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/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)
 
@@ -210,7 +225,7 @@ 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

.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
@@ -548,7 +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')) }}
@@ -627,7 +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
@@ -649,7 +649,7 @@ jobs:
 
     strategy:
       matrix:
-        version: [ v14, v15, v16 ]
+        version: [ v14, v15, v16, v17 ]
 
     steps:
       - uses: docker/login-action@v3
@@ -671,7 +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 \
@@ -689,7 +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 }}
@@ -700,7 +700,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        version: [ v14, v15, v16 ]
+        version: [ v14, v15, v16, v17 ]
     env:
       VM_BUILDER_VERSION: v0.29.3
 
@@ -794,14 +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
@@ -823,28 +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}:${{ needs.tag.outputs.build-tag }}
-          done
-
       - name: Add latest tag to images
         if: github.ref_name == 'main'
         run: |
@@ -864,7 +839,7 @@ jobs:
             done
           done
           docker buildx imagetools create -t neondatabase/neon-test-extensions-v16:latest \
-                                             neondatabase/neon-test-extensions-v16:${{ 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
@@ -877,11 +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
@@ -957,8 +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
@@ -972,7 +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
@@ -1118,6 +1117,7 @@ jobs:
 
               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

.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",
@@ -1363,8 +1360,8 @@ dependencies = [
  "tokio",
  "tokio-postgres",
  "tokio-util",
- "toml 0.7.4",
- "toml_edit 0.19.10",
+ "toml",
+ "toml_edit",
  "tracing",
  "url",
  "utils",
@@ -1488,25 +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",
 ]
@@ -2949,12 +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 = "libc"
 version = "0.2.150"
@@ -3153,7 +3141,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "fd01039851e82f8799046eabbb354056283fb265c8ec0996af940f4e85a380ff"
 dependencies = [
  "serde",
- "toml 0.8.14",
+ "toml",
 ]
 
 [[package]]
@@ -3669,7 +3657,7 @@ dependencies = [
  "thiserror",
  "tokio",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "utils",
  "workspace_hack",
 ]
@@ -3756,7 +3744,7 @@ dependencies = [
  "tokio-stream",
  "tokio-tar",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "tracing",
  "twox-hash",
  "url",
@@ -3919,8 +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",
@@ -3939,8 +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",
@@ -3977,12 +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"
@@ -4136,7 +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",
@@ -4149,7 +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",
@@ -4168,7 +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",
@@ -4280,9 +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",
@@ -4827,7 +4811,7 @@ dependencies = [
  "tokio",
  "tokio-stream",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "tracing",
  "utils",
 ]
@@ -5337,7 +5321,7 @@ dependencies = [
  "tokio-stream",
  "tokio-tar",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "tracing",
  "tracing-subscriber",
  "url",
@@ -5746,17 +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"
@@ -6069,21 +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]]
@@ -6094,8 +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"
@@ -6423,7 +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",
@@ -6534,18 +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"
@@ -6555,7 +6517,7 @@ dependencies = [
  "serde",
  "serde_spanned",
  "toml_datetime",
- "toml_edit 0.22.14",
+ "toml_edit",
 ]
 
 [[package]]
@@ -6567,19 +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"
@@ -6590,7 +6539,7 @@ dependencies = [
  "serde",
  "serde_spanned",
  "toml_datetime",
- "winnow 0.6.13",
+ "winnow",
 ]
 
 [[package]]
@@ -7003,7 +6952,7 @@ dependencies = [
  "tokio-stream",
  "tokio-tar",
  "tokio-util",
- "toml_edit 0.19.10",
+ "toml_edit",
  "tracing",
  "tracing-error",
  "tracing-subscriber",
@@ -7549,15 +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"
@@ -7627,6 +7567,7 @@ dependencies = [
  "hyper 0.14.26",
  "indexmap 1.9.3",
  "itertools 0.10.5",
+ "itertools 0.12.1",
  "lazy_static",
  "libc",
  "log",
@@ -7664,6 +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"
@@ -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
 
 

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/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

@@ -342,7 +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());

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",

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/src/config.rs

@@ -104,7 +104,9 @@ pub struct ConfigToml {
     pub image_compression: ImageCompressionAlgorithm,
     pub ephemeral_bytes_per_memory_kb: usize,
     pub l0_flush: Option<crate::models::L0FlushConfig>,
-    pub compact_level0_phase1_value_access: CompactL0Phase1ValueAccess,
+    #[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,
 }
@@ -209,43 +211,6 @@ pub enum GetImpl {
 #[serde(transparent)]
 pub struct MaxVectoredReadBytes(pub NonZeroUsize);
 
-#[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize, serde::Serialize)]
-#[serde(tag = "mode", rename_all = "kebab-case", deny_unknown_fields)]
-pub enum CompactL0Phase1ValueAccess {
-    /// The old way.
-    PageCachedBlobIo,
-    /// The new way.
-    StreamingKmerge {
-        /// If set, we run both the old way and the new way, validate that
-        /// they are identical (=> [`CompactL0BypassPageCacheValidation`]),
-        /// and if the validation fails,
-        /// - in tests: fail them with a panic or
-        /// - in prod, log a rate-limited warning and use the old way's results.
-        ///
-        /// If not set, we only run the new way and trust its results.
-        validate: Option<CompactL0BypassPageCacheValidation>,
-    },
-}
-
-/// See [`CompactL0Phase1ValueAccess::StreamingKmerge`].
-#[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize, serde::Serialize)]
-#[serde(rename_all = "kebab-case")]
-pub enum CompactL0BypassPageCacheValidation {
-    /// Validate that the series of (key, lsn) pairs are the same.
-    KeyLsn,
-    /// Validate that the entire output of old and new way is identical.
-    KeyLsnValue,
-}
-
-impl Default for CompactL0Phase1ValueAccess {
-    fn default() -> Self {
-        CompactL0Phase1ValueAccess::StreamingKmerge {
-            // TODO(https://github.com/neondatabase/neon/issues/8184): change to None once confident
-            validate: Some(CompactL0BypassPageCacheValidation::KeyLsnValue),
-        }
-    }
-}
-
 /// A tenant's calcuated configuration, which is the result of merging a
 /// tenant's TenantConfOpt with the global TenantConf from PageServerConf.
 ///
@@ -452,7 +417,7 @@ impl Default for ConfigToml {
             image_compression: (DEFAULT_IMAGE_COMPRESSION),
             ephemeral_bytes_per_memory_kb: (DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB),
             l0_flush: None,
-            compact_level0_phase1_value_access: CompactL0Phase1ValueAccess::default(),
+            compact_level0_phase1_value_access: Default::default(),
             virtual_file_direct_io: crate::models::virtual_file::DirectIoMode::default(),
 
             io_buffer_alignment: DEFAULT_IO_BUFFER_ALIGNMENT,

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};
 
@@ -263,15 +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 {
@@ -359,7 +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
@@ -591,35 +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

@@ -62,7 +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,
 )]
@@ -305,8 +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:
@@ -896,7 +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)]

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;
 

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/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())
     }
 

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/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/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

@@ -3,11 +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, strum_macros::Display, 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,
@@ -190,7 +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();

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/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,

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/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)]
@@ -454,6 +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

