Cache initdb output to speed up tenant creation in tests

initdb takes about 1 s. Our tests create and destroy a lot of tenants,
so that adds up. Cache the initdb result to speed it up.

This is currently only enabled in tests. Out of caution, mostly. But
also because when you reuse the initdb result, all the postgres
clusters end up having the same system_identifier, which is supposed
to be unique. It's not necessary for it to be unique for correctness,
nothing critical relies on it and you can easily end up with duplicate
system_identifiers in standalone PostgreSQL too, if you e.g. create a
backup and restore it on a different system. But it is used in various
checks to reduce the chance that you e.g. accidentally apply WAL
belonging to a different cluster.

Because this is aimed at tests, there are a few things that might be
surprising:

- The initdb cache directory is configurable, and can be outside the
  pageserver's repo directory. This allows reuse across different
  pageservers running on the same host. In production use, that'd be
  pointless, but our tests create a lot of pageservers.

- The cache is not automatically purged at start / shutdown. For
  production use, we'd probably want that, so that we'd pick up any
  changes in what an empty cluster looks like after a Postgres minor
  version upgrade, for example. But again tests create and destroy a
  lot of pageservers, so it's important to retain the cache.

- The locking on the cache directory relies purely on filesystem
  operations and atomic rename(). Using e.g. a rust Mutex() would be
  more straightforward, but that's not enough because the cache needs
  to be shared between different pageservers running on the same
  system.

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1,6 @@
+
+blank_issues_enabled: true
+contact_links:
+  - name: Feature request
+    url: https://console.neon.tech/app/projects?modal=feedback
+    about: For feature requests in the Neon product, please submit via the feedback form on `https://console.neon.tech`

.github/actionlint.yml

@@ -7,6 +7,13 @@ self-hosted-runner:
     - small-arm64
     - us-east-2
 config-variables:
+  - AZURE_DEV_CLIENT_ID
+  - AZURE_DEV_REGISTRY_NAME
+  - AZURE_DEV_SUBSCRIPTION_ID
+  - AZURE_PROD_CLIENT_ID
+  - AZURE_PROD_REGISTRY_NAME
+  - AZURE_PROD_SUBSCRIPTION_ID
+  - AZURE_TENANT_ID
   - BENCHMARK_PROJECT_ID_PUB
   - BENCHMARK_PROJECT_ID_SUB
   - REMOTE_STORAGE_AZURE_CONTAINER

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

@@ -71,7 +71,7 @@ runs:
       if: inputs.build_type != 'remote'
       uses: ./.github/actions/download
       with:
-        name: compatibility-snapshot-${{ inputs.build_type }}-pg${{ inputs.pg_version }}
+        name: compatibility-snapshot-${{ runner.arch }}-${{ inputs.build_type }}-pg${{ inputs.pg_version }}
         path: /tmp/compatibility_snapshot_pg${{ inputs.pg_version }}
         prefix: latest
         # The lack of compatibility snapshot (for example, for the new Postgres version)
@@ -211,13 +211,13 @@ runs:
         fi
 
     - name: Upload compatibility snapshot
-      if: github.ref_name == 'release'
+      # Note, that we use `github.base_ref` which is a target branch for a PR
+      if: github.event_name == 'pull_request' && github.base_ref == 'release'
       uses: ./.github/actions/upload
       with:
-        name: compatibility-snapshot-${{ inputs.build_type }}-pg${{ inputs.pg_version }}-${{ github.run_id }}
+        name: compatibility-snapshot-${{ runner.arch }}-${{ inputs.build_type }}-pg${{ inputs.pg_version }}
         # Directory is created by test_compatibility.py::test_create_snapshot, keep the path in sync with the test
         path: /tmp/test_output/compatibility_snapshot_pg${{ inputs.pg_version }}/
-        prefix: latest
 
     - name: Upload test results
       if: ${{ !cancelled() }}

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

@@ -62,7 +62,7 @@ jobs:
           #
           git config --global --add safe.directory ${{ github.workspace }}
           git config --global --add safe.directory ${GITHUB_WORKSPACE}
-          for r in 14 15 16; do
+          for r in 14 15 16 17; do
             git config --global --add safe.directory "${{ github.workspace }}/vendor/postgres-v$r"
             git config --global --add safe.directory "${GITHUB_WORKSPACE}/vendor/postgres-v$r"
           done
@@ -83,6 +83,10 @@ jobs:
         id: pg_v16_rev
         run: echo pg_rev=$(git rev-parse HEAD:vendor/postgres-v16) >> $GITHUB_OUTPUT
 
+      - name: Set pg 17 revision for caching
+        id: pg_v17_rev
+        run: echo pg_rev=$(git rev-parse HEAD:vendor/postgres-v17) >> $GITHUB_OUTPUT
+
       # Set some environment variables used by all the steps.
       #
       # CARGO_FLAGS is extra options to pass to "cargo build", "cargo test" etc.
@@ -136,6 +140,13 @@ jobs:
           path: pg_install/v16
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ inputs.build-type }}-pg-${{ steps.pg_v16_rev.outputs.pg_rev }}-${{ hashFiles('Makefile', 'Dockerfile.build-tools') }}
 
+      - name: Cache postgres v17 build
+        id: cache_pg_17
+        uses: actions/cache@v4
+        with:
+          path: pg_install/v17
+          key: v1-${{ runner.os }}-${{ runner.arch }}-${{ inputs.build-type }}-pg-${{ steps.pg_v17_rev.outputs.pg_rev }}-${{ hashFiles('Makefile', 'Dockerfile.build-tools') }}
+
       - name: Build postgres v14
         if: steps.cache_pg_14.outputs.cache-hit != 'true'
         run: mold -run make postgres-v14 -j$(nproc)
@@ -148,6 +159,10 @@ jobs:
         if: steps.cache_pg_16.outputs.cache-hit != 'true'
         run: mold -run make postgres-v16 -j$(nproc)
 
+      - name: Build postgres v17
+        if: steps.cache_pg_17.outputs.cache-hit != 'true'
+        run: mold -run make postgres-v17 -j$(nproc)
+
       - name: Build neon extensions
         run: mold -run make neon-pg-ext -j$(nproc)
 
@@ -210,14 +225,20 @@ jobs:
         run: |
           PQ_LIB_DIR=$(pwd)/pg_install/v16/lib
           export PQ_LIB_DIR
-          LD_LIBRARY_PATH=$(pwd)/pg_install/v16/lib
+          LD_LIBRARY_PATH=$(pwd)/pg_install/v17/lib
           export LD_LIBRARY_PATH
 
           #nextest does not yet support running doctests
           ${cov_prefix} cargo test --doc $CARGO_FLAGS $CARGO_FEATURES
 
+          # run all non-pageserver tests
+          ${cov_prefix} cargo nextest run $CARGO_FLAGS $CARGO_FEATURES -E '!package(pageserver)'
+
+          # run pageserver tests with different settings
           for io_engine in std-fs tokio-epoll-uring ; do
-            NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOENGINE=$io_engine ${cov_prefix} cargo nextest run $CARGO_FLAGS $CARGO_FEATURES
+            for io_buffer_alignment in 0 1 512 ; do
+              NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOENGINE=$io_engine NEON_PAGESERVER_UNIT_TEST_IO_BUFFER_ALIGNMENT=$io_buffer_alignment ${cov_prefix} cargo nextest run $CARGO_FLAGS $CARGO_FEATURES  -E 'package(pageserver)'
+            done
           done
 
           # Run separate tests for real S3

.github/workflows/_push-to-acr.yml

@@ -0,0 +1,56 @@
+name: Push images to ACR
+on:
+  workflow_call:
+    inputs:
+      client_id:
+        description: Client ID of Azure managed identity or Entra app
+        required: true
+        type: string
+      image_tag:
+        description: Tag for the container image
+        required: true
+        type: string
+      images:
+        description: Images to push
+        required: true
+        type: string
+      registry_name:
+        description: Name of the container registry
+        required: true
+        type: string
+      subscription_id:
+        description: Azure subscription ID
+        required: true
+        type: string
+      tenant_id:
+        description: Azure tenant ID
+        required: true
+        type: string
+
+jobs:
+  push-to-acr:
+    runs-on: ubuntu-22.04
+    permissions:
+      contents: read  # This is required for actions/checkout
+      id-token: write # This is required for Azure Login to work.
+
+    steps:
+      - name: Azure login
+        uses: azure/login@6c251865b4e6290e7b78be643ea2d005bc51f69a  # @v2.1.1
+        with:
+          client-id: ${{ inputs.client_id }}
+          subscription-id: ${{ inputs.subscription_id }}
+          tenant-id: ${{ inputs.tenant_id }}
+
+      - name: Login to ACR
+        run: |
+          az acr login --name=${{ inputs.registry_name }}
+
+      - name: Copy docker images to ACR ${{ inputs.registry_name }}
+        run: |
+          images='${{ inputs.images }}'
+          for image in ${images}; do
+            docker buildx imagetools create \
+              -t ${{ inputs.registry_name }}.azurecr.io/neondatabase/${image}:${{ inputs.image_tag }} \
+                                                        neondatabase/${image}:${{ inputs.image_tag }}
+          done

.github/workflows/build_and_test.yml

@@ -54,8 +54,8 @@ jobs:
       build-tag: ${{steps.build-tag.outputs.tag}}
 
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
+      # Need `fetch-depth: 0` to count the number of commits in the branch
+      - uses: actions/checkout@v4
         with:
           fetch-depth: 0
 
@@ -159,6 +159,10 @@ jobs:
       # This will catch compiler & clippy warnings in all feature combinations.
       # TODO: use cargo hack for build and test as well, but, that's quite expensive.
       # NB: keep clippy args in sync with ./run_clippy.sh
+      #
+      # The only difference between "clippy --debug" and "clippy --release" is that in --release mode,
+      # #[cfg(debug_assertions)] blocks are not built. It's not worth building everything for second
+      # time just for that, so skip "clippy --release".
       - run: |
           CLIPPY_COMMON_ARGS="$( source .neon_clippy_args; echo "$CLIPPY_COMMON_ARGS")"
           if [ "$CLIPPY_COMMON_ARGS" = "" ]; then
@@ -168,8 +172,6 @@ jobs:
           echo "CLIPPY_COMMON_ARGS=${CLIPPY_COMMON_ARGS}" >> $GITHUB_ENV
       - name: Run cargo clippy (debug)
         run: cargo hack --feature-powerset clippy $CLIPPY_COMMON_ARGS
-      - name: Run cargo clippy (release)
-        run: cargo hack --feature-powerset clippy --release $CLIPPY_COMMON_ARGS
 
       - name: Check documentation generation
         run: cargo doc --workspace --no-deps --document-private-items
@@ -211,7 +213,7 @@ jobs:
       build-tag: ${{ needs.tag.outputs.build-tag }}
       build-type: ${{ matrix.build-type }}
       # Run tests on all Postgres versions in release builds and only on the latest version in debug builds
-      pg-versions: ${{ matrix.build-type == 'release' && '["v14", "v15", "v16"]' || '["v16"]' }}
+      pg-versions: ${{ matrix.build-type == 'release' && '["v14", "v15", "v16", "v17"]' || '["v17"]' }}
     secrets: inherit
 
   # Keep `benchmarks` job outside of `build-and-test-locally` workflow to make job failures non-blocking
@@ -286,6 +288,7 @@ jobs:
           PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
           TEST_RESULT_CONNSTR: "${{ secrets.REGRESS_TEST_RESULT_CONNSTR_NEW }}"
           PAGESERVER_VIRTUAL_FILE_IO_ENGINE: tokio-epoll-uring
+          SYNC_AFTER_EACH_TEST: true
       # XXX: no coverage data handling here, since benchmarks are run on release builds,
       # while coverage is currently collected for the debug ones
 
@@ -356,6 +359,7 @@ jobs:
             })
 
   coverage-report:
+    if: ${{ !startsWith(github.ref_name, 'release') }}
     needs: [ check-permissions, build-build-tools-image, build-and-test-locally ]
     runs-on: [ self-hosted, small ]
     container:
@@ -372,8 +376,8 @@ jobs:
         coverage-html: ${{ steps.upload-coverage-report-new.outputs.report-url }}
         coverage-json: ${{ steps.upload-coverage-report-new.outputs.summary-json }}
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
+      # Need `fetch-depth: 0` for differential coverage (to get diff between two commits)
+      - uses: actions/checkout@v4
         with:
           submodules: true
           fetch-depth: 0
@@ -474,11 +478,9 @@ jobs:
     runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}
 
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
+      - uses: actions/checkout@v4
         with:
           submodules: true
-          fetch-depth: 0
 
       - uses: ./.github/actions/set-docker-config-dir
       - uses: docker/setup-buildx-action@v3
@@ -547,17 +549,15 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        version: [ v14, v15, v16 ]
+        version: [ v14, v15, v16, v17 ]
         arch: [ x64, arm64 ]
 
     runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}
 
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
+      - uses: actions/checkout@v4
         with:
           submodules: true
-          fetch-depth: 0
 
       - uses: ./.github/actions/set-docker-config-dir
       - uses: docker/setup-buildx-action@v3
@@ -626,7 +626,7 @@ jobs:
 
       - name: Build compute-tools image
         # compute-tools are Postgres independent, so build it only once
-        if: matrix.version == 'v16'
+        if: matrix.version == 'v17'
         uses: docker/build-push-action@v6
         with:
           target: compute-tools-image
@@ -648,7 +648,7 @@ jobs:
 
     strategy:
       matrix:
-        version: [ v14, v15, v16 ]
+        version: [ v14, v15, v16, v17 ]
 
     steps:
       - uses: docker/login-action@v3
@@ -670,7 +670,7 @@ jobs:
                                              neondatabase/neon-test-extensions-${{ matrix.version }}:${{ needs.tag.outputs.build-tag }}-arm64
 
       - name: Create multi-arch compute-tools image
-        if: matrix.version == 'v16'
+        if: matrix.version == 'v17'
         run: |
           docker buildx imagetools create -t neondatabase/compute-tools:${{ needs.tag.outputs.build-tag }} \
                                              neondatabase/compute-tools:${{ needs.tag.outputs.build-tag }}-x64 \
@@ -688,7 +688,7 @@ jobs:
                                                                                 neondatabase/compute-node-${{ matrix.version }}:${{ needs.tag.outputs.build-tag }}
 
       - name: Push multi-arch compute-tools image to ECR
-        if: matrix.version == 'v16'
+        if: matrix.version == 'v17'
         run: |
           docker buildx imagetools create -t 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:${{ needs.tag.outputs.build-tag }} \
                                                                                 neondatabase/compute-tools:${{ needs.tag.outputs.build-tag }}
@@ -699,15 +699,12 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        version: [ v14, v15, v16 ]
+        version: [ v14, v15, v16, v17 ]
     env:
       VM_BUILDER_VERSION: v0.29.3
 
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
+      - uses: actions/checkout@v4
 
       - name: Downloading vm-builder
         run: |
@@ -747,10 +744,7 @@ jobs:
     runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', matrix.arch == 'arm64' && 'small-arm64' || 'small')) }}
 
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
+      - uses: actions/checkout@v4
 
       - uses: ./.github/actions/set-docker-config-dir
       - uses: docker/login-action@v3
@@ -793,14 +787,11 @@ jobs:
           docker compose -f ./docker-compose/docker-compose.yml down
 
   promote-images:
-    permissions:
-      contents: read  # This is required for actions/checkout
-      id-token: write # This is required for Azure Login to work.
     needs: [ check-permissions, tag, test-images, vm-compute-node-image ]
     runs-on: ubuntu-22.04
 
     env:
-      VERSIONS: v14 v15 v16
+      VERSIONS: v14 v15 v16 v17
 
     steps:
       - uses: docker/login-action@v3
@@ -822,28 +813,6 @@ jobs:
                                                neondatabase/vm-compute-node-${version}:${{ needs.tag.outputs.build-tag }}
           done
 
-      - name: Azure login
-        if: github.ref_name == 'main'
-        uses: azure/login@6c251865b4e6290e7b78be643ea2d005bc51f69a  # @v2.1.1
-        with:
-          client-id: ${{ secrets.AZURE_DEV_CLIENT_ID }}
-          tenant-id: ${{ secrets.AZURE_TENANT_ID }}
-          subscription-id: ${{ secrets.AZURE_DEV_SUBSCRIPTION_ID }}
-
-      - name: Login to ACR
-        if: github.ref_name == 'main'
-        run: |
-          az acr login --name=neoneastus2
-
-      - name: Copy docker images to ACR-dev
-        if: github.ref_name == 'main'
-        run: |
-          for image in neon compute-tools {vm-,}compute-node-{v14,v15,v16}; do
-            docker buildx imagetools create \
-              -t neoneastus2.azurecr.io/neondatabase/${image}:${{ needs.tag.outputs.build-tag }} \
-                                        neondatabase/${image}:${{ needs.tag.outputs.build-tag }}
-          done
-
       - name: Add latest tag to images
         if: github.ref_name == 'main'
         run: |
@@ -863,7 +832,7 @@ jobs:
             done
           done
           docker buildx imagetools create -t neondatabase/neon-test-extensions-v16:latest \
-                                             neondatabase/neon-test-extensions-v16:${{ needs.tag.outputs.build-tag }}
+                                              neondatabase/neon-test-extensions-v16:${{ needs.tag.outputs.build-tag }}
 
       - name: Login to prod ECR
         uses: docker/login-action@v3
@@ -876,11 +845,35 @@ jobs:
       - name: Copy all images to prod ECR
         if: github.ref_name == 'release'|| github.ref_name == 'release-proxy'
         run: |
-          for image in neon compute-tools {vm-,}compute-node-{v14,v15,v16}; do
+          for image in neon compute-tools {vm-,}compute-node-{v14,v15,v16,v17}; do
             docker buildx imagetools create -t 093970136003.dkr.ecr.eu-central-1.amazonaws.com/${image}:${{ needs.tag.outputs.build-tag }} \
                                                369495373322.dkr.ecr.eu-central-1.amazonaws.com/${image}:${{ needs.tag.outputs.build-tag }}
           done
 
+  push-to-acr-dev:
+    if: github.ref_name == 'main'
+    needs: [ tag, promote-images ]
+    uses: ./.github/workflows/_push-to-acr.yml
+    with:
+      client_id: ${{ vars.AZURE_DEV_CLIENT_ID }}
+      image_tag: ${{ needs.tag.outputs.build-tag }}
+      images: neon compute-tools vm-compute-node-v14 vm-compute-node-v15 vm-compute-node-v16 vm-compute-node-v17 compute-node-v14 compute-node-v15 compute-node-v16 compute-node-v17
+      registry_name: ${{ vars.AZURE_DEV_REGISTRY_NAME }}
+      subscription_id: ${{ vars.AZURE_DEV_SUBSCRIPTION_ID }}
+      tenant_id: ${{ vars.AZURE_TENANT_ID }}
+
+  push-to-acr-prod:
+    if: github.ref_name == 'release'|| github.ref_name == 'release-proxy'
+    needs: [ tag, promote-images ]
+    uses: ./.github/workflows/_push-to-acr.yml
+    with:
+      client_id: ${{ vars.AZURE_PROD_CLIENT_ID }}
+      image_tag: ${{ needs.tag.outputs.build-tag }}
+      images: neon compute-tools vm-compute-node-v14 vm-compute-node-v15 vm-compute-node-v16 vm-compute-node-v17 compute-node-v14 compute-node-v15 compute-node-v16 compute-node-v17
+      registry_name: ${{ vars.AZURE_PROD_REGISTRY_NAME }}
+      subscription_id: ${{ vars.AZURE_PROD_SUBSCRIPTION_ID }}
+      tenant_id: ${{ vars.AZURE_TENANT_ID }}
+
   trigger-custom-extensions-build-and-wait:
     needs: [ check-permissions, tag ]
     runs-on: ubuntu-22.04
@@ -956,8 +949,9 @@ jobs:
           exit 1
 
   deploy:
-    needs: [ check-permissions, promote-images, tag, build-and-test-locally, trigger-custom-extensions-build-and-wait ]
-    if: github.ref_name == 'main' || github.ref_name == 'release'|| github.ref_name == 'release-proxy'
+    needs: [ check-permissions, promote-images, tag, build-and-test-locally, trigger-custom-extensions-build-and-wait, push-to-acr-dev, push-to-acr-prod ]
+    # `!failure() && !cancelled()` is required because the workflow depends on the job that can be skipped: `push-to-acr-dev` and `push-to-acr-prod`
+    if: (github.ref_name == 'main' || github.ref_name == 'release' || github.ref_name == 'release-proxy') && !failure() && !cancelled()
 
     runs-on: [ self-hosted, small ]
     container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/ansible:latest
@@ -971,15 +965,12 @@ jobs:
           #
           git config --global --add safe.directory ${{ github.workspace }}
           git config --global --add safe.directory ${GITHUB_WORKSPACE}
-          for r in 14 15 16; do
+          for r in 14 15 16 17; do
             git config --global --add safe.directory "${{ github.workspace }}/vendor/postgres-v$r"
             git config --global --add safe.directory "${GITHUB_WORKSPACE}/vendor/postgres-v$r"
           done
 
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
+      - uses: actions/checkout@v4
 
       - name: Trigger deploy workflow
         env:
@@ -1055,43 +1046,90 @@ jobs:
               generate_release_notes: true,
             })
 
+  # The job runs on `release` branch and copies compatibility data and Neon artifact from the last *release PR* to the latest directory
   promote-compatibility-data:
-    needs: [ check-permissions, promote-images, tag, build-and-test-locally ]
-    if: github.ref_name == 'release'
+    needs: [ deploy ]
+    # `!failure() && !cancelled()` is required because the workflow transitively depends on the job that can be skipped: `push-to-acr-dev` and `push-to-acr-prod`
+    if: github.ref_name == 'release' && !failure() && !cancelled()
 
-    runs-on: [ self-hosted, small ]
-    container:
-      image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/base:pinned
-      options: --init
+    runs-on: ubuntu-22.04
     steps:
-      - name: Promote compatibility snapshot for the release
+      - name: Fetch GITHUB_RUN_ID and COMMIT_SHA for the last merged release PR
+        id: fetch-last-release-pr-info
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          branch_name_and_pr_number=$(gh pr list \
+            --repo "${GITHUB_REPOSITORY}" \
+            --base release \
+            --state merged \
+            --limit 10 \
+            --json mergeCommit,headRefName,number \
+            --jq ".[] | select(.mergeCommit.oid==\"${GITHUB_SHA}\") | { branch_name: .headRefName, pr_number: .number }")
+          branch_name=$(echo "${branch_name_and_pr_number}" | jq -r '.branch_name')
+          pr_number=$(echo "${branch_name_and_pr_number}" | jq -r '.pr_number')
+
+          run_id=$(gh run list \
+            --repo "${GITHUB_REPOSITORY}" \
+            --workflow build_and_test.yml \
+            --branch "${branch_name}" \
+            --json databaseId \
+            --limit 1 \
+            --jq '.[].databaseId')
+
+          last_commit_sha=$(gh pr view "${pr_number}" \
+            --repo "${GITHUB_REPOSITORY}" \
+            --json commits \
+            --jq '.commits[-1].oid')
+
+          echo "run-id=${run_id}" | tee -a ${GITHUB_OUTPUT}
+          echo "commit-sha=${last_commit_sha}" | tee -a ${GITHUB_OUTPUT}
+
+      - name: Promote compatibility snapshot and Neon artifact
         env:
           BUCKET: neon-github-public-dev
-          PREFIX: artifacts/latest
-          COMMIT_SHA: ${{ github.event.pull_request.head.sha || github.sha }}
+          AWS_REGION: eu-central-1
+          COMMIT_SHA: ${{ steps.fetch-last-release-pr-info.outputs.commit-sha }}
+          RUN_ID: ${{ steps.fetch-last-release-pr-info.outputs.run-id }}
         run: |
-          # Update compatibility snapshot for the release
-          for pg_version in v14 v15 v16; do
-            for build_type in debug release; do
-              OLD_FILENAME=compatibility-snapshot-${build_type}-pg${pg_version}-${GITHUB_RUN_ID}.tar.zst
-              NEW_FILENAME=compatibility-snapshot-${build_type}-pg${pg_version}.tar.zst
+          old_prefix="artifacts/${COMMIT_SHA}/${RUN_ID}"
+          new_prefix="artifacts/latest"
 
-              time aws s3 mv --only-show-errors s3://${BUCKET}/${PREFIX}/${OLD_FILENAME} s3://${BUCKET}/${PREFIX}/${NEW_FILENAME}
+          files_to_promote=()
+          files_on_s3=$(aws s3api list-objects-v2 --bucket ${BUCKET} --prefix ${old_prefix} | jq -r '.Contents[]?.Key' || true)
+
+          for arch in X64 ARM64; do
+            for build_type in debug release; do
+              neon_artifact_filename="neon-Linux-${arch}-${build_type}-artifact.tar.zst"
+              s3_key=$(echo "${files_on_s3}" | grep ${neon_artifact_filename} | sort --version-sort | tail -1 || true)
+              if [ -z "${s3_key}" ]; then
+                echo >&2 "Neither s3://${BUCKET}/${old_prefix}/${neon_artifact_filename} nor its version from previous attempts exist"
+                exit 1
+              fi
+
+              files_to_promote+=("s3://${BUCKET}/${s3_key}")
+
+              # TODO Add v17
+              for pg_version in v14 v15 v16; do
+                # We run less tests for debug builds, so we don't need to promote them
+                if [ "${build_type}" == "debug" ] && { [ "${arch}" == "ARM64" ] || [ "${pg_version}" != "v16" ] ; }; then
+                  continue
+                fi
+
+                compatibility_data_filename="compatibility-snapshot-${arch}-${build_type}-pg${pg_version}.tar.zst"
+                s3_key=$(echo "${files_on_s3}" | grep ${compatibility_data_filename} | sort --version-sort | tail -1 || true)
+                if [ -z "${s3_key}" ]; then
+                  echo >&2 "Neither s3://${BUCKET}/${old_prefix}/${compatibility_data_filename} nor its version from previous attempts exist"
+                  exit 1
+                fi
+
+                files_to_promote+=("s3://${BUCKET}/${s3_key}")
+              done
             done
           done
 
-          # Update Neon artifact for the release (reuse already uploaded artifact)
-          for build_type in debug release; do
-            OLD_PREFIX=artifacts/${COMMIT_SHA}/${GITHUB_RUN_ID}
-            FILENAME=neon-${{ runner.os }}-${{ runner.arch }}-${build_type}-artifact.tar.zst
-
-            S3_KEY=$(aws s3api list-objects-v2 --bucket ${BUCKET} --prefix ${OLD_PREFIX} | jq -r '.Contents[]?.Key' | grep ${FILENAME} | sort --version-sort | tail -1 || true)
-            if [ -z "${S3_KEY}" ]; then
-              echo >&2 "Neither s3://${BUCKET}/${OLD_PREFIX}/${FILENAME} nor its version from previous attempts exist"
-              exit 1
-            fi
-
-            time aws s3 cp --only-show-errors s3://${BUCKET}/${S3_KEY} s3://${BUCKET}/${PREFIX}/${FILENAME}
+          for f in "${files_to_promote[@]}"; do
+            time aws s3 cp --only-show-errors ${f} s3://${BUCKET}/${new_prefix}/
           done
 
   pin-build-tools-image:

.github/workflows/label-for-external-users.yml

@@ -7,6 +7,11 @@ on:
   pull_request_target:
     types:
       - opened
+  workflow_dispatch:
+    inputs:
+      github-actor:
+        description: 'GitHub username. If empty, the username of the current user will be used'
+        required: false
 
 # No permission for GITHUB_TOKEN by default; the **minimal required** set of permissions should be granted in each job.
 permissions: {}
@@ -26,12 +31,31 @@ jobs:
       id: check-user
       env:
         GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
+        ACTOR: ${{ inputs.github-actor || github.actor }}
       run: |
-        if gh api -H "Accept: application/vnd.github+json" -H "X-GitHub-Api-Version: 2022-11-28" "/orgs/${GITHUB_REPOSITORY_OWNER}/members/${GITHUB_ACTOR}"; then
-          is_member=true
-        else
-          is_member=false
-        fi
+        expected_error="User does not exist or is not a member of the organization"
+        output_file=output.txt
+
+        for i in $(seq 1 10); do
+          if gh api "/orgs/${GITHUB_REPOSITORY_OWNER}/members/${ACTOR}" \
+              -H "Accept: application/vnd.github+json" \
+              -H "X-GitHub-Api-Version: 2022-11-28" > ${output_file}; then
+
+            is_member=true
+            break
+          elif grep -q "${expected_error}" ${output_file}; then
+            is_member=false
+            break
+          elif [ $i -eq 10 ]; then
+            title="Failed to get memmbership status for ${ACTOR}"
+            message="The latest GitHub API error message: '$(cat ${output_file})'"
+            echo "::error file=.github/workflows/label-for-external-users.yml,title=${title}::${message}"
+
+            exit 1
+          fi
+
+          sleep 1
+        done
 
         echo "is-member=${is_member}" | tee -a ${GITHUB_OUTPUT}
 

.github/workflows/neon_extra_builds.yml

@@ -72,6 +72,10 @@ jobs:
         id: pg_v16_rev
         run: echo pg_rev=$(git rev-parse HEAD:vendor/postgres-v16) >> $GITHUB_OUTPUT
 
+      - name: Set pg 17 revision for caching
+        id: pg_v17_rev
+        run: echo pg_rev=$(git rev-parse HEAD:vendor/postgres-v17) >> $GITHUB_OUTPUT
+
       - name: Cache postgres v14 build
         id: cache_pg_14
         uses: actions/cache@v4
@@ -93,6 +97,13 @@ jobs:
           path: pg_install/v16
           key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-pg-${{ steps.pg_v16_rev.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
 
+      - name: Cache postgres v17 build
+        id: cache_pg_17
+        uses: actions/cache@v4
+        with:
+          path: pg_install/v17
+          key: v1-${{ runner.os }}-${{ runner.arch }}-${{ env.BUILD_TYPE }}-pg-${{ steps.pg_v17_rev.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
+
       - name: Set extra env for macOS
         run: |
           echo 'LDFLAGS=-L/usr/local/opt/openssl@3/lib' >> $GITHUB_ENV
@@ -120,6 +131,10 @@ jobs:
         if: steps.cache_pg_16.outputs.cache-hit != 'true'
         run: make postgres-v16 -j$(sysctl -n hw.ncpu)
 
+      - name: Build postgres v17
+        if: steps.cache_pg_17.outputs.cache-hit != 'true'
+        run: make postgres-v17 -j$(sysctl -n hw.ncpu)
+
       - name: Build neon extensions
         run: make neon-pg-ext -j$(sysctl -n hw.ncpu)
 
@@ -166,7 +181,7 @@ jobs:
         run: make walproposer-lib -j$(nproc)
 
       - name: Produce the build stats
-        run: PQ_LIB_DIR=$(pwd)/pg_install/v16/lib cargo build --all --release --timings -j$(nproc)
+        run: PQ_LIB_DIR=$(pwd)/pg_install/v17/lib cargo build --all --release --timings -j$(nproc)
 
       - name: Upload the build stats
         id: upload-stats

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

@@ -34,8 +34,8 @@ jobs:
       build-tag: ${{ steps.build-tag.outputs.tag }}
 
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
+      # Need `fetch-depth: 0` to count the number of commits in the branch
+      - uses: actions/checkout@v4
         with:
           fetch-depth: 0
 

.gitmodules

@@ -10,3 +10,7 @@
 	path = vendor/postgres-v16
 	url = https://github.com/neondatabase/postgres.git
 	branch = REL_16_STABLE_neon
+[submodule "vendor/postgres-v17"]
+	path = vendor/postgres-v17
+	url = https://github.com/neondatabase/postgres.git
+	branch = REL_17_STABLE_neon

Cargo.toml

@@ -64,7 +64,8 @@ aws-types = "1.2.0"
 axum = { version = "0.6.20", features = ["ws"] }
 base64 = "0.13.0"
 bincode = "1.3"
-bindgen = "0.65"
+bindgen = "0.70"
+bit_field = "0.10.2"
 bstr = "1.0"
 byteorder = "1.4"
 bytes = "1.0"
@@ -72,7 +73,7 @@ camino = "1.1.6"
 cfg-if = "1.0.0"
 chrono = { version = "0.4", default-features = false, features = ["clock"] }
 clap = { version = "4.0", features = ["derive"] }
-comfy-table = "6.1"
+comfy-table = "7.1"
 const_format = "0.2"
 crc32c = "0.6"
 crossbeam-deque = "0.8.5"
@@ -102,18 +103,18 @@ humantime-serde = "1.1.1"
 hyper = "0.14"
 tokio-tungstenite = "0.20.0"
 indexmap = "2"
+indoc = "2"
 inotify = "0.10.2"
 ipnet = "2.9.0"
 itertools = "0.10"
 jsonwebtoken = "9"
 lasso = "0.7"
-leaky-bucket = "1.0.1"
 libc = "0.2"
 md5 = "0.7.0"
 measured = { version = "0.0.22", features=["lasso"] }
 measured-process = { version = "0.0.22" }
 memoffset = "0.8"
-nix = { version = "0.27", features = ["fs", "process", "socket", "signal", "poll"] }
+nix = { version = "0.27", features = ["dir", "fs", "process", "socket", "signal", "poll"] }
 notify = "6.0.0"
 num_cpus = "1.15"
 num-traits = "0.2.15"
@@ -122,8 +123,8 @@ opentelemetry = "0.20.0"
 opentelemetry-otlp = { version = "0.13.0", default-features=false, features = ["http-proto", "trace", "http", "reqwest-client"] }
 opentelemetry-semantic-conventions = "0.12.0"
 parking_lot = "0.12"
-parquet = { version = "51.0.0", default-features = false, features = ["zstd"] }
-parquet_derive = "51.0.0"
+parquet = { version = "53", default-features = false, features = ["zstd"] }
+parquet_derive = "53"
 pbkdf2 = { version = "0.12.1", features = ["simple", "std"] }
 pin-project-lite = "0.2"
 procfs = "0.16"
@@ -145,6 +146,7 @@ rustls-split = "0.3"
 scopeguard = "1.1"
 sysinfo = "0.29.2"
 sd-notify = "0.4.1"
+send-future = "0.1.0"
 sentry = { version = "0.32", default-features = false, features = ["backtrace", "contexts", "panic", "rustls", "reqwest" ] }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1"
@@ -156,11 +158,10 @@ signal-hook = "0.3"
 smallvec = "1.11"
 smol_str = { version = "0.2.0", features = ["serde"] }
 socket2 = "0.5"
-strum = "0.24"
-strum_macros = "0.24"
+strum = "0.26"
+strum_macros = "0.26"
 "subtle"  = "2.5.0"
-# Our PR https://github.com/nical/rust_debug/pull/4 has been merged but no new version released yet
-svg_fmt = { git = "https://github.com/nical/rust_debug", rev = "28a7d96eecff2f28e75b1ea09f2d499a60d0e3b4" }
+svg_fmt = "0.4.3"
 sync_wrapper = "0.1.2"
 tar = "0.4"
 task-local-extensions = "0.1.4"
@@ -176,8 +177,8 @@ tokio-rustls = "0.25"
 tokio-stream = "0.1"
 tokio-tar = "0.3"
 tokio-util = { version = "0.7.10", features = ["io", "rt"] }
-toml = "0.7"
-toml_edit = "0.19"
+toml = "0.8"
+toml_edit = "0.22"
 tonic = {version = "0.9", features = ["tls", "tls-roots"]}
 tower-service = "0.3.2"
 tracing = "0.1"
@@ -200,10 +201,21 @@ env_logger = "0.10"
 log = "0.4"
 
 ## Libraries from neondatabase/ git forks, ideally with changes to be upstreamed
-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" }
+
+# 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" }
 
 ## Local libraries
 compute_api = { version = "0.1", path = "./libs/compute_api/" }
@@ -240,11 +252,7 @@ tonic-build = "0.9"
 [patch.crates-io]
 
 # Needed to get `tokio-postgres-rustls` to depend on our fork.
-tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", branch="neon" }
-
-# bug fixes for UUID
-parquet = { git = "https://github.com/apache/arrow-rs", branch = "master" }
-parquet_derive = { git = "https://github.com/apache/arrow-rs", branch = "master" }
+tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev = "20031d7a9ee1addeae6e0968e3899ae6bf01cee2" }
 
 ################# Binary contents sections
 

Dockerfile

@@ -5,6 +5,8 @@
 ARG REPOSITORY=neondatabase
 ARG IMAGE=build-tools
 ARG TAG=pinned
+ARG DEFAULT_PG_VERSION=17
+ARG STABLE_PG_VERSION=16
 
 # Build Postgres
 FROM $REPOSITORY/$IMAGE:$TAG AS pg-build
@@ -13,6 +15,7 @@ WORKDIR /home/nonroot
 COPY --chown=nonroot vendor/postgres-v14 vendor/postgres-v14
 COPY --chown=nonroot vendor/postgres-v15 vendor/postgres-v15
 COPY --chown=nonroot vendor/postgres-v16 vendor/postgres-v16
+COPY --chown=nonroot vendor/postgres-v17 vendor/postgres-v17
 COPY --chown=nonroot pgxn pgxn
 COPY --chown=nonroot Makefile Makefile
 COPY --chown=nonroot scripts/ninstall.sh scripts/ninstall.sh
@@ -28,16 +31,19 @@ FROM $REPOSITORY/$IMAGE:$TAG AS build
 WORKDIR /home/nonroot
 ARG GIT_VERSION=local
 ARG BUILD_TAG
+ARG STABLE_PG_VERSION
 
 COPY --from=pg-build /home/nonroot/pg_install/v14/include/postgresql/server pg_install/v14/include/postgresql/server
 COPY --from=pg-build /home/nonroot/pg_install/v15/include/postgresql/server pg_install/v15/include/postgresql/server
 COPY --from=pg-build /home/nonroot/pg_install/v16/include/postgresql/server pg_install/v16/include/postgresql/server
+COPY --from=pg-build /home/nonroot/pg_install/v17/include/postgresql/server pg_install/v17/include/postgresql/server
 COPY --from=pg-build /home/nonroot/pg_install/v16/lib                       pg_install/v16/lib
+COPY --from=pg-build /home/nonroot/pg_install/v17/lib                       pg_install/v17/lib
 COPY --chown=nonroot . .
 
