fixup: doc reference to renamed field

## Problem

`gather-rust-build-stats` extra CI job fails with 
```
"PQ_LIB_DIR" doesn't exist in the configured path: "/__w/neon/neon/pg_install/v16/lib"
```

## Summary of changes
- Use the path to Postgres 17 for the `gather-rust-build-stats` job. 
The job uses Postgres built by `make walproposer-lib`

.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)
 
@@ -210,15 +225,19 @@ 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 $CARGO_FEATURES -E '!package(pageserver)'
+
+          # run pageserver tests with different settings
           for io_engine in std-fs tokio-epoll-uring ; do
             for io_buffer_alignment in 0 1 512 ; do
-              NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOENGINE=$io_engine NEON_PAGESERVER_UNIT_TEST_IO_BUFFER_ALIGNMENT=$io_buffer_alignment ${cov_prefix} cargo nextest run $CARGO_FLAGS $CARGO_FEATURES
+              NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOENGINE=$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
 

.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

@@ -211,7 +211,7 @@ 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 +286,7 @@ 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
 
@@ -547,7 +548,7 @@ 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')) }}
@@ -626,7 +627,7 @@ 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 +649,7 @@ jobs:
 
     strategy:
       matrix:
-        version: [ v14, v15, v16 ]
+        version: [ v14, v15, v16, v17 ]
 
     steps:
       - uses: docker/login-action@v3
@@ -670,7 +671,7 @@ 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 +689,7 @@ 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 }}
@@ -699,7 +700,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        version: [ v14, v15, v16 ]
+        version: [ v14, v15, v16, v17 ]
     env:
       VM_BUILDER_VERSION: v0.29.3
 
@@ -793,14 +794,11 @@ 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 +820,6 @@ 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}:${{ needs.tag.outputs.build-tag }}
-          done
-
       - name: Add latest tag to images
         if: github.ref_name == 'main'
         run: |
@@ -863,7 +839,7 @@ jobs:
             done
           done
           docker buildx imagetools create -t neondatabase/neon-test-extensions-v16:latest \
-                                             neondatabase/neon-test-extensions-v16:${{ 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 +852,35 @@ 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 +956,8 @@ 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 ]
+    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
@@ -971,7 +971,7 @@ 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
@@ -1055,43 +1055,89 @@ 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 ]
+    needs: [ deploy ]
     if: github.ref_name == 'release'
 
-    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

.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.lock

@@ -915,25 +915,22 @@ dependencies = [
 
 [[package]]
 name = "bindgen"
-version = "0.65.1"
+version = "0.70.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "cfdf7b466f9a4903edc73f95d6d2bcd5baf8ae620638762244d3f60143643cc5"
+checksum = "f49d8fed880d473ea71efb9bf597651e77201bdd4893efe54c9e5d65ae04ce6f"
 dependencies = [
- "bitflags 1.3.2",
+ "bitflags 2.4.1",
  "cexpr",
  "clang-sys",
- "lazy_static",
- "lazycell",
+ "itertools 0.12.1",
  "log",
- "peeking_take_while",
- "prettyplease 0.2.6",
+ "prettyplease 0.2.17",
  "proc-macro2",
  "quote",
  "regex",
  "rustc-hash",
  "shlex",
  "syn 2.0.52",
- "which",
 ]
 
 [[package]]
@@ -1192,9 +1189,9 @@ dependencies = [
 
 [[package]]
 name = "comfy-table"
-version = "6.1.4"
+version = "7.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6e7b787b0dc42e8111badfdbe4c3059158ccb2db8780352fa1b01e8ccf45cc4d"
+checksum = "b34115915337defe99b2aff5c2ce6771e5fbc4079f4b506301f5cf394c8452f7"
 dependencies = [
  "crossterm",
  "strum",
@@ -1249,7 +1246,7 @@ dependencies = [
  "tokio-postgres",
  "tokio-stream",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "tracing",
  "tracing-opentelemetry",
  "tracing-subscriber",
@@ -1333,7 +1330,6 @@ name = "control_plane"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "async-trait",
  "camino",
  "clap",
  "comfy-table",
@@ -1364,8 +1360,8 @@ dependencies = [
  "tokio",
  "tokio-postgres",
  "tokio-util",
- "toml 0.7.4",
- "toml_edit 0.19.10",
+ "toml",
+ "toml_edit",
  "tracing",
  "url",
  "utils",
@@ -1489,25 +1485,22 @@ checksum = "248e3bacc7dc6baa3b21e405ee045c3047101a49145e7e9eca583ab4c2ca5345"
 
 [[package]]
 name = "crossterm"
-version = "0.25.0"
+version = "0.27.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e64e6c0fbe2c17357405f7c758c1ef960fce08bdfb2c03d88d2a18d7e09c4b67"
+checksum = "f476fe445d41c9e991fd07515a6f463074b782242ccf4a5b7b1d1012e70824df"
 dependencies = [
- "bitflags 1.3.2",
+ "bitflags 2.4.1",
  "crossterm_winapi",
  "libc",
- "mio",
  "parking_lot 0.12.1",
- "signal-hook",
- "signal-hook-mio",
  "winapi",
 ]
 
 [[package]]
 name = "crossterm_winapi"
-version = "0.9.0"
+version = "0.9.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2ae1b35a484aa10e07fe0638d02301c5ad24de82d310ccbd2f3693da5f09bf1c"
+checksum = "acdd7c62a3665c7f6830a51635d9ac9b23ed385797f70a83bb8bafe9c572ab2b"
 dependencies = [
  "winapi",
 ]
@@ -2728,6 +2721,12 @@ dependencies = [
  "hashbrown 0.14.5",
 ]
 
+[[package]]
+name = "indoc"
+version = "2.0.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b248f5224d1d606005e02c97f5aa4e88eeb230488bcc03bc9ca4d7991399f2b5"
+
 [[package]]
 name = "infer"
 version = "0.2.3"
@@ -2944,23 +2943,6 @@ dependencies = [
  "spin 0.5.2",
 ]
 
-[[package]]
-name = "lazycell"
-version = "1.3.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "830d08ce1d1d941e6b30645f1a0eb5643013d835ce3779a5fc208261dbe10f55"
-
-[[package]]
-name = "leaky-bucket"
-version = "1.0.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8eb491abd89e9794d50f93c8db610a29509123e3fbbc9c8c67a528e9391cd853"
-dependencies = [
- "parking_lot 0.12.1",
- "tokio",
- "tracing",
-]
-
 [[package]]
 name = "libc"
 version = "0.2.150"
@@ -3159,7 +3141,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "fd01039851e82f8799046eabbb354056283fb265c8ec0996af940f4e85a380ff"
 dependencies = [
  "serde",
- "toml 0.8.14",
+ "toml",
 ]
 
 [[package]]
@@ -3675,7 +3657,7 @@ dependencies = [
  "thiserror",
  "tokio",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "utils",
  "workspace_hack",
 ]
@@ -3713,8 +3695,8 @@ dependencies = [
  "humantime",
  "humantime-serde",
  "hyper 0.14.26",
+ "indoc",
  "itertools 0.10.5",
- "leaky-bucket",
  "md5",
  "metrics",
  "nix 0.27.1",
@@ -3762,7 +3744,7 @@ dependencies = [
  "tokio-stream",
  "tokio-tar",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "tracing",
  "twox-hash",
  "url",
@@ -3779,6 +3761,7 @@ dependencies = [
  "bincode",
  "byteorder",
  "bytes",
+ "camino",
  "chrono",
  "const_format",
  "enum-map",
@@ -3786,11 +3769,16 @@ dependencies = [
  "humantime",
  "humantime-serde",
  "itertools 0.10.5",
+ "nix 0.27.1",
+ "postgres_backend",
  "postgres_ffi",
  "rand 0.8.5",
+ "remote_storage",
+ "reqwest 0.12.4",
  "serde",
  "serde_json",
  "serde_with",
+ "storage_broker",
  "strum",
  "strum_macros",
  "thiserror",
@@ -3802,7 +3790,6 @@ name = "pageserver_client"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "async-trait",
  "bytes",
  "futures",
  "pageserver_api",
@@ -3920,8 +3907,9 @@ dependencies = [
 
 [[package]]
 name = "parquet"
-version = "51.0.0"
-source = "git+https://github.com/apache/arrow-rs?branch=master#2534976a564be3d2d56312dc88fb1b6ed4cef829"
+version = "53.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f0fbf928021131daaa57d334ca8e3904fe9ae22f73c56244fc7db9b04eedc3d8"
 dependencies = [
  "ahash",
  "bytes",
@@ -3940,8 +3928,9 @@ dependencies = [
 
 [[package]]
 name = "parquet_derive"
-version = "51.0.0"
-source = "git+https://github.com/apache/arrow-rs?branch=master#2534976a564be3d2d56312dc88fb1b6ed4cef829"
+version = "53.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "86e9fcfae007533a06b580429a3f7e07cb833ec8aa37c041c16563e7918f057e"
 dependencies = [
  "parquet",
  "proc-macro2",
@@ -3978,12 +3967,6 @@ dependencies = [
  "sha2",
 ]
 
-[[package]]
-name = "peeking_take_while"
-version = "0.1.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "19b17cddbe7ec3f8bc800887bab5e717348c95ea2ca0b1bf0837fb964dc67099"
-
 [[package]]
 name = "pem"
 version = "3.0.3"
@@ -4137,7 +4120,7 @@ dependencies = [
 [[package]]
 name = "postgres"
 version = "0.19.4"
-source = "git+https://github.com/neondatabase/rust-postgres.git?branch=neon#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
+source = "git+https://github.com/neondatabase/rust-postgres.git?rev=20031d7a9ee1addeae6e0968e3899ae6bf01cee2#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
 dependencies = [
  "bytes",
  "fallible-iterator",
@@ -4150,7 +4133,7 @@ dependencies = [
 [[package]]
 name = "postgres-protocol"
 version = "0.6.4"
-source = "git+https://github.com/neondatabase/rust-postgres.git?branch=neon#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
+source = "git+https://github.com/neondatabase/rust-postgres.git?rev=20031d7a9ee1addeae6e0968e3899ae6bf01cee2#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
 dependencies = [
  "base64 0.20.0",
  "byteorder",
@@ -4169,7 +4152,7 @@ dependencies = [
 [[package]]
 name = "postgres-types"
 version = "0.2.4"
-source = "git+https://github.com/neondatabase/rust-postgres.git?branch=neon#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
+source = "git+https://github.com/neondatabase/rust-postgres.git?rev=20031d7a9ee1addeae6e0968e3899ae6bf01cee2#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
 dependencies = [
  "bytes",
  "fallible-iterator",
@@ -4281,9 +4264,9 @@ dependencies = [
 
 [[package]]
 name = "prettyplease"
-version = "0.2.6"
+version = "0.2.17"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3b69d39aab54d069e7f2fe8cb970493e7834601ca2d8c65fd7bbd183578080d1"
+checksum = "8d3928fb5db768cb86f891ff014f0144589297e3c6a1aba6ed7cecfdace270c7"
 dependencies = [
  "proc-macro2",
  "syn 2.0.52",
@@ -4828,7 +4811,7 @@ dependencies = [
  "tokio",
  "tokio-stream",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "tracing",
  "utils",
 ]
@@ -5338,7 +5321,7 @@ dependencies = [
  "tokio-stream",
  "tokio-tar",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "tracing",
  "tracing-subscriber",
  "url",
@@ -5747,17 +5730,6 @@ dependencies = [
  "signal-hook-registry",
 ]
 
-[[package]]
-name = "signal-hook-mio"
-version = "0.2.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "29ad2e15f37ec9a6cc544097b78a1ec90001e9f71b81338ca39f430adaca99af"
-dependencies = [
- "libc",
- "mio",
- "signal-hook",
-]
-
 [[package]]
 name = "signal-hook-registry"
 version = "1.4.1"
@@ -5964,7 +5936,6 @@ name = "storage_controller_client"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "async-trait",
  "bytes",
  "futures",
  "pageserver_api",
@@ -6071,21 +6042,21 @@ checksum = "73473c0e59e6d5812c5dfe2a064a6444949f089e20eec9a2e5506596494e4623"
 
 [[package]]
 name = "strum"
-version = "0.24.1"
+version = "0.26.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "063e6045c0e62079840579a7e47a355ae92f60eb74daaf156fb1e84ba164e63f"
+checksum = "8fec0f0aef304996cf250b31b5a10dee7980c85da9d759361292b8bca5a18f06"
 
 [[package]]
 name = "strum_macros"
-version = "0.24.3"
+version = "0.26.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1e385be0d24f186b4ce2f9982191e7101bb737312ad61c1f2f984f34bcf85d59"
+checksum = "4c6bee85a5a24955dc440386795aa378cd9cf82acd5f764469152d2270e581be"
 dependencies = [
- "heck 0.4.1",
+ "heck 0.5.0",
  "proc-macro2",
  "quote",
  "rustversion",
- "syn 1.0.109",
+ "syn 2.0.52",
 ]
 
 [[package]]
@@ -6096,8 +6067,9 @@ checksum = "81cdd64d312baedb58e21336b31bc043b77e01cc99033ce76ef539f78e965ebc"
 
 [[package]]
 name = "svg_fmt"
-version = "0.4.2"
-source = "git+https://github.com/nical/rust_debug?rev=28a7d96eecff2f28e75b1ea09f2d499a60d0e3b4#28a7d96eecff2f28e75b1ea09f2d499a60d0e3b4"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "20e16a0f46cf5fd675563ef54f26e83e20f2366bcf027bcb3cc3ed2b98aaf2ca"
 
 [[package]]
 name = "syn"
@@ -6425,7 +6397,7 @@ dependencies = [
 [[package]]
 name = "tokio-postgres"
 version = "0.7.7"
-source = "git+https://github.com/neondatabase/rust-postgres.git?branch=neon#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
+source = "git+https://github.com/neondatabase/rust-postgres.git?rev=20031d7a9ee1addeae6e0968e3899ae6bf01cee2#20031d7a9ee1addeae6e0968e3899ae6bf01cee2"
 dependencies = [
  "async-trait",
  "byteorder",
@@ -6536,18 +6508,6 @@ dependencies = [
  "tracing",
 ]
 