@@ -79,16 +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,

pageserver/ctl/src/main.rs

@@ -174,7 +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

@@ -37,6 +37,7 @@ use pageserver::{
     virtual_file,
 };
 use postgres_backend::AuthType;
+use utils::crashsafe::syncfs;
 use utils::failpoint_support;
 use utils::logging::TracingErrorLayerEnablement;
 use utils::{
@@ -125,7 +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.
@@ -156,23 +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(),

pageserver/src/config.rs

@@ -174,16 +174,14 @@ pub struct PageServerConf {
 
     pub l0_flush: crate::l0_flush::L0FlushConfig,
 
-    /// This flag is temporary and will be removed after gradual rollout.
-    /// See <https://github.com/neondatabase/neon/issues/8184>.
-    pub compact_level0_phase1_value_access: pageserver_api::config::CompactL0Phase1ValueAccess,
-
     /// Direct IO settings
     pub virtual_file_direct_io: virtual_file::DirectIoMode,
 
     pub io_buffer_alignment: usize,
 }
 
+/// Token for authentication to safekeepers
+///
 /// We do not want to store this in a PageServerConf because the latter may be logged
 /// and/or serialized at a whim, while the token is secret. Currently this token is the
 /// same for accessing all tenants/timelines, but may become per-tenant/per-timeline in
@@ -283,7 +281,7 @@ impl PageServerConf {
 
         #[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),
         }
     }
@@ -338,7 +336,7 @@ impl PageServerConf {
             max_vectored_read_bytes,
             image_compression,
             ephemeral_bytes_per_memory_kb,
-            compact_level0_phase1_value_access,
+            compact_level0_phase1_value_access: _,
             l0_flush,
             virtual_file_direct_io,
             concurrent_tenant_warmup,
@@ -383,7 +381,6 @@ impl PageServerConf {
             max_vectored_read_bytes,
             image_compression,
             ephemeral_bytes_per_memory_kb,
-            compact_level0_phase1_value_access,
             virtual_file_direct_io,
             io_buffer_alignment,
 
@@ -561,6 +558,16 @@ mod tests {
             .expect("parse_and_validate");
     }
 
+    #[test]
+    fn test_compactl0_phase1_access_mode_is_ignored_silently() {
+        let input = indoc::indoc! {r#"
+            [compact_level0_phase1_value_access]
+            mode = "streaming-kmerge"
+            validate = "key-lsn-value"
+        "#};
+        toml_edit::de::from_str::<pageserver_api::config::ConfigToml>(input).unwrap();
+    }
+
     /// If there's a typo in the pageserver config, we'd rather catch that typo
     /// and fail pageserver startup than silently ignoring the typo, leaving whoever
     /// made it in the believe that their config change is effective.
@@ -637,14 +644,5 @@ mod tests {
         //         some_invalid_field = 23
         //     "#}
         // );
-
-        test!(
-            compact_level0_phase1_value_access,
-            indoc! {r#"
-                [compact_level0_phase1_value_access]
-                mode = "streaming-kmerge"
-                some_invalid_field = 23
-            "#}
-        );
     }
 }

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?

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/import_datadir.rs

@@ -580,9 +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/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

@@ -633,7 +633,7 @@ impl Timeline {
 
     pub(crate) async fn get_twophase_file(
         &self,
-        xid: TransactionId,
+        xid: u64,
         lsn: Lsn,
         ctx: &RequestContext,
     ) -> Result<Bytes, PageReconstructError> {
@@ -646,11 +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(
@@ -902,9 +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));
@@ -1021,9 +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
@@ -1126,9 +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()));
@@ -1204,6 +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(())
     }
@@ -1215,14 +1241,34 @@ 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) {
-        self.pending_zero_data_pages
-            .insert(rel_block_to_key(rel, blknum).to_compact());
+    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(
@@ -1230,10 +1276,18 @@ impl<'a> DatadirModification<'a> {
         kind: SlruKind,
         segno: u32,
         blknum: BlockNumber,
-    ) {
-        self.pending_zero_data_pages
-            .insert(slru_block_to_key(kind, segno, blknum).to_compact());
+    ) -> 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.
@@ -1285,22 +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(())
@@ -1603,22 +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));
@@ -2048,6 +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
@@ -2087,11 +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/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.
@@ -7090,13 +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
                 },
@@ -8469,4 +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

@@ -452,7 +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

@@ -136,10 +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);
 

pageserver/src/tenant/storage_layer/image_layer.rs

@@ -1,7 +1,9 @@
 //! An ImageLayer represents an image or a snapshot of a key-range at
-//! one particular LSN. It contains an image of all key-value pairs
-//! in its key-range. Any key that falls into the image layer's range
-//! but does not exist in the layer, does not exist.
+//! one particular LSN.
+//!
+//! It contains an image of all key-value pairs in its key-range. Any key
+//! that falls into the image layer's range but does not exist in the layer,
+//! does not exist.
 //!
 //! An image layer is stored in a file on disk. The file is stored in
 //! timelines/<timeline_id> directory.  Currently, there are no

pageserver/src/tenant/storage_layer/layer_desc.rs

@@ -12,8 +12,10 @@ use serde::{Deserialize, Serialize};
 #[cfg(test)]
 use utils::id::TenantId;
 
-/// A unique identifier of a persistent layer. This is different from `LayerDescriptor`, which is only used in the
-/// benchmarks. This struct contains all necessary information to find the image / delta layer. It also provides
+/// A unique identifier of a persistent layer.
+///
+/// This is different from `LayerDescriptor`, which is only used in the benchmarks.
+/// This struct contains all necessary information to find the image / delta layer. It also provides
 /// a unified way to generate layer information like file name.
 #[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize, Hash)]
 pub struct PersistentLayerDesc {

pageserver/src/tenant/storage_layer/layer_name.rs

@@ -217,8 +217,9 @@ impl fmt::Display for ImageLayerName {
     }
 }
 
-/// LayerName is the logical identity of a layer within a LayerMap at a moment in time.  The
-/// LayerName is not a unique filename, as the same LayerName may have multiple physical incarnations
+/// LayerName is the logical identity of a layer within a LayerMap at a moment in time.
+///
+/// The LayerName is not a unique filename, as the same LayerName may have multiple physical incarnations
 /// over time (e.g. across shard splits or compression). The physical filenames of layers in local
 /// storage and object names in remote storage consist of the LayerName plus some extra qualifiers
 /// that uniquely identify the physical incarnation of a layer (see [crate::tenant::remote_timeline_client::remote_layer_path])

pageserver/src/tenant/storage_layer/merge_iterator.rs

@@ -226,9 +226,11 @@ impl<'a> IteratorWrapper<'a> {
     }
 }
 
-/// A merge iterator over delta/image layer iterators. When duplicated records are
-/// found, the iterator will not perform any deduplication, and the caller should handle
-/// these situation. By saying duplicated records, there are many possibilities:
+/// A merge iterator over delta/image layer iterators.
+///
+/// When duplicated records are found, the iterator will not perform any
+/// deduplication, and the caller should handle these situation. By saying
+/// duplicated records, there are many possibilities:
 ///
 /// * Two same delta at the same LSN.
 /// * Two same image at the same LSN.

pageserver/src/tenant/storage_layer/split_writer.rs

@@ -34,9 +34,10 @@ impl SplitWriterResult {
     }
 }
 
-/// An image writer that takes images and produces multiple image layers. The interface does not
-/// guarantee atomicity (i.e., if the image layer generation fails, there might be leftover files
-/// to be cleaned up)
+/// An image writer that takes images and produces multiple image layers.
+///
+/// The interface does not guarantee atomicity (i.e., if the image layer generation
+/// fails, there might be leftover files to be cleaned up)
 #[must_use]
 pub struct SplitImageLayerWriter {
     inner: ImageLayerWriter,
@@ -187,22 +188,23 @@ impl SplitImageLayerWriter {
             .await
     }
 
-    /// When split writer fails, the caller should call this function and handle partially generated layers.
+    /// This function will be deprecated with #8841.
     pub(crate) fn take(self) -> anyhow::Result<(Vec<SplitWriterResult>, ImageLayerWriter)> {
         Ok((self.generated_layers, self.inner))
     }
 }
 