 ARG ADDITIONAL_RUSTFLAGS
 RUN set -e \
-    && PQ_LIB_DIR=$(pwd)/pg_install/v16/lib RUSTFLAGS="-Clinker=clang -Clink-arg=-fuse-ld=mold -Clink-arg=-Wl,--no-rosegment ${ADDITIONAL_RUSTFLAGS}" cargo build \
+    && PQ_LIB_DIR=$(pwd)/pg_install/v${STABLE_PG_VERSION}/lib RUSTFLAGS="-Clinker=clang -Clink-arg=-fuse-ld=mold -Clink-arg=-Wl,--no-rosegment ${ADDITIONAL_RUSTFLAGS}" cargo build \
       --bin pg_sni_router  \
       --bin pageserver  \
       --bin pagectl  \
@@ -52,6 +58,7 @@ RUN set -e \
 # Build final image
 #
 FROM debian:bullseye-slim
+ARG DEFAULT_PG_VERSION
 WORKDIR /data
 
 RUN set -e \
@@ -77,6 +84,7 @@ COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_scrubbe
 COPY --from=pg-build /home/nonroot/pg_install/v14 /usr/local/v14/
 COPY --from=pg-build /home/nonroot/pg_install/v15 /usr/local/v15/
 COPY --from=pg-build /home/nonroot/pg_install/v16 /usr/local/v16/
+COPY --from=pg-build /home/nonroot/pg_install/v17 /usr/local/v17/
 COPY --from=pg-build /home/nonroot/postgres_install.tar.gz /data/
 
 # By default, pageserver uses `.neon/` working directory in WORKDIR, so create one and fill it with the dummy config.
@@ -87,12 +95,13 @@ RUN mkdir -p /data/.neon/ && \
        "pg_distrib_dir='/usr/local/'\n" \
        "listen_pg_addr='0.0.0.0:6400'\n" \
        "listen_http_addr='0.0.0.0:9898'\n" \
+       "availability_zone='local'\n" \
   > /data/.neon/pageserver.toml && \
   chown -R neon:neon /data/.neon
 
 # When running a binary that links with libpq, default to using our most recent postgres version.  Binaries
 # that want a particular postgres version will select it explicitly: this is just a default.
-ENV LD_LIBRARY_PATH=/usr/local/v16/lib
+ENV LD_LIBRARY_PATH=/usr/local/v${DEFAULT_PG_VERSION}/lib
 
 
 VOLUME ["/data"]

Dockerfile.build-tools

@@ -192,7 +192,7 @@ WORKDIR /home/nonroot
 
 # Rust
 # Please keep the version of llvm (installed above) in sync with rust llvm (`rustc --version --verbose | grep LLVM`)
-ENV RUSTC_VERSION=1.80.1
+ENV RUSTC_VERSION=1.81.0
 ENV RUSTUP_HOME="/home/nonroot/.rustup"
 ENV PATH="/home/nonroot/.cargo/bin:${PATH}"
 ARG RUSTFILT_VERSION=0.2.1
@@ -207,7 +207,7 @@ RUN curl -sSO https://static.rust-lang.org/rustup/dist/$(uname -m)-unknown-linux
     export PATH="$HOME/.cargo/bin:$PATH" && \
     . "$HOME/.cargo/env" && \
     cargo --version && rustup --version && \
-    rustup component add llvm-tools-preview rustfmt clippy && \
+    rustup component add llvm-tools rustfmt clippy && \
     cargo install rustfilt            --version ${RUSTFILT_VERSION} && \
     cargo install cargo-hakari        --version ${CARGO_HAKARI_VERSION} && \
     cargo install cargo-deny --locked --version ${CARGO_DENY_VERSION} && \

Dockerfile.compute-node

@@ -55,22 +55,27 @@ RUN cd postgres && \
     # We could add the additional grant statements to the postgres repository but it would be hard to maintain,
     # whenever we need to pick up a new postgres version and we want to limit the changes in our postgres fork,
     # so we do it here.
-    old_list="pg_stat_statements--1.0--1.1.sql pg_stat_statements--1.1--1.2.sql pg_stat_statements--1.2--1.3.sql pg_stat_statements--1.3--1.4.sql pg_stat_statements--1.4--1.5.sql pg_stat_statements--1.4.sql pg_stat_statements--1.5--1.6.sql"; \
-    # the first loop is for pg_stat_statement extension version <= 1.6
     for file in /usr/local/pgsql/share/extension/pg_stat_statements--*.sql; do \
         filename=$(basename "$file"); \
-        if echo "$old_list" | grep -q -F "$filename"; then \
+        # Note that there are no downgrade scripts for pg_stat_statements, so we \
+        # don't have to modify any downgrade paths or (much) older versions: we only \
+        # have to make sure every creation of the pg_stat_statements_reset function \
+        # also adds execute permissions to the neon_superuser.
+        case $filename in \
+          pg_stat_statements--1.4.sql) \
+            # pg_stat_statements_reset is first created with 1.4
             echo 'GRANT EXECUTE ON FUNCTION pg_stat_statements_reset() TO neon_superuser;' >> $file; \
-        fi; \
-    done; \
-    # the second loop is for pg_stat_statement extension versions >= 1.7,
-    # where pg_stat_statement_reset() got 3 additional arguments
-    for file in /usr/local/pgsql/share/extension/pg_stat_statements--*.sql; do \
-        filename=$(basename "$file"); \
-        if ! echo "$old_list" | grep -q -F "$filename"; then \
+            ;; \
+          pg_stat_statements--1.6--1.7.sql) \
+            # Then with the 1.6-1.7 migration it is re-created with a new signature, thus add the permissions back
             echo 'GRANT EXECUTE ON FUNCTION pg_stat_statements_reset(Oid, Oid, bigint) TO neon_superuser;' >> $file; \
-        fi; \
-    done
+            ;; \
+          pg_stat_statements--1.10--1.11.sql) \
+            # Then with the 1.10-1.11 migration it is re-created with a new signature again, thus add the permissions back
+            echo 'GRANT EXECUTE ON FUNCTION pg_stat_statements_reset(Oid, Oid, bigint, boolean) TO neon_superuser;' >> $file; \
+            ;; \
+        esac; \
+    done;
 
 #########################################################################################
 #
@@ -79,6 +84,7 @@ RUN cd postgres && \
 #
 #########################################################################################
 FROM build-deps AS postgis-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 RUN apt update && \
     apt install -y cmake gdal-bin libboost-dev libboost-thread-dev libboost-filesystem-dev \
@@ -87,7 +93,11 @@ RUN apt update && \
     protobuf-c-compiler xsltproc
 
 # SFCGAL > 1.3 requires CGAL > 5.2, Bullseye's libcgal-dev is 5.2
-RUN wget https://gitlab.com/Oslandia/SFCGAL/-/archive/v1.3.10/SFCGAL-v1.3.10.tar.gz -O SFCGAL.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    mkdir -p /sfcgal && \
+    echo "Postgis doensn't yet support PG17 (needs 3.4.3, if not higher)" && exit 0;; \
+    esac && \
+    wget https://gitlab.com/Oslandia/SFCGAL/-/archive/v1.3.10/SFCGAL-v1.3.10.tar.gz -O SFCGAL.tar.gz && \
     echo "4e39b3b2adada6254a7bdba6d297bb28e1a9835a9f879b74f37e2dab70203232 SFCGAL.tar.gz" | sha256sum --check && \
     mkdir sfcgal-src && cd sfcgal-src && tar xzf ../SFCGAL.tar.gz --strip-components=1 -C . && \
     cmake -DCMAKE_BUILD_TYPE=Release . && make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -96,7 +106,10 @@ RUN wget https://gitlab.com/Oslandia/SFCGAL/-/archive/v1.3.10/SFCGAL-v1.3.10.tar
 
 ENV PATH="/usr/local/pgsql/bin:$PATH"
 
-RUN wget https://download.osgeo.org/postgis/source/postgis-3.3.3.tar.gz -O postgis.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "Postgis doensn't yet support PG17 (needs 3.4.3, if not higher)" && exit 0;; \
+    esac && \
+    wget https://download.osgeo.org/postgis/source/postgis-3.3.3.tar.gz -O postgis.tar.gz && \
     echo "74eb356e3f85f14233791013360881b6748f78081cc688ff9d6f0f673a762d13 postgis.tar.gz" | sha256sum --check && \
     mkdir postgis-src && cd postgis-src && tar xzf ../postgis.tar.gz --strip-components=1 -C . && \
     find /usr/local/pgsql -type f | sed 's|^/usr/local/pgsql/||' > /before.txt &&\
@@ -122,7 +135,10 @@ RUN wget https://download.osgeo.org/postgis/source/postgis-3.3.3.tar.gz -O postg
     cp /usr/local/pgsql/share/extension/address_standardizer.control /extensions/postgis && \
     cp /usr/local/pgsql/share/extension/address_standardizer_data_us.control /extensions/postgis
 
-RUN wget https://github.com/pgRouting/pgrouting/archive/v3.4.2.tar.gz -O pgrouting.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/pgRouting/pgrouting/archive/v3.4.2.tar.gz -O pgrouting.tar.gz && \
     echo "cac297c07d34460887c4f3b522b35c470138760fe358e351ad1db4edb6ee306e pgrouting.tar.gz" | sha256sum --check && \
     mkdir pgrouting-src && cd pgrouting-src && tar xzf ../pgrouting.tar.gz --strip-components=1 -C . && \
     mkdir build && cd build && \
@@ -142,12 +158,19 @@ RUN wget https://github.com/pgRouting/pgrouting/archive/v3.4.2.tar.gz -O pgrouti
 #
 #########################################################################################
 FROM build-deps AS plv8-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN apt update && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    apt update && \
     apt install -y ninja-build python3-dev libncurses5 binutils clang
 
-RUN wget https://github.com/plv8/plv8/archive/refs/tags/v3.1.10.tar.gz -O plv8.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/plv8/plv8/archive/refs/tags/v3.1.10.tar.gz -O plv8.tar.gz && \
     echo "7096c3290928561f0d4901b7a52794295dc47f6303102fae3f8e42dd575ad97d plv8.tar.gz" | sha256sum --check && \
     mkdir plv8-src && cd plv8-src && tar xzf ../plv8.tar.gz --strip-components=1 -C . && \
     # generate and copy upgrade scripts
@@ -172,9 +195,13 @@ RUN wget https://github.com/plv8/plv8/archive/refs/tags/v3.1.10.tar.gz -O plv8.t
 #
 #########################################################################################
 FROM build-deps AS h3-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN case "$(uname -m)" in \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    case "$(uname -m)" in \
       "x86_64") \
         export CMAKE_CHECKSUM=739d372726cb23129d57a539ce1432453448816e345e1545f6127296926b6754 \
         ;; \
@@ -192,7 +219,11 @@ RUN case "$(uname -m)" in \
       && /tmp/cmake-install.sh --skip-license --prefix=/usr/local/ \
       && rm /tmp/cmake-install.sh
 
-RUN wget https://github.com/uber/h3/archive/refs/tags/v4.1.0.tar.gz -O h3.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+        mkdir -p /h3/usr/ && \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/uber/h3/archive/refs/tags/v4.1.0.tar.gz -O h3.tar.gz && \
     echo "ec99f1f5974846bde64f4513cf8d2ea1b8d172d2218ab41803bf6a63532272bc h3.tar.gz" | sha256sum --check && \
     mkdir h3-src && cd h3-src && tar xzf ../h3.tar.gz --strip-components=1 -C . && \
     mkdir build && cd build && \
@@ -202,7 +233,10 @@ RUN wget https://github.com/uber/h3/archive/refs/tags/v4.1.0.tar.gz -O h3.tar.gz
     cp -R /h3/usr / && \
     rm -rf build
 
-RUN wget https://github.com/zachasme/h3-pg/archive/refs/tags/v4.1.3.tar.gz -O h3-pg.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/zachasme/h3-pg/archive/refs/tags/v4.1.3.tar.gz -O h3-pg.tar.gz && \
     echo "5c17f09a820859ffe949f847bebf1be98511fb8f1bd86f94932512c00479e324 h3-pg.tar.gz" | sha256sum --check && \
     mkdir h3-pg-src && cd h3-pg-src && tar xzf ../h3-pg.tar.gz --strip-components=1 -C . && \
     export PATH="/usr/local/pgsql/bin:$PATH" && \
@@ -218,9 +252,13 @@ RUN wget https://github.com/zachasme/h3-pg/archive/refs/tags/v4.1.3.tar.gz -O h3
 #
 #########################################################################################
 FROM build-deps AS unit-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/df7cb/postgresql-unit/archive/refs/tags/7.7.tar.gz -O postgresql-unit.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/df7cb/postgresql-unit/archive/refs/tags/7.7.tar.gz -O postgresql-unit.tar.gz && \
     echo "411d05beeb97e5a4abf17572bfcfbb5a68d98d1018918feff995f6ee3bb03e79 postgresql-unit.tar.gz" | sha256sum --check && \
     mkdir postgresql-unit-src && cd postgresql-unit-src && tar xzf ../postgresql-unit.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -239,6 +277,7 @@ RUN wget https://github.com/df7cb/postgresql-unit/archive/refs/tags/7.7.tar.gz -
 #
 #########################################################################################
 FROM build-deps AS vector-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 COPY patches/pgvector.patch /pgvector.patch
@@ -246,7 +285,10 @@ COPY patches/pgvector.patch /pgvector.patch
 # By default, pgvector Makefile uses `-march=native`. We don't want that,
 # because we build the images on different machines than where we run them.
 # Pass OPTFLAGS="" to remove it.
-RUN wget https://github.com/pgvector/pgvector/archive/refs/tags/v0.7.2.tar.gz -O pgvector.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/pgvector/pgvector/archive/refs/tags/v0.7.2.tar.gz -O pgvector.tar.gz && \
     echo "617fba855c9bcb41a2a9bc78a78567fd2e147c72afd5bf9d37b31b9591632b30 pgvector.tar.gz" | sha256sum --check && \
     mkdir pgvector-src && cd pgvector-src && tar xzf ../pgvector.tar.gz --strip-components=1 -C . && \
     patch -p1 < /pgvector.patch && \
@@ -261,10 +303,14 @@ RUN wget https://github.com/pgvector/pgvector/archive/refs/tags/v0.7.2.tar.gz -O
 #
 #########################################################################################
 FROM build-deps AS pgjwt-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 # 9742dab1b2f297ad3811120db7b21451bca2d3c9 made on 13/11/2021
-RUN wget https://github.com/michelp/pgjwt/archive/9742dab1b2f297ad3811120db7b21451bca2d3c9.tar.gz -O pgjwt.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/michelp/pgjwt/archive/9742dab1b2f297ad3811120db7b21451bca2d3c9.tar.gz -O pgjwt.tar.gz && \
     echo "cfdefb15007286f67d3d45510f04a6a7a495004be5b3aecb12cda667e774203f pgjwt.tar.gz" | sha256sum --check && \
     mkdir pgjwt-src && cd pgjwt-src && tar xzf ../pgjwt.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -277,9 +323,13 @@ RUN wget https://github.com/michelp/pgjwt/archive/9742dab1b2f297ad3811120db7b214
 #
 #########################################################################################
 FROM build-deps AS hypopg-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/HypoPG/hypopg/archive/refs/tags/1.4.0.tar.gz -O hypopg.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/HypoPG/hypopg/archive/refs/tags/1.4.0.tar.gz -O hypopg.tar.gz && \
     echo "0821011743083226fc9b813c1f2ef5897a91901b57b6bea85a78e466187c6819 hypopg.tar.gz" | sha256sum --check && \
     mkdir hypopg-src && cd hypopg-src && tar xzf ../hypopg.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -293,9 +343,13 @@ RUN wget https://github.com/HypoPG/hypopg/archive/refs/tags/1.4.0.tar.gz -O hypo
 #
 #########################################################################################
 FROM build-deps AS pg-hashids-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/iCyberon/pg_hashids/archive/refs/tags/v1.2.1.tar.gz -O pg_hashids.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/iCyberon/pg_hashids/archive/refs/tags/v1.2.1.tar.gz -O pg_hashids.tar.gz && \
     echo "74576b992d9277c92196dd8d816baa2cc2d8046fe102f3dcd7f3c3febed6822a pg_hashids.tar.gz" | sha256sum --check && \
     mkdir pg_hashids-src && cd pg_hashids-src && tar xzf ../pg_hashids.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
@@ -309,11 +363,15 @@ RUN wget https://github.com/iCyberon/pg_hashids/archive/refs/tags/v1.2.1.tar.gz
 #
 #########################################################################################
 FROM build-deps AS rum-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 COPY patches/rum.patch /rum.patch
 
-RUN wget https://github.com/postgrespro/rum/archive/refs/tags/1.3.13.tar.gz -O rum.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/postgrespro/rum/archive/refs/tags/1.3.13.tar.gz -O rum.tar.gz && \
     echo "6ab370532c965568df6210bd844ac6ba649f53055e48243525b0b7e5c4d69a7d rum.tar.gz" | sha256sum --check && \
     mkdir rum-src && cd rum-src && tar xzf ../rum.tar.gz --strip-components=1 -C . && \
     patch -p1 < /rum.patch && \
@@ -328,9 +386,13 @@ RUN wget https://github.com/postgrespro/rum/archive/refs/tags/1.3.13.tar.gz -O r
 #
 #########################################################################################
 FROM build-deps AS pgtap-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/theory/pgtap/archive/refs/tags/v1.2.0.tar.gz -O pgtap.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/theory/pgtap/archive/refs/tags/v1.2.0.tar.gz -O pgtap.tar.gz && \
     echo "9c7c3de67ea41638e14f06da5da57bac6f5bd03fea05c165a0ec862205a5c052 pgtap.tar.gz" | sha256sum --check && \
     mkdir pgtap-src && cd pgtap-src && tar xzf ../pgtap.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -344,9 +406,13 @@ RUN wget https://github.com/theory/pgtap/archive/refs/tags/v1.2.0.tar.gz -O pgta
 #
 #########################################################################################
 FROM build-deps AS ip4r-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/RhodiumToad/ip4r/archive/refs/tags/2.4.2.tar.gz -O ip4r.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/RhodiumToad/ip4r/archive/refs/tags/2.4.2.tar.gz -O ip4r.tar.gz && \
     echo "0f7b1f159974f49a47842a8ab6751aecca1ed1142b6d5e38d81b064b2ead1b4b ip4r.tar.gz" | sha256sum --check && \
     mkdir ip4r-src && cd ip4r-src && tar xzf ../ip4r.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -360,9 +426,13 @@ RUN wget https://github.com/RhodiumToad/ip4r/archive/refs/tags/2.4.2.tar.gz -O i
 #
 #########################################################################################
 FROM build-deps AS prefix-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/dimitri/prefix/archive/refs/tags/v1.2.10.tar.gz -O prefix.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/dimitri/prefix/archive/refs/tags/v1.2.10.tar.gz -O prefix.tar.gz && \
     echo "4342f251432a5f6fb05b8597139d3ccde8dcf87e8ca1498e7ee931ca057a8575 prefix.tar.gz" | sha256sum --check && \
     mkdir prefix-src && cd prefix-src && tar xzf ../prefix.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -376,9 +446,13 @@ RUN wget https://github.com/dimitri/prefix/archive/refs/tags/v1.2.10.tar.gz -O p
 #
 #########################################################################################
 FROM build-deps AS hll-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/citusdata/postgresql-hll/archive/refs/tags/v2.18.tar.gz -O hll.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/citusdata/postgresql-hll/archive/refs/tags/v2.18.tar.gz -O hll.tar.gz && \
     echo "e2f55a6f4c4ab95ee4f1b4a2b73280258c5136b161fe9d059559556079694f0e hll.tar.gz" | sha256sum --check && \
     mkdir hll-src && cd hll-src && tar xzf ../hll.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -392,9 +466,13 @@ RUN wget https://github.com/citusdata/postgresql-hll/archive/refs/tags/v2.18.tar
 #
 #########################################################################################
 FROM build-deps AS plpgsql-check-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/okbob/plpgsql_check/archive/refs/tags/v2.5.3.tar.gz -O plpgsql_check.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/okbob/plpgsql_check/archive/refs/tags/v2.5.3.tar.gz -O plpgsql_check.tar.gz && \
     echo "6631ec3e7fb3769eaaf56e3dfedb829aa761abf163d13dba354b4c218508e1c0 plpgsql_check.tar.gz" | sha256sum --check && \
     mkdir plpgsql_check-src && cd plpgsql_check-src && tar xzf ../plpgsql_check.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
@@ -413,7 +491,10 @@ COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 ARG PG_VERSION
 ENV PATH="/usr/local/pgsql/bin:$PATH"
 
-RUN case "${PG_VERSION}" in \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    case "${PG_VERSION}" in \
       "v14" | "v15") \
         export TIMESCALEDB_VERSION=2.10.1 \
         export TIMESCALEDB_CHECKSUM=6fca72a6ed0f6d32d2b3523951ede73dc5f9b0077b38450a029a5f411fdb8c73 \
@@ -446,7 +527,10 @@ COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 ARG PG_VERSION
 ENV PATH="/usr/local/pgsql/bin:$PATH"
 
-RUN case "${PG_VERSION}" in \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    case "${PG_VERSION}" in \
       "v14") \
         export PG_HINT_PLAN_VERSION=14_1_4_1 \
         export PG_HINT_PLAN_CHECKSUM=c3501becf70ead27f70626bce80ea401ceac6a77e2083ee5f3ff1f1444ec1ad1 \
@@ -459,6 +543,9 @@ RUN case "${PG_VERSION}" in \
         export PG_HINT_PLAN_VERSION=16_1_6_0 \
         export PG_HINT_PLAN_CHECKSUM=fc85a9212e7d2819d4ae4ac75817481101833c3cfa9f0fe1f980984e12347d00 \
         ;; \
+      "v17") \
+        echo "TODO: PG17 pg_hint_plan support" && exit 0 \
+        ;; \
       *) \
         echo "Export the valid PG_HINT_PLAN_VERSION variable" && exit 1 \
         ;; \
@@ -478,10 +565,14 @@ RUN case "${PG_VERSION}" in \
 #
 #########################################################################################
 FROM build-deps AS pg-cron-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH="/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/citusdata/pg_cron/archive/refs/tags/v1.6.0.tar.gz -O pg_cron.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/citusdata/pg_cron/archive/refs/tags/v1.6.0.tar.gz -O pg_cron.tar.gz && \
     echo "383a627867d730222c272bfd25cd5e151c578d73f696d32910c7db8c665cc7db pg_cron.tar.gz" | sha256sum --check && \
     mkdir pg_cron-src && cd pg_cron-src && tar xzf ../pg_cron.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -495,9 +586,13 @@ RUN wget https://github.com/citusdata/pg_cron/archive/refs/tags/v1.6.0.tar.gz -O
 #
 #########################################################################################
 FROM build-deps AS rdkit-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN apt-get update && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    apt-get update && \
     apt-get install -y \
         cmake \
         libboost-iostreams1.74-dev \
@@ -507,7 +602,10 @@ RUN apt-get update && \
         libeigen3-dev
 
 ENV PATH="/usr/local/pgsql/bin/:/usr/local/pgsql/:$PATH"
-RUN wget https://github.com/rdkit/rdkit/archive/refs/tags/Release_2023_03_3.tar.gz -O rdkit.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/rdkit/rdkit/archive/refs/tags/Release_2023_03_3.tar.gz -O rdkit.tar.gz && \
     echo "bdbf9a2e6988526bfeb8c56ce3cdfe2998d60ac289078e2215374288185e8c8d rdkit.tar.gz" | sha256sum --check && \
     mkdir rdkit-src && cd rdkit-src && tar xzf ../rdkit.tar.gz --strip-components=1 -C . && \
     cmake \
@@ -544,10 +642,14 @@ RUN wget https://github.com/rdkit/rdkit/archive/refs/tags/Release_2023_03_3.tar.
 #
 #########################################################################################
 FROM build-deps AS pg-uuidv7-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH="/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/fboulnois/pg_uuidv7/archive/refs/tags/v1.0.1.tar.gz -O pg_uuidv7.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/fboulnois/pg_uuidv7/archive/refs/tags/v1.0.1.tar.gz -O pg_uuidv7.tar.gz && \
     echo "0d0759ab01b7fb23851ecffb0bce27822e1868a4a5819bfd276101c716637a7a pg_uuidv7.tar.gz" | sha256sum --check && \
     mkdir pg_uuidv7-src && cd pg_uuidv7-src && tar xzf ../pg_uuidv7.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -561,10 +663,14 @@ RUN wget https://github.com/fboulnois/pg_uuidv7/archive/refs/tags/v1.0.1.tar.gz
 #
 #########################################################################################
 FROM build-deps AS pg-roaringbitmap-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH="/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/ChenHuajun/pg_roaringbitmap/archive/refs/tags/v0.5.4.tar.gz -O pg_roaringbitmap.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions is not supported yet by pg_roaringbitmap. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/ChenHuajun/pg_roaringbitmap/archive/refs/tags/v0.5.4.tar.gz -O pg_roaringbitmap.tar.gz && \
     echo "b75201efcb1c2d1b014ec4ae6a22769cc7a224e6e406a587f5784a37b6b5a2aa pg_roaringbitmap.tar.gz" | sha256sum --check && \
     mkdir pg_roaringbitmap-src && cd pg_roaringbitmap-src && tar xzf ../pg_roaringbitmap.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -578,10 +684,14 @@ RUN wget https://github.com/ChenHuajun/pg_roaringbitmap/archive/refs/tags/v0.5.4
 #
 #########################################################################################
 FROM build-deps AS pg-semver-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH="/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/theory/pg-semver/archive/refs/tags/v0.32.1.tar.gz -O pg_semver.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 is not supported yet by pg_semver. Quit" && exit 0;; \
+    esac && \
+    wget https://github.com/theory/pg-semver/archive/refs/tags/v0.32.1.tar.gz -O pg_semver.tar.gz && \
     echo "fbdaf7512026d62eec03fad8687c15ed509b6ba395bff140acd63d2e4fbe25d7 pg_semver.tar.gz" | sha256sum --check && \
     mkdir pg_semver-src && cd pg_semver-src && tar xzf ../pg_semver.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -620,10 +730,14 @@ RUN case "${PG_VERSION}" in \
 #
 #########################################################################################
 FROM build-deps AS pg-anon-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH="/usr/local/pgsql/bin/:$PATH"
-RUN wget  https://github.com/neondatabase/postgresql_anonymizer/archive/refs/tags/neon_1.1.1.tar.gz -O pg_anon.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "postgresql_anonymizer does not yet support PG17" && exit 0;; \
+    esac && \
+    wget  https://github.com/neondatabase/postgresql_anonymizer/archive/refs/tags/neon_1.1.1.tar.gz -O pg_anon.tar.gz && \
     echo "321ea8d5c1648880aafde850a2c576e4a9e7b9933a34ce272efc839328999fa9  pg_anon.tar.gz" | sha256sum --check && \
     mkdir pg_anon-src && cd pg_anon-src && tar xzf ../pg_anon.tar.gz --strip-components=1 -C . && \
     find /usr/local/pgsql -type f | sed 's|^/usr/local/pgsql/||' > /before.txt &&\
@@ -641,6 +755,7 @@ RUN wget  https://github.com/neondatabase/postgresql_anonymizer/archive/refs/tag
 #
 #########################################################################################
 FROM build-deps AS rust-extensions-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 RUN apt-get update && \
@@ -651,9 +766,11 @@ ENV HOME=/home/nonroot
 ENV PATH="/home/nonroot/.cargo/bin:/usr/local/pgsql/bin/:$PATH"
 USER nonroot
 WORKDIR /home/nonroot
-ARG PG_VERSION
 
-RUN curl -sSO https://static.rust-lang.org/rustup/dist/$(uname -m)-unknown-linux-gnu/rustup-init && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 is not supported yet by pgrx. Quit" && exit 0;; \
+    esac && \
+    curl -sSO https://static.rust-lang.org/rustup/dist/$(uname -m)-unknown-linux-gnu/rustup-init && \
     chmod +x rustup-init && \
     ./rustup-init -y --no-modify-path --profile minimal --default-toolchain stable && \
     rm rustup-init && \
@@ -672,7 +789,10 @@ USER root
 FROM rust-extensions-build AS pg-jsonschema-pg-build
 ARG PG_VERSION
 
-RUN wget https://github.com/supabase/pg_jsonschema/archive/refs/tags/v0.3.1.tar.gz -O pg_jsonschema.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "pg_jsonschema does not yet have a release that supports pg17" && exit 0;; \
+    esac && \
+    wget https://github.com/supabase/pg_jsonschema/archive/refs/tags/v0.3.1.tar.gz -O pg_jsonschema.tar.gz && \
     echo "61df3db1ed83cf24f6aa39c826f8818bfa4f0bd33b587fd6b2b1747985642297 pg_jsonschema.tar.gz" | sha256sum --check && \
     mkdir pg_jsonschema-src && cd pg_jsonschema-src && tar xzf ../pg_jsonschema.tar.gz --strip-components=1 -C . && \
     # see commit 252b3685a27a0f4c31a0f91e983c6314838e89e8
@@ -694,7 +814,10 @@ RUN wget https://github.com/supabase/pg_jsonschema/archive/refs/tags/v0.3.1.tar.
 FROM rust-extensions-build AS pg-graphql-pg-build
 ARG PG_VERSION
 
-RUN wget https://github.com/supabase/pg_graphql/archive/refs/tags/v1.5.7.tar.gz -O pg_graphql.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "pg_graphql does not yet have a release that supports pg17 as of now" && exit 0;; \
+    esac && \
+    wget https://github.com/supabase/pg_graphql/archive/refs/tags/v1.5.7.tar.gz -O pg_graphql.tar.gz && \
     echo "2b3e567a5b31019cb97ae0e33263c1bcc28580be5a444ac4c8ece5c4be2aea41 pg_graphql.tar.gz" | sha256sum --check && \
     mkdir pg_graphql-src && cd pg_graphql-src && tar xzf ../pg_graphql.tar.gz --strip-components=1 -C . && \
     sed -i 's/pgrx = "=0.11.3"/pgrx = { version = "0.11.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
@@ -714,7 +837,10 @@ FROM rust-extensions-build AS pg-tiktoken-pg-build
 ARG PG_VERSION
 
