broker::BrokerClientChannel;
 use tokio::{
@@ -272,7 +273,7 @@ pub struct Timeline {
 
     /// Remote storage client.
     /// See [`remote_timeline_client`](super::remote_timeline_client) module comment for details.
-    pub remote_client: Arc<RemoteTimelineClient>,
+    pub(crate) remote_client: Arc<RemoteTimelineClient>,
 
     // What page versions do we hold in the repository? If we get a
     // request > last_record_lsn, we need to wait until we receive all
@@ -477,8 +478,31 @@ impl GcInfo {
         self.retain_lsns.sort_by_key(|i| i.0);
     }
 
-    pub(super) fn remove_child(&mut self, child_id: TimelineId) {
-        self.retain_lsns.retain(|i| i.1 != child_id);
+    pub(super) fn remove_child_maybe_offloaded(
+        &mut self,
+        child_id: TimelineId,
+        maybe_offloaded: MaybeOffloaded,
+    ) -> bool {
+        // Remove at most one element. Needed for correctness if there is two live `Timeline` objects referencing
+        // the same timeline. Shouldn't but maybe can occur when Arc's live longer than intended.
+        let mut removed = false;
+        self.retain_lsns.retain(|i| {
+            if removed {
+                return true;
+            }
+            let remove = i.1 == child_id && i.2 == maybe_offloaded;
+            removed |= remove;
+            !remove
+        });
+        removed
+    }
+
+    pub(super) fn remove_child_not_offloaded(&mut self, child_id: TimelineId) -> bool {
+        self.remove_child_maybe_offloaded(child_id, MaybeOffloaded::No)
+    }
+
+    pub(super) fn remove_child_offloaded(&mut self, child_id: TimelineId) -> bool {
+        self.remove_child_maybe_offloaded(child_id, MaybeOffloaded::Yes)
     }
 }
 
@@ -751,6 +775,21 @@ pub(crate) enum CompactFlags {
     DryRun,
 }
 
+#[serde_with::serde_as]
+#[derive(Debug, Clone, serde::Deserialize)]
+pub(crate) struct CompactRange {
+    #[serde_as(as = "serde_with::DisplayFromStr")]
+    pub start: Key,
+    #[serde_as(as = "serde_with::DisplayFromStr")]
+    pub end: Key,
+}
+
+#[derive(Clone, Default)]
+pub(crate) struct CompactOptions {
+    pub flags: EnumSet<CompactFlags>,
+    pub compact_range: Option<CompactRange>,
+}
+
 impl std::fmt::Debug for Timeline {
     fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
         write!(f, "Timeline<{}>", self.timeline_id)
@@ -1589,6 +1628,25 @@ impl Timeline {
         cancel: &CancellationToken,
         flags: EnumSet<CompactFlags>,
         ctx: &RequestContext,
+    ) -> Result<bool, CompactionError> {
+        self.compact_with_options(
+            cancel,
+            CompactOptions {
+                flags,
+                compact_range: None,
+            },
+            ctx,
+        )
+        .await
+    }
+
+    /// Outermost timeline compaction operation; downloads needed layers. Returns whether we have pending
+    /// compaction tasks.
+    pub(crate) async fn compact_with_options(
+        self: &Arc<Self>,
+        cancel: &CancellationToken,
+        options: CompactOptions,
+        ctx: &RequestContext,
     ) -> Result<bool, CompactionError> {
         // most likely the cancellation token is from background task, but in tests it could be the
         // request task as well.
@@ -1626,7 +1684,7 @@ impl Timeline {
                 self.compact_tiered(cancel, ctx).await?;
                 Ok(false)
             }
-            CompactionAlgorithm::Legacy => self.compact_legacy(cancel, flags, ctx).await,
+            CompactionAlgorithm::Legacy => self.compact_legacy(cancel, options, ctx).await,
         }
     }
 
@@ -2114,14 +2172,14 @@ impl Timeline {
             )
     }
 
-    pub(super) fn tenant_conf_updated(&self, new_conf: &TenantConfOpt) {
+    pub(super) fn tenant_conf_updated(&self, new_conf: &AttachedTenantConf) {
         // NB: Most tenant conf options are read by background loops, so,
         // changes will automatically be picked up.
 
         // The threshold is embedded in the metric. So, we need to update it.
         {
             let new_threshold = Self::get_evictions_low_residence_duration_metric_threshold(
-                new_conf,
+                &new_conf.tenant_conf,
                 &self.conf.default_tenant_conf,
             );
 
@@ -2129,6 +2187,9 @@ impl Timeline {
             let shard_id_str = format!("{}", self.tenant_shard_id.shard_slug());
 
             let timeline_id_str = self.timeline_id.to_string();
+
+            self.remote_client.update_config(&new_conf.location);
+
             self.metrics
                 .evictions_with_low_residence_duration
                 .write()
@@ -4501,7 +4562,10 @@ impl Drop for Timeline {
             // This lock should never be poisoned, but in case it is we do a .map() instead of
             // an unwrap(), to avoid panicking in a destructor and thereby aborting the process.
             if let Ok(mut gc_info) = ancestor.gc_info.write() {
-                gc_info.remove_child(self.timeline_id)
+                if !gc_info.remove_child_not_offloaded(self.timeline_id) {
+                    tracing::error!(tenant_id = %self.tenant_shard_id.tenant_id, shard_id = %self.tenant_shard_id.shard_slug(), timeline_id = %self.timeline_id,
+                        "Couldn't remove retain_lsn entry from offloaded timeline's parent: already removed");
+                }
             }
         }
     }
@@ -4761,6 +4825,86 @@ impl Timeline {
         Ok(())
     }
 
+    async fn find_gc_time_cutoff(
+        &self,
+        pitr: Duration,
+        cancel: &CancellationToken,
+        ctx: &RequestContext,
+    ) -> Result<Option<Lsn>, PageReconstructError> {
+        debug_assert_current_span_has_tenant_and_timeline_id();
+        if self.shard_identity.is_shard_zero() {
+            // Shard Zero has SLRU data and can calculate the PITR time -> LSN mapping itself
+            let now = SystemTime::now();
+            let time_range = if pitr == Duration::ZERO {
+                humantime::parse_duration(DEFAULT_PITR_INTERVAL).expect("constant is invalid")
+            } else {
+                pitr
+            };
+
+            // If PITR is so large or `now` is so small that this underflows, we will retain no history (highly unexpected case)
+            let time_cutoff = now.checked_sub(time_range).unwrap_or(now);
+            let timestamp = to_pg_timestamp(time_cutoff);
+
+            let time_cutoff = match self.find_lsn_for_timestamp(timestamp, cancel, ctx).await? {
+                LsnForTimestamp::Present(lsn) => Some(lsn),
+                LsnForTimestamp::Future(lsn) => {
+                    // The timestamp is in the future. That sounds impossible,
+                    // but what it really means is that there hasn't been
+                    // any commits since the cutoff timestamp.
+                    //
+                    // In this case we should use the LSN of the most recent commit,
+                    // which is implicitly the last LSN in the log.
+                    debug!("future({})", lsn);
+                    Some(self.get_last_record_lsn())
+                }
+                LsnForTimestamp::Past(lsn) => {
+                    debug!("past({})", lsn);
+                    None
+                }
+                LsnForTimestamp::NoData(lsn) => {
+                    debug!("nodata({})", lsn);
+                    None
+                }
+            };
+            Ok(time_cutoff)
+        } else {
+            // Shards other than shard zero cannot do timestamp->lsn lookups, and must instead learn their GC cutoff
+            // from shard zero's index.  The index doesn't explicitly tell us the time cutoff, but we may assume that
+            // the point up to which shard zero's last_gc_cutoff has advanced will either be the time cutoff, or a
+            // space cutoff that we would also have respected ourselves.
+            match self
+                .remote_client
+                .download_foreign_index(ShardNumber(0), cancel)
+                .await
+            {
+                Ok((index_part, index_generation, _index_mtime)) => {
+                    tracing::info!("GC loaded shard zero metadata (gen {index_generation:?}): latest_gc_cutoff_lsn: {}",
+                        index_part.metadata.latest_gc_cutoff_lsn());
+                    Ok(Some(index_part.metadata.latest_gc_cutoff_lsn()))
+                }
+                Err(DownloadError::NotFound) => {
+                    // This is unexpected, because during timeline creations shard zero persists to remote
+                    // storage before other shards are called, and during timeline deletion non-zeroth shards are
+                    // deleted before the zeroth one.  However, it should be harmless: if we somehow end up in this
+                    // state, then shard zero should _eventually_ write an index when it GCs.
+                    tracing::warn!("GC couldn't find shard zero's index for timeline");
+                    Ok(None)
+                }
+                Err(e) => {
+                    // TODO: this function should return a different error type than page reconstruct error
+                    Err(PageReconstructError::Other(anyhow::anyhow!(e)))
+                }
+            }
+
+            // TODO: after reading shard zero's GC cutoff, we should validate its generation with the storage
+            // controller.  Otherwise, it is possible that we see the GC cutoff go backwards while shard zero
+            // is going through a migration if we read the old location's index and it has GC'd ahead of the
+            // new location.  This is legal in principle, but problematic in practice because it might result
+            // in a timeline creation succeeding on shard zero ('s new location) but then failing on other shards
+            // because they have GC'd past the branch point.
+        }
+    }
+
     /// Find the Lsns above which layer files need to be retained on
     /// garbage collection.
     ///
@@ -4803,40 +4947,7 @@ impl Timeline {
         // - if PITR interval is set, then this is our cutoff.
         // - if PITR interval is not set, then we do a lookup
         //   based on DEFAULT_PITR_INTERVAL, so that size-based retention does not result in keeping history around permanently on idle databases.
-        let time_cutoff = {
-            let now = SystemTime::now();
-            let time_range = if pitr == Duration::ZERO {
-                humantime::parse_duration(DEFAULT_PITR_INTERVAL).expect("constant is invalid")
-            } else {
-                pitr
-            };
-
-            // If PITR is so large or `now` is so small that this underflows, we will retain no history (highly unexpected case)
-            let time_cutoff = now.checked_sub(time_range).unwrap_or(now);
-            let timestamp = to_pg_timestamp(time_cutoff);
-
-            match self.find_lsn_for_timestamp(timestamp, cancel, ctx).await? {
-                LsnForTimestamp::Present(lsn) => Some(lsn),
-                LsnForTimestamp::Future(lsn) => {
-                    // The timestamp is in the future. That sounds impossible,
-                    // but what it really means is that there hasn't been
-                    // any commits since the cutoff timestamp.
-                    //
-                    // In this case we should use the LSN of the most recent commit,
-                    // which is implicitly the last LSN in the log.
-                    debug!("future({})", lsn);
-                    Some(self.get_last_record_lsn())
-                }
-                LsnForTimestamp::Past(lsn) => {
-                    debug!("past({})", lsn);
-                    None
-                }
-                LsnForTimestamp::NoData(lsn) => {
-                    debug!("nodata({})", lsn);
-                    None
-                }
-            }
-        };
+        let time_cutoff = self.find_gc_time_cutoff(pitr, cancel, ctx).await?;
 
         Ok(match (pitr, time_cutoff) {
             (Duration::ZERO, Some(time_cutoff)) => {
@@ -5030,7 +5141,7 @@ impl Timeline {
 
             // 1. Is it newer than GC horizon cutoff point?
             if l.get_lsn_range().end > space_cutoff {
-                debug!(
+                info!(
                     "keeping {} because it's newer than space_cutoff {}",
                     l.layer_name(),
                     space_cutoff,
@@ -5041,7 +5152,7 @@ impl Timeline {
 
             // 2. It is newer than PiTR cutoff point?
             if l.get_lsn_range().end > time_cutoff {
-                debug!(
+                info!(
                     "keeping {} because it's newer than time_cutoff {}",
                     l.layer_name(),
                     time_cutoff,
@@ -5060,7 +5171,7 @@ impl Timeline {
             for retain_lsn in &retain_lsns {
                 // start_lsn is inclusive
                 if &l.get_lsn_range().start <= retain_lsn {
-                    debug!(
+                    info!(
                         "keeping {} because it's still might be referenced by child branch forked at {} is_dropped: xx is_incremental: {}",
                         l.layer_name(),
                         retain_lsn,
@@ -5075,7 +5186,7 @@ impl Timeline {
             if let Some(lsn) = &max_lsn_with_valid_lease {
                 // keep if layer start <= any of the lease
                 if &l.get_lsn_range().start <= lsn {
-                    debug!(
+                    info!(
                         "keeping {} because there is a valid lease preventing GC at {}",
                         l.layer_name(),
                         lsn,
@@ -5107,13 +5218,13 @@ impl Timeline {
             if !layers
                 .image_layer_exists(&l.get_key_range(), &(l.get_lsn_range().end..new_gc_cutoff))
             {
-                debug!("keeping {} because it is the latest layer", l.layer_name());
+                info!("keeping {} because it is the latest layer", l.layer_name());
                 result.layers_not_updated += 1;
                 continue 'outer;
             }
 
             // We didn't find any reason to keep this file, so remove it.
-            debug!(
+            info!(
                 "garbage collecting {} is_dropped: xx is_incremental: {}",
                 l.layer_name(),
                 l.is_incremental(),

pageserver/src/tenant/timeline/compaction.rs

@@ -10,7 +10,7 @@ use std::sync::Arc;
 
 use super::layer_manager::LayerManager;
 use super::{
-    CompactFlags, CreateImageLayersError, DurationRecorder, ImageLayerCreationMode,
+    CompactFlags, CompactOptions, CreateImageLayersError, DurationRecorder, ImageLayerCreationMode,
     RecordedDuration, Timeline,
 };
 
@@ -273,22 +273,32 @@ impl Timeline {
     pub(crate) async fn compact_legacy(
         self: &Arc<Self>,
         cancel: &CancellationToken,
-        flags: EnumSet<CompactFlags>,
+        options: CompactOptions,
         ctx: &RequestContext,
     ) -> Result<bool, CompactionError> {
-        if flags.contains(CompactFlags::EnhancedGcBottomMostCompaction) {
-            self.compact_with_gc(cancel, flags, ctx)
+        if options
+            .flags
+            .contains(CompactFlags::EnhancedGcBottomMostCompaction)
+        {
+            self.compact_with_gc(cancel, options, ctx)
                 .await
                 .map_err(CompactionError::Other)?;
             return Ok(false);
         }
 
-        if flags.contains(CompactFlags::DryRun) {
+        if options.flags.contains(CompactFlags::DryRun) {
             return Err(CompactionError::Other(anyhow!(
                 "dry-run mode is not supported for legacy compaction for now"
             )));
         }
 
+        if options.compact_range.is_some() {
+            // maybe useful in the future? could implement this at some point
+            return Err(CompactionError::Other(anyhow!(
+                "compaction range is not supported for legacy compaction for now"
+            )));
+        }
+
         // High level strategy for compaction / image creation:
         //
         // 1. First, calculate the desired "partitioning" of the
@@ -338,7 +348,7 @@ impl Timeline {
             .repartition(
                 self.get_last_record_lsn(),
                 self.get_compaction_target_size(),
-                flags,
+                options.flags,
                 ctx,
             )
             .await
@@ -354,7 +364,7 @@ impl Timeline {
                 let fully_compacted = self
                     .compact_level0(
                         target_file_size,
-                        flags.contains(CompactFlags::ForceL0Compaction),
+                        options.flags.contains(CompactFlags::ForceL0Compaction),
                         ctx,
                     )
                     .await?;
@@ -372,7 +382,10 @@ impl Timeline {
                         .create_image_layers(
                             &partitioning,
                             lsn,
-                            if flags.contains(CompactFlags::ForceImageLayerCreation) {
+                            if options
+                                .flags
+                                .contains(CompactFlags::ForceImageLayerCreation)
+                            {
                                 ImageLayerCreationMode::Force
                             } else {
                                 ImageLayerCreationMode::Try
@@ -1736,11 +1749,19 @@ impl Timeline {
     pub(crate) async fn compact_with_gc(
         self: &Arc<Self>,
         cancel: &CancellationToken,
-        flags: EnumSet<CompactFlags>,
+        options: CompactOptions,
         ctx: &RequestContext,
     ) -> anyhow::Result<()> {
-        self.partial_compact_with_gc(Key::MIN..Key::MAX, cancel, flags, ctx)
-            .await
+        self.partial_compact_with_gc(
+            options
+                .compact_range
+                .map(|range| range.start..range.end)
+                .unwrap_or_else(|| Key::MIN..Key::MAX),
+            cancel,
+            options.flags,
+            ctx,
+        )
+        .await
     }
 
     /// An experimental compaction building block that combines compaction with garbage collection.
@@ -2021,6 +2042,14 @@ impl Timeline {
             if cancel.is_cancelled() {
                 return Err(anyhow!("cancelled")); // TODO: refactor to CompactionError and pass cancel error
             }
+            if self.shard_identity.is_key_disposable(&key) {
+                // If this shard does not need to store this key, simply skip it.
+                //
+                // This is not handled in the filter iterator because shard is determined by hash.
+                // Therefore, it does not give us any performance benefit to do things like skip
+                // a whole layer file as handling key spaces (ranges).
+                continue;
+            }
             if !job_desc.compaction_key_range.contains(&key) {
                 if !desc.is_delta {
                     continue;

pageserver/src/tenant/timeline/delete.rs

@@ -5,6 +5,7 @@ use std::{
 
 use anyhow::Context;
 use pageserver_api::{models::TimelineState, shard::TenantShardId};
+use remote_storage::DownloadError;
 use tokio::sync::OwnedMutexGuard;
 use tracing::{error, info, info_span, instrument, Instrument};
 use utils::{crashsafe, fs_ext, id::TimelineId, pausable_failpoint};
@@ -16,7 +17,7 @@ use crate::{
         metadata::TimelineMetadata,
         remote_timeline_client::{PersistIndexPartWithDeletedFlagError, RemoteTimelineClient},
         CreateTimelineCause, DeleteTimelineError, MaybeDeletedIndexPart, Tenant,
-        TimelineOrOffloaded,
+        TenantManifestError, TimelineOrOffloaded,
     },
     virtual_file::MaybeFatalIo,
 };
@@ -110,13 +111,6 @@ pub(super) async fn delete_local_timeline_directory(
     info!("finished deleting layer files, releasing locks");
 }
 
-/// Removes remote layers and an index file after them.
-async fn delete_remote_layers_and_index(
-    remote_client: &Arc<RemoteTimelineClient>,
-) -> anyhow::Result<()> {
-    remote_client.delete_all().await.context("delete_all")
-}
-
 /// It is important that this gets called when DeletionGuard is being held.
 /// For more context see comments in [`DeleteTimelineFlow::prepare`]
 async fn remove_maybe_offloaded_timeline_from_tenant(
@@ -147,9 +141,10 @@ async fn remove_maybe_offloaded_timeline_from_tenant(
             );
         }
         TimelineOrOffloaded::Offloaded(timeline) => {
-            timelines_offloaded
+            let offloaded_timeline = timelines_offloaded
                 .remove(&timeline.timeline_id)
                 .expect("timeline that we were deleting was concurrently removed from 'timelines_offloaded' map");
+            offloaded_timeline.delete_from_ancestor_with_timelines(&timelines);
         }
     }
 
@@ -221,11 +216,24 @@ impl DeleteTimelineFlow {
             None => {
                 let remote_client = tenant
                     .build_timeline_client(timeline.timeline_id(), tenant.remote_storage.clone());
-                let result = remote_client
+                let result = match remote_client
                     .download_index_file(&tenant.cancel)
                     .instrument(info_span!("download_index_file"))
                     .await
-                    .map_err(|e| DeleteTimelineError::Other(anyhow::anyhow!("error: {:?}", e)))?;
+                {
+                    Ok(r) => r,
+                    Err(DownloadError::NotFound) => {
+                        // Deletion is already complete
+                        tracing::info!("Timeline already deleted in remote storage");
+                        return Ok(());
+                    }
+                    Err(e) => {
+                        return Err(DeleteTimelineError::Other(anyhow::anyhow!(
+                            "error: {:?}",
+                            e
+                        )));
+                    }
+                };
                 let index_part = match result {
                     MaybeDeletedIndexPart::Deleted(p) => {
                         tracing::info!("Timeline already set as deleted in remote index");
@@ -275,7 +283,7 @@ impl DeleteTimelineFlow {
 
     /// Shortcut to create Timeline in stopping state and spawn deletion task.
     #[instrument(skip_all, fields(%timeline_id))]
-    pub async fn resume_deletion(
+    pub(crate) async fn resume_deletion(
         tenant: Arc<Tenant>,
         timeline_id: TimelineId,
         local_metadata: &TimelineMetadata,
@@ -406,7 +414,12 @@ impl DeleteTimelineFlow {
             "timeline_delete",
             async move {
                 if let Err(err) = Self::background(guard, conf, &tenant, &timeline, remote_client).await {
-                    error!("Error: {err:#}");
+                    // Only log as an error if it's not a cancellation.
+                    if matches!(err, DeleteTimelineError::Cancelled) {
+                        info!("Shutdown during timeline deletion");
+                    }else {
+                        error!("Error: {err:#}");
+                    }
                     if let TimelineOrOffloaded::Timeline(timeline) = timeline {
                         timeline.set_broken(format!("{err:#}"))
                     }
@@ -438,7 +451,7 @@ impl DeleteTimelineFlow {
             Err(anyhow::anyhow!("failpoint: timeline-delete-after-rm"))?
         });
 
-        delete_remote_layers_and_index(&remote_client).await?;
+        remote_client.delete_all().await?;
 
         pausable_failpoint!("in_progress_delete");
 
@@ -449,10 +462,10 @@ impl DeleteTimelineFlow {
         // So indeed, the tenant manifest might refer to an offloaded timeline which has already been deleted.
         // However, we handle this case in tenant loading code so the next time we attach, the issue is
         // resolved.
-        tenant
-            .store_tenant_manifest()
-            .await
-            .map_err(|e| DeleteTimelineError::Other(anyhow::anyhow!(e)))?;
+        tenant.store_tenant_manifest().await.map_err(|e| match e {
+            TenantManifestError::Cancelled => DeleteTimelineError::Cancelled,
+            _ => DeleteTimelineError::Other(e.into()),
+        })?;
 
         *guard = Self::Finished;
 

pageserver/src/tenant/timeline/offload.rs

@@ -66,7 +66,7 @@ pub(crate) async fn offload_timeline(
     let conf = &tenant.conf;
     delete_local_timeline_directory(conf, tenant.tenant_shard_id, &timeline).await;
 
-    remove_timeline_from_tenant(tenant, &timeline, &guard);
+    let remaining_refcount = remove_timeline_from_tenant(tenant, &timeline, &guard);
 
     {
         let mut offloaded_timelines = tenant.timelines_offloaded.lock().unwrap();
@@ -87,16 +87,20 @@ pub(crate) async fn offload_timeline(
     // not our actual state of offloaded timelines.
     tenant.store_tenant_manifest().await?;
 
+    tracing::info!("Timeline offload complete (remaining arc refcount: {remaining_refcount})");
+
     Ok(())
 }
 