-/// A delta writer that takes key-lsn-values and produces multiple delta layers. The interface does not
-/// guarantee atomicity (i.e., if the delta layer generation fails, there might be leftover files
-/// to be cleaned up).
+/// A delta writer that takes key-lsn-values and produces multiple delta layers.
+///
+/// The interface does not guarantee atomicity (i.e., if the delta layer generation fails,
+/// there might be leftover files to be cleaned up).
 ///
 /// Note that if updates of a single key exceed the target size limit, all of the updates will be batched
 /// into a single file. This behavior might change in the future. For reference, the legacy compaction algorithm
 /// will split them into multiple files based on size.
 #[must_use]
 pub struct SplitDeltaLayerWriter {
-    inner: DeltaLayerWriter,
+    inner: Option<(Key, DeltaLayerWriter)>,
     target_layer_size: u64,
     generated_layers: Vec<SplitWriterResult>,
     conf: &'static PageServerConf,
@@ -210,7 +212,6 @@ pub struct SplitDeltaLayerWriter {
     tenant_shard_id: TenantShardId,
     lsn_range: Range<Lsn>,
     last_key_written: Key,
-    start_key: Key,
 }
 
 impl SplitDeltaLayerWriter {
@@ -218,29 +219,18 @@ impl SplitDeltaLayerWriter {
         conf: &'static PageServerConf,
         timeline_id: TimelineId,
         tenant_shard_id: TenantShardId,
-        start_key: Key,
         lsn_range: Range<Lsn>,
         target_layer_size: u64,
-        ctx: &RequestContext,
     ) -> anyhow::Result<Self> {
         Ok(Self {
             target_layer_size,
-            inner: DeltaLayerWriter::new(
-                conf,
-                timeline_id,
-                tenant_shard_id,
-                start_key,
-                lsn_range.clone(),
-                ctx,
-            )
-            .await?,
+            inner: None,
             generated_layers: Vec::new(),
             conf,
             timeline_id,
             tenant_shard_id,
             lsn_range,
             last_key_written: Key::MIN,
-            start_key,
         })
     }
 
@@ -263,9 +253,26 @@ impl SplitDeltaLayerWriter {
         //
         // Also, keep all updates of a single key in a single file. TODO: split them using the legacy compaction
         // strategy. https://github.com/neondatabase/neon/issues/8837
+
+        if self.inner.is_none() {
+            self.inner = Some((
+                key,
+                DeltaLayerWriter::new(
+                    self.conf,
+                    self.timeline_id,
+                    self.tenant_shard_id,
+                    key,
+                    self.lsn_range.clone(),
+                    ctx,
+                )
+                .await?,
+            ));
+        }
+        let (_, inner) = self.inner.as_mut().unwrap();
+
         let addition_size_estimation = KEY_SIZE as u64 + 8 /* LSN u64 size */ + 80 /* value size estimation */;
-        if self.inner.num_keys() >= 1
-            && self.inner.estimated_size() + addition_size_estimation >= self.target_layer_size
+        if inner.num_keys() >= 1
+            && inner.estimated_size() + addition_size_estimation >= self.target_layer_size
         {
             if key != self.last_key_written {
                 let next_delta_writer = DeltaLayerWriter::new(
@@ -277,13 +284,13 @@ impl SplitDeltaLayerWriter {
                     ctx,
                 )
                 .await?;
-                let prev_delta_writer = std::mem::replace(&mut self.inner, next_delta_writer);
+                let (start_key, prev_delta_writer) =
+                    std::mem::replace(&mut self.inner, Some((key, next_delta_writer))).unwrap();
                 let layer_key = PersistentLayerKey {
-                    key_range: self.start_key..key,
+                    key_range: start_key..key,
                     lsn_range: self.lsn_range.clone(),
                     is_delta: true,
                 };
-                self.start_key = key;
                 if discard(&layer_key).await {
                     drop(prev_delta_writer);
                     self.generated_layers
@@ -294,17 +301,18 @@ impl SplitDeltaLayerWriter {
                     self.generated_layers
                         .push(SplitWriterResult::Produced(delta_layer));
                 }
-            } else if self.inner.estimated_size() >= S3_UPLOAD_LIMIT {
+            } else if inner.estimated_size() >= S3_UPLOAD_LIMIT {
                 // We have to produce a very large file b/c a key is updated too often.
                 anyhow::bail!(
                     "a single key is updated too often: key={}, estimated_size={}, and the layer file cannot be produced",
                     key,
-                    self.inner.estimated_size()
+                    inner.estimated_size()
                 );
             }
         }
         self.last_key_written = key;
-        self.inner.put_value(key, lsn, val, ctx).await
+        let (_, inner) = self.inner.as_mut().unwrap();
+        inner.put_value(key, lsn, val, ctx).await
     }
 
     pub async fn put_value(
@@ -323,7 +331,6 @@ impl SplitDeltaLayerWriter {
         self,
         tline: &Arc<Timeline>,
         ctx: &RequestContext,
-        end_key: Key,
         discard: D,
     ) -> anyhow::Result<Vec<SplitWriterResult>>
     where
@@ -335,11 +342,15 @@ impl SplitDeltaLayerWriter {
             inner,
             ..
         } = self;
+        let Some((start_key, inner)) = inner else {
+            return Ok(generated_layers);
+        };
         if inner.num_keys() == 0 {
             return Ok(generated_layers);
         }
+        let end_key = self.last_key_written.next();
         let layer_key = PersistentLayerKey {
-            key_range: self.start_key..end_key,
+            key_range: start_key..end_key,
             lsn_range: self.lsn_range.clone(),
             is_delta: true,
         };
@@ -358,15 +369,14 @@ impl SplitDeltaLayerWriter {
         self,
         tline: &Arc<Timeline>,
         ctx: &RequestContext,
-        end_key: Key,
     ) -> anyhow::Result<Vec<SplitWriterResult>> {
-        self.finish_with_discard_fn(tline, ctx, end_key, |_| async { false })
+        self.finish_with_discard_fn(tline, ctx, |_| async { false })
             .await
     }
 
-    /// When split writer fails, the caller should call this function and handle partially generated layers.
-    pub(crate) fn take(self) -> anyhow::Result<(Vec<SplitWriterResult>, DeltaLayerWriter)> {
-        Ok((self.generated_layers, self.inner))
+    /// This function will be deprecated with #8841.
+    pub(crate) fn take(self) -> anyhow::Result<(Vec<SplitWriterResult>, Option<DeltaLayerWriter>)> {
+        Ok((self.generated_layers, self.inner.map(|x| x.1)))
     }
 }
 
@@ -430,10 +440,8 @@ mod tests {
             tenant.conf,
             tline.timeline_id,
             tenant.tenant_shard_id,
-            get_key(0),
             Lsn(0x18)..Lsn(0x20),
             4 * 1024 * 1024,
-            &ctx,
         )
         .await
         .unwrap();
@@ -458,11 +466,22 @@ mod tests {
             )
             .await
             .unwrap();
-        let layers = delta_writer
-            .finish(&tline, &ctx, get_key(10))
-            .await
-            .unwrap();
+        let layers = delta_writer.finish(&tline, &ctx).await.unwrap();
         assert_eq!(layers.len(), 1);
+        assert_eq!(
+            layers
+                .into_iter()
+                .next()
+                .unwrap()
+                .into_resident_layer()
+                .layer_desc()
+                .key(),
+            PersistentLayerKey {
+                key_range: get_key(0)..get_key(1),
+                lsn_range: Lsn(0x18)..Lsn(0x20),
+                is_delta: true
+            }
+        );
     }
 