-[[package]]
-name = "toml"
-version = "0.7.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d6135d499e69981f9ff0ef2167955a5333c35e36f6937d382974566b3d5b94ec"
-dependencies = [
- "serde",
- "serde_spanned",
- "toml_datetime",
- "toml_edit 0.19.10",
-]
-
 [[package]]
 name = "toml"
 version = "0.8.14"
@@ -6557,7 +6517,7 @@ dependencies = [
  "serde",
  "serde_spanned",
  "toml_datetime",
- "toml_edit 0.22.14",
+ "toml_edit",
 ]
 
 [[package]]
@@ -6569,19 +6529,6 @@ dependencies = [
  "serde",
 ]
 
-[[package]]
-name = "toml_edit"
-version = "0.19.10"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2380d56e8670370eee6566b0bfd4265f65b3f432e8c6d85623f728d4fa31f739"
-dependencies = [
- "indexmap 1.9.3",
- "serde",
- "serde_spanned",
- "toml_datetime",
- "winnow 0.4.6",
-]
-
 [[package]]
 name = "toml_edit"
 version = "0.22.14"
@@ -6592,7 +6539,7 @@ dependencies = [
  "serde",
  "serde_spanned",
  "toml_datetime",
- "winnow 0.6.13",
+ "winnow",
 ]
 
 [[package]]
@@ -6967,7 +6914,6 @@ dependencies = [
  "anyhow",
  "arc-swap",
  "async-compression",
- "async-trait",
  "bincode",
  "byteorder",
  "bytes",
@@ -6983,7 +6929,6 @@ dependencies = [
  "humantime",
  "hyper 0.14.26",
  "jsonwebtoken",
- "leaky-bucket",
  "metrics",
  "nix 0.27.1",
  "once_cell",
@@ -7007,7 +6952,7 @@ dependencies = [
  "tokio-stream",
  "tokio-tar",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "tracing",
  "tracing-error",
  "tracing-subscriber",
@@ -7553,15 +7498,6 @@ version = "0.52.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "32b752e52a2da0ddfbdbcc6fceadfeede4c939ed16d13e648833a61dfb611ed8"
 
-[[package]]
-name = "winnow"
-version = "0.4.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "61de7bac303dc551fe038e2b3cef0f571087a47571ea6e79a87692ac99b99699"
-dependencies = [
- "memchr",
-]
-
 [[package]]
 name = "winnow"
 version = "0.6.13"
@@ -7631,6 +7567,7 @@ dependencies = [
  "hyper 0.14.26",
  "indexmap 1.9.3",
  "itertools 0.10.5",
+ "itertools 0.12.1",
  "lazy_static",
  "libc",
  "log",
@@ -7668,6 +7605,7 @@ dependencies = [
  "tokio",
  "tokio-rustls 0.24.0",
  "tokio-util",
+ "toml_edit",
  "tonic",
  "tower",
  "tracing",

Cargo.toml

@@ -64,7 +64,7 @@ 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"
@@ -73,7 +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"
@@ -103,12 +103,12 @@ 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"] }
@@ -123,8 +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"
@@ -158,11 +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"
@@ -178,8 +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"
@@ -202,10 +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/" }
@@ -242,11 +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

@@ -79,6 +79,7 @@ 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 +88,11 @@ 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 +101,10 @@ 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 +130,10 @@ 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 +153,19 @@ 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 +190,13 @@ 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 "$(uname -m)" in \
       "x86_64") \
         export CMAKE_CHECKSUM=739d372726cb23129d57a539ce1432453448816e345e1545f6127296926b6754 \
         ;; \
@@ -192,7 +214,11 @@ RUN case "$(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 +228,10 @@ 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 +247,13 @@ 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 +272,7 @@ 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 +280,10 @@ 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 +298,14 @@ 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 +318,13 @@ 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 +338,13 @@ 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 +358,15 @@ 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 +381,13 @@ 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 +401,13 @@ 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 +421,13 @@ 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 +441,13 @@ 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 +461,13 @@ 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 $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
@@ -413,7 +486,10 @@ COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 ARG PG_VERSION
 ENV PATH="/usr/local/pgsql/bin:$PATH"
 
-RUN case "${PG_VERSION}" in \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    case "${PG_VERSION}" in \
       "v14" | "v15") \
         export TIMESCALEDB_VERSION=2.10.1 \
         export TIMESCALEDB_CHECKSUM=6fca72a6ed0f6d32d2b3523951ede73dc5f9b0077b38450a029a5f411fdb8c73 \
@@ -446,7 +522,10 @@ COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 ARG PG_VERSION
 ENV PATH="/usr/local/pgsql/bin:$PATH"
 
-RUN case "${PG_VERSION}" in \
+RUN case "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    case "${PG_VERSION}" in \
       "v14") \
         export PG_HINT_PLAN_VERSION=14_1_4_1 \
         export PG_HINT_PLAN_CHECKSUM=c3501becf70ead27f70626bce80ea401ceac6a77e2083ee5f3ff1f1444ec1ad1 \
@@ -459,6 +538,9 @@ RUN case "${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 +560,14 @@ 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 +581,13 @@ 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 +597,10 @@ 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 +637,14 @@ 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 +658,14 @@ 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 +679,14 @@ 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 $(getconf _NPROCESSORS_ONLN) && \
@@ -620,10 +725,14 @@ RUN case "${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 +750,7 @@ 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 +761,11 @@ 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 +784,10 @@ 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 +809,10 @@ 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 +832,10 @@ 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 +854,10 @@ 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 +872,14 @@ 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 +892,14 @@ 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 +913,14 @@ 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 +990,8 @@ 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 +1035,7 @@ 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 +1054,10 @@ 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
@@ -956,18 +1095,39 @@ 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 "${PG_VERSION}" in "v17") \
+    echo "v17 extensions are not supported yet. Quit" && exit 0;; \
+    esac && \
+    cd /ext-src/pgvector-src && patch -p1 <../pgvector.patch
 # cmake is required for the h3 test
-RUN apt-get update && apt-get install -y cmake
-RUN cd /ext-src/pg_hint_plan-src && patch -p1 < /ext-src/pg_hint_plan.patch
+RUN case "${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

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/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 {

control_plane/Cargo.toml

@@ -6,7 +6,6 @@ license.workspace = true
 
 [dependencies]
 anyhow.workspace = true
-async-trait.workspace = true
 camino.workspace = true
 clap.workspace = true
 comfy-table.workspace = true

control_plane/src/bin/neon_local.rs

@@ -640,6 +640,8 @@ 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 +660,6 @@ 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,
@@ -1570,7 +1571,6 @@ fn cli() -> Command {
                         .value_parser(value_parser!(PathBuf))
                         .value_name("config")
                 )
-                .arg(pg_version_arg.clone())
                 .arg(force_arg)
         )
         .subcommand(
@@ -1583,6 +1583,7 @@ 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))

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

@@ -75,14 +75,14 @@ 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 +137,9 @@ 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 +181,23 @@ 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 +233,7 @@ 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 +244,10 @@ 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(),
         )

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/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,
@@ -80,7 +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)]
@@ -336,7 +339,7 @@ async fn main() -> anyhow::Result<()> {
                         listen_pg_port,
                         listen_http_addr,
                         listen_http_port,
-                        availability_zone_id: Some(availability_zone_id),
+                        availability_zone_id,
                     }),
                 )
                 .await?;
@@ -403,7 +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,

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 +42,476 @@ 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 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(
+    Eq,
+    PartialEq,
+    Debug,
+    Copy,
+    Clone,
+    strum_macros::EnumString,
+    strum_macros::Display,
+    serde_with::DeserializeFromStr,
+    serde_with::SerializeDisplay,
+)]
+#[strum(serialize_all = "kebab-case")]
+pub enum GetVectoredImpl {
+    Sequential,
+    Vectored,
+}
+
+#[derive(
+    Eq,
+    PartialEq,
+    Debug,
+    Copy,
+    Clone,
+    strum_macros::EnumString,
+    strum_macros::Display,
+    serde_with::DeserializeFromStr,
+    serde_with::SerializeDisplay,
+)]
+#[strum(serialize_all = "kebab-case")]
+pub enum GetImpl {
+    Legacy,
+    Vectored,
+}
+
+#[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_VALIDATE_VECTORED_GET: bool = false;
+
+    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()),
+            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(1).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;
+
+    pub const DEFAULT_INGEST_BATCH_SIZE: u64 = 100;
+}
+
+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};
 
@@ -57,7 +57,7 @@ pub struct NodeRegisterRequest {
     pub listen_http_addr: String,
     pub listen_http_port: u16,
 
-    pub availability_zone_id: Option<String>,
+    pub availability_zone_id: String,
 }
 
 #[derive(Serialize, Deserialize)]
@@ -74,6 +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,
@@ -101,6 +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,
@@ -132,8 +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)

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)
@@ -236,15 +263,6 @@ impl Key {
         field5: u8::MAX,
         field6: u32::MAX,
     };
-    /// A key slightly smaller than [`Key::MAX`] for use in layer key ranges to avoid them to be confused with L0 layers
-    pub const NON_L0_MAX: Key = Key {
-        field1: u8::MAX,
-        field2: u32::MAX,
-        field3: u32::MAX,
-        field4: u32::MAX,
-        field5: u8::MAX,
-        field6: u32::MAX - 1,
-    };
 
     pub fn from_hex(s: &str) -> Result<Self> {
         if s.len() != 36 {
@@ -332,7 +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
@@ -564,35 +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, serde::Deserialize, serde::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.
@@ -718,12 +743,17 @@ pub struct TimelineInfo {
     pub pg_version: u32,
 
     pub state: TimelineState,
-    pub is_archived: bool,
 
     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)]
@@ -868,7 +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)]
@@ -1653,21 +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

@@ -89,8 +89,19 @@ impl PageserverUtilization {
 
     /// 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 {
-        score >= Self::UTILIZATION_FULL
+        // 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) {

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.
     ///

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/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

@@ -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;
 
@@ -219,15 +222,20 @@ 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 +253,6 @@ 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

@@ -135,6 +135,8 @@ 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 +156,18 @@ 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
@@ -545,14 +547,14 @@ 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/src/lib.rs

@@ -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),
         }
     }

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>
@@ -300,7 +302,9 @@ 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/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/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

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/id.rs

@@ -249,8 +249,10 @@ 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 +1,280 @@
+//! 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::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 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 {
+            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 mut throttled = false;
+
+        let start = tokio::time::Instant::now();
+
+        // wait until we are the first in the queue
+        let mut notified = std::pin::pin!(self.queue.notified());
+        if !notified.as_mut().enable() {
+            throttled = true;
+            notified.await;
+        }
+
+        // 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 throttled,
+                Err(ready_at) => {
+                    throttled = true;
+                    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

@@ -71,6 +71,7 @@ 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/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/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`]

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/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,

pageserver/Cargo.toml

@@ -8,7 +8,7 @@ license.workspace = true
 default = []
 # Enables test-only APIs, incuding failpoints. In particular, enables the `fail_point!` macro,
 # which adds some runtime cost to run tests on outage conditions
-testing = ["fail/failpoints"]
+testing = ["fail/failpoints", "pageserver_api/testing" ]
 
 [dependencies]
 anyhow.workspace = true
@@ -37,7 +37,6 @@ humantime.workspace = true
 humantime-serde.workspace = true
 hyper.workspace = true
 itertools.workspace = true
-leaky-bucket.workspace = true
 md5.workspace = true
 nix.workspace = true
 # hack to get the number of worker threads tokio uses
@@ -102,6 +101,7 @@ procfs.workspace = true
 criterion.workspace = true
 hex-literal.workspace = true
 tokio = { workspace = true, features = ["process", "sync", "fs", "rt", "io-util", "time", "test-util"] }
+indoc.workspace = true
 
 [[bench]]
 name = "bench_layer_map"

pageserver/benches/bench_ingest.rs

@@ -4,7 +4,7 @@ use bytes::Bytes;
 use camino::Utf8PathBuf;
 use criterion::{criterion_group, criterion_main, Criterion};
 use pageserver::{
-    config::{defaults::DEFAULT_IO_BUFFER_ALIGNMENT, PageServerConf},
+    config::PageServerConf,
     context::{DownloadBehavior, RequestContext},
     l0_flush::{L0FlushConfig, L0FlushGlobalState},
     page_cache,
@@ -167,7 +167,7 @@ fn criterion_benchmark(c: &mut Criterion) {
     virtual_file::init(
         16384,
         virtual_file::io_engine_for_bench(),
-        DEFAULT_IO_BUFFER_ALIGNMENT,
+        pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
     );
     page_cache::init(conf.page_cache_size);
 

pageserver/client/Cargo.toml

@@ -7,7 +7,6 @@ license.workspace = true
 [dependencies]
 pageserver_api.workspace = true
 thiserror.workspace = true
-async-trait.workspace = true
 reqwest = { workspace = true, features = [ "stream" ] }
 utils.workspace = true
 serde.workspace = true

pageserver/client/src/lib.rs

@@ -1,2 +1,20 @@
 pub mod mgmt_api;
 pub mod page_service;
+
+/// For timeline_block_unblock_gc, distinguish the two different operations. This could be a bool.
+// If file structure is per-kind not per-feature then where to put this?
+#[derive(Clone, Copy)]
+pub enum BlockUnblock {
+    Block,
+    Unblock,
+}
+
+impl std::fmt::Display for BlockUnblock {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        let s = match self {
+            BlockUnblock::Block => "block",
+            BlockUnblock::Unblock => "unblock",
+        };
+        f.write_str(s)
+    }
+}

pageserver/client/src/mgmt_api.rs

@@ -12,6 +12,8 @@ use utils::{
 
 pub use reqwest::Body as ReqwestBody;
 
+use crate::BlockUnblock;
+
 pub mod util;
 
 #[derive(Debug, Clone)]
@@ -419,6 +421,24 @@ impl Client {
         }
     }
 
+    pub async fn timeline_archival_config(
+        &self,
+        tenant_shard_id: TenantShardId,
+        timeline_id: TimelineId,
+        req: &TimelineArchivalConfigRequest,
+    ) -> Result<()> {
+        let uri = format!(
+            "{}/v1/tenant/{tenant_shard_id}/timeline/{timeline_id}/archival_config",
+            self.mgmt_api_endpoint
+        );
+
+        self.request(Method::POST, &uri, req)
+            .await?
+            .json()
+            .await
+            .map_err(Error::ReceiveBody)
+    }
+
     pub async fn timeline_detach_ancestor(
         &self,
         tenant_shard_id: TenantShardId,
@@ -436,6 +456,20 @@ impl Client {
             .map_err(Error::ReceiveBody)
     }
 