 # 26806147b17b60763039c6a6878884c41a262318 made on 26/09/2023
-RUN wget https://github.com/kelvich/pg_tiktoken/archive/26806147b17b60763039c6a6878884c41a262318.tar.gz -O pg_tiktoken.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "pg_tiktoken does not have versions, nor support for pg17" && exit 0;; \
+    esac && \
+    wget https://github.com/kelvich/pg_tiktoken/archive/26806147b17b60763039c6a6878884c41a262318.tar.gz -O pg_tiktoken.tar.gz && \
     echo "e64e55aaa38c259512d3e27c572da22c4637418cf124caba904cd50944e5004e pg_tiktoken.tar.gz" | sha256sum --check && \
     mkdir pg_tiktoken-src && cd pg_tiktoken-src && tar xzf ../pg_tiktoken.tar.gz --strip-components=1 -C . && \
     # TODO update pgrx version in the pg_tiktoken repo and remove this line
@@ -733,7 +859,10 @@ RUN wget https://github.com/kelvich/pg_tiktoken/archive/26806147b17b60763039c6a6
 FROM rust-extensions-build AS pg-pgx-ulid-build
 ARG PG_VERSION
 
-RUN wget https://github.com/pksunkara/pgx_ulid/archive/refs/tags/v0.1.5.tar.gz -O pgx_ulid.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "pgx_ulid does not support pg17 as of the latest version (0.1.5)" && exit 0;; \
+    esac && \
+    wget https://github.com/pksunkara/pgx_ulid/archive/refs/tags/v0.1.5.tar.gz -O pgx_ulid.tar.gz && \
     echo "9d1659a2da65af0133d5451c454de31b37364e3502087dadf579f790bc8bef17 pgx_ulid.tar.gz" | sha256sum --check && \
     mkdir pgx_ulid-src && cd pgx_ulid-src && tar xzf ../pgx_ulid.tar.gz --strip-components=1 -C . && \
     sed -i 's/pgrx       = "^0.11.2"/pgrx = { version = "=0.11.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
@@ -748,10 +877,14 @@ RUN wget https://github.com/pksunkara/pgx_ulid/archive/refs/tags/v0.1.5.tar.gz -
 #########################################################################################
 
 FROM build-deps AS wal2json-pg-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH="/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/eulerto/wal2json/archive/refs/tags/wal2json_2_5.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "We'll need to update wal2json to 2.6+ for pg17 support" && exit 0;; \
+    esac && \
+    wget https://github.com/eulerto/wal2json/archive/refs/tags/wal2json_2_5.tar.gz && \
     echo "b516653575541cf221b99cf3f8be9b6821f6dbcfc125675c85f35090f824f00e wal2json_2_5.tar.gz" | sha256sum --check && \
     mkdir wal2json-src && cd wal2json-src && tar xzf ../wal2json_2_5.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -764,10 +897,14 @@ RUN wget https://github.com/eulerto/wal2json/archive/refs/tags/wal2json_2_5.tar.
 #
 #########################################################################################
 FROM build-deps AS pg-ivm-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH="/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/sraoss/pg_ivm/archive/refs/tags/v1.7.tar.gz -O pg_ivm.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "We'll need to update pg_ivm to 1.9+ for pg17 support" && exit 0;; \
+    esac && \
+    wget https://github.com/sraoss/pg_ivm/archive/refs/tags/v1.7.tar.gz -O pg_ivm.tar.gz && \
     echo "ebfde04f99203c7be4b0e873f91104090e2e83e5429c32ac242d00f334224d5e pg_ivm.tar.gz" | sha256sum --check && \
     mkdir pg_ivm-src && cd pg_ivm-src && tar xzf ../pg_ivm.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -781,10 +918,14 @@ RUN wget https://github.com/sraoss/pg_ivm/archive/refs/tags/v1.7.tar.gz -O pg_iv
 #
 #########################################################################################
 FROM build-deps AS pg-partman-build
+ARG PG_VERSION
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH="/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/pgpartman/pg_partman/archive/refs/tags/v5.0.1.tar.gz -O pg_partman.tar.gz && \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "pg_partman doesn't support PG17 yet" && exit 0;; \
+    esac && \
+    wget https://github.com/pgpartman/pg_partman/archive/refs/tags/v5.0.1.tar.gz -O pg_partman.tar.gz && \
     echo "75b541733a9659a6c90dbd40fccb904a630a32880a6e3044d0c4c5f4c8a65525 pg_partman.tar.gz" | sha256sum --check && \
     mkdir pg_partman-src && cd pg_partman-src && tar xzf ../pg_partman.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -854,8 +995,8 @@ RUN make -j $(getconf _NPROCESSORS_ONLN) \
     case "${PG_VERSION}" in \
         "v14" | "v15") \
         ;; \
-        "v16") \
-            echo "Skipping HNSW for PostgreSQL 16" && exit 0 \
+        "v16" | "v17") \
+            echo "Skipping HNSW for PostgreSQL ${PG_VERSION}" && exit 0 \
         ;; \
         *) \
             echo "unexpected PostgreSQL version" && exit 1 \
@@ -899,7 +1040,7 @@ FROM neon-pg-ext-build AS postgres-cleanup-layer
 COPY --from=neon-pg-ext-build /usr/local/pgsql /usr/local/pgsql
 
 # Remove binaries from /bin/ that we won't use (or would manually copy & install otherwise)
-RUN cd /usr/local/pgsql/bin && rm ecpg raster2pgsql shp2pgsql pgtopo_export pgtopo_import pgsql2shp
+RUN cd /usr/local/pgsql/bin && rm -f ecpg raster2pgsql shp2pgsql pgtopo_export pgtopo_import pgsql2shp
 
 # Remove headers that we won't need anymore - we've completed installation of all extensions
 RUN rm -r /usr/local/pgsql/include
@@ -918,7 +1059,10 @@ RUN rm /usr/local/pgsql/lib/lib*.a
 
 FROM neon-pg-ext-build AS neon-pg-ext-test
 ARG PG_VERSION
-RUN mkdir /ext-src
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    mkdir /ext-src
 
 #COPY --from=postgis-build /postgis.tar.gz /ext-src/
 #COPY --from=postgis-build /sfcgal/* /usr
@@ -942,7 +1086,7 @@ COPY --from=hll-pg-build /hll.tar.gz /ext-src
 COPY --from=plpgsql-check-pg-build /plpgsql_check.tar.gz /ext-src
 #COPY --from=timescaledb-pg-build /timescaledb.tar.gz /ext-src
 COPY --from=pg-hint-plan-pg-build /pg_hint_plan.tar.gz /ext-src
-COPY patches/pg_hintplan.patch /ext-src
+COPY patches/pg_hint_plan.patch /ext-src
 COPY --from=pg-cron-pg-build /pg_cron.tar.gz /ext-src
 COPY patches/pg_cron.patch /ext-src
 #COPY --from=pg-pgx-ulid-build /home/nonroot/pgx_ulid.tar.gz /ext-src
@@ -956,18 +1100,39 @@ COPY --from=pg-anon-pg-build /pg_anon.tar.gz /ext-src
 COPY patches/pg_anon.patch /ext-src
 COPY --from=pg-ivm-build /pg_ivm.tar.gz /ext-src
 COPY --from=pg-partman-build /pg_partman.tar.gz /ext-src
-RUN cd /ext-src/ && for f in *.tar.gz; \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    cd /ext-src/ && for f in *.tar.gz; \
     do echo $f; dname=$(echo $f | sed 's/\.tar.*//')-src; \
     rm -rf $dname; mkdir $dname; tar xzf $f --strip-components=1 -C $dname \
     || exit 1; rm -f $f; done
-RUN cd /ext-src/pgvector-src && patch -p1 <../pgvector.patch
-RUN cd /ext-src/rum-src && patch -p1 <../rum.patch
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    cd /ext-src/rum-src && patch -p1 <../rum.patch
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    cd /ext-src/pgvector-src && patch -p1 <../pgvector.patch
 # cmake is required for the h3 test
-RUN apt-get update && apt-get install -y cmake
-RUN patch -p1 < /ext-src/pg_hintplan.patch
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    apt-get update && apt-get install -y cmake
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    cd /ext-src/pg_hint_plan-src && patch -p1 < /ext-src/pg_hint_plan.patch
 COPY --chmod=755 docker-compose/run-tests.sh /run-tests.sh
-RUN patch -p1 </ext-src/pg_anon.patch
-RUN patch -p1 </ext-src/pg_cron.patch
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    patch -p1 </ext-src/pg_anon.patch
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    patch -p1 </ext-src/pg_cron.patch
 ENV PATH=/usr/local/pgsql/bin:$PATH
 ENV PGHOST=compute
 ENV PGPORT=55433

Makefile

@@ -119,6 +119,8 @@ $(POSTGRES_INSTALL_DIR)/build/%/config.status:
 # I'm not sure why it wouldn't work, but this is the only place (apart from
 # the "build-all-versions" entry points) where direct mention of PostgreSQL
 # versions is used.
+.PHONY: postgres-configure-v17
+postgres-configure-v17: $(POSTGRES_INSTALL_DIR)/build/v17/config.status
 .PHONY: postgres-configure-v16
 postgres-configure-v16: $(POSTGRES_INSTALL_DIR)/build/v16/config.status
 .PHONY: postgres-configure-v15
@@ -215,29 +217,31 @@ neon-pg-clean-ext-%:
 # they depend on openssl and other libraries that are not included in our
 # Rust build.
 .PHONY: walproposer-lib
-walproposer-lib: neon-pg-ext-v16
+walproposer-lib: neon-pg-ext-v17
 	+@echo "Compiling walproposer-lib"
 	mkdir -p $(POSTGRES_INSTALL_DIR)/build/walproposer-lib
-	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/v16/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
+	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/v17/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
 		-C $(POSTGRES_INSTALL_DIR)/build/walproposer-lib \
 		-f $(ROOT_PROJECT_DIR)/pgxn/neon/Makefile walproposer-lib
-	cp $(POSTGRES_INSTALL_DIR)/v16/lib/libpgport.a $(POSTGRES_INSTALL_DIR)/build/walproposer-lib
-	cp $(POSTGRES_INSTALL_DIR)/v16/lib/libpgcommon.a $(POSTGRES_INSTALL_DIR)/build/walproposer-lib
-ifeq ($(UNAME_S),Linux)
+	cp $(POSTGRES_INSTALL_DIR)/v17/lib/libpgport.a $(POSTGRES_INSTALL_DIR)/build/walproposer-lib
+	cp $(POSTGRES_INSTALL_DIR)/v17/lib/libpgcommon.a $(POSTGRES_INSTALL_DIR)/build/walproposer-lib
 	$(AR) d $(POSTGRES_INSTALL_DIR)/build/walproposer-lib/libpgport.a \
 		pg_strong_random.o
 	$(AR) d $(POSTGRES_INSTALL_DIR)/build/walproposer-lib/libpgcommon.a \
-		pg_crc32c.o \
-		hmac_openssl.o \
+		checksum_helper.o \
 		cryptohash_openssl.o \
-		scram-common.o \
+		hmac_openssl.o \
 		md5_common.o \
-		checksum_helper.o
+		parse_manifest.o \
+		scram-common.o
+ifeq ($(UNAME_S),Linux)
+	$(AR) d $(POSTGRES_INSTALL_DIR)/build/walproposer-lib/libpgcommon.a \
+		pg_crc32c.o
 endif
 
 .PHONY: walproposer-lib-clean
 walproposer-lib-clean:
-	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/v16/bin/pg_config \
+	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/v17/bin/pg_config \
 		-C $(POSTGRES_INSTALL_DIR)/build/walproposer-lib \
 		-f $(ROOT_PROJECT_DIR)/pgxn/neon/Makefile clean
 
@@ -245,38 +249,44 @@ walproposer-lib-clean:
 neon-pg-ext: \
 	neon-pg-ext-v14 \
 	neon-pg-ext-v15 \
-	neon-pg-ext-v16
+	neon-pg-ext-v16 \
+	neon-pg-ext-v17
 
 .PHONY: neon-pg-clean-ext
 neon-pg-clean-ext: \
 	neon-pg-clean-ext-v14 \
 	neon-pg-clean-ext-v15 \
-	neon-pg-clean-ext-v16
+	neon-pg-clean-ext-v16 \
+	neon-pg-clean-ext-v17
 
 # shorthand to build all Postgres versions
 .PHONY: postgres
 postgres: \
 	postgres-v14 \
 	postgres-v15 \
-	postgres-v16
+	postgres-v16 \
+	postgres-v17
 
 .PHONY: postgres-headers
 postgres-headers: \
 	postgres-headers-v14 \
 	postgres-headers-v15 \
-	postgres-headers-v16
+	postgres-headers-v16 \
+	postgres-headers-v17
 
 .PHONY: postgres-clean
 postgres-clean: \
 	postgres-clean-v14 \
 	postgres-clean-v15 \
-	postgres-clean-v16
+	postgres-clean-v16 \
+	postgres-clean-v17
 
 .PHONY: postgres-check
 postgres-check: \
 	postgres-check-v14 \
 	postgres-check-v15 \
-	postgres-check-v16
+	postgres-check-v16 \
+	postgres-check-v17
 
 # This doesn't remove the effects of 'configure'.
 .PHONY: clean
@@ -321,13 +331,13 @@ postgres-%-pgindent: postgres-%-pg-bsd-indent postgres-%-typedefs.list
 	rm -f pg*.BAK
 
 # Indent pxgn/neon.
-.PHONY: pgindent
-neon-pgindent: postgres-v16-pg-bsd-indent neon-pg-ext-v16
-	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/v16/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
-		FIND_TYPEDEF=$(ROOT_PROJECT_DIR)/vendor/postgres-v16/src/tools/find_typedef \
-		INDENT=$(POSTGRES_INSTALL_DIR)/build/v16/src/tools/pg_bsd_indent/pg_bsd_indent \
-		PGINDENT_SCRIPT=$(ROOT_PROJECT_DIR)/vendor/postgres-v16/src/tools/pgindent/pgindent \
-		-C $(POSTGRES_INSTALL_DIR)/build/neon-v16 \
+.PHONY: neon-pgindent
+neon-pgindent: postgres-v17-pg-bsd-indent neon-pg-ext-v17
+	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/v17/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
+		FIND_TYPEDEF=$(ROOT_PROJECT_DIR)/vendor/postgres-v17/src/tools/find_typedef \
+		INDENT=$(POSTGRES_INSTALL_DIR)/build/v17/src/tools/pg_bsd_indent/pg_bsd_indent \
+		PGINDENT_SCRIPT=$(ROOT_PROJECT_DIR)/vendor/postgres-v17/src/tools/pgindent/pgindent \
+		-C $(POSTGRES_INSTALL_DIR)/build/neon-v17 \
 		-f $(ROOT_PROJECT_DIR)/pgxn/neon/Makefile pgindent
 
 

README.md

@@ -64,6 +64,12 @@ brew install protobuf openssl flex bison icu4c pkg-config
 echo 'export PATH="$(brew --prefix openssl)/bin:$PATH"' >> ~/.zshrc
 ```
 
+If you get errors about missing `m4` you may have to install it manually:
+```
+brew install m4
+brew link --force m4
+```
+
 2. [Install Rust](https://www.rust-lang.org/tools/install)
 ```
 # recommended approach from https://www.rust-lang.org/tools/install
@@ -126,7 +132,7 @@ make -j`sysctl -n hw.logicalcpu` -s
 To run the `psql` client, install the `postgresql-client` package or modify `PATH` and `LD_LIBRARY_PATH` to include `pg_install/bin` and `pg_install/lib`, respectively.
 
 To run the integration tests or Python scripts (not required to use the code), install
-Python (3.9 or higher), and install the python3 packages using `./scripts/pysync` (requires [poetry>=1.3](https://python-poetry.org/)) in the project directory.
+Python (3.9 or higher), and install the python3 packages using `./scripts/pysync` (requires [poetry>=1.8](https://python-poetry.org/)) in the project directory.
 
 
 #### Running neon database

compute_tools/Cargo.toml

@@ -11,7 +11,6 @@ testing = []
 
 [dependencies]
 anyhow.workspace = true
-async-compression.workspace = true
 chrono.workspace = true
 cfg-if.workspace = true
 clap.workspace = true
@@ -24,7 +23,6 @@ num_cpus.workspace = true
 opentelemetry.workspace = true
 postgres.workspace = true
 regex.workspace = true
-serde.workspace = true
 serde_json.workspace = true
 signal-hook.workspace = true
 tar.workspace = true
@@ -43,7 +41,6 @@ url.workspace = true
 compute_api.workspace = true
 utils.workspace = true
 workspace_hack.workspace = true
-toml_edit.workspace = true
 remote_storage = { version = "0.1", path = "../libs/remote_storage/" }
 vm_monitor = { version = "0.1", path = "../libs/vm_monitor/" }
 zstd = "0.13"

compute_tools/src/bin/compute_ctl.rs

@@ -44,6 +44,7 @@ use std::{thread, time::Duration};
 use anyhow::{Context, Result};
 use chrono::Utc;
 use clap::Arg;
+use compute_tools::lsn_lease::launch_lsn_lease_bg_task_for_static;
 use signal_hook::consts::{SIGQUIT, SIGTERM};
 use signal_hook::{consts::SIGINT, iterator::Signals};
 use tracing::{error, info, warn};
@@ -366,6 +367,8 @@ fn wait_spec(
         state.start_time = now;
     }
 
+    launch_lsn_lease_bg_task_for_static(&compute);
+
     Ok(WaitSpecResult {
         compute,
         http_port,

compute_tools/src/compute.rs

@@ -1052,26 +1052,19 @@ impl ComputeNode {
         let pg_process = self.start_postgres(pspec.storage_auth_token.clone())?;
 
         let config_time = Utc::now();
-        if pspec.spec.mode == ComputeMode::Primary {
-            if !pspec.spec.skip_pg_catalog_updates {
-                let pgdata_path = Path::new(&self.pgdata);
-                // temporarily reset max_cluster_size in config
-                // to avoid the possibility of hitting the limit, while we are applying config:
-                // creating new extensions, roles, etc...
-                config::with_compute_ctl_tmp_override(
-                    pgdata_path,
-                    "neon.max_cluster_size=-1",
-                    || {
-                        self.pg_reload_conf()?;
-
-                        self.apply_config(&compute_state)?;
-
-                        Ok(())
-                    },
-                )?;
+        if pspec.spec.mode == ComputeMode::Primary && !pspec.spec.skip_pg_catalog_updates {
+            let pgdata_path = Path::new(&self.pgdata);
+            // temporarily reset max_cluster_size in config
+            // to avoid the possibility of hitting the limit, while we are applying config:
+            // creating new extensions, roles, etc...
+            config::with_compute_ctl_tmp_override(pgdata_path, "neon.max_cluster_size=-1", || {
                 self.pg_reload_conf()?;
-            }
-            self.post_apply_config()?;
+
+                self.apply_config(&compute_state)?;
+
+                Ok(())
+            })?;
+            self.pg_reload_conf()?;
         }
 
         let startup_end_time = Utc::now();

compute_tools/src/extension_server.rs

@@ -124,6 +124,7 @@ fn parse_pg_version(human_version: &str) -> &str {
             "14" => return "v14",
             "15" => return "v15",
             "16" => return "v16",
+            "17" => return "v17",
             _ => {}
         },
         _ => {}

compute_tools/src/lib.rs

@@ -11,6 +11,7 @@ pub mod logger;
 pub mod catalog;
 pub mod compute;
 pub mod extension_server;
+pub mod lsn_lease;
 mod migration;
 pub mod monitor;
 pub mod params;

compute_tools/src/lsn_lease.rs

@@ -0,0 +1,186 @@
+use anyhow::bail;
+use anyhow::Result;
+use postgres::{NoTls, SimpleQueryMessage};
+use std::time::SystemTime;
+use std::{str::FromStr, sync::Arc, thread, time::Duration};
+use utils::id::TenantId;
+use utils::id::TimelineId;
+
+use compute_api::spec::ComputeMode;
+use tracing::{info, warn};
+use utils::{
+    lsn::Lsn,
+    shard::{ShardCount, ShardNumber, TenantShardId},
+};
+
+use crate::compute::ComputeNode;
+
+/// Spawns a background thread to periodically renew LSN leases for static compute.
+/// Do nothing if the compute is not in static mode.
+pub fn launch_lsn_lease_bg_task_for_static(compute: &Arc<ComputeNode>) {
+    let (tenant_id, timeline_id, lsn) = {
+        let state = compute.state.lock().unwrap();
+        let spec = state.pspec.as_ref().expect("Spec must be set");
+        match spec.spec.mode {
+            ComputeMode::Static(lsn) => (spec.tenant_id, spec.timeline_id, lsn),
+            _ => return,
+        }
+    };
+    let compute = compute.clone();
+
+    let span = tracing::info_span!("lsn_lease_bg_task", %tenant_id, %timeline_id, %lsn);
+    thread::spawn(move || {
+        let _entered = span.entered();
+        if let Err(e) = lsn_lease_bg_task(compute, tenant_id, timeline_id, lsn) {
+            // TODO: might need stronger error feedback than logging an warning.
+            warn!("Exited with error: {e}");
+        }
+    });
+}
+
+/// Renews lsn lease periodically so static compute are not affected by GC.
+fn lsn_lease_bg_task(
+    compute: Arc<ComputeNode>,
+    tenant_id: TenantId,
+    timeline_id: TimelineId,
+    lsn: Lsn,
+) -> Result<()> {
+    loop {
+        let valid_until = acquire_lsn_lease_with_retry(&compute, tenant_id, timeline_id, lsn)?;
+        let valid_duration = valid_until
+            .duration_since(SystemTime::now())
+            .unwrap_or(Duration::ZERO);
+
+        // Sleep for 60 seconds less than the valid duration but no more than half of the valid duration.
+        let sleep_duration = valid_duration
+            .saturating_sub(Duration::from_secs(60))
+            .max(valid_duration / 2);
+
+        info!(
+            "Succeeded, sleeping for {} seconds",
+            sleep_duration.as_secs()
+        );
+        thread::sleep(sleep_duration);
+    }
+}
+
+/// Acquires lsn lease in a retry loop. Returns the expiration time if a lease is granted.
+/// Returns an error if a lease is explicitly not granted. Otherwise, we keep sending requests.
+fn acquire_lsn_lease_with_retry(
+    compute: &Arc<ComputeNode>,
+    tenant_id: TenantId,
+    timeline_id: TimelineId,
+    lsn: Lsn,
+) -> Result<SystemTime> {
+    let mut attempts = 0usize;
+    let mut retry_period_ms: f64 = 500.0;
+    const MAX_RETRY_PERIOD_MS: f64 = 60.0 * 1000.0;
+
+    loop {
+        // Note: List of pageservers is dynamic, need to re-read configs before each attempt.
+        let configs = {
+            let state = compute.state.lock().unwrap();
+
+            let spec = state.pspec.as_ref().expect("spec must be set");
+
+            let conn_strings = spec.pageserver_connstr.split(',');
+
+            conn_strings
+                .map(|connstr| {
+                    let mut config = postgres::Config::from_str(connstr).expect("Invalid connstr");
+                    if let Some(storage_auth_token) = &spec.storage_auth_token {
+                        info!("Got storage auth token from spec file");
+                        config.password(storage_auth_token.clone());
+                    } else {
+                        info!("Storage auth token not set");
+                    }
+                    config
+                })
+                .collect::<Vec<_>>()
+        };
+
+        let result = try_acquire_lsn_lease(tenant_id, timeline_id, lsn, &configs);
+        match result {
+            Ok(Some(res)) => {
+                return Ok(res);
+            }
+            Ok(None) => {
+                bail!("Permanent error: lease could not be obtained, LSN is behind the GC cutoff");
+            }
+            Err(e) => {
+                warn!("Failed to acquire lsn lease: {e} (attempt {attempts}");
+
+                thread::sleep(Duration::from_millis(retry_period_ms as u64));
+                retry_period_ms *= 1.5;
+                retry_period_ms = retry_period_ms.min(MAX_RETRY_PERIOD_MS);
+            }
+        }
+        attempts += 1;
+    }
+}
+
+/// Tries to acquire an LSN lease through PS page_service API.
+fn try_acquire_lsn_lease(
+    tenant_id: TenantId,
+    timeline_id: TimelineId,
+    lsn: Lsn,
+    configs: &[postgres::Config],
+) -> Result<Option<SystemTime>> {
+    fn get_valid_until(
+        config: &postgres::Config,
+        tenant_shard_id: TenantShardId,
+        timeline_id: TimelineId,
+        lsn: Lsn,
+    ) -> Result<Option<SystemTime>> {
+        let mut client = config.connect(NoTls)?;
+        let cmd = format!("lease lsn {} {} {} ", tenant_shard_id, timeline_id, lsn);
+        let res = client.simple_query(&cmd)?;
+        let msg = match res.first() {
+            Some(msg) => msg,
+            None => bail!("empty response"),
+        };
+        let row = match msg {
+            SimpleQueryMessage::Row(row) => row,
+            _ => bail!("error parsing lsn lease response"),
+        };
+
+        // Note: this will be None if a lease is explicitly not granted.
+        let valid_until_str = row.get("valid_until");
+
+        let valid_until = valid_until_str.map(|s| {
+            SystemTime::UNIX_EPOCH
+                .checked_add(Duration::from_millis(u128::from_str(s).unwrap() as u64))
+                .expect("Time larger than max SystemTime could handle")
+        });
+        Ok(valid_until)
+    }
+
+    let shard_count = configs.len();
+
+    let valid_until = if shard_count > 1 {
+        configs
+            .iter()
+            .enumerate()
+            .map(|(shard_number, config)| {
+                let tenant_shard_id = TenantShardId {
+                    tenant_id,
+                    shard_count: ShardCount::new(shard_count as u8),
+                    shard_number: ShardNumber(shard_number as u8),
+                };
+                get_valid_until(config, tenant_shard_id, timeline_id, lsn)
+            })
+            .collect::<Result<Vec<Option<SystemTime>>>>()?
+            .into_iter()
+            .min()
+            .unwrap()
+    } else {
+        get_valid_until(
+            &configs[0],
+            TenantShardId::unsharded(tenant_id),
+            timeline_id,
+            lsn,
+        )?
+    };
+
+    Ok(valid_until)
+}

compute_tools/src/migrations/0011-grant_pg_show_replication_origin_status_to_neon_superuser.sql

@@ -0,0 +1 @@
+GRANT EXECUTE ON FUNCTION pg_show_replication_origin_status TO neon_superuser;

compute_tools/src/pg_helpers.rs

@@ -22,9 +22,10 @@ use compute_api::spec::{Database, GenericOption, GenericOptions, PgIdent, Role};
 
 const POSTGRES_WAIT_TIMEOUT: Duration = Duration::from_millis(60 * 1000); // milliseconds
 
-/// Escape a string for including it in a SQL literal. Wrapping the result
-/// with `E'{}'` or `'{}'` is not required, as it returns a ready-to-use
-/// SQL string literal, e.g. `'db'''` or `E'db\\'`.
+/// Escape a string for including it in a SQL literal.
+///
+/// Wrapping the result with `E'{}'` or `'{}'` is not required,
+/// as it returns a ready-to-use SQL string literal, e.g. `'db'''` or `E'db\\'`.
 /// See <https://github.com/postgres/postgres/blob/da98d005cdbcd45af563d0c4ac86d0e9772cd15f/src/backend/utils/adt/quote.c#L47>
 /// for the original implementation.
 pub fn escape_literal(s: &str) -> String {

compute_tools/src/spec.rs

@@ -793,6 +793,9 @@ pub fn handle_migrations(client: &mut Client) -> Result<()> {
         include_str!(
             "./migrations/0010-grant_snapshot_synchronization_funcs_to_neon_superuser.sql"
         ),
+        include_str!(
+            "./migrations/0011-grant_pg_show_replication_origin_status_to_neon_superuser.sql"
+        ),
     ];
 
     MigrationRunner::new(client, &migrations).run_migrations()?;

control_plane/Cargo.toml

@@ -6,17 +6,13 @@ license.workspace = true
 
 [dependencies]
 anyhow.workspace = true
-async-trait.workspace = true
 camino.workspace = true
 clap.workspace = true
 comfy-table.workspace = true
-futures.workspace = true
 git-version.workspace = true
 humantime.workspace = true
 nix.workspace = true
 once_cell.workspace = true
-postgres.workspace = true
-hex.workspace = true
 humantime-serde.workspace = true
 hyper.workspace = true
 regex.workspace = true
@@ -24,8 +20,6 @@ reqwest = { workspace = true, features = ["blocking", "json"] }
 scopeguard.workspace = true
 serde.workspace = true
 serde_json.workspace = true
-serde_with.workspace = true
-tar.workspace = true
 thiserror.workspace = true
 toml.workspace = true
 toml_edit.workspace = true

control_plane/src/background_process.rs

@@ -151,7 +151,7 @@ where
                     print!(".");
                     io::stdout().flush().unwrap();
                 }
-                thread::sleep(RETRY_INTERVAL);
+                tokio::time::sleep(RETRY_INTERVAL).await;
             }
             Err(e) => {
                 println!("error starting process {process_name:?}: {e:#}");

control_plane/src/bin/neon_local.rs

@@ -34,12 +34,14 @@ use safekeeper_api::{
     DEFAULT_HTTP_LISTEN_PORT as DEFAULT_SAFEKEEPER_HTTP_PORT,
     DEFAULT_PG_LISTEN_PORT as DEFAULT_SAFEKEEPER_PG_PORT,
 };
+use std::borrow::Cow;
 use std::collections::{BTreeSet, HashMap};
 use std::path::PathBuf;
 use std::process::exit;
 use std::str::FromStr;
 use std::time::Duration;
 use storage_broker::DEFAULT_LISTEN_ADDR as DEFAULT_BROKER_ADDR;
+use tokio::task::JoinSet;
 use url::Host;
 use utils::{
     auth::{Claims, Scope},
@@ -87,34 +89,35 @@ fn main() -> Result<()> {
 
     // Check for 'neon init' command first.
     let subcommand_result = if sub_name == "init" {
-        handle_init(sub_args).map(Some)
+        handle_init(sub_args).map(|env| Some(Cow::Owned(env)))
     } else {
         // all other commands need an existing config
-        let mut env =
-            LocalEnv::load_config(&local_env::base_path()).context("Error loading config")?;
-        let original_env = env.clone();
 
+        let env = LocalEnv::load_config(&local_env::base_path()).context("Error loading config")?;
+        let original_env = env.clone();
+        let env = Box::leak(Box::new(env));
         let rt = tokio::runtime::Builder::new_current_thread()
             .enable_all()
             .build()
             .unwrap();
 
         let subcommand_result = match sub_name {
-            "tenant" => rt.block_on(handle_tenant(sub_args, &mut env)),
-            "timeline" => rt.block_on(handle_timeline(sub_args, &mut env)),
-            "start" => rt.block_on(handle_start_all(&env, get_start_timeout(sub_args))),
-            "stop" => rt.block_on(handle_stop_all(sub_args, &env)),
-            "pageserver" => rt.block_on(handle_pageserver(sub_args, &env)),
-            "storage_controller" => rt.block_on(handle_storage_controller(sub_args, &env)),
-            "safekeeper" => rt.block_on(handle_safekeeper(sub_args, &env)),
-            "endpoint" => rt.block_on(handle_endpoint(sub_args, &env)),
-            "mappings" => handle_mappings(sub_args, &mut env),
+            "tenant" => rt.block_on(handle_tenant(sub_args, env)),
+            "timeline" => rt.block_on(handle_timeline(sub_args, env)),
+            "start" => rt.block_on(handle_start_all(env, get_start_timeout(sub_args))),
+            "stop" => rt.block_on(handle_stop_all(sub_args, env)),
+            "pageserver" => rt.block_on(handle_pageserver(sub_args, env)),
+            "storage_controller" => rt.block_on(handle_storage_controller(sub_args, env)),
+            "storage_broker" => rt.block_on(handle_storage_broker(sub_args, env)),
+            "safekeeper" => rt.block_on(handle_safekeeper(sub_args, env)),
+            "endpoint" => rt.block_on(handle_endpoint(sub_args, env)),
+            "mappings" => handle_mappings(sub_args, env),
             "pg" => bail!("'pg' subcommand has been renamed to 'endpoint'"),
             _ => bail!("unexpected subcommand {sub_name}"),
         };
 
-        if original_env != env {
-            subcommand_result.map(|()| Some(env))
+        if &original_env != env {
+            subcommand_result.map(|()| Some(Cow::Borrowed(env)))
         } else {
             subcommand_result.map(|()| None)
         }
@@ -640,6 +643,8 @@ async fn handle_timeline(timeline_match: &ArgMatches, env: &mut local_env::Local
         }
         Some(("branch", branch_match)) => {
             let tenant_id = get_tenant_id(branch_match, env)?;
+            let new_timeline_id =
+                parse_timeline_id(branch_match)?.unwrap_or(TimelineId::generate());
             let new_branch_name = branch_match
                 .get_one::<String>("branch-name")
                 .ok_or_else(|| anyhow!("No branch name provided"))?;
@@ -658,7 +663,6 @@ async fn handle_timeline(timeline_match: &ArgMatches, env: &mut local_env::Local
                 .map(|lsn_str| Lsn::from_str(lsn_str))
                 .transpose()
                 .context("Failed to parse ancestor start Lsn from the request")?;
-            let new_timeline_id = TimelineId::generate();
             let storage_controller = StorageController::from_env(env);
             let create_req = TimelineCreateRequest {
                 new_timeline_id,
@@ -1244,49 +1248,122 @@ async fn handle_safekeeper(sub_match: &ArgMatches, env: &local_env::LocalEnv) ->
     Ok(())
 }
 