     #[tokio::test]
@@ -499,10 +518,8 @@ mod tests {
             tenant.conf,
             tline.timeline_id,
             tenant.tenant_shard_id,
-            get_key(0),
             Lsn(0x18)..Lsn(0x20),
             4 * 1024 * 1024,
-            &ctx,
         )
         .await
         .unwrap();
@@ -531,10 +548,7 @@ mod tests {
             .finish(&tline, &ctx, get_key(N as u32))
             .await
             .unwrap();
-        let delta_layers = delta_writer
-            .finish(&tline, &ctx, get_key(N as u32))
-            .await
-            .unwrap();
+        let delta_layers = delta_writer.finish(&tline, &ctx).await.unwrap();
         if discard {
             for layer in image_layers {
                 layer.into_discarded_layer();
@@ -553,6 +567,14 @@ mod tests {
                 .collect_vec();
             assert_eq!(image_layers.len(), N / 512 + 1);
             assert_eq!(delta_layers.len(), N / 512 + 1);
+            assert_eq!(
+                delta_layers.first().unwrap().layer_desc().key_range.start,
+                get_key(0)
+            );
+            assert_eq!(
+                delta_layers.last().unwrap().layer_desc().key_range.end,
+                get_key(N as u32)
+            );
             for idx in 0..image_layers.len() {
                 assert_ne!(image_layers[idx].layer_desc().key_range.start, Key::MIN);
                 assert_ne!(image_layers[idx].layer_desc().key_range.end, Key::MAX);
@@ -600,10 +622,8 @@ mod tests {
             tenant.conf,
             tline.timeline_id,
             tenant.tenant_shard_id,
-            get_key(0),
             Lsn(0x18)..Lsn(0x20),
             4 * 1024,
-            &ctx,
         )
         .await
         .unwrap();
@@ -642,11 +662,35 @@ mod tests {
             )
             .await
             .unwrap();
-        let layers = delta_writer
-            .finish(&tline, &ctx, get_key(10))
-            .await
-            .unwrap();
+        let layers = delta_writer.finish(&tline, &ctx).await.unwrap();
         assert_eq!(layers.len(), 2);
+        let mut layers_iter = layers.into_iter();
+        assert_eq!(
+            layers_iter
+                .next()
+                .unwrap()
+                .into_resident_layer()
+                .layer_desc()
+                .key(),
+            PersistentLayerKey {
+                key_range: get_key(0)..get_key(1),
+                lsn_range: Lsn(0x18)..Lsn(0x20),
+                is_delta: true
+            }
+        );
+        assert_eq!(
+            layers_iter
+                .next()
+                .unwrap()
+                .into_resident_layer()
+                .layer_desc()
+                .key(),
+            PersistentLayerKey {
+                key_range: get_key(1)..get_key(2),
+                lsn_range: Lsn(0x18)..Lsn(0x20),
+                is_delta: true
+            }
+        );
     }
 
     #[tokio::test]
@@ -666,10 +710,8 @@ mod tests {
             tenant.conf,
             tline.timeline_id,
             tenant.tenant_shard_id,
-            get_key(0),
             Lsn(0x10)..Lsn(N as u64 * 16 + 0x10),
             4 * 1024 * 1024,
-            &ctx,
         )
         .await
         .unwrap();
@@ -687,10 +729,20 @@ mod tests {
                 .await
                 .unwrap();
         }
-        let delta_layers = delta_writer
-            .finish(&tline, &ctx, get_key(N as u32))
-            .await
-            .unwrap();
+        let delta_layers = delta_writer.finish(&tline, &ctx).await.unwrap();
         assert_eq!(delta_layers.len(), 1);
+        let delta_layer = delta_layers
+            .into_iter()
+            .next()
+            .unwrap()
+            .into_resident_layer();
+        assert_eq!(
+            delta_layer.layer_desc().key(),
+            PersistentLayerKey {
+                key_range: get_key(0)..get_key(1),
+                lsn_range: Lsn(0x10)..Lsn(N as u64 * 16 + 0x10),
+                is_delta: true
+            }
+        );
     }
 }

pageserver/src/tenant/timeline/compaction.rs

@@ -19,7 +19,6 @@ use bytes::Bytes;
 use enumset::EnumSet;
 use fail::fail_point;
 use itertools::Itertools;
-use pageserver_api::config::{CompactL0BypassPageCacheValidation, CompactL0Phase1ValueAccess};
 use pageserver_api::key::KEY_SIZE;
 use pageserver_api::keyspace::ShardedRange;
 use pageserver_api::shard::{ShardCount, ShardIdentity, TenantShardId};
@@ -912,137 +911,13 @@ impl Timeline {
         // we're compacting, in key, LSN order.
         // If there's both a Value::Image and Value::WalRecord for the same (key,lsn),
         // then the Value::Image is ordered before Value::WalRecord.
-        //
-        // TODO(https://github.com/neondatabase/neon/issues/8184): remove the page cached blob_io
-        // option and validation code once we've reached confidence.
-        enum AllValuesIter<'a> {
-            PageCachedBlobIo {
-                all_keys_iter: VecIter<'a>,
-            },
-            StreamingKmergeBypassingPageCache {
-                merge_iter: MergeIterator<'a>,
-            },
-            ValidatingStreamingKmergeBypassingPageCache {
-                mode: CompactL0BypassPageCacheValidation,
-                merge_iter: MergeIterator<'a>,
-                all_keys_iter: VecIter<'a>,
-            },
-        }
-        type VecIter<'a> = std::slice::Iter<'a, DeltaEntry<'a>>; // TODO: distinguished lifetimes
-        impl AllValuesIter<'_> {
-            async fn next_all_keys_iter(
-                iter: &mut VecIter<'_>,
-                ctx: &RequestContext,
-            ) -> anyhow::Result<Option<(Key, Lsn, Value)>> {
-                let Some(DeltaEntry {
-                    key,
-                    lsn,
-                    val: value_ref,
-                    ..
-                }) = iter.next()
-                else {
-                    return Ok(None);
-                };
-                let value = value_ref.load(ctx).await?;
-                Ok(Some((*key, *lsn, value)))
-            }
-            async fn next(
-                &mut self,
-                ctx: &RequestContext,
-            ) -> anyhow::Result<Option<(Key, Lsn, Value)>> {
-                match self {
-                    AllValuesIter::PageCachedBlobIo { all_keys_iter: iter } => {
-                      Self::next_all_keys_iter(iter, ctx).await
-                    }
-                    AllValuesIter::StreamingKmergeBypassingPageCache { merge_iter } => merge_iter.next().await,
-                    AllValuesIter::ValidatingStreamingKmergeBypassingPageCache { mode, merge_iter, all_keys_iter } => async {
-                        // advance both iterators
-                        let all_keys_iter_item = Self::next_all_keys_iter(all_keys_iter, ctx).await;
-                        let merge_iter_item = merge_iter.next().await;
-                        // compare results & log warnings as needed
-                        macro_rules! rate_limited_warn {
-                            ($($arg:tt)*) => {{
-                                if cfg!(debug_assertions) || cfg!(feature = "testing") {
-                                    warn!($($arg)*);
-                                    panic!("CompactL0BypassPageCacheValidation failure, check logs");
-                                }
-                                use once_cell::sync::Lazy;
-                                use utils::rate_limit::RateLimit;
-                                use std::sync::Mutex;
-                                use std::time::Duration;
-                                static LOGGED: Lazy<Mutex<RateLimit>> =
-                                    Lazy::new(|| Mutex::new(RateLimit::new(Duration::from_secs(10))));
-                                let mut rate_limit = LOGGED.lock().unwrap();
-                                rate_limit.call(|| {
-                                    warn!($($arg)*);
-                                });
-                            }}
-                        }
-                        match (&all_keys_iter_item, &merge_iter_item) {
-                            (Err(_), Err(_)) => {
-                                // don't bother asserting equivality of the errors
-                            }
-                            (Err(all_keys), Ok(merge)) => {
-                                rate_limited_warn!(?merge, "all_keys_iter returned an error where merge did not: {all_keys:?}");
-                            },
-                            (Ok(all_keys), Err(merge)) => {
-                                rate_limited_warn!(?all_keys, "merge returned an error where all_keys_iter did not: {merge:?}");
-                            },
-                            (Ok(None), Ok(None)) => { }
-                            (Ok(Some(all_keys)), Ok(None)) => {
-                                rate_limited_warn!(?all_keys, "merge returned None where all_keys_iter returned Some");
-                            }
-                            (Ok(None), Ok(Some(merge))) => {
-                                rate_limited_warn!(?merge, "all_keys_iter returned None where merge returned Some");
-                            }
-                            (Ok(Some((all_keys_key, all_keys_lsn, all_keys_value))), Ok(Some((merge_key, merge_lsn, merge_value)))) => {
-                                match mode {
-                                    // TODO: in this mode, we still load the value from disk for both iterators, even though we only need the all_keys_iter one
-                                    CompactL0BypassPageCacheValidation::KeyLsn => {
-                                        let all_keys = (all_keys_key, all_keys_lsn);
-                                        let merge = (merge_key, merge_lsn);
-                                        if all_keys != merge {
-                                            rate_limited_warn!(?all_keys, ?merge, "merge returned a different (Key,LSN) than all_keys_iter");
-                                        }
-                                    }
-                                    CompactL0BypassPageCacheValidation::KeyLsnValue => {
-                                        let all_keys = (all_keys_key, all_keys_lsn, all_keys_value);
-                                        let merge = (merge_key, merge_lsn, merge_value);
-                                        if all_keys != merge {
-                                            rate_limited_warn!(?all_keys, ?merge, "merge returned a different (Key,LSN,Value) than all_keys_iter");
-                                        }
-                                    }
-                                }
-                            }
-                        }
-                        // in case of mismatch, trust the legacy all_keys_iter_item
-                        all_keys_iter_item
-                    }.instrument(info_span!("next")).await
-                }
-            }
-        }
-        let mut all_values_iter = match &self.conf.compact_level0_phase1_value_access {
-            CompactL0Phase1ValueAccess::PageCachedBlobIo => AllValuesIter::PageCachedBlobIo {
-                all_keys_iter: all_keys.iter(),
-            },
-            CompactL0Phase1ValueAccess::StreamingKmerge { validate } => {
-                let merge_iter = {
-                    let mut deltas = Vec::with_capacity(deltas_to_compact.len());
-                    for l in deltas_to_compact.iter() {
-                        let l = l.get_as_delta(ctx).await.map_err(CompactionError::Other)?;
-                        deltas.push(l);
-                    }
-                    MergeIterator::create(&deltas, &[], ctx)
-                };
-                match validate {
-                    None => AllValuesIter::StreamingKmergeBypassingPageCache { merge_iter },
-                    Some(validate) => AllValuesIter::ValidatingStreamingKmergeBypassingPageCache {
-                        mode: validate.clone(),
-                        merge_iter,
-                        all_keys_iter: all_keys.iter(),
-                    },
-                }
+        let mut all_values_iter = {
+            let mut deltas = Vec::with_capacity(deltas_to_compact.len());
+            for l in deltas_to_compact.iter() {
+                let l = l.get_as_delta(ctx).await.map_err(CompactionError::Other)?;
+                deltas.push(l);
             }
+            MergeIterator::create(&deltas, &[], ctx)
         };
 