+    pub async fn timeline_block_unblock_gc(
+        &self,
+        tenant_shard_id: TenantShardId,
+        timeline_id: TimelineId,
+        dir: BlockUnblock,
+    ) -> Result<()> {
+        let uri = format!(
+            "{}/v1/tenant/{tenant_shard_id}/timeline/{timeline_id}/{dir}_gc",
+            self.mgmt_api_endpoint,
+        );
+
+        self.request(Method::POST, &uri, ()).await.map(|_| ())
+    }
+
     pub async fn tenant_reset(&self, tenant_shard_id: TenantShardId) -> Result<()> {
         let uri = format!(
             "{}/v1/tenant/{}/reset",

pageserver/ctl/src/layer_map_analyzer.rs

@@ -4,7 +4,6 @@
 
 use anyhow::Result;
 use camino::{Utf8Path, Utf8PathBuf};
-use pageserver::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
 use pageserver::context::{DownloadBehavior, RequestContext};
 use pageserver::task_mgr::TaskKind;
 use pageserver::tenant::{TENANTS_SEGMENT_NAME, TIMELINES_SEGMENT_NAME};
@@ -80,16 +79,24 @@ pub(crate) fn parse_filename(name: &str) -> Option<LayerFile> {
         return None;
     }
     let keys: Vec<&str> = split[0].split('-').collect();
-    let mut lsns: Vec<&str> = split[1].split('-').collect();
-    let is_delta = if lsns.len() == 1 {
-        lsns.push(lsns[0]);
+    let lsn_and_opt_generation: Vec<&str> = split[1].split('v').collect();
+    let lsns: Vec<&str> = lsn_and_opt_generation[0].split('-').collect();
+    let the_lsns: [&str; 2];
+
+    /*
+     * Generations add a -vX-XXXXXX postfix, which causes issues when we try to
+     * parse 'vX' as an LSN.
+     */
+    let is_delta = if lsns.len() == 1 || lsns[1].is_empty() {
+        the_lsns = [lsns[0], lsns[0]];
         false
     } else {
+        the_lsns = [lsns[0], lsns[1]];
         true
     };
 
     let key_range = Key::from_hex(keys[0]).unwrap()..Key::from_hex(keys[1]).unwrap();
-    let lsn_range = Lsn::from_hex(lsns[0]).unwrap()..Lsn::from_hex(lsns[1]).unwrap();
+    let lsn_range = Lsn::from_hex(the_lsns[0]).unwrap()..Lsn::from_hex(the_lsns[1]).unwrap();
     let holes = Vec::new();
     Some(LayerFile {
         key_range,
@@ -148,7 +155,7 @@ pub(crate) async fn main(cmd: &AnalyzeLayerMapCmd) -> Result<()> {
     pageserver::virtual_file::init(
         10,
         virtual_file::api::IoEngineKind::StdFs,
-        DEFAULT_IO_BUFFER_ALIGNMENT,
+        pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
     );
     pageserver::page_cache::init(100);
 

pageserver/ctl/src/layers.rs

@@ -3,7 +3,6 @@ use std::path::{Path, PathBuf};
 use anyhow::Result;
 use camino::{Utf8Path, Utf8PathBuf};
 use clap::Subcommand;
-use pageserver::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
 use pageserver::context::{DownloadBehavior, RequestContext};
 use pageserver::task_mgr::TaskKind;
 use pageserver::tenant::block_io::BlockCursor;
@@ -90,6 +89,7 @@ async fn read_delta_file(path: impl AsRef<Path>, ctx: &RequestContext) -> Result
     for (k, v) in all {
         let value = cursor.read_blob(v.pos(), ctx).await?;
         println!("key:{} value_len:{}", k, value.len());
+        assert!(k.is_i128_representable(), "invalid key: ");
     }
     // TODO(chi): special handling for last key?
     Ok(())
@@ -193,7 +193,7 @@ pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
             pageserver::virtual_file::init(
                 10,
                 virtual_file::api::IoEngineKind::StdFs,
-                DEFAULT_IO_BUFFER_ALIGNMENT,
+                pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
             );
             pageserver::page_cache::init(100);
 

pageserver/ctl/src/main.rs

@@ -20,14 +20,13 @@ use clap::{Parser, Subcommand};
 use index_part::IndexPartCmd;
 use layers::LayerCmd;
 use pageserver::{
-    config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
     context::{DownloadBehavior, RequestContext},
     page_cache,
     task_mgr::TaskKind,
     tenant::{dump_layerfile_from_path, metadata::TimelineMetadata},
     virtual_file,
 };
-use pageserver_api::shard::TenantShardId;
+use pageserver_api::{config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT, shard::TenantShardId};
 use postgres_ffi::ControlFileData;
 use remote_storage::{RemotePath, RemoteStorageConfig};
 use tokio_util::sync::CancellationToken;
@@ -175,7 +174,7 @@ async fn main() -> anyhow::Result<()> {
                 println!("specified prefix '{}' failed validation", cmd.prefix);
                 return Ok(());
             };
-            let toml_document = toml_edit::Document::from_str(&cmd.config_toml_str)?;
+            let toml_document = toml_edit::DocumentMut::from_str(&cmd.config_toml_str)?;
             let toml_item = toml_document
                 .get("remote_storage")
                 .expect("need remote_storage");

pageserver/src/basebackup.rs

@@ -30,9 +30,8 @@ use pageserver_api::reltag::{RelTag, SlruKind};
 
 use postgres_ffi::dispatch_pgversion;
 use postgres_ffi::pg_constants::{DEFAULTTABLESPACE_OID, GLOBALTABLESPACE_OID};
-use postgres_ffi::pg_constants::{PGDATA_SPECIAL_FILES, PGDATA_SUBDIRS, PG_HBA};
+use postgres_ffi::pg_constants::{PGDATA_SPECIAL_FILES, PG_HBA};
 use postgres_ffi::relfile_utils::{INIT_FORKNUM, MAIN_FORKNUM};
-use postgres_ffi::TransactionId;
 use postgres_ffi::XLogFileName;
 use postgres_ffi::PG_TLI;
 use postgres_ffi::{BLCKSZ, RELSEG_SIZE, WAL_SEGMENT_SIZE};
@@ -255,8 +254,11 @@ where
 
         let lazy_slru_download = self.timeline.get_lazy_slru_download() && !self.full_backup;
 
+        let pgversion = self.timeline.pg_version;
+        let subdirs = dispatch_pgversion!(pgversion, &pgv::bindings::PGDATA_SUBDIRS[..]);
+
         // Create pgdata subdirs structure
-        for dir in PGDATA_SUBDIRS.iter() {
+        for dir in subdirs.iter() {
             let header = new_tar_header_dir(dir)?;
             self.ar
                 .append(&header, &mut io::empty())
@@ -606,7 +608,7 @@ where
     //
     // Extract twophase state files
     //
-    async fn add_twophase_file(&mut self, xid: TransactionId) -> Result<(), BasebackupError> {
+    async fn add_twophase_file(&mut self, xid: u64) -> Result<(), BasebackupError> {
         let img = self
             .timeline
             .get_twophase_file(xid, self.lsn, self.ctx)
@@ -617,7 +619,11 @@ where
         buf.extend_from_slice(&img[..]);
         let crc = crc32c::crc32c(&img[..]);
         buf.put_u32_le(crc);
-        let path = format!("pg_twophase/{:>08X}", xid);
+        let path = if self.timeline.pg_version < 17 {
+            format!("pg_twophase/{:>08X}", xid)
+        } else {
+            format!("pg_twophase/{:>016X}", xid)
+        };
         let header = new_tar_header(&path, buf.len() as u64)?;
         self.ar
             .append(&header, &buf[..])

pageserver/src/bin/pageserver.rs

@@ -5,6 +5,7 @@
 use std::env;
 use std::env::{var, VarError};
 use std::io::Read;
+use std::str::FromStr;
 use std::sync::Arc;
 use std::time::Duration;
 
@@ -36,6 +37,7 @@ use pageserver::{
     virtual_file,
 };
 use postgres_backend::AuthType;
+use utils::crashsafe::syncfs;
 use utils::failpoint_support;
 use utils::logging::TracingErrorLayerEnablement;
 use utils::{
@@ -124,7 +126,6 @@ fn main() -> anyhow::Result<()> {
     // after setting up logging, log the effective IO engine choice and read path implementations
     info!(?conf.virtual_file_io_engine, "starting with virtual_file IO engine");
     info!(?conf.virtual_file_direct_io, "starting with virtual_file Direct IO settings");
-    info!(?conf.compact_level0_phase1_value_access, "starting with setting for compact_level0_phase1_value_access");
     info!(?conf.io_buffer_alignment, "starting with setting for IO buffer alignment");
 
     // The tenants directory contains all the pageserver local disk state.
@@ -155,23 +156,7 @@ fn main() -> anyhow::Result<()> {
         };
 
         let started = Instant::now();
-        // Linux guarantees durability for syncfs.
-        // POSIX doesn't have syncfs, and further does not actually guarantee durability of sync().
-        #[cfg(target_os = "linux")]
-        {
-            use std::os::fd::AsRawFd;
-            nix::unistd::syncfs(dirfd.as_raw_fd()).context("syncfs")?;
-        }
-        #[cfg(target_os = "macos")]
-        {
-            // macOS is not a production platform for Neon, don't even bother.
-            drop(dirfd);
-        }
-        #[cfg(not(any(target_os = "linux", target_os = "macos")))]
-        {
-            compile_error!("Unsupported OS");
-        }
-
+        syncfs(dirfd)?;
         let elapsed = started.elapsed();
         info!(
             elapsed_ms = elapsed.as_millis(),
@@ -223,27 +208,15 @@ fn initialize_config(
         }
     };
 
-    let config: toml_edit::Document = match std::fs::File::open(cfg_file_path) {
-        Ok(mut f) => {
-            let md = f.metadata().context("stat config file")?;
-            if md.is_file() {
-                let mut s = String::new();
-                f.read_to_string(&mut s).context("read config file")?;
-                s.parse().context("parse config file toml")?
-            } else {
-                anyhow::bail!("directory entry exists but is not a file: {cfg_file_path}");
-            }
-        }
-        Err(e) => {
-            anyhow::bail!("open pageserver config: {e}: {cfg_file_path}");
-        }
-    };
-
-    debug!("Using pageserver toml: {config}");
-
-    // Construct the runtime representation
-    let conf = PageServerConf::parse_and_validate(identity.id, &config, workdir)
-        .context("Failed to parse pageserver configuration")?;
+    let config_file_contents =
+        std::fs::read_to_string(cfg_file_path).context("read config file from filesystem")?;
+    let config_toml = serde_path_to_error::deserialize(
+        toml_edit::de::Deserializer::from_str(&config_file_contents)
+            .context("build toml deserializer")?,
+    )
+    .context("deserialize config toml")?;
+    let conf = PageServerConf::parse_and_validate(identity.id, config_toml, workdir)
+        .context("runtime-validation of config toml")?;
 
     Ok(Box::leak(Box::new(conf)))
 }

pageserver/src/consumption_metrics.rs

@@ -1,6 +1,8 @@
 //! Periodically collect consumption metrics for all active tenants
 //! and push them to a HTTP endpoint.
 use crate::config::PageServerConf;
+use crate::consumption_metrics::metrics::MetricsKey;
+use crate::consumption_metrics::upload::KeyGen as _;
 use crate::context::{DownloadBehavior, RequestContext};
 use crate::task_mgr::{self, TaskKind, BACKGROUND_RUNTIME};
 use crate::tenant::size::CalculateSyntheticSizeError;
@@ -8,6 +10,7 @@ use crate::tenant::tasks::BackgroundLoopKind;
 use crate::tenant::{mgr::TenantManager, LogicalSizeCalculationCause, Tenant};
 use camino::Utf8PathBuf;
 use consumption_metrics::EventType;
+use itertools::Itertools as _;
 use pageserver_api::models::TenantState;
 use remote_storage::{GenericRemoteStorage, RemoteStorageConfig};
 use reqwest::Url;
@@ -19,9 +22,8 @@ use tokio_util::sync::CancellationToken;
 use tracing::*;
 use utils::id::NodeId;
 
-mod metrics;
-use crate::consumption_metrics::metrics::MetricsKey;
 mod disk_cache;
+mod metrics;
 mod upload;
 
 const DEFAULT_HTTP_REPORTING_TIMEOUT: Duration = Duration::from_secs(60);
@@ -143,6 +145,12 @@ async fn collect_metrics(
         // these are point in time, with variable "now"
         let metrics = metrics::collect_all_metrics(&tenant_manager, &cached_metrics, &ctx).await;
 
+        // Pre-generate event idempotency keys, to reuse them across the bucket
+        // and HTTP sinks.
+        let idempotency_keys = std::iter::repeat_with(|| node_id.as_str().generate())
+            .take(metrics.len())
+            .collect_vec();
+
         let metrics = Arc::new(metrics);
 
         // why not race cancellation here? because we are one of the last tasks, and if we are
@@ -161,8 +169,14 @@ async fn collect_metrics(
             }
 
             if let Some(bucket_client) = &bucket_client {
-                let res =
-                    upload::upload_metrics_bucket(bucket_client, &cancel, &node_id, &metrics).await;
+                let res = upload::upload_metrics_bucket(
+                    bucket_client,
+                    &cancel,
+                    &node_id,
+                    &metrics,
+                    &idempotency_keys,
+                )
+                .await;
                 if let Err(e) = res {
                     tracing::error!("failed to upload to S3: {e:#}");
                 }
@@ -174,9 +188,9 @@ async fn collect_metrics(
                 &client,
                 metric_collection_endpoint,
                 &cancel,
-                &node_id,
                 &metrics,
                 &mut cached_metrics,
+                &idempotency_keys,
             )
             .await;
             if let Err(e) = res {

pageserver/src/consumption_metrics/upload.rs

@@ -24,16 +24,16 @@ pub(super) async fn upload_metrics_http(
     client: &reqwest::Client,
     metric_collection_endpoint: &reqwest::Url,
     cancel: &CancellationToken,
-    node_id: &str,
     metrics: &[RawMetric],
     cached_metrics: &mut Cache,
+    idempotency_keys: &[IdempotencyKey<'_>],
 ) -> anyhow::Result<()> {
     let mut uploaded = 0;
     let mut failed = 0;
 
     let started_at = std::time::Instant::now();
 