+async fn handle_storage_broker(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
+    let (sub_name, sub_args) = match sub_match.subcommand() {
+        Some(broker_command_data) => broker_command_data,
+        None => bail!("no broker subcommand provided"),
+    };
+
+    match sub_name {
+        "start" => {
+            if let Err(e) = broker::start_broker_process(env, get_start_timeout(sub_args)).await {
+                eprintln!("broker start failed: {e}");
+                exit(1);
+            }
+        }
+
+        "stop" => {
+            if let Err(e) = broker::stop_broker_process(env) {
+                eprintln!("broker stop failed: {e}");
+                exit(1);
+            }
+        }
+
+        _ => bail!("Unexpected broker subcommand '{}'", sub_name),
+    }
+    Ok(())
+}
+
 async fn handle_start_all(
-    env: &local_env::LocalEnv,
+    env: &'static local_env::LocalEnv,
     retry_timeout: &Duration,
 ) -> anyhow::Result<()> {
+    let Err(errors) = handle_start_all_impl(env, *retry_timeout).await else {
+        neon_start_status_check(env, retry_timeout)
+            .await
+            .context("status check after successful startup of all services")?;
+        return Ok(());
+    };
+
+    eprintln!("startup failed because one or more services could not be started");
+
+    for e in errors {
+        eprintln!("{e}");
+        let debug_repr = format!("{e:?}");
+        for line in debug_repr.lines() {
+            eprintln!("  {line}");
+        }
+    }
+
+    try_stop_all(env, true).await;
+
+    exit(2);
+}
+
+/// Returns Ok() if and only if all services could be started successfully.
+/// Otherwise, returns the list of errors that occurred during startup.
+async fn handle_start_all_impl(
+    env: &'static local_env::LocalEnv,
+    retry_timeout: Duration,
+) -> Result<(), Vec<anyhow::Error>> {
     // Endpoints are not started automatically
 
-    broker::start_broker_process(env, retry_timeout).await?;
+    let mut js = JoinSet::new();
 
-    // Only start the storage controller if the pageserver is configured to need it
-    if env.control_plane_api.is_some() {
-        let storage_controller = StorageController::from_env(env);
-        if let Err(e) = storage_controller
-            .start(NeonStorageControllerStartArgs::with_default_instance_id(
-                (*retry_timeout).into(),
-            ))
-            .await
-        {
-            eprintln!("storage_controller start failed: {:#}", e);
-            try_stop_all(env, true).await;
-            exit(1);
+    // force infalliblity through closure
+    #[allow(clippy::redundant_closure_call)]
+    (|| {
+        js.spawn(async move {
+            let retry_timeout = retry_timeout;
+            broker::start_broker_process(env, &retry_timeout).await
+        });
+
+        // Only start the storage controller if the pageserver is configured to need it
+        if env.control_plane_api.is_some() {
+            js.spawn(async move {
+                let storage_controller = StorageController::from_env(env);
+                storage_controller
+                    .start(NeonStorageControllerStartArgs::with_default_instance_id(
+                        retry_timeout.into(),
+                    ))
+                    .await
+                    .map_err(|e| e.context("start storage_controller"))
+            });
+        }
+
+        for ps_conf in &env.pageservers {
+            js.spawn(async move {
+                let pageserver = PageServerNode::from_env(env, ps_conf);
+                pageserver
+                    .start(&retry_timeout)
+                    .await
+                    .map_err(|e| e.context(format!("start pageserver {}", ps_conf.id)))
+            });
+        }
+
+        for node in env.safekeepers.iter() {
+            js.spawn(async move {
+                let safekeeper = SafekeeperNode::from_env(env, node);
+                safekeeper
+                    .start(vec![], &retry_timeout)
+                    .await
+                    .map_err(|e| e.context(format!("start safekeeper {}", safekeeper.id)))
+            });
+        }
+    })();
+
+    let mut errors = Vec::new();
+    while let Some(result) = js.join_next().await {
+        let result = result.expect("we don't panic or cancel the tasks");
+        if let Err(e) = result {
+            errors.push(e);
         }
     }
 
-    for ps_conf in &env.pageservers {
-        let pageserver = PageServerNode::from_env(env, ps_conf);
-        if let Err(e) = pageserver.start(retry_timeout).await {
-            eprintln!("pageserver {} start failed: {:#}", ps_conf.id, e);
-            try_stop_all(env, true).await;
-            exit(1);
-        }
+    if !errors.is_empty() {
+        return Err(errors);
     }
 
-    for node in env.safekeepers.iter() {
-        let safekeeper = SafekeeperNode::from_env(env, node);
-        if let Err(e) = safekeeper.start(vec![], retry_timeout).await {
-            eprintln!("safekeeper {} start failed: {:#}", safekeeper.id, e);
-            try_stop_all(env, false).await;
-            exit(1);
-        }
-    }
-
-    neon_start_status_check(env, retry_timeout).await?;
-
     Ok(())
 }
 
@@ -1570,7 +1647,6 @@ fn cli() -> Command {
                         .value_parser(value_parser!(PathBuf))
                         .value_name("config")
                 )
-                .arg(pg_version_arg.clone())
                 .arg(force_arg)
         )
         .subcommand(
@@ -1583,6 +1659,7 @@ fn cli() -> Command {
             .subcommand(Command::new("branch")
                 .about("Create a new timeline, using another timeline as a base, copying its data")
                 .arg(tenant_id_arg.clone())
+                .arg(timeline_id_arg.clone())
                 .arg(branch_name_arg.clone())
                 .arg(Arg::new("ancestor-branch-name").long("ancestor-branch-name")
                     .help("Use last Lsn of another timeline (and its data) as base when creating the new timeline. The timeline gets resolved by its branch name.").required(false))
@@ -1671,6 +1748,19 @@ fn cli() -> Command {
                             .arg(stop_mode_arg.clone())
                             .arg(instance_id))
         )
+        .subcommand(
+            Command::new("storage_broker")
+                .arg_required_else_help(true)
+                .about("Manage broker")
+                .subcommand(Command::new("start")
+                            .about("Start broker")
+                            .arg(timeout_arg.clone())
+                )
+                .subcommand(Command::new("stop")
+                            .about("Stop broker")
+                            .arg(stop_mode_arg.clone())
+                )
+        )
         .subcommand(
             Command::new("safekeeper")
                 .arg_required_else_help(true)

control_plane/src/endpoint.rs

@@ -702,7 +702,7 @@ impl Endpoint {
                     }
                 }
             }
-            std::thread::sleep(ATTEMPT_INTERVAL);
+            tokio::time::sleep(ATTEMPT_INTERVAL).await;
         }
 
         // disarm the scopeguard, let the child outlive this function (and neon_local invoction)

control_plane/src/local_env.rs

@@ -165,6 +165,9 @@ pub struct NeonStorageControllerConf {
     pub split_threshold: Option<u64>,
 
     pub max_secondary_lag_bytes: Option<u64>,
+
+    #[serde(with = "humantime_serde")]
+    pub heartbeat_interval: Duration,
 }
 
 impl NeonStorageControllerConf {
@@ -172,6 +175,9 @@ impl NeonStorageControllerConf {
     const DEFAULT_MAX_OFFLINE_INTERVAL: std::time::Duration = std::time::Duration::from_secs(10);
 
     const DEFAULT_MAX_WARMING_UP_INTERVAL: std::time::Duration = std::time::Duration::from_secs(30);
+
+    // Very tight heartbeat interval to speed up tests
+    const DEFAULT_HEARTBEAT_INTERVAL: std::time::Duration = std::time::Duration::from_millis(100);
 }
 
 impl Default for NeonStorageControllerConf {
@@ -183,6 +189,7 @@ impl Default for NeonStorageControllerConf {
             database_url: None,
             split_threshold: None,
             max_secondary_lag_bytes: None,
+            heartbeat_interval: Self::DEFAULT_HEARTBEAT_INTERVAL,
         }
     }
 }
@@ -335,7 +342,7 @@ impl LocalEnv {
 
         #[allow(clippy::manual_range_patterns)]
         match pg_version {
-            14 | 15 | 16 => Ok(path.join(format!("v{pg_version}"))),
+            14 | 15 | 16 | 17 => Ok(path.join(format!("v{pg_version}"))),
             _ => bail!("Unsupported postgres version: {}", pg_version),
         }
     }

control_plane/src/pageserver.rs

@@ -17,9 +17,7 @@ use std::time::Duration;
 
 use anyhow::{bail, Context};
 use camino::Utf8PathBuf;
-use pageserver_api::models::{
-    self, AuxFilePolicy, LocationConfig, TenantHistorySize, TenantInfo, TimelineInfo,
-};
+use pageserver_api::models::{self, AuxFilePolicy, TenantInfo, TimelineInfo};
 use pageserver_api::shard::TenantShardId;
 use pageserver_client::mgmt_api;
 use postgres_backend::AuthType;
@@ -75,14 +73,14 @@ impl PageServerNode {
         }
     }
 
-    fn pageserver_make_identity_toml(&self, node_id: NodeId) -> toml_edit::Document {
-        toml_edit::Document::from_str(&format!("id={node_id}")).unwrap()
+    fn pageserver_make_identity_toml(&self, node_id: NodeId) -> toml_edit::DocumentMut {
+        toml_edit::DocumentMut::from_str(&format!("id={node_id}")).unwrap()
     }
 
     fn pageserver_init_make_toml(
         &self,
         conf: NeonLocalInitPageserverConf,
-    ) -> anyhow::Result<toml_edit::Document> {
+    ) -> anyhow::Result<toml_edit::DocumentMut> {
         assert_eq!(&PageServerConf::from(&conf), &self.conf, "during neon_local init, we derive the runtime state of ps conf (self.conf) from the --config flag fully");
 
         // TODO(christian): instead of what we do here, create a pageserver_api::config::ConfigToml (PR #7656)
@@ -137,9 +135,9 @@ impl PageServerNode {
 
         // Turn `overrides` into a toml document.
         // TODO: above code is legacy code, it should be refactored to use toml_edit directly.
-        let mut config_toml = toml_edit::Document::new();
+        let mut config_toml = toml_edit::DocumentMut::new();
         for fragment_str in overrides {
-            let fragment = toml_edit::Document::from_str(&fragment_str)
+            let fragment = toml_edit::DocumentMut::from_str(&fragment_str)
                 .expect("all fragments in `overrides` are valid toml documents, this function controls that");
             for (key, item) in fragment.iter() {
                 config_toml.insert(key, item.clone());
@@ -181,6 +179,23 @@ impl PageServerNode {
         );
         io::stdout().flush()?;
 
+        // If the config file we got as a CLI argument includes the `availability_zone`
+        // config, then use that to populate the `metadata.json` file for the pageserver.
+        // In production the deployment orchestrator does this for us.
+        let az_id = conf
+            .other
+            .get("availability_zone")
+            .map(|toml| {
+                let az_str = toml.to_string();
+                // Trim the (") chars from the toml representation
+                if az_str.starts_with('"') && az_str.ends_with('"') {
+                    az_str[1..az_str.len() - 1].to_string()
+                } else {
+                    az_str
+                }
+            })
+            .unwrap_or("local".to_string());
+
         let config = self
             .pageserver_init_make_toml(conf)
             .context("make pageserver toml")?;
@@ -216,6 +231,7 @@ impl PageServerNode {
         let (_http_host, http_port) =
             parse_host_port(&self.conf.listen_http_addr).expect("Unable to parse listen_http_addr");
         let http_port = http_port.unwrap_or(9898);
+
         // Intentionally hand-craft JSON: this acts as an implicit format compat test
         // in case the pageserver-side structure is edited, and reflects the real life
         // situation: the metadata is written by some other script.
@@ -226,7 +242,10 @@ impl PageServerNode {
                 postgres_port: self.pg_connection_config.port(),
                 http_host: "localhost".to_string(),
                 http_port,
-                other: HashMap::new(),
+                other: HashMap::from([(
+                    "availability_zone_id".to_string(),
+                    serde_json::json!(az_id),
+                )]),
             })
             .unwrap(),
         )
@@ -303,22 +322,6 @@ impl PageServerNode {
         background_process::stop_process(immediate, "pageserver", &self.pid_file())
     }
 
-    pub async fn page_server_psql_client(
-        &self,
-    ) -> anyhow::Result<(
-        tokio_postgres::Client,
-        tokio_postgres::Connection<tokio_postgres::Socket, tokio_postgres::tls::NoTlsStream>,
-    )> {
-        let mut config = self.pg_connection_config.clone();
-        if self.conf.pg_auth_type == AuthType::NeonJWT {
-            let token = self
-                .env
-                .generate_auth_token(&Claims::new(None, Scope::PageServerApi))?;
-            config = config.set_password(Some(token));
-        }
-        Ok(config.connect_no_tls().await?)
-    }
-
     pub async fn check_status(&self) -> mgmt_api::Result<()> {
         self.http_client.status().await
     }
@@ -519,19 +522,6 @@ impl PageServerNode {
         Ok(())
     }
 
-    pub async fn location_config(
-        &self,
-        tenant_shard_id: TenantShardId,
-        config: LocationConfig,
-        flush_ms: Option<Duration>,
-        lazy: bool,
-    ) -> anyhow::Result<()> {
-        Ok(self
-            .http_client
-            .location_config(tenant_shard_id, config, flush_ms, lazy)
-            .await?)
-    }
-
     pub async fn timeline_list(
         &self,
         tenant_shard_id: &TenantShardId,
@@ -615,14 +605,4 @@ impl PageServerNode {
 
         Ok(())
     }
-
-    pub async fn tenant_synthetic_size(
-        &self,
-        tenant_shard_id: TenantShardId,
-    ) -> anyhow::Result<TenantHistorySize> {
-        Ok(self
-            .http_client
-            .tenant_synthetic_size(tenant_shard_id)
-            .await?)
-    }
 }

control_plane/src/postgresql_conf.rs

@@ -4,13 +4,10 @@
 /// NOTE: This doesn't implement the full, correct postgresql.conf syntax. Just
 /// enough to extract a few settings we need in Neon, assuming you don't do
 /// funny stuff like include-directives or funny escaping.
-use anyhow::{bail, Context, Result};
 use once_cell::sync::Lazy;
 use regex::Regex;
 use std::collections::HashMap;
 use std::fmt;
-use std::io::BufRead;
-use std::str::FromStr;
 
 /// In-memory representation of a postgresql.conf file
 #[derive(Default, Debug)]
@@ -19,84 +16,16 @@ pub struct PostgresConf {
     hash: HashMap<String, String>,
 }
 
-static CONF_LINE_RE: Lazy<Regex> = Lazy::new(|| Regex::new(r"^((?:\w|\.)+)\s*=\s*(\S+)$").unwrap());
-
 impl PostgresConf {
     pub fn new() -> PostgresConf {
         PostgresConf::default()
     }
 
-    /// Read file into memory
-    pub fn read(read: impl std::io::Read) -> Result<PostgresConf> {
-        let mut result = Self::new();
-
-        for line in std::io::BufReader::new(read).lines() {
-            let line = line?;
-
-            // Store each line in a vector, in original format
-            result.lines.push(line.clone());
-
-            // Also parse each line and insert key=value lines into a hash map.
-            //
-            // FIXME: This doesn't match exactly the flex/bison grammar in PostgreSQL.
-            // But it's close enough for our usage.
-            let line = line.trim();
-            if line.starts_with('#') {
-                // comment, ignore
-                continue;
-            } else if let Some(caps) = CONF_LINE_RE.captures(line) {
-                let name = caps.get(1).unwrap().as_str();
-                let raw_val = caps.get(2).unwrap().as_str();
-
-                if let Ok(val) = deescape_str(raw_val) {
-                    // Note: if there's already an entry in the hash map for
-                    // this key, this will replace it. That's the behavior what
-                    // we want; when PostgreSQL reads the file, each line
-                    // overrides any previous value for the same setting.
-                    result.hash.insert(name.to_string(), val.to_string());
-                }
-            }
-        }
-        Ok(result)
-    }
-
     /// Return the current value of 'option'
     pub fn get(&self, option: &str) -> Option<&str> {
         self.hash.get(option).map(|x| x.as_ref())
     }
 
-    /// Return the current value of a field, parsed to the right datatype.
-    ///
-    /// This calls the FromStr::parse() function on the value of the field. If
-    /// the field does not exist, or parsing fails, returns an error.
-    ///
-    pub fn parse_field<T>(&self, field_name: &str, context: &str) -> Result<T>
-    where
-        T: FromStr,
-        <T as FromStr>::Err: std::error::Error + Send + Sync + 'static,
-    {
-        self.get(field_name)
-            .with_context(|| format!("could not find '{}' option {}", field_name, context))?
-            .parse::<T>()
-            .with_context(|| format!("could not parse '{}' option {}", field_name, context))
-    }
-
-    pub fn parse_field_optional<T>(&self, field_name: &str, context: &str) -> Result<Option<T>>
-    where
-        T: FromStr,
-        <T as FromStr>::Err: std::error::Error + Send + Sync + 'static,
-    {
-        if let Some(val) = self.get(field_name) {
-            let result = val
-                .parse::<T>()
-                .with_context(|| format!("could not parse '{}' option {}", field_name, context))?;
-
-            Ok(Some(result))
-        } else {
-            Ok(None)
-        }
-    }
-
     ///
     /// Note: if you call this multiple times for the same option, the config
     /// file will a line for each call. It would be nice to have a function
@@ -154,48 +83,8 @@ fn escape_str(s: &str) -> String {
     }
 }
 
-/// De-escape a possibly-quoted value.
-///
-/// See `DeescapeQuotedString` function in PostgreSQL sources for how PostgreSQL
-/// does this.
-fn deescape_str(s: &str) -> Result<String> {
-    // If the string has a quote at the beginning and end, strip them out.
-    if s.len() >= 2 && s.starts_with('\'') && s.ends_with('\'') {
-        let mut result = String::new();
-
-        let mut iter = s[1..(s.len() - 1)].chars().peekable();
-        while let Some(c) = iter.next() {
-            let newc = if c == '\\' {
-                match iter.next() {
-                    Some('b') => '\x08',
-                    Some('f') => '\x0c',
-                    Some('n') => '\n',
-                    Some('r') => '\r',
-                    Some('t') => '\t',
-                    Some('0'..='7') => {
-                        // TODO
-                        bail!("octal escapes not supported");
-                    }
-                    Some(n) => n,
-                    None => break,
-                }
-            } else if c == '\'' && iter.peek() == Some(&'\'') {
-                // doubled quote becomes just one quote
-                iter.next().unwrap()
-            } else {
-                c
-            };
-
-            result.push(newc);
-        }
-        Ok(result)
-    } else {
-        Ok(s.to_string())
-    }
-}
-
 #[test]
-fn test_postgresql_conf_escapes() -> Result<()> {
+fn test_postgresql_conf_escapes() -> anyhow::Result<()> {
     assert_eq!(escape_str("foo bar"), "'foo bar'");
     // these don't need to be quoted
     assert_eq!(escape_str("foo"), "foo");
@@ -214,13 +103,5 @@ fn test_postgresql_conf_escapes() -> Result<()> {
     assert_eq!(escape_str("fo\\o"), "'fo\\\\o'");
     assert_eq!(escape_str("10 cats"), "'10 cats'");
 
-    // Test de-escaping
-    assert_eq!(deescape_str(&escape_str("foo"))?, "foo");
-    assert_eq!(deescape_str(&escape_str("fo'o\nba\\r"))?, "fo'o\nba\\r");
-    assert_eq!(deescape_str("'\\b\\f\\n\\r\\t'")?, "\x08\x0c\n\r\t");
-
-    // octal-escapes are currently not supported
-    assert!(deescape_str("'foo\\7\\07\\007'").is_err());
-
     Ok(())
 }

control_plane/src/safekeeper.rs

@@ -5,6 +5,7 @@
 //! ```text
 //!   .neon/safekeepers/<safekeeper id>
 //! ```
+use std::future::Future;
 use std::io::Write;
 use std::path::PathBuf;
 use std::time::Duration;
@@ -34,12 +35,10 @@ pub enum SafekeeperHttpError {
 
 type Result<T> = result::Result<T, SafekeeperHttpError>;
 
-#[async_trait::async_trait]
-pub trait ResponseErrorMessageExt: Sized {
-    async fn error_from_body(self) -> Result<Self>;
+pub(crate) trait ResponseErrorMessageExt: Sized {
+    fn error_from_body(self) -> impl Future<Output = Result<Self>> + Send;
 }
 
-#[async_trait::async_trait]
 impl ResponseErrorMessageExt for reqwest::Response {
     async fn error_from_body(self) -> Result<Self> {
         let status = self.status();

control_plane/src/storage_controller.rs

@@ -28,6 +28,7 @@ use utils::{
     auth::{encode_from_key_file, Claims, Scope},
     id::{NodeId, TenantId},
 };
+use whoami::username;
 
 pub struct StorageController {
     env: LocalEnv,
@@ -183,7 +184,7 @@ impl StorageController {
     /// to other versions if that one isn't found.  Some automated tests create circumstances
     /// where only one version is available in pg_distrib_dir, such as `test_remote_extensions`.
     async fn get_pg_dir(&self, dir_name: &str) -> anyhow::Result<Utf8PathBuf> {
-        let prefer_versions = [STORAGE_CONTROLLER_POSTGRES_VERSION, 15, 14];
+        let prefer_versions = [STORAGE_CONTROLLER_POSTGRES_VERSION, 16, 15, 14];
 
         for v in prefer_versions {
             let path = Utf8PathBuf::from_path_buf(self.env.pg_dir(v, dir_name)?).unwrap();
@@ -211,7 +212,16 @@ impl StorageController {
     /// Readiness check for our postgres process
     async fn pg_isready(&self, pg_bin_dir: &Utf8Path, postgres_port: u16) -> anyhow::Result<bool> {
         let bin_path = pg_bin_dir.join("pg_isready");
-        let args = ["-h", "localhost", "-p", &format!("{}", postgres_port)];
+        let args = [
+            "-h",
+            "localhost",
+            "-U",
+            &username(),
+            "-d",
+            DB_NAME,
+            "-p",
+            &format!("{}", postgres_port),
+        ];
         let exitcode = Command::new(bin_path).args(args).spawn()?.wait().await?;
 
         Ok(exitcode.success())
@@ -225,7 +235,11 @@ impl StorageController {
     ///
     /// Returns the database url
     pub async fn setup_database(&self, postgres_port: u16) -> anyhow::Result<String> {
-        let database_url = format!("postgresql://localhost:{}/{DB_NAME}", postgres_port);
+        let database_url = format!(
+            "postgresql://{}@localhost:{}/{DB_NAME}",
+            &username(),
+            postgres_port
+        );
 
         let pg_bin_dir = self.get_pg_bin_dir().await?;
         let createdb_path = pg_bin_dir.join("createdb");
@@ -235,6 +249,10 @@ impl StorageController {
                 "localhost",
                 "-p",
                 &format!("{}", postgres_port),
+                "-U",
+                &username(),
+                "-O",
+                &username(),
                 DB_NAME,
             ])
             .output()
@@ -271,7 +289,7 @@ impl StorageController {
             // But tokio-postgres fork doesn't have this upstream commit:
             // https://github.com/sfackler/rust-postgres/commit/cb609be758f3fb5af537f04b584a2ee0cebd5e79
             // => we should rebase our fork => TODO https://github.com/neondatabase/neon/issues/8399
-            .user(&whoami::username())
+            .user(&username())
             .dbname(DB_NAME)
             .connect(tokio_postgres::NoTls)
             .await
@@ -328,6 +346,12 @@ impl StorageController {
             let pg_log_path = pg_data_path.join("postgres.log");
 
             if !tokio::fs::try_exists(&pg_data_path).await? {
+                let initdb_args = ["-D", pg_data_path.as_ref(), "--username", &username()];
+                tracing::info!(
+                    "Initializing storage controller database with args: {:?}",
+                    initdb_args
+                );
+
                 // Initialize empty database
                 let initdb_path = pg_bin_dir.join("initdb");
                 let mut child = Command::new(&initdb_path)
@@ -335,7 +359,7 @@ impl StorageController {
                         ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
                         ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
                     ])
-                    .args(["-D", pg_data_path.as_ref()])
+                    .args(initdb_args)
                     .spawn()
                     .expect("Failed to spawn initdb");
                 let status = child.wait().await?;
@@ -364,8 +388,14 @@ impl StorageController {
                 pg_data_path.as_ref(),
                 "-l",
                 pg_log_path.as_ref(),
+                "-U",
+                &username(),
                 "start",
             ];
+            tracing::info!(
+                "Starting storage controller database with args: {:?}",
+                db_start_args
+            );
 
             background_process::start_process(
                 "storage_controller_db",
@@ -437,6 +467,8 @@ impl StorageController {
             &humantime::Duration::from(self.config.max_offline).to_string(),
             "--max-warming-up-interval",
             &humantime::Duration::from(self.config.max_warming_up).to_string(),
+            "--heartbeat-interval",
+            &humantime::Duration::from(self.config.heartbeat_interval).to_string(),
             "--address-for-peers",
             &address_for_peers.to_string(),
         ]

control_plane/storcon_cli/Cargo.toml

@@ -11,14 +11,11 @@ clap.workspace = true
 comfy-table.workspace = true
 futures.workspace = true
 humantime.workspace = true
-hyper.workspace = true
 pageserver_api.workspace = true
 pageserver_client.workspace = true
 reqwest.workspace = true
-serde.workspace = true
 serde_json = { workspace = true, features = ["raw_value"] }
 storage_controller_client.workspace = true
-thiserror.workspace = true
 tokio.workspace = true
 tracing.workspace = true
 utils.workspace = true

control_plane/storcon_cli/src/main.rs

@@ -4,8 +4,8 @@ use std::{str::FromStr, time::Duration};
 use clap::{Parser, Subcommand};
 use pageserver_api::{
     controller_api::{
-        NodeAvailabilityWrapper, NodeDescribeResponse, ShardSchedulingPolicy, TenantCreateRequest,
-        TenantDescribeResponse, TenantPolicyRequest,
+        NodeAvailabilityWrapper, NodeDescribeResponse, NodeShardResponse, ShardSchedulingPolicy,
+        TenantCreateRequest, TenantDescribeResponse, TenantPolicyRequest,
     },
     models::{
         EvictionPolicy, EvictionPolicyLayerAccessThreshold, LocationConfigSecondary,
@@ -41,6 +41,8 @@ enum Command {
         listen_http_addr: String,
         #[arg(long)]
         listen_http_port: u16,
+        #[arg(long)]
+        availability_zone_id: String,
     },
 
     /// Modify a node's configuration in the storage controller
@@ -78,7 +80,10 @@ enum Command {
     /// List nodes known to the storage controller
     Nodes {},
     /// List tenants known to the storage controller
-    Tenants {},
+    Tenants {
+        /// If this field is set, it will list the tenants on a specific node
+        node_id: Option<NodeId>,
+    },
     /// Create a new tenant in the storage controller, and by extension on pageservers.
     TenantCreate {
         #[arg(long)]
@@ -147,9 +152,9 @@ enum Command {
         #[arg(long)]
         threshold: humantime::Duration,
     },
-    // Drain a set of specified pageservers by moving the primary attachments to pageservers
+    // Migrate away from a set of specified pageservers by moving the primary attachments to pageservers
     // outside of the specified set.
-    Drain {
+    BulkMigrate {
         // Set of pageserver node ids to drain.
         #[arg(long)]
         nodes: Vec<NodeId>,
@@ -163,6 +168,34 @@ enum Command {
         #[arg(long)]
         dry_run: Option<bool>,
     },
+    /// Start draining the specified pageserver.
+    /// The drain is complete when the schedulling policy returns to active.
+    StartDrain {
+        #[arg(long)]
+        node_id: NodeId,
+    },
+    /// Cancel draining the specified pageserver and wait for `timeout`
+    /// for the operation to be canceled. May be retried.
+    CancelDrain {
+        #[arg(long)]
+        node_id: NodeId,
+        #[arg(long)]
+        timeout: humantime::Duration,
+    },
+    /// Start filling the specified pageserver.
+    /// The drain is complete when the schedulling policy returns to active.
+    StartFill {
+        #[arg(long)]
+        node_id: NodeId,
+    },
+    /// Cancel filling the specified pageserver and wait for `timeout`
+    /// for the operation to be canceled. May be retried.
+    CancelFill {
+        #[arg(long)]
+        node_id: NodeId,
+        #[arg(long)]
+        timeout: humantime::Duration,
+    },
 }
 
 #[derive(Parser)]
@@ -249,6 +282,34 @@ impl FromStr for NodeAvailabilityArg {
     }
 }
 
+async fn wait_for_scheduling_policy<F>(
+    client: Client,
+    node_id: NodeId,
+    timeout: Duration,
+    f: F,
+) -> anyhow::Result<NodeSchedulingPolicy>
+where
+    F: Fn(NodeSchedulingPolicy) -> bool,
+{
+    let waiter = tokio::time::timeout(timeout, async move {
+        loop {
+            let node = client
+                .dispatch::<(), NodeDescribeResponse>(
+                    Method::GET,
+                    format!("control/v1/node/{node_id}"),
+                    None,
+                )
+                .await?;
+
+            if f(node.scheduling) {
+                return Ok::<NodeSchedulingPolicy, mgmt_api::Error>(node.scheduling);
+            }
+        }
+    });
+
+    Ok(waiter.await??)
+}
+
 #[tokio::main]
 async fn main() -> anyhow::Result<()> {
     let cli = Cli::parse();
@@ -266,6 +327,7 @@ async fn main() -> anyhow::Result<()> {
             listen_pg_port,
             listen_http_addr,
             listen_http_port,
+            availability_zone_id,
         } => {
             storcon_client
                 .dispatch::<_, ()>(
@@ -277,6 +339,7 @@ async fn main() -> anyhow::Result<()> {
                         listen_pg_port,
                         listen_http_addr,
                         listen_http_port,
+                        availability_zone_id,
                     }),
                 )
                 .await?;
@@ -343,7 +406,41 @@ async fn main() -> anyhow::Result<()> {
                 )
                 .await?;
         }
-        Command::Tenants {} => {
+        Command::Tenants {
+            node_id: Some(node_id),
+        } => {
+            let describe_response = storcon_client
+                .dispatch::<(), NodeShardResponse>(
+                    Method::GET,
+                    format!("control/v1/node/{node_id}/shards"),
+                    None,
+                )
+                .await?;
+            let shards = describe_response.shards;
+            let mut table = comfy_table::Table::new();
+            table.set_header([
+                "Shard",
+                "Intended Primary/Secondary",
+                "Observed Primary/Secondary",
+            ]);
+            for shard in shards {
+                table.add_row([
+                    format!("{}", shard.tenant_shard_id),
+                    match shard.is_intended_secondary {
+                        None => "".to_string(),
+                        Some(true) => "Secondary".to_string(),
+                        Some(false) => "Primary".to_string(),
+                    },
+                    match shard.is_observed_secondary {
+                        None => "".to_string(),
+                        Some(true) => "Secondary".to_string(),
+                        Some(false) => "Primary".to_string(),
+                    },
+                ]);
+            }
+            println!("{table}");
+        }
+        Command::Tenants { node_id: None } => {
             let mut resp = storcon_client
                 .dispatch::<(), Vec<TenantDescribeResponse>>(
                     Method::GET,
@@ -628,7 +725,7 @@ async fn main() -> anyhow::Result<()> {
                 })
                 .await?;
         }
-        Command::Drain {
+        Command::BulkMigrate {
             nodes,
             concurrency,
             max_shards,
@@ -657,7 +754,7 @@ async fn main() -> anyhow::Result<()> {
             }
 
             if nodes.len() != node_to_drain_descs.len() {
-                anyhow::bail!("Drain requested for node which doesn't exist.")
+                anyhow::bail!("Bulk migration requested away from node which doesn't exist.")
             }
 
             node_to_fill_descs.retain(|desc| {
@@ -669,7 +766,7 @@ async fn main() -> anyhow::Result<()> {
             });
 
             if node_to_fill_descs.is_empty() {
-                anyhow::bail!("There are no nodes to drain to")
+                anyhow::bail!("There are no nodes to migrate to")
             }
 
             // Set the node scheduling policy to draining for the nodes which
@@ -690,7 +787,7 @@ async fn main() -> anyhow::Result<()> {
                     .await?;
             }
 
-            // Perform the drain: move each tenant shard scheduled on a node to
+            // Perform the migration: move each tenant shard scheduled on a node to
             // be drained to a node which is being filled. A simple round robin
             // strategy is used to pick the new node.
             let tenants = storcon_client
@@ -703,13 +800,13 @@ async fn main() -> anyhow::Result<()> {
 
             let mut selected_node_idx = 0;
 
-            struct DrainMove {
+            struct MigrationMove {
                 tenant_shard_id: TenantShardId,
                 from: NodeId,
                 to: NodeId,
             }
 
-            let mut moves: Vec<DrainMove> = Vec::new();
+            let mut moves: Vec<MigrationMove> = Vec::new();
 
             let shards = tenants
                 .into_iter()
@@ -739,7 +836,7 @@ async fn main() -> anyhow::Result<()> {
                     continue;
                 }
 
-                moves.push(DrainMove {
+                moves.push(MigrationMove {
                     tenant_shard_id: shard.tenant_shard_id,
                     from: shard
                         .node_attached
@@ -816,6 +913,67 @@ async fn main() -> anyhow::Result<()> {
                 failure
             );
         }
+        Command::StartDrain { node_id } => {
+            storcon_client
+                .dispatch::<(), ()>(
+                    Method::PUT,
+                    format!("control/v1/node/{node_id}/drain"),
+                    None,
+                )
+                .await?;
+            println!("Drain started for {node_id}");
+        }
+        Command::CancelDrain { node_id, timeout } => {
+            storcon_client
+                .dispatch::<(), ()>(
+                    Method::DELETE,
+                    format!("control/v1/node/{node_id}/drain"),
+                    None,
+                )
+                .await?;
+
+            println!("Waiting for node {node_id} to quiesce on scheduling policy ...");
+
+            let final_policy =
+                wait_for_scheduling_policy(storcon_client, node_id, *timeout, |sched| {
+                    use NodeSchedulingPolicy::*;
+                    matches!(sched, Active | PauseForRestart)
+                })
+                .await?;
+
+            println!(
+                "Drain was cancelled for node {node_id}. Schedulling policy is now {final_policy:?}"
+            );
+        }
+        Command::StartFill { node_id } => {
+            storcon_client
+                .dispatch::<(), ()>(Method::PUT, format!("control/v1/node/{node_id}/fill"), None)
+                .await?;
+
+            println!("Fill started for {node_id}");
+        }
+        Command::CancelFill { node_id, timeout } => {
+            storcon_client
+                .dispatch::<(), ()>(
+                    Method::DELETE,
+                    format!("control/v1/node/{node_id}/fill"),
+                    None,
+                )
+                .await?;
+
+            println!("Waiting for node {node_id} to quiesce on scheduling policy ...");
+
+            let final_policy =
+                wait_for_scheduling_policy(storcon_client, node_id, *timeout, |sched| {
+                    use NodeSchedulingPolicy::*;
+                    matches!(sched, Active)
+                })
+                .await?;
+
+            println!(
+                "Fill was cancelled for node {node_id}. Schedulling policy is now {final_policy:?}"
+            );
+        }
     }
 