 /// It is important that this gets called when DeletionGuard is being held.
 /// For more context see comments in [`DeleteTimelineFlow::prepare`]
+///
+/// Returns the strong count of the timeline `Arc`
 fn remove_timeline_from_tenant(
     tenant: &Tenant,
     timeline: &Timeline,
     _: &DeletionGuard, // using it as a witness
-) {
+) -> usize {
     // Remove the timeline from the map.
     let mut timelines = tenant.timelines.lock().unwrap();
     let children_exist = timelines
@@ -109,7 +113,9 @@ fn remove_timeline_from_tenant(
         panic!("Timeline grew children while we removed layer files");
     }
 
-    timelines
+    let timeline = timelines
         .remove(&timeline.timeline_id)
         .expect("timeline that we were deleting was concurrently removed from 'timelines' map");
+
+    Arc::strong_count(&timeline)
 }

pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs

@@ -331,11 +331,11 @@ pub(super) async fn handle_walreceiver_connection(
                         Ok(())
                     }
 
-                    while let Some((record_end_lsn, recdata)) = waldecoder.poll_decode()? {
+                    while let Some((next_record_lsn, recdata)) = waldecoder.poll_decode()? {
                         // It is important to deal with the aligned records as lsn in getPage@LSN is
                         // aligned and can be several bytes bigger. Without this alignment we are
                         // at risk of hitting a deadlock.
-                        if !record_end_lsn.is_aligned() {
+                        if !next_record_lsn.is_aligned() {
                             return Err(WalReceiverError::Other(anyhow!("LSN not aligned")));
                         }
 
@@ -343,7 +343,7 @@ pub(super) async fn handle_walreceiver_connection(
                         let interpreted = InterpretedWalRecord::from_bytes_filtered(
                             recdata,
                             modification.tline.get_shard_identity(),
-                            record_end_lsn,
+                            next_record_lsn,
                             modification.tline.pg_version,
                         )?;
 
@@ -367,10 +367,10 @@ pub(super) async fn handle_walreceiver_connection(
                             .ingest_record(interpreted, &mut modification, &ctx)
                             .await
                             .with_context(|| {
-                                format!("could not ingest record at {record_end_lsn}")
+                                format!("could not ingest record at {next_record_lsn}")
                             })?;
                         if !ingested {
-                            tracing::debug!("ingest: filtered out record @ LSN {record_end_lsn}");
+                            tracing::debug!("ingest: filtered out record @ LSN {next_record_lsn}");
                             WAL_INGEST.records_filtered.inc();
                             filtered_records += 1;
                         }
@@ -380,7 +380,7 @@ pub(super) async fn handle_walreceiver_connection(
                         // to timeout the tests.
                         fail_point!("walreceiver-after-ingest");
 
-                        last_rec_lsn = record_end_lsn;
+                        last_rec_lsn = next_record_lsn;
 
                         // Commit every ingest_batch_size records. Even if we filtered out
                         // all records, we still need to call commit to advance the LSN.

pageserver/src/tenant/upload_queue.rs

@@ -88,6 +88,9 @@ pub(crate) struct UploadQueueInitialized {
     #[cfg(feature = "testing")]
     pub(crate) dangling_files: HashMap<LayerName, Generation>,
 
+    /// Deletions that are blocked by the tenant configuration
+    pub(crate) blocked_deletions: Vec<Delete>,
+
     /// Set to true when we have inserted the `UploadOp::Shutdown` into the `inprogress_tasks`.
     pub(crate) shutting_down: bool,
 
@@ -180,6 +183,7 @@ impl UploadQueue {
             queued_operations: VecDeque::new(),
             #[cfg(feature = "testing")]
             dangling_files: HashMap::new(),
+            blocked_deletions: Vec::new(),
             shutting_down: false,
             shutdown_ready: Arc::new(tokio::sync::Semaphore::new(0)),
         };
@@ -220,6 +224,7 @@ impl UploadQueue {
             queued_operations: VecDeque::new(),
             #[cfg(feature = "testing")]
             dangling_files: HashMap::new(),
+            blocked_deletions: Vec::new(),
             shutting_down: false,
             shutdown_ready: Arc::new(tokio::sync::Semaphore::new(0)),
         };
@@ -270,7 +275,7 @@ pub(crate) struct UploadTask {
 
 /// A deletion of some layers within the lifetime of a timeline.  This is not used
 /// for timeline deletion, which skips this queue and goes directly to DeletionQueue.
-#[derive(Debug)]
+#[derive(Debug, Clone)]
 pub(crate) struct Delete {
     pub(crate) layers: Vec<(LayerName, LayerFileMetadata)>,
 }

pageserver/src/virtual_file.rs

@@ -175,10 +175,16 @@ impl VirtualFile {
     }
 
     pub async fn sync_all(&self) -> Result<(), Error> {
+        if SYNC_MODE.load(std::sync::atomic::Ordering::Relaxed) == SyncMode::UnsafeNoSync as u8 {
+            return Ok(());
+        }
         self.inner.sync_all().await
     }
 
     pub async fn sync_data(&self) -> Result<(), Error> {
+        if SYNC_MODE.load(std::sync::atomic::Ordering::Relaxed) == SyncMode::UnsafeNoSync as u8 {
+            return Ok(());
+        }
         self.inner.sync_data().await
     }
 
@@ -233,6 +239,27 @@ impl VirtualFile {
     }
 }
 
+/// Indicates whether to enable fsync, fdatasync, or O_SYNC/O_DSYNC when writing
+/// files. Switching this off is unsafe and only used for testing on machines
+/// with slow drives.
+#[repr(u8)]
+pub enum SyncMode {
+    Sync,
+    UnsafeNoSync,
+}
+
+impl TryFrom<u8> for SyncMode {
+    type Error = u8;
+
+    fn try_from(value: u8) -> Result<Self, Self::Error> {
+        Ok(match value {
+            v if v == (SyncMode::Sync as u8) => SyncMode::Sync,
+            v if v == (SyncMode::UnsafeNoSync as u8) => SyncMode::UnsafeNoSync,
+            x => return Err(x),
+        })
+    }
+}
+
 ///
 /// A virtual file descriptor. You can use this just like std::fs::File, but internally
 /// the underlying file is closed if the system is low on file descriptors,
@@ -1332,12 +1359,13 @@ impl OpenFiles {
 /// server startup.
 ///
 #[cfg(not(test))]
-pub fn init(num_slots: usize, engine: IoEngineKind, mode: IoMode) {
+pub fn init(num_slots: usize, engine: IoEngineKind, mode: IoMode, sync_mode: SyncMode) {
     if OPEN_FILES.set(OpenFiles::new(num_slots)).is_err() {
         panic!("virtual_file::init called twice");
     }
     set_io_mode(mode);
     io_engine::init(engine);
+    SYNC_MODE.store(sync_mode as u8, std::sync::atomic::Ordering::Relaxed);
     crate::metrics::virtual_file_descriptor_cache::SIZE_MAX.set(num_slots as u64);
 }
 
@@ -1379,6 +1407,9 @@ pub(crate) fn set_io_mode(mode: IoMode) {
 pub(crate) fn get_io_mode() -> IoMode {
     IoMode::try_from(IO_MODE.load(Ordering::Relaxed)).unwrap()
 }
+
+static SYNC_MODE: AtomicU8 = AtomicU8::new(SyncMode::Sync as u8);
+
 #[cfg(test)]
 mod tests {
     use crate::context::DownloadBehavior;

pageserver/src/walingest.rs

@@ -154,7 +154,7 @@ impl WalIngest {
         WAL_INGEST.records_received.inc();
         let prev_len = modification.len();
 
-        modification.set_lsn(interpreted.end_lsn)?;
+        modification.set_lsn(interpreted.next_record_lsn)?;
 
         if matches!(interpreted.flush_uncommitted, FlushUncommittedRecords::Yes) {
             // Records of this type should always be preceded by a commit(), as they
@@ -587,11 +587,29 @@ impl WalIngest {
                 forknum: VISIBILITYMAP_FORKNUM,
             };
 
-            let mut vm_page_no = blkno / pg_constants::VM_HEAPBLOCKS_PER_PAGE;
-            if blkno % pg_constants::VM_HEAPBLOCKS_PER_PAGE != 0 {
-                // Tail of last remaining vm page has to be zeroed.
-                // We are not precise here and instead of digging in VM bitmap format just clear the whole page.
-                modification.put_rel_page_image_zero(rel, vm_page_no)?;
+            // last remaining block, byte, and bit
+            let mut vm_page_no = blkno / (pg_constants::VM_HEAPBLOCKS_PER_PAGE as u32);
+            let trunc_byte = blkno as usize % pg_constants::VM_HEAPBLOCKS_PER_PAGE
+                / pg_constants::VM_HEAPBLOCKS_PER_BYTE;
+            let trunc_offs = blkno as usize % pg_constants::VM_HEAPBLOCKS_PER_BYTE
+                * pg_constants::VM_BITS_PER_HEAPBLOCK;
+
+            // Unless the new size is exactly at a visibility map page boundary, the
+            // tail bits in the last remaining map page, representing truncated heap
+            // blocks, need to be cleared. This is not only tidy, but also necessary
+            // because we don't get a chance to clear the bits if the heap is extended
+            // again.
+            if (trunc_byte != 0 || trunc_offs != 0)
+                && self.shard.is_key_local(&rel_block_to_key(rel, vm_page_no))
+            {
+                modification.put_rel_wal_record(
+                    rel,
+                    vm_page_no,
+                    NeonWalRecord::TruncateVisibilityMap {
+                        trunc_byte,
+                        trunc_offs,
+                    },
+                )?;
                 vm_page_no += 1;
             }
             let nblocks = get_relsize(modification, rel, ctx).await?;
@@ -1510,6 +1528,11 @@ mod tests {
 
         assert_current_logical_size(&tline, Lsn(0x50));
 
+        let test_span = tracing::info_span!(parent: None, "test",
+                                            tenant_id=%tline.tenant_shard_id.tenant_id,
+                                            shard_id=%tline.tenant_shard_id.shard_slug(),
+                                            timeline_id=%tline.timeline_id);
+
         // The relation was created at LSN 2, not visible at LSN 1 yet.
         assert_eq!(
             tline
@@ -1544,6 +1567,7 @@ mod tests {
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x20)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 0 at 2")
         );
@@ -1551,6 +1575,7 @@ mod tests {
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x30)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 0 at 3")
         );
@@ -1558,12 +1583,14 @@ mod tests {
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x40)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 0 at 3")
         );
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 1, Version::Lsn(Lsn(0x40)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 1 at 4")
         );
@@ -1571,18 +1598,21 @@ mod tests {
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x50)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 0 at 3")
         );
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 1, Version::Lsn(Lsn(0x50)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 1 at 4")
         );
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 2, Version::Lsn(Lsn(0x50)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 2 at 5")
         );
@@ -1605,12 +1635,14 @@ mod tests {
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x60)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 0 at 3")
         );
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 1, Version::Lsn(Lsn(0x60)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 1 at 4")
         );
@@ -1625,6 +1657,7 @@ mod tests {
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 2, Version::Lsn(Lsn(0x50)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 2 at 5")
         );
@@ -1657,12 +1690,14 @@ mod tests {
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x70)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             ZERO_PAGE
         );
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 1, Version::Lsn(Lsn(0x70)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 1")
         );
@@ -1683,6 +1718,7 @@ mod tests {
             assert_eq!(
                 tline
                     .get_rel_page_at_lsn(TESTREL_A, blk, Version::Lsn(Lsn(0x80)), &ctx)
+                    .instrument(test_span.clone())
                     .await?,
                 ZERO_PAGE
             );
@@ -1690,6 +1726,7 @@ mod tests {
         assert_eq!(
             tline
                 .get_rel_page_at_lsn(TESTREL_A, 1500, Version::Lsn(Lsn(0x80)), &ctx)
+                .instrument(test_span.clone())
                 .await?,
             test_img("foo blk 1500")
         );
@@ -1797,6 +1834,11 @@ mod tests {
         }
         m.commit(&ctx).await?;
 
+        let test_span = tracing::info_span!(parent: None, "test",
+                                            tenant_id=%tline.tenant_shard_id.tenant_id,
+                                            shard_id=%tline.tenant_shard_id.shard_slug(),
+                                            timeline_id=%tline.timeline_id);
+
         // The relation was created at LSN 20, not visible at LSN 1 yet.
         assert_eq!(
             tline
@@ -1829,6 +1871,7 @@ mod tests {
             assert_eq!(
                 tline
                     .get_rel_page_at_lsn(TESTREL_A, blkno, Version::Lsn(lsn), &ctx)
+                    .instrument(test_span.clone())
                     .await?,
                 test_img(&data)
             );
@@ -1856,6 +1899,7 @@ mod tests {
             assert_eq!(
                 tline
                     .get_rel_page_at_lsn(TESTREL_A, blkno, Version::Lsn(Lsn(0x60)), &ctx)
+                    .instrument(test_span.clone())
                     .await?,
                 test_img(&data)
             );
@@ -1874,6 +1918,7 @@ mod tests {
             assert_eq!(
                 tline
                     .get_rel_page_at_lsn(TESTREL_A, blkno, Version::Lsn(Lsn(0x50)), &ctx)
+                    .instrument(test_span.clone())
                     .await?,
                 test_img(&data)
             );
@@ -1910,6 +1955,7 @@ mod tests {
             assert_eq!(
                 tline
                     .get_rel_page_at_lsn(TESTREL_A, blkno, Version::Lsn(Lsn(0x80)), &ctx)
+                    .instrument(test_span.clone())
                     .await?,
                 test_img(&data)
             );

pageserver/src/walredo/apply_neon.rs

@@ -42,6 +42,34 @@ pub(crate) fn apply_in_neon(
         } => {
             anyhow::bail!("tried to pass postgres wal record to neon WAL redo");
         }
+        //
+        // Code copied from PostgreSQL `visibilitymap_prepare_truncate` function in `visibilitymap.c`
+        //
+        NeonWalRecord::TruncateVisibilityMap {
+            trunc_byte,
+            trunc_offs,
+        } => {
+            // sanity check that this is modifying the correct relation
+            let (rel, _) = key.to_rel_block().context("invalid record")?;
+            assert!(
+                rel.forknum == VISIBILITYMAP_FORKNUM,
+                "TruncateVisibilityMap record on unexpected rel {}",
+                rel
+            );
+            let map = &mut page[pg_constants::MAXALIGN_SIZE_OF_PAGE_HEADER_DATA..];
+            map[*trunc_byte + 1..].fill(0u8);
+            /*----
+             * Mask out the unwanted bits of the last remaining byte.
+             *
+             * ((1 << 0) - 1) = 00000000
+             * ((1 << 1) - 1) = 00000001
+             * ...
+             * ((1 << 6) - 1) = 00111111
+             * ((1 << 7) - 1) = 01111111
+             *----
+             */
+            map[*trunc_byte] &= (1 << *trunc_offs) - 1;
+        }
         NeonWalRecord::ClearVisibilityMapFlags {
             new_heap_blkno,
             old_heap_blkno,

pgxn/neon/neon.c

@@ -421,9 +421,7 @@ _PG_init(void)
 
 	pg_init_libpagestore();
 	pg_init_walproposer();
-	WalSender_Custom_XLogReaderRoutines = NeonOnDemandXLogReaderRoutines;
-	LogicalFuncs_Custom_XLogReaderRoutines = NeonOnDemandXLogReaderRoutines;
-	SlotFuncs_Custom_XLogReaderRoutines = NeonOnDemandXLogReaderRoutines;
+	Custom_XLogReaderRoutines = NeonOnDemandXLogReaderRoutines;
 
 	InitUnstableExtensionsSupport();
 	InitLogicalReplicationMonitor();
@@ -512,7 +510,7 @@ neon_shmem_startup_hook(void)
 	if (prev_shmem_startup_hook)
 		prev_shmem_startup_hook();
 
-#if PG_PG_MAJORVERSION_NUM >= 17
+#if PG_MAJORVERSION_NUM >= 17
 	WAIT_EVENT_NEON_LFC_MAINTENANCE = WaitEventExtensionNew("Neon/FileCache_Maintenance");
 	WAIT_EVENT_NEON_LFC_READ = WaitEventExtensionNew("Neon/FileCache_Read");
 	WAIT_EVENT_NEON_LFC_TRUNCATE = WaitEventExtensionNew("Neon/FileCache_Truncate");

pgxn/neon_walredo/walredoproc.c

@@ -453,7 +453,6 @@ WalRedoMain(int argc, char *argv[])
 static void
 CreateFakeSharedMemoryAndSemaphores(void)
 {
-	PGShmemHeader *shim = NULL;
 	PGShmemHeader *hdr;
 	Size		size;
 	int			numSemas;
@@ -486,7 +485,6 @@ CreateFakeSharedMemoryAndSemaphores(void)
 		hdr->totalsize = size;
 		hdr->freeoffset = MAXALIGN(sizeof(PGShmemHeader));
 
-		shim = hdr;
 		UsedShmemSegAddr = hdr;
 		UsedShmemSegID = (unsigned long) 42; /* not relevant for non-shared memory */
 	}
@@ -523,8 +521,6 @@ CreateFakeSharedMemoryAndSemaphores(void)
 	 */
 	InitShmemIndex();
 
-	dsm_shmem_init();
-
 	/*
 	 * Set up xlog, clog, and buffers
 	 */
@@ -599,10 +595,6 @@ CreateFakeSharedMemoryAndSemaphores(void)
 		ShmemBackendArrayAllocation();
 #endif
 
-	/* Initialize dynamic shared memory facilities. */
-	if (!IsUnderPostmaster)
-		dsm_postmaster_startup(shim);
-
 	/*
 	 * Now give loadable modules a chance to set up their shmem allocations
 	 */