-    let mut iter = serialize_in_chunks(CHUNK_SIZE, metrics, node_id);
+    let mut iter = serialize_in_chunks(CHUNK_SIZE, metrics, idempotency_keys);
 
     while let Some(res) = iter.next() {
         let (chunk, body) = res?;
@@ -87,6 +87,7 @@ pub(super) async fn upload_metrics_bucket(
     cancel: &CancellationToken,
     node_id: &str,
     metrics: &[RawMetric],
+    idempotency_keys: &[IdempotencyKey<'_>],
 ) -> anyhow::Result<()> {
     if metrics.is_empty() {
         // Skip uploads if we have no metrics, so that readers don't have to handle the edge case
@@ -106,7 +107,7 @@ pub(super) async fn upload_metrics_bucket(
 
     // Serialize and write into compressed buffer
     let started_at = std::time::Instant::now();
-    for res in serialize_in_chunks(CHUNK_SIZE, metrics, node_id) {
+    for res in serialize_in_chunks(CHUNK_SIZE, metrics, idempotency_keys) {
         let (_chunk, body) = res?;
         gzip_writer.write_all(&body).await?;
     }
@@ -134,29 +135,31 @@ pub(super) async fn upload_metrics_bucket(
     Ok(())
 }
 
-// The return type is quite ugly, but we gain testability in isolation
-fn serialize_in_chunks<'a, F>(
+/// Serializes the input metrics as JSON in chunks of chunk_size. The provided
+/// idempotency keys are injected into the corresponding metric events (reused
+/// across different metrics sinks), and must have the same length as input.
+fn serialize_in_chunks<'a>(
     chunk_size: usize,
     input: &'a [RawMetric],
-    factory: F,
+    idempotency_keys: &'a [IdempotencyKey<'a>],
 ) -> impl ExactSizeIterator<Item = Result<(&'a [RawMetric], bytes::Bytes), serde_json::Error>> + 'a
-where
-    F: KeyGen<'a> + 'a,
 {
     use bytes::BufMut;
 
-    struct Iter<'a, F> {
+    assert_eq!(input.len(), idempotency_keys.len());
+
+    struct Iter<'a> {
         inner: std::slice::Chunks<'a, RawMetric>,
+        idempotency_keys: std::slice::Iter<'a, IdempotencyKey<'a>>,
         chunk_size: usize,
 
         // write to a BytesMut so that we can cheaply clone the frozen Bytes for retries
         buffer: bytes::BytesMut,
         // chunk amount of events are reused to produce the serialized document
         scratch: Vec<Event<Ids, Name>>,
-        factory: F,
     }
 
-    impl<'a, F: KeyGen<'a>> Iterator for Iter<'a, F> {
+    impl<'a> Iterator for Iter<'a> {
         type Item = Result<(&'a [RawMetric], bytes::Bytes), serde_json::Error>;
 
         fn next(&mut self) -> Option<Self::Item> {
@@ -167,17 +170,14 @@ where
                 self.scratch.extend(
                     chunk
                         .iter()
-                        .map(|raw_metric| raw_metric.as_event(&self.factory.generate())),
+                        .zip(&mut self.idempotency_keys)
+                        .map(|(raw_metric, key)| raw_metric.as_event(key)),
                 );
             } else {
                 // next rounds: update_in_place to reuse allocations
                 assert_eq!(self.scratch.len(), self.chunk_size);
-                self.scratch
-                    .iter_mut()
-                    .zip(chunk.iter())
-                    .for_each(|(slot, raw_metric)| {
-                        raw_metric.update_in_place(slot, &self.factory.generate())
-                    });
+                itertools::izip!(self.scratch.iter_mut(), chunk, &mut self.idempotency_keys)
+                    .for_each(|(slot, raw_metric, key)| raw_metric.update_in_place(slot, key));
             }
 
             let res = serde_json::to_writer(
@@ -198,18 +198,19 @@ where
         }
     }
 
-    impl<'a, F: KeyGen<'a>> ExactSizeIterator for Iter<'a, F> {}
+    impl<'a> ExactSizeIterator for Iter<'a> {}
 
     let buffer = bytes::BytesMut::new();
     let inner = input.chunks(chunk_size);
+    let idempotency_keys = idempotency_keys.iter();
     let scratch = Vec::new();
 
     Iter {
         inner,
+        idempotency_keys,
         chunk_size,
         buffer,
         scratch,
-        factory,
     }
 }
 
@@ -268,7 +269,7 @@ impl RawMetricExt for RawMetric {
     }
 }
 
-trait KeyGen<'a>: Copy {
+pub(crate) trait KeyGen<'a> {
     fn generate(&self) -> IdempotencyKey<'a>;
 }
 
@@ -389,7 +390,10 @@ mod tests {
         let examples = metric_samples();
         assert!(examples.len() > 1);
 
-        let factory = FixedGen::new(Utc::now(), "1", 42);
+        let now = Utc::now();
+        let idempotency_keys = (0..examples.len())
+            .map(|i| FixedGen::new(now, "1", i as u16).generate())
+            .collect::<Vec<_>>();
 
         // need to use Event here because serde_json::Value uses default hashmap, not linked
         // hashmap
@@ -398,13 +402,13 @@ mod tests {
             events: Vec<Event<Ids, Name>>,
         }
 
-        let correct = serialize_in_chunks(examples.len(), &examples, factory)
+        let correct = serialize_in_chunks(examples.len(), &examples, &idempotency_keys)
             .map(|res| res.unwrap().1)
             .flat_map(|body| serde_json::from_slice::<EventChunk>(&body).unwrap().events)
             .collect::<Vec<_>>();
 
         for chunk_size in 1..examples.len() {
-            let actual = serialize_in_chunks(chunk_size, &examples, factory)
+            let actual = serialize_in_chunks(chunk_size, &examples, &idempotency_keys)
                 .map(|res| res.unwrap().1)
                 .flat_map(|body| serde_json::from_slice::<EventChunk>(&body).unwrap().events)
                 .collect::<Vec<_>>();

pageserver/src/context.rs

@@ -1,7 +1,9 @@
-//! This module defines `RequestContext`, a structure that we use throughout
-//! the pageserver to propagate high-level context from places
-//! that _originate_ activity down to the shared code paths at the
-//! heart of the pageserver. It's inspired by Golang's `context.Context`.
+//! Defines [`RequestContext`].
+//!
+//! It is a structure that we use throughout the pageserver to propagate
+//! high-level context from places that _originate_ activity down to the
+//! shared code paths at the heart of the pageserver. It's inspired by
+//! Golang's `context.Context`.
 //!
 //! For example, in `Timeline::get(page_nr, lsn)` we need to answer the following questions:
 //! 1. What high-level activity ([`TaskKind`]) needs this page?
@@ -105,8 +107,10 @@ pub struct RequestContext {
 #[derive(Clone, Copy, PartialEq, Eq, Debug, enum_map::Enum, strum_macros::IntoStaticStr)]
 pub enum PageContentKind {
     Unknown,
+    DeltaLayerSummary,
     DeltaLayerBtreeNode,
     DeltaLayerValue,
+    ImageLayerSummary,
     ImageLayerBtreeNode,
     ImageLayerValue,
     InMemoryLayer,

pageserver/src/control_plane_client.rs

@@ -141,10 +141,24 @@ impl ControlPlaneGenerationsApi for ControlPlaneClient {
                         m.other
                     );
 
-                    let az_id = m
-                        .other
-                        .get("availability_zone_id")
-                        .and_then(|jv| jv.as_str().map(|str| str.to_owned()));
+                    let az_id = {
+                        let az_id_from_metadata = m
+                            .other
+                            .get("availability_zone_id")
+                            .and_then(|jv| jv.as_str().map(|str| str.to_owned()));
+
+                        match az_id_from_metadata {
+                            Some(az_id) => Some(az_id),
+                            None => {
+                                tracing::warn!("metadata.json does not contain an 'availability_zone_id' field");
+                                conf.availability_zone.clone()
+                            }
+                        }
+                    };
+
+                    if az_id.is_none() {
+                        panic!("Availablity zone id could not be inferred from metadata.json or pageserver config");
+                    }
 
                     Some(NodeRegisterRequest {
                         node_id: conf.id,
@@ -152,7 +166,7 @@ impl ControlPlaneGenerationsApi for ControlPlaneClient {
                         listen_pg_port: m.postgres_port,
                         listen_http_addr: m.http_host,
                         listen_http_port: m.http_port,
-                        availability_zone_id: az_id,
+                        availability_zone_id: az_id.expect("Checked above"),
                     })
                 }
                 Err(e) => {

pageserver/src/disk_usage_eviction_task.rs

@@ -41,19 +41,15 @@
 // - The `#[allow(dead_code)]` above various structs are to suppress warnings about only the Debug impl
 //   reading these fields. We use the Debug impl for semi-structured logging, though.
 
-use std::{
-    sync::Arc,
-    time::{Duration, SystemTime},
-};
+use std::{sync::Arc, time::SystemTime};
 
 use anyhow::Context;
-use pageserver_api::shard::TenantShardId;
+use pageserver_api::{config::DiskUsageEvictionTaskConfig, shard::TenantShardId};
 use remote_storage::GenericRemoteStorage;
-use serde::{Deserialize, Serialize};
+use serde::Serialize;
 use tokio::time::Instant;
 use tokio_util::sync::CancellationToken;
 use tracing::{debug, error, info, instrument, warn, Instrument};
-use utils::serde_percent::Percent;
 use utils::{completion, id::TimelineId};
 
 use crate::{
@@ -69,23 +65,9 @@ use crate::{
     CancellableTask, DiskUsageEvictionTask,
 };
 
-#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
-pub struct DiskUsageEvictionTaskConfig {
-    pub max_usage_pct: Percent,
-    pub min_avail_bytes: u64,
-    #[serde(with = "humantime_serde")]
-    pub period: Duration,
-    #[cfg(feature = "testing")]
-    pub mock_statvfs: Option<crate::statvfs::mock::Behavior>,
-    /// Select sorting for evicted layers
-    #[serde(default)]
-    pub eviction_order: EvictionOrder,
-}
-
 /// Selects the sort order for eviction candidates *after* per tenant `min_resident_size`
 /// partitioning.
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
-#[serde(tag = "type", content = "args")]
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
 pub enum EvictionOrder {
     /// Order the layers to be evicted by how recently they have been accessed relatively within
     /// the set of resident layers of a tenant.
@@ -96,23 +78,22 @@ pub enum EvictionOrder {
         /// we read tenants is deterministic. If we find the need to use this as `false`, we need
         /// to ensure nondeterminism by adding in a random number to break the
         /// `relative_last_activity==0.0` ties.
-        #[serde(default = "default_highest_layer_count_loses_first")]
         highest_layer_count_loses_first: bool,
     },
 }
 
-impl Default for EvictionOrder {
-    fn default() -> Self {
-        Self::RelativeAccessed {
-            highest_layer_count_loses_first: true,
+impl From<pageserver_api::config::EvictionOrder> for EvictionOrder {
+    fn from(value: pageserver_api::config::EvictionOrder) -> Self {
+        match value {
+            pageserver_api::config::EvictionOrder::RelativeAccessed {
+                highest_layer_count_loses_first,
+            } => Self::RelativeAccessed {
+                highest_layer_count_loses_first,
+            },
         }
     }
 }
 
-fn default_highest_layer_count_loses_first() -> bool {
-    true
-}
-
 impl EvictionOrder {
     fn sort(&self, candidates: &mut [(EvictionPartition, EvictionCandidate)]) {
         use EvictionOrder::*;
@@ -295,7 +276,7 @@ async fn disk_usage_eviction_task_iteration(
         storage,
         usage_pre,
         tenant_manager,
-        task_config.eviction_order,
+        task_config.eviction_order.into(),
         cancel,
     )
     .await;
@@ -1257,7 +1238,6 @@ mod filesystem_level_usage {
 
     #[test]
     fn max_usage_pct_pressure() {
-        use super::EvictionOrder;
         use super::Usage as _;
         use std::time::Duration;
         use utils::serde_percent::Percent;
@@ -1269,7 +1249,7 @@ mod filesystem_level_usage {
                 period: Duration::MAX,
                 #[cfg(feature = "testing")]
                 mock_statvfs: None,
-                eviction_order: EvictionOrder::default(),
+                eviction_order: pageserver_api::config::EvictionOrder::default(),
             },
             total_bytes: 100_000,
             avail_bytes: 0,

pageserver/src/http/routes.rs

@@ -324,6 +324,9 @@ impl From<crate::tenant::TimelineArchivalError> for ApiError {
         match value {
             NotFound => ApiError::NotFound(anyhow::anyhow!("timeline not found").into()),
             Timeout => ApiError::Timeout("hit pageserver internal timeout".into()),
+            e @ HasArchivedParent(_) => {
+                ApiError::PreconditionFailed(e.to_string().into_boxed_str())
+            }
             HasUnarchivedChildren(children) => ApiError::PreconditionFailed(
                 format!(
                     "Cannot archive timeline which has non-archived child timelines: {children:?}"
@@ -465,7 +468,7 @@ async fn build_timeline_info_common(
         pg_version: timeline.pg_version,
 
         state,
-        is_archived,
+        is_archived: Some(is_archived),
 
         walreceiver_status,
 
@@ -871,7 +874,10 @@ async fn get_timestamp_of_lsn_handler(
 
     match result {
         Some(time) => {
-            let time = format_rfc3339(postgres_ffi::from_pg_timestamp(time)).to_string();
+            let time = format_rfc3339(
+                postgres_ffi::try_from_pg_timestamp(time).map_err(ApiError::InternalServerError)?,
+            )
+            .to_string();
             json_response(StatusCode::OK, time)
         }
         None => Err(ApiError::NotFound(
@@ -1727,6 +1733,10 @@ async fn timeline_compact_handler(
     if Some(true) == parse_query_param::<_, bool>(&request, "enhanced_gc_bottom_most_compaction")? {
         flags |= CompactFlags::EnhancedGcBottomMostCompaction;
     }
+    if Some(true) == parse_query_param::<_, bool>(&request, "dry_run")? {
+        flags |= CompactFlags::DryRun;
+    }
+
     let wait_until_uploaded =
         parse_query_param::<_, bool>(&request, "wait_until_uploaded")?.unwrap_or(false);
 
@@ -2066,7 +2076,7 @@ async fn disk_usage_eviction_run(
         evict_bytes: u64,
 