     Ok(())

docker-compose/run-tests.sh

@@ -3,7 +3,7 @@ set -x
 
 cd /ext-src || exit 2
 FAILED=
-LIST=$( (echo "${SKIP//","/"\n"}"; ls -d -- *-src) | sort | uniq -u)
+LIST=$( (echo -e "${SKIP//","/"\n"}"; ls -d -- *-src) | sort | uniq -u)
 for d in ${LIST}
 do
        [ -d "${d}" ] || continue

docs/rfcs/037-storage-controller-restarts.md

@@ -0,0 +1,259 @@
+# Rolling Storage Controller Restarts
+
+## Summary
+
+This RFC describes the issues around the current storage controller restart procedure
+and describes an implementation which reduces downtime to a few milliseconds on the happy path.
+
+## Motivation
+
+Storage controller upgrades (restarts, more generally) can cause multi-second availability gaps.
+While the storage controller does not sit on the main data path, it's generally not acceptable
+to block management requests for extended periods of time (e.g. https://github.com/neondatabase/neon/issues/8034).
+
+### Current Implementation
+
+The storage controller runs in a Kubernetes Deployment configured for one replica and strategy set to [Recreate](https://kubernetes.io/docs/concepts/workloads/controllers/deployment/#recreate-deployment).
+In non Kubernetes terms, during an upgrade, the currently running storage controller is stopped and, only after,
+a new instance is created.
+
+At start-up, the storage controller calls into all the pageservers it manages (retrieved from DB) to learn the
+latest locations of all tenant shards present on them. This is usually fast, but can push into tens of seconds
+under unfavourable circumstances: pageservers are heavily loaded or unavailable.
+
+## Prior Art
+
+There's probably as many ways of handling restarts gracefully as there are distributed systems. Some examples include:
+* Active/Standby architectures: Two or more instance of the same service run, but traffic is only routed to one of them.
+For fail-over, traffic is routed to one of the standbys (which becomes active).
+* Consensus Algorithms (Raft, Paxos and friends): The part of consensus we care about here is leader election: peers communicate to each other
+and use a voting scheme that ensures the existence of a single leader (e.g. Raft epochs).
+
+## Requirements
+
+* Reduce storage controller unavailability during upgrades to milliseconds
+* Minimize the interval in which it's possible for more than one storage controller
+to issue reconciles.
+* Have one uniform implementation for restarts and upgrades
+* Fit in with the current Kubernetes deployment scheme
+
+## Non Goals
+
+* Implement our own consensus algorithm from scratch
+* Completely eliminate downtime storage controller downtime. Instead we aim to reduce it to the point where it looks
+like a transient error to the control plane
+
+## Impacted Components
+
+* storage controller
+* deployment orchestration (i.e. Ansible)
+* helm charts
+
+## Terminology
+
+* Observed State: in-memory mapping between tenant shards and their current pageserver locations - currently built up
+at start-up by quering pageservers
+* Deployment: Kubernetes [primitive](https://kubernetes.io/docs/concepts/workloads/controllers/deployment/) that models
+a set of replicas
+
+## Implementation
+
+### High Level Flow
+
+At a very high level the proposed idea is to start a new storage controller instance while
+the previous one is still running and cut-over to it when it becomes ready. The new instance,
+should coordinate with the existing one and transition responsibility gracefully. While the controller
+has built in safety against split-brain situations (via generation numbers), we'd like to avoid such
+scenarios since they can lead to availability issues for tenants that underwent changes while two controllers
+were operating at the same time and require operator intervention to remedy.
+
+### Kubernetes Deployment Configuration
+
+On the Kubernetes configuration side, the proposal is to update the storage controller `Deployment`
+to use `spec.strategy.type = RollingUpdate`, `spec.strategy.rollingUpdate.maxSurge=1` and `spec.strategy.maxUnavailable=0`.
+Under the hood, Kubernetes creates a new replica set and adds one pod to it (`maxSurge=1`). The old replica set does not
+scale down until the new replica set has one replica in the ready state (`maxUnavailable=0`).
+
+The various possible failure scenarios are investigated in the [Handling Failures](#handling-failures) section.
+
+### Storage Controller Start-Up
+
+This section describes the primitives required on the storage controller side and the flow of the happy path.
+
+#### Database Table For Leader Synchronization
+
+A new table should be added to the storage controller database for leader synchronization during startup.
+This table will always contain at most one row. The proposed name for the table is `leader` and the schema
+contains two elements:
+* `hostname`: represents the hostname for the current storage controller leader - should be addressible
+from other pods in the deployment
+* `start_timestamp`: holds the start timestamp for the current storage controller leader (UTC timezone) - only required
+for failure case handling: see [Previous Leader Crashes Before New Leader Readiness](#previous-leader-crashes-before-new-leader-readiness)
+
+Storage controllers will read the leader row at start-up and then update it to mark themselves as the leader
+at the end of the start-up sequence. We want compare-and-exchange semantics for the update: avoid the
+situation where two concurrent updates succeed and overwrite each other. The default Postgres isolation
+level is `READ COMMITTED`, which isn't strict enough here. This update transaction should use at least `REPEATABLE
+READ` isolation level in order to [prevent lost updates](https://www.interdb.jp/pg/pgsql05/08.html). Currently,
+the storage controller uses the stricter `SERIALIZABLE` isolation level for all transactions. This more than suits
+our needs here.
+
+```
+START TRANSACTION ISOLATION LEVEL REPEATABLE READ
+UPDATE leader SET hostname=<new_hostname>, start_timestamp=<new_start_ts>
+WHERE hostname=<old_hostname>, start_timestampt=<old_start_ts>;
+```
+
+If the transaction fails or if no rows have been updated, then the compare-and-exchange is regarded as a failure.
+
+#### Step Down API
+
+A new HTTP endpoint should be added to the storage controller: `POST /control/v1/step_down`. Upon receiving this
+request the leader cancels any pending reconciles and goes into a mode where it replies with 503 to all other APIs
+and does not issue any location configurations to its pageservers. The successful HTTP response will return a serialized
+snapshot of the observed state.
+
+If other step down requests come in after the initial one, the request is handled and the observed state is returned (required
+for failure scenario handling - see [Handling Failures](#handling-failures)).
+
+#### Graceful Restart Happy Path
+
+At start-up, the first thing the storage controller does is retrieve the sole row from the new
+`leader` table. If such an entry exists, send a `/step_down` PUT API call to the current leader.
+This should be retried a few times with a short backoff (see [1]). The aspiring leader loads the
+observed state into memory and the start-up sequence proceeds as usual, but *without* querying the
+pageservers in order to build up the observed state.
+
+Before doing any reconciliations or persistence change, update the `leader` database table as described in the [Database Table For Leader Synchronization](database-table-for-leader-synchronization)
+section. If this step fails, the storage controller process exits.
+
+Note that no row will exist in the `leaders` table for the first graceful restart. In that case, force update the `leader` table
+(without the WHERE clause) and perform with the pre-existing start-up procedure (i.e. build observed state by querying pageservers).
+
+Summary of proposed new start-up sequence:
+1. Call `/step_down`
+2. Perform any pending database migrations
+3. Load state from database
+4. Load observed state returned in step (1) into memory
+5. Do initial heartbeat round (may be moved after 5)
+7. Mark self as leader by updating the database
+8. Reschedule and reconcile everything
+
+Some things to note from the steps above:
+* The storage controller makes no changes to the cluster state before step (5) (i.e. no location config
+calls to the pageserver and no compute notifications)
+* Ask the current leader to step down before loading state from database so we don't get a lost update
+if the transactions overlap.
+* Before loading the observed state at step (3), cross-validate against the database. If validation fails,
+fall back to asking the pageservers about their current locations.
+* Database migrations should only run **after** the previous instance steps down (or the step down times out).
+
+
+[1] The API call might fail because there's no storage controller running (i.e. [restart](#storage-controller-crash-or-restart)),
+so we don't want to extend the unavailability period by much. We still want to retry since that's not the common case.
+
+### Handling Failures
+
+#### Storage Controller Crash Or Restart
+
+The storage controller may crash or be restarted outside of roll-outs. When a new pod is created, its call to
+`/step_down` will fail since the previous leader is no longer reachable. In this case perform the pre-existing
+start-up procedure and update the leader table (with the WHERE clause). If the update fails, the storage controller
+exists and consistency is maintained.
+
+#### Previous Leader Crashes Before New Leader Readiness
+
+When the previous leader (P1) crashes before the new leader (P2) passses the readiness check, Kubernetes will
+reconcile the old replica set and create a new pod for it (P1'). The `/step_down` API call will fail for P1'
+(see [2]).
+
+Now we have two cases to consider:
+* P2 updates the `leader` table first: The database update from P1' will fail and P1' will exit, or be terminated
+by Kubernetes depending on timings.
+* P1' updates the `leader` table first: The `hostname` field of the `leader` row stays the same, but the `start_timestamp` field changes.
+The database update from P2 will fail (since `start_timestamp` does not match). P2 will exit and Kubernetes will
+create a new replacement pod for it (P2'). Now the entire dance starts again, but with P1' as the leader and P2' as the incumbent.
+
+[2] P1 and P1' may (more likely than not) be the same pod and have the same hostname. The implementation
+should avoid this self reference and fail the API call at the client if the persisted hostname matches
+the current one.
+
+#### Previous Leader Crashes After New Leader Readiness
+
+The deployment's replica sets already satisfy the deployment's replica count requirements and the
+Kubernetes deployment rollout will just clean up the dead pod.
+
+#### New Leader Crashes Before Pasing Readiness Check
+
+The deployment controller scales up the new replica sets by creating a new pod. The entire procedure is repeated
+with the new pod.
+
+#### Network Partition Between New Pod and Previous Leader
+
+This feels very unlikely, but should be considered in any case. P2 (the new aspiring leader) fails the `/step_down`
+API call into P1 (the current leader). P2 proceeds with the pre-existing startup procedure and updates the `leader` table.
+Kubernetes will terminate P1, but there may be a brief period where both storage controller can drive reconciles.
+
+### Dealing With Split Brain Scenarios
+
+As we've seen in the previous section, we can end up with two storage controller running at the same time. The split brain
+duration is not bounded since the Kubernetes controller might become partitioned from the pods (unlikely though). While these
+scenarios are not fatal, they can cause tenant unavailability, so we'd like to reduce the chances of this happening.
+The rest of this section sketches some safety measure. It's likely overkill to implement all of them however.
+
+### Ensure Leadership Before Producing Side Effects
+
+The storage controller has two types of side effects: location config requests into pageservers and compute notifications into the control plane.
+Before issuing either, the storage controller could check that it is indeed still the leader by querying the database. Side effects might still be
+applied if they race with the database updatem, but the situation will eventually be detected. The storage controller process should terminate in these cases.
+
+### Leadership Lease
+
+Up until now, the leadership defined by this RFC is static. In order to bound the length of the split brain scenario, we could require the leadership
+to be renewed periodically. Two new columns would be added to the leaders table:
+1. `last_renewed` - timestamp indicating when the lease was last renewed
+2. `lease_duration` - duration indicating the amount of time after which the lease expires
+
+The leader periodically attempts to renew the lease by checking that it is in fact still the legitimate leader and updating `last_renewed` in the
+same transaction. If the update fails, the process exits. New storage controller instances wishing to become leaders must wait for the current lease
+to expire before acquiring leadership if they have not succesfully received a response to the `/step_down` request.
+
+### Notify Pageserver Of Storage Controller Term
+
+Each time that leadership changes, we can bump a `term` integer column in the `leader` table. This term uniquely identifies a leader.
+Location config requests and re-attach responses can include this term. On the pageserver side, keep the latest term in memory and refuse
+anything which contains a stale term (i.e. smaller than the current one).
+
+### Observability
+
+* The storage controller should expose a metric which describes it's state (`Active | WarmingUp | SteppedDown`).
+Per region alerts should be added on this metric which triggers when:
+  + no storage controller has been in the `Active` state for an extended period of time
+  + more than one storage controllers are in the `Active` state
+
+* An alert that periodically verifies that the `leader` table is in sync with the metric above would be very useful.
+We'd have to expose the storage controller read only database to Grafana (perhaps it is already done).
+
+## Alternatives
+
+### Kubernetes Leases
+
+Kubernetes has a [lease primitive](https://kubernetes.io/docs/concepts/architecture/leases/) which can be used to implement leader election.
+Only one instance may hold a lease at any given time. This lease needs to be periodically renewed and has an expiration period.
+
+In our case, it would work something like this:
+* `/step_down` deletes the lease or stops it from renewing
+* lease acquisition becomes part of the start-up procedure
+
+The kubert crate implements a [lightweight lease API](https://docs.rs/kubert/latest/kubert/lease/struct.LeaseManager.html), but it's still
+not exactly trivial to implement.
+
+This approach has the benefit of baked in observability (`kubectl describe lease`), but:
+* We offload the responsibility to Kubernetes which makes it harder to debug when things go wrong.
+* More code surface than the simple "row in database" approach. Also, most of this code would be in
+a dependency not subject to code review, etc.
+* Hard to test. Our testing infra does not run the storage controller in Kubernetes and changing it do
+so is not simple and complictes and the test set-up.
+
+To my mind, the "row in database" approach is straightforward enough that we don't have to offload this
+to something external.

docs/updating-postgres.md

@@ -21,30 +21,21 @@ _Example: 15.4 is the new minor version to upgrade to from 15.3._
 1. Create a new branch based on the stable branch you are updating.
 
     ```shell
-    git checkout -b my-branch REL_15_STABLE_neon
+    git checkout -b my-branch-15 REL_15_STABLE_neon
     ```
 
-1. Tag the last commit on the stable branch you are updating.
+1. Find the upstream release tags you're looking for. They are of the form `REL_X_Y`.
 
-    ```shell
-    git tag REL_15_3_neon
-    ```
-
-1. Push the new tag to the Neon Postgres repository.
-
-    ```shell
-    git push origin REL_15_3_neon
-    ```
-
-1. Find the release tags you're looking for. They are of the form `REL_X_Y`.
-
-1. Rebase the branch you created on the tag and resolve any conflicts.
+1. Merge the upstream tag into the branch you created on the tag and resolve any conflicts.
 
     ```shell
     git fetch upstream REL_15_4
-    git rebase REL_15_4
+    git merge REL_15_4
     ```
 
+    In the commit message of the merge commit, mention if there were
+    any non-trivial conflicts or other issues.
+
 1. Run the Postgres test suite to make sure our commits have not affected
 Postgres in a negative way.
 
@@ -57,7 +48,7 @@ Postgres in a negative way.
 1. Push your branch to the Neon Postgres repository.
 
     ```shell
-    git push origin my-branch
+    git push origin my-branch-15
     ```
 
 1. Clone the Neon repository if you have not done so already.
@@ -74,7 +65,7 @@ branch.
 1. Update the Git submodule.
 
     ```shell
-    git submodule set-branch --branch my-branch vendor/postgres-v15
+    git submodule set-branch --branch my-branch-15 vendor/postgres-v15
     git submodule update --remote vendor/postgres-v15
     ```
 
@@ -89,14 +80,12 @@ minor Postgres release.
 
 1. Create a pull request, and wait for CI to go green.
 
-1. Force push the rebased Postgres branches into the Neon Postgres repository.
+1. Push the Postgres branches with the merge commits into the Neon Postgres repository.
 
     ```shell
-    git push --force origin my-branch:REL_15_STABLE_neon
+    git push origin my-branch-15:REL_15_STABLE_neon
     ```
 
-    It may require disabling various branch protections.
-
 1. Update your Neon PR to point at the branches.
 
     ```shell

libs/compute_api/Cargo.toml

@@ -8,7 +8,6 @@ license.workspace = true
 anyhow.workspace = true
 chrono.workspace = true
 serde.workspace = true
-serde_with.workspace = true
 serde_json.workspace = true
 regex.workspace = true
 

libs/consumption_metrics/Cargo.toml

@@ -5,9 +5,6 @@ edition = "2021"
 license = "Apache-2.0"
 
 [dependencies]
-anyhow.workspace = true
 chrono = { workspace = true, features = ["serde"] }
 rand.workspace = true
 serde.workspace = true
-serde_with.workspace = true
-utils.workspace = true

libs/consumption_metrics/src/lib.rs

@@ -5,7 +5,7 @@ use chrono::{DateTime, Utc};
 use rand::Rng;
 use serde::{Deserialize, Serialize};
 
-#[derive(Serialize, serde::Deserialize, Debug, Clone, Copy, Eq, PartialEq, Ord, PartialOrd)]
+#[derive(Serialize, Deserialize, Debug, Clone, Copy, Eq, PartialEq, Ord, PartialOrd)]
 #[serde(tag = "type")]
 pub enum EventType {
     #[serde(rename = "absolute")]
@@ -107,7 +107,7 @@ pub const CHUNK_SIZE: usize = 1000;
 
 // Just a wrapper around a slice of events
 // to serialize it as `{"events" : [ ] }
-#[derive(serde::Serialize, serde::Deserialize)]
+#[derive(serde::Serialize, Deserialize)]
 pub struct EventChunk<'a, T: Clone> {
     pub events: std::borrow::Cow<'a, [T]>,
 }

libs/desim/Cargo.toml

@@ -12,5 +12,4 @@ bytes.workspace = true
 utils.workspace = true
 parking_lot.workspace = true
 hex.workspace = true
-scopeguard.workspace = true
 smallvec = { workspace = true, features = ["write"] }

libs/metrics/src/lib.rs

@@ -68,6 +68,7 @@ macro_rules! register_uint_gauge {
 static INTERNAL_REGISTRY: Lazy<Registry> = Lazy::new(Registry::new);
 
 /// Register a collector in the internal registry. MUST be called before the first call to `gather()`.
+///
 /// Otherwise, we can have a deadlock in the `gather()` call, trying to register a new collector
 /// while holding the lock.
 pub fn register_internal(c: Box<dyn Collector>) -> prometheus::Result<()> {

libs/pageserver_api/Cargo.toml

@@ -4,6 +4,10 @@ version = "0.1.0"
 edition.workspace = true
 license.workspace = true
 
+[features]
+# See pageserver/Cargo.toml
+testing = ["dep:nix"]
+
 [dependencies]
 serde.workspace = true
 serde_with.workspace = true
@@ -23,6 +27,12 @@ thiserror.workspace = true
 humantime-serde.workspace = true
 chrono = { workspace = true, features = ["serde"] }
 itertools.workspace = true
+storage_broker.workspace = true
+camino = {workspace = true, features = ["serde1"]}
+remote_storage.workspace = true
+postgres_backend.workspace = true
+nix = {workspace = true, optional = true}
+reqwest.workspace = true
 
 [dev-dependencies]
 bincode.workspace = true

libs/pageserver_api/src/config.rs

@@ -1,15 +1,28 @@
-use std::collections::HashMap;
-
-use const_format::formatcp;
+use camino::Utf8PathBuf;
 
 #[cfg(test)]
 mod tests;
 
+use const_format::formatcp;
 pub const DEFAULT_PG_LISTEN_PORT: u16 = 64000;
 pub const DEFAULT_PG_LISTEN_ADDR: &str = formatcp!("127.0.0.1:{DEFAULT_PG_LISTEN_PORT}");
 pub const DEFAULT_HTTP_LISTEN_PORT: u16 = 9898;
 pub const DEFAULT_HTTP_LISTEN_ADDR: &str = formatcp!("127.0.0.1:{DEFAULT_HTTP_LISTEN_PORT}");
 
+use postgres_backend::AuthType;
+use remote_storage::RemoteStorageConfig;
+use serde_with::serde_as;
+use std::{
+    collections::HashMap,
+    num::{NonZeroU64, NonZeroUsize},
+    str::FromStr,
+    time::Duration,
+};
+use utils::logging::LogFormat;
+
+use crate::models::ImageCompressionAlgorithm;
+use crate::models::LsnLease;
+
 // Certain metadata (e.g. externally-addressable name, AZ) is delivered
 // as a separate structure.  This information is not neeed by the pageserver
 // itself, it is only used for registering the pageserver with the control
@@ -29,3 +42,443 @@ pub struct NodeMetadata {
     #[serde(flatten)]
     pub other: HashMap<String, serde_json::Value>,
 }
+
+/// `pageserver.toml`
+///
+/// We use serde derive with `#[serde(default)]` to generate a deserializer
+/// that fills in the default values for each config field.
+///
+/// If there cannot be a static default value because we need to make runtime
+/// checks to determine the default, make it an `Option` (which defaults to None).
+/// The runtime check should be done in the consuming crate, i.e., `pageserver`.
+#[serde_as]
+#[derive(Clone, Debug, serde::Deserialize, serde::Serialize)]
+#[serde(default, deny_unknown_fields)]
+pub struct ConfigToml {
+    // types mapped 1:1 into the runtime PageServerConfig type
+    pub listen_pg_addr: String,
+    pub listen_http_addr: String,
+    pub availability_zone: Option<String>,
+    #[serde(with = "humantime_serde")]
+    pub wait_lsn_timeout: Duration,
+    #[serde(with = "humantime_serde")]
+    pub wal_redo_timeout: Duration,
+    pub superuser: String,
+    pub initdb_cache_dir: Option<Utf8PathBuf>,
+    pub page_cache_size: usize,
+    pub max_file_descriptors: usize,
+    pub pg_distrib_dir: Option<Utf8PathBuf>,
+    #[serde_as(as = "serde_with::DisplayFromStr")]
+    pub http_auth_type: AuthType,
+    #[serde_as(as = "serde_with::DisplayFromStr")]
+    pub pg_auth_type: AuthType,
+    pub auth_validation_public_key_path: Option<Utf8PathBuf>,
+    pub remote_storage: Option<RemoteStorageConfig>,
+    pub tenant_config: TenantConfigToml,
+    #[serde_as(as = "serde_with::DisplayFromStr")]
+    pub broker_endpoint: storage_broker::Uri,
+    #[serde(with = "humantime_serde")]
+    pub broker_keepalive_interval: Duration,
+    #[serde_as(as = "serde_with::DisplayFromStr")]
+    pub log_format: LogFormat,
+    pub concurrent_tenant_warmup: NonZeroUsize,
+    pub concurrent_tenant_size_logical_size_queries: NonZeroUsize,
+    #[serde(with = "humantime_serde")]
+    pub metric_collection_interval: Duration,
+    pub metric_collection_endpoint: Option<reqwest::Url>,
+    pub metric_collection_bucket: Option<RemoteStorageConfig>,
+    #[serde(with = "humantime_serde")]
+    pub synthetic_size_calculation_interval: Duration,
+    pub disk_usage_based_eviction: Option<DiskUsageEvictionTaskConfig>,
+    pub test_remote_failures: u64,
+    pub ondemand_download_behavior_treat_error_as_warn: bool,
+    #[serde(with = "humantime_serde")]
+    pub background_task_maximum_delay: Duration,
+    pub control_plane_api: Option<reqwest::Url>,
+    pub control_plane_api_token: Option<String>,
+    pub control_plane_emergency_mode: bool,
+    pub heatmap_upload_concurrency: usize,
+    pub secondary_download_concurrency: usize,
+    pub virtual_file_io_engine: Option<crate::models::virtual_file::IoEngineKind>,
+    pub ingest_batch_size: u64,
+    pub max_vectored_read_bytes: MaxVectoredReadBytes,
+    pub image_compression: ImageCompressionAlgorithm,
+    pub ephemeral_bytes_per_memory_kb: usize,
+    pub l0_flush: Option<crate::models::L0FlushConfig>,
+    #[serde(skip_serializing)]
+    // TODO(https://github.com/neondatabase/neon/issues/8184): remove after this field is removed from all pageserver.toml's
+    pub compact_level0_phase1_value_access: serde::de::IgnoredAny,
+    pub virtual_file_direct_io: crate::models::virtual_file::DirectIoMode,
+    pub io_buffer_alignment: usize,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
+#[serde(deny_unknown_fields)]
+pub struct DiskUsageEvictionTaskConfig {
+    pub max_usage_pct: utils::serde_percent::Percent,
+    pub min_avail_bytes: u64,
+    #[serde(with = "humantime_serde")]
+    pub period: Duration,
+    #[cfg(feature = "testing")]
+    pub mock_statvfs: Option<statvfs::mock::Behavior>,
+    /// Select sorting for evicted layers
+    #[serde(default)]
+    pub eviction_order: EvictionOrder,
+}
+
+pub mod statvfs {
+    pub mod mock {
+        #[derive(Debug, Clone, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
+        #[serde(tag = "type")]
+        pub enum Behavior {
+            Success {
+                blocksize: u64,
+                total_blocks: u64,
+                name_filter: Option<utils::serde_regex::Regex>,
+            },
+            #[cfg(feature = "testing")]
+            Failure { mocked_error: MockedError },
+        }
+
+        #[cfg(feature = "testing")]
+        #[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
+        #[allow(clippy::upper_case_acronyms)]
+        pub enum MockedError {
+            EIO,
+        }
+
+        #[cfg(feature = "testing")]
+        impl From<MockedError> for nix::Error {
+            fn from(e: MockedError) -> Self {
+                match e {
+                    MockedError::EIO => nix::Error::EIO,
+                }
+            }
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
+#[serde(tag = "type", content = "args")]
+pub enum EvictionOrder {
+    RelativeAccessed {
+        highest_layer_count_loses_first: bool,
+    },
+}
+
+impl Default for EvictionOrder {
+    fn default() -> Self {
+        Self::RelativeAccessed {
+            highest_layer_count_loses_first: true,
+        }
+    }
+}
+
+#[derive(Copy, Clone, Debug, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
+#[serde(transparent)]
+pub struct MaxVectoredReadBytes(pub NonZeroUsize);
+
+/// A tenant's calcuated configuration, which is the result of merging a
+/// tenant's TenantConfOpt with the global TenantConf from PageServerConf.
+///
+/// For storing and transmitting individual tenant's configuration, see
+/// TenantConfOpt.
+#[derive(Debug, Clone, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
+#[serde(deny_unknown_fields, default)]
+pub struct TenantConfigToml {
+    // Flush out an inmemory layer, if it's holding WAL older than this
+    // This puts a backstop on how much WAL needs to be re-digested if the
+    // page server crashes.
+    // This parameter actually determines L0 layer file size.
+    pub checkpoint_distance: u64,
+    // Inmemory layer is also flushed at least once in checkpoint_timeout to
+    // eventually upload WAL after activity is stopped.
+    #[serde(with = "humantime_serde")]
+    pub checkpoint_timeout: Duration,
+    // Target file size, when creating image and delta layers.
+    // This parameter determines L1 layer file size.
+    pub compaction_target_size: u64,
+    // How often to check if there's compaction work to be done.
+    // Duration::ZERO means automatic compaction is disabled.
+    #[serde(with = "humantime_serde")]
+    pub compaction_period: Duration,
+    // Level0 delta layer threshold for compaction.
+    pub compaction_threshold: usize,
+    pub compaction_algorithm: crate::models::CompactionAlgorithmSettings,
+    // Determines how much history is retained, to allow
+    // branching and read replicas at an older point in time.
+    // The unit is #of bytes of WAL.
+    // Page versions older than this are garbage collected away.
+    pub gc_horizon: u64,
+    // Interval at which garbage collection is triggered.
+    // Duration::ZERO means automatic GC is disabled
+    #[serde(with = "humantime_serde")]
+    pub gc_period: Duration,
+    // Delta layer churn threshold to create L1 image layers.
+    pub image_creation_threshold: usize,
+    // Determines how much history is retained, to allow
+    // branching and read replicas at an older point in time.
+    // The unit is time.
+    // Page versions older than this are garbage collected away.
+    #[serde(with = "humantime_serde")]
+    pub pitr_interval: Duration,
+    /// Maximum amount of time to wait while opening a connection to receive wal, before erroring.
+    #[serde(with = "humantime_serde")]
+    pub walreceiver_connect_timeout: Duration,
+    /// Considers safekeepers stalled after no WAL updates were received longer than this threshold.
+    /// A stalled safekeeper will be changed to a newer one when it appears.
+    #[serde(with = "humantime_serde")]
+    pub lagging_wal_timeout: Duration,
+    /// Considers safekeepers lagging when their WAL is behind another safekeeper for more than this threshold.
+    /// A lagging safekeeper will be changed after `lagging_wal_timeout` time elapses since the last WAL update,
+    /// to avoid eager reconnects.
+    pub max_lsn_wal_lag: NonZeroU64,
+    pub eviction_policy: crate::models::EvictionPolicy,
+    pub min_resident_size_override: Option<u64>,
+    // See the corresponding metric's help string.
+    #[serde(with = "humantime_serde")]
+    pub evictions_low_residence_duration_metric_threshold: Duration,
+
+    /// If non-zero, the period between uploads of a heatmap from attached tenants.  This
+    /// may be disabled if a Tenant will not have secondary locations: only secondary
+    /// locations will use the heatmap uploaded by attached locations.
+    #[serde(with = "humantime_serde")]
+    pub heatmap_period: Duration,
+
+    /// If true then SLRU segments are dowloaded on demand, if false SLRU segments are included in basebackup
+    pub lazy_slru_download: bool,
+
+    pub timeline_get_throttle: crate::models::ThrottleConfig,
+
+    // How much WAL must be ingested before checking again whether a new image layer is required.
+    // Expresed in multiples of checkpoint distance.
+    pub image_layer_creation_check_threshold: u8,
+
+    /// Switch to a new aux file policy. Switching this flag requires the user has not written any aux file into
+    /// the storage before, and this flag cannot be switched back. Otherwise there will be data corruptions.
+    /// There is a `last_aux_file_policy` flag which gets persisted in `index_part.json` once the first aux
+    /// file is written.
+    pub switch_aux_file_policy: crate::models::AuxFilePolicy,
+
+    /// The length for an explicit LSN lease request.
+    /// Layers needed to reconstruct pages at LSN will not be GC-ed during this interval.
+    #[serde(with = "humantime_serde")]
+    pub lsn_lease_length: Duration,
+
+    /// The length for an implicit LSN lease granted as part of `get_lsn_by_timestamp` request.
+    /// Layers needed to reconstruct pages at LSN will not be GC-ed during this interval.
+    #[serde(with = "humantime_serde")]
+    pub lsn_lease_length_for_ts: Duration,
+}
+
+pub mod defaults {
+    use crate::models::ImageCompressionAlgorithm;
+
+    pub use storage_broker::DEFAULT_ENDPOINT as BROKER_DEFAULT_ENDPOINT;
+
+    pub const DEFAULT_WAIT_LSN_TIMEOUT: &str = "300 s";
+    pub const DEFAULT_WAL_REDO_TIMEOUT: &str = "60 s";
+
+    pub const DEFAULT_SUPERUSER: &str = "cloud_admin";
+
+    pub const DEFAULT_PAGE_CACHE_SIZE: usize = 8192;
+    pub const DEFAULT_MAX_FILE_DESCRIPTORS: usize = 100;
+
+    pub const DEFAULT_LOG_FORMAT: &str = "plain";
+
+    pub const DEFAULT_CONCURRENT_TENANT_WARMUP: usize = 8;
+
+    pub const DEFAULT_CONCURRENT_TENANT_SIZE_LOGICAL_SIZE_QUERIES: usize = 1;
+
+    pub const DEFAULT_METRIC_COLLECTION_INTERVAL: &str = "10 min";
+    pub const DEFAULT_METRIC_COLLECTION_ENDPOINT: Option<reqwest::Url> = None;
+    pub const DEFAULT_SYNTHETIC_SIZE_CALCULATION_INTERVAL: &str = "10 min";
+    pub const DEFAULT_BACKGROUND_TASK_MAXIMUM_DELAY: &str = "10s";
+
+    pub const DEFAULT_HEATMAP_UPLOAD_CONCURRENCY: usize = 8;
+    pub const DEFAULT_SECONDARY_DOWNLOAD_CONCURRENCY: usize = 1;
+
+    pub const DEFAULT_INGEST_BATCH_SIZE: u64 = 100;
+
+    pub const DEFAULT_MAX_VECTORED_READ_BYTES: usize = 128 * 1024; // 128 KiB
+
+    pub const DEFAULT_IMAGE_COMPRESSION: ImageCompressionAlgorithm =
+        ImageCompressionAlgorithm::Zstd { level: Some(1) };
+
+    pub const DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB: usize = 0;
+
+    pub const DEFAULT_IO_BUFFER_ALIGNMENT: usize = 512;
+}
+
+impl Default for ConfigToml {
+    fn default() -> Self {
+        use defaults::*;
+
+        Self {
+            listen_pg_addr: (DEFAULT_PG_LISTEN_ADDR.to_string()),
+            listen_http_addr: (DEFAULT_HTTP_LISTEN_ADDR.to_string()),
+            availability_zone: (None),
+            wait_lsn_timeout: (humantime::parse_duration(DEFAULT_WAIT_LSN_TIMEOUT)
+                .expect("cannot parse default wait lsn timeout")),
+            wal_redo_timeout: (humantime::parse_duration(DEFAULT_WAL_REDO_TIMEOUT)
+                .expect("cannot parse default wal redo timeout")),
+            superuser: (DEFAULT_SUPERUSER.to_string()),
+            initdb_cache_dir: None,
+            page_cache_size: (DEFAULT_PAGE_CACHE_SIZE),
+            max_file_descriptors: (DEFAULT_MAX_FILE_DESCRIPTORS),
+            pg_distrib_dir: None, // Utf8PathBuf::from("./pg_install"), // TODO: formely, this was std::env::current_dir()
+            http_auth_type: (AuthType::Trust),
+            pg_auth_type: (AuthType::Trust),
+            auth_validation_public_key_path: (None),
+            remote_storage: None,
+            broker_endpoint: (storage_broker::DEFAULT_ENDPOINT
+                .parse()
+                .expect("failed to parse default broker endpoint")),
+            broker_keepalive_interval: (humantime::parse_duration(
+                storage_broker::DEFAULT_KEEPALIVE_INTERVAL,
+            )
+            .expect("cannot parse default keepalive interval")),
+            log_format: (LogFormat::from_str(DEFAULT_LOG_FORMAT).unwrap()),
+
+            concurrent_tenant_warmup: (NonZeroUsize::new(DEFAULT_CONCURRENT_TENANT_WARMUP)
+                .expect("Invalid default constant")),
+            concurrent_tenant_size_logical_size_queries: NonZeroUsize::new(
+                DEFAULT_CONCURRENT_TENANT_SIZE_LOGICAL_SIZE_QUERIES,
+            )
+            .unwrap(),
+            metric_collection_interval: (humantime::parse_duration(
+                DEFAULT_METRIC_COLLECTION_INTERVAL,
+            )
+            .expect("cannot parse default metric collection interval")),
+            synthetic_size_calculation_interval: (humantime::parse_duration(
+                DEFAULT_SYNTHETIC_SIZE_CALCULATION_INTERVAL,
+            )
+            .expect("cannot parse default synthetic size calculation interval")),
+            metric_collection_endpoint: (DEFAULT_METRIC_COLLECTION_ENDPOINT),
+
+            metric_collection_bucket: (None),
+
+            disk_usage_based_eviction: (None),
+
+            test_remote_failures: (0),
+
+            ondemand_download_behavior_treat_error_as_warn: (false),
+
+            background_task_maximum_delay: (humantime::parse_duration(
+                DEFAULT_BACKGROUND_TASK_MAXIMUM_DELAY,
+            )
+            .unwrap()),
+
+            control_plane_api: (None),
+            control_plane_api_token: (None),
+            control_plane_emergency_mode: (false),
+
+            heatmap_upload_concurrency: (DEFAULT_HEATMAP_UPLOAD_CONCURRENCY),
+            secondary_download_concurrency: (DEFAULT_SECONDARY_DOWNLOAD_CONCURRENCY),
+
+            ingest_batch_size: (DEFAULT_INGEST_BATCH_SIZE),
+
+            virtual_file_io_engine: None,
+
+            max_vectored_read_bytes: (MaxVectoredReadBytes(
+                NonZeroUsize::new(DEFAULT_MAX_VECTORED_READ_BYTES).unwrap(),
+            )),
+            image_compression: (DEFAULT_IMAGE_COMPRESSION),
+            ephemeral_bytes_per_memory_kb: (DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB),
+            l0_flush: None,
+            compact_level0_phase1_value_access: Default::default(),
+            virtual_file_direct_io: crate::models::virtual_file::DirectIoMode::default(),
+
+            io_buffer_alignment: DEFAULT_IO_BUFFER_ALIGNMENT,
+
+            tenant_config: TenantConfigToml::default(),
+        }
+    }
+}
+
+pub mod tenant_conf_defaults {
+
+    // FIXME: This current value is very low. I would imagine something like 1 GB or 10 GB
+    // would be more appropriate. But a low value forces the code to be exercised more,
+    // which is good for now to trigger bugs.
+    // This parameter actually determines L0 layer file size.
+    pub const DEFAULT_CHECKPOINT_DISTANCE: u64 = 256 * 1024 * 1024;
+    pub const DEFAULT_CHECKPOINT_TIMEOUT: &str = "10 m";
+
+    // FIXME the below configs are only used by legacy algorithm. The new algorithm
+    // has different parameters.
+
+    // Target file size, when creating image and delta layers.
+    // This parameter determines L1 layer file size.
+    pub const DEFAULT_COMPACTION_TARGET_SIZE: u64 = 128 * 1024 * 1024;
+
+    pub const DEFAULT_COMPACTION_PERIOD: &str = "20 s";
+    pub const DEFAULT_COMPACTION_THRESHOLD: usize = 10;
+    pub const DEFAULT_COMPACTION_ALGORITHM: crate::models::CompactionAlgorithm =
+        crate::models::CompactionAlgorithm::Legacy;
+
+    pub const DEFAULT_GC_HORIZON: u64 = 64 * 1024 * 1024;
+
+    // Large DEFAULT_GC_PERIOD is fine as long as PITR_INTERVAL is larger.
+    // If there's a need to decrease this value, first make sure that GC
+    // doesn't hold a layer map write lock for non-trivial operations.
+    // Relevant: https://github.com/neondatabase/neon/issues/3394
+    pub const DEFAULT_GC_PERIOD: &str = "1 hr";
+    pub const DEFAULT_IMAGE_CREATION_THRESHOLD: usize = 3;
+    pub const DEFAULT_PITR_INTERVAL: &str = "7 days";
+    pub const DEFAULT_WALRECEIVER_CONNECT_TIMEOUT: &str = "10 seconds";
+    pub const DEFAULT_WALRECEIVER_LAGGING_WAL_TIMEOUT: &str = "10 seconds";
+    // The default limit on WAL lag should be set to avoid causing disconnects under high throughput
+    // scenarios: since the broker stats are updated ~1/s, a value of 1GiB should be sufficient for
+    // throughputs up to 1GiB/s per timeline.
+    pub const DEFAULT_MAX_WALRECEIVER_LSN_WAL_LAG: u64 = 1024 * 1024 * 1024;
+    pub const DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD: &str = "24 hour";
+    // By default ingest enough WAL for two new L0 layers before checking if new image
+    // image layers should be created.
+    pub const DEFAULT_IMAGE_LAYER_CREATION_CHECK_THRESHOLD: u8 = 2;
+}
+
+impl Default for TenantConfigToml {
+    fn default() -> Self {
+        use tenant_conf_defaults::*;
+        Self {
+            checkpoint_distance: DEFAULT_CHECKPOINT_DISTANCE,
+            checkpoint_timeout: humantime::parse_duration(DEFAULT_CHECKPOINT_TIMEOUT)
+                .expect("cannot parse default checkpoint timeout"),
+            compaction_target_size: DEFAULT_COMPACTION_TARGET_SIZE,
+            compaction_period: humantime::parse_duration(DEFAULT_COMPACTION_PERIOD)
+                .expect("cannot parse default compaction period"),
+            compaction_threshold: DEFAULT_COMPACTION_THRESHOLD,
+            compaction_algorithm: crate::models::CompactionAlgorithmSettings {
+                kind: DEFAULT_COMPACTION_ALGORITHM,
+            },
+            gc_horizon: DEFAULT_GC_HORIZON,
+            gc_period: humantime::parse_duration(DEFAULT_GC_PERIOD)
+                .expect("cannot parse default gc period"),
+            image_creation_threshold: DEFAULT_IMAGE_CREATION_THRESHOLD,
+            pitr_interval: humantime::parse_duration(DEFAULT_PITR_INTERVAL)
+                .expect("cannot parse default PITR interval"),
+            walreceiver_connect_timeout: humantime::parse_duration(
+                DEFAULT_WALRECEIVER_CONNECT_TIMEOUT,
+            )
+            .expect("cannot parse default walreceiver connect timeout"),
+            lagging_wal_timeout: humantime::parse_duration(DEFAULT_WALRECEIVER_LAGGING_WAL_TIMEOUT)
+                .expect("cannot parse default walreceiver lagging wal timeout"),
+            max_lsn_wal_lag: NonZeroU64::new(DEFAULT_MAX_WALRECEIVER_LSN_WAL_LAG)
+                .expect("cannot parse default max walreceiver Lsn wal lag"),
+            eviction_policy: crate::models::EvictionPolicy::NoEviction,
+            min_resident_size_override: None,
+            evictions_low_residence_duration_metric_threshold: humantime::parse_duration(
+                DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD,
+            )
+            .expect("cannot parse default evictions_low_residence_duration_metric_threshold"),
+            heatmap_period: Duration::ZERO,
+            lazy_slru_download: false,
+            timeline_get_throttle: crate::models::ThrottleConfig::disabled(),
+            image_layer_creation_check_threshold: DEFAULT_IMAGE_LAYER_CREATION_CHECK_THRESHOLD,
+            switch_aux_file_policy: crate::models::AuxFilePolicy::default_tenant_config(),
+            lsn_lease_length: LsnLease::DEFAULT_LENGTH,
+            lsn_lease_length_for_ts: LsnLease::DEFAULT_LENGTH_FOR_TS,
+        }
+    }
+}