poetry.lock

@@ -13,97 +13,112 @@ files = [
 
 [[package]]
 name = "aiohttp"
-version = "3.10.2"
+version = "3.10.11"
 description = "Async http client/server framework (asyncio)"
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "aiohttp-3.10.2-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:95213b3d79c7e387144e9cb7b9d2809092d6ff2c044cb59033aedc612f38fb6d"},
-    {file = "aiohttp-3.10.2-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:1aa005f060aff7124cfadaa2493f00a4e28ed41b232add5869e129a2e395935a"},
-    {file = "aiohttp-3.10.2-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:eabe6bf4c199687592f5de4ccd383945f485779c7ffb62a9b9f1f8a3f9756df8"},
-    {file = "aiohttp-3.10.2-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:96e010736fc16d21125c7e2dc5c350cd43c528b85085c04bf73a77be328fe944"},
-    {file = "aiohttp-3.10.2-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:99f81f9c1529fd8e03be4a7bd7df32d14b4f856e90ef6e9cbad3415dbfa9166c"},
-    {file = "aiohttp-3.10.2-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:d611d1a01c25277bcdea06879afbc11472e33ce842322496b211319aa95441bb"},
-    {file = "aiohttp-3.10.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:e00191d38156e09e8c81ef3d75c0d70d4f209b8381e71622165f22ef7da6f101"},
-    {file = "aiohttp-3.10.2-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:74c091a5ded6cb81785de2d7a8ab703731f26de910dbe0f3934eabef4ae417cc"},
-    {file = "aiohttp-3.10.2-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:18186a80ec5a701816adbf1d779926e1069392cf18504528d6e52e14b5920525"},
-    {file = "aiohttp-3.10.2-cp310-cp310-musllinux_1_2_i686.whl", hash = "sha256:5a7ceb2a0d2280f23a02c64cd0afdc922079bb950400c3dd13a1ab2988428aac"},
-    {file = "aiohttp-3.10.2-cp310-cp310-musllinux_1_2_ppc64le.whl", hash = "sha256:8bd7be6ff6c162a60cb8fce65ee879a684fbb63d5466aba3fa5b9288eb04aefa"},
-    {file = "aiohttp-3.10.2-cp310-cp310-musllinux_1_2_s390x.whl", hash = "sha256:fae962b62944eaebff4f4fddcf1a69de919e7b967136a318533d82d93c3c6bd1"},
-    {file = "aiohttp-3.10.2-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:a0fde16d284efcacbe15fb0c1013f0967b6c3e379649239d783868230bf1db42"},
-    {file = "aiohttp-3.10.2-cp310-cp310-win32.whl", hash = "sha256:f81cd85a0e76ec7b8e2b6636fe02952d35befda4196b8c88f3cec5b4fb512839"},
-    {file = "aiohttp-3.10.2-cp310-cp310-win_amd64.whl", hash = "sha256:54ba10eb5a3481c28282eb6afb5f709aedf53cf9c3a31875ffbdc9fc719ffd67"},
-    {file = "aiohttp-3.10.2-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:87fab7f948e407444c2f57088286e00e2ed0003ceaf3d8f8cc0f60544ba61d91"},
-    {file = "aiohttp-3.10.2-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:ec6ad66ed660d46503243cbec7b2b3d8ddfa020f984209b3b8ef7d98ce69c3f2"},
-    {file = "aiohttp-3.10.2-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:a4be88807283bd96ae7b8e401abde4ca0bab597ba73b5e9a2d98f36d451e9aac"},
-    {file = "aiohttp-3.10.2-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:01c98041f90927c2cbd72c22a164bb816fa3010a047d264969cf82e1d4bcf8d1"},
-    {file = "aiohttp-3.10.2-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:54e36c67e1a9273ecafab18d6693da0fb5ac48fd48417e4548ac24a918c20998"},
-    {file = "aiohttp-3.10.2-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:7de3ddb6f424af54535424082a1b5d1ae8caf8256ebd445be68c31c662354720"},
-    {file = "aiohttp-3.10.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:7dd9c7db94b4692b827ce51dcee597d61a0e4f4661162424faf65106775b40e7"},
-    {file = "aiohttp-3.10.2-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:e57e21e1167705f8482ca29cc5d02702208d8bf4aff58f766d94bcd6ead838cd"},
-    {file = "aiohttp-3.10.2-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:a1a50e59b720060c29e2951fd9f13c01e1ea9492e5a527b92cfe04dd64453c16"},
-    {file = "aiohttp-3.10.2-cp311-cp311-musllinux_1_2_i686.whl", hash = "sha256:686c87782481fda5ee6ba572d912a5c26d9f98cc5c243ebd03f95222af3f1b0f"},
-    {file = "aiohttp-3.10.2-cp311-cp311-musllinux_1_2_ppc64le.whl", hash = "sha256:dafb4abb257c0ed56dc36f4e928a7341b34b1379bd87e5a15ce5d883c2c90574"},
-    {file = "aiohttp-3.10.2-cp311-cp311-musllinux_1_2_s390x.whl", hash = "sha256:494a6f77560e02bd7d1ab579fdf8192390567fc96a603f21370f6e63690b7f3d"},
-    {file = "aiohttp-3.10.2-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:6fe8503b1b917508cc68bf44dae28823ac05e9f091021e0c41f806ebbb23f92f"},
-    {file = "aiohttp-3.10.2-cp311-cp311-win32.whl", hash = "sha256:4ddb43d06ce786221c0dfd3c91b4892c318eaa36b903f7c4278e7e2fa0dd5102"},
-    {file = "aiohttp-3.10.2-cp311-cp311-win_amd64.whl", hash = "sha256:ca2f5abcb0a9a47e56bac173c01e9f6c6e7f27534d91451c5f22e6a35a5a2093"},
-    {file = "aiohttp-3.10.2-cp312-cp312-macosx_10_9_universal2.whl", hash = "sha256:14eb6b17f6246959fb0b035d4f4ae52caa870c4edfb6170aad14c0de5bfbf478"},
-    {file = "aiohttp-3.10.2-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:465e445ec348d4e4bd349edd8b22db75f025da9d7b6dc1369c48e7935b85581e"},
-    {file = "aiohttp-3.10.2-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:341f8ece0276a828d95b70cd265d20e257f5132b46bf77d759d7f4e0443f2906"},
-    {file = "aiohttp-3.10.2-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:c01fbb87b5426381cd9418b3ddcf4fc107e296fa2d3446c18ce6c76642f340a3"},
-    {file = "aiohttp-3.10.2-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:2c474af073e1a6763e1c5522bbb2d85ff8318197e4c6c919b8d7886e16213345"},
-    {file = "aiohttp-3.10.2-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:d9076810a5621236e29b2204e67a68e1fe317c8727ee4c9abbfbb1083b442c38"},
-    {file = "aiohttp-3.10.2-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:e8f515d6859e673940e08de3922b9c4a2249653b0ac181169313bd6e4b1978ac"},
-    {file = "aiohttp-3.10.2-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:655e583afc639bef06f3b2446972c1726007a21003cd0ef57116a123e44601bc"},
-    {file = "aiohttp-3.10.2-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:8da9449a575133828cc99985536552ea2dcd690e848f9d41b48d8853a149a959"},
-    {file = "aiohttp-3.10.2-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:19073d57d0feb1865d12361e2a1f5a49cb764bf81a4024a3b608ab521568093a"},
-    {file = "aiohttp-3.10.2-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:c8e98e1845805f184d91fda6f9ab93d7c7b0dddf1c07e0255924bfdb151a8d05"},
-    {file = "aiohttp-3.10.2-cp312-cp312-musllinux_1_2_s390x.whl", hash = "sha256:377220a5efde6f9497c5b74649b8c261d3cce8a84cb661be2ed8099a2196400a"},
-    {file = "aiohttp-3.10.2-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:92f7f4a4dc9cdb5980973a74d43cdbb16286dacf8d1896b6c3023b8ba8436f8e"},
-    {file = "aiohttp-3.10.2-cp312-cp312-win32.whl", hash = "sha256:9bb2834a6f11d65374ce97d366d6311a9155ef92c4f0cee543b2155d06dc921f"},
-    {file = "aiohttp-3.10.2-cp312-cp312-win_amd64.whl", hash = "sha256:518dc3cb37365255708283d1c1c54485bbacccd84f0a0fb87ed8917ba45eda5b"},
-    {file = "aiohttp-3.10.2-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:7f98e70bbbf693086efe4b86d381efad8edac040b8ad02821453083d15ec315f"},
-    {file = "aiohttp-3.10.2-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:9f6f0b252a009e98fe84028a4ec48396a948e7a65b8be06ccfc6ef68cf1f614d"},
-    {file = "aiohttp-3.10.2-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:9360e3ffc7b23565600e729e8c639c3c50d5520e05fdf94aa2bd859eef12c407"},
-    {file = "aiohttp-3.10.2-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:3988044d1635c7821dd44f0edfbe47e9875427464e59d548aece447f8c22800a"},
-    {file = "aiohttp-3.10.2-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:30a9d59da1543a6f1478c3436fd49ec59be3868bca561a33778b4391005e499d"},
-    {file = "aiohttp-3.10.2-cp38-cp38-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:f9f49bdb94809ac56e09a310a62f33e5f22973d6fd351aac72a39cd551e98194"},
-    {file = "aiohttp-3.10.2-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ddfd2dca3f11c365d6857a07e7d12985afc59798458a2fdb2ffa4a0332a3fd43"},
-    {file = "aiohttp-3.10.2-cp38-cp38-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:685c1508ec97b2cd3e120bfe309a4ff8e852e8a7460f1ef1de00c2c0ed01e33c"},
-    {file = "aiohttp-3.10.2-cp38-cp38-musllinux_1_2_aarch64.whl", hash = "sha256:49904f38667c44c041a0b44c474b3ae36948d16a0398a8f8cd84e2bb3c42a069"},
-    {file = "aiohttp-3.10.2-cp38-cp38-musllinux_1_2_i686.whl", hash = "sha256:352f3a4e5f11f3241a49b6a48bc5b935fabc35d1165fa0d87f3ca99c1fcca98b"},
-    {file = "aiohttp-3.10.2-cp38-cp38-musllinux_1_2_ppc64le.whl", hash = "sha256:fc61f39b534c5d5903490478a0dd349df397d2284a939aa3cbaa2fb7a19b8397"},
-    {file = "aiohttp-3.10.2-cp38-cp38-musllinux_1_2_s390x.whl", hash = "sha256:ad2274e707be37420d0b6c3d26a8115295fe9d8e6e530fa6a42487a8ca3ad052"},
-    {file = "aiohttp-3.10.2-cp38-cp38-musllinux_1_2_x86_64.whl", hash = "sha256:c836bf3c7512100219fe1123743fd8dd9a2b50dd7cfb0c3bb10d041309acab4b"},
-    {file = "aiohttp-3.10.2-cp38-cp38-win32.whl", hash = "sha256:53e8898adda402be03ff164b0878abe2d884e3ea03a4701e6ad55399d84b92dc"},
-    {file = "aiohttp-3.10.2-cp38-cp38-win_amd64.whl", hash = "sha256:7cc8f65f5b22304693de05a245b6736b14cb5bc9c8a03da6e2ae9ef15f8b458f"},
-    {file = "aiohttp-3.10.2-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:9dfc906d656e14004c5bc672399c1cccc10db38df2b62a13fb2b6e165a81c316"},
-    {file = "aiohttp-3.10.2-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:91b10208b222ddf655c3a3d5b727879d7163db12b634492df41a9182a76edaae"},
-    {file = "aiohttp-3.10.2-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:9fd16b5e1a7bdd14668cd6bde60a2a29b49147a535c74f50d8177d11b38433a7"},
-    {file = "aiohttp-3.10.2-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:b2bfdda4971bd79201f59adbad24ec2728875237e1c83bba5221284dbbf57bda"},
-    {file = "aiohttp-3.10.2-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:69d73f869cf29e8a373127fc378014e2b17bcfbe8d89134bc6fb06a2f67f3cb3"},
-    {file = "aiohttp-3.10.2-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:df59f8486507c421c0620a2c3dce81fbf1d54018dc20ff4fecdb2c106d6e6abc"},
-    {file = "aiohttp-3.10.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:0df930015db36b460aa9badbf35eccbc383f00d52d4b6f3de2ccb57d064a6ade"},
-    {file = "aiohttp-3.10.2-cp39-cp39-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:562b1153ab7f766ee6b8b357ec777a302770ad017cf18505d34f1c088fccc448"},
-    {file = "aiohttp-3.10.2-cp39-cp39-musllinux_1_2_aarch64.whl", hash = "sha256:d984db6d855de58e0fde1ef908d48fe9a634cadb3cf715962722b4da1c40619d"},
-    {file = "aiohttp-3.10.2-cp39-cp39-musllinux_1_2_i686.whl", hash = "sha256:14dc3fcb0d877911d775d511eb617a486a8c48afca0a887276e63db04d3ee920"},
-    {file = "aiohttp-3.10.2-cp39-cp39-musllinux_1_2_ppc64le.whl", hash = "sha256:b52a27a5c97275e254704e1049f4b96a81e67d6205f52fa37a4777d55b0e98ef"},
-    {file = "aiohttp-3.10.2-cp39-cp39-musllinux_1_2_s390x.whl", hash = "sha256:cd33d9de8cfd006a0d0fe85f49b4183c57e91d18ffb7e9004ce855e81928f704"},
-    {file = "aiohttp-3.10.2-cp39-cp39-musllinux_1_2_x86_64.whl", hash = "sha256:1238fc979160bc03a92fff9ad021375ff1c8799c6aacb0d8ea1b357ea40932bb"},
-    {file = "aiohttp-3.10.2-cp39-cp39-win32.whl", hash = "sha256:e2f43d238eae4f0b04f58d4c0df4615697d4ca3e9f9b1963d49555a94f0f5a04"},
-    {file = "aiohttp-3.10.2-cp39-cp39-win_amd64.whl", hash = "sha256:947847f07a8f81d7b39b2d0202fd73e61962ebe17ac2d8566f260679e467da7b"},
-    {file = "aiohttp-3.10.2.tar.gz", hash = "sha256:4d1f694b5d6e459352e5e925a42e05bac66655bfde44d81c59992463d2897014"},
+    {file = "aiohttp-3.10.11-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:5077b1a5f40ffa3ba1f40d537d3bec4383988ee51fbba6b74aa8fb1bc466599e"},
+    {file = "aiohttp-3.10.11-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:8d6a14a4d93b5b3c2891fca94fa9d41b2322a68194422bef0dd5ec1e57d7d298"},
+    {file = "aiohttp-3.10.11-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:ffbfde2443696345e23a3c597049b1dd43049bb65337837574205e7368472177"},
+    {file = "aiohttp-3.10.11-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:20b3d9e416774d41813bc02fdc0663379c01817b0874b932b81c7f777f67b217"},
+    {file = "aiohttp-3.10.11-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:2b943011b45ee6bf74b22245c6faab736363678e910504dd7531a58c76c9015a"},
+    {file = "aiohttp-3.10.11-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:48bc1d924490f0d0b3658fe5c4b081a4d56ebb58af80a6729d4bd13ea569797a"},
+    {file = "aiohttp-3.10.11-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:e12eb3f4b1f72aaaf6acd27d045753b18101524f72ae071ae1c91c1cd44ef115"},
+    {file = "aiohttp-3.10.11-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:f14ebc419a568c2eff3c1ed35f634435c24ead2fe19c07426af41e7adb68713a"},
+    {file = "aiohttp-3.10.11-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:72b191cdf35a518bfc7ca87d770d30941decc5aaf897ec8b484eb5cc8c7706f3"},
+    {file = "aiohttp-3.10.11-cp310-cp310-musllinux_1_2_i686.whl", hash = "sha256:5ab2328a61fdc86424ee540d0aeb8b73bbcad7351fb7cf7a6546fc0bcffa0038"},
+    {file = "aiohttp-3.10.11-cp310-cp310-musllinux_1_2_ppc64le.whl", hash = "sha256:aa93063d4af05c49276cf14e419550a3f45258b6b9d1f16403e777f1addf4519"},
+    {file = "aiohttp-3.10.11-cp310-cp310-musllinux_1_2_s390x.whl", hash = "sha256:30283f9d0ce420363c24c5c2421e71a738a2155f10adbb1a11a4d4d6d2715cfc"},
+    {file = "aiohttp-3.10.11-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:e5358addc8044ee49143c546d2182c15b4ac3a60be01c3209374ace05af5733d"},
+    {file = "aiohttp-3.10.11-cp310-cp310-win32.whl", hash = "sha256:e1ffa713d3ea7cdcd4aea9cddccab41edf6882fa9552940344c44e59652e1120"},
+    {file = "aiohttp-3.10.11-cp310-cp310-win_amd64.whl", hash = "sha256:778cbd01f18ff78b5dd23c77eb82987ee4ba23408cbed233009fd570dda7e674"},
+    {file = "aiohttp-3.10.11-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:80ff08556c7f59a7972b1e8919f62e9c069c33566a6d28586771711e0eea4f07"},
+    {file = "aiohttp-3.10.11-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:2c8f96e9ee19f04c4914e4e7a42a60861066d3e1abf05c726f38d9d0a466e695"},
+    {file = "aiohttp-3.10.11-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:fb8601394d537da9221947b5d6e62b064c9a43e88a1ecd7414d21a1a6fba9c24"},
+    {file = "aiohttp-3.10.11-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:2ea224cf7bc2d8856d6971cea73b1d50c9c51d36971faf1abc169a0d5f85a382"},
+    {file = "aiohttp-3.10.11-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:db9503f79e12d5d80b3efd4d01312853565c05367493379df76d2674af881caa"},
+    {file = "aiohttp-3.10.11-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:0f449a50cc33f0384f633894d8d3cd020e3ccef81879c6e6245c3c375c448625"},
+    {file = "aiohttp-3.10.11-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:82052be3e6d9e0c123499127782a01a2b224b8af8c62ab46b3f6197035ad94e9"},
+    {file = "aiohttp-3.10.11-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:20063c7acf1eec550c8eb098deb5ed9e1bb0521613b03bb93644b810986027ac"},
+    {file = "aiohttp-3.10.11-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:489cced07a4c11488f47aab1f00d0c572506883f877af100a38f1fedaa884c3a"},
+    {file = "aiohttp-3.10.11-cp311-cp311-musllinux_1_2_i686.whl", hash = "sha256:ea9b3bab329aeaa603ed3bf605f1e2a6f36496ad7e0e1aa42025f368ee2dc07b"},
+    {file = "aiohttp-3.10.11-cp311-cp311-musllinux_1_2_ppc64le.whl", hash = "sha256:ca117819d8ad113413016cb29774b3f6d99ad23c220069789fc050267b786c16"},
+    {file = "aiohttp-3.10.11-cp311-cp311-musllinux_1_2_s390x.whl", hash = "sha256:2dfb612dcbe70fb7cdcf3499e8d483079b89749c857a8f6e80263b021745c730"},
+    {file = "aiohttp-3.10.11-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:f9b615d3da0d60e7d53c62e22b4fd1c70f4ae5993a44687b011ea3a2e49051b8"},
+    {file = "aiohttp-3.10.11-cp311-cp311-win32.whl", hash = "sha256:29103f9099b6068bbdf44d6a3d090e0a0b2be6d3c9f16a070dd9d0d910ec08f9"},
+    {file = "aiohttp-3.10.11-cp311-cp311-win_amd64.whl", hash = "sha256:236b28ceb79532da85d59aa9b9bf873b364e27a0acb2ceaba475dc61cffb6f3f"},
+    {file = "aiohttp-3.10.11-cp312-cp312-macosx_10_9_universal2.whl", hash = "sha256:7480519f70e32bfb101d71fb9a1f330fbd291655a4c1c922232a48c458c52710"},
+    {file = "aiohttp-3.10.11-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:f65267266c9aeb2287a6622ee2bb39490292552f9fbf851baabc04c9f84e048d"},
+    {file = "aiohttp-3.10.11-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:7400a93d629a0608dc1d6c55f1e3d6e07f7375745aaa8bd7f085571e4d1cee97"},
+    {file = "aiohttp-3.10.11-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f34b97e4b11b8d4eb2c3a4f975be626cc8af99ff479da7de49ac2c6d02d35725"},
+    {file = "aiohttp-3.10.11-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:1e7b825da878464a252ccff2958838f9caa82f32a8dbc334eb9b34a026e2c636"},
+    {file = "aiohttp-3.10.11-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:f9f92a344c50b9667827da308473005f34767b6a2a60d9acff56ae94f895f385"},
+    {file = "aiohttp-3.10.11-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:bc6f1ab987a27b83c5268a17218463c2ec08dbb754195113867a27b166cd6087"},
+    {file = "aiohttp-3.10.11-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:1dc0f4ca54842173d03322793ebcf2c8cc2d34ae91cc762478e295d8e361e03f"},
+    {file = "aiohttp-3.10.11-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:7ce6a51469bfaacff146e59e7fb61c9c23006495d11cc24c514a455032bcfa03"},
+    {file = "aiohttp-3.10.11-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:aad3cd91d484d065ede16f3cf15408254e2469e3f613b241a1db552c5eb7ab7d"},
+    {file = "aiohttp-3.10.11-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:f4df4b8ca97f658c880fb4b90b1d1ec528315d4030af1ec763247ebfd33d8b9a"},
+    {file = "aiohttp-3.10.11-cp312-cp312-musllinux_1_2_s390x.whl", hash = "sha256:2e4e18a0a2d03531edbc06c366954e40a3f8d2a88d2b936bbe78a0c75a3aab3e"},
+    {file = "aiohttp-3.10.11-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:6ce66780fa1a20e45bc753cda2a149daa6dbf1561fc1289fa0c308391c7bc0a4"},
+    {file = "aiohttp-3.10.11-cp312-cp312-win32.whl", hash = "sha256:a919c8957695ea4c0e7a3e8d16494e3477b86f33067478f43106921c2fef15bb"},
+    {file = "aiohttp-3.10.11-cp312-cp312-win_amd64.whl", hash = "sha256:b5e29706e6389a2283a91611c91bf24f218962717c8f3b4e528ef529d112ee27"},
+    {file = "aiohttp-3.10.11-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:703938e22434d7d14ec22f9f310559331f455018389222eed132808cd8f44127"},
+    {file = "aiohttp-3.10.11-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:9bc50b63648840854e00084c2b43035a62e033cb9b06d8c22b409d56eb098413"},
+    {file = "aiohttp-3.10.11-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:5f0463bf8b0754bc744e1feb61590706823795041e63edf30118a6f0bf577461"},
+    {file = "aiohttp-3.10.11-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f6c6dec398ac5a87cb3a407b068e1106b20ef001c344e34154616183fe684288"},
+    {file = "aiohttp-3.10.11-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:bcaf2d79104d53d4dcf934f7ce76d3d155302d07dae24dff6c9fffd217568067"},
+    {file = "aiohttp-3.10.11-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:25fd5470922091b5a9aeeb7e75be609e16b4fba81cdeaf12981393fb240dd10e"},
+    {file = "aiohttp-3.10.11-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:bbde2ca67230923a42161b1f408c3992ae6e0be782dca0c44cb3206bf330dee1"},
+    {file = "aiohttp-3.10.11-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:249c8ff8d26a8b41a0f12f9df804e7c685ca35a207e2410adbd3e924217b9006"},
+    {file = "aiohttp-3.10.11-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:878ca6a931ee8c486a8f7b432b65431d095c522cbeb34892bee5be97b3481d0f"},
+    {file = "aiohttp-3.10.11-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:8663f7777ce775f0413324be0d96d9730959b2ca73d9b7e2c2c90539139cbdd6"},
+    {file = "aiohttp-3.10.11-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:6cd3f10b01f0c31481fba8d302b61603a2acb37b9d30e1d14e0f5a58b7b18a31"},
+    {file = "aiohttp-3.10.11-cp313-cp313-musllinux_1_2_s390x.whl", hash = "sha256:4e8d8aad9402d3aa02fdc5ca2fe68bcb9fdfe1f77b40b10410a94c7f408b664d"},
+    {file = "aiohttp-3.10.11-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:38e3c4f80196b4f6c3a85d134a534a56f52da9cb8d8e7af1b79a32eefee73a00"},
+    {file = "aiohttp-3.10.11-cp313-cp313-win32.whl", hash = "sha256:fc31820cfc3b2863c6e95e14fcf815dc7afe52480b4dc03393c4873bb5599f71"},
+    {file = "aiohttp-3.10.11-cp313-cp313-win_amd64.whl", hash = "sha256:4996ff1345704ffdd6d75fb06ed175938c133425af616142e7187f28dc75f14e"},
+    {file = "aiohttp-3.10.11-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:74baf1a7d948b3d640badeac333af581a367ab916b37e44cf90a0334157cdfd2"},
+    {file = "aiohttp-3.10.11-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:473aebc3b871646e1940c05268d451f2543a1d209f47035b594b9d4e91ce8339"},
+    {file = "aiohttp-3.10.11-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:c2f746a6968c54ab2186574e15c3f14f3e7f67aef12b761e043b33b89c5b5f95"},
+    {file = "aiohttp-3.10.11-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d110cabad8360ffa0dec8f6ec60e43286e9d251e77db4763a87dcfe55b4adb92"},
+    {file = "aiohttp-3.10.11-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:e0099c7d5d7afff4202a0c670e5b723f7718810000b4abcbc96b064129e64bc7"},
+    {file = "aiohttp-3.10.11-cp38-cp38-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:0316e624b754dbbf8c872b62fe6dcb395ef20c70e59890dfa0de9eafccd2849d"},
+    {file = "aiohttp-3.10.11-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:5a5f7ab8baf13314e6b2485965cbacb94afff1e93466ac4d06a47a81c50f9cca"},
+    {file = "aiohttp-3.10.11-cp38-cp38-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:c891011e76041e6508cbfc469dd1a8ea09bc24e87e4c204e05f150c4c455a5fa"},
+    {file = "aiohttp-3.10.11-cp38-cp38-musllinux_1_2_aarch64.whl", hash = "sha256:9208299251370ee815473270c52cd3f7069ee9ed348d941d574d1457d2c73e8b"},
+    {file = "aiohttp-3.10.11-cp38-cp38-musllinux_1_2_i686.whl", hash = "sha256:459f0f32c8356e8125f45eeff0ecf2b1cb6db1551304972702f34cd9e6c44658"},
+    {file = "aiohttp-3.10.11-cp38-cp38-musllinux_1_2_ppc64le.whl", hash = "sha256:14cdc8c1810bbd4b4b9f142eeee23cda528ae4e57ea0923551a9af4820980e39"},
+    {file = "aiohttp-3.10.11-cp38-cp38-musllinux_1_2_s390x.whl", hash = "sha256:971aa438a29701d4b34e4943e91b5e984c3ae6ccbf80dd9efaffb01bd0b243a9"},
+    {file = "aiohttp-3.10.11-cp38-cp38-musllinux_1_2_x86_64.whl", hash = "sha256:9a309c5de392dfe0f32ee57fa43ed8fc6ddf9985425e84bd51ed66bb16bce3a7"},
+    {file = "aiohttp-3.10.11-cp38-cp38-win32.whl", hash = "sha256:9ec1628180241d906a0840b38f162a3215114b14541f1a8711c368a8739a9be4"},
+    {file = "aiohttp-3.10.11-cp38-cp38-win_amd64.whl", hash = "sha256:9c6e0ffd52c929f985c7258f83185d17c76d4275ad22e90aa29f38e211aacbec"},
+    {file = "aiohttp-3.10.11-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:cdc493a2e5d8dc79b2df5bec9558425bcd39aff59fc949810cbd0832e294b106"},
+    {file = "aiohttp-3.10.11-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:b3e70f24e7d0405be2348da9d5a7836936bf3a9b4fd210f8c37e8d48bc32eca6"},
+    {file = "aiohttp-3.10.11-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:968b8fb2a5eee2770eda9c7b5581587ef9b96fbdf8dcabc6b446d35ccc69df01"},
+    {file = "aiohttp-3.10.11-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:deef4362af9493d1382ef86732ee2e4cbc0d7c005947bd54ad1a9a16dd59298e"},
+    {file = "aiohttp-3.10.11-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:686b03196976e327412a1b094f4120778c7c4b9cff9bce8d2fdfeca386b89829"},
+    {file = "aiohttp-3.10.11-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:3bf6d027d9d1d34e1c2e1645f18a6498c98d634f8e373395221121f1c258ace8"},
+    {file = "aiohttp-3.10.11-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:099fd126bf960f96d34a760e747a629c27fb3634da5d05c7ef4d35ef4ea519fc"},
+    {file = "aiohttp-3.10.11-cp39-cp39-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:c73c4d3dae0b4644bc21e3de546530531d6cdc88659cdeb6579cd627d3c206aa"},
+    {file = "aiohttp-3.10.11-cp39-cp39-musllinux_1_2_aarch64.whl", hash = "sha256:0c5580f3c51eea91559db3facd45d72e7ec970b04528b4709b1f9c2555bd6d0b"},
+    {file = "aiohttp-3.10.11-cp39-cp39-musllinux_1_2_i686.whl", hash = "sha256:fdf6429f0caabfd8a30c4e2eaecb547b3c340e4730ebfe25139779b9815ba138"},
+    {file = "aiohttp-3.10.11-cp39-cp39-musllinux_1_2_ppc64le.whl", hash = "sha256:d97187de3c276263db3564bb9d9fad9e15b51ea10a371ffa5947a5ba93ad6777"},
+    {file = "aiohttp-3.10.11-cp39-cp39-musllinux_1_2_s390x.whl", hash = "sha256:0acafb350cfb2eba70eb5d271f55e08bd4502ec35e964e18ad3e7d34d71f7261"},
+    {file = "aiohttp-3.10.11-cp39-cp39-musllinux_1_2_x86_64.whl", hash = "sha256:c13ed0c779911c7998a58e7848954bd4d63df3e3575f591e321b19a2aec8df9f"},
+    {file = "aiohttp-3.10.11-cp39-cp39-win32.whl", hash = "sha256:22b7c540c55909140f63ab4f54ec2c20d2635c0289cdd8006da46f3327f971b9"},
+    {file = "aiohttp-3.10.11-cp39-cp39-win_amd64.whl", hash = "sha256:7b26b1551e481012575dab8e3727b16fe7dd27eb2711d2e63ced7368756268fb"},
+    {file = "aiohttp-3.10.11.tar.gz", hash = "sha256:9dc2b8f3dcab2e39e0fa309c8da50c3b55e6f34ab25f1a71d3288f24924d33a7"},
 ]
 