         #[serde(default)]
-        eviction_order: crate::disk_usage_eviction_task::EvictionOrder,
+        eviction_order: pageserver_api::config::EvictionOrder,
     }
 
     #[derive(Debug, Clone, Copy, serde::Serialize)]
@@ -2102,7 +2112,7 @@ async fn disk_usage_eviction_run(
         &state.remote_storage,
         usage,
         &state.tenant_manager,
-        config.eviction_order,
+        config.eviction_order.into(),
         &cancel,
     )
     .await;

pageserver/src/import_datadir.rs

@@ -19,6 +19,7 @@ use crate::metrics::WAL_INGEST;
 use crate::pgdatadir_mapping::*;
 use crate::tenant::Timeline;
 use crate::walingest::WalIngest;
+use crate::walrecord::decode_wal_record;
 use crate::walrecord::DecodedWALRecord;
 use pageserver_api::reltag::{RelTag, SlruKind};
 use postgres_ffi::pg_constants;
@@ -310,11 +311,13 @@ async fn import_wal(
 
         let mut nrecords = 0;
         let mut modification = tline.begin_modification(last_lsn);
-        let mut decoded = DecodedWALRecord::default();
         while last_lsn <= endpoint {
             if let Some((lsn, recdata)) = waldecoder.poll_decode()? {
+                let mut decoded = DecodedWALRecord::default();
+                decode_wal_record(recdata, &mut decoded, tline.pg_version)?;
+
                 walingest
-                    .ingest_record(recdata, lsn, &mut modification, &mut decoded, ctx)
+                    .ingest_record(decoded, lsn, &mut modification, ctx)
                     .await?;
                 WAL_INGEST.records_committed.inc();
 
@@ -449,11 +452,12 @@ pub async fn import_wal_from_tar(
         waldecoder.feed_bytes(&bytes[offset..]);
 
         let mut modification = tline.begin_modification(last_lsn);
-        let mut decoded = DecodedWALRecord::default();
         while last_lsn <= end_lsn {
             if let Some((lsn, recdata)) = waldecoder.poll_decode()? {
+                let mut decoded = DecodedWALRecord::default();
+                decode_wal_record(recdata, &mut decoded, tline.pg_version)?;
                 walingest
-                    .ingest_record(recdata, lsn, &mut modification, &mut decoded, ctx)
+                    .ingest_record(decoded, lsn, &mut modification, ctx)
                     .await?;
                 modification.commit(ctx).await?;
                 last_lsn = lsn;
@@ -576,9 +580,11 @@ async fn import_file(
         import_slru(modification, slru, file_path, reader, len, ctx).await?;
         debug!("imported multixact members slru");
     } else if file_path.starts_with("pg_twophase") {
-        let xid = u32::from_str_radix(file_name.as_ref(), 16)?;
-
         let bytes = read_all_bytes(reader).await?;
+
+        // In PostgreSQL v17, this is a 64-bit FullTransactionid. In previous versions,
+        // it's a 32-bit TransactionId, which fits in u64 anyway.
+        let xid = u64::from_str_radix(file_name.as_ref(), 16)?;
         modification
             .put_twophase_file(xid, Bytes::copy_from_slice(&bytes[..]), ctx)
             .await?;

pageserver/src/l0_flush.rs

@@ -1,9 +1,7 @@
 use std::{num::NonZeroUsize, sync::Arc};
 
-#[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize)]
-#[serde(tag = "mode", rename_all = "kebab-case", deny_unknown_fields)]
+#[derive(Debug, PartialEq, Eq, Clone)]
 pub enum L0FlushConfig {
-    #[serde(rename_all = "snake_case")]
     Direct { max_concurrency: NonZeroUsize },
 }
 
@@ -16,6 +14,16 @@ impl Default for L0FlushConfig {
     }
 }
 
+impl From<pageserver_api::models::L0FlushConfig> for L0FlushConfig {
+    fn from(config: pageserver_api::models::L0FlushConfig) -> Self {
+        match config {
+            pageserver_api::models::L0FlushConfig::Direct { max_concurrency } => {
+                Self::Direct { max_concurrency }
+            }
+        }
+    }
+}
+
 #[derive(Clone)]
 pub struct L0FlushGlobalState(Arc<Inner>);
 

pageserver/src/metrics.rs

@@ -9,7 +9,7 @@ use metrics::{
 use once_cell::sync::Lazy;
 use pageserver_api::shard::TenantShardId;
 use strum::{EnumCount, VariantNames};
-use strum_macros::{EnumVariantNames, IntoStaticStr};
+use strum_macros::{IntoStaticStr, VariantNames};
 use tracing::warn;
 use utils::id::TimelineId;
 
@@ -27,7 +27,7 @@ const CRITICAL_OP_BUCKETS: &[f64] = &[
 ];
 
 // Metrics collected on operations on the storage repository.
-#[derive(Debug, EnumVariantNames, IntoStaticStr)]
+#[derive(Debug, VariantNames, IntoStaticStr)]
 #[strum(serialize_all = "kebab_case")]
 pub(crate) enum StorageTimeOperation {
     #[strum(serialize = "layer flush")]

pageserver/src/page_service.rs

@@ -1199,7 +1199,6 @@ impl PageServerHandler {
     }
 }
 
-#[async_trait::async_trait]
 impl<IO> postgres_backend::Handler<IO> for PageServerHandler
 where
     IO: AsyncRead + AsyncWrite + Send + Sync + Unpin,

pageserver/src/pgdatadir_mapping.rs

@@ -168,7 +168,9 @@ impl Timeline {
         DatadirModification {
             tline: self,
             pending_lsns: Vec::new(),
-            pending_updates: HashMap::new(),
+            pending_metadata_pages: HashMap::new(),
+            pending_data_pages: Vec::new(),
+            pending_zero_data_pages: Default::default(),
             pending_deletions: Vec::new(),
             pending_nblocks: 0,
             pending_directory_entries: Vec::new(),
@@ -631,7 +633,7 @@ impl Timeline {
 
     pub(crate) async fn get_twophase_file(
         &self,
-        xid: TransactionId,
+        xid: u64,
         lsn: Lsn,
         ctx: &RequestContext,
     ) -> Result<Bytes, PageReconstructError> {
@@ -644,11 +646,19 @@ impl Timeline {
         &self,
         lsn: Lsn,
         ctx: &RequestContext,
-    ) -> Result<HashSet<TransactionId>, PageReconstructError> {
+    ) -> Result<HashSet<u64>, PageReconstructError> {
         // fetch directory entry
         let buf = self.get(TWOPHASEDIR_KEY, lsn, ctx).await?;
 
-        Ok(TwoPhaseDirectory::des(&buf)?.xids)
+        if self.pg_version >= 17 {
+            Ok(TwoPhaseDirectoryV17::des(&buf)?.xids)
+        } else {
+            Ok(TwoPhaseDirectory::des(&buf)?
+                .xids
+                .iter()
+                .map(|x| u64::from(*x))
+                .collect())
+        }
     }
 
     pub(crate) async fn get_control_file(
@@ -727,8 +737,12 @@ impl Timeline {
         let current_policy = self.last_aux_file_policy.load();
         match current_policy {
             Some(AuxFilePolicy::V1) => {
-                warn!("this timeline is using deprecated aux file policy V1 (policy=V1)");
-                self.list_aux_files_v1(lsn, ctx).await
+                let res = self.list_aux_files_v1(lsn, ctx).await?;
+                let empty_str = if res.is_empty() { ", empty" } else { "" };
+                warn!(
+                    "this timeline is using deprecated aux file policy V1 (policy=v1{empty_str})"
+                );
+                Ok(res)
             }
             None => {
                 let res = self.list_aux_files_v1(lsn, ctx).await?;
@@ -896,9 +910,13 @@ impl Timeline {
 
         // Then pg_twophase
         result.add_key(TWOPHASEDIR_KEY);
-        let buf = self.get(TWOPHASEDIR_KEY, lsn, ctx).await?;
-        let twophase_dir = TwoPhaseDirectory::des(&buf)?;
-        let mut xids: Vec<TransactionId> = twophase_dir.xids.iter().cloned().collect();
+
+        let mut xids: Vec<u64> = self
+            .list_twophase_files(lsn, ctx)
+            .await?
+            .iter()
+            .cloned()
+            .collect();
         xids.sort_unstable();
         for xid in xids {
             result.add_key(twophase_file_key(xid));
@@ -1015,9 +1033,10 @@ impl Timeline {
 }
 
 /// DatadirModification represents an operation to ingest an atomic set of
-/// updates to the repository. It is created by the 'begin_record'
-/// function. It is called for each WAL record, so that all the modifications
-/// by a one WAL record appear atomic.
+/// updates to the repository.
+///
+/// It is created by the 'begin_record' function. It is called for each WAL
+/// record, so that all the modifications by a one WAL record appear atomic.
 pub struct DatadirModification<'a> {
     /// The timeline this modification applies to. You can access this to
     /// read the state, but note that any pending updates are *not* reflected
@@ -1031,10 +1050,24 @@ pub struct DatadirModification<'a> {
     // The put-functions add the modifications here, and they are flushed to the
     // underlying key-value store by the 'finish' function.
     pending_lsns: Vec<Lsn>,
-    pending_updates: HashMap<Key, Vec<(Lsn, usize, Value)>>,
     pending_deletions: Vec<(Range<Key>, Lsn)>,
     pending_nblocks: i64,
 
+    /// Metadata writes, indexed by key so that they can be read from not-yet-committed modifications
+    /// while ingesting subsequent records. See [`Self::is_data_key`] for the definition of 'metadata'.
+    pending_metadata_pages: HashMap<CompactKey, Vec<(Lsn, usize, Value)>>,
+
+    /// Data writes, ready to be flushed into an ephemeral layer. See [`Self::is_data_key`] for
+    /// which keys are stored here.
+    pending_data_pages: Vec<(CompactKey, Lsn, usize, Value)>,
+
+    // Sometimes during ingest, for example when extending a relation, we would like to write a zero page.  However,
+    // if we encounter a write from postgres in the same wal record, we will drop this entry.
+    //
+    // Unlike other 'pending' fields, this does not last until the next call to commit(): it is flushed
+    // at the end of each wal record, and all these writes implicitly are at lsn Self::lsn
+    pending_zero_data_pages: HashSet<CompactKey>,
+
     /// For special "directory" keys that store key-value maps, track the size of the map
     /// if it was updated in this modification.
     pending_directory_entries: Vec<(DirectoryKind, usize)>,
@@ -1058,6 +1091,10 @@ impl<'a> DatadirModification<'a> {
         self.pending_bytes
     }
 
+    pub(crate) fn has_dirty_data_pages(&self) -> bool {
+        (!self.pending_data_pages.is_empty()) || (!self.pending_zero_data_pages.is_empty())
+    }
+
     /// Set the current lsn
     pub(crate) fn set_lsn(&mut self, lsn: Lsn) -> anyhow::Result<()> {
         ensure!(
@@ -1066,6 +1103,10 @@ impl<'a> DatadirModification<'a> {
             lsn,
             self.lsn
         );
+
+        // If we are advancing LSN, then state from previous wal record should have been flushed.
+        assert!(self.pending_zero_data_pages.is_empty());
+
         if lsn > self.lsn {
             self.pending_lsns.push(self.lsn);
             self.lsn = lsn;
@@ -1073,6 +1114,17 @@ impl<'a> DatadirModification<'a> {
         Ok(())
     }
 
+    /// In this context, 'metadata' means keys that are only read by the pageserver internally, and 'data' means
+    /// keys that represent literal blocks that postgres can read.  So data includes relation blocks and
+    /// SLRU blocks, which are read directly by postgres, and everything else is considered metadata.
+    ///
+    /// The distinction is important because data keys are handled on a fast path where dirty writes are
+    /// not readable until this modification is committed, whereas metadata keys are visible for read
+    /// via [`Self::get`] as soon as their record has been ingested.
+    fn is_data_key(key: &Key) -> bool {
+        key.is_rel_block_key() || key.is_slru_block_key()
+    }
+
     /// Initialize a completely new repository.
     ///
     /// This inserts the directory metadata entries that are assumed to
@@ -1087,9 +1139,15 @@ impl<'a> DatadirModification<'a> {
         // Create AuxFilesDirectory
         self.init_aux_dir()?;
 
-        let buf = TwoPhaseDirectory::ser(&TwoPhaseDirectory {
-            xids: HashSet::new(),
-        })?;
+        let buf = if self.tline.pg_version >= 17 {
+            TwoPhaseDirectoryV17::ser(&TwoPhaseDirectoryV17 {
+                xids: HashSet::new(),
+            })
+        } else {
+            TwoPhaseDirectory::ser(&TwoPhaseDirectory {
+                xids: HashSet::new(),
+            })
+        }?;
         self.pending_directory_entries
             .push((DirectoryKind::TwoPhase, 0));
         self.put(TWOPHASEDIR_KEY, Value::Image(buf.into()));
@@ -1165,6 +1223,13 @@ impl<'a> DatadirModification<'a> {
         img: Bytes,
     ) -> anyhow::Result<()> {
         anyhow::ensure!(rel.relnode != 0, RelationError::InvalidRelnode);
+        let key = rel_block_to_key(rel, blknum);
+        if !key.is_valid_key_on_write_path() {
+            anyhow::bail!(
+                "the request contains data not supported by pageserver at {}",
+                key
+            );
+        }
         self.put(rel_block_to_key(rel, blknum), Value::Image(img));
         Ok(())
     }
@@ -1176,10 +1241,63 @@ impl<'a> DatadirModification<'a> {
         blknum: BlockNumber,
         img: Bytes,
     ) -> anyhow::Result<()> {
-        self.put(slru_block_to_key(kind, segno, blknum), Value::Image(img));
+        let key = slru_block_to_key(kind, segno, blknum);
+        if !key.is_valid_key_on_write_path() {
+            anyhow::bail!(
+                "the request contains data not supported by pageserver at {}",
+                key
+            );
+        }
+        self.put(key, Value::Image(img));
         Ok(())
     }
 