libs/pageserver_api/src/controller_api.rs

@@ -1,4 +1,4 @@
-use std::collections::HashSet;
+use std::collections::{HashMap, HashSet};
 use std::str::FromStr;
 use std::time::{Duration, Instant};
 
@@ -8,6 +8,7 @@ use std::time::{Duration, Instant};
 use serde::{Deserialize, Serialize};
 use utils::id::{NodeId, TenantId};
 
+use crate::models::PageserverUtilization;
 use crate::{
     models::{ShardParameters, TenantConfig},
     shard::{ShardStripeSize, TenantShardId},
@@ -55,6 +56,8 @@ pub struct NodeRegisterRequest {
 
     pub listen_http_addr: String,
     pub listen_http_port: u16,
+
+    pub availability_zone_id: String,
 }
 
 #[derive(Serialize, Deserialize)]
@@ -71,6 +74,17 @@ pub struct TenantPolicyRequest {
     pub scheduling: Option<ShardSchedulingPolicy>,
 }
 
+#[derive(Serialize, Deserialize)]
+pub struct ShardsPreferredAzsRequest {
+    #[serde(flatten)]
+    pub preferred_az_ids: HashMap<TenantShardId, String>,
+}
+
+#[derive(Serialize, Deserialize)]
+pub struct ShardsPreferredAzsResponse {
+    pub updated: Vec<TenantShardId>,
+}
+
 #[derive(Serialize, Deserialize, Debug)]
 pub struct TenantLocateResponseShard {
     pub shard_id: TenantShardId,
@@ -98,6 +112,21 @@ pub struct TenantDescribeResponse {
     pub config: TenantConfig,
 }
 
+#[derive(Serialize, Deserialize, Debug)]
+pub struct NodeShardResponse {
+    pub node_id: NodeId,
+    pub shards: Vec<NodeShard>,
+}
+
+#[derive(Serialize, Deserialize, Debug)]
+pub struct NodeShard {
+    pub tenant_shard_id: TenantShardId,
+    /// Whether the shard is observed secondary on a specific node. True = yes, False = no, None = not on this node.
+    pub is_observed_secondary: Option<bool>,
+    /// Whether the shard is intended to be a secondary on a specific node. True = yes, False = no, None = not on this node.
+    pub is_intended_secondary: Option<bool>,
+}
+
 #[derive(Serialize, Deserialize)]
 pub struct NodeDescribeResponse {
     pub id: NodeId,
@@ -129,8 +158,12 @@ pub struct TenantDescribeResponseShard {
     pub is_splitting: bool,
 
     pub scheduling_policy: ShardSchedulingPolicy,
+
+    pub preferred_az_id: Option<String>,
 }
 
+/// Migration request for a given tenant shard to a given node.
+///
 /// Explicitly migrating a particular shard is a low level operation
 /// TODO: higher level "Reschedule tenant" operation where the request
 /// specifies some constraints, e.g. asking it to get off particular node(s)
@@ -140,23 +173,11 @@ pub struct TenantShardMigrateRequest {
     pub node_id: NodeId,
 }
 
-/// Utilisation score indicating how good a candidate a pageserver
-/// is for scheduling the next tenant. See [`crate::models::PageserverUtilization`].
-/// Lower values are better.
-#[derive(Serialize, Deserialize, Clone, Copy, Eq, PartialEq, PartialOrd, Ord, Debug)]
-pub struct UtilizationScore(pub u64);
-
-impl UtilizationScore {
-    pub fn worst() -> Self {
-        UtilizationScore(u64::MAX)
-    }
-}
-
-#[derive(Serialize, Clone, Copy, Debug)]
+#[derive(Serialize, Clone, Debug)]
 #[serde(into = "NodeAvailabilityWrapper")]
 pub enum NodeAvailability {
     // Normal, happy state
-    Active(UtilizationScore),
+    Active(PageserverUtilization),
     // Node is warming up, but we expect it to become available soon. Covers
     // the time span between the re-attach response being composed on the storage controller
     // and the first successful heartbeat after the processing of the re-attach response
@@ -195,7 +216,9 @@ impl From<NodeAvailabilityWrapper> for NodeAvailability {
         match val {
             // Assume the worst utilisation score to begin with. It will later be updated by
             // the heartbeats.
-            NodeAvailabilityWrapper::Active => NodeAvailability::Active(UtilizationScore::worst()),
+            NodeAvailabilityWrapper::Active => {
+                NodeAvailability::Active(PageserverUtilization::full())
+            }
             NodeAvailabilityWrapper::WarmingUp => NodeAvailability::WarmingUp(Instant::now()),
             NodeAvailabilityWrapper::Offline => NodeAvailability::Offline,
         }

libs/pageserver_api/src/key.rs

@@ -1,8 +1,8 @@
 use anyhow::{bail, Result};
 use byteorder::{ByteOrder, BE};
 use postgres_ffi::relfile_utils::{FSM_FORKNUM, VISIBILITYMAP_FORKNUM};
+use postgres_ffi::Oid;
 use postgres_ffi::RepOriginId;
-use postgres_ffi::{Oid, TransactionId};
 use serde::{Deserialize, Serialize};
 use std::{fmt, ops::Range};
 
@@ -108,14 +108,41 @@ impl Key {
         }
     }
 
+    /// This function checks more extensively what keys we can take on the write path.
+    /// If a key beginning with 00 does not have a global/default tablespace OID, it
+    /// will be rejected on the write path.
+    #[allow(dead_code)]
+    pub fn is_valid_key_on_write_path_strong(&self) -> bool {
+        use postgres_ffi::pg_constants::{DEFAULTTABLESPACE_OID, GLOBALTABLESPACE_OID};
+        if !self.is_i128_representable() {
+            return false;
+        }
+        if self.field1 == 0
+            && !(self.field2 == GLOBALTABLESPACE_OID
+                || self.field2 == DEFAULTTABLESPACE_OID
+                || self.field2 == 0)
+        {
+            return false; // User defined tablespaces are not supported
+        }
+        true
+    }
+
+    /// This is a weaker version of `is_valid_key_on_write_path_strong` that simply
+    /// checks if the key is i128 representable. Note that some keys can be successfully
+    /// ingested into the pageserver, but will cause errors on generating basebackup.
+    pub fn is_valid_key_on_write_path(&self) -> bool {
+        self.is_i128_representable()
+    }
+
+    pub fn is_i128_representable(&self) -> bool {
+        self.field2 <= 0xFFFF || self.field2 == 0xFFFFFFFF || self.field2 == 0x22222222
+    }
+
     /// 'field2' is used to store tablespaceid for relations and small enum numbers for other relish.
     /// As long as Neon does not support tablespace (because of lack of access to local file system),
     /// we can assume that only some predefined namespace OIDs are used which can fit in u16
     pub fn to_i128(&self) -> i128 {
-        assert!(
-            self.field2 <= 0xFFFF || self.field2 == 0xFFFFFFFF || self.field2 == 0x22222222,
-            "invalid key: {self}",
-        );
+        assert!(self.is_i128_representable(), "invalid key: {self}");
         (((self.field1 & 0x7F) as i128) << 120)
             | (((self.field2 & 0xFFFF) as i128) << 104)
             | ((self.field3 as i128) << 72)
@@ -323,7 +350,17 @@ impl Key {
 // 02 00000000 00000000 00000000 00   00000000
 //
 // TwoPhaseFile:
-// 02 00000000 00000000 00000000 00   XID
+//
+// 02 00000000 00000000 00XXXXXX XX   XXXXXXXX
+//
+//                        \______XID_________/
+//
+// The 64-bit XID is stored a little awkwardly in field6, field5 and
+// field4. PostgreSQL v16 and below only stored a 32-bit XID, which
+// fit completely in field6, but starting with PostgreSQL v17, a full
+// 64-bit XID is used. Most pageserver code that accesses
+// TwoPhaseFiles now deals with 64-bit XIDs even on v16, the high bits
+// are just unused.
 //
 // ControlFile:
 // 03 00000000 00000000 00000000 00   00000000
@@ -555,35 +592,36 @@ pub const TWOPHASEDIR_KEY: Key = Key {
 };
 
 #[inline(always)]
-pub fn twophase_file_key(xid: TransactionId) -> Key {
+pub fn twophase_file_key(xid: u64) -> Key {
     Key {
         field1: 0x02,
         field2: 0,
         field3: 0,
-        field4: 0,
-        field5: 0,
-        field6: xid,
+        field4: ((xid & 0xFFFFFF0000000000) >> 40) as u32,
+        field5: ((xid & 0x000000FF00000000) >> 32) as u8,
+        field6: (xid & 0x00000000FFFFFFFF) as u32,
     }
 }
 
 #[inline(always)]
-pub fn twophase_key_range(xid: TransactionId) -> Range<Key> {
+pub fn twophase_key_range(xid: u64) -> Range<Key> {
+    // 64-bit XIDs really should not overflow
     let (next_xid, overflowed) = xid.overflowing_add(1);
 
     Key {
         field1: 0x02,
         field2: 0,
         field3: 0,
-        field4: 0,
-        field5: 0,
-        field6: xid,
+        field4: ((xid & 0xFFFFFF0000000000) >> 40) as u32,
+        field5: ((xid & 0x000000FF00000000) >> 32) as u8,
+        field6: (xid & 0x00000000FFFFFFFF) as u32,
     }..Key {
         field1: 0x02,
         field2: 0,
-        field3: 0,
-        field4: 0,
-        field5: u8::from(overflowed),
-        field6: next_xid,
+        field3: u32::from(overflowed),
+        field4: ((next_xid & 0xFFFFFF0000000000) >> 40) as u32,
+        field5: ((next_xid & 0x000000FF00000000) >> 32) as u8,
+        field6: (next_xid & 0x00000000FFFFFFFF) as u32,
     }
 }
 

libs/pageserver_api/src/models.rs

@@ -6,8 +6,9 @@ pub use utilization::PageserverUtilization;
 
 use std::{
     collections::HashMap,
+    fmt::Display,
     io::{BufRead, Read},
-    num::{NonZeroU64, NonZeroUsize},
+    num::{NonZeroU32, NonZeroU64, NonZeroUsize},
     str::FromStr,
     sync::atomic::AtomicUsize,
     time::{Duration, SystemTime},
@@ -61,7 +62,7 @@ use bytes::{Buf, BufMut, Bytes, BytesMut};
     serde::Serialize,
     serde::Deserialize,
     strum_macros::Display,
-    strum_macros::EnumVariantNames,
+    strum_macros::VariantNames,
     strum_macros::AsRefStr,
     strum_macros::IntoStaticStr,
 )]
@@ -304,8 +305,10 @@ pub struct TenantConfig {
     pub lsn_lease_length_for_ts: Option<String>,
 }
 
-/// The policy for the aux file storage. It can be switched through `switch_aux_file_policy`
-/// tenant config. When the first aux file written, the policy will be persisted in the
+/// The policy for the aux file storage.
+///
+/// It can be switched through `switch_aux_file_policy` tenant config.
+/// When the first aux file written, the policy will be persisted in the
 /// `index_part.json` file and has a limited migration path.
 ///
 /// Currently, we only allow the following migration path:
@@ -435,7 +438,9 @@ pub enum CompactionAlgorithm {
     Tiered,
 }
 
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[derive(
+    Debug, Clone, Copy, PartialEq, Eq, serde_with::DeserializeFromStr, serde_with::SerializeDisplay,
+)]
 pub enum ImageCompressionAlgorithm {
     // Disabled for writes, support decompressing during read path
     Disabled,
@@ -470,11 +475,33 @@ impl FromStr for ImageCompressionAlgorithm {
     }
 }
 
+impl Display for ImageCompressionAlgorithm {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            ImageCompressionAlgorithm::Disabled => write!(f, "disabled"),
+            ImageCompressionAlgorithm::Zstd { level } => {
+                if let Some(level) = level {
+                    write!(f, "zstd({})", level)
+                } else {
+                    write!(f, "zstd")
+                }
+            }
+        }
+    }
+}
+
 #[derive(Eq, PartialEq, Debug, Clone, Serialize, Deserialize)]
 pub struct CompactionAlgorithmSettings {
     pub kind: CompactionAlgorithm,
 }
 
+#[derive(Debug, PartialEq, Eq, Clone, Deserialize, Serialize)]
+#[serde(tag = "mode", rename_all = "kebab-case", deny_unknown_fields)]
+pub enum L0FlushConfig {
+    #[serde(rename_all = "snake_case")]
+    Direct { max_concurrency: NonZeroUsize },
+}
+
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
 pub struct EvictionPolicyLayerAccessThreshold {
     #[serde(with = "humantime_serde")]
@@ -486,12 +513,11 @@ pub struct EvictionPolicyLayerAccessThreshold {
 #[derive(Debug, Serialize, Deserialize, Clone, PartialEq, Eq)]
 pub struct ThrottleConfig {
     pub task_kinds: Vec<String>, // TaskKind
-    pub initial: usize,
+    pub initial: u32,
     #[serde(with = "humantime_serde")]
     pub refill_interval: Duration,
-    pub refill_amount: NonZeroUsize,
-    pub max: usize,
-    pub fair: bool,
+    pub refill_amount: NonZeroU32,
+    pub max: u32,
 }
 
 impl ThrottleConfig {
@@ -501,9 +527,8 @@ impl ThrottleConfig {
             // other values don't matter with emtpy `task_kinds`.
             initial: 0,
             refill_interval: Duration::from_millis(1),
-            refill_amount: NonZeroUsize::new(1).unwrap(),
+            refill_amount: NonZeroU32::new(1).unwrap(),
             max: 1,
-            fair: true,
         }
     }
     /// The requests per second allowed  by the given config.
@@ -721,8 +746,14 @@ pub struct TimelineInfo {
 
     pub walreceiver_status: String,
 
+    // ALWAYS add new fields at the end of the struct with `Option` to ensure forward/backward compatibility.
+    // Backward compatibility: you will get a JSON not containing the newly-added field.
+    // Forward compatibility: a previous version of the pageserver will receive a JSON. serde::Deserialize does
+    // not deny unknown fields by default so it's safe to set the field to some value, though it won't be
+    // read.
     /// The last aux file policy being used on this timeline
     pub last_aux_file_policy: Option<AuxFilePolicy>,
+    pub is_archived: Option<bool>,
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -867,7 +898,9 @@ pub struct WalRedoManagerStatus {
     pub process: Option<WalRedoManagerProcessStatus>,
 }
 
-/// The progress of a secondary tenant is mostly useful when doing a long running download: e.g. initiating
+/// The progress of a secondary tenant.
+///
+/// It is mostly useful when doing a long running download: e.g. initiating
 /// a download job, timing out while waiting for it to run, and then inspecting this status to understand
 /// what's happening.
 #[derive(Default, Debug, Serialize, Deserialize, Clone)]
@@ -1062,7 +1095,7 @@ impl TryFrom<u8> for PagestreamBeMessageTag {
     }
 }
 
-// In the V2 protocol version, a GetPage request contains two LSN values:
+// A GetPage request contains two LSN values:
 //
 // request_lsn: Get the page version at this point in time.  Lsn::Max is a special value that means
 // "get the latest version present". It's used by the primary server, which knows that no one else
@@ -1075,7 +1108,7 @@ impl TryFrom<u8> for PagestreamBeMessageTag {
 // passing an earlier LSN can speed up the request, by allowing the pageserver to process the
 // request without waiting for 'request_lsn' to arrive.
 //
-// The legacy V1 interface contained only one LSN, and a boolean 'latest' flag. The V1 interface was
+// The now-defunct V1 interface contained only one LSN, and a boolean 'latest' flag. The V1 interface was
 // sufficient for the primary; the 'lsn' was equivalent to the 'not_modified_since' value, and
 // 'latest' was set to true. The V2 interface was added because there was no correct way for a
 // standby to request a page at a particular non-latest LSN, and also include the
@@ -1083,15 +1116,11 @@ impl TryFrom<u8> for PagestreamBeMessageTag {
 // request, if the standby knows that the page hasn't been modified since, and risk getting an error
 // if that LSN has fallen behind the GC horizon, or requesting the current replay LSN, which could
 // require the pageserver unnecessarily to wait for the WAL to arrive up to that point. The new V2
-// interface allows sending both LSNs, and let the pageserver do the right thing. There is no
+// interface allows sending both LSNs, and let the pageserver do the right thing. There was no
 // difference in the responses between V1 and V2.
 //
-// The Request structs below reflect the V2 interface. If V1 is used, the parse function
-// maps the old format requests to the new format.
-//
 #[derive(Clone, Copy)]
 pub enum PagestreamProtocolVersion {
-    V1,
     V2,
 }
 
@@ -1230,36 +1259,17 @@ impl PagestreamFeMessage {
         bytes.into()
     }
 
-    pub fn parse<R: std::io::Read>(
-        body: &mut R,
-        protocol_version: PagestreamProtocolVersion,
-    ) -> anyhow::Result<PagestreamFeMessage> {
+    pub fn parse<R: std::io::Read>(body: &mut R) -> anyhow::Result<PagestreamFeMessage> {
         // these correspond to the NeonMessageTag enum in pagestore_client.h
         //
         // TODO: consider using protobuf or serde bincode for less error prone
         // serialization.
         let msg_tag = body.read_u8()?;
 
-        let (request_lsn, not_modified_since) = match protocol_version {
-            PagestreamProtocolVersion::V2 => (
-                Lsn::from(body.read_u64::<BigEndian>()?),
-                Lsn::from(body.read_u64::<BigEndian>()?),
-            ),
-            PagestreamProtocolVersion::V1 => {
-                // In the old protocol, each message starts with a boolean 'latest' flag,
-                // followed by 'lsn'. Convert that to the two LSNs, 'request_lsn' and
-                // 'not_modified_since', used in the new protocol version.
-                let latest = body.read_u8()? != 0;
-                let request_lsn = Lsn::from(body.read_u64::<BigEndian>()?);
-                if latest {
-                    (Lsn::MAX, request_lsn) // get latest version
-                } else {
-                    (request_lsn, request_lsn) // get version at specified LSN
-                }
-            }
-        };
+        // these two fields are the same for every request type
+        let request_lsn = Lsn::from(body.read_u64::<BigEndian>()?);
+        let not_modified_since = Lsn::from(body.read_u64::<BigEndian>()?);
 
-        // The rest of the messages are the same between V1 and V2
         match msg_tag {
             0 => Ok(PagestreamFeMessage::Exists(PagestreamExistsRequest {
                 request_lsn,
@@ -1467,9 +1477,7 @@ mod tests {
         ];
         for msg in messages {
             let bytes = msg.serialize();
-            let reconstructed =
-                PagestreamFeMessage::parse(&mut bytes.reader(), PagestreamProtocolVersion::V2)
-                    .unwrap();
+            let reconstructed = PagestreamFeMessage::parse(&mut bytes.reader()).unwrap();
             assert!(msg == reconstructed);
         }
     }
@@ -1677,21 +1685,33 @@ mod tests {
     #[test]
     fn test_image_compression_algorithm_parsing() {
         use ImageCompressionAlgorithm::*;
-        assert_eq!(
-            ImageCompressionAlgorithm::from_str("disabled").unwrap(),
-            Disabled
-        );
-        assert_eq!(
-            ImageCompressionAlgorithm::from_str("zstd").unwrap(),
-            Zstd { level: None }
-        );
-        assert_eq!(
-            ImageCompressionAlgorithm::from_str("zstd(18)").unwrap(),
-            Zstd { level: Some(18) }
-        );
-        assert_eq!(
-            ImageCompressionAlgorithm::from_str("zstd(-3)").unwrap(),
-            Zstd { level: Some(-3) }
-        );
+        let cases = [
+            ("disabled", Disabled),
+            ("zstd", Zstd { level: None }),
+            ("zstd(18)", Zstd { level: Some(18) }),
+            ("zstd(-3)", Zstd { level: Some(-3) }),
+        ];
+
+        for (display, expected) in cases {
+            assert_eq!(
+                ImageCompressionAlgorithm::from_str(display).unwrap(),
+                expected,
+                "parsing works"
+            );
+            assert_eq!(format!("{expected}"), display, "Display FromStr roundtrip");
+
+            let ser = serde_json::to_string(&expected).expect("serialization");
+            assert_eq!(
+                serde_json::from_str::<ImageCompressionAlgorithm>(&ser).unwrap(),
+                expected,
+                "serde roundtrip"
+            );
+
+            assert_eq!(
+                serde_json::Value::String(display.to_string()),
+                serde_json::to_value(expected).unwrap(),
+                "Display is the serde serialization"
+            );
+        }
     }
 }

libs/pageserver_api/src/models/utilization.rs

@@ -38,7 +38,7 @@ pub struct PageserverUtilization {
     pub max_shard_count: u32,
 
     /// Cached result of [`Self::score`]
-    pub utilization_score: u64,
+    pub utilization_score: Option<u64>,
 
     /// When was this snapshot captured, pageserver local time.
     ///
@@ -50,6 +50,8 @@ fn unity_percent() -> Percent {
     Percent::new(0).unwrap()
 }
 
+pub type RawScore = u64;
+
 impl PageserverUtilization {
     const UTILIZATION_FULL: u64 = 1000000;
 
@@ -62,7 +64,7 @@ impl PageserverUtilization {
     /// - Negative values are forbidden
     /// - Values over UTILIZATION_FULL indicate an overloaded node, which may show degraded performance due to
     ///   layer eviction.
-    pub fn score(&self) -> u64 {
+    pub fn score(&self) -> RawScore {
         let disk_usable_capacity = ((self.disk_usage_bytes + self.free_space_bytes)
             * self.disk_usable_pct.get() as u64)
             / 100;
@@ -74,8 +76,41 @@ impl PageserverUtilization {
         std::cmp::max(disk_utilization_score, shard_utilization_score)
     }
 
-    pub fn refresh_score(&mut self) {
-        self.utilization_score = self.score();
+    pub fn cached_score(&mut self) -> RawScore {
+        match self.utilization_score {
+            None => {
+                let s = self.score();
+                self.utilization_score = Some(s);
+                s
+            }
+            Some(s) => s,
+        }
+    }
+
+    /// If a node is currently hosting more work than it can comfortably handle.  This does not indicate that
+    /// it will fail, but it is a strong signal that more work should not be added unless there is no alternative.
+    ///
+    /// When a node is overloaded, we may override soft affinity preferences and do things like scheduling
+    /// into a node in a less desirable AZ, if all the nodes in the preferred AZ are overloaded.
+    pub fn is_overloaded(score: RawScore) -> bool {
+        // Why the factor of two?  This is unscientific but reflects behavior of real systems:
+        // - In terms of shard counts, a node's preferred max count is a soft limit intended to keep
+        //   startup and housekeeping jobs nice and responsive.  We can go to double this limit if needed
+        //   until some more nodes are deployed.
+        // - In terms of disk space, the node's utilization heuristic assumes every tenant needs to
+        //   hold its biggest timeline fully on disk, which is tends to be an over estimate when
+        //   some tenants are very idle and have dropped layers from disk.  In practice going up to
+        //   double is generally better than giving up and scheduling in a sub-optimal AZ.
+        score >= 2 * Self::UTILIZATION_FULL
+    }
+
+    pub fn adjust_shard_count_max(&mut self, shard_count: u32) {
+        if self.shard_count < shard_count {
+            self.shard_count = shard_count;
+
+            // Dirty cache: this will be calculated next time someone retrives the score
+            self.utilization_score = None;
+        }
     }
 
     /// A utilization structure that has a full utilization score: use this as a placeholder when
@@ -88,7 +123,38 @@ impl PageserverUtilization {
             disk_usable_pct: Percent::new(100).unwrap(),
             shard_count: 1,
             max_shard_count: 1,
-            utilization_score: Self::UTILIZATION_FULL,
+            utilization_score: Some(Self::UTILIZATION_FULL),
+            captured_at: serde_system_time::SystemTime(SystemTime::now()),
+        }
+    }
+}
+
+/// Test helper
+pub mod test_utilization {
+    use super::PageserverUtilization;
+    use std::time::SystemTime;
+    use utils::{
+        serde_percent::Percent,
+        serde_system_time::{self},
+    };
+
+    // Parameters of the imaginary node used for test utilization instances
+    const TEST_DISK_SIZE: u64 = 1024 * 1024 * 1024 * 1024;
+    const TEST_SHARDS_MAX: u32 = 1000;
+
+    /// Unit test helper.  Unconditionally compiled because cfg(test) doesn't carry across crates.  Do
+    /// not abuse this function from non-test code.
+    ///
+    /// Emulates a node with a 1000 shard limit and a 1TB disk.
+    pub fn simple(shard_count: u32, disk_wanted_bytes: u64) -> PageserverUtilization {
+        PageserverUtilization {
+            disk_usage_bytes: disk_wanted_bytes,
+            free_space_bytes: TEST_DISK_SIZE - std::cmp::min(disk_wanted_bytes, TEST_DISK_SIZE),
+            disk_wanted_bytes,
+            disk_usable_pct: Percent::new(100).unwrap(),
+            shard_count,
+            max_shard_count: TEST_SHARDS_MAX,
+            utilization_score: None,
             captured_at: serde_system_time::SystemTime(SystemTime::now()),
         }
     }
@@ -120,7 +186,7 @@ mod tests {
             disk_usage_bytes: u64::MAX,
             free_space_bytes: 0,
             disk_wanted_bytes: u64::MAX,
-            utilization_score: 13,
+            utilization_score: Some(13),
             disk_usable_pct: Percent::new(90).unwrap(),
             shard_count: 100,
             max_shard_count: 200,

libs/postgres_backend/Cargo.toml

@@ -5,10 +5,8 @@ edition.workspace = true
 license.workspace = true
 
 [dependencies]
-async-trait.workspace = true
 anyhow.workspace = true
 bytes.workspace = true
-futures.workspace = true
 rustls.workspace = true
 serde.workspace = true
 thiserror.workspace = true

libs/postgres_backend/src/lib.rs

@@ -69,8 +69,10 @@ impl QueryError {
 }
 
 /// Returns true if the given error is a normal consequence of a network issue,
-/// or the client closing the connection. These errors can happen during normal
-/// operations, and don't indicate a bug in our code.
+/// or the client closing the connection.
+///
+/// These errors can happen during normal operations,
+/// and don't indicate a bug in our code.
 pub fn is_expected_io_error(e: &io::Error) -> bool {
     use io::ErrorKind::*;
     matches!(
@@ -79,17 +81,16 @@ pub fn is_expected_io_error(e: &io::Error) -> bool {
     )
 }
 
-#[async_trait::async_trait]
 pub trait Handler<IO> {
     /// Handle single query.
     /// postgres_backend will issue ReadyForQuery after calling this (this
     /// might be not what we want after CopyData streaming, but currently we don't
     /// care). It will also flush out the output buffer.
-    async fn process_query(
+    fn process_query(
         &mut self,
         pgb: &mut PostgresBackend<IO>,
         query_string: &str,
-    ) -> Result<(), QueryError>;
+    ) -> impl Future<Output = Result<(), QueryError>>;
 