         // This iterator walks through all keys and is needed to calculate size used by each key
@@ -1119,7 +994,7 @@ impl Timeline {
         let mut keys = 0;
 
         while let Some((key, lsn, value)) = all_values_iter
-            .next(ctx)
+            .next()
             .await
             .map_err(CompactionError::Other)?
         {
@@ -1934,7 +1809,6 @@ impl Timeline {
             .unwrap();
         // We don't want any of the produced layers to cover the full key range (i.e., MIN..MAX) b/c it will then be recognized
         // as an L0 layer.
-        let hack_end_key = Key::NON_L0_MAX;
         let mut delta_layers = Vec::new();
         let mut image_layers = Vec::new();
         let mut downloaded_layers = Vec::new();
@@ -1980,10 +1854,8 @@ impl Timeline {
             self.conf,
             self.timeline_id,
             self.tenant_shard_id,
-            Key::MIN,
             lowest_retain_lsn..end_lsn,
             self.get_compaction_target_size(),
-            ctx,
         )
         .await?;
 
@@ -2090,7 +1962,7 @@ impl Timeline {
         let produced_image_layers = if let Some(writer) = image_layer_writer {
             if !dry_run {
                 writer
-                    .finish_with_discard_fn(self, ctx, hack_end_key, discard)
+                    .finish_with_discard_fn(self, ctx, Key::MAX, discard)
                     .await?
             } else {
                 let (layers, _) = writer.take()?;
@@ -2103,7 +1975,7 @@ impl Timeline {
 
         let produced_delta_layers = if !dry_run {
             delta_layer_writer
-                .finish_with_discard_fn(self, ctx, hack_end_key, discard)
+                .finish_with_discard_fn(self, ctx, discard)
                 .await?
         } else {
             let (layers, _) = delta_layer_writer.take()?;

pageserver/src/tenant/vectored_blob_io.rs

@@ -593,8 +593,10 @@ impl<'a> VectoredBlobReader<'a> {
     }
 }
 
-/// Read planner used in [`crate::tenant::storage_layer::image_layer::ImageLayerIterator`]. It provides a streaming API for
-/// getting read blobs. It returns a batch when `handle` gets called and when the current key would just exceed the read_size and
+/// Read planner used in [`crate::tenant::storage_layer::image_layer::ImageLayerIterator`].
+///
+/// It provides a streaming API for getting read blobs. It returns a batch when
+/// `handle` gets called and when the current key would just exceed the read_size and
 /// max_cnt constraints.
 pub struct StreamingVectoredReadPlanner {
     read_builder: Option<VectoredReadBuilder>,

pageserver/src/virtual_file.rs

@@ -1,6 +1,7 @@
-//!
 //! VirtualFile is like a normal File, but it's not bound directly to
-//! a file descriptor. Instead, the file is opened when it's read from,
+//! a file descriptor.
+//!
+//! Instead, the file is opened when it's read from,
 //! and if too many files are open globally in the system, least-recently
 //! used ones are closed.
 //!

pageserver/src/walingest.rs

@@ -25,9 +25,7 @@ use std::time::Duration;
 use std::time::SystemTime;
 
 use pageserver_api::shard::ShardIdentity;
-use postgres_ffi::v14::nonrelfile_utils::clogpage_precedes;
-use postgres_ffi::v14::nonrelfile_utils::slru_may_delete_clogsegment;
-use postgres_ffi::TimestampTz;
+use postgres_ffi::{dispatch_pgversion, enum_pgversion, enum_pgversion_dispatch, TimestampTz};
 use postgres_ffi::{fsm_logical_to_physical, page_is_new, page_set_lsn};
 
 use anyhow::{bail, Context, Result};
@@ -48,16 +46,31 @@ use pageserver_api::key::rel_block_to_key;
 use pageserver_api::reltag::{BlockNumber, RelTag, SlruKind};
 use postgres_ffi::pg_constants;
 use postgres_ffi::relfile_utils::{FSM_FORKNUM, INIT_FORKNUM, MAIN_FORKNUM, VISIBILITYMAP_FORKNUM};
-use postgres_ffi::v14::nonrelfile_utils::mx_offset_to_member_segment;
-use postgres_ffi::v14::xlog_utils::*;
-use postgres_ffi::v14::CheckPoint;
 use postgres_ffi::TransactionId;
 use postgres_ffi::BLCKSZ;
+use utils::bin_ser::SerializeError;
 use utils::lsn::Lsn;
 