+    pub(crate) fn put_rel_page_image_zero(
+        &mut self,
+        rel: RelTag,
+        blknum: BlockNumber,
+    ) -> anyhow::Result<()> {
+        anyhow::ensure!(rel.relnode != 0, RelationError::InvalidRelnode);
+        let key = rel_block_to_key(rel, blknum);
+        if !key.is_valid_key_on_write_path() {
+            anyhow::bail!(
+                "the request contains data not supported by pageserver: {} @ {}",
+                key,
+                self.lsn
+            );
+        }
+        self.pending_zero_data_pages.insert(key.to_compact());
+        self.pending_bytes += ZERO_PAGE.len();
+        Ok(())
+    }
+
+    pub(crate) fn put_slru_page_image_zero(
+        &mut self,
+        kind: SlruKind,
+        segno: u32,
+        blknum: BlockNumber,
+    ) -> anyhow::Result<()> {
+        let key = slru_block_to_key(kind, segno, blknum);
+        if !key.is_valid_key_on_write_path() {
+            anyhow::bail!(
+                "the request contains data not supported by pageserver: {} @ {}",
+                key,
+                self.lsn
+            );
+        }
+        self.pending_zero_data_pages.insert(key.to_compact());
+        self.pending_bytes += ZERO_PAGE.len();
+        Ok(())
+    }
+
+    /// Call this at the end of each WAL record.
+    pub(crate) fn on_record_end(&mut self) {
+        let pending_zero_data_pages = std::mem::take(&mut self.pending_zero_data_pages);
+        for key in pending_zero_data_pages {
+            self.put_data(key, Value::Image(ZERO_PAGE.clone()));
+        }
+    }
+
     /// Store a relmapper file (pg_filenode.map) in the repository
     pub async fn put_relmap_file(
         &mut self,
@@ -1221,22 +1339,31 @@ impl<'a> DatadirModification<'a> {
 
     pub async fn put_twophase_file(
         &mut self,
-        xid: TransactionId,
+        xid: u64,
         img: Bytes,
         ctx: &RequestContext,
     ) -> anyhow::Result<()> {
         // Add it to the directory entry
-        let buf = self.get(TWOPHASEDIR_KEY, ctx).await?;
-        let mut dir = TwoPhaseDirectory::des(&buf)?;
-        if !dir.xids.insert(xid) {
-            anyhow::bail!("twophase file for xid {} already exists", xid);
-        }
-        self.pending_directory_entries
-            .push((DirectoryKind::TwoPhase, dir.xids.len()));
-        self.put(
-            TWOPHASEDIR_KEY,
-            Value::Image(Bytes::from(TwoPhaseDirectory::ser(&dir)?)),
-        );
+        let dirbuf = self.get(TWOPHASEDIR_KEY, ctx).await?;
+        let newdirbuf = if self.tline.pg_version >= 17 {
+            let mut dir = TwoPhaseDirectoryV17::des(&dirbuf)?;
+            if !dir.xids.insert(xid) {
+                anyhow::bail!("twophase file for xid {} already exists", xid);
+            }
+            self.pending_directory_entries
+                .push((DirectoryKind::TwoPhase, dir.xids.len()));
+            Bytes::from(TwoPhaseDirectoryV17::ser(&dir)?)
+        } else {
+            let xid = xid as u32;
+            let mut dir = TwoPhaseDirectory::des(&dirbuf)?;
+            if !dir.xids.insert(xid) {
+                anyhow::bail!("twophase file for xid {} already exists", xid);
+            }
+            self.pending_directory_entries
+                .push((DirectoryKind::TwoPhase, dir.xids.len()));
+            Bytes::from(TwoPhaseDirectory::ser(&dir)?)
+        };
+        self.put(TWOPHASEDIR_KEY, Value::Image(newdirbuf));
 
         self.put(twophase_file_key(xid), Value::Image(img));
         Ok(())
@@ -1539,22 +1666,32 @@ impl<'a> DatadirModification<'a> {
     /// This method is used for marking truncated SLRU files
     pub async fn drop_twophase_file(
         &mut self,
-        xid: TransactionId,
+        xid: u64,
         ctx: &RequestContext,
     ) -> anyhow::Result<()> {
         // Remove it from the directory entry
         let buf = self.get(TWOPHASEDIR_KEY, ctx).await?;
-        let mut dir = TwoPhaseDirectory::des(&buf)?;
+        let newdirbuf = if self.tline.pg_version >= 17 {
+            let mut dir = TwoPhaseDirectoryV17::des(&buf)?;
 
-        if !dir.xids.remove(&xid) {
-            warn!("twophase file for xid {} does not exist", xid);
-        }
-        self.pending_directory_entries
-            .push((DirectoryKind::TwoPhase, dir.xids.len()));
-        self.put(
-            TWOPHASEDIR_KEY,
-            Value::Image(Bytes::from(TwoPhaseDirectory::ser(&dir)?)),
-        );
+            if !dir.xids.remove(&xid) {
+                warn!("twophase file for xid {} does not exist", xid);
+            }
+            self.pending_directory_entries
+                .push((DirectoryKind::TwoPhase, dir.xids.len()));
+            Bytes::from(TwoPhaseDirectoryV17::ser(&dir)?)
+        } else {
+            let xid: u32 = u32::try_from(xid)?;
+            let mut dir = TwoPhaseDirectory::des(&buf)?;
+
+            if !dir.xids.remove(&xid) {
+                warn!("twophase file for xid {} does not exist", xid);
+            }
+            self.pending_directory_entries
+                .push((DirectoryKind::TwoPhase, dir.xids.len()));
+            Bytes::from(TwoPhaseDirectory::ser(&dir)?)
+        };
+        self.put(TWOPHASEDIR_KEY, Value::Image(newdirbuf));
 
         // Delete it
         self.delete(twophase_key_range(xid));
@@ -1597,7 +1734,7 @@ impl<'a> DatadirModification<'a> {
                 if aux_files_key_v1.is_empty() {
                     None
                 } else {
-                    warn!("this timeline is using deprecated aux file policy V1");
+                    warn!("this timeline is using deprecated aux file policy V1 (detected existing v1 files)");
                     self.tline.do_switch_aux_policy(AuxFilePolicy::V1)?;
                     Some(AuxFilePolicy::V1)
                 }
@@ -1778,7 +1915,7 @@ impl<'a> DatadirModification<'a> {
     /// retains all the metadata, but data pages are flushed. That's again OK
     /// for bulk import, where you are just loading data pages and won't try to
     /// modify the same pages twice.
-    pub async fn flush(&mut self, ctx: &RequestContext) -> anyhow::Result<()> {
+    pub(crate) async fn flush(&mut self, ctx: &RequestContext) -> anyhow::Result<()> {
         // Unless we have accumulated a decent amount of changes, it's not worth it
         // to scan through the pending_updates list.
         let pending_nblocks = self.pending_nblocks;
@@ -1789,26 +1926,11 @@ impl<'a> DatadirModification<'a> {
         let mut writer = self.tline.writer().await;
 
         // Flush relation and  SLRU data blocks, keep metadata.
-        let mut retained_pending_updates = HashMap::<_, Vec<_>>::new();
-        for (key, values) in self.pending_updates.drain() {
-            let mut write_batch = Vec::new();
-            for (lsn, value_ser_size, value) in values {
-                if key.is_rel_block_key() || key.is_slru_block_key() {
-                    // This bails out on first error without modifying pending_updates.
-                    // That's Ok, cf this function's doc comment.
-                    write_batch.push((key.to_compact(), lsn, value_ser_size, value));
-                } else {
-                    retained_pending_updates.entry(key).or_default().push((
-                        lsn,
-                        value_ser_size,
-                        value,
-                    ));
-                }
-            }
-            writer.put_batch(write_batch, ctx).await?;
-        }
+        let pending_data_pages = std::mem::take(&mut self.pending_data_pages);
 
-        self.pending_updates = retained_pending_updates;
+        // This bails out on first error without modifying pending_updates.
+        // That's Ok, cf this function's doc comment.
+        writer.put_batch(pending_data_pages, ctx).await?;
         self.pending_bytes = 0;
 
         if pending_nblocks != 0 {
@@ -1829,26 +1951,31 @@ impl<'a> DatadirModification<'a> {
     /// All the modifications in this atomic update are stamped by the specified LSN.
     ///
     pub async fn commit(&mut self, ctx: &RequestContext) -> anyhow::Result<()> {
+        // Commit should never be called mid-wal-record
+        assert!(self.pending_zero_data_pages.is_empty());
+
         let mut writer = self.tline.writer().await;
 
         let pending_nblocks = self.pending_nblocks;
         self.pending_nblocks = 0;
 
-        if !self.pending_updates.is_empty() {
-            // Ordering: the items in this batch do not need to be in any global order, but values for
-            // a particular Key must be in Lsn order relative to one another.  InMemoryLayer relies on
-            // this to do efficient updates to its index.
-            let batch: Vec<(CompactKey, Lsn, usize, Value)> = self
-                .pending_updates
+        // Ordering: the items in this batch do not need to be in any global order, but values for
+        // a particular Key must be in Lsn order relative to one another.  InMemoryLayer relies on
+        // this to do efficient updates to its index.
+        let mut write_batch = std::mem::take(&mut self.pending_data_pages);
+
+        write_batch.extend(
+            self.pending_metadata_pages
                 .drain()
                 .flat_map(|(key, values)| {
-                    values.into_iter().map(move |(lsn, val_ser_size, value)| {
-                        (key.to_compact(), lsn, val_ser_size, value)
-                    })
-                })
-                .collect::<Vec<_>>();
+                    values
+                        .into_iter()
+                        .map(move |(lsn, value_size, value)| (key, lsn, value_size, value))
+                }),
+        );
 
-            writer.put_batch(batch, ctx).await?;
+        if !write_batch.is_empty() {
+            writer.put_batch(write_batch, ctx).await?;
         }
 
         if !self.pending_deletions.is_empty() {
@@ -1879,33 +2006,58 @@ impl<'a> DatadirModification<'a> {
     }
 
     pub(crate) fn len(&self) -> usize {
-        self.pending_updates.len() + self.pending_deletions.len()
+        self.pending_metadata_pages.len()
+            + self.pending_data_pages.len()
+            + self.pending_deletions.len()
     }
 
-    // Internal helper functions to batch the modifications
-
+    /// Read a page from the Timeline we are writing to.  For metadata pages, this passes through
+    /// a cache in Self, which makes writes earlier in this modification visible to WAL records later
+    /// in the modification.
+    ///
+    /// For data pages, reads pass directly to the owning Timeline: any ingest code which reads a data
+    /// page must ensure that the pages they read are already committed in Timeline, for example
+    /// DB create operations are always preceded by a call to commit().  This is special cased because
+    /// it's rare: all the 'normal' WAL operations will only read metadata pages such as relation sizes,
+    /// and not data pages.
     async fn get(&self, key: Key, ctx: &RequestContext) -> Result<Bytes, PageReconstructError> {
-        // Have we already updated the same key? Read the latest pending updated
-        // version in that case.
-        //
-        // Note: we don't check pending_deletions. It is an error to request a
-        // value that has been removed, deletion only avoids leaking storage.
-        if let Some(values) = self.pending_updates.get(&key) {
-            if let Some((_, _, value)) = values.last() {
-                return if let Value::Image(img) = value {
-                    Ok(img.clone())
-                } else {
-                    // Currently, we never need to read back a WAL record that we
-                    // inserted in the same "transaction". All the metadata updates
-                    // work directly with Images, and we never need to read actual
-                    // data pages. We could handle this if we had to, by calling
-                    // the walredo manager, but let's keep it simple for now.
-                    Err(PageReconstructError::Other(anyhow::anyhow!(
-                        "unexpected pending WAL record"
-                    )))
-                };
+        if !Self::is_data_key(&key) {
+            // Have we already updated the same key? Read the latest pending updated
+            // version in that case.
+            //
+            // Note: we don't check pending_deletions. It is an error to request a
+            // value that has been removed, deletion only avoids leaking storage.
+            if let Some(values) = self.pending_metadata_pages.get(&key.to_compact()) {
+                if let Some((_, _, value)) = values.last() {
+                    return if let Value::Image(img) = value {
+                        Ok(img.clone())
+                    } else {
+                        // Currently, we never need to read back a WAL record that we
+                        // inserted in the same "transaction". All the metadata updates
+                        // work directly with Images, and we never need to read actual
+                        // data pages. We could handle this if we had to, by calling
+                        // the walredo manager, but let's keep it simple for now.
+                        Err(PageReconstructError::Other(anyhow::anyhow!(
+                            "unexpected pending WAL record"
+                        )))
+                    };
+                }
+            }
+        } else {
+            // This is an expensive check, so we only do it in debug mode. If reading a data key,
+            // this key should never be present in pending_data_pages. We ensure this by committing
+            // modifications before ingesting DB create operations, which are the only kind that reads
+            // data pages during ingest.
+            if cfg!(debug_assertions) {
+                for (dirty_key, _, _, _) in &self.pending_data_pages {
+                    debug_assert!(&key.to_compact() != dirty_key);
+                }
+
+                debug_assert!(!self.pending_zero_data_pages.contains(&key.to_compact()))
             }
         }
+
+        // Metadata page cache miss, or we're reading a data page.
         let lsn = Lsn::max(self.tline.get_last_record_lsn(), self.lsn);
         self.tline.get(key, lsn, ctx).await
     }
@@ -1917,11 +2069,40 @@ impl<'a> DatadirModification<'a> {
     }
 
     fn put(&mut self, key: Key, val: Value) {
-        let values = self.pending_updates.entry(key).or_default();
+        if Self::is_data_key(&key) {
+            self.put_data(key.to_compact(), val)
+        } else {
+            self.put_metadata(key.to_compact(), val)
+        }
+    }
+
+    fn put_data(&mut self, key: CompactKey, val: Value) {
+        let val_serialized_size = val.serialized_size().unwrap() as usize;
+
+        // If this page was previously zero'd in the same WalRecord, then drop the previous zero page write.  This
+        // is an optimization that avoids persisting both the zero page generated by us (e.g. during a relation extend),
+        // and the subsequent postgres-originating write
+        if self.pending_zero_data_pages.remove(&key) {
+            self.pending_bytes -= ZERO_PAGE.len();
+        }
+
+        self.pending_bytes += val_serialized_size;
+        self.pending_data_pages
+            .push((key, self.lsn, val_serialized_size, val))
+    }
+
+    fn put_metadata(&mut self, key: CompactKey, val: Value) {
+        let values = self.pending_metadata_pages.entry(key).or_default();
         // Replace the previous value if it exists at the same lsn
         if let Some((last_lsn, last_value_ser_size, last_value)) = values.last_mut() {
             if *last_lsn == self.lsn {
+                // Update the pending_bytes contribution from this entry, and update the serialized size in place
+                self.pending_bytes -= *last_value_ser_size;
                 *last_value_ser_size = val.serialized_size().unwrap() as usize;
+                self.pending_bytes += *last_value_ser_size;
+
+                // Use the latest value, this replaces any earlier write to the same (key,lsn), such as much
+                // have been generated by synthesized zero page writes prior to the first real write to a page.
                 *last_value = val;
                 return;
             }
@@ -1940,6 +2121,7 @@ impl<'a> DatadirModification<'a> {
 