     /// Called on startup packet receival, allows to process params.
     ///
@@ -279,16 +280,6 @@ pub struct PostgresBackend<IO> {
 
 pub type PostgresBackendTCP = PostgresBackend<tokio::net::TcpStream>;
 
-pub fn query_from_cstring(query_string: Bytes) -> Vec<u8> {
-    let mut query_string = query_string.to_vec();
-    if let Some(ch) = query_string.last() {
-        if *ch == 0 {
-            query_string.pop();
-        }
-    }
-    query_string
-}
-
 /// Cast a byte slice to a string slice, dropping null terminator if there's one.
 fn cstr_to_str(bytes: &[u8]) -> anyhow::Result<&str> {
     let without_null = bytes.strip_suffix(&[0]).unwrap_or(bytes);

libs/postgres_backend/tests/simple_select.rs

@@ -23,7 +23,6 @@ async fn make_tcp_pair() -> (TcpStream, TcpStream) {
 
 struct TestHandler {}
 
-#[async_trait::async_trait]
 impl<IO: AsyncRead + AsyncWrite + Unpin + Send> Handler<IO> for TestHandler {
     // return single col 'hey' for any query
     async fn process_query(

libs/postgres_connection/src/lib.rs

@@ -7,6 +7,7 @@ use std::fmt;
 use url::Host;
 
 /// Parses a string of format either `host:port` or `host` into a corresponding pair.
+///
 /// The `host` part should be a correct `url::Host`, while `port` (if present) should be
 /// a valid decimal u16 of digits only.
 pub fn parse_host_port<S: AsRef<str>>(host_port: S) -> Result<(Host, Option<u16>), anyhow::Error> {

libs/postgres_ffi/Cargo.toml

@@ -5,13 +5,10 @@ edition.workspace = true
 license.workspace = true
 
 [dependencies]
-rand.workspace = true
 regex.workspace = true
 bytes.workspace = true
-byteorder.workspace = true
 anyhow.workspace = true
 crc32c.workspace = true
-hex.workspace = true
 once_cell.workspace = true
 log.workspace = true
 memoffset.workspace = true

libs/postgres_ffi/build.rs

@@ -14,7 +14,7 @@ impl ParseCallbacks for PostgresFfiCallbacks {
     fn include_file(&self, filename: &str) {
         // This does the equivalent of passing bindgen::CargoCallbacks
         // to the builder .parse_callbacks() method.
-        let cargo_callbacks = bindgen::CargoCallbacks;
+        let cargo_callbacks = bindgen::CargoCallbacks::new();
         cargo_callbacks.include_file(filename)
     }
 
@@ -56,7 +56,7 @@ fn main() -> anyhow::Result<()> {
         PathBuf::from("pg_install")
     };
 
-    for pg_version in &["v14", "v15", "v16"] {
+    for pg_version in &["v14", "v15", "v16", "v17"] {
         let mut pg_install_dir_versioned = pg_install_dir.join(pg_version);
         if pg_install_dir_versioned.is_relative() {
             let cwd = env::current_dir().context("Failed to get current_dir")?;
@@ -121,6 +121,7 @@ fn main() -> anyhow::Result<()> {
             .allowlist_type("XLogPageHeaderData")
             .allowlist_type("XLogLongPageHeaderData")
             .allowlist_var("XLOG_PAGE_MAGIC")
+            .allowlist_var("PG_MAJORVERSION_NUM")
             .allowlist_var("PG_CONTROL_FILE_SIZE")
             .allowlist_var("PG_CONTROLFILEDATA_OFFSETOF_CRC")
             .allowlist_type("PageHeaderData")

libs/postgres_ffi/src/lib.rs

@@ -44,6 +44,9 @@ macro_rules! postgres_ffi {
             // Re-export some symbols from bindings
             pub use bindings::DBState_DB_SHUTDOWNED;
             pub use bindings::{CheckPoint, ControlFileData, XLogRecord};
+
+            pub const ZERO_CHECKPOINT: bytes::Bytes =
+                bytes::Bytes::from_static(&[0u8; xlog_utils::SIZEOF_CHECKPOINT]);
         }
     };
 }
@@ -54,6 +57,7 @@ macro_rules! for_all_postgres_versions {
         $macro!(v14);
         $macro!(v15);
         $macro!(v16);
+        $macro!(v17);
     };
 }
 
@@ -88,6 +92,7 @@ macro_rules! dispatch_pgversion {
                 14 : v14,
                 15 : v15,
                 16 : v16,
+                17 : v17,
             ]
         )
     };
@@ -106,6 +111,110 @@ macro_rules! dispatch_pgversion {
     };
 }
 
+#[macro_export]
+macro_rules! enum_pgversion_dispatch {
+    ($name:expr, $typ:ident, $bind:ident, $code:block) => {
+        enum_pgversion_dispatch!(
+            name = $name,
+            bind = $bind,
+            typ = $typ,
+            code = $code,
+            pgversions = [
+                V14 : v14,
+                V15 : v15,
+                V16 : v16,
+                V17 : v17,
+            ]
+        )
+    };
+    (name = $name:expr,
+     bind = $bind:ident,
+     typ = $typ:ident,
+     code = $code:block,
+     pgversions = [$($variant:ident : $md:ident),+ $(,)?]) => {
+        match $name {
+            $(
+            self::$typ::$variant($bind) => {
+                use $crate::$md as pgv;
+                $code
+            }
+            ),+,
+        }
+    };
+}
+
+#[macro_export]
+macro_rules! enum_pgversion {
+    {$name:ident, pgv :: $t:ident} => {
+        enum_pgversion!{
+            name = $name,
+            typ = $t,
+            pgversions = [
+                V14 : v14,
+                V15 : v15,
+                V16 : v16,
+                V17 : v17,
+            ]
+        }
+    };
+    {$name:ident, pgv :: $p:ident :: $t:ident} => {
+        enum_pgversion!{
+            name = $name,
+            path = $p,
+            typ = $t,
+            pgversions = [
+                V14 : v14,
+                V15 : v15,
+                V16 : v16,
+                V17 : v17,
+            ]
+        }
+    };
+    {name = $name:ident,
+     typ = $t:ident,
+     pgversions = [$($variant:ident : $md:ident),+ $(,)?]} => {
+        pub enum $name {
+            $($variant ( $crate::$md::$t )),+
+        }
+        impl self::$name {
+            pub fn pg_version(&self) -> u32 {
+                enum_pgversion_dispatch!(self, $name, _ign, {
+                    pgv::bindings::PG_MAJORVERSION_NUM
+                })
+            }
+        }
+        $(
+        impl Into<self::$name> for $crate::$md::$t {
+            fn into(self) -> self::$name {
+                self::$name::$variant (self)
+            }
+        }
+        )+
+    };
+    {name = $name:ident,
+     path = $p:ident,
+     typ = $t:ident,
+     pgversions = [$($variant:ident : $md:ident),+ $(,)?]} => {
+        pub enum $name {
+            $($variant ($crate::$md::$p::$t)),+
+        }
+        impl $name {
+            pub fn pg_version(&self) -> u32 {
+                enum_pgversion_dispatch!(self, $name, _ign, {
+                    pgv::bindings::PG_MAJORVERSION_NUM
+                })
+            }
+        }
+        $(
+        impl Into<$name> for $crate::$md::$p::$t {
+            fn into(self) -> $name {
+                $name::$variant (self)
+            }
+        }
+        )+
+    };
+}
+
 pub mod pg_constants;
 pub mod relfile_utils;
 
@@ -136,9 +245,9 @@ pub const MAX_SEND_SIZE: usize = XLOG_BLCKSZ * 16;
 
 // Export some version independent functions that are used outside of this mod
 pub use v14::xlog_utils::encode_logical_message;
-pub use v14::xlog_utils::from_pg_timestamp;
 pub use v14::xlog_utils::get_current_timestamp;
 pub use v14::xlog_utils::to_pg_timestamp;
+pub use v14::xlog_utils::try_from_pg_timestamp;
 pub use v14::xlog_utils::XLogFileName;
 
 pub use v14::bindings::DBState_DB_SHUTDOWNED;

libs/postgres_ffi/src/pg_constants.rs

@@ -9,8 +9,8 @@
 //! comments on them.
 //!
 
+use crate::PageHeaderData;
 use crate::BLCKSZ;
-use crate::{PageHeaderData, XLogRecord};
 
 //
 // From pg_tablespace_d.h
@@ -152,6 +152,9 @@ pub const XLH_UPDATE_OLD_ALL_VISIBLE_CLEARED: u8 = (1 << 0) as u8;
 pub const XLH_UPDATE_NEW_ALL_VISIBLE_CLEARED: u8 = (1 << 1) as u8;
 pub const XLH_DELETE_ALL_VISIBLE_CLEARED: u8 = (1 << 0) as u8;
 
+// From heapam_xlog.h
+pub const XLOG_HEAP2_REWRITE: u8 = 0x00;
+
 // From replication/message.h
 pub const XLOG_LOGICAL_MESSAGE: u8 = 0x00;
 
@@ -191,8 +194,6 @@ pub const XLR_RMGR_INFO_MASK: u8 = 0xF0;
 pub const XLOG_TBLSPC_CREATE: u8 = 0x00;
 pub const XLOG_TBLSPC_DROP: u8 = 0x10;
 
-pub const SIZEOF_XLOGRECORD: u32 = size_of::<XLogRecord>() as u32;
-
 //
 // from xlogrecord.h
 //
@@ -216,18 +217,21 @@ pub const BKPIMAGE_HAS_HOLE: u8 = 0x01; /* page image has "hole" */
 /* From transam.h */
 pub const FIRST_NORMAL_TRANSACTION_ID: u32 = 3;
 pub const INVALID_TRANSACTION_ID: u32 = 0;
-pub const FIRST_BOOTSTRAP_OBJECT_ID: u32 = 12000;
-pub const FIRST_NORMAL_OBJECT_ID: u32 = 16384;
 
+/* pg_control.h */
 pub const XLOG_CHECKPOINT_SHUTDOWN: u8 = 0x00;
 pub const XLOG_CHECKPOINT_ONLINE: u8 = 0x10;
-pub const XLP_FIRST_IS_CONTRECORD: u16 = 0x0001;
-pub const XLP_LONG_HEADER: u16 = 0x0002;
+pub const XLOG_PARAMETER_CHANGE: u8 = 0x60;
+pub const XLOG_END_OF_RECOVERY: u8 = 0x90;
 
 /* From xlog.h */
 pub const XLOG_REPLORIGIN_SET: u8 = 0x00;
 pub const XLOG_REPLORIGIN_DROP: u8 = 0x10;
 
+/* xlog_internal.h */
+pub const XLP_FIRST_IS_CONTRECORD: u16 = 0x0001;
+pub const XLP_LONG_HEADER: u16 = 0x0002;
+
 /* From replication/slot.h */
 pub const REPL_SLOT_ON_DISK_OFFSETOF_RESTART_LSN: usize = 4*4  /* offset of `slotdata` in ReplicationSlotOnDisk  */
    + 64 /* NameData */  + 4*4;
@@ -245,33 +249,6 @@ pub const VM_HEAPBLOCKS_PER_PAGE: u32 =
 /* From origin.c */
 pub const REPLICATION_STATE_MAGIC: u32 = 0x1257DADE;
 
-// List of subdirectories inside pgdata.
-// Copied from src/bin/initdb/initdb.c
-pub const PGDATA_SUBDIRS: [&str; 22] = [
-    "global",
-    "pg_wal/archive_status",
-    "pg_commit_ts",
-    "pg_dynshmem",
-    "pg_notify",
-    "pg_serial",
-    "pg_snapshots",
-    "pg_subtrans",
-    "pg_twophase",
-    "pg_multixact",
-    "pg_multixact/members",
-    "pg_multixact/offsets",
-    "base",
-    "base/1",
-    "pg_replslot",
-    "pg_tblspc",
-    "pg_stat",
-    "pg_stat_tmp",
-    "pg_xact",
-    "pg_logical",
-    "pg_logical/snapshots",
-    "pg_logical/mappings",
-];
-
 // Don't include postgresql.conf as it is inconvenient on node start:
 // we need postgresql.conf before basebackup to synchronize safekeepers
 // so no point in overwriting it during backup restore. Rest of the files

libs/postgres_ffi/src/pg_constants_v14.rs

@@ -5,6 +5,33 @@ pub const BKPIMAGE_IS_COMPRESSED: u8 = 0x02; /* page image is compressed */
 pub const BKPIMAGE_APPLY: u8 = 0x04; /* page image should be restored during replay */
 pub const SIZEOF_RELMAPFILE: usize = 512; /* sizeof(RelMapFile) in relmapper.c */
 
+// List of subdirectories inside pgdata.
+// Copied from src/bin/initdb/initdb.c
+pub const PGDATA_SUBDIRS: [&str; 22] = [
+    "global",
+    "pg_wal/archive_status",
+    "pg_commit_ts",
+    "pg_dynshmem",
+    "pg_notify",
+    "pg_serial",
+    "pg_snapshots",
+    "pg_subtrans",
+    "pg_twophase",
+    "pg_multixact",
+    "pg_multixact/members",
+    "pg_multixact/offsets",
+    "base",
+    "base/1",
+    "pg_replslot",
+    "pg_tblspc",
+    "pg_stat",
+    "pg_stat_tmp",
+    "pg_xact",
+    "pg_logical",
+    "pg_logical/snapshots",
+    "pg_logical/mappings",
+];
+
 pub fn bkpimg_is_compressed(bimg_info: u8) -> bool {
     (bimg_info & BKPIMAGE_IS_COMPRESSED) != 0
 }

libs/postgres_ffi/src/pg_constants_v15.rs

@@ -11,6 +11,8 @@ pub const BKPIMAGE_COMPRESS_ZSTD: u8 = 0x10; /* page image is compressed */
 
 pub const SIZEOF_RELMAPFILE: usize = 512; /* sizeof(RelMapFile) in relmapper.c */
 
+pub use super::super::v14::bindings::PGDATA_SUBDIRS;
+
 pub fn bkpimg_is_compressed(bimg_info: u8) -> bool {
     const ANY_COMPRESS_FLAG: u8 = BKPIMAGE_COMPRESS_PGLZ | BKPIMAGE_COMPRESS_LZ4 | BKPIMAGE_COMPRESS_ZSTD;
 

libs/postgres_ffi/src/pg_constants_v16.rs

@@ -11,6 +11,8 @@ pub const BKPIMAGE_COMPRESS_ZSTD: u8 = 0x10; /* page image is compressed */
 
 pub const SIZEOF_RELMAPFILE: usize = 524; /* sizeof(RelMapFile) in relmapper.c */
 
+pub use super::super::v14::bindings::PGDATA_SUBDIRS;
+
 pub fn bkpimg_is_compressed(bimg_info: u8) -> bool {
     const ANY_COMPRESS_FLAG: u8 = BKPIMAGE_COMPRESS_PGLZ | BKPIMAGE_COMPRESS_LZ4 | BKPIMAGE_COMPRESS_ZSTD;
 

libs/postgres_ffi/src/pg_constants_v17.rs

@@ -0,0 +1,55 @@
+pub const XACT_XINFO_HAS_DROPPED_STATS: u32 = 1u32 << 8;
+
+pub const XLOG_DBASE_CREATE_FILE_COPY: u8 = 0x00;
+pub const XLOG_DBASE_CREATE_WAL_LOG: u8 = 0x10;
+pub const XLOG_DBASE_DROP: u8 = 0x20;
+
+pub const BKPIMAGE_APPLY: u8 = 0x02; /* page image should be restored during replay */
+pub const BKPIMAGE_COMPRESS_PGLZ: u8 = 0x04; /* page image is compressed */
+pub const BKPIMAGE_COMPRESS_LZ4: u8 = 0x08; /* page image is compressed */
+pub const BKPIMAGE_COMPRESS_ZSTD: u8 = 0x10; /* page image is compressed */
+
+pub const SIZEOF_RELMAPFILE: usize = 524; /* sizeof(RelMapFile) in relmapper.c */
+
+// List of subdirectories inside pgdata.
+// Copied from src/bin/initdb/initdb.c
+pub const PGDATA_SUBDIRS: [&str; 23] = [
+    "global",
+    "pg_wal/archive_status",
+    "pg_wal/summaries",
+    "pg_commit_ts",
+    "pg_dynshmem",
+    "pg_notify",
+    "pg_serial",
+    "pg_snapshots",
+    "pg_subtrans",
+    "pg_twophase",
+    "pg_multixact",
+    "pg_multixact/members",
+    "pg_multixact/offsets",
+    "base",
+    "base/1",
+    "pg_replslot",
+    "pg_tblspc",
+    "pg_stat",
+    "pg_stat_tmp",
+    "pg_xact",
+    "pg_logical",
+    "pg_logical/snapshots",
+    "pg_logical/mappings",
+];
+
+pub fn bkpimg_is_compressed(bimg_info: u8) -> bool {
+    const ANY_COMPRESS_FLAG: u8 = BKPIMAGE_COMPRESS_PGLZ | BKPIMAGE_COMPRESS_LZ4 | BKPIMAGE_COMPRESS_ZSTD;
+
+    (bimg_info & ANY_COMPRESS_FLAG) != 0
+}
+
+
+pub const XLOG_HEAP2_PRUNE_ON_ACCESS: u8 = 0x10;
+pub const XLOG_HEAP2_PRUNE_VACUUM_SCAN: u8 = 0x20;
+pub const XLOG_HEAP2_PRUNE_VACUUM_CLEANUP: u8 = 0x30;
+
+
+pub const XLOG_OVERWRITE_CONTRECORD: u8 = 0xD0;
+pub const XLOG_CHECKPOINT_REDO: u8 = 0xE0;

libs/postgres_ffi/src/xlog_utils.rs

@@ -26,11 +26,12 @@ use bytes::{Buf, Bytes};
 use log::*;
 
 use serde::Serialize;
+use std::ffi::OsStr;
 use std::fs::File;
 use std::io::prelude::*;
 use std::io::ErrorKind;
 use std::io::SeekFrom;
-use std::path::{Path, PathBuf};
+use std::path::Path;
 use std::time::SystemTime;
 use utils::bin_ser::DeserializeError;
 use utils::bin_ser::SerializeError;
@@ -78,19 +79,34 @@ pub fn XLogFileName(tli: TimeLineID, logSegNo: XLogSegNo, wal_segsz_bytes: usize
     )
 }
 
-pub fn XLogFromFileName(fname: &str, wal_seg_size: usize) -> (XLogSegNo, TimeLineID) {
-    let tli = u32::from_str_radix(&fname[0..8], 16).unwrap();
-    let log = u32::from_str_radix(&fname[8..16], 16).unwrap() as XLogSegNo;
-    let seg = u32::from_str_radix(&fname[16..24], 16).unwrap() as XLogSegNo;
-    (log * XLogSegmentsPerXLogId(wal_seg_size) + seg, tli)
+pub fn XLogFromFileName(
+    fname: &OsStr,
+    wal_seg_size: usize,
+) -> anyhow::Result<(XLogSegNo, TimeLineID)> {
+    if let Some(fname_str) = fname.to_str() {
+        let tli = u32::from_str_radix(&fname_str[0..8], 16)?;
+        let log = u32::from_str_radix(&fname_str[8..16], 16)? as XLogSegNo;
+        let seg = u32::from_str_radix(&fname_str[16..24], 16)? as XLogSegNo;
+        Ok((log * XLogSegmentsPerXLogId(wal_seg_size) + seg, tli))
+    } else {
+        anyhow::bail!("non-ut8 filename: {:?}", fname);
+    }
 }
 
-pub fn IsXLogFileName(fname: &str) -> bool {
-    return fname.len() == XLOG_FNAME_LEN && fname.chars().all(|c| c.is_ascii_hexdigit());
+pub fn IsXLogFileName(fname: &OsStr) -> bool {
+    if let Some(fname) = fname.to_str() {
+        fname.len() == XLOG_FNAME_LEN && fname.chars().all(|c| c.is_ascii_hexdigit())
+    } else {
+        false
+    }
 }
 
-pub fn IsPartialXLogFileName(fname: &str) -> bool {
-    fname.ends_with(".partial") && IsXLogFileName(&fname[0..fname.len() - 8])
+pub fn IsPartialXLogFileName(fname: &OsStr) -> bool {
+    if let Some(fname) = fname.to_str() {
+        fname.ends_with(".partial") && IsXLogFileName(OsStr::new(&fname[0..fname.len() - 8]))
+    } else {
+        false
+    }
 }
 
 /// If LSN points to the beginning of the page, then shift it to first record,
@@ -135,6 +151,8 @@ pub fn get_current_timestamp() -> TimestampTz {
 mod timestamp_conversions {
     use std::time::Duration;
 
+    use anyhow::Context;
+
     use super::*;
 
     const UNIX_EPOCH_JDATE: u64 = 2440588; // == date2j(1970, 1, 1)
@@ -154,18 +172,18 @@ mod timestamp_conversions {
         }
     }
 
-    pub fn from_pg_timestamp(time: TimestampTz) -> SystemTime {
+    pub fn try_from_pg_timestamp(time: TimestampTz) -> anyhow::Result<SystemTime> {
         let time: u64 = time
             .try_into()
-            .expect("timestamp before millenium (postgres epoch)");
+            .context("timestamp before millenium (postgres epoch)")?;
         let since_unix_epoch = time + SECS_DIFF_UNIX_TO_POSTGRES_EPOCH * USECS_PER_SEC;
         SystemTime::UNIX_EPOCH
             .checked_add(Duration::from_micros(since_unix_epoch))
-            .expect("SystemTime overflow")
+            .context("SystemTime overflow")
     }
 }
 
-pub use timestamp_conversions::{from_pg_timestamp, to_pg_timestamp};
+pub use timestamp_conversions::{to_pg_timestamp, try_from_pg_timestamp};
 
 // Returns (aligned) end_lsn of the last record in data_dir with WAL segments.
 // start_lsn must point to some previously known record boundary (beginning of
@@ -258,13 +276,6 @@ fn open_wal_segment(seg_file_path: &Path) -> anyhow::Result<Option<File>> {
     }
 }
 
-pub fn main() {
-    let mut data_dir = PathBuf::new();
-    data_dir.push(".");
-    let wal_end = find_end_of_wal(&data_dir, WAL_SEGMENT_SIZE, Lsn(0)).unwrap();
-    println!("wal_end={:?}", wal_end);
-}
-
 impl XLogRecord {
     pub fn from_slice(buf: &[u8]) -> Result<XLogRecord, DeserializeError> {
         use utils::bin_ser::LeSer;
@@ -545,14 +556,14 @@ mod tests {
     #[test]
     fn test_ts_conversion() {
         let now = SystemTime::now();
-        let round_trip = from_pg_timestamp(to_pg_timestamp(now));
+        let round_trip = try_from_pg_timestamp(to_pg_timestamp(now)).unwrap();
 
         let now_since = now.duration_since(SystemTime::UNIX_EPOCH).unwrap();
         let round_trip_since = round_trip.duration_since(SystemTime::UNIX_EPOCH).unwrap();
         assert_eq!(now_since.as_micros(), round_trip_since.as_micros());
 
         let now_pg = get_current_timestamp();
-        let round_trip_pg = to_pg_timestamp(from_pg_timestamp(now_pg));
+        let round_trip_pg = to_pg_timestamp(try_from_pg_timestamp(now_pg).unwrap());
 
         assert_eq!(now_pg, round_trip_pg);
     }

libs/postgres_ffi/wal_craft/Cargo.toml

@@ -9,7 +9,6 @@ anyhow.workspace = true
 clap.workspace = true
 env_logger.workspace = true
 log.workspace = true
-once_cell.workspace = true
 postgres.workspace = true
 postgres_ffi.workspace = true
 camino-tempfile.workspace = true

libs/postgres_ffi/wal_craft/src/lib.rs

@@ -7,6 +7,7 @@ use postgres_ffi::{WAL_SEGMENT_SIZE, XLOG_BLCKSZ};
 use postgres_ffi::{
     XLOG_SIZE_OF_XLOG_LONG_PHD, XLOG_SIZE_OF_XLOG_RECORD, XLOG_SIZE_OF_XLOG_SHORT_PHD,
 };
+use std::ffi::OsStr;
 use std::path::{Path, PathBuf};
 use std::process::Command;
 use std::time::{Duration, Instant};
@@ -26,7 +27,6 @@ macro_rules! xlog_utils_test {
 
 postgres_ffi::for_all_postgres_versions! { xlog_utils_test }
 
-#[derive(Debug, Clone, PartialEq, Eq)]
 pub struct Conf {
     pub pg_version: u32,
     pub pg_distrib_dir: PathBuf,
@@ -53,7 +53,7 @@ impl Conf {
 
         #[allow(clippy::manual_range_patterns)]
         match self.pg_version {
-            14 | 15 | 16 => Ok(path.join(format!("v{}", self.pg_version))),
+            14 | 15 | 16 | 17 => Ok(path.join(format!("v{}", self.pg_version))),
             _ => bail!("Unsupported postgres version: {}", self.pg_version),
         }
     }
@@ -136,8 +136,8 @@ impl Conf {
 
     pub fn pg_waldump(
         &self,
-        first_segment_name: &str,
-        last_segment_name: &str,
+        first_segment_name: &OsStr,
+        last_segment_name: &OsStr,
     ) -> anyhow::Result<std::process::Output> {
         let first_segment_file = self.datadir.join(first_segment_name);
         let last_segment_file = self.datadir.join(last_segment_name);

libs/postgres_ffi/wal_craft/src/xlog_utils_test.rs

@@ -4,6 +4,7 @@ use super::*;
 use crate::{error, info};
 use regex::Regex;
 use std::cmp::min;
+use std::ffi::OsStr;
 use std::fs::{self, File};
 use std::io::Write;
 use std::{env, str::FromStr};
@@ -54,7 +55,7 @@ fn test_end_of_wal<C: crate::Crafter>(test_name: &str) {
         .wal_dir()
         .read_dir()
         .unwrap()
-        .map(|f| f.unwrap().file_name().into_string().unwrap())
+        .map(|f| f.unwrap().file_name())
         .filter(|fname| IsXLogFileName(fname))
         .max()
         .unwrap();
@@ -70,11 +71,11 @@ fn test_end_of_wal<C: crate::Crafter>(test_name: &str) {
             start_lsn
         );
         for file in fs::read_dir(cfg.wal_dir()).unwrap().flatten() {
-            let fname = file.file_name().into_string().unwrap();
+            let fname = file.file_name();
             if !IsXLogFileName(&fname) {
                 continue;
             }
-            let (segno, _) = XLogFromFileName(&fname, WAL_SEGMENT_SIZE);
+            let (segno, _) = XLogFromFileName(&fname, WAL_SEGMENT_SIZE).unwrap();
             let seg_start_lsn = XLogSegNoOffsetToRecPtr(segno, 0, WAL_SEGMENT_SIZE);
             if seg_start_lsn > u64::from(*start_lsn) {
                 continue;
@@ -93,10 +94,10 @@ fn test_end_of_wal<C: crate::Crafter>(test_name: &str) {
     }
 }
 
-fn find_pg_waldump_end_of_wal(cfg: &crate::Conf, last_segment: &str) -> Lsn {
+fn find_pg_waldump_end_of_wal(cfg: &crate::Conf, last_segment: &OsStr) -> Lsn {
     // Get the actual end of WAL by pg_waldump
     let waldump_output = cfg
-        .pg_waldump("000000010000000000000001", last_segment)
+        .pg_waldump(OsStr::new("000000010000000000000001"), last_segment)
         .unwrap()
         .stderr;
     let waldump_output = std::str::from_utf8(&waldump_output).unwrap();
@@ -117,7 +118,7 @@ fn find_pg_waldump_end_of_wal(cfg: &crate::Conf, last_segment: &str) -> Lsn {
 
 fn check_end_of_wal(
     cfg: &crate::Conf,
-    last_segment: &str,
+    last_segment: &OsStr,
     start_lsn: Lsn,
     expected_end_of_wal: Lsn,
 ) {
@@ -132,7 +133,8 @@ fn check_end_of_wal(
     // Rename file to partial to actually find last valid lsn, then rename it back.
     fs::rename(
         cfg.wal_dir().join(last_segment),
-        cfg.wal_dir().join(format!("{}.partial", last_segment)),
+        cfg.wal_dir()
+            .join(format!("{}.partial", last_segment.to_str().unwrap())),
     )
     .unwrap();
     let wal_end = find_end_of_wal(&cfg.wal_dir(), WAL_SEGMENT_SIZE, start_lsn).unwrap();
@@ -142,7 +144,8 @@ fn check_end_of_wal(
     );
     assert_eq!(wal_end, expected_end_of_wal);
     fs::rename(
-        cfg.wal_dir().join(format!("{}.partial", last_segment)),
+        cfg.wal_dir()
+            .join(format!("{}.partial", last_segment.to_str().unwrap())),
         cfg.wal_dir().join(last_segment),
     )
     .unwrap();

libs/pq_proto/Cargo.toml

@@ -8,10 +8,8 @@ license.workspace = true
 bytes.workspace = true
 byteorder.workspace = true
 itertools.workspace = true
-pin-project-lite.workspace = true
 postgres-protocol.workspace = true
 rand.workspace = true
 tokio = { workspace = true, features = ["io-util"] }
-tracing.workspace = true
 thiserror.workspace = true
 serde.workspace = true

libs/remote_storage/Cargo.toml

@@ -13,14 +13,11 @@ aws-smithy-async.workspace = true
 aws-smithy-types.workspace = true
 aws-config.workspace = true
 aws-sdk-s3.workspace = true
-aws-credential-types.workspace = true
 bytes.workspace = true
 camino = { workspace = true, features = ["serde1"] }
-humantime.workspace = true
 humantime-serde.workspace = true
 hyper = { workspace = true, features = ["stream"] }
 futures.workspace = true
-rand.workspace = true
 serde.workspace = true
 serde_json.workspace = true
 tokio = { workspace = true, features = ["sync", "fs", "io-util"] }

libs/remote_storage/src/config.rs

@@ -185,7 +185,7 @@ mod tests {
     use super::*;
 
     fn parse(input: &str) -> anyhow::Result<RemoteStorageConfig> {
-        let toml = input.parse::<toml_edit::Document>().unwrap();
+        let toml = input.parse::<toml_edit::DocumentMut>().unwrap();
         RemoteStorageConfig::from_toml(toml.as_item())
     }
 
@@ -235,6 +235,31 @@ timeout = '5s'";
         );
     }
 
+    #[test]
+    fn test_storage_class_serde_roundtrip() {
+        let classes = [
+            None,
+            Some(StorageClass::Standard),
+            Some(StorageClass::IntelligentTiering),
+        ];
+        for class in classes {
+            #[derive(Serialize, Deserialize)]
+            struct Wrapper {
+                #[serde(
+                    deserialize_with = "deserialize_storage_class",
+                    serialize_with = "serialize_storage_class"
+                )]
+                class: Option<StorageClass>,
+            }
+            let wrapped = Wrapper {
+                class: class.clone(),
+            };
+            let serialized = serde_json::to_string(&wrapped).unwrap();
+            let deserialized: Wrapper = serde_json::from_str(&serialized).unwrap();
+            assert_eq!(class, deserialized.class);
+        }
+    }
+
     #[test]
     fn test_azure_parsing() {
         let toml = "\

libs/remote_storage/src/lib.rs

@@ -45,6 +45,8 @@ pub use azure_core::Etag;
 
 pub use error::{DownloadError, TimeTravelError, TimeoutOrCancel};
 
+/// Default concurrency limit for S3 operations
+///
 /// Currently, sync happens with AWS S3, that has two limits on requests per second:
 /// ~200 RPS for IAM services
 /// <https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/UsingWithRDS.IAMDBAuth.html>
@@ -125,10 +127,6 @@ impl RemotePath {
         &self.0
     }
 
-    pub fn extension(&self) -> Option<&str> {
-        self.0.extension()
-    }
-
     pub fn strip_prefix(&self, p: &RemotePath) -> Result<&Utf8Path, std::path::StripPrefixError> {
         self.0.strip_prefix(&p.0)
     }
@@ -300,7 +298,9 @@ pub trait RemoteStorage: Send + Sync + 'static {
     ) -> Result<(), TimeTravelError>;
 }
 
-/// DownloadStream is sensitive to the timeout and cancellation used with the original
+/// Data part of an ongoing [`Download`].
+///
+/// `DownloadStream` is sensitive to the timeout and cancellation used with the original
 /// [`RemoteStorage::download`] request. The type yields `std::io::Result<Bytes>` to be compatible
 /// with `tokio::io::copy_buf`.
 // This has 'static because safekeepers do not use cancellation tokens (yet)

libs/safekeeper_api/Cargo.toml

@@ -6,6 +6,5 @@ license.workspace = true
 
 [dependencies]
 serde.workspace = true
-serde_with.workspace = true
 const_format.workspace = true
 utils.workspace = true

libs/safekeeper_api/src/models.rs

@@ -60,3 +60,16 @@ pub struct TimelineCopyRequest {
     pub target_timeline_id: TimelineId,
     pub until_lsn: Lsn,
 }
+
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct TimelineTermBumpRequest {
+    /// bump to
+    pub term: Option<u64>,
+}
+
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct TimelineTermBumpResponse {
+    // before the request
+    pub previous_term: u64,
+    pub current_term: u64,
+}

libs/tenant_size_model/src/lib.rs

@@ -5,9 +5,10 @@
 mod calculation;
 pub mod svg;
 