 [package.dependencies]
 aiohappyeyeballs = ">=2.3.0"
 aiosignal = ">=1.1.2"
-async-timeout = {version = ">=4.0,<5.0", markers = "python_version < \"3.11\""}
+async-timeout = {version = ">=4.0,<6.0", markers = "python_version < \"3.11\""}
 attrs = ">=17.3.0"
 frozenlist = ">=1.1.1"
 multidict = ">=4.5,<7.0"
-yarl = ">=1.0,<2.0"
+yarl = ">=1.12.0,<2.0"
 
 [package.extras]
 speedups = ["Brotli", "aiodns (>=3.2.0)", "brotlicffi"]
@@ -2078,6 +2093,113 @@ files = [
 [package.extras]
 twisted = ["twisted"]
 
+[[package]]
+name = "propcache"
+version = "0.2.0"
+description = "Accelerated property cache"
+optional = false
+python-versions = ">=3.8"
+files = [
+    {file = "propcache-0.2.0-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:c5869b8fd70b81835a6f187c5fdbe67917a04d7e52b6e7cc4e5fe39d55c39d58"},
+    {file = "propcache-0.2.0-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:952e0d9d07609d9c5be361f33b0d6d650cd2bae393aabb11d9b719364521984b"},
+    {file = "propcache-0.2.0-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:33ac8f098df0585c0b53009f039dfd913b38c1d2edafed0cedcc0c32a05aa110"},
+    {file = "propcache-0.2.0-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:97e48e8875e6c13909c800fa344cd54cc4b2b0db1d5f911f840458a500fde2c2"},
+    {file = "propcache-0.2.0-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:388f3217649d6d59292b722d940d4d2e1e6a7003259eb835724092a1cca0203a"},
+    {file = "propcache-0.2.0-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:f571aea50ba5623c308aa146eb650eebf7dbe0fd8c5d946e28343cb3b5aad577"},
+    {file = "propcache-0.2.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:3dfafb44f7bb35c0c06eda6b2ab4bfd58f02729e7c4045e179f9a861b07c9850"},
+    {file = "propcache-0.2.0-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:a3ebe9a75be7ab0b7da2464a77bb27febcb4fab46a34f9288f39d74833db7f61"},
+    {file = "propcache-0.2.0-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:d2f0d0f976985f85dfb5f3d685697ef769faa6b71993b46b295cdbbd6be8cc37"},
+    {file = "propcache-0.2.0-cp310-cp310-musllinux_1_2_armv7l.whl", hash = "sha256:a3dc1a4b165283bd865e8f8cb5f0c64c05001e0718ed06250d8cac9bec115b48"},
+    {file = "propcache-0.2.0-cp310-cp310-musllinux_1_2_i686.whl", hash = "sha256:9e0f07b42d2a50c7dd2d8675d50f7343d998c64008f1da5fef888396b7f84630"},
+    {file = "propcache-0.2.0-cp310-cp310-musllinux_1_2_ppc64le.whl", hash = "sha256:e63e3e1e0271f374ed489ff5ee73d4b6e7c60710e1f76af5f0e1a6117cd26394"},
+    {file = "propcache-0.2.0-cp310-cp310-musllinux_1_2_s390x.whl", hash = "sha256:56bb5c98f058a41bb58eead194b4db8c05b088c93d94d5161728515bd52b052b"},
+    {file = "propcache-0.2.0-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:7665f04d0c7f26ff8bb534e1c65068409bf4687aa2534faf7104d7182debb336"},
+    {file = "propcache-0.2.0-cp310-cp310-win32.whl", hash = "sha256:7cf18abf9764746b9c8704774d8b06714bcb0a63641518a3a89c7f85cc02c2ad"},
+    {file = "propcache-0.2.0-cp310-cp310-win_amd64.whl", hash = "sha256:cfac69017ef97db2438efb854edf24f5a29fd09a536ff3a992b75990720cdc99"},
+    {file = "propcache-0.2.0-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:63f13bf09cc3336eb04a837490b8f332e0db41da66995c9fd1ba04552e516354"},
+    {file = "propcache-0.2.0-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:608cce1da6f2672a56b24a015b42db4ac612ee709f3d29f27a00c943d9e851de"},
+    {file = "propcache-0.2.0-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:466c219deee4536fbc83c08d09115249db301550625c7fef1c5563a584c9bc87"},
+    {file = "propcache-0.2.0-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:fc2db02409338bf36590aa985a461b2c96fce91f8e7e0f14c50c5fcc4f229016"},
+    {file = "propcache-0.2.0-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:a6ed8db0a556343d566a5c124ee483ae113acc9a557a807d439bcecc44e7dfbb"},
+    {file = "propcache-0.2.0-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:91997d9cb4a325b60d4e3f20967f8eb08dfcb32b22554d5ef78e6fd1dda743a2"},
+    {file = "propcache-0.2.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:4c7dde9e533c0a49d802b4f3f218fa9ad0a1ce21f2c2eb80d5216565202acab4"},
+    {file = "propcache-0.2.0-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:ffcad6c564fe6b9b8916c1aefbb37a362deebf9394bd2974e9d84232e3e08504"},
+    {file = "propcache-0.2.0-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:97a58a28bcf63284e8b4d7b460cbee1edaab24634e82059c7b8c09e65284f178"},
+    {file = "propcache-0.2.0-cp311-cp311-musllinux_1_2_armv7l.whl", hash = "sha256:945db8ee295d3af9dbdbb698cce9bbc5c59b5c3fe328bbc4387f59a8a35f998d"},
+    {file = "propcache-0.2.0-cp311-cp311-musllinux_1_2_i686.whl", hash = "sha256:39e104da444a34830751715f45ef9fc537475ba21b7f1f5b0f4d71a3b60d7fe2"},
+    {file = "propcache-0.2.0-cp311-cp311-musllinux_1_2_ppc64le.whl", hash = "sha256:c5ecca8f9bab618340c8e848d340baf68bcd8ad90a8ecd7a4524a81c1764b3db"},
+    {file = "propcache-0.2.0-cp311-cp311-musllinux_1_2_s390x.whl", hash = "sha256:c436130cc779806bdf5d5fae0d848713105472b8566b75ff70048c47d3961c5b"},
+    {file = "propcache-0.2.0-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:191db28dc6dcd29d1a3e063c3be0b40688ed76434622c53a284e5427565bbd9b"},
+    {file = "propcache-0.2.0-cp311-cp311-win32.whl", hash = "sha256:5f2564ec89058ee7c7989a7b719115bdfe2a2fb8e7a4543b8d1c0cc4cf6478c1"},
+    {file = "propcache-0.2.0-cp311-cp311-win_amd64.whl", hash = "sha256:6e2e54267980349b723cff366d1e29b138b9a60fa376664a157a342689553f71"},
+    {file = "propcache-0.2.0-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:2ee7606193fb267be4b2e3b32714f2d58cad27217638db98a60f9efb5efeccc2"},
+    {file = "propcache-0.2.0-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:91ee8fc02ca52e24bcb77b234f22afc03288e1dafbb1f88fe24db308910c4ac7"},
+    {file = "propcache-0.2.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:2e900bad2a8456d00a113cad8c13343f3b1f327534e3589acc2219729237a2e8"},
+    {file = "propcache-0.2.0-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f52a68c21363c45297aca15561812d542f8fc683c85201df0bebe209e349f793"},
+    {file = "propcache-0.2.0-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:1e41d67757ff4fbc8ef2af99b338bfb955010444b92929e9e55a6d4dcc3c4f09"},
+    {file = "propcache-0.2.0-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:a64e32f8bd94c105cc27f42d3b658902b5bcc947ece3c8fe7bc1b05982f60e89"},
+    {file = "propcache-0.2.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:55346705687dbd7ef0d77883ab4f6fabc48232f587925bdaf95219bae072491e"},
+    {file = "propcache-0.2.0-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:00181262b17e517df2cd85656fcd6b4e70946fe62cd625b9d74ac9977b64d8d9"},
+    {file = "propcache-0.2.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:6994984550eaf25dd7fc7bd1b700ff45c894149341725bb4edc67f0ffa94efa4"},
+    {file = "propcache-0.2.0-cp312-cp312-musllinux_1_2_armv7l.whl", hash = "sha256:56295eb1e5f3aecd516d91b00cfd8bf3a13991de5a479df9e27dd569ea23959c"},
+    {file = "propcache-0.2.0-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:439e76255daa0f8151d3cb325f6dd4a3e93043e6403e6491813bcaaaa8733887"},
+    {file = "propcache-0.2.0-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:f6475a1b2ecb310c98c28d271a30df74f9dd436ee46d09236a6b750a7599ce57"},
+    {file = "propcache-0.2.0-cp312-cp312-musllinux_1_2_s390x.whl", hash = "sha256:3444cdba6628accf384e349014084b1cacd866fbb88433cd9d279d90a54e0b23"},
+    {file = "propcache-0.2.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:4a9d9b4d0a9b38d1c391bb4ad24aa65f306c6f01b512e10a8a34a2dc5675d348"},
+    {file = "propcache-0.2.0-cp312-cp312-win32.whl", hash = "sha256:69d3a98eebae99a420d4b28756c8ce6ea5a29291baf2dc9ff9414b42676f61d5"},
+    {file = "propcache-0.2.0-cp312-cp312-win_amd64.whl", hash = "sha256:ad9c9b99b05f163109466638bd30ada1722abb01bbb85c739c50b6dc11f92dc3"},
+    {file = "propcache-0.2.0-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:ecddc221a077a8132cf7c747d5352a15ed763b674c0448d811f408bf803d9ad7"},
+    {file = "propcache-0.2.0-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:0e53cb83fdd61cbd67202735e6a6687a7b491c8742dfc39c9e01e80354956763"},
+    {file = "propcache-0.2.0-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:92fe151145a990c22cbccf9ae15cae8ae9eddabfc949a219c9f667877e40853d"},
+    {file = "propcache-0.2.0-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d6a21ef516d36909931a2967621eecb256018aeb11fc48656e3257e73e2e247a"},
+    {file = "propcache-0.2.0-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:3f88a4095e913f98988f5b338c1d4d5d07dbb0b6bad19892fd447484e483ba6b"},
+    {file = "propcache-0.2.0-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:5a5b3bb545ead161be780ee85a2b54fdf7092815995661947812dde94a40f6fb"},
+    {file = "propcache-0.2.0-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:67aeb72e0f482709991aa91345a831d0b707d16b0257e8ef88a2ad246a7280bf"},
+    {file = "propcache-0.2.0-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:3c997f8c44ec9b9b0bcbf2d422cc00a1d9b9c681f56efa6ca149a941e5560da2"},
+    {file = "propcache-0.2.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:2a66df3d4992bc1d725b9aa803e8c5a66c010c65c741ad901e260ece77f58d2f"},
+    {file = "propcache-0.2.0-cp313-cp313-musllinux_1_2_armv7l.whl", hash = "sha256:3ebbcf2a07621f29638799828b8d8668c421bfb94c6cb04269130d8de4fb7136"},
+    {file = "propcache-0.2.0-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:1235c01ddaa80da8235741e80815ce381c5267f96cc49b1477fdcf8c047ef325"},
+    {file = "propcache-0.2.0-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:3947483a381259c06921612550867b37d22e1df6d6d7e8361264b6d037595f44"},
+    {file = "propcache-0.2.0-cp313-cp313-musllinux_1_2_s390x.whl", hash = "sha256:d5bed7f9805cc29c780f3aee05de3262ee7ce1f47083cfe9f77471e9d6777e83"},
+    {file = "propcache-0.2.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:e4a91d44379f45f5e540971d41e4626dacd7f01004826a18cb048e7da7e96544"},
+    {file = "propcache-0.2.0-cp313-cp313-win32.whl", hash = "sha256:f902804113e032e2cdf8c71015651c97af6418363bea8d78dc0911d56c335032"},
+    {file = "propcache-0.2.0-cp313-cp313-win_amd64.whl", hash = "sha256:8f188cfcc64fb1266f4684206c9de0e80f54622c3f22a910cbd200478aeae61e"},
+    {file = "propcache-0.2.0-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:53d1bd3f979ed529f0805dd35ddaca330f80a9a6d90bc0121d2ff398f8ed8861"},
+    {file = "propcache-0.2.0-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:83928404adf8fb3d26793665633ea79b7361efa0287dfbd372a7e74311d51ee6"},
+    {file = "propcache-0.2.0-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:77a86c261679ea5f3896ec060be9dc8e365788248cc1e049632a1be682442063"},
+    {file = "propcache-0.2.0-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:218db2a3c297a3768c11a34812e63b3ac1c3234c3a086def9c0fee50d35add1f"},
+    {file = "propcache-0.2.0-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:7735e82e3498c27bcb2d17cb65d62c14f1100b71723b68362872bca7d0913d90"},
+    {file = "propcache-0.2.0-cp38-cp38-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:20a617c776f520c3875cf4511e0d1db847a076d720714ae35ffe0df3e440be68"},
+    {file = "propcache-0.2.0-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:67b69535c870670c9f9b14a75d28baa32221d06f6b6fa6f77a0a13c5a7b0a5b9"},
+    {file = "propcache-0.2.0-cp38-cp38-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:4569158070180c3855e9c0791c56be3ceeb192defa2cdf6a3f39e54319e56b89"},
+    {file = "propcache-0.2.0-cp38-cp38-musllinux_1_2_aarch64.whl", hash = "sha256:db47514ffdbd91ccdc7e6f8407aac4ee94cc871b15b577c1c324236b013ddd04"},
+    {file = "propcache-0.2.0-cp38-cp38-musllinux_1_2_armv7l.whl", hash = "sha256:2a60ad3e2553a74168d275a0ef35e8c0a965448ffbc3b300ab3a5bb9956c2162"},
+    {file = "propcache-0.2.0-cp38-cp38-musllinux_1_2_i686.whl", hash = "sha256:662dd62358bdeaca0aee5761de8727cfd6861432e3bb828dc2a693aa0471a563"},
+    {file = "propcache-0.2.0-cp38-cp38-musllinux_1_2_ppc64le.whl", hash = "sha256:25a1f88b471b3bc911d18b935ecb7115dff3a192b6fef46f0bfaf71ff4f12418"},
+    {file = "propcache-0.2.0-cp38-cp38-musllinux_1_2_s390x.whl", hash = "sha256:f60f0ac7005b9f5a6091009b09a419ace1610e163fa5deaba5ce3484341840e7"},
+    {file = "propcache-0.2.0-cp38-cp38-musllinux_1_2_x86_64.whl", hash = "sha256:74acd6e291f885678631b7ebc85d2d4aec458dd849b8c841b57ef04047833bed"},
+    {file = "propcache-0.2.0-cp38-cp38-win32.whl", hash = "sha256:d9b6ddac6408194e934002a69bcaadbc88c10b5f38fb9307779d1c629181815d"},
+    {file = "propcache-0.2.0-cp38-cp38-win_amd64.whl", hash = "sha256:676135dcf3262c9c5081cc8f19ad55c8a64e3f7282a21266d05544450bffc3a5"},
+    {file = "propcache-0.2.0-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:25c8d773a62ce0451b020c7b29a35cfbc05de8b291163a7a0f3b7904f27253e6"},
+    {file = "propcache-0.2.0-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:375a12d7556d462dc64d70475a9ee5982465fbb3d2b364f16b86ba9135793638"},
+    {file = "propcache-0.2.0-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:1ec43d76b9677637a89d6ab86e1fef70d739217fefa208c65352ecf0282be957"},
+    {file = "propcache-0.2.0-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f45eec587dafd4b2d41ac189c2156461ebd0c1082d2fe7013571598abb8505d1"},
+    {file = "propcache-0.2.0-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:bc092ba439d91df90aea38168e11f75c655880c12782facf5cf9c00f3d42b562"},
+    {file = "propcache-0.2.0-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:fa1076244f54bb76e65e22cb6910365779d5c3d71d1f18b275f1dfc7b0d71b4d"},
+    {file = "propcache-0.2.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:682a7c79a2fbf40f5dbb1eb6bfe2cd865376deeac65acf9beb607505dced9e12"},
+    {file = "propcache-0.2.0-cp39-cp39-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:8e40876731f99b6f3c897b66b803c9e1c07a989b366c6b5b475fafd1f7ba3fb8"},
+    {file = "propcache-0.2.0-cp39-cp39-musllinux_1_2_aarch64.whl", hash = "sha256:363ea8cd3c5cb6679f1c2f5f1f9669587361c062e4899fce56758efa928728f8"},
+    {file = "propcache-0.2.0-cp39-cp39-musllinux_1_2_armv7l.whl", hash = "sha256:140fbf08ab3588b3468932974a9331aff43c0ab8a2ec2c608b6d7d1756dbb6cb"},
+    {file = "propcache-0.2.0-cp39-cp39-musllinux_1_2_i686.whl", hash = "sha256:e70fac33e8b4ac63dfc4c956fd7d85a0b1139adcfc0d964ce288b7c527537fea"},
+    {file = "propcache-0.2.0-cp39-cp39-musllinux_1_2_ppc64le.whl", hash = "sha256:b33d7a286c0dc1a15f5fc864cc48ae92a846df287ceac2dd499926c3801054a6"},
+    {file = "propcache-0.2.0-cp39-cp39-musllinux_1_2_s390x.whl", hash = "sha256:f6d5749fdd33d90e34c2efb174c7e236829147a2713334d708746e94c4bde40d"},
+    {file = "propcache-0.2.0-cp39-cp39-musllinux_1_2_x86_64.whl", hash = "sha256:22aa8f2272d81d9317ff5756bb108021a056805ce63dd3630e27d042c8092798"},
+    {file = "propcache-0.2.0-cp39-cp39-win32.whl", hash = "sha256:73e4b40ea0eda421b115248d7e79b59214411109a5bc47d0d48e4c73e3b8fcf9"},
+    {file = "propcache-0.2.0-cp39-cp39-win_amd64.whl", hash = "sha256:9517d5e9e0731957468c29dbfd0f976736a0e55afaea843726e887f36fe017df"},
+    {file = "propcache-0.2.0-py3-none-any.whl", hash = "sha256:2ccc28197af5313706511fab3a8b66dcd6da067a1331372c82ea1cb74285e036"},
+    {file = "propcache-0.2.0.tar.gz", hash = "sha256:df81779732feb9d01e5d513fad0122efb3d53bbc75f61b2a4f29a020bc985e70"},
+]
+
 [[package]]
 name = "psutil"
 version = "5.9.4"
@@ -3307,106 +3429,99 @@ files = [
 