 /// This struct facilitates accessing either a committed key from the timeline at a
 /// specific LSN, or the latest uncommitted key from a pending modification.
+///
 /// During WAL ingestion, the records from multiple LSNs may be batched in the same
 /// modification before being flushed to the timeline. Hence, the routines in WalIngest
 /// need to look up the keys in the modification first before looking them up in the
@@ -1979,11 +2161,21 @@ struct DbDirectory {
     dbdirs: HashMap<(Oid, Oid), bool>,
 }
 
+// The format of TwoPhaseDirectory changed in PostgreSQL v17, because the filenames of
+// pg_twophase files was expanded from 32-bit XIDs to 64-bit XIDs.  Previously, the files
+// were named like "pg_twophase/000002E5", now they're like
+// "pg_twophsae/0000000A000002E4".
+
 #[derive(Debug, Serialize, Deserialize)]
 struct TwoPhaseDirectory {
     xids: HashSet<TransactionId>,
 }
 
+#[derive(Debug, Serialize, Deserialize)]
+struct TwoPhaseDirectoryV17 {
+    xids: HashSet<u64>,
+}
+
 #[derive(Debug, Serialize, Deserialize, Default)]
 struct RelDirectory {
     // Set of relations that exist. (relfilenode, forknum)

pageserver/src/statvfs.rs

@@ -60,32 +60,7 @@ pub mod mock {
     use regex::Regex;
     use tracing::log::info;
 
-    #[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>,
-        },
-        Failure {
-            mocked_error: MockedError,
-        },
-    }
-
-    #[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
-    #[allow(clippy::upper_case_acronyms)]
-    pub enum MockedError {
-        EIO,
-    }
-
-    impl From<MockedError> for nix::Error {
-        fn from(e: MockedError) -> Self {
-            match e {
-                MockedError::EIO => nix::Error::EIO,
-            }
-        }
-    }
+    pub use pageserver_api::config::statvfs::mock::Behavior;
 
     pub fn get(tenants_dir: &Utf8Path, behavior: &Behavior) -> nix::Result<Statvfs> {
         info!("running mocked statvfs");
@@ -116,6 +91,7 @@ pub mod mock {
                     block_size: *blocksize,
                 })
             }
+            #[cfg(feature = "testing")]
             Behavior::Failure { mocked_error } => Err((*mocked_error).into()),
         }
     }

pageserver/src/task_mgr.rs

@@ -146,6 +146,12 @@ impl FromStr for TokioRuntimeMode {
     }
 }
 
+static TOKIO_THREAD_STACK_SIZE: Lazy<NonZeroUsize> = Lazy::new(|| {
+    env::var("NEON_PAGESERVER_TOKIO_THREAD_STACK_SIZE")
+        // the default 2MiB are insufficent, especially in debug mode
+        .unwrap_or_else(|| NonZeroUsize::new(4 * 1024 * 1024).unwrap())
+});
+
 static ONE_RUNTIME: Lazy<Option<tokio::runtime::Runtime>> = Lazy::new(|| {
     let thread_name = "pageserver-tokio";
     let Some(mode) = env::var("NEON_PAGESERVER_USE_ONE_RUNTIME") else {
@@ -164,6 +170,7 @@ static ONE_RUNTIME: Lazy<Option<tokio::runtime::Runtime>> = Lazy::new(|| {
             tokio::runtime::Builder::new_current_thread()
                 .thread_name(thread_name)
                 .enable_all()
+                .thread_stack_size(TOKIO_THREAD_STACK_SIZE.get())
                 .build()
                 .expect("failed to create one single runtime")
         }
@@ -173,6 +180,7 @@ static ONE_RUNTIME: Lazy<Option<tokio::runtime::Runtime>> = Lazy::new(|| {
                 .thread_name(thread_name)
                 .enable_all()
                 .worker_threads(num_workers.get())
+                .thread_stack_size(TOKIO_THREAD_STACK_SIZE.get())
                 .build()
                 .expect("failed to create one multi-threaded runtime")
         }
@@ -199,6 +207,7 @@ macro_rules! pageserver_runtime {
                     .thread_name($name)
                     .worker_threads(TOKIO_WORKER_THREADS.get())
                     .enable_all()
+                    .thread_stack_size(TOKIO_THREAD_STACK_SIZE.get())
                     .build()
                     .expect(std::concat!("Failed to create runtime ", $name))
             });

pageserver/src/tenant.rs

@@ -1,8 +1,9 @@
+//! Timeline repository implementation that keeps old data in layer files, and
+//! the recent changes in ephemeral files.
 //!
-//! Timeline repository implementation that keeps old data in files on disk, and
-//! the recent changes in memory. See tenant/*_layer.rs files.
-//! The functions here are responsible for locating the correct layer for the
-//! get/put call, walking back the timeline branching history as needed.
+//! See tenant/*_layer.rs files. The functions here are responsible for locating
+//! the correct layer for the get/put call, walking back the timeline branching
+//! history as needed.
 //!
 //! The files are stored in the .neon/tenants/<tenant_id>/timelines/<timeline_id>
 //! directory. See docs/pageserver-storage.md for how the files are managed.
@@ -509,6 +510,9 @@ pub enum TimelineArchivalError {
     #[error("Timeout")]
     Timeout,
 
+    #[error("ancestor is archived: {}", .0)]
+    HasArchivedParent(TimelineId),
+
     #[error("HasUnarchivedChildren")]
     HasUnarchivedChildren(Vec<TimelineId>),
 
@@ -524,6 +528,7 @@ impl Debug for TimelineArchivalError {
         match self {
             Self::NotFound => write!(f, "NotFound"),
             Self::Timeout => write!(f, "Timeout"),
+            Self::HasArchivedParent(p) => f.debug_tuple("HasArchivedParent").field(p).finish(),
             Self::HasUnarchivedChildren(c) => {
                 f.debug_tuple("HasUnarchivedChildren").field(c).finish()
             }
@@ -1369,11 +1374,20 @@ impl Tenant {
         let timeline = {
             let timelines = self.timelines.lock().unwrap();
 
-            let timeline = match timelines.get(&timeline_id) {
-                Some(t) => t,
-                None => return Err(TimelineArchivalError::NotFound),
+            let Some(timeline) = timelines.get(&timeline_id) else {
+                return Err(TimelineArchivalError::NotFound);
             };
 
+            if state == TimelineArchivalState::Unarchived {
+                if let Some(ancestor_timeline) = timeline.ancestor_timeline() {
+                    if ancestor_timeline.is_archived() == Some(true) {
+                        return Err(TimelineArchivalError::HasArchivedParent(
+                            ancestor_timeline.timeline_id,
+                        ));
+                    }
+                }
+            }
+
             // Ensure that there are no non-archived child timelines
             let children: Vec<TimelineId> = timelines
                 .iter()
@@ -7077,13 +7091,13 @@ mod tests {
             vec![
                 // Image layer at GC horizon
                 PersistentLayerKey {
-                    key_range: Key::MIN..Key::NON_L0_MAX,
+                    key_range: Key::MIN..Key::MAX,
                     lsn_range: Lsn(0x30)..Lsn(0x31),
                     is_delta: false
                 },
-                // The delta layer covers the full range (with the layer key hack to avoid being recognized as L0)
+                // The delta layer below the horizon
                 PersistentLayerKey {
-                    key_range: Key::MIN..Key::NON_L0_MAX,
+                    key_range: get_key(3)..get_key(4),
                     lsn_range: Lsn(0x30)..Lsn(0x48),
                     is_delta: true
                 },
@@ -8456,4 +8470,127 @@ mod tests {
 
         Ok(())
     }
+
+    // Regression test for https://github.com/neondatabase/neon/issues/9012
+    // Create an image arrangement where we have to read at different LSN ranges
+    // from a delta layer. This is achieved by overlapping an image layer on top of
+    // a delta layer. Like so:
+    //
+    //     A      B
+    // +----------------+ -> delta_layer
+    // |                |                           ^ lsn
+    // |       =========|-> nested_image_layer      |
+    // |       C        |                           |
+    // +----------------+                           |
+    // ======== -> baseline_image_layer             +-------> key
+    //
+    //
+    // When querying the key range [A, B) we need to read at different LSN ranges
+    // for [A, C) and [C, B). This test checks that the described edge case is handled correctly.
+    #[tokio::test]
+    async fn test_vectored_read_with_nested_image_layer() -> anyhow::Result<()> {
+        let harness = TenantHarness::create("test_vectored_read_with_nested_image_layer").await?;
+        let (tenant, ctx) = harness.load().await;
+
+        fn get_key(id: u32) -> Key {
+            let mut key = Key::from_hex("110000000033333333444444445500000000").unwrap();
+            key.field6 = id;
+            key
+        }
+
+        let mut expected_key_values = HashMap::new();
+
+        let baseline_image_layer_lsn = Lsn(0x10);
+        let mut baseline_img_layer = Vec::new();
+        for i in 0..5 {
+            let key = get_key(i);
+            let value = format!("value {i}@{baseline_image_layer_lsn}");
+
+            let removed = expected_key_values.insert(key, value.clone());
+            assert!(removed.is_none());
+
+            baseline_img_layer.push((key, Bytes::from(value)));
+        }
+
+        let nested_image_layer_lsn = Lsn(0x50);
+        let mut nested_img_layer = Vec::new();
+        for i in 5..10 {
+            let key = get_key(i);
+            let value = format!("value {i}@{nested_image_layer_lsn}");
+
+            let removed = expected_key_values.insert(key, value.clone());
+            assert!(removed.is_none());
+
+            nested_img_layer.push((key, Bytes::from(value)));
+        }
+
+        let mut delta_layer_spec = Vec::default();
+        let delta_layer_start_lsn = Lsn(0x20);
+        let mut delta_layer_end_lsn = delta_layer_start_lsn;
+
+        for i in 0..10 {
+            let key = get_key(i);
+            let key_in_nested = nested_img_layer
+                .iter()
+                .any(|(key_with_img, _)| *key_with_img == key);
+            let lsn = {
+                if key_in_nested {
+                    Lsn(nested_image_layer_lsn.0 + 0x10)
+                } else {
+                    delta_layer_start_lsn
+                }
+            };
+
+            let delta = format!("@{lsn}");
+            delta_layer_spec.push((
+                key,
+                lsn,
+                Value::WalRecord(NeonWalRecord::wal_append(&delta)),
+            ));
+            delta_layer_end_lsn = std::cmp::max(delta_layer_start_lsn, lsn);
+
+            expected_key_values
+                .get_mut(&key)
+                .expect("An image exists for each key")
+                .push_str(delta.as_str());
+        }
+
+        delta_layer_end_lsn = Lsn(delta_layer_end_lsn.0 + 1);
+
+        assert!(
+            nested_image_layer_lsn > delta_layer_start_lsn
+                && nested_image_layer_lsn < delta_layer_end_lsn
+        );
+
+        let tline = tenant
+            .create_test_timeline_with_layers(
+                TIMELINE_ID,
+                baseline_image_layer_lsn,
+                DEFAULT_PG_VERSION,
+                &ctx,
+                vec![DeltaLayerTestDesc::new_with_inferred_key_range(
+                    delta_layer_start_lsn..delta_layer_end_lsn,
+                    delta_layer_spec,
+                )], // delta layers
+                vec![
+                    (baseline_image_layer_lsn, baseline_img_layer),
+                    (nested_image_layer_lsn, nested_img_layer),
+                ], // image layers
+                delta_layer_end_lsn,
+            )
+            .await?;
+
+        let keyspace = KeySpace::single(get_key(0)..get_key(10));
+        let results = tline
+            .get_vectored(keyspace, delta_layer_end_lsn, &ctx)
+            .await
+            .expect("No vectored errors");
+        for (key, res) in results {
+            let value = res.expect("No key errors");
+            let expected_value = expected_key_values.remove(&key).expect("No unknown keys");
+            assert_eq!(value, Bytes::from(expected_value));
+        }
+
+        Ok(())
+    }
 }

pageserver/src/tenant/config.rs

@@ -9,11 +9,10 @@
 //! may lead to a data loss.
 //!
 use anyhow::bail;
+pub(crate) use pageserver_api::config::TenantConfigToml as TenantConf;
 use pageserver_api::models::AuxFilePolicy;
-use pageserver_api::models::CompactionAlgorithm;
 use pageserver_api::models::CompactionAlgorithmSettings;
 use pageserver_api::models::EvictionPolicy;
-use pageserver_api::models::LsnLease;
 use pageserver_api::models::{self, ThrottleConfig};
 use pageserver_api::shard::{ShardCount, ShardIdentity, ShardNumber, ShardStripeSize};
 use serde::de::IntoDeserializer;
@@ -23,50 +22,6 @@ use std::num::NonZeroU64;
 use std::time::Duration;
 use utils::generation::Generation;
 
-pub mod 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: super::CompactionAlgorithm =
-        super::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;
-
-    pub const DEFAULT_INGEST_BATCH_SIZE: u64 = 100;
-}
-
 #[derive(Debug, Copy, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub(crate) enum AttachmentMode {
     /// Our generation is current as far as we know, and as far as we know we are the only attached
@@ -281,96 +236,20 @@ impl LocationConf {
     }
 }
 
-/// 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, Serialize, Deserialize)]
-pub struct TenantConf {
-    // 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: 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: 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: pageserver_api::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: 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,
+impl Default for LocationConf {
+    // TODO: this should be removed once tenant loading can guarantee that we are never
+    // loading from a directory without a configuration.
+    // => tech debt since https://github.com/neondatabase/neon/issues/1555
+    fn default() -> Self {
+        Self {
+            mode: LocationMode::Attached(AttachedLocationConfig {
+                generation: Generation::none(),
+                attach_mode: AttachmentMode::Single,
+            }),
+            tenant_conf: TenantConfOpt::default(),
+            shard: ShardIdentity::unsharded(),
+        }
+    }
 }
 
 /// Same as TenantConf, but this struct preserves the information about
@@ -545,51 +424,6 @@ impl TenantConfOpt {
     }
 }
 