+enum_pgversion! {CheckPoint, pgv::CheckPoint}
+
+impl CheckPoint {
+    fn encode(&self) -> Result<Bytes, SerializeError> {
+        enum_pgversion_dispatch!(self, CheckPoint, cp, { cp.encode() })
+    }
+
+    fn update_next_xid(&mut self, xid: u32) -> bool {
+        enum_pgversion_dispatch!(self, CheckPoint, cp, { cp.update_next_xid(xid) })
+    }
+
+    pub fn update_next_multixid(&mut self, multi_xid: u32, multi_offset: u32) -> bool {
+        enum_pgversion_dispatch!(self, CheckPoint, cp, {
+            cp.update_next_multixid(multi_xid, multi_offset)
+        })
+    }
+}
+
 pub struct WalIngest {
     shard: ShardIdentity,
-    pg_version: u32,
     checkpoint: CheckPoint,
     checkpoint_modified: bool,
     warn_ingest_lag: WarnIngestLag,
@@ -78,12 +91,16 @@ impl WalIngest {
         // Fetch the latest checkpoint into memory, so that we can compare with it
         // quickly in `ingest_record` and update it when it changes.
         let checkpoint_bytes = timeline.get_checkpoint(startpoint, ctx).await?;
-        let checkpoint = CheckPoint::decode(&checkpoint_bytes)?;
-        trace!("CheckPoint.nextXid = {}", checkpoint.nextXid.value);
+        let pgversion = timeline.pg_version;
+
+        let checkpoint = dispatch_pgversion!(pgversion, {
+            let checkpoint = pgv::CheckPoint::decode(&checkpoint_bytes)?;
+            trace!("CheckPoint.nextXid = {}", checkpoint.nextXid.value);
+            <pgv::CheckPoint as Into<CheckPoint>>::into(checkpoint)
+        });
 
         Ok(WalIngest {
             shard: *timeline.get_shard_identity(),
-            pg_version: timeline.pg_version,
             checkpoint,
             checkpoint_modified: false,
             warn_ingest_lag: WarnIngestLag {
@@ -117,7 +134,7 @@ impl WalIngest {
 
         modification.set_lsn(lsn)?;
 
-        if decoded.is_dbase_create_copy(self.pg_version) {
+        if decoded.is_dbase_create_copy(pg_version) {
             // Records of this type should always be preceded by a commit(), as they
             // rely on reading data pages back from the Timeline.
             assert!(!modification.has_dirty_data_pages());
@@ -220,6 +237,26 @@ impl WalIngest {
                                 .await?;
                         }
                     }
+                } else if pg_version == 17 {
+                    if info == postgres_ffi::v17::bindings::XLOG_DBASE_CREATE_WAL_LOG {
+                        debug!("XLOG_DBASE_CREATE_WAL_LOG: noop");
+                    } else if info == postgres_ffi::v17::bindings::XLOG_DBASE_CREATE_FILE_COPY {
+                        // The XLOG record was renamed between v14 and v15,
+                        // but the record format is the same.
+                        // So we can reuse XlCreateDatabase here.
+                        debug!("XLOG_DBASE_CREATE_FILE_COPY");
+                        let createdb = XlCreateDatabase::decode(&mut buf);
+                        self.ingest_xlog_dbase_create(modification, &createdb, ctx)
+                            .await?;
+                    } else if info == postgres_ffi::v17::bindings::XLOG_DBASE_DROP {
+                        let dropdb = XlDropDatabase::decode(&mut buf);
+                        for tablespace_id in dropdb.tablespace_ids {
+                            trace!("Drop db {}, {}", tablespace_id, dropdb.db_id);
+                            modification
+                                .drop_dbdir(tablespace_id, dropdb.db_id, ctx)
+                                .await?;
+                        }
+                    }
                 }
             }
             pg_constants::RM_TBLSPC_ID => {
@@ -229,7 +266,11 @@ impl WalIngest {
                 let info = decoded.xl_info & !pg_constants::XLR_INFO_MASK;
 
                 if info == pg_constants::CLOG_ZEROPAGE {
-                    let pageno = buf.get_u32_le();
+                    let pageno = if pg_version < 17 {
+                        buf.get_u32_le()
+                    } else {
+                        buf.get_u64_le() as u32
+                    };
                     let segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
                     let rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
                     self.put_slru_page_image(
@@ -243,7 +284,7 @@ impl WalIngest {
                     .await?;
                 } else {
                     assert!(info == pg_constants::CLOG_TRUNCATE);
-                    let xlrec = XlClogTruncate::decode(&mut buf);
+                    let xlrec = XlClogTruncate::decode(&mut buf, pg_version);
                     self.ingest_clog_truncate_record(modification, &xlrec, ctx)
                         .await?;
                 }
@@ -282,12 +323,21 @@ impl WalIngest {
                         parsed_xact.xid,
                         lsn,
                     );
-                    modification
-                        .drop_twophase_file(parsed_xact.xid, ctx)
-                        .await?;
+
+                    let xid: u64 = if pg_version >= 17 {
+                        self.adjust_to_full_transaction_id(parsed_xact.xid)?
+                    } else {
+                        parsed_xact.xid as u64
+                    };
+                    modification.drop_twophase_file(xid, ctx).await?;
                 } else if info == pg_constants::XLOG_XACT_PREPARE {
+                    let xid: u64 = if pg_version >= 17 {
+                        self.adjust_to_full_transaction_id(decoded.xl_xid)?
+                    } else {
+                        decoded.xl_xid as u64
+                    };
                     modification
-                        .put_twophase_file(decoded.xl_xid, Bytes::copy_from_slice(&buf[..]), ctx)
+                        .put_twophase_file(xid, Bytes::copy_from_slice(&buf[..]), ctx)
                         .await?;
                 }
             }
@@ -295,7 +345,11 @@ impl WalIngest {
                 let info = decoded.xl_info & pg_constants::XLR_RMGR_INFO_MASK;
 
                 if info == pg_constants::XLOG_MULTIXACT_ZERO_OFF_PAGE {
-                    let pageno = buf.get_u32_le();
+                    let pageno = if pg_version < 17 {
+                        buf.get_u32_le()
+                    } else {
+                        buf.get_u64_le() as u32
+                    };
                     let segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
                     let rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
                     self.put_slru_page_image(
@@ -308,7 +362,11 @@ impl WalIngest {
                     )
                     .await?;
                 } else if info == pg_constants::XLOG_MULTIXACT_ZERO_MEM_PAGE {
-                    let pageno = buf.get_u32_le();
+                    let pageno = if pg_version < 17 {
+                        buf.get_u32_le()
+                    } else {
+                        buf.get_u64_le() as u32
+                    };
                     let segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
                     let rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
                     self.put_slru_page_image(
@@ -337,70 +395,93 @@ impl WalIngest {
             pg_constants::RM_XLOG_ID => {
                 let info = decoded.xl_info & pg_constants::XLR_RMGR_INFO_MASK;
 