 [[package]]
 name = "yarl"
-version = "1.9.4"
+version = "1.17.2"
 description = "Yet another URL library"
 optional = false
-python-versions = ">=3.7"
+python-versions = ">=3.9"
 files = [
-    {file = "yarl-1.9.4-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:a8c1df72eb746f4136fe9a2e72b0c9dc1da1cbd23b5372f94b5820ff8ae30e0e"},
-    {file = "yarl-1.9.4-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:a3a6ed1d525bfb91b3fc9b690c5a21bb52de28c018530ad85093cc488bee2dd2"},
-    {file = "yarl-1.9.4-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:c38c9ddb6103ceae4e4498f9c08fac9b590c5c71b0370f98714768e22ac6fa66"},
-    {file = "yarl-1.9.4-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d9e09c9d74f4566e905a0b8fa668c58109f7624db96a2171f21747abc7524234"},
-    {file = "yarl-1.9.4-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:b8477c1ee4bd47c57d49621a062121c3023609f7a13b8a46953eb6c9716ca392"},
-    {file = "yarl-1.9.4-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:d5ff2c858f5f6a42c2a8e751100f237c5e869cbde669a724f2062d4c4ef93551"},
-    {file = "yarl-1.9.4-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:357495293086c5b6d34ca9616a43d329317feab7917518bc97a08f9e55648455"},
-    {file = "yarl-1.9.4-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:54525ae423d7b7a8ee81ba189f131054defdb122cde31ff17477951464c1691c"},
-    {file = "yarl-1.9.4-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:801e9264d19643548651b9db361ce3287176671fb0117f96b5ac0ee1c3530d53"},
-    {file = "yarl-1.9.4-cp310-cp310-musllinux_1_1_i686.whl", hash = "sha256:e516dc8baf7b380e6c1c26792610230f37147bb754d6426462ab115a02944385"},
-    {file = "yarl-1.9.4-cp310-cp310-musllinux_1_1_ppc64le.whl", hash = "sha256:7d5aaac37d19b2904bb9dfe12cdb08c8443e7ba7d2852894ad448d4b8f442863"},
-    {file = "yarl-1.9.4-cp310-cp310-musllinux_1_1_s390x.whl", hash = "sha256:54beabb809ffcacbd9d28ac57b0db46e42a6e341a030293fb3185c409e626b8b"},
-    {file = "yarl-1.9.4-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:bac8d525a8dbc2a1507ec731d2867025d11ceadcb4dd421423a5d42c56818541"},
-    {file = "yarl-1.9.4-cp310-cp310-win32.whl", hash = "sha256:7855426dfbddac81896b6e533ebefc0af2f132d4a47340cee6d22cac7190022d"},
-    {file = "yarl-1.9.4-cp310-cp310-win_amd64.whl", hash = "sha256:848cd2a1df56ddbffeb375535fb62c9d1645dde33ca4d51341378b3f5954429b"},
-    {file = "yarl-1.9.4-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:35a2b9396879ce32754bd457d31a51ff0a9d426fd9e0e3c33394bf4b9036b099"},
-    {file = "yarl-1.9.4-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:4c7d56b293cc071e82532f70adcbd8b61909eec973ae9d2d1f9b233f3d943f2c"},
-    {file = "yarl-1.9.4-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:d8a1c6c0be645c745a081c192e747c5de06e944a0d21245f4cf7c05e457c36e0"},
-    {file = "yarl-1.9.4-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:4b3c1ffe10069f655ea2d731808e76e0f452fc6c749bea04781daf18e6039525"},
-    {file = "yarl-1.9.4-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:549d19c84c55d11687ddbd47eeb348a89df9cb30e1993f1b128f4685cd0ebbf8"},
-    {file = "yarl-1.9.4-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:a7409f968456111140c1c95301cadf071bd30a81cbd7ab829169fb9e3d72eae9"},
-    {file = "yarl-1.9.4-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:e23a6d84d9d1738dbc6e38167776107e63307dfc8ad108e580548d1f2c587f42"},
-    {file = "yarl-1.9.4-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:d8b889777de69897406c9fb0b76cdf2fd0f31267861ae7501d93003d55f54fbe"},
-    {file = "yarl-1.9.4-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:03caa9507d3d3c83bca08650678e25364e1843b484f19986a527630ca376ecce"},
-    {file = "yarl-1.9.4-cp311-cp311-musllinux_1_1_i686.whl", hash = "sha256:4e9035df8d0880b2f1c7f5031f33f69e071dfe72ee9310cfc76f7b605958ceb9"},
-    {file = "yarl-1.9.4-cp311-cp311-musllinux_1_1_ppc64le.whl", hash = "sha256:c0ec0ed476f77db9fb29bca17f0a8fcc7bc97ad4c6c1d8959c507decb22e8572"},
-    {file = "yarl-1.9.4-cp311-cp311-musllinux_1_1_s390x.whl", hash = "sha256:ee04010f26d5102399bd17f8df8bc38dc7ccd7701dc77f4a68c5b8d733406958"},
-    {file = "yarl-1.9.4-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:49a180c2e0743d5d6e0b4d1a9e5f633c62eca3f8a86ba5dd3c471060e352ca98"},
-    {file = "yarl-1.9.4-cp311-cp311-win32.whl", hash = "sha256:81eb57278deb6098a5b62e88ad8281b2ba09f2f1147c4767522353eaa6260b31"},
-    {file = "yarl-1.9.4-cp311-cp311-win_amd64.whl", hash = "sha256:d1d2532b340b692880261c15aee4dc94dd22ca5d61b9db9a8a361953d36410b1"},
-    {file = "yarl-1.9.4-cp312-cp312-macosx_10_9_universal2.whl", hash = "sha256:0d2454f0aef65ea81037759be5ca9947539667eecebca092733b2eb43c965a81"},
-    {file = "yarl-1.9.4-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:44d8ffbb9c06e5a7f529f38f53eda23e50d1ed33c6c869e01481d3fafa6b8142"},
-    {file = "yarl-1.9.4-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:aaaea1e536f98754a6e5c56091baa1b6ce2f2700cc4a00b0d49eca8dea471074"},
-    {file = "yarl-1.9.4-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:3777ce5536d17989c91696db1d459574e9a9bd37660ea7ee4d3344579bb6f129"},
-    {file = "yarl-1.9.4-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:9fc5fc1eeb029757349ad26bbc5880557389a03fa6ada41703db5e068881e5f2"},
-    {file = "yarl-1.9.4-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:ea65804b5dc88dacd4a40279af0cdadcfe74b3e5b4c897aa0d81cf86927fee78"},
-    {file = "yarl-1.9.4-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:aa102d6d280a5455ad6a0f9e6d769989638718e938a6a0a2ff3f4a7ff8c62cc4"},
-    {file = "yarl-1.9.4-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:09efe4615ada057ba2d30df871d2f668af661e971dfeedf0c159927d48bbeff0"},
-    {file = "yarl-1.9.4-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:008d3e808d03ef28542372d01057fd09168419cdc8f848efe2804f894ae03e51"},
-    {file = "yarl-1.9.4-cp312-cp312-musllinux_1_1_i686.whl", hash = "sha256:6f5cb257bc2ec58f437da2b37a8cd48f666db96d47b8a3115c29f316313654ff"},
-    {file = "yarl-1.9.4-cp312-cp312-musllinux_1_1_ppc64le.whl", hash = "sha256:992f18e0ea248ee03b5a6e8b3b4738850ae7dbb172cc41c966462801cbf62cf7"},
-    {file = "yarl-1.9.4-cp312-cp312-musllinux_1_1_s390x.whl", hash = "sha256:0e9d124c191d5b881060a9e5060627694c3bdd1fe24c5eecc8d5d7d0eb6faabc"},
-    {file = "yarl-1.9.4-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:3986b6f41ad22988e53d5778f91855dc0399b043fc8946d4f2e68af22ee9ff10"},
-    {file = "yarl-1.9.4-cp312-cp312-win32.whl", hash = "sha256:4b21516d181cd77ebd06ce160ef8cc2a5e9ad35fb1c5930882baff5ac865eee7"},
-    {file = "yarl-1.9.4-cp312-cp312-win_amd64.whl", hash = "sha256:a9bd00dc3bc395a662900f33f74feb3e757429e545d831eef5bb280252631984"},
-    {file = "yarl-1.9.4-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:63b20738b5aac74e239622d2fe30df4fca4942a86e31bf47a81a0e94c14df94f"},
-    {file = "yarl-1.9.4-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d7d7f7de27b8944f1fee2c26a88b4dabc2409d2fea7a9ed3df79b67277644e17"},
-    {file = "yarl-1.9.4-cp37-cp37m-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:c74018551e31269d56fab81a728f683667e7c28c04e807ba08f8c9e3bba32f14"},
-    {file = "yarl-1.9.4-cp37-cp37m-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:ca06675212f94e7a610e85ca36948bb8fc023e458dd6c63ef71abfd482481aa5"},
-    {file = "yarl-1.9.4-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:5aef935237d60a51a62b86249839b51345f47564208c6ee615ed2a40878dccdd"},
-    {file = "yarl-1.9.4-cp37-cp37m-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:2b134fd795e2322b7684155b7855cc99409d10b2e408056db2b93b51a52accc7"},
-    {file = "yarl-1.9.4-cp37-cp37m-musllinux_1_1_aarch64.whl", hash = "sha256:d25039a474c4c72a5ad4b52495056f843a7ff07b632c1b92ea9043a3d9950f6e"},
-    {file = "yarl-1.9.4-cp37-cp37m-musllinux_1_1_i686.whl", hash = "sha256:f7d6b36dd2e029b6bcb8a13cf19664c7b8e19ab3a58e0fefbb5b8461447ed5ec"},
-    {file = "yarl-1.9.4-cp37-cp37m-musllinux_1_1_ppc64le.whl", hash = "sha256:957b4774373cf6f709359e5c8c4a0af9f6d7875db657adb0feaf8d6cb3c3964c"},
-    {file = "yarl-1.9.4-cp37-cp37m-musllinux_1_1_s390x.whl", hash = "sha256:d7eeb6d22331e2fd42fce928a81c697c9ee2d51400bd1a28803965883e13cead"},
-    {file = "yarl-1.9.4-cp37-cp37m-musllinux_1_1_x86_64.whl", hash = "sha256:6a962e04b8f91f8c4e5917e518d17958e3bdee71fd1d8b88cdce74dd0ebbf434"},
-    {file = "yarl-1.9.4-cp37-cp37m-win32.whl", hash = "sha256:f3bc6af6e2b8f92eced34ef6a96ffb248e863af20ef4fde9448cc8c9b858b749"},
-    {file = "yarl-1.9.4-cp37-cp37m-win_amd64.whl", hash = "sha256:ad4d7a90a92e528aadf4965d685c17dacff3df282db1121136c382dc0b6014d2"},
-    {file = "yarl-1.9.4-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:ec61d826d80fc293ed46c9dd26995921e3a82146feacd952ef0757236fc137be"},
-    {file = "yarl-1.9.4-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:8be9e837ea9113676e5754b43b940b50cce76d9ed7d2461df1af39a8ee674d9f"},
-    {file = "yarl-1.9.4-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:bef596fdaa8f26e3d66af846bbe77057237cb6e8efff8cd7cc8dff9a62278bbf"},
-    {file = "yarl-1.9.4-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:2d47552b6e52c3319fede1b60b3de120fe83bde9b7bddad11a69fb0af7db32f1"},
-    {file = "yarl-1.9.4-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:84fc30f71689d7fc9168b92788abc977dc8cefa806909565fc2951d02f6b7d57"},
-    {file = "yarl-1.9.4-cp38-cp38-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:4aa9741085f635934f3a2583e16fcf62ba835719a8b2b28fb2917bb0537c1dfa"},
-    {file = "yarl-1.9.4-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:206a55215e6d05dbc6c98ce598a59e6fbd0c493e2de4ea6cc2f4934d5a18d130"},
-    {file = "yarl-1.9.4-cp38-cp38-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:07574b007ee20e5c375a8fe4a0789fad26db905f9813be0f9fef5a68080de559"},
-    {file = "yarl-1.9.4-cp38-cp38-musllinux_1_1_aarch64.whl", hash = "sha256:5a2e2433eb9344a163aced6a5f6c9222c0786e5a9e9cac2c89f0b28433f56e23"},
-    {file = "yarl-1.9.4-cp38-cp38-musllinux_1_1_i686.whl", hash = "sha256:6ad6d10ed9b67a382b45f29ea028f92d25bc0bc1daf6c5b801b90b5aa70fb9ec"},
-    {file = "yarl-1.9.4-cp38-cp38-musllinux_1_1_ppc64le.whl", hash = "sha256:6fe79f998a4052d79e1c30eeb7d6c1c1056ad33300f682465e1b4e9b5a188b78"},
-    {file = "yarl-1.9.4-cp38-cp38-musllinux_1_1_s390x.whl", hash = "sha256:a825ec844298c791fd28ed14ed1bffc56a98d15b8c58a20e0e08c1f5f2bea1be"},
-    {file = "yarl-1.9.4-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:8619d6915b3b0b34420cf9b2bb6d81ef59d984cb0fde7544e9ece32b4b3043c3"},
-    {file = "yarl-1.9.4-cp38-cp38-win32.whl", hash = "sha256:686a0c2f85f83463272ddffd4deb5e591c98aac1897d65e92319f729c320eece"},
-    {file = "yarl-1.9.4-cp38-cp38-win_amd64.whl", hash = "sha256:a00862fb23195b6b8322f7d781b0dc1d82cb3bcac346d1e38689370cc1cc398b"},
-    {file = "yarl-1.9.4-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:604f31d97fa493083ea21bd9b92c419012531c4e17ea6da0f65cacdcf5d0bd27"},
-    {file = "yarl-1.9.4-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:8a854227cf581330ffa2c4824d96e52ee621dd571078a252c25e3a3b3d94a1b1"},
-    {file = "yarl-1.9.4-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:ba6f52cbc7809cd8d74604cce9c14868306ae4aa0282016b641c661f981a6e91"},
-    {file = "yarl-1.9.4-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a6327976c7c2f4ee6816eff196e25385ccc02cb81427952414a64811037bbc8b"},
-    {file = "yarl-1.9.4-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:8397a3817d7dcdd14bb266283cd1d6fc7264a48c186b986f32e86d86d35fbac5"},
-    {file = "yarl-1.9.4-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:e0381b4ce23ff92f8170080c97678040fc5b08da85e9e292292aba67fdac6c34"},
-    {file = "yarl-1.9.4-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:23d32a2594cb5d565d358a92e151315d1b2268bc10f4610d098f96b147370136"},
-    {file = "yarl-1.9.4-cp39-cp39-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:ddb2a5c08a4eaaba605340fdee8fc08e406c56617566d9643ad8bf6852778fc7"},
-    {file = "yarl-1.9.4-cp39-cp39-musllinux_1_1_aarch64.whl", hash = "sha256:26a1dc6285e03f3cc9e839a2da83bcbf31dcb0d004c72d0730e755b33466c30e"},
-    {file = "yarl-1.9.4-cp39-cp39-musllinux_1_1_i686.whl", hash = "sha256:18580f672e44ce1238b82f7fb87d727c4a131f3a9d33a5e0e82b793362bf18b4"},
-    {file = "yarl-1.9.4-cp39-cp39-musllinux_1_1_ppc64le.whl", hash = "sha256:29e0f83f37610f173eb7e7b5562dd71467993495e568e708d99e9d1944f561ec"},
-    {file = "yarl-1.9.4-cp39-cp39-musllinux_1_1_s390x.whl", hash = "sha256:1f23e4fe1e8794f74b6027d7cf19dc25f8b63af1483d91d595d4a07eca1fb26c"},
-    {file = "yarl-1.9.4-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:db8e58b9d79200c76956cefd14d5c90af54416ff5353c5bfd7cbe58818e26ef0"},
-    {file = "yarl-1.9.4-cp39-cp39-win32.whl", hash = "sha256:c7224cab95645c7ab53791022ae77a4509472613e839dab722a72abe5a684575"},
-    {file = "yarl-1.9.4-cp39-cp39-win_amd64.whl", hash = "sha256:824d6c50492add5da9374875ce72db7a0733b29c2394890aef23d533106e2b15"},
-    {file = "yarl-1.9.4-py3-none-any.whl", hash = "sha256:928cecb0ef9d5a7946eb6ff58417ad2fe9375762382f1bf5c55e61645f2c43ad"},
-    {file = "yarl-1.9.4.tar.gz", hash = "sha256:566db86717cf8080b99b58b083b773a908ae40f06681e87e589a976faf8246bf"},
+    {file = "yarl-1.17.2-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:93771146ef048b34201bfa382c2bf74c524980870bb278e6df515efaf93699ff"},
+    {file = "yarl-1.17.2-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:8281db240a1616af2f9c5f71d355057e73a1409c4648c8949901396dc0a3c151"},
+    {file = "yarl-1.17.2-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:170ed4971bf9058582b01a8338605f4d8c849bd88834061e60e83b52d0c76870"},
+    {file = "yarl-1.17.2-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:bc61b005f6521fcc00ca0d1243559a5850b9dd1e1fe07b891410ee8fe192d0c0"},
+    {file = "yarl-1.17.2-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:871e1b47eec7b6df76b23c642a81db5dd6536cbef26b7e80e7c56c2fd371382e"},
+    {file = "yarl-1.17.2-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:3a58a2f2ca7aaf22b265388d40232f453f67a6def7355a840b98c2d547bd037f"},
+    {file = "yarl-1.17.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:736bb076f7299c5c55dfef3eb9e96071a795cb08052822c2bb349b06f4cb2e0a"},
+    {file = "yarl-1.17.2-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:8fd51299e21da709eabcd5b2dd60e39090804431292daacbee8d3dabe39a6bc0"},
+    {file = "yarl-1.17.2-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:358dc7ddf25e79e1cc8ee16d970c23faee84d532b873519c5036dbb858965795"},
+    {file = "yarl-1.17.2-cp310-cp310-musllinux_1_2_armv7l.whl", hash = "sha256:50d866f7b1a3f16f98603e095f24c0eeba25eb508c85a2c5939c8b3870ba2df8"},
+    {file = "yarl-1.17.2-cp310-cp310-musllinux_1_2_i686.whl", hash = "sha256:8b9c4643e7d843a0dca9cd9d610a0876e90a1b2cbc4c5ba7930a0d90baf6903f"},
+    {file = "yarl-1.17.2-cp310-cp310-musllinux_1_2_ppc64le.whl", hash = "sha256:d63123bfd0dce5f91101e77c8a5427c3872501acece8c90df457b486bc1acd47"},
+    {file = "yarl-1.17.2-cp310-cp310-musllinux_1_2_s390x.whl", hash = "sha256:4e76381be3d8ff96a4e6c77815653063e87555981329cf8f85e5be5abf449021"},
+    {file = "yarl-1.17.2-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:734144cd2bd633a1516948e477ff6c835041c0536cef1d5b9a823ae29899665b"},
+    {file = "yarl-1.17.2-cp310-cp310-win32.whl", hash = "sha256:26bfb6226e0c157af5da16d2d62258f1ac578d2899130a50433ffee4a5dfa673"},
+    {file = "yarl-1.17.2-cp310-cp310-win_amd64.whl", hash = "sha256:76499469dcc24759399accd85ec27f237d52dec300daaca46a5352fcbebb1071"},
+    {file = "yarl-1.17.2-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:792155279dc093839e43f85ff7b9b6493a8eaa0af1f94f1f9c6e8f4de8c63500"},
+    {file = "yarl-1.17.2-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:38bc4ed5cae853409cb193c87c86cd0bc8d3a70fd2268a9807217b9176093ac6"},
+    {file = "yarl-1.17.2-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:4a8c83f6fcdc327783bdc737e8e45b2e909b7bd108c4da1892d3bc59c04a6d84"},
+    {file = "yarl-1.17.2-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:8c6d5fed96f0646bfdf698b0a1cebf32b8aae6892d1bec0c5d2d6e2df44e1e2d"},
+    {file = "yarl-1.17.2-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:782ca9c58f5c491c7afa55518542b2b005caedaf4685ec814fadfcee51f02493"},
+    {file = "yarl-1.17.2-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:ff6af03cac0d1a4c3c19e5dcc4c05252411bf44ccaa2485e20d0a7c77892ab6e"},
+    {file = "yarl-1.17.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6a3f47930fbbed0f6377639503848134c4aa25426b08778d641491131351c2c8"},
+    {file = "yarl-1.17.2-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:d1fa68a3c921365c5745b4bd3af6221ae1f0ea1bf04b69e94eda60e57958907f"},
+    {file = "yarl-1.17.2-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:187df91395c11e9f9dc69b38d12406df85aa5865f1766a47907b1cc9855b6303"},
+    {file = "yarl-1.17.2-cp311-cp311-musllinux_1_2_armv7l.whl", hash = "sha256:93d1c8cc5bf5df401015c5e2a3ce75a5254a9839e5039c881365d2a9dcfc6dc2"},
+    {file = "yarl-1.17.2-cp311-cp311-musllinux_1_2_i686.whl", hash = "sha256:11d86c6145ac5c706c53d484784cf504d7d10fa407cb73b9d20f09ff986059ef"},
+    {file = "yarl-1.17.2-cp311-cp311-musllinux_1_2_ppc64le.whl", hash = "sha256:c42774d1d1508ec48c3ed29e7b110e33f5e74a20957ea16197dbcce8be6b52ba"},
+    {file = "yarl-1.17.2-cp311-cp311-musllinux_1_2_s390x.whl", hash = "sha256:0c8e589379ef0407b10bed16cc26e7392ef8f86961a706ade0a22309a45414d7"},
+    {file = "yarl-1.17.2-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:1056cadd5e850a1c026f28e0704ab0a94daaa8f887ece8dfed30f88befb87bb0"},
+    {file = "yarl-1.17.2-cp311-cp311-win32.whl", hash = "sha256:be4c7b1c49d9917c6e95258d3d07f43cfba2c69a6929816e77daf322aaba6628"},
+    {file = "yarl-1.17.2-cp311-cp311-win_amd64.whl", hash = "sha256:ac8eda86cc75859093e9ce390d423aba968f50cf0e481e6c7d7d63f90bae5c9c"},
+    {file = "yarl-1.17.2-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:dd90238d3a77a0e07d4d6ffdebc0c21a9787c5953a508a2231b5f191455f31e9"},
+    {file = "yarl-1.17.2-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:c74f0b0472ac40b04e6d28532f55cac8090e34c3e81f118d12843e6df14d0909"},
+    {file = "yarl-1.17.2-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:4d486ddcaca8c68455aa01cf53d28d413fb41a35afc9f6594a730c9779545876"},
+    {file = "yarl-1.17.2-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f25b7e93f5414b9a983e1a6c1820142c13e1782cc9ed354c25e933aebe97fcf2"},
+    {file = "yarl-1.17.2-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:3a0baff7827a632204060f48dca9e63fbd6a5a0b8790c1a2adfb25dc2c9c0d50"},
+    {file = "yarl-1.17.2-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:460024cacfc3246cc4d9f47a7fc860e4fcea7d1dc651e1256510d8c3c9c7cde0"},
+    {file = "yarl-1.17.2-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:5870d620b23b956f72bafed6a0ba9a62edb5f2ef78a8849b7615bd9433384171"},
+    {file = "yarl-1.17.2-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:2941756754a10e799e5b87e2319bbec481ed0957421fba0e7b9fb1c11e40509f"},
+    {file = "yarl-1.17.2-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:9611b83810a74a46be88847e0ea616794c406dbcb4e25405e52bff8f4bee2d0a"},
+    {file = "yarl-1.17.2-cp312-cp312-musllinux_1_2_armv7l.whl", hash = "sha256:cd7e35818d2328b679a13268d9ea505c85cd773572ebb7a0da7ccbca77b6a52e"},
+    {file = "yarl-1.17.2-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:6b981316fcd940f085f646b822c2ff2b8b813cbd61281acad229ea3cbaabeb6b"},
+    {file = "yarl-1.17.2-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:688058e89f512fb7541cb85c2f149c292d3fa22f981d5a5453b40c5da49eb9e8"},
+    {file = "yarl-1.17.2-cp312-cp312-musllinux_1_2_s390x.whl", hash = "sha256:56afb44a12b0864d17b597210d63a5b88915d680f6484d8d202ed68ade38673d"},
+    {file = "yarl-1.17.2-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:17931dfbb84ae18b287279c1f92b76a3abcd9a49cd69b92e946035cff06bcd20"},
+    {file = "yarl-1.17.2-cp312-cp312-win32.whl", hash = "sha256:ff8d95e06546c3a8c188f68040e9d0360feb67ba8498baf018918f669f7bc39b"},
+    {file = "yarl-1.17.2-cp312-cp312-win_amd64.whl", hash = "sha256:4c840cc11163d3c01a9d8aad227683c48cd3e5be5a785921bcc2a8b4b758c4f3"},
+    {file = "yarl-1.17.2-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:3294f787a437cb5d81846de3a6697f0c35ecff37a932d73b1fe62490bef69211"},
+    {file = "yarl-1.17.2-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:f1e7fedb09c059efee2533119666ca7e1a2610072076926fa028c2ba5dfeb78c"},
+    {file = "yarl-1.17.2-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:da9d3061e61e5ae3f753654813bc1cd1c70e02fb72cf871bd6daf78443e9e2b1"},
+    {file = "yarl-1.17.2-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:91c012dceadc695ccf69301bfdccd1fc4472ad714fe2dd3c5ab4d2046afddf29"},
+    {file = "yarl-1.17.2-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f11fd61d72d93ac23718d393d2a64469af40be2116b24da0a4ca6922df26807e"},
+    {file = "yarl-1.17.2-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:46c465ad06971abcf46dd532f77560181387b4eea59084434bdff97524444032"},
+    {file = "yarl-1.17.2-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ef6eee1a61638d29cd7c85f7fd3ac7b22b4c0fabc8fd00a712b727a3e73b0685"},
+    {file = "yarl-1.17.2-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:4434b739a8a101a837caeaa0137e0e38cb4ea561f39cb8960f3b1e7f4967a3fc"},
+    {file = "yarl-1.17.2-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:752485cbbb50c1e20908450ff4f94217acba9358ebdce0d8106510859d6eb19a"},
+    {file = "yarl-1.17.2-cp313-cp313-musllinux_1_2_armv7l.whl", hash = "sha256:17791acaa0c0f89323c57da7b9a79f2174e26d5debbc8c02d84ebd80c2b7bff8"},
+    {file = "yarl-1.17.2-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:5c6ea72fe619fee5e6b5d4040a451d45d8175f560b11b3d3e044cd24b2720526"},
+    {file = "yarl-1.17.2-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:db5ac3871ed76340210fe028f535392f097fb31b875354bcb69162bba2632ef4"},
+    {file = "yarl-1.17.2-cp313-cp313-musllinux_1_2_s390x.whl", hash = "sha256:7a1606ba68e311576bcb1672b2a1543417e7e0aa4c85e9e718ba6466952476c0"},
+    {file = "yarl-1.17.2-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:9bc27dd5cfdbe3dc7f381b05e6260ca6da41931a6e582267d5ca540270afeeb2"},
+    {file = "yarl-1.17.2-cp313-cp313-win32.whl", hash = "sha256:52492b87d5877ec405542f43cd3da80bdcb2d0c2fbc73236526e5f2c28e6db28"},
+    {file = "yarl-1.17.2-cp313-cp313-win_amd64.whl", hash = "sha256:8e1bf59e035534ba4077f5361d8d5d9194149f9ed4f823d1ee29ef3e8964ace3"},
+    {file = "yarl-1.17.2-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:c556fbc6820b6e2cda1ca675c5fa5589cf188f8da6b33e9fc05b002e603e44fa"},
+    {file = "yarl-1.17.2-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:f2f44a4247461965fed18b2573f3a9eb5e2c3cad225201ee858726cde610daca"},
+    {file = "yarl-1.17.2-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:3a3ede8c248f36b60227eb777eac1dbc2f1022dc4d741b177c4379ca8e75571a"},
+    {file = "yarl-1.17.2-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:2654caaf5584449d49c94a6b382b3cb4a246c090e72453493ea168b931206a4d"},
+    {file = "yarl-1.17.2-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:0d41c684f286ce41fa05ab6af70f32d6da1b6f0457459a56cf9e393c1c0b2217"},
+    {file = "yarl-1.17.2-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:2270d590997445a0dc29afa92e5534bfea76ba3aea026289e811bf9ed4b65a7f"},
+    {file = "yarl-1.17.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:18662443c6c3707e2fc7fad184b4dc32dd428710bbe72e1bce7fe1988d4aa654"},
+    {file = "yarl-1.17.2-cp39-cp39-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:75ac158560dec3ed72f6d604c81090ec44529cfb8169b05ae6fcb3e986b325d9"},
+    {file = "yarl-1.17.2-cp39-cp39-musllinux_1_2_aarch64.whl", hash = "sha256:1fee66b32e79264f428dc8da18396ad59cc48eef3c9c13844adec890cd339db5"},
+    {file = "yarl-1.17.2-cp39-cp39-musllinux_1_2_armv7l.whl", hash = "sha256:585ce7cd97be8f538345de47b279b879e091c8b86d9dbc6d98a96a7ad78876a3"},
+    {file = "yarl-1.17.2-cp39-cp39-musllinux_1_2_i686.whl", hash = "sha256:c019abc2eca67dfa4d8fb72ba924871d764ec3c92b86d5b53b405ad3d6aa56b0"},
+    {file = "yarl-1.17.2-cp39-cp39-musllinux_1_2_ppc64le.whl", hash = "sha256:c6e659b9a24d145e271c2faf3fa6dd1fcb3e5d3f4e17273d9e0350b6ab0fe6e2"},
+    {file = "yarl-1.17.2-cp39-cp39-musllinux_1_2_s390x.whl", hash = "sha256:d17832ba39374134c10e82d137e372b5f7478c4cceeb19d02ae3e3d1daed8721"},
+    {file = "yarl-1.17.2-cp39-cp39-musllinux_1_2_x86_64.whl", hash = "sha256:bc3003710e335e3f842ae3fd78efa55f11a863a89a72e9a07da214db3bf7e1f8"},
+    {file = "yarl-1.17.2-cp39-cp39-win32.whl", hash = "sha256:f5ffc6b7ace5b22d9e73b2a4c7305740a339fbd55301d52735f73e21d9eb3130"},
+    {file = "yarl-1.17.2-cp39-cp39-win_amd64.whl", hash = "sha256:48e424347a45568413deec6f6ee2d720de2cc0385019bedf44cd93e8638aa0ed"},
+    {file = "yarl-1.17.2-py3-none-any.whl", hash = "sha256:dd7abf4f717e33b7487121faf23560b3a50924f80e4bef62b22dab441ded8f3b"},
+    {file = "yarl-1.17.2.tar.gz", hash = "sha256:753eaaa0c7195244c84b5cc159dc8204b7fd99f716f11198f999f2332a86b178"},
 ]
 