-impl Default for TenantConf {
-    fn default() -> Self {
-        use 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: 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: 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::tenant::throttle::Config::disabled(),
-            image_layer_creation_check_threshold: DEFAULT_IMAGE_LAYER_CREATION_CHECK_THRESHOLD,
-            switch_aux_file_policy: AuxFilePolicy::default_tenant_config(),
-            lsn_lease_length: LsnLease::DEFAULT_LENGTH,
-            lsn_lease_length_for_ts: LsnLease::DEFAULT_LENGTH_FOR_TS,
-        }
-    }
-}
-
 impl TryFrom<&'_ models::TenantConfig> for TenantConfOpt {
     type Error = anyhow::Error;
 
@@ -618,7 +452,8 @@ impl TryFrom<toml_edit::Item> for TenantConfOpt {
                     .map_err(|e| anyhow::anyhow!("{}: {}", e.path(), e.inner().message()));
             }
             toml_edit::Item::Table(table) => {
-                let deserializer = toml_edit::de::Deserializer::new(table.into());
+                let deserializer =
+                    toml_edit::de::Deserializer::from(toml_edit::DocumentMut::from(table));
                 return serde_path_to_error::deserialize(deserializer)
                     .map_err(|e| anyhow::anyhow!("{}: {}", e.path(), e.inner().message()));
             }

pageserver/src/tenant/metadata.rs

@@ -1,7 +1,8 @@
-//! Describes the legacy now hopefully no longer modified per-timeline metadata stored in
-//! `index_part.json` managed by [`remote_timeline_client`]. For many tenants and their timelines,
-//! this struct and it's original serialization format is still needed because they were written a
-//! long time ago.
+//! Describes the legacy now hopefully no longer modified per-timeline metadata.
+//!
+//! It is stored in `index_part.json` managed by [`remote_timeline_client`]. For many tenants and
+//! their timelines, this struct and its original serialization format is still needed because
+//! they were written a long time ago.
 //!
 //! Instead of changing and adding versioning to this, just change [`IndexPart`] with soft json
 //! versioning.

pageserver/src/tenant/mgr.rs

@@ -282,9 +282,10 @@ impl BackgroundPurges {
 static TENANTS: Lazy<std::sync::RwLock<TenantsMap>> =
     Lazy::new(|| std::sync::RwLock::new(TenantsMap::Initializing));
 
-/// The TenantManager is responsible for storing and mutating the collection of all tenants
-/// that this pageserver process has state for.  Every Tenant and SecondaryTenant instance
-/// lives inside the TenantManager.
+/// Responsible for storing and mutating the collection of all tenants
+/// that this pageserver has state for.
+///
+/// Every Tenant and SecondaryTenant instance lives inside the TenantManager.
 ///
 /// The most important role of the TenantManager is to prevent conflicts: e.g. trying to attach
 /// the same tenant twice concurrently, or trying to configure the same tenant into secondary
@@ -2346,8 +2347,9 @@ pub enum TenantMapError {
     ShuttingDown,
 }
 
-/// Guards a particular tenant_id's content in the TenantsMap.  While this
-/// structure exists, the TenantsMap will contain a [`TenantSlot::InProgress`]
+/// Guards a particular tenant_id's content in the TenantsMap.
+///
+/// While this structure exists, the TenantsMap will contain a [`TenantSlot::InProgress`]
 /// for this tenant, which acts as a marker for any operations targeting
 /// this tenant to retry later, or wait for the InProgress state to end.
 ///

pageserver/src/tenant/remote_timeline_client.rs

@@ -2184,6 +2184,8 @@ pub fn remote_timeline_path(
     remote_timelines_path(tenant_shard_id).join(Utf8Path::new(&timeline_id.to_string()))
 }
 
+/// Obtains the path of the given Layer in the remote
+///
 /// Note that the shard component of a remote layer path is _not_ always the same
 /// as in the TenantShardId of the caller: tenants may reference layers from a different
 /// ShardIndex.  Use the ShardIndex from the layer's metadata.

pageserver/src/tenant/remote_timeline_client/download.rs

@@ -548,7 +548,7 @@ pub(crate) async fn download_initdb_tar_zst(
         cancel,
     )
     .await
-    .map_err(|e| {
+    .inspect_err(|_e| {
         // Do a best-effort attempt at deleting the temporary file upon encountering an error.
         // We don't have async here nor do we want to pile on any extra errors.
         if let Err(e) = std::fs::remove_file(&temp_path) {
@@ -556,7 +556,6 @@ pub(crate) async fn download_initdb_tar_zst(
                 warn!("error deleting temporary file {temp_path}: {e}");
             }
         }
-        e
     })?;
 
     Ok((temp_path, file))

pageserver/src/tenant/remote_timeline_client/index.rs

@@ -1,4 +1,5 @@
 //! In-memory index to track the tenant files on the remote storage.
+//!
 //! Able to restore itself from the storage index parts, that are located in every timeline's remote directory and contain all data about
 //! remote timeline layers and its metadata.
 

pageserver/src/tenant/storage_layer.rs

@@ -279,7 +279,7 @@ pub(crate) enum LayerId {
 /// Layer wrapper for the read path. Note that it is valid
 /// to use these layers even after external operations have
 /// been performed on them (compaction, freeze, etc.).
-#[derive(Debug)]
+#[derive(Debug, Clone)]
 pub(crate) enum ReadableLayer {
     PersistentLayer(Layer),
     InMemoryLayer(Arc<InMemoryLayer>),
@@ -292,6 +292,8 @@ struct ReadDesc {
     layer_id: LayerId,
     /// Lsn range for the read, used for selecting the next read
     lsn_range: Range<Lsn>,
+    /// This read's index in [`LayerKeyspace::reads`];
+    read_id: LayerKeyspaceReadId,
 }
 
 /// Data structure which maintains a fringe of layers for the
@@ -310,9 +312,13 @@ pub(crate) struct LayerFringe {
 #[derive(Debug)]
 struct LayerKeyspace {
     layer: ReadableLayer,
-    target_keyspace: KeySpaceRandomAccum,
+    next_read_id: LayerKeyspaceReadId,
+    reads: HashMap<LayerKeyspaceReadId, (Range<Lsn>, KeySpace)>,
 }
 
+#[derive(PartialEq, Eq, Hash, Debug, Clone, Copy)]
+struct LayerKeyspaceReadId(usize);
+
 impl LayerFringe {
     pub(crate) fn new() -> Self {
         LayerFringe {
@@ -327,22 +333,24 @@ impl LayerFringe {
             None => return None,
         };
 
-        let removed = self.layers.remove_entry(&read_desc.layer_id);
+        let mut entry = match self.layers.entry(read_desc.layer_id) {
+            Entry::Occupied(o) => o,
+            Entry::Vacant(_) => unreachable!("fringe internals are always consistent"),
+        };
 
-        match removed {
-            Some((
-                _,
-                LayerKeyspace {
-                    layer,
-                    mut target_keyspace,
-                },
-            )) => Some((
-                layer,
-                target_keyspace.consume_keyspace(),
-                read_desc.lsn_range,
-            )),
-            None => unreachable!("fringe internals are always consistent"),
+        let (lsn_range, keyspace) = entry
+            .get_mut()
+            .reads
+            .remove(&read_desc.read_id)
+            .expect("fringe internals are always consistent");
+
+        let layer = entry.get().layer.clone();
+
+        if entry.get().reads.is_empty() {
+            entry.remove();
         }
+
+        Some((layer, keyspace, lsn_range))
     }
 
     pub(crate) fn update(
@@ -355,18 +363,31 @@ impl LayerFringe {
         let entry = self.layers.entry(layer_id.clone());
         match entry {
             Entry::Occupied(mut entry) => {
-                entry.get_mut().target_keyspace.add_keyspace(keyspace);
+                let read_id = {
+                    let r = &mut entry.get_mut().next_read_id;
+                    let read_id = *r;
+                    *r = LayerKeyspaceReadId(r.0 + 1);
+                    read_id
+                };
+                self.planned_reads_by_lsn.push(ReadDesc {
+                    lsn_range: lsn_range.clone(),
+                    layer_id: layer_id.clone(),
+                    read_id,
+                });
+                let replaced = entry.get_mut().reads.insert(read_id, (lsn_range, keyspace));
+                assert!(replaced.is_none());
             }
             Entry::Vacant(entry) => {
+                let read_id = LayerKeyspaceReadId(0);
                 self.planned_reads_by_lsn.push(ReadDesc {
-                    lsn_range,
+                    lsn_range: lsn_range.clone(),
                     layer_id: layer_id.clone(),
+                    read_id,
                 });
-                let mut accum = KeySpaceRandomAccum::new();
-                accum.add_keyspace(keyspace);
                 entry.insert(LayerKeyspace {
                     layer,
-                    target_keyspace: accum,
+                    next_read_id: LayerKeyspaceReadId(1),
+                    reads: [(read_id, (lsn_range, keyspace))].into(),
                 });
             }
         }
@@ -434,10 +455,11 @@ impl ReadableLayer {
     }
 }
 
-/// Layers contain a hint indicating whether they are likely to be used for reads.  This is a hint rather
-/// than an authoritative value, so that we do not have to update it synchronously when changing the visibility
-/// of layers (for example when creating a branch that makes some previously covered layers visible).  It should
-/// be used for cache management but not for correctness-critical checks.
+/// Layers contain a hint indicating whether they are likely to be used for reads.
+///
+/// This is a hint rather than an authoritative value, so that we do not have to update it synchronously
+/// when changing the visibility of layers (for example when creating a branch that makes some previously
+/// covered layers visible).  It should be used for cache management but not for correctness-critical checks.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub enum LayerVisibilityHint {
     /// A Visible layer might be read while serving a read, because there is not an image layer between it

pageserver/src/tenant/storage_layer/delta_layer.rs

@@ -39,7 +39,7 @@ use crate::tenant::disk_btree::{
 use crate::tenant::storage_layer::layer::S3_UPLOAD_LIMIT;
 use crate::tenant::timeline::GetVectoredError;
 use crate::tenant::vectored_blob_io::{
-    BlobFlag, MaxVectoredReadBytes, StreamingVectoredReadPlanner, VectoredBlobReader, VectoredRead,
+    BlobFlag, StreamingVectoredReadPlanner, VectoredBlobReader, VectoredRead,
     VectoredReadCoalesceMode, VectoredReadPlanner,
 };
 use crate::tenant::PageReconstructError;
@@ -52,6 +52,7 @@ use bytes::BytesMut;
 use camino::{Utf8Path, Utf8PathBuf};
 use futures::StreamExt;
 use itertools::Itertools;
+use pageserver_api::config::MaxVectoredReadBytes;
 use pageserver_api::keyspace::KeySpace;
 use pageserver_api::models::ImageCompressionAlgorithm;
 use pageserver_api::shard::TenantShardId;
@@ -135,10 +136,11 @@ impl Summary {
 // Flag indicating that this version initialize the page
 const WILL_INIT: u64 = 1;
 
-/// Struct representing reference to BLOB in layers. Reference contains BLOB
-/// offset, and for WAL records it also contains `will_init` flag. The flag
-/// helps to determine the range of records that needs to be applied, without
-/// reading/deserializing records themselves.
+/// Struct representing reference to BLOB in layers.
+///
+/// Reference contains BLOB offset, and for WAL records it also contains
+/// `will_init` flag. The flag helps to determine the range of records
+/// that needs to be applied, without reading/deserializing records themselves.
 #[derive(Debug, Serialize, Deserialize, Copy, Clone)]
 pub struct BlobRef(pub u64);
 
@@ -225,9 +227,7 @@ pub struct DeltaLayerInner {
     file: VirtualFile,
     file_id: FileId,
 
-    #[allow(dead_code)]
     layer_key_range: Range<Key>,
-    #[allow(dead_code)]
     layer_lsn_range: Range<Lsn>,
 
     max_vectored_read_bytes: Option<MaxVectoredReadBytes>,
@@ -882,44 +882,6 @@ impl DeltaLayerInner {
         Ok(())
     }
 
-    /// Load all key-values in the delta layer, should be replaced by an iterator-based interface in the future.
-    pub(super) async fn load_key_values(
-        &self,
-        ctx: &RequestContext,
-    ) -> anyhow::Result<Vec<(Key, Lsn, Value)>> {
-        let block_reader = FileBlockReader::new(&self.file, self.file_id);
-        let index_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
-            self.index_start_blk,
-            self.index_root_blk,
-            block_reader,
-        );
-        let mut result = Vec::new();
-        let mut stream =
-            Box::pin(self.stream_index_forwards(index_reader, &[0; DELTA_KEY_SIZE], ctx));
-        let block_reader = FileBlockReader::new(&self.file, self.file_id);
-        let cursor = block_reader.block_cursor();
-        let mut buf = Vec::new();
-        while let Some(item) = stream.next().await {
-            let (key, lsn, pos) = item?;
-            // TODO: dedup code with get_reconstruct_value
-            // TODO: ctx handling and sharding
-            cursor
-                .read_blob_into_buf(pos.pos(), &mut buf, ctx)
-                .await
-                .with_context(|| {
-                    format!("Failed to read blob from virtual file {}", self.file.path)
-                })?;
-            let val = Value::des(&buf).with_context(|| {
-                format!(
-                    "Failed to deserialize file blob from virtual file {}",
-                    self.file.path
-                )
-            })?;
-            result.push((key, lsn, val));
-        }
-        Ok(result)
-    }
-
     async fn plan_reads<Reader>(
         keyspace: &KeySpace,
         lsn_range: Range<Lsn>,
@@ -2283,7 +2245,7 @@ pub(crate) mod test {
             .await
             .unwrap();
         let delta_layer = resident_layer.get_as_delta(&ctx).await.unwrap();
-        for max_read_size in [1, 2048] {
+        for max_read_size in [1, 1024] {
             for batch_size in [1, 2, 4, 8, 3, 7, 13] {
                 println!("running with batch_size={batch_size} max_read_size={max_read_size}");
                 // Test if the batch size is correctly determined
@@ -2297,7 +2259,7 @@ pub(crate) mod test {
                         // every key should be a batch b/c the value is larger than max_read_size
                         assert_eq!(iter.key_values_batch.len(), 1);
                     } else {
-                        assert_eq!(iter.key_values_batch.len(), batch_size);
+                        assert!(iter.key_values_batch.len() <= batch_size);
                     }
                     if num_items >= N {
                         break;