-                if info == pg_constants::XLOG_NEXTOID {
-                    let next_oid = buf.get_u32_le();
-                    if self.checkpoint.nextOid != next_oid {
-                        self.checkpoint.nextOid = next_oid;
+                if info == pg_constants::XLOG_PARAMETER_CHANGE {
+                    if let CheckPoint::V17(cp) = &mut self.checkpoint {
+                        let rec = v17::XlParameterChange::decode(&mut buf);
+                        cp.wal_level = rec.wal_level;
                         self.checkpoint_modified = true;
                     }
-                } else if info == pg_constants::XLOG_CHECKPOINT_ONLINE
-                    || info == pg_constants::XLOG_CHECKPOINT_SHUTDOWN
-                {
-                    let mut checkpoint_bytes = [0u8; SIZEOF_CHECKPOINT];
-                    buf.copy_to_slice(&mut checkpoint_bytes);
-                    let xlog_checkpoint = CheckPoint::decode(&checkpoint_bytes)?;
-                    trace!(
-                        "xlog_checkpoint.oldestXid={}, checkpoint.oldestXid={}",
-                        xlog_checkpoint.oldestXid,
-                        self.checkpoint.oldestXid
-                    );
-                    if (self
-                        .checkpoint
-                        .oldestXid
-                        .wrapping_sub(xlog_checkpoint.oldestXid) as i32)
-                        < 0
-                    {
-                        self.checkpoint.oldestXid = xlog_checkpoint.oldestXid;
+                } else if info == pg_constants::XLOG_END_OF_RECOVERY {
+                    if let CheckPoint::V17(cp) = &mut self.checkpoint {
+                        let rec = v17::XlEndOfRecovery::decode(&mut buf);
+                        cp.wal_level = rec.wal_level;
+                        self.checkpoint_modified = true;
                     }
-                    trace!(
-                        "xlog_checkpoint.oldestActiveXid={}, checkpoint.oldestActiveXid={}",
-                        xlog_checkpoint.oldestActiveXid,
-                        self.checkpoint.oldestActiveXid
-                    );
-
-                    // A shutdown checkpoint has `oldestActiveXid == InvalidTransactionid`,
-                    // because at shutdown, all in-progress transactions will implicitly
-                    // end. Postgres startup code knows that, and allows hot standby to start
-                    // immediately from a shutdown checkpoint.
-                    //
-                    // In Neon, Postgres hot standby startup always behaves as if starting from
-                    // an online checkpoint. It needs a valid `oldestActiveXid` value, so
-                    // instead of overwriting self.checkpoint.oldestActiveXid with
-                    // InvalidTransactionid from the checkpoint WAL record, update it to a
-                    // proper value, knowing that there are no in-progress transactions at this
-                    // point, except for prepared transactions.
-                    //
-                    // See also the neon code changes in the InitWalRecovery() function.
-                    if xlog_checkpoint.oldestActiveXid == pg_constants::INVALID_TRANSACTION_ID
-                        && info == pg_constants::XLOG_CHECKPOINT_SHUTDOWN
-                    {
-                        let mut oldest_active_xid = self.checkpoint.nextXid.value as u32;
-                        for xid in modification.tline.list_twophase_files(lsn, ctx).await? {
-                            if (xid.wrapping_sub(oldest_active_xid) as i32) < 0 {
-                                oldest_active_xid = xid;
-                            }
-                        }
-                        self.checkpoint.oldestActiveXid = oldest_active_xid;
-                    } else {
-                        self.checkpoint.oldestActiveXid = xlog_checkpoint.oldestActiveXid;
-                    }
-
-                    // Write a new checkpoint key-value pair on every checkpoint record, even
-                    // if nothing really changed. Not strictly required, but it seems nice to
-                    // have some trace of the checkpoint records in the layer files at the same
-                    // LSNs.
-                    self.checkpoint_modified = true;
                 }
+
+                enum_pgversion_dispatch!(&mut self.checkpoint, CheckPoint, cp, {
+                    if info == pg_constants::XLOG_NEXTOID {
+                        let next_oid = buf.get_u32_le();
+                        if cp.nextOid != next_oid {
+                            cp.nextOid = next_oid;
+                            self.checkpoint_modified = true;
+                        }
+                    } else if info == pg_constants::XLOG_CHECKPOINT_ONLINE
+                        || info == pg_constants::XLOG_CHECKPOINT_SHUTDOWN
+                    {
+                        let mut checkpoint_bytes = [0u8; pgv::xlog_utils::SIZEOF_CHECKPOINT];
+                        buf.copy_to_slice(&mut checkpoint_bytes);
+                        let xlog_checkpoint = pgv::CheckPoint::decode(&checkpoint_bytes)?;
+                        trace!(
+                            "xlog_checkpoint.oldestXid={}, checkpoint.oldestXid={}",
+                            xlog_checkpoint.oldestXid,
+                            cp.oldestXid
+                        );
+                        if (cp.oldestXid.wrapping_sub(xlog_checkpoint.oldestXid) as i32) < 0 {
+                            cp.oldestXid = xlog_checkpoint.oldestXid;
+                        }
+                        trace!(
+                            "xlog_checkpoint.oldestActiveXid={}, checkpoint.oldestActiveXid={}",
+                            xlog_checkpoint.oldestActiveXid,
+                            cp.oldestActiveXid
+                        );
+
+                        // A shutdown checkpoint has `oldestActiveXid == InvalidTransactionid`,
+                        // because at shutdown, all in-progress transactions will implicitly
+                        // end. Postgres startup code knows that, and allows hot standby to start
+                        // immediately from a shutdown checkpoint.
+                        //
+                        // In Neon, Postgres hot standby startup always behaves as if starting from
+                        // an online checkpoint. It needs a valid `oldestActiveXid` value, so
+                        // instead of overwriting self.checkpoint.oldestActiveXid with
+                        // InvalidTransactionid from the checkpoint WAL record, update it to a
+                        // proper value, knowing that there are no in-progress transactions at this
+                        // point, except for prepared transactions.
+                        //
+                        // See also the neon code changes in the InitWalRecovery() function.
+                        if xlog_checkpoint.oldestActiveXid == pg_constants::INVALID_TRANSACTION_ID
+                            && info == pg_constants::XLOG_CHECKPOINT_SHUTDOWN
+                        {
+                            let oldest_active_xid = if pg_version >= 17 {
+                                let mut oldest_active_full_xid = cp.nextXid.value;
+                                for xid in modification.tline.list_twophase_files(lsn, ctx).await? {
+                                    if xid < oldest_active_full_xid {
+                                        oldest_active_full_xid = xid;
+                                    }
+                                }
+                                oldest_active_full_xid as u32
+                            } else {
+                                let mut oldest_active_xid = cp.nextXid.value as u32;
+                                for xid in modification.tline.list_twophase_files(lsn, ctx).await? {
+                                    let narrow_xid = xid as u32;
+                                    if (narrow_xid.wrapping_sub(oldest_active_xid) as i32) < 0 {
+                                        oldest_active_xid = narrow_xid;
+                                    }
+                                }
+                                oldest_active_xid
+                            };
+                            cp.oldestActiveXid = oldest_active_xid;
+                        } else {
+                            cp.oldestActiveXid = xlog_checkpoint.oldestActiveXid;
+                        }
+
+                        // Write a new checkpoint key-value pair on every checkpoint record, even
+                        // if nothing really changed. Not strictly required, but it seems nice to
+                        // have some trace of the checkpoint records in the layer files at the same
+                        // LSNs.
+                        self.checkpoint_modified = true;
+                    }
+                });
             }
             pg_constants::RM_LOGICALMSG_ID => {
                 let info = decoded.xl_info & pg_constants::XLR_RMGR_INFO_MASK;
@@ -424,7 +505,11 @@ impl WalIngest {
                 let info = decoded.xl_info & pg_constants::XLR_RMGR_INFO_MASK;
                 if info == pg_constants::XLOG_RUNNING_XACTS {
                     let xlrec = crate::walrecord::XlRunningXacts::decode(&mut buf);
-                    self.checkpoint.oldestActiveXid = xlrec.oldest_running_xid;
+
+                    enum_pgversion_dispatch!(&mut self.checkpoint, CheckPoint, cp, {
+                        cp.oldestActiveXid = xlrec.oldest_running_xid;
+                    });
+
                     self.checkpoint_modified = true;
                 }
             }
@@ -497,6 +582,25 @@ impl WalIngest {
         Ok(modification.len() > prev_len)
     }
 