 [package.dependencies]
 idna = ">=2.0"
 multidict = ">=4.0"
+propcache = ">=0.2.0"
 
 [[package]]
 name = "zipp"
@@ -3484,4 +3599,4 @@ cffi = ["cffi (>=1.11)"]
 [metadata]
 lock-version = "2.0"
 python-versions = "^3.9"
-content-hash = "c656496f9fbb7c29b2df3143c1d72c95b5e121cb6340134c0b8d070f54a08508"
+content-hash = "8cb9c38d83eec441391c0528ac2fbefde18c734373b2399e07c69382044e8ced"

proxy/src/auth/backend/classic.rs

@@ -1,17 +1,17 @@
 use tokio::io::{AsyncRead, AsyncWrite};
-use tracing::{info, warn};
+use tracing::{debug, info, warn};
 
 use super::{ComputeCredentials, ComputeUserInfo};
 use crate::auth::backend::ComputeCredentialKeys;
 use crate::auth::{self, AuthFlow};
 use crate::config::AuthenticationConfig;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::control_plane::AuthSecret;
 use crate::stream::{PqStream, Stream};
 use crate::{compute, sasl};
 
 pub(super) async fn authenticate(
-    ctx: &RequestMonitoring,
+    ctx: &RequestContext,
     creds: ComputeUserInfo,
     client: &mut PqStream<Stream<impl AsyncRead + AsyncWrite + Unpin>>,
     config: &'static AuthenticationConfig,
@@ -21,11 +21,11 @@ pub(super) async fn authenticate(
     let scram_keys = match secret {
         #[cfg(any(test, feature = "testing"))]
         AuthSecret::Md5(_) => {
-            info!("auth endpoint chooses MD5");
+            debug!("auth endpoint chooses MD5");
             return Err(auth::AuthError::bad_auth_method("MD5"));
         }
         AuthSecret::Scram(secret) => {
-            info!("auth endpoint chooses SCRAM");
+            debug!("auth endpoint chooses SCRAM");
             let scram = auth::Scram(&secret, ctx);
 
             let auth_outcome = tokio::time::timeout(
@@ -50,6 +50,8 @@ pub(super) async fn authenticate(
             let client_key = match auth_outcome {
                 sasl::Outcome::Success(key) => key,
                 sasl::Outcome::Failure(reason) => {
+                    // TODO: warnings?
+                    // TODO: should we get rid of this because double logging?
                     info!("auth backend failed with an error: {reason}");
                     return Err(auth::AuthError::password_failed(&*creds.user));
                 }

proxy/src/auth/backend/console_redirect.rs

@@ -8,7 +8,7 @@ use tracing::{info, info_span};
 use super::ComputeCredentialKeys;
 use crate::cache::Cached;
 use crate::config::AuthenticationConfig;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::control_plane::{self, CachedNodeInfo, NodeInfo};
 use crate::error::{ReportableError, UserFacingError};
 use crate::proxy::connect_compute::ComputeConnectBackend;
@@ -71,7 +71,7 @@ impl ConsoleRedirectBackend {
 
     pub(crate) async fn authenticate(
         &self,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         auth_config: &'static AuthenticationConfig,
         client: &mut PqStream<impl AsyncRead + AsyncWrite + Unpin>,
     ) -> auth::Result<ConsoleRedirectNodeInfo> {
@@ -87,7 +87,7 @@ pub struct ConsoleRedirectNodeInfo(pub(super) NodeInfo);
 impl ComputeConnectBackend for ConsoleRedirectNodeInfo {
     async fn wake_compute(
         &self,
-        _ctx: &RequestMonitoring,
+        _ctx: &RequestContext,
     ) -> Result<CachedNodeInfo, control_plane::errors::WakeComputeError> {
         Ok(Cached::new_uncached(self.0.clone()))
     }
@@ -98,7 +98,7 @@ impl ComputeConnectBackend for ConsoleRedirectNodeInfo {
 }
 
 async fn authenticate(
-    ctx: &RequestMonitoring,
+    ctx: &RequestContext,
     auth_config: &'static AuthenticationConfig,
     link_uri: &reqwest::Url,
     client: &mut PqStream<impl AsyncRead + AsyncWrite + Unpin>,

proxy/src/auth/backend/hacks.rs

@@ -4,7 +4,7 @@ use tracing::{debug, info};
 use super::{ComputeCredentials, ComputeUserInfo, ComputeUserInfoNoEndpoint};
 use crate::auth::{self, AuthFlow};
 use crate::config::AuthenticationConfig;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::control_plane::AuthSecret;
 use crate::intern::EndpointIdInt;
 use crate::sasl;
@@ -15,7 +15,7 @@ use crate::stream::{self, Stream};
 /// These properties are benefical for serverless JS workers, so we
 /// use this mechanism for websocket connections.
 pub(crate) async fn authenticate_cleartext(
-    ctx: &RequestMonitoring,
+    ctx: &RequestContext,
     info: ComputeUserInfo,
     client: &mut stream::PqStream<Stream<impl AsyncRead + AsyncWrite + Unpin>>,
     secret: AuthSecret,
@@ -57,7 +57,7 @@ pub(crate) async fn authenticate_cleartext(
 /// Similar to [`authenticate_cleartext`], but there's a specific password format,
 /// and passwords are not yet validated (we don't know how to validate them!)
 pub(crate) async fn password_hack_no_authentication(
-    ctx: &RequestMonitoring,
+    ctx: &RequestContext,
     info: ComputeUserInfoNoEndpoint,
     client: &mut stream::PqStream<Stream<impl AsyncRead + AsyncWrite + Unpin>>,
 ) -> auth::Result<(ComputeUserInfo, Vec<u8>)> {
@@ -73,7 +73,7 @@ pub(crate) async fn password_hack_no_authentication(
         .get_password()
         .await?;
 
-    info!(project = &*payload.endpoint, "received missing parameter");
+    debug!(project = &*payload.endpoint, "received missing parameter");
 
     // Report tentative success; compute node will check the password anyway.
     Ok((

proxy/src/auth/backend/jwt.rs

@@ -17,7 +17,7 @@ use thiserror::Error;
 use tokio::time::Instant;
 
 use crate::auth::backend::ComputeCredentialKeys;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::control_plane::errors::GetEndpointJwksError;
 use crate::http::read_body_with_limit;
 use crate::intern::RoleNameInt;
@@ -39,7 +39,7 @@ const JWKS_FETCH_RETRIES: u32 = 3;
 pub(crate) trait FetchAuthRules: Clone + Send + Sync + 'static {
     fn fetch_auth_rules(
         &self,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         endpoint: EndpointId,
     ) -> impl Future<Output = Result<Vec<AuthRule>, FetchAuthRulesError>> + Send;
 }
@@ -144,7 +144,7 @@ impl JwkCacheEntryLock {
     async fn renew_jwks<F: FetchAuthRules>(
         &self,
         _permit: JwkRenewalPermit<'_>,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         client: &reqwest_middleware::ClientWithMiddleware,
         endpoint: EndpointId,
         auth_rules: &F,
@@ -261,7 +261,7 @@ impl JwkCacheEntryLock {
 
     async fn get_or_update_jwk_cache<F: FetchAuthRules>(
         self: &Arc<Self>,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         client: &reqwest_middleware::ClientWithMiddleware,
         endpoint: EndpointId,
         fetch: &F,
@@ -314,7 +314,7 @@ impl JwkCacheEntryLock {
 
     async fn check_jwt<F: FetchAuthRules>(
         self: &Arc<Self>,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         jwt: &str,
         client: &reqwest_middleware::ClientWithMiddleware,
         endpoint: EndpointId,
@@ -409,7 +409,7 @@ impl JwkCacheEntryLock {
 impl JwkCache {
     pub(crate) async fn check_jwt<F: FetchAuthRules>(
         &self,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         endpoint: EndpointId,
         role_name: &RoleName,
         fetch: &F,
@@ -941,7 +941,7 @@ X0n5X2/pBLJzxZc62ccvZYVnctBiFs6HbSnxpuMQCfkt/BcR/ttIepBQQIW86wHL
     impl FetchAuthRules for Fetch {
         async fn fetch_auth_rules(
             &self,
-            _ctx: &RequestMonitoring,
+            _ctx: &RequestContext,
             _endpoint: EndpointId,
         ) -> Result<Vec<AuthRule>, FetchAuthRulesError> {
             Ok(self.0.clone())
@@ -1039,7 +1039,7 @@ X0n5X2/pBLJzxZc62ccvZYVnctBiFs6HbSnxpuMQCfkt/BcR/ttIepBQQIW86wHL
             for token in &tokens {
                 jwk_cache
                     .check_jwt(
-                        &RequestMonitoring::test(),
+                        &RequestContext::test(),
                         endpoint.clone(),
                         role,
                         &fetch,
@@ -1097,7 +1097,7 @@ X0n5X2/pBLJzxZc62ccvZYVnctBiFs6HbSnxpuMQCfkt/BcR/ttIepBQQIW86wHL
 
         jwk_cache
             .check_jwt(
-                &RequestMonitoring::test(),
+                &RequestContext::test(),
                 endpoint.clone(),
                 &role_name,
                 &fetch,
@@ -1136,7 +1136,7 @@ X0n5X2/pBLJzxZc62ccvZYVnctBiFs6HbSnxpuMQCfkt/BcR/ttIepBQQIW86wHL
 
         let ep = EndpointId::from("ep");
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let err = jwk_cache
             .check_jwt(&ctx, ep, &role, &fetch, &bad_jwt)
             .await
@@ -1175,7 +1175,7 @@ X0n5X2/pBLJzxZc62ccvZYVnctBiFs6HbSnxpuMQCfkt/BcR/ttIepBQQIW86wHL
         // this role_name is not accepted
         let bad_role_name = RoleName::from("cloud_admin");
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let err = jwk_cache
             .check_jwt(&ctx, ep, &bad_role_name, &fetch, &jwt)
             .await
@@ -1268,7 +1268,7 @@ X0n5X2/pBLJzxZc62ccvZYVnctBiFs6HbSnxpuMQCfkt/BcR/ttIepBQQIW86wHL
 
         let ep = EndpointId::from("ep");
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         for test in table {
             let jwt = new_custom_ec_jwt("1".into(), &key, test.body);
 
@@ -1336,7 +1336,7 @@ X0n5X2/pBLJzxZc62ccvZYVnctBiFs6HbSnxpuMQCfkt/BcR/ttIepBQQIW86wHL
 
         jwk_cache
             .check_jwt(
-                &RequestMonitoring::test(),
+                &RequestContext::test(),
                 endpoint.clone(),
                 &role_name,
                 &fetch,

proxy/src/auth/backend/local.rs

@@ -7,7 +7,7 @@ use super::jwt::{AuthRule, FetchAuthRules};
 use crate::auth::backend::jwt::FetchAuthRulesError;
 use crate::compute::ConnCfg;
 use crate::compute_ctl::ComputeCtlApi;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::control_plane::messages::{ColdStartInfo, EndpointJwksResponse, MetricsAuxInfo};
 use crate::control_plane::NodeInfo;
 use crate::http;
@@ -56,7 +56,7 @@ pub static JWKS_ROLE_MAP: ArcSwapOption<EndpointJwksResponse> = ArcSwapOption::c
 impl FetchAuthRules for StaticAuthRules {
     async fn fetch_auth_rules(
         &self,
-        _ctx: &RequestMonitoring,
+        _ctx: &RequestContext,
         _endpoint: EndpointId,
     ) -> Result<Vec<AuthRule>, FetchAuthRulesError> {
         let mappings = JWKS_ROLE_MAP.load();

proxy/src/auth/backend/mod.rs

@@ -14,13 +14,13 @@ use ipnet::{Ipv4Net, Ipv6Net};
 use local::LocalBackend;
 use tokio::io::{AsyncRead, AsyncWrite};
 use tokio_postgres::config::AuthKeys;
-use tracing::{info, warn};
+use tracing::{debug, info, warn};
 
 use crate::auth::credentials::check_peer_addr_is_in_list;
 use crate::auth::{self, validate_password_and_exchange, AuthError, ComputeUserInfoMaybeEndpoint};
 use crate::cache::Cached;
 use crate::config::AuthenticationConfig;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::control_plane::client::ControlPlaneClient;
 use crate::control_plane::errors::GetAuthInfoError;
 use crate::control_plane::{
@@ -210,7 +210,7 @@ impl RateBucketInfo {
 impl AuthenticationConfig {
     pub(crate) fn check_rate_limit(
         &self,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         secret: AuthSecret,
         endpoint: &EndpointId,
         is_cleartext: bool,
@@ -265,7 +265,7 @@ impl AuthenticationConfig {
 ///
 /// All authentication flows will emit an AuthenticationOk message if successful.
 async fn auth_quirks(
-    ctx: &RequestMonitoring,
+    ctx: &RequestContext,
     api: &impl control_plane::ControlPlaneApi,
     user_info: ComputeUserInfoMaybeEndpoint,
     client: &mut stream::PqStream<Stream<impl AsyncRead + AsyncWrite + Unpin>>,
@@ -286,7 +286,7 @@ async fn auth_quirks(
         Ok(info) => (info, None),
     };
 
-    info!("fetching user's authentication info");
+    debug!("fetching user's authentication info");
     let (allowed_ips, maybe_secret) = api.get_allowed_ips_and_secret(ctx, &info).await?;
 
     // check allowed list
@@ -343,7 +343,7 @@ async fn auth_quirks(
 }
 
 async fn authenticate_with_secret(
-    ctx: &RequestMonitoring,
+    ctx: &RequestContext,
     secret: AuthSecret,
     info: ComputeUserInfo,
     client: &mut stream::PqStream<Stream<impl AsyncRead + AsyncWrite + Unpin>>,
@@ -396,7 +396,7 @@ impl<'a> Backend<'a, ComputeUserInfoMaybeEndpoint> {
     #[tracing::instrument(fields(allow_cleartext = allow_cleartext), skip_all)]
     pub(crate) async fn authenticate(
         self,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         client: &mut stream::PqStream<Stream<impl AsyncRead + AsyncWrite + Unpin>>,
         allow_cleartext: bool,
         config: &'static AuthenticationConfig,
@@ -404,7 +404,7 @@ impl<'a> Backend<'a, ComputeUserInfoMaybeEndpoint> {
     ) -> auth::Result<Backend<'a, ComputeCredentials>> {
         let res = match self {
             Self::ControlPlane(api, user_info) => {
-                info!(
+                debug!(
                     user = &*user_info.user,
                     project = user_info.endpoint(),
                     "performing authentication using the console"
@@ -427,6 +427,7 @@ impl<'a> Backend<'a, ComputeUserInfoMaybeEndpoint> {
             }
         };
 
+        // TODO: replace with some metric
         info!("user successfully authenticated");
         Ok(res)
     }
@@ -435,7 +436,7 @@ impl<'a> Backend<'a, ComputeUserInfoMaybeEndpoint> {
 impl Backend<'_, ComputeUserInfo> {
     pub(crate) async fn get_role_secret(
         &self,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
     ) -> Result<CachedRoleSecret, GetAuthInfoError> {
         match self {
             Self::ControlPlane(api, user_info) => api.get_role_secret(ctx, user_info).await,
@@ -445,7 +446,7 @@ impl Backend<'_, ComputeUserInfo> {
 
     pub(crate) async fn get_allowed_ips_and_secret(
         &self,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
     ) -> Result<(CachedAllowedIps, Option<CachedRoleSecret>), GetAuthInfoError> {
         match self {
             Self::ControlPlane(api, user_info) => {
@@ -460,7 +461,7 @@ impl Backend<'_, ComputeUserInfo> {
 impl ComputeConnectBackend for Backend<'_, ComputeCredentials> {
     async fn wake_compute(
         &self,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
     ) -> Result<CachedNodeInfo, control_plane::errors::WakeComputeError> {
         match self {
             Self::ControlPlane(api, creds) => api.wake_compute(ctx, &creds.info).await,
@@ -496,7 +497,7 @@ mod tests {
     use crate::auth::backend::MaskedIp;
     use crate::auth::{ComputeUserInfoMaybeEndpoint, IpPattern};
     use crate::config::AuthenticationConfig;
-    use crate::context::RequestMonitoring;
+    use crate::context::RequestContext;
     use crate::control_plane::{self, CachedAllowedIps, CachedNodeInfo, CachedRoleSecret};
     use crate::proxy::NeonOptions;
     use crate::rate_limiter::{EndpointRateLimiter, RateBucketInfo};
@@ -512,7 +513,7 @@ mod tests {
     impl control_plane::ControlPlaneApi for Auth {
         async fn get_role_secret(
             &self,
-            _ctx: &RequestMonitoring,
+            _ctx: &RequestContext,
             _user_info: &super::ComputeUserInfo,
         ) -> Result<CachedRoleSecret, control_plane::errors::GetAuthInfoError> {
             Ok(CachedRoleSecret::new_uncached(Some(self.secret.clone())))
@@ -520,7 +521,7 @@ mod tests {
 
         async fn get_allowed_ips_and_secret(
             &self,
-            _ctx: &RequestMonitoring,
+            _ctx: &RequestContext,
             _user_info: &super::ComputeUserInfo,
         ) -> Result<
             (CachedAllowedIps, Option<CachedRoleSecret>),
@@ -534,7 +535,7 @@ mod tests {
 
         async fn get_endpoint_jwks(
             &self,
-            _ctx: &RequestMonitoring,
+            _ctx: &RequestContext,
             _endpoint: crate::types::EndpointId,
         ) -> Result<Vec<super::jwt::AuthRule>, control_plane::errors::GetEndpointJwksError>
         {
@@ -543,7 +544,7 @@ mod tests {
 
         async fn wake_compute(
             &self,
-            _ctx: &RequestMonitoring,
+            _ctx: &RequestContext,
             _user_info: &super::ComputeUserInfo,
         ) -> Result<CachedNodeInfo, control_plane::errors::WakeComputeError> {
             unimplemented!()
@@ -622,7 +623,7 @@ mod tests {
         let (mut client, server) = tokio::io::duplex(1024);
         let mut stream = PqStream::new(Stream::from_raw(server));
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let api = Auth {
             ips: vec![],
             secret: AuthSecret::Scram(ServerSecret::build("my-secret-password").await.unwrap()),
@@ -699,7 +700,7 @@ mod tests {
         let (mut client, server) = tokio::io::duplex(1024);
         let mut stream = PqStream::new(Stream::from_raw(server));
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let api = Auth {
             ips: vec![],
             secret: AuthSecret::Scram(ServerSecret::build("my-secret-password").await.unwrap()),
@@ -751,7 +752,7 @@ mod tests {
         let (mut client, server) = tokio::io::duplex(1024);
         let mut stream = PqStream::new(Stream::from_raw(server));
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let api = Auth {
             ips: vec![],
             secret: AuthSecret::Scram(ServerSecret::build("my-secret-password").await.unwrap()),