-/// StorageModel is the input to the synthetic size calculation. It represents
-/// a tree of timelines, with just the information that's needed for the
-/// calculation. This doesn't track timeline names or where each timeline
+/// StorageModel is the input to the synthetic size calculation.
+///
+/// It represents a tree of timelines, with just the information that's needed
+/// for the calculation. This doesn't track timeline names or where each timeline
 /// begins and ends, for example. Instead, it consists of "points of interest"
 /// on the timelines. A point of interest could be the timeline start or end point,
 /// the oldest point on a timeline that needs to be retained because of PITR

libs/tracing-utils/Cargo.toml

@@ -9,8 +9,9 @@ hyper.workspace = true
 opentelemetry = { workspace = true, features=["rt-tokio"] }
 opentelemetry-otlp = { workspace = true, default-features=false, features = ["http-proto", "trace", "http", "reqwest-client"] }
 opentelemetry-semantic-conventions.workspace = true
-reqwest = { workspace = true, default-features = false, features = ["rustls-tls"] }
 tokio = { workspace = true, features = ["rt", "rt-multi-thread"] }
 tracing.workspace = true
 tracing-opentelemetry.workspace = true
-tracing-subscriber.workspace = true
+
+[dev-dependencies]
+tracing-subscriber.workspace = true    # For examples in docs

libs/utils/Cargo.toml

@@ -14,7 +14,6 @@ testing = ["fail/failpoints"]
 arc-swap.workspace = true
 sentry.workspace = true
 async-compression.workspace = true
-async-trait.workspace = true
 anyhow.workspace = true
 bincode.workspace = true
 bytes.workspace = true
@@ -26,7 +25,6 @@ hyper = { workspace = true, features = ["full"] }
 fail.workspace = true
 futures = { workspace = true}
 jsonwebtoken.workspace = true
-leaky-bucket.workspace = true
 nix.workspace = true
 once_cell.workspace = true
 pin-project-lite.workspace = true
@@ -44,7 +42,6 @@ tracing.workspace = true
 tracing-error.workspace = true
 tracing-subscriber = { workspace = true, features = ["json", "registry"] }
 rand.workspace = true
-serde_with.workspace = true
 strum.workspace = true
 strum_macros.workspace = true
 url.workspace = true

libs/utils/src/accum.rs

@@ -1,33 +0,0 @@
-/// A helper to "accumulate" a value similar to `Iterator::reduce`, but lets you
-/// feed the accumulated values by calling the 'accum' function, instead of having an
-/// iterator.
-///
-/// For example, to calculate the smallest value among some integers:
-///
-/// ```
-/// use utils::accum::Accum;
-///
-/// let values = [1, 2, 3];
-///
-/// let mut min_value: Accum<u32> = Accum(None);
-/// for new_value in &values {
-///     min_value.accum(std::cmp::min, *new_value);
-/// }
-///
-/// assert_eq!(min_value.0.unwrap(), 1);
-/// ```
-pub struct Accum<T>(pub Option<T>);
-impl<T: Copy> Accum<T> {
-    pub fn accum<F>(&mut self, func: F, new_value: T)
-    where
-        F: FnOnce(T, T) -> T,
-    {
-        // If there is no previous value, just store the new value.
-        // Otherwise call the function to decide which one to keep.
-        self.0 = Some(if let Some(accum) = self.0 {
-            func(accum, new_value)
-        } else {
-            new_value
-        });
-    }
-}

libs/utils/src/circuit_breaker.rs

@@ -5,8 +5,10 @@ use std::{
 
 use metrics::IntCounter;
 
-/// Circuit breakers are for operations that are expensive and fallible: if they fail repeatedly,
-/// we will stop attempting them for some period of time, to avoid denial-of-service from retries, and
+/// Circuit breakers are for operations that are expensive and fallible.
+///
+/// If a circuit breaker fails repeatedly, we will stop attempting it for some
+/// period of time, to avoid denial-of-service from retries, and
 /// to mitigate the log spam from repeated failures.
 pub struct CircuitBreaker {
     /// An identifier that enables us to log useful errors when a circuit is broken

libs/utils/src/crashsafe.rs

@@ -1,3 +1,4 @@
+use std::os::fd::AsRawFd;
 use std::{
     borrow::Cow,
     fs::{self, File},
@@ -203,6 +204,27 @@ pub fn overwrite(
     Ok(())
 }
 
+/// Syncs the filesystem for the given file descriptor.
+#[cfg_attr(target_os = "macos", allow(unused_variables))]
+pub fn syncfs(fd: impl AsRawFd) -> anyhow::Result<()> {
+    // Linux guarantees durability for syncfs.
+    // POSIX doesn't have syncfs, and further does not actually guarantee durability of sync().
+    #[cfg(target_os = "linux")]
+    {
+        use anyhow::Context;
+        nix::unistd::syncfs(fd.as_raw_fd()).context("syncfs")?;
+    }
+    #[cfg(target_os = "macos")]
+    {
+        // macOS is not a production platform for Neon, don't even bother.
+    }
+    #[cfg(not(any(target_os = "linux", target_os = "macos")))]
+    {
+        compile_error!("Unsupported OS");
+    }
+    Ok(())
+}
+
 #[cfg(test)]
 mod tests {
 

libs/utils/src/http/error.rs

@@ -82,7 +82,7 @@ impl ApiError {
                 StatusCode::INTERNAL_SERVER_ERROR,
             ),
             ApiError::InternalServerError(err) => HttpErrorBody::response_from_msg_and_status(
-                err.to_string(),
+                format!("{err:#}"), // use alternative formatting so that we give the cause without backtrace
                 StatusCode::INTERNAL_SERVER_ERROR,
             ),
         }

libs/utils/src/id.rs

@@ -88,12 +88,6 @@ impl<'de> Deserialize<'de> for Id {
 }
 
 impl Id {
-    pub fn get_from_buf(buf: &mut impl bytes::Buf) -> Id {
-        let mut arr = [0u8; 16];
-        buf.copy_to_slice(&mut arr);
-        Id::from(arr)
-    }
-
     pub fn from_slice(src: &[u8]) -> Result<Id, IdError> {
         if src.len() != 16 {
             return Err(IdError::SliceParseError(src.len()));
@@ -179,10 +173,6 @@ impl fmt::Debug for Id {
 macro_rules! id_newtype {
     ($t:ident) => {
         impl $t {
-            pub fn get_from_buf(buf: &mut impl bytes::Buf) -> $t {
-                $t(Id::get_from_buf(buf))
-            }
-
             pub fn from_slice(src: &[u8]) -> Result<$t, IdError> {
                 Ok($t(Id::from_slice(src)?))
             }
@@ -249,8 +239,10 @@ macro_rules! id_newtype {
     };
 }
 
-/// Neon timeline IDs are different from PostgreSQL timeline
-/// IDs. They serve a similar purpose though: they differentiate
+/// Neon timeline ID.
+///
+/// They are different from PostgreSQL timeline
+/// IDs, but serve a similar purpose: they differentiate
 /// between different "histories" of the same cluster.  However,
 /// PostgreSQL timeline IDs are a bit cumbersome, because they are only
 /// 32-bits wide, and they must be in ascending order in any given

libs/utils/src/leaky_bucket.rs

@@ -0,0 +1,302 @@
+//! This module implements the Generic Cell Rate Algorithm for a simplified
+//! version of the Leaky Bucket rate limiting system.
+//!
+//! # Leaky Bucket
+//!
+//! If the bucket is full, no new requests are allowed and are throttled/errored.
+//! If the bucket is partially full/empty, new requests are added to the bucket in
+//! terms of "tokens".
+//!
+//! Over time, tokens are removed from the bucket, naturally allowing new requests at a steady rate.
+//!
+//! The bucket size tunes the burst support. The drain rate tunes the steady-rate requests per second.
+//!
+//! # [GCRA](https://en.wikipedia.org/wiki/Generic_cell_rate_algorithm)
+//!
+//! GCRA is a continuous rate leaky-bucket impl that stores minimal state and requires
+//! no background jobs to drain tokens, as the design utilises timestamps to drain automatically over time.
+//!
+//! We store an "empty_at" timestamp as the only state. As time progresses, we will naturally approach
+//! the empty state. The full-bucket state is calculated from `empty_at - config.bucket_width`.
+//!
+//! Another explaination can be found here: <https://brandur.org/rate-limiting>
+
+use std::{
+    sync::{
+        atomic::{AtomicU64, Ordering},
+        Mutex,
+    },
+    time::Duration,
+};
+
+use tokio::{sync::Notify, time::Instant};
+
+pub struct LeakyBucketConfig {
+    /// This is the "time cost" of a single request unit.
+    /// Should loosely represent how long it takes to handle a request unit in active resource time.
+    /// Loosely speaking this is the inverse of the steady-rate requests-per-second
+    pub cost: Duration,
+
+    /// total size of the bucket
+    pub bucket_width: Duration,
+}
+
+impl LeakyBucketConfig {
+    pub fn new(rps: f64, bucket_size: f64) -> Self {
+        let cost = Duration::from_secs_f64(rps.recip());
+        let bucket_width = cost.mul_f64(bucket_size);
+        Self { cost, bucket_width }
+    }
+}
+
+pub struct LeakyBucketState {
+    /// Bucket is represented by `allow_at..empty_at` where `allow_at = empty_at - config.bucket_width`.
+    ///
+    /// At any given time, `empty_at - now` represents the number of tokens in the bucket, multiplied by the "time_cost".
+    /// Adding `n` tokens to the bucket is done by moving `empty_at` forward by `n * config.time_cost`.
+    /// If `now < allow_at`, the bucket is considered filled and cannot accept any more tokens.
+    /// Draining the bucket will happen naturally as `now` moves forward.
+    ///
+    /// Let `n` be some "time cost" for the request,
+    /// If now is after empty_at, the bucket is empty and the empty_at is reset to now,
+    /// If now is within the `bucket window + n`, we are within time budget.
+    /// If now is before the `bucket window + n`, we have run out of budget.
+    ///
+    /// This is inspired by the generic cell rate algorithm (GCRA) and works
+    /// exactly the same as a leaky-bucket.
+    pub empty_at: Instant,
+}
+
+impl LeakyBucketState {
+    pub fn with_initial_tokens(config: &LeakyBucketConfig, initial_tokens: f64) -> Self {
+        LeakyBucketState {
+            empty_at: Instant::now() + config.cost.mul_f64(initial_tokens),
+        }
+    }
+
+    pub fn bucket_is_empty(&self, now: Instant) -> bool {
+        // if self.end is after now, the bucket is not empty
+        self.empty_at <= now
+    }
+
+    /// Immediately adds tokens to the bucket, if there is space.
+    ///
+    /// In a scenario where you are waiting for available rate,
+    /// rather than just erroring immediately, `started` corresponds to when this waiting started.
+    ///
+    /// `n` is the number of tokens that will be filled in the bucket.
+    ///
+    /// # Errors
+    ///
+    /// If there is not enough space, no tokens are added. Instead, an error is returned with the time when
+    /// there will be space again.
+    pub fn add_tokens(
+        &mut self,
+        config: &LeakyBucketConfig,
+        started: Instant,
+        n: f64,
+    ) -> Result<(), Instant> {
+        let now = Instant::now();
+
+        // invariant: started <= now
+        debug_assert!(started <= now);
+
+        // If the bucket was empty when we started our search,
+        // we should update the `empty_at` value accordingly.
+        // this prevents us from having negative tokens in the bucket.
+        let mut empty_at = self.empty_at;
+        if empty_at < started {
+            empty_at = started;
+        }
+
+        let n = config.cost.mul_f64(n);
+        let new_empty_at = empty_at + n;
+        let allow_at = new_empty_at.checked_sub(config.bucket_width);
+
+        //                     empty_at
+        //          allow_at    |   new_empty_at
+        //           /          |   /
+        // -------o-[---------o-|--]---------
+        //   now1 ^      now2 ^
+        //
+        // at now1, the bucket would be completely filled if we add n tokens.
+        // at now2, the bucket would be partially filled if we add n tokens.
+
+        match allow_at {
+            Some(allow_at) if now < allow_at => Err(allow_at),
+            _ => {
+                self.empty_at = new_empty_at;
+                Ok(())
+            }
+        }
+    }
+}
+
+pub struct RateLimiter {
+    pub config: LeakyBucketConfig,
+    pub sleep_counter: AtomicU64,
+    pub state: Mutex<LeakyBucketState>,
+    /// a queue to provide this fair ordering.
+    pub queue: Notify,
+}
+
+struct Requeue<'a>(&'a Notify);
+
+impl Drop for Requeue<'_> {
+    fn drop(&mut self) {
+        self.0.notify_one();
+    }
+}
+
+impl RateLimiter {
+    pub fn with_initial_tokens(config: LeakyBucketConfig, initial_tokens: f64) -> Self {
+        RateLimiter {
+            sleep_counter: AtomicU64::new(0),
+            state: Mutex::new(LeakyBucketState::with_initial_tokens(
+                &config,
+                initial_tokens,
+            )),
+            config,
+            queue: {
+                let queue = Notify::new();
+                queue.notify_one();
+                queue
+            },
+        }
+    }
+
+    pub fn steady_rps(&self) -> f64 {
+        self.config.cost.as_secs_f64().recip()
+    }
+
+    /// returns true if we did throttle
+    pub async fn acquire(&self, count: usize) -> bool {
+        let start = tokio::time::Instant::now();
+
+        let start_count = self.sleep_counter.load(Ordering::Acquire);
+        let mut end_count = start_count;
+
+        // wait until we are the first in the queue
+        let mut notified = std::pin::pin!(self.queue.notified());
+        if !notified.as_mut().enable() {
+            notified.await;
+            end_count = self.sleep_counter.load(Ordering::Acquire);
+        }
+
+        // notify the next waiter in the queue when we are done.
+        let _guard = Requeue(&self.queue);
+
+        loop {
+            let res = self
+                .state
+                .lock()
+                .unwrap()
+                .add_tokens(&self.config, start, count as f64);
+            match res {
+                Ok(()) => return end_count > start_count,
+                Err(ready_at) => {
+                    struct Increment<'a>(&'a AtomicU64);
+
+                    impl Drop for Increment<'_> {
+                        fn drop(&mut self) {
+                            self.0.fetch_add(1, Ordering::AcqRel);
+                        }
+                    }
+
+                    // increment the counter after we finish sleeping (or cancel this task).
+                    // this ensures that tasks that have already started the acquire will observe
+                    // the new sleep count when they are allowed to resume on the notify.
+                    let _inc = Increment(&self.sleep_counter);
+                    end_count += 1;
+
+                    tokio::time::sleep_until(ready_at).await;
+                }
+            }
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::time::Duration;
+
+    use tokio::time::Instant;
+
+    use super::{LeakyBucketConfig, LeakyBucketState};
+
+    #[tokio::test(start_paused = true)]
+    async fn check() {
+        let config = LeakyBucketConfig {
+            // average 100rps
+            cost: Duration::from_millis(10),
+            // burst up to 100 requests
+            bucket_width: Duration::from_millis(1000),
+        };
+
+        let mut state = LeakyBucketState {
+            empty_at: Instant::now(),
+        };
+
+        // supports burst
+        {
+            // should work for 100 requests this instant
+            for _ in 0..100 {
+                state.add_tokens(&config, Instant::now(), 1.0).unwrap();
+            }
+            let ready = state.add_tokens(&config, Instant::now(), 1.0).unwrap_err();
+            assert_eq!(ready - Instant::now(), Duration::from_millis(10));
+        }
+
+        // doesn't overfill
+        {
+            // after 1s we should have an empty bucket again.
+            tokio::time::advance(Duration::from_secs(1)).await;
+            assert!(state.bucket_is_empty(Instant::now()));
+
+            // after 1s more, we should not over count the tokens and allow more than 200 requests.
+            tokio::time::advance(Duration::from_secs(1)).await;
+            for _ in 0..100 {
+                state.add_tokens(&config, Instant::now(), 1.0).unwrap();
+            }
+            let ready = state.add_tokens(&config, Instant::now(), 1.0).unwrap_err();
+            assert_eq!(ready - Instant::now(), Duration::from_millis(10));
+        }
+
+        // supports sustained rate over a long period
+        {
+            tokio::time::advance(Duration::from_secs(1)).await;
+
+            // should sustain 100rps
+            for _ in 0..2000 {
+                tokio::time::advance(Duration::from_millis(10)).await;
+                state.add_tokens(&config, Instant::now(), 1.0).unwrap();
+            }
+        }
+
+        // supports requesting more tokens than can be stored in the bucket
+        // we just wait a little bit longer upfront.
+        {
+            // start the bucket completely empty
+            tokio::time::advance(Duration::from_secs(5)).await;
+            assert!(state.bucket_is_empty(Instant::now()));
+
+            // requesting 200 tokens of space should take 200*cost = 2s
+            // but we already have 1s available, so we wait 1s from start.
+            let start = Instant::now();
+
+            let ready = state.add_tokens(&config, start, 200.0).unwrap_err();
+            assert_eq!(ready - Instant::now(), Duration::from_secs(1));
+
+            tokio::time::advance(Duration::from_millis(500)).await;
+            let ready = state.add_tokens(&config, start, 200.0).unwrap_err();
+            assert_eq!(ready - Instant::now(), Duration::from_millis(500));
+
+            tokio::time::advance(Duration::from_millis(500)).await;
+            state.add_tokens(&config, start, 200.0).unwrap();
+
+            // bucket should be completely full now
+            let ready = state.add_tokens(&config, Instant::now(), 1.0).unwrap_err();
+            assert_eq!(ready - Instant::now(), Duration::from_millis(10));
+        }
+    }
+}

libs/utils/src/lib.rs

@@ -43,16 +43,9 @@ pub mod logging;
 pub mod lock_file;
 pub mod pid_file;
 
-// Misc
-pub mod accum;
-pub mod shutdown;
-
 // Utility for binding TcpListeners with proper socket options.
 pub mod tcp_listener;
 
-// Utility for putting a raw file descriptor into non-blocking mode
-pub mod nonblock;
-
 // Default signal handling
 pub mod sentry_init;
 pub mod signals;
@@ -71,6 +64,7 @@ pub mod postgres_client;
 
 pub mod tracing_span_assert;
 
+pub mod leaky_bucket;
 pub mod rate_limit;
 
 /// Simple once-barrier and a guard which keeps barrier awaiting.

libs/utils/src/lock_file.rs

@@ -100,7 +100,9 @@ pub enum LockFileRead {
 }
 
 /// Open & try to lock the lock file at the given `path`, returning a [handle][`LockFileRead`] to
-/// inspect its content. It is not an `Err(...)` if the file does not exist or is already locked.
+/// inspect its content.
+///
+/// It is not an `Err(...)` if the file does not exist or is already locked.
 /// Check the [`LockFileRead`] variants for details.
 pub fn read_and_hold_lock_file(path: &Utf8Path) -> anyhow::Result<LockFileRead> {
     let res = fs::OpenOptions::new().read(true).open(path);

libs/utils/src/logging.rs

@@ -3,9 +3,9 @@ use std::str::FromStr;
 use anyhow::Context;
 use metrics::{IntCounter, IntCounterVec};
 use once_cell::sync::Lazy;
-use strum_macros::{EnumString, EnumVariantNames};
+use strum_macros::{EnumString, VariantNames};
 
-#[derive(EnumString, EnumVariantNames, Eq, PartialEq, Debug, Clone, Copy)]
+#[derive(EnumString, strum_macros::Display, VariantNames, Eq, PartialEq, Debug, Clone, Copy)]
 #[strum(serialize_all = "snake_case")]
 pub enum LogFormat {
     Plain,
@@ -188,7 +188,7 @@ impl Drop for TracingPanicHookGuard {
 }
 
 /// Named symbol for our panic hook, which logs the panic.
-fn tracing_panic_hook(info: &std::panic::PanicInfo) {
+fn tracing_panic_hook(info: &std::panic::PanicHookInfo) {
     // following rust 1.66.1 std implementation:
     // https://github.com/rust-lang/rust/blob/90743e7298aca107ddaa0c202a4d3604e29bfeb6/library/std/src/panicking.rs#L235-L288
     let location = info.location();
@@ -274,6 +274,14 @@ impl From<String> for SecretString {
     }
 }
 
+impl FromStr for SecretString {
+    type Err = std::convert::Infallible;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        Ok(Self(s.to_string()))
+    }
+}
+
 impl std::fmt::Debug for SecretString {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
         write!(f, "[SECRET]")

libs/utils/src/lsn.rs

@@ -1,6 +1,5 @@
 #![warn(missing_docs)]
 
-use camino::Utf8Path;
 use serde::{de::Visitor, Deserialize, Serialize};
 use std::fmt;
 use std::ops::{Add, AddAssign};
@@ -145,14 +144,6 @@ impl Lsn {
         i128::from(self.0) - i128::from(other)
     }
 
-    /// Parse an LSN from a filename in the form `0000000000000000`
-    pub fn from_filename<F>(filename: F) -> Result<Self, LsnParseError>
-    where
-        F: AsRef<Utf8Path>,
-    {
-        Lsn::from_hex(filename.as_ref().as_str())
-    }
-
     /// Parse an LSN from a string in the form `0000000000000000`
     pub fn from_hex<S>(s: S) -> Result<Self, LsnParseError>
     where

libs/utils/src/nonblock.rs

@@ -1,17 +0,0 @@
-use nix::fcntl::{fcntl, OFlag, F_GETFL, F_SETFL};
-use std::os::unix::io::RawFd;
-
-/// Put a file descriptor into non-blocking mode
-pub fn set_nonblock(fd: RawFd) -> Result<(), std::io::Error> {
-    let bits = fcntl(fd, F_GETFL)?;
-
-    // If F_GETFL returns some unknown bits, they should be valid
-    // for passing back to F_SETFL, too. If we left them out, the F_SETFL
-    // would effectively clear them, which is not what we want.
-    let mut flags = OFlag::from_bits_retain(bits);
-    flags |= OFlag::O_NONBLOCK;
-
-    fcntl(fd, F_SETFL(flags))?;
-
-    Ok(())
-}

libs/utils/src/pageserver_feedback.rs

@@ -8,6 +8,7 @@ use tracing::{trace, warn};
 use crate::lsn::Lsn;
 
 /// Feedback pageserver sends to safekeeper and safekeeper resends to compute.
+///
 /// Serialized in custom flexible key/value format. In replication protocol, it
 /// is marked with NEON_STATUS_UPDATE_TAG_BYTE to differentiate from postgres
 /// Standby status update / Hot standby feedback messages.

libs/utils/src/poison.rs

@@ -65,6 +65,8 @@ impl<T> Poison<T> {
     }
 }
 
+/// Armed pointer to a [`Poison`].
+///
 /// Use [`Self::data`] and [`Self::data_mut`] to access the wrapped state.
 /// Once modifications are done, use [`Self::disarm`].
 /// If [`Guard`] gets dropped instead of calling [`Self::disarm`], the state is poisoned

libs/utils/src/rate_limit.rs

@@ -5,6 +5,15 @@ use std::time::{Duration, Instant};
 pub struct RateLimit {
     last: Option<Instant>,
     interval: Duration,
+    dropped: u64,
+}
+
+pub struct RateLimitStats(u64);
+
+impl std::fmt::Display for RateLimitStats {
+    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+        write!(f, "{} dropped calls", self.0)
+    }
 }
 
 impl RateLimit {
@@ -12,20 +21,27 @@ impl RateLimit {
         Self {
             last: None,
             interval,
+            dropped: 0,
         }
     }
 
     /// Call `f` if the rate limit allows.
     /// Don't call it otherwise.
     pub fn call<F: FnOnce()>(&mut self, f: F) {
+        self.call2(|_| f())
+    }
+
+    pub fn call2<F: FnOnce(RateLimitStats)>(&mut self, f: F) {
         let now = Instant::now();
         match self.last {
             Some(last) if now - last <= self.interval => {
                 // ratelimit
+                self.dropped += 1;
             }
             _ => {
                 self.last = Some(now);
-                f();
+                f(RateLimitStats(self.dropped));
+                self.dropped = 0;
             }
         }
     }

libs/utils/src/shard.rs

@@ -13,10 +13,11 @@ pub struct ShardNumber(pub u8);
 #[derive(Ord, PartialOrd, Eq, PartialEq, Clone, Copy, Serialize, Deserialize, Debug, Hash)]
 pub struct ShardCount(pub u8);
 
-/// Combination of ShardNumber and ShardCount.  For use within the context of a particular tenant,
-/// when we need to know which shard we're dealing with, but do not need to know the full
-/// ShardIdentity (because we won't be doing any page->shard mapping), and do not need to know
-/// the fully qualified TenantShardId.
+/// Combination of ShardNumber and ShardCount.
+///
+/// For use within the context of a particular tenant, when we need to know which shard we're
+/// dealing with, but do not need to know the full ShardIdentity (because we won't be doing
+/// any page->shard mapping), and do not need to know the fully qualified TenantShardId.
 #[derive(Eq, PartialEq, PartialOrd, Ord, Clone, Copy, Hash)]
 pub struct ShardIndex {
     pub shard_number: ShardNumber,

libs/utils/src/shutdown.rs

@@ -1,7 +0,0 @@
-/// Immediately terminate the calling process without calling
-/// atexit callbacks, C runtime destructors etc. We mainly use
-/// this to protect coverage data from concurrent writes.
-pub fn exit_now(code: u8) -> ! {
-    // SAFETY: exiting is safe, the ffi is not safe
-    unsafe { nix::libc::_exit(code as _) };
-}

libs/utils/src/simple_rcu.rs

@@ -49,12 +49,11 @@ use std::sync::{RwLock, RwLockWriteGuard};
 
 use tokio::sync::watch;
 
-///
 /// Rcu allows multiple readers to read and hold onto a value without blocking
-/// (for very long).  Storing to the Rcu updates the value, making new readers
-/// immediately see the new value, but it also waits for all current readers to
-/// finish.
+/// (for very long).
 ///
+/// Storing to the Rcu updates the value, making new readers immediately see
+/// the new value, but it also waits for all current readers to finish.
 pub struct Rcu<V> {
     inner: RwLock<RcuInner<V>>,
 }

libs/utils/src/sync/heavier_once_cell.rs

@@ -5,7 +5,9 @@ use std::sync::{
 use tokio::sync::Semaphore;
 
 /// Custom design like [`tokio::sync::OnceCell`] but using [`OwnedSemaphorePermit`] instead of
-/// `SemaphorePermit`, allowing use of `take` which does not require holding an outer mutex guard
+/// `SemaphorePermit`.
+///
+/// Allows use of `take` which does not require holding an outer mutex guard
 /// for the duration of initialization.
 ///
 /// Has no unsafe, builds upon [`tokio::sync::Semaphore`] and [`std::sync::Mutex`].

libs/utils/src/toml_edit_ext.rs

@@ -10,7 +10,7 @@ pub fn deserialize_item<T>(item: &toml_edit::Item) -> Result<T, Error>
 where
     T: serde::de::DeserializeOwned,
 {
-    let document: toml_edit::Document = match item {
+    let document: toml_edit::DocumentMut = match item {
         toml_edit::Item::Table(toml) => toml.clone().into(),
         toml_edit::Item::Value(toml_edit::Value::InlineTable(toml)) => {
             toml.clone().into_table().into()

libs/utils/src/vec_map.rs

@@ -7,6 +7,7 @@ pub enum VecMapOrdering {
 }
 
 /// Ordered map datastructure implemented in a Vec.
+///
 /// Append only - can only add keys that are larger than the
 /// current max key.
 /// Ordering can be adjusted using [`VecMapOrdering`]
@@ -119,32 +120,6 @@ impl<K: Ord, V> VecMap<K, V> {
         Ok((None, delta_size))
     }
 
-    /// Split the map into two.
-    ///
-    /// The left map contains everything before `cutoff` (exclusive).
-    /// Right map contains `cutoff` and everything after (inclusive).
-    pub fn split_at(&self, cutoff: &K) -> (Self, Self)
-    where
-        K: Clone,
-        V: Clone,
-    {
-        let split_idx = self
-            .data
-            .binary_search_by_key(&cutoff, extract_key)
-            .unwrap_or_else(std::convert::identity);
-
-        (
-            VecMap {
-                data: self.data[..split_idx].to_vec(),
-                ordering: self.ordering,
-            },
-            VecMap {
-                data: self.data[split_idx..].to_vec(),
-                ordering: self.ordering,
-            },
-        )
-    }
-
     /// Move items from `other` to the end of `self`, leaving `other` empty.
     /// If the `other` ordering is different from `self` ordering
     /// `ExtendOrderingError` error will be returned.

libs/utils/src/yielding_loop.rs

@@ -6,9 +6,10 @@ pub enum YieldingLoopError {
     Cancelled,
 }
 
-/// Helper for long synchronous loops, e.g. over all tenants in the system.  Periodically
-/// yields to avoid blocking the executor, and after resuming checks the provided
-/// cancellation token to drop out promptly on shutdown.
+/// Helper for long synchronous loops, e.g. over all tenants in the system.
+///
+/// Periodically yields to avoid blocking the executor, and after resuming
+/// checks the provided cancellation token to drop out promptly on shutdown.
 #[inline(always)]
 pub async fn yielding_loop<I, T, F>(
     interval: usize,
@@ -23,7 +24,7 @@ where
     for (i, item) in iter.enumerate() {
         visitor(item);
 
-        if i + 1 % interval == 0 {
+        if (i + 1) % interval == 0 {
             tokio::task::yield_now().await;
             if cancel.is_cancelled() {
                 return Err(YieldingLoopError::Cancelled);

libs/vm_monitor/Cargo.toml

@@ -15,13 +15,11 @@ anyhow.workspace = true
 axum.workspace = true
 clap.workspace = true
 futures.workspace = true
-inotify.workspace = true
 serde.workspace = true
 serde_json.workspace = true
 sysinfo.workspace = true
 tokio = { workspace = true, features = ["rt-multi-thread"] }
 tokio-postgres.workspace = true
-tokio-stream.workspace = true
 tokio-util.workspace = true
 tracing.workspace = true
 tracing-subscriber.workspace = true

libs/walproposer/build.rs

@@ -4,7 +4,8 @@
 use std::{env, path::PathBuf, process::Command};
 
 use anyhow::{anyhow, Context};
-use bindgen::CargoCallbacks;
+
+const WALPROPOSER_PG_VERSION: &str = "v17";
 
 fn main() -> anyhow::Result<()> {
     // Tell cargo to invalidate the built crate whenever the wrapper changes
@@ -37,7 +38,10 @@ fn main() -> anyhow::Result<()> {
     // Rebuild crate when libwalproposer.a changes
     println!("cargo:rerun-if-changed={walproposer_lib_search_str}/libwalproposer.a");
 
-    let pg_config_bin = pg_install_abs.join("v16").join("bin").join("pg_config");
+    let pg_config_bin = pg_install_abs
+        .join(WALPROPOSER_PG_VERSION)
+        .join("bin")
+        .join("pg_config");
     let inc_server_path: String = if pg_config_bin.exists() {
         let output = Command::new(pg_config_bin)
             .arg("--includedir-server")
@@ -54,7 +58,7 @@ fn main() -> anyhow::Result<()> {
             .into()
     } else {
         let server_path = pg_install_abs
-            .join("v16")
+            .join(WALPROPOSER_PG_VERSION)
             .join("include")
             .join("postgresql")
             .join("server")
@@ -64,16 +68,25 @@ fn main() -> anyhow::Result<()> {
             .map_err(|s| anyhow!("Bad postgres server path {s:?}"))?
     };
 
+    let unwind_abi_functions = [
+        "log_internal",
+        "recovery_download",
+        "start_streaming",
+        "finish_sync_safekeepers",
+        "wait_event_set",
+        "WalProposerStart",
+    ];
+
     // The bindgen::Builder is the main entry point
     // to bindgen, and lets you build up options for
     // the resulting bindings.
-    let bindings = bindgen::Builder::default()
+    let mut builder = bindgen::Builder::default()
         // The input header we would like to generate
         // bindings for.
         .header("bindgen_deps.h")
         // Tell cargo to invalidate the built crate whenever any of the
         // included header files changed.
-        .parse_callbacks(Box::new(CargoCallbacks))
+        .parse_callbacks(Box::new(bindgen::CargoCallbacks::new()))
         .allowlist_type("WalProposer")
         .allowlist_type("WalProposerConfig")
         .allowlist_type("walproposer_api")
@@ -105,7 +118,12 @@ fn main() -> anyhow::Result<()> {
         .allowlist_var("WL_SOCKET_MASK")
         .clang_arg("-DWALPROPOSER_LIB")
         .clang_arg(format!("-I{pgxn_neon}"))
-        .clang_arg(format!("-I{inc_server_path}"))
+        .clang_arg(format!("-I{inc_server_path}"));
+
+    for name in unwind_abi_functions {
+        builder = builder.override_abi(bindgen::Abi::CUnwind, name);
+    }
+    let bindings = builder
         // Finish the builder and generate the bindings.
         .generate()
         // Unwrap the Result and panic on failure.

libs/walproposer/src/api_bindings.rs

@@ -33,7 +33,7 @@ extern "C" fn get_shmem_state(wp: *mut WalProposer) -> *mut WalproposerShmemStat
     }
 }
 
-extern "C" fn start_streaming(wp: *mut WalProposer, startpos: XLogRecPtr) {
+extern "C-unwind" fn start_streaming(wp: *mut WalProposer, startpos: XLogRecPtr) {
     unsafe {
         let callback_data = (*(*wp).config).callback_data;
         let api = callback_data as *mut Box<dyn ApiImpl>;
@@ -187,7 +187,7 @@ extern "C" fn conn_blocking_write(
     }
 }
 
-extern "C" fn recovery_download(wp: *mut WalProposer, sk: *mut Safekeeper) -> bool {
+extern "C-unwind" fn recovery_download(wp: *mut WalProposer, sk: *mut Safekeeper) -> bool {
     unsafe {
         let callback_data = (*(*(*sk).wp).config).callback_data;
         let api = callback_data as *mut Box<dyn ApiImpl>;
@@ -272,7 +272,7 @@ extern "C" fn rm_safekeeper_event_set(sk: *mut Safekeeper) {
     }
 }
 
-extern "C" fn wait_event_set(
+extern "C-unwind" fn wait_event_set(
     wp: *mut WalProposer,
     timeout: ::std::os::raw::c_long,
     event_sk: *mut *mut Safekeeper,
@@ -324,7 +324,7 @@ extern "C" fn get_redo_start_lsn(wp: *mut WalProposer) -> XLogRecPtr {
     }
 }
 
-extern "C" fn finish_sync_safekeepers(wp: *mut WalProposer, lsn: XLogRecPtr) {
+extern "C-unwind" fn finish_sync_safekeepers(wp: *mut WalProposer, lsn: XLogRecPtr) {
     unsafe {
         let callback_data = (*(*wp).config).callback_data;
         let api = callback_data as *mut Box<dyn ApiImpl>;
@@ -340,7 +340,7 @@ extern "C" fn process_safekeeper_feedback(wp: *mut WalProposer, sk: *mut Safekee
     }
 }
 
-extern "C" fn log_internal(
+extern "C-unwind" fn log_internal(
     wp: *mut WalProposer,
     level: ::std::os::raw::c_int,
     line: *const ::std::os::raw::c_char,