proxy/src/auth/credentials.rs

@@ -7,10 +7,10 @@ use std::str::FromStr;
 use itertools::Itertools;
 use pq_proto::StartupMessageParams;
 use thiserror::Error;
-use tracing::{info, warn};
+use tracing::{debug, warn};
 
 use crate::auth::password_hack::parse_endpoint_param;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::error::{ReportableError, UserFacingError};
 use crate::metrics::{Metrics, SniKind};
 use crate::proxy::NeonOptions;
@@ -86,7 +86,7 @@ pub(crate) fn endpoint_sni(
 
 impl ComputeUserInfoMaybeEndpoint {
     pub(crate) fn parse(
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         params: &StartupMessageParams,
         sni: Option<&str>,
         common_names: Option<&HashSet<String>>,
@@ -147,22 +147,22 @@ impl ComputeUserInfoMaybeEndpoint {
         }
 
         let metrics = Metrics::get();
-        info!(%user, "credentials");
+        debug!(%user, "credentials");
         if sni.is_some() {
-            info!("Connection with sni");
+            debug!("Connection with sni");
             metrics.proxy.accepted_connections_by_sni.inc(SniKind::Sni);
         } else if endpoint.is_some() {
             metrics
                 .proxy
                 .accepted_connections_by_sni
                 .inc(SniKind::NoSni);
-            info!("Connection without sni");
+            debug!("Connection without sni");
         } else {
             metrics
                 .proxy
                 .accepted_connections_by_sni
                 .inc(SniKind::PasswordHack);
-            info!("Connection with password hack");
+            debug!("Connection with password hack");
         }
 
         let options = NeonOptions::parse_params(params);
@@ -260,7 +260,7 @@ mod tests {
     fn parse_bare_minimum() -> anyhow::Result<()> {
         // According to postgresql, only `user` should be required.
         let options = StartupMessageParams::new([("user", "john_doe")]);
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info = ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, None, None)?;
         assert_eq!(user_info.user, "john_doe");
         assert_eq!(user_info.endpoint_id, None);
@@ -275,7 +275,7 @@ mod tests {
             ("database", "world"), // should be ignored
             ("foo", "bar"),        // should be ignored
         ]);
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info = ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, None, None)?;
         assert_eq!(user_info.user, "john_doe");
         assert_eq!(user_info.endpoint_id, None);
@@ -290,7 +290,7 @@ mod tests {
         let sni = Some("foo.localhost");
         let common_names = Some(["localhost".into()].into());
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info =
             ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, sni, common_names.as_ref())?;
         assert_eq!(user_info.user, "john_doe");
@@ -307,7 +307,7 @@ mod tests {
             ("options", "-ckey=1 project=bar -c geqo=off"),
         ]);
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info = ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, None, None)?;
         assert_eq!(user_info.user, "john_doe");
         assert_eq!(user_info.endpoint_id.as_deref(), Some("bar"));
@@ -322,7 +322,7 @@ mod tests {
             ("options", "-ckey=1 endpoint=bar -c geqo=off"),
         ]);
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info = ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, None, None)?;
         assert_eq!(user_info.user, "john_doe");
         assert_eq!(user_info.endpoint_id.as_deref(), Some("bar"));
@@ -340,7 +340,7 @@ mod tests {
             ),
         ]);
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info = ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, None, None)?;
         assert_eq!(user_info.user, "john_doe");
         assert!(user_info.endpoint_id.is_none());
@@ -355,7 +355,7 @@ mod tests {
             ("options", "-ckey=1 endpoint=bar project=foo -c geqo=off"),
         ]);
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info = ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, None, None)?;
         assert_eq!(user_info.user, "john_doe");
         assert!(user_info.endpoint_id.is_none());
@@ -370,7 +370,7 @@ mod tests {
         let sni = Some("baz.localhost");
         let common_names = Some(["localhost".into()].into());
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info =
             ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, sni, common_names.as_ref())?;
         assert_eq!(user_info.user, "john_doe");
@@ -385,14 +385,14 @@ mod tests {
 
         let common_names = Some(["a.com".into(), "b.com".into()].into());
         let sni = Some("p1.a.com");
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info =
             ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, sni, common_names.as_ref())?;
         assert_eq!(user_info.endpoint_id.as_deref(), Some("p1"));
 
         let common_names = Some(["a.com".into(), "b.com".into()].into());
         let sni = Some("p1.b.com");
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info =
             ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, sni, common_names.as_ref())?;
         assert_eq!(user_info.endpoint_id.as_deref(), Some("p1"));
@@ -408,7 +408,7 @@ mod tests {
         let sni = Some("second.localhost");
         let common_names = Some(["localhost".into()].into());
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let err = ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, sni, common_names.as_ref())
             .expect_err("should fail");
         match err {
@@ -427,7 +427,7 @@ mod tests {
         let sni = Some("project.localhost");
         let common_names = Some(["example.com".into()].into());
 
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let err = ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, sni, common_names.as_ref())
             .expect_err("should fail");
         match err {
@@ -447,7 +447,7 @@ mod tests {
 
         let sni = Some("project.localhost");
         let common_names = Some(["localhost".into()].into());
-        let ctx = RequestMonitoring::test();
+        let ctx = RequestContext::test();
         let user_info =
             ComputeUserInfoMaybeEndpoint::parse(&ctx, &options, sni, common_names.as_ref())?;
         assert_eq!(user_info.endpoint_id.as_deref(), Some("project"));

proxy/src/auth/flow.rs

@@ -11,7 +11,7 @@ use tracing::info;
 use super::backend::ComputeCredentialKeys;
 use super::{AuthError, PasswordHackPayload};
 use crate::config::TlsServerEndPoint;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::control_plane::AuthSecret;
 use crate::intern::EndpointIdInt;
 use crate::sasl;
@@ -32,7 +32,7 @@ pub(crate) struct Begin;
 /// Use [SCRAM](crate::scram)-based auth in [`AuthFlow`].
 pub(crate) struct Scram<'a>(
     pub(crate) &'a scram::ServerSecret,
-    pub(crate) &'a RequestMonitoring,
+    pub(crate) &'a RequestContext,
 );
 
 impl AuthMethod for Scram<'_> {
@@ -178,6 +178,8 @@ impl<S: AsyncRead + AsyncWrite + Unpin> AuthFlow<'_, S, Scram<'_>> {
             SCRAM_SHA_256_PLUS => ctx.set_auth_method(crate::context::AuthMethod::ScramSha256Plus),
             _ => {}
         }
+
+        // TODO: make this a metric instead
         info!("client chooses {}", sasl.method);
 
         let outcome = sasl::SaslStream::new(self.stream, sasl.message)

proxy/src/bin/local_proxy.rs

@@ -32,11 +32,12 @@ project_git_version!(GIT_VERSION);
 project_build_tag!(BUILD_TAG);
 
 use clap::Parser;
+use thiserror::Error;
 use tokio::net::TcpListener;
 use tokio::sync::Notify;
 use tokio::task::JoinSet;
 use tokio_util::sync::CancellationToken;
-use tracing::{error, info, warn};
+use tracing::{debug, error, info, warn};
 use utils::sentry_init::init_sentry;
 use utils::{pid_file, project_build_tag, project_git_version};
 
@@ -110,7 +111,7 @@ struct SqlOverHttpArgs {
     sql_over_http_cancel_set_shards: usize,
 
     #[clap(long, default_value_t = 10 * 1024 * 1024)] // 10 MiB
-    sql_over_http_max_request_size_bytes: u64,
+    sql_over_http_max_request_size_bytes: usize,
 
     #[clap(long, default_value_t = 10 * 1024 * 1024)] // 10 MiB
     sql_over_http_max_response_size_bytes: usize,
@@ -124,8 +125,9 @@ async fn main() -> anyhow::Result<()> {
 
     Metrics::install(Arc::new(ThreadPoolMetrics::new(0)));
 
-    info!("Version: {GIT_VERSION}");
-    info!("Build_tag: {BUILD_TAG}");
+    // TODO: refactor these to use labels
+    debug!("Version: {GIT_VERSION}");
+    debug!("Build_tag: {BUILD_TAG}");
     let neon_metrics = ::metrics::NeonMetrics::new(::metrics::BuildInfo {
         revision: GIT_VERSION,
         build_tag: BUILD_TAG,
@@ -305,26 +307,46 @@ fn build_auth_backend(
     Ok(Box::leak(Box::new(auth_backend)))
 }
 
+#[derive(Error, Debug)]
+enum RefreshConfigError {
+    #[error(transparent)]
+    Read(#[from] std::io::Error),
+    #[error(transparent)]
+    Parse(#[from] serde_json::Error),
+    #[error(transparent)]
+    Validate(anyhow::Error),
+}
+
 async fn refresh_config_loop(path: Utf8PathBuf, rx: Arc<Notify>) {
+    let mut init = true;
     loop {
         rx.notified().await;
 
         match refresh_config_inner(&path).await {
             Ok(()) => {}
+            // don't log for file not found errors if this is the first time we are checking
+            // for computes that don't use local_proxy, this is not an error.
+            Err(RefreshConfigError::Read(e))
+                if init && e.kind() == std::io::ErrorKind::NotFound =>
+            {
+                debug!(error=?e, ?path, "could not read config file");
+            }
             Err(e) => {
                 error!(error=?e, ?path, "could not read config file");
             }
         }
+
+        init = false;
     }
 }
 
-async fn refresh_config_inner(path: &Utf8Path) -> anyhow::Result<()> {
+async fn refresh_config_inner(path: &Utf8Path) -> Result<(), RefreshConfigError> {
     let bytes = tokio::fs::read(&path).await?;
     let data: LocalProxySpec = serde_json::from_slice(&bytes)?;
 
     let mut jwks_set = vec![];
 
-    for jwks in data.jwks.into_iter().flatten() {
+    fn parse_jwks_settings(jwks: compute_api::spec::JwksSettings) -> anyhow::Result<JwksSettings> {
         let mut jwks_url = url::Url::from_str(&jwks.jwks_url).context("parsing JWKS url")?;
 
         ensure!(
@@ -367,7 +389,7 @@ async fn refresh_config_inner(path: &Utf8Path) -> anyhow::Result<()> {
             }
         }
 
-        jwks_set.push(JwksSettings {
+        Ok(JwksSettings {
             id: jwks.id,
             jwks_url,
             provider_name: jwks.provider_name,
@@ -381,6 +403,10 @@ async fn refresh_config_inner(path: &Utf8Path) -> anyhow::Result<()> {
         })
     }
 
+    for jwks in data.jwks.into_iter().flatten() {
+        jwks_set.push(parse_jwks_settings(jwks).map_err(RefreshConfigError::Validate)?);
+    }
+
     info!("successfully loaded new config");
     JWKS_ROLE_MAP.store(Some(Arc::new(EndpointJwksResponse { jwks: jwks_set })));
 

proxy/src/bin/pg_sni_router.rs

@@ -11,7 +11,7 @@ use futures::future::Either;
 use futures::TryFutureExt;
 use itertools::Itertools;
 use proxy::config::TlsServerEndPoint;
-use proxy::context::RequestMonitoring;
+use proxy::context::RequestContext;
 use proxy::metrics::{Metrics, ThreadPoolMetrics};
 use proxy::protocol2::ConnectionInfo;
 use proxy::proxy::{copy_bidirectional_client_compute, run_until_cancelled, ErrorSource};
@@ -177,7 +177,7 @@ async fn task_main(
                     .context("failed to set socket option")?;
 
                 info!(%peer_addr, "serving");
-                let ctx = RequestMonitoring::new(
+                let ctx = RequestContext::new(
                     session_id,
                     ConnectionInfo {
                         addr: peer_addr,
@@ -208,7 +208,7 @@ async fn task_main(
 const ERR_INSECURE_CONNECTION: &str = "connection is insecure (try using `sslmode=require`)";
 
 async fn ssl_handshake<S: AsyncRead + AsyncWrite + Unpin>(
-    ctx: &RequestMonitoring,
+    ctx: &RequestContext,
     raw_stream: S,
     tls_config: Arc<rustls::ServerConfig>,
     tls_server_end_point: TlsServerEndPoint,
@@ -259,7 +259,7 @@ async fn ssl_handshake<S: AsyncRead + AsyncWrite + Unpin>(
 }
 
 async fn handle_client(
-    ctx: RequestMonitoring,
+    ctx: RequestContext,
     dest_suffix: Arc<String>,
     tls_config: Arc<rustls::ServerConfig>,
     tls_server_end_point: TlsServerEndPoint,

proxy/src/bin/proxy.rs

@@ -276,7 +276,7 @@ struct SqlOverHttpArgs {
     sql_over_http_cancel_set_shards: usize,
 
     #[clap(long, default_value_t = 10 * 1024 * 1024)] // 10 MiB
-    sql_over_http_max_request_size_bytes: u64,
+    sql_over_http_max_request_size_bytes: usize,
 
     #[clap(long, default_value_t = 10 * 1024 * 1024)] // 10 MiB
     sql_over_http_max_response_size_bytes: usize,
@@ -288,6 +288,7 @@ async fn main() -> anyhow::Result<()> {
     let _panic_hook_guard = utils::logging::replace_panic_hook_with_tracing_panic_hook();
     let _sentry_guard = init_sentry(Some(GIT_VERSION.into()), &[]);
 
+    // TODO: refactor these to use labels
     info!("Version: {GIT_VERSION}");
     info!("Build_tag: {BUILD_TAG}");
     let neon_metrics = ::metrics::NeonMetrics::new(::metrics::BuildInfo {

proxy/src/bin/stream_events.rs

@@ -1,450 +0,0 @@
-use std::sync::Arc;
-
-use anyhow::bail;
-use aws_config::environment::EnvironmentVariableCredentialsProvider;
-use aws_config::imds::credentials::ImdsCredentialsProvider;
-use aws_config::meta::credentials::CredentialsProviderChain;
-use aws_config::meta::region::RegionProviderChain;
-use aws_config::profile::ProfileFileCredentialsProvider;
-use aws_config::provider_config::ProviderConfig;
-use aws_config::web_identity_token::WebIdentityTokenCredentialsProvider;
-use aws_config::Region;
-use clap::{Parser, ValueEnum};
-use proxy::config::{self, remote_storage_from_toml, ProxyProtocolV2};
-use proxy::context::parquet::ParquetUploadArgs;
-use proxy::rate_limiter::RateBucketInfo;
-use proxy::redis::connection_with_credentials_provider::ConnectionWithCredentialsProvider;
-use proxy::redis::elasticache;
-use redis::streams::{StreamReadOptions, StreamReadReply};
-use redis::{AsyncCommands, FromRedisValue, Value};
-use remote_storage::RemoteStorageConfig;
-use serde::{Deserialize, Serialize};
-use tracing::warn;
-
-#[global_allocator]
-static GLOBAL: tikv_jemallocator::Jemalloc = tikv_jemallocator::Jemalloc;
-
-#[derive(Clone, Debug, ValueEnum)]
-enum AuthBackendType {
-    #[value(name("console"), alias("cplane"))]
-    ControlPlane,
-
-    #[value(name("link"), alias("control-redirect"))]
-    ConsoleRedirect,
-
-    #[cfg(feature = "testing")]
-    Postgres,
-}
-
-/// Neon proxy/router
-#[derive(Parser)]
-struct ProxyCliArgs {
-    /// Name of the region this proxy is deployed in
-    #[clap(long, default_value_t = String::new())]
-    region: String,
-    /// listen for incoming client connections on ip:port
-    #[clap(short, long, default_value = "127.0.0.1:4432")]
-    proxy: String,
-    #[clap(value_enum, long, default_value_t = AuthBackendType::ConsoleRedirect)]
-    auth_backend: AuthBackendType,
-    /// listen for management callback connection on ip:port
-    #[clap(short, long, default_value = "127.0.0.1:7000")]
-    mgmt: String,
-    /// listen for incoming http connections (metrics, etc) on ip:port
-    #[clap(long, default_value = "127.0.0.1:7001")]
-    http: String,
-    /// listen for incoming wss connections on ip:port
-    #[clap(long)]
-    wss: Option<String>,
-    /// redirect unauthenticated users to the given uri in case of console redirect auth
-    #[clap(short, long, default_value = "http://localhost:3000/psql_session/")]
-    uri: String,
-    /// cloud API endpoint for authenticating users
-    #[clap(
-        short,
-        long,
-        default_value = "http://localhost:3000/authenticate_proxy_request/"
-    )]
-    auth_endpoint: String,
-    /// JWT used to connect to control plane.
-    #[clap(
-        long,
-        value_name = "JWT",
-        default_value = "",
-        env = "NEON_PROXY_TO_CONTROLPLANE_TOKEN"
-    )]
-    control_plane_token: Arc<str>,
-    /// if this is not local proxy, this toggles whether we accept jwt or passwords for http
-    #[clap(long, default_value_t = false, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
-    is_auth_broker: bool,
-    /// path to TLS key for client postgres connections
-    ///
-    /// tls-key and tls-cert are for backwards compatibility, we can put all certs in one dir
-    #[clap(short = 'k', long, alias = "ssl-key")]
-    tls_key: Option<String>,
-    /// path to TLS cert for client postgres connections
-    ///
-    /// tls-key and tls-cert are for backwards compatibility, we can put all certs in one dir
-    #[clap(short = 'c', long, alias = "ssl-cert")]
-    tls_cert: Option<String>,
-    /// path to directory with TLS certificates for client postgres connections
-    #[clap(long)]
-    certs_dir: Option<String>,
-    /// timeout for the TLS handshake
-    #[clap(long, default_value = "15s", value_parser = humantime::parse_duration)]
-    handshake_timeout: tokio::time::Duration,
-    /// http endpoint to receive periodic metric updates
-    #[clap(long)]
-    metric_collection_endpoint: Option<String>,
-    /// how often metrics should be sent to a collection endpoint
-    #[clap(long)]
-    metric_collection_interval: Option<String>,
-    /// cache for `wake_compute` api method (use `size=0` to disable)
-    #[clap(long, default_value = config::CacheOptions::CACHE_DEFAULT_OPTIONS)]
-    wake_compute_cache: String,
-    /// lock for `wake_compute` api method. example: "shards=32,permits=4,epoch=10m,timeout=1s". (use `permits=0` to disable).
-    #[clap(long, default_value = config::ConcurrencyLockOptions::DEFAULT_OPTIONS_WAKE_COMPUTE_LOCK)]
-    wake_compute_lock: String,
-    /// lock for `connect_compute` api method. example: "shards=32,permits=4,epoch=10m,timeout=1s". (use `permits=0` to disable).
-    #[clap(long, default_value = config::ConcurrencyLockOptions::DEFAULT_OPTIONS_CONNECT_COMPUTE_LOCK)]
-    connect_compute_lock: String,
-    /// Allow self-signed certificates for compute nodes (for testing)
-    #[clap(long, default_value_t = false, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
-    allow_self_signed_compute: bool,
-    #[clap(flatten)]
-    sql_over_http: SqlOverHttpArgs,
-    /// timeout for scram authentication protocol
-    #[clap(long, default_value = "15s", value_parser = humantime::parse_duration)]
-    scram_protocol_timeout: tokio::time::Duration,
-    /// size of the threadpool for password hashing
-    #[clap(long, default_value_t = 4)]
-    scram_thread_pool_size: u8,
-    /// Endpoint rate limiter max number of requests per second.
-    ///
-    /// Provided in the form `<Requests Per Second>@<Bucket Duration Size>`.
-    /// Can be given multiple times for different bucket sizes.
-    #[clap(long, default_values_t = RateBucketInfo::DEFAULT_ENDPOINT_SET)]
-    endpoint_rps_limit: Vec<RateBucketInfo>,
-    /// Wake compute rate limiter max number of requests per second.
-    #[clap(long, default_values_t = RateBucketInfo::DEFAULT_SET)]
-    wake_compute_limit: Vec<RateBucketInfo>,
-    /// Whether the auth rate limiter actually takes effect (for testing)
-    #[clap(long, default_value_t = false, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
-    auth_rate_limit_enabled: bool,
-    /// Authentication rate limiter max number of hashes per second.
-    #[clap(long, default_values_t = RateBucketInfo::DEFAULT_AUTH_SET)]
-    auth_rate_limit: Vec<RateBucketInfo>,
-    /// The IP subnet to use when considering whether two IP addresses are considered the same.
-    #[clap(long, default_value_t = 64)]
-    auth_rate_limit_ip_subnet: u8,
-    /// Redis rate limiter max number of requests per second.
-    #[clap(long, default_values_t = RateBucketInfo::DEFAULT_SET)]
-    redis_rps_limit: Vec<RateBucketInfo>,
-    /// cache for `allowed_ips` (use `size=0` to disable)
-    #[clap(long, default_value = config::CacheOptions::CACHE_DEFAULT_OPTIONS)]
-    allowed_ips_cache: String,
-    /// cache for `role_secret` (use `size=0` to disable)
-    #[clap(long, default_value = config::CacheOptions::CACHE_DEFAULT_OPTIONS)]
-    role_secret_cache: String,
-    /// redis url for notifications (if empty, redis_host:port will be used for both notifications and streaming connections)
-    #[clap(long)]
-    redis_notifications: Option<String>,
-    /// what from the available authentications type to use for the regional redis we have. Supported are "irsa" and "plain".
-    #[clap(long, default_value = "irsa")]
-    redis_auth_type: String,
-    /// redis host for streaming connections (might be different from the notifications host)
-    #[clap(long)]
-    redis_host: Option<String>,
-    /// redis port for streaming connections (might be different from the notifications host)
-    #[clap(long)]
-    redis_port: Option<u16>,
-    /// redis cluster name, used in aws elasticache
-    #[clap(long)]
-    redis_cluster_name: Option<String>,
-    /// redis user_id, used in aws elasticache
-    #[clap(long)]
-    redis_user_id: Option<String>,
-    /// aws region to retrieve credentials
-    #[clap(long, default_value_t = String::new())]
-    aws_region: String,
-    /// cache for `project_info` (use `size=0` to disable)
-    #[clap(long, default_value = config::ProjectInfoCacheOptions::CACHE_DEFAULT_OPTIONS)]
-    project_info_cache: String,
-    /// cache for all valid endpoints
-    #[clap(long, default_value = config::EndpointCacheConfig::CACHE_DEFAULT_OPTIONS)]
-    endpoint_cache_config: String,
-    #[clap(flatten)]
-    parquet_upload: ParquetUploadArgs,
-
-    /// interval for backup metric collection
-    #[clap(long, default_value = "10m", value_parser = humantime::parse_duration)]
-    metric_backup_collection_interval: std::time::Duration,
-    /// remote storage configuration for backup metric collection
-    /// Encoded as toml (same format as pageservers), eg
-    /// `{bucket_name='the-bucket',bucket_region='us-east-1',prefix_in_bucket='proxy',endpoint='http://minio:9000'}`
-    #[clap(long, value_parser = remote_storage_from_toml)]
-    metric_backup_collection_remote_storage: Option<RemoteStorageConfig>,
-    /// chunk size for backup metric collection
-    /// Size of each event is no more than 400 bytes, so 2**22 is about 200MB before the compression.
-    #[clap(long, default_value = "4194304")]
-    metric_backup_collection_chunk_size: usize,
-    /// Whether to retry the connection to the compute node
-    #[clap(long, default_value = config::RetryConfig::CONNECT_TO_COMPUTE_DEFAULT_VALUES)]
-    connect_to_compute_retry: String,
-    /// Whether to retry the wake_compute request
-    #[clap(long, default_value = config::RetryConfig::WAKE_COMPUTE_DEFAULT_VALUES)]
-    wake_compute_retry: String,
-
-    /// Configure if this is a private access proxy for the POC: In that case the proxy will ignore the IP allowlist
-    #[clap(long, default_value_t = false, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
-    is_private_access_proxy: bool,
-
-    /// Configure whether all incoming requests have a Proxy Protocol V2 packet.
-    // TODO(conradludgate): switch default to rejected or required once we've updated all deployments
-    #[clap(value_enum, long, default_value_t = ProxyProtocolV2::Supported)]
-    proxy_protocol_v2: ProxyProtocolV2,
-
-    /// Time the proxy waits for the webauth session to be confirmed by the control plane.
-    // TODO: rename to `console_redirect_confirmation_timeout`.
-    #[clap(long, default_value = "2m", value_parser = humantime::parse_duration)]
-    webauth_confirmation_timeout: std::time::Duration,
-}
-
-#[derive(clap::Args, Clone, Copy, Debug)]
-struct SqlOverHttpArgs {
-    /// timeout for http connection requests
-    #[clap(long, default_value = "15s", value_parser = humantime::parse_duration)]
-    sql_over_http_timeout: tokio::time::Duration,
-
-    /// Whether the SQL over http pool is opt-in
-    #[clap(long, default_value_t = true, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
-    sql_over_http_pool_opt_in: bool,
-
-    /// How many connections to pool for each endpoint. Excess connections are discarded
-    #[clap(long, default_value_t = 20)]
-    sql_over_http_pool_max_conns_per_endpoint: usize,
-
-    /// How many connections to pool for each endpoint. Excess connections are discarded
-    #[clap(long, default_value_t = 20000)]
-    sql_over_http_pool_max_total_conns: usize,
-
-    /// How long pooled connections should remain idle for before closing
-    #[clap(long, default_value = "5m", value_parser = humantime::parse_duration)]
-    sql_over_http_idle_timeout: tokio::time::Duration,
-
-    /// Duration each shard will wait on average before a GC sweep.
-    /// A longer time will causes sweeps to take longer but will interfere less frequently.
-    #[clap(long, default_value = "10m", value_parser = humantime::parse_duration)]
-    sql_over_http_pool_gc_epoch: tokio::time::Duration,
-
-    /// How many shards should the global pool have. Must be a power of two.
-    /// More shards will introduce less contention for pool operations, but can
-    /// increase memory used by the pool
-    #[clap(long, default_value_t = 128)]
-    sql_over_http_pool_shards: usize,
-
-    #[clap(long, default_value_t = 10000)]
-    sql_over_http_client_conn_threshold: u64,
-
-    #[clap(long, default_value_t = 64)]
-    sql_over_http_cancel_set_shards: usize,
-
-    #[clap(long, default_value_t = 10 * 1024 * 1024)] // 10 MiB
-    sql_over_http_max_request_size_bytes: u64,
-
-    #[clap(long, default_value_t = 10 * 1024 * 1024)] // 10 MiB
-    sql_over_http_max_response_size_bytes: usize,
-}
-
-#[tokio::main]
-async fn main() -> anyhow::Result<()> {
-    let _logging_guard = proxy::logging::init().await?;
-    let _panic_hook_guard = utils::logging::replace_panic_hook_with_tracing_panic_hook();
-
-    let args = ProxyCliArgs::parse();
-
-    let region_provider =
-        RegionProviderChain::default_provider().or_else(Region::new(args.aws_region.clone()));
-    let provider_conf =
-        ProviderConfig::without_region().with_region(region_provider.region().await);
-    let aws_credentials_provider = {
-        // uses "AWS_ACCESS_KEY_ID", "AWS_SECRET_ACCESS_KEY"
-        CredentialsProviderChain::first_try("env", EnvironmentVariableCredentialsProvider::new())
-            // uses "AWS_PROFILE" / `aws sso login --profile <profile>`
-            .or_else(
-                "profile-sso",
-                ProfileFileCredentialsProvider::builder()
-                    .configure(&provider_conf)
-                    .build(),
-            )
-            // uses "AWS_WEB_IDENTITY_TOKEN_FILE", "AWS_ROLE_ARN", "AWS_ROLE_SESSION_NAME"
-            // needed to access remote extensions bucket
-            .or_else(
-                "token",
-                WebIdentityTokenCredentialsProvider::builder()
-                    .configure(&provider_conf)
-                    .build(),
-            )
-            // uses imds v2
-            .or_else("imds", ImdsCredentialsProvider::builder().build())
-    };
-    let elasticache_credentials_provider = Arc::new(elasticache::CredentialsProvider::new(
-        elasticache::AWSIRSAConfig::new(
-            args.aws_region.clone(),
-            args.redis_cluster_name,
-            args.redis_user_id,
-        ),
-        aws_credentials_provider,
-    ));
-    let regional_redis_client = match (args.redis_auth_type.as_str(), &args.redis_notifications) {
-        ("plain", redis_url) => match redis_url {
-            None => {
-                bail!("plain auth requires redis_notifications to be set");
-            }
-            Some(url) => Some(
-                ConnectionWithCredentialsProvider::new_with_static_credentials(url.to_string()),
-            ),
-        },
-        ("irsa", _) => match (&args.redis_host, args.redis_port) {
-            (Some(host), Some(port)) => Some(
-                ConnectionWithCredentialsProvider::new_with_credentials_provider(
-                    host.to_string(),
-                    port,
-                    elasticache_credentials_provider.clone(),
-                ),
-            ),
-            (None, None) => {
-                warn!("irsa auth requires redis-host and redis-port to be set, continuing without regional_redis_client");
-                None
-            }
-            _ => {
-                bail!("redis-host and redis-port must be specified together");
-            }
-        },
-        _ => {
-            bail!("unknown auth type given");
-        }
-    };
-
-    let endpoint_cache_config: config::EndpointCacheConfig = args.endpoint_cache_config.parse()?;
-
-    let Some(mut regional_redis_client) = regional_redis_client else {
-        bail!("no regional_redis_client");
-    };
-
-    if let Err(e) = regional_redis_client.connect().await {
-        bail!("error connecting to redis: {:?}", e);
-    }
-
-    let mut last_id = "0-0".to_string();
-    batch_read(
-        &mut regional_redis_client,
-        endpoint_cache_config.stream_name,
-        StreamReadOptions::default().count(endpoint_cache_config.default_batch_size),
-        &mut last_id,
-        true,
-        |event| {
-            let json = serde_json::to_string(&event)?;
-            println!("{}", json);
-            Ok(())
-        },
-    )
-    .await?;
-
-    Ok(())
-}
-
-// TODO: this could be an enum, but events in Redis need to be fixed first.
-// ProjectCreated was sent with type:branch_created. So we ignore type.
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
-struct CPlaneEvent {
-    id: Option<String>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    endpoint_created: Option<EndpointCreated>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    branch_created: Option<BranchCreated>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    project_created: Option<ProjectCreated>,
-
-    #[serde(rename = "type")]
-    _type: Option<String>,
-}
-
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
-struct EndpointCreated {
-    endpoint_id: String,
-}
-
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
-struct BranchCreated {
-    branch_id: String,
-}
-
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
-struct ProjectCreated {
-    project_id: String,
-}
-
-impl TryFrom<&Value> for CPlaneEvent {
-    type Error = anyhow::Error;
-    fn try_from(value: &Value) -> Result<Self, Self::Error> {
-        let json = String::from_redis_value(value)?;
-        Ok(serde_json::from_str(&json)?)
-    }
-}
-
-async fn batch_read(
-    conn: &mut ConnectionWithCredentialsProvider,
-    stream_name: String,
-    opts: StreamReadOptions,
-    last_id: &mut String,
-    return_when_finish: bool,
-    mut insert_event: impl FnMut(CPlaneEvent) -> anyhow::Result<()>,
-) -> anyhow::Result<()> {
-    let mut total: usize = 0;
-    loop {
-        let mut res: StreamReadReply = conn
-            .xread_options(&[&stream_name], &[last_id.as_str()], &opts)
-            .await?;
-
-        if res.keys.is_empty() {
-            if return_when_finish {
-                if total != 0 {
-                    break;
-                }
-                anyhow::bail!(
-                    "Redis stream {} is empty, cannot be used to filter endpoints",
-                    stream_name
-                );
-            }
-            // If we are not returning when finish, we should wait for more data.
-            continue;
-        }
-        if res.keys.len() != 1 {
-            anyhow::bail!("Cannot read from redis stream {}", stream_name);
-        }
-
-        let key = res.keys.pop().expect("Checked length above");
-
-        for stream_id in key.ids {
-            total += 1;
-            for value in stream_id.map.values() {
-                match value.try_into() {
-                    Ok::<CPlaneEvent, _>(mut event) => {
-                        event.id = Some(stream_id.id.clone());
-                        insert_event(event)?;
-                    }
-                    Err(err) => {
-                        tracing::error!("error parsing value {value:?}: {err:?}");
-                    }
-                };
-            }
-            if total.is_power_of_two() {
-                tracing::debug!("endpoints read {}", total);
-            }
-            *last_id = stream_id.id;
-        }
-    }
-    tracing::info!("read {} endpoints/branches/projects from redis", total);
-    Ok(())
-}

proxy/src/cache/endpoints.rs

@@ -11,7 +11,7 @@ use tokio_util::sync::CancellationToken;
 use tracing::info;
 
 use crate::config::EndpointCacheConfig;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::intern::{BranchIdInt, EndpointIdInt, ProjectIdInt};
 use crate::metrics::{Metrics, RedisErrors, RedisEventsCount};
 use crate::rate_limiter::GlobalRateLimiter;
@@ -75,7 +75,7 @@ impl EndpointsCache {
         }
     }
 
-    pub(crate) fn is_valid(&self, ctx: &RequestMonitoring, endpoint: &EndpointId) -> bool {
+    pub(crate) fn is_valid(&self, ctx: &RequestContext, endpoint: &EndpointId) -> bool {
         if !self.ready.load(Ordering::Acquire) {
             // the endpoint cache is not yet fully initialised.
             return true;

proxy/src/cancellation.rs

@@ -7,7 +7,7 @@ use thiserror::Error;
 use tokio::net::TcpStream;
 use tokio::sync::Mutex;
 use tokio_postgres::{CancelToken, NoTls};
-use tracing::info;
+use tracing::{debug, info};
 use uuid::Uuid;
 
 use crate::error::ReportableError;
@@ -73,7 +73,7 @@ impl<P: CancellationPublisher> CancellationHandler<P> {
             break key;
         };
 
-        info!("registered new query cancellation key {key}");
+        debug!("registered new query cancellation key {key}");
         Session {
             key,
             cancellation_handler: self,
@@ -165,7 +165,7 @@ impl CancelClosure {
     pub(crate) async fn try_cancel_query(self) -> Result<(), CancelError> {
         let socket = TcpStream::connect(self.socket_addr).await?;
         self.cancel_token.cancel_query_raw(socket, NoTls).await?;
-        info!("query was cancelled");
+        debug!("query was cancelled");
         Ok(())
     }
 }
@@ -182,7 +182,7 @@ impl<P> Session<P> {
     /// Store the cancel token for the given session.
     /// This enables query cancellation in `crate::proxy::prepare_client_connection`.
     pub(crate) fn enable_query_cancellation(&self, cancel_closure: CancelClosure) -> CancelKeyData {
-        info!("enabling query cancellation for this session");
+        debug!("enabling query cancellation for this session");
         self.cancellation_handler
             .map
             .insert(self.key, Some(cancel_closure));
@@ -194,7 +194,7 @@ impl<P> Session<P> {
 impl<P> Drop for Session<P> {
     fn drop(&mut self) {
         self.cancellation_handler.map.remove(&self.key);
-        info!("dropped query cancellation key {}", &self.key);
+        debug!("dropped query cancellation key {}", &self.key);
     }
 }
 

proxy/src/compute.rs

@@ -14,11 +14,11 @@ use thiserror::Error;
 use tokio::net::TcpStream;
 use tokio_postgres::tls::MakeTlsConnect;
 use tokio_postgres_rustls::MakeRustlsConnect;
-use tracing::{error, info, warn};
+use tracing::{debug, error, info, warn};
 
 use crate::auth::parse_endpoint_param;
 use crate::cancellation::CancelClosure;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::control_plane::client::ApiLockError;
 use crate::control_plane::errors::WakeComputeError;
 use crate::control_plane::messages::MetricsAuxInfo;
@@ -213,7 +213,7 @@ impl ConnCfg {
         };
 
         let connect_once = |host, port| {
-            info!("trying to connect to compute node at {host}:{port}");
+            debug!("trying to connect to compute node at {host}:{port}");
             connect_with_timeout(host, port).and_then(|socket| async {
                 let socket_addr = socket.peer_addr()?;
                 // This prevents load balancer from severing the connection.
@@ -286,7 +286,7 @@ impl ConnCfg {
     /// Connect to a corresponding compute node.
     pub(crate) async fn connect(
         &self,
-        ctx: &RequestMonitoring,
+        ctx: &RequestContext,
         allow_self_signed_compute: bool,
         aux: MetricsAuxInfo,
         timeout: Duration,
@@ -328,6 +328,7 @@ impl ConnCfg {
         tracing::Span::current().record("pid", tracing::field::display(client.get_process_id()));
         let stream = connection.stream.into_inner();
 
+        // TODO: lots of useful info but maybe we can move it elsewhere (eg traces?)
         info!(
             cold_start_info = ctx.cold_start_info().as_str(),
             "connected to compute node at {host} ({socket_addr}) sslmode={:?}",

proxy/src/config.rs

@@ -64,7 +64,7 @@ pub struct HttpConfig {
     pub pool_options: GlobalConnPoolOptions,
     pub cancel_set: CancelSet,
     pub client_conn_threshold: u64,
-    pub max_request_size_bytes: u64,
+    pub max_request_size_bytes: usize,
     pub max_response_size_bytes: usize,
 }
 

proxy/src/console_redirect_proxy.rs

@@ -8,7 +8,7 @@ use tracing::{debug, error, info, Instrument};
 use crate::auth::backend::ConsoleRedirectBackend;
 use crate::cancellation::{CancellationHandlerMain, CancellationHandlerMainInternal};
 use crate::config::{ProxyConfig, ProxyProtocolV2};
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::error::ReportableError;
 use crate::metrics::{Metrics, NumClientConnectionsGuard};
 use crate::protocol2::{read_proxy_protocol, ConnectHeader, ConnectionInfo};
@@ -82,7 +82,7 @@ pub async fn task_main(
                 }
             };
 
-            let ctx = RequestMonitoring::new(
+            let ctx = RequestContext::new(
                 session_id,
                 peer_addr,
                 crate::metrics::Protocol::Tcp,
@@ -141,12 +141,12 @@ pub async fn task_main(
 pub(crate) async fn handle_client<S: AsyncRead + AsyncWrite + Unpin>(
     config: &'static ProxyConfig,
     backend: &'static ConsoleRedirectBackend,
-    ctx: &RequestMonitoring,
+    ctx: &RequestContext,
     cancellation_handler: Arc<CancellationHandlerMain>,
     stream: S,
     conn_gauge: NumClientConnectionsGuard<'static>,
 ) -> Result<Option<ProxyPassthrough<CancellationHandlerMainInternal, S>>, ClientRequestError> {
-    info!(
+    debug!(
         protocol = %ctx.protocol(),
         "handling interactive connection from client"
     );

proxy/src/context/mod.rs

@@ -8,7 +8,7 @@ use pq_proto::StartupMessageParams;
 use smol_str::SmolStr;
 use tokio::sync::mpsc;
 use tracing::field::display;
-use tracing::{debug, info, info_span, Span};
+use tracing::{debug, info_span, Span};
 use try_lock::TryLock;
 use uuid::Uuid;
 
@@ -32,15 +32,15 @@ pub(crate) static LOG_CHAN_DISCONNECT: OnceCell<mpsc::WeakUnboundedSender<Reques
 ///
 /// This data should **not** be used for connection logic, only for observability and limiting purposes.
 /// All connection logic should instead use strongly typed state machines, not a bunch of Options.
-pub struct RequestMonitoring(
+pub struct RequestContext(
     /// To allow easier use of the ctx object, we have interior mutability.
     /// I would typically use a RefCell but that would break the `Send` requirements
     /// so we need something with thread-safety. `TryLock` is a cheap alternative
     /// that offers similar semantics to a `RefCell` but with synchronisation.
-    TryLock<RequestMonitoringInner>,
+    TryLock<RequestContextInner>,
 );
 
-struct RequestMonitoringInner {
+struct RequestContextInner {
     pub(crate) conn_info: ConnectionInfo,
     pub(crate) session_id: Uuid,
     pub(crate) protocol: Protocol,
@@ -81,10 +81,10 @@ pub(crate) enum AuthMethod {
     Cleartext,
 }
 
-impl Clone for RequestMonitoring {
+impl Clone for RequestContext {
     fn clone(&self) -> Self {
         let inner = self.0.try_lock().expect("should not deadlock");
-        let new = RequestMonitoringInner {
+        let new = RequestContextInner {
             conn_info: inner.conn_info.clone(),
             session_id: inner.session_id,
             protocol: inner.protocol,
@@ -115,13 +115,14 @@ impl Clone for RequestMonitoring {
     }
 }
 
-impl RequestMonitoring {
+impl RequestContext {
     pub fn new(
         session_id: Uuid,
         conn_info: ConnectionInfo,
         protocol: Protocol,
         region: &'static str,
     ) -> Self {
+        // TODO: be careful with long lived spans
         let span = info_span!(
             "connect_request",
             %protocol,
@@ -131,7 +132,7 @@ impl RequestMonitoring {
             role = tracing::field::Empty,
         );
 
-        let inner = RequestMonitoringInner {
+        let inner = RequestContextInner {
             conn_info,
             session_id,
             protocol,
@@ -167,7 +168,7 @@ impl RequestMonitoring {
         let ip = IpAddr::from([127, 0, 0, 1]);
         let addr = SocketAddr::new(ip, 5432);
         let conn_info = ConnectionInfo { addr, extra: None };
-        RequestMonitoring::new(Uuid::now_v7(), conn_info, Protocol::Tcp, "test")
+        RequestContext::new(Uuid::now_v7(), conn_info, Protocol::Tcp, "test")
     }
 
     pub(crate) fn console_application_name(&self) -> String {
@@ -324,7 +325,7 @@ impl RequestMonitoring {
 }
 
 pub(crate) struct LatencyTimerPause<'a> {
-    ctx: &'a RequestMonitoring,
+    ctx: &'a RequestContext,
     start: tokio::time::Instant,
     waiting_for: Waiting,
 }
@@ -340,7 +341,7 @@ impl Drop for LatencyTimerPause<'_> {
     }
 }
 
-impl RequestMonitoringInner {
+impl RequestContextInner {
     fn set_cold_start_info(&mut self, info: ColdStartInfo) {
         self.cold_start_info = info;
         self.latency_timer.cold_start_info(info);
@@ -384,6 +385,10 @@ impl RequestMonitoringInner {
         } else {
             ConnectOutcome::Failed
         };
+
+        // TODO: get rid of entirely/refactor
+        // check for false positives
+        // AND false negatives
         if let Some(rejected) = self.rejected {
             let ep = self
                 .endpoint_id
@@ -391,7 +396,7 @@ impl RequestMonitoringInner {
                 .map(|x| x.as_str())
                 .unwrap_or_default();
             // This makes sense only if cache is disabled
-            info!(
+            debug!(
                 ?outcome,
                 ?rejected,
                 ?ep,
@@ -425,7 +430,7 @@ impl RequestMonitoringInner {
     }
 }
 
-impl Drop for RequestMonitoringInner {
+impl Drop for RequestContextInner {
     fn drop(&mut self) {
         if self.sender.is_some() {
             self.log_connect();

proxy/src/context/parquet.rs

@@ -20,7 +20,7 @@ use tokio_util::sync::CancellationToken;
 use tracing::{debug, info, Span};
 use utils::backoff;
 
-use super::{RequestMonitoringInner, LOG_CHAN};
+use super::{RequestContextInner, LOG_CHAN};
 use crate::config::remote_storage_from_toml;
 use crate::context::LOG_CHAN_DISCONNECT;
 
@@ -117,8 +117,8 @@ impl serde::Serialize for Options<'_> {
     }
 }
 
-impl From<&RequestMonitoringInner> for RequestData {
-    fn from(value: &RequestMonitoringInner) -> Self {
+impl From<&RequestContextInner> for RequestData {
+    fn from(value: &RequestContextInner) -> Self {
         Self {
             session_id: value.session_id,
             peer_addr: value.conn_info.addr.ip().to_string(),

proxy/src/control_plane/client/mock.rs

@@ -13,7 +13,7 @@ use crate::auth::backend::jwt::AuthRule;
 use crate::auth::backend::ComputeUserInfo;
 use crate::auth::IpPattern;
 use crate::cache::Cached;
-use crate::context::RequestMonitoring;
+use crate::context::RequestContext;
 use crate::control_plane::client::{CachedAllowedIps, CachedRoleSecret};
 use crate::control_plane::errors::{
     ControlPlaneError, GetAuthInfoError, GetEndpointJwksError, WakeComputeError,
@@ -206,7 +206,7 @@ impl super::ControlPlaneApi for MockControlPlane {
     #[tracing::instrument(skip_all)]
     async fn get_role_secret(
         &self,
-        _ctx: &RequestMonitoring,
+        _ctx: &RequestContext,
         user_info: &ComputeUserInfo,
     ) -> Result<CachedRoleSecret, GetAuthInfoError> {
         Ok(CachedRoleSecret::new_uncached(
@@ -216,7 +216,7 @@ impl super::ControlPlaneApi for MockControlPlane {
 
     async fn get_allowed_ips_and_secret(
         &self,
-        _ctx: &RequestMonitoring,
+        _ctx: &RequestContext,
         user_info: &ComputeUserInfo,
     ) -> Result<(CachedAllowedIps, Option<CachedRoleSecret>), GetAuthInfoError> {
         Ok((
@@ -229,7 +229,7 @@ impl super::ControlPlaneApi for MockControlPlane {
 
     async fn get_endpoint_jwks(
         &self,
-        _ctx: &RequestMonitoring,
+        _ctx: &RequestContext,
         endpoint: EndpointId,
     ) -> Result<Vec<AuthRule>, GetEndpointJwksError> {
         self.do_get_endpoint_jwks(endpoint).await
@@ -238,7 +238,7 @@ impl super::ControlPlaneApi for MockControlPlane {
     #[tracing::instrument(skip_all)]
     async fn wake_compute(
         &self,
-        _ctx: &RequestMonitoring,
+        _ctx: &RequestContext,
         _user_info: &ComputeUserInfo,
     ) -> Result<CachedNodeInfo, WakeComputeError> {
         self.do_wake_compute().map_ok(Cached::new_uncached).await