+    /// This is the same as AdjustToFullTransactionId(xid) in PostgreSQL
+    fn adjust_to_full_transaction_id(&self, xid: TransactionId) -> Result<u64> {
+        let next_full_xid =
+            enum_pgversion_dispatch!(&self.checkpoint, CheckPoint, cp, { cp.nextXid.value });
+
+        let next_xid = (next_full_xid) as u32;
+        let mut epoch = (next_full_xid >> 32) as u32;
+
+        if xid > next_xid {
+            // Wraparound occurred, must be from a prev epoch.
+            if epoch == 0 {
+                bail!("apparent XID wraparound with prepared transaction XID {xid}, nextXid is {next_full_xid}");
+            }
+            epoch -= 1;
+        }
+
+        Ok((epoch as u64) << 32 | xid as u64)
+    }
+
     /// Do not store this block, but observe it for the purposes of updating our relation size state.
     async fn observe_decoded_block(
         &mut self,
@@ -539,7 +643,7 @@ impl WalIngest {
             && blk.has_image
             && decoded.xl_rmid == pg_constants::RM_XLOG_ID
             && (decoded.xl_info == pg_constants::XLOG_FPI
-                || decoded.xl_info == pg_constants::XLOG_FPI_FOR_HINT)
+            || decoded.xl_info == pg_constants::XLOG_FPI_FOR_HINT)
             // compression of WAL is not yet supported: fall back to storing the original WAL record
             && !postgres_ffi::bkpimage_is_compressed(blk.bimg_info, modification.tline.pg_version)
             // do not materialize null pages because them most likely be soon replaced with real data
@@ -797,6 +901,73 @@ impl WalIngest {
                     bail!("Unknown RMGR {} for Heap decoding", decoded.xl_rmid);
                 }
             }
+            17 => {
+                if decoded.xl_rmid == pg_constants::RM_HEAP_ID {
+                    let info = decoded.xl_info & pg_constants::XLOG_HEAP_OPMASK;
+
+                    if info == pg_constants::XLOG_HEAP_INSERT {
+                        let xlrec = v17::XlHeapInsert::decode(buf);
+                        assert_eq!(0, buf.remaining());
+                        if (xlrec.flags & pg_constants::XLH_INSERT_ALL_VISIBLE_CLEARED) != 0 {
+                            new_heap_blkno = Some(decoded.blocks[0].blkno);
+                        }
+                    } else if info == pg_constants::XLOG_HEAP_DELETE {
+                        let xlrec = v17::XlHeapDelete::decode(buf);
+                        if (xlrec.flags & pg_constants::XLH_DELETE_ALL_VISIBLE_CLEARED) != 0 {
+                            new_heap_blkno = Some(decoded.blocks[0].blkno);
+                        }
+                    } else if info == pg_constants::XLOG_HEAP_UPDATE
+                        || info == pg_constants::XLOG_HEAP_HOT_UPDATE
+                    {
+                        let xlrec = v17::XlHeapUpdate::decode(buf);
+                        // the size of tuple data is inferred from the size of the record.
+                        // we can't validate the remaining number of bytes without parsing
+                        // the tuple data.
+                        if (xlrec.flags & pg_constants::XLH_UPDATE_OLD_ALL_VISIBLE_CLEARED) != 0 {
+                            old_heap_blkno = Some(decoded.blocks.last().unwrap().blkno);
+                        }
+                        if (xlrec.flags & pg_constants::XLH_UPDATE_NEW_ALL_VISIBLE_CLEARED) != 0 {
+                            // PostgreSQL only uses XLH_UPDATE_NEW_ALL_VISIBLE_CLEARED on a
+                            // non-HOT update where the new tuple goes to different page than
+                            // the old one. Otherwise, only XLH_UPDATE_OLD_ALL_VISIBLE_CLEARED is
+                            // set.
+                            new_heap_blkno = Some(decoded.blocks[0].blkno);
+                        }
+                    } else if info == pg_constants::XLOG_HEAP_LOCK {
+                        let xlrec = v17::XlHeapLock::decode(buf);
+                        if (xlrec.flags & pg_constants::XLH_LOCK_ALL_FROZEN_CLEARED) != 0 {
+                            old_heap_blkno = Some(decoded.blocks[0].blkno);
+                            flags = pg_constants::VISIBILITYMAP_ALL_FROZEN;
+                        }
+                    }
+                } else if decoded.xl_rmid == pg_constants::RM_HEAP2_ID {
+                    let info = decoded.xl_info & pg_constants::XLOG_HEAP_OPMASK;
+                    if info == pg_constants::XLOG_HEAP2_MULTI_INSERT {
+                        let xlrec = v17::XlHeapMultiInsert::decode(buf);
+
+                        let offset_array_len =
+                            if decoded.xl_info & pg_constants::XLOG_HEAP_INIT_PAGE > 0 {
+                                // the offsets array is omitted if XLOG_HEAP_INIT_PAGE is set
+                                0
+                            } else {
+                                size_of::<u16>() * xlrec.ntuples as usize
+                            };
+                        assert_eq!(offset_array_len, buf.remaining());
+
+                        if (xlrec.flags & pg_constants::XLH_INSERT_ALL_VISIBLE_CLEARED) != 0 {
+                            new_heap_blkno = Some(decoded.blocks[0].blkno);
+                        }
+                    } else if info == pg_constants::XLOG_HEAP2_LOCK_UPDATED {
+                        let xlrec = v17::XlHeapLockUpdated::decode(buf);
+                        if (xlrec.flags & pg_constants::XLH_LOCK_ALL_FROZEN_CLEARED) != 0 {
+                            old_heap_blkno = Some(decoded.blocks[0].blkno);
+                            flags = pg_constants::VISIBILITYMAP_ALL_FROZEN;
+                        }
+                    }
+                } else {
+                    bail!("Unknown RMGR {} for Heap decoding", decoded.xl_rmid);
+                }
+            }
             _ => {}
         }
 
@@ -905,26 +1076,26 @@ impl WalIngest {
         assert_eq!(decoded.xl_rmid, pg_constants::RM_NEON_ID);
 
         match pg_version {
-            16 => {
+            16 | 17 => {
                 let info = decoded.xl_info & pg_constants::XLOG_HEAP_OPMASK;
 
                 match info {
                     pg_constants::XLOG_NEON_HEAP_INSERT => {
-                        let xlrec = v16::rm_neon::XlNeonHeapInsert::decode(buf);
+                        let xlrec = v17::rm_neon::XlNeonHeapInsert::decode(buf);
                         assert_eq!(0, buf.remaining());
                         if (xlrec.flags & pg_constants::XLH_INSERT_ALL_VISIBLE_CLEARED) != 0 {
                             new_heap_blkno = Some(decoded.blocks[0].blkno);
                         }
                     }
                     pg_constants::XLOG_NEON_HEAP_DELETE => {
-                        let xlrec = v16::rm_neon::XlNeonHeapDelete::decode(buf);
+                        let xlrec = v17::rm_neon::XlNeonHeapDelete::decode(buf);
                         if (xlrec.flags & pg_constants::XLH_DELETE_ALL_VISIBLE_CLEARED) != 0 {
                             new_heap_blkno = Some(decoded.blocks[0].blkno);
                         }
                     }
                     pg_constants::XLOG_NEON_HEAP_UPDATE
                     | pg_constants::XLOG_NEON_HEAP_HOT_UPDATE => {
-                        let xlrec = v16::rm_neon::XlNeonHeapUpdate::decode(buf);
+                        let xlrec = v17::rm_neon::XlNeonHeapUpdate::decode(buf);
                         // the size of tuple data is inferred from the size of the record.
                         // we can't validate the remaining number of bytes without parsing
                         // the tuple data.
@@ -940,7 +1111,7 @@ impl WalIngest {
                         }
                     }
                     pg_constants::XLOG_NEON_HEAP_MULTI_INSERT => {
-                        let xlrec = v16::rm_neon::XlNeonHeapMultiInsert::decode(buf);
+                        let xlrec = v17::rm_neon::XlNeonHeapMultiInsert::decode(buf);
 
                         let offset_array_len =
                             if decoded.xl_info & pg_constants::XLOG_HEAP_INIT_PAGE > 0 {
@@ -956,7 +1127,7 @@ impl WalIngest {
                         }
                     }
                     pg_constants::XLOG_NEON_HEAP_LOCK => {
-                        let xlrec = v16::rm_neon::XlNeonHeapLock::decode(buf);
+                        let xlrec = v17::rm_neon::XlNeonHeapLock::decode(buf);
                         if (xlrec.flags & pg_constants::XLH_LOCK_ALL_FROZEN_CLEARED) != 0 {
                             old_heap_blkno = Some(decoded.blocks[0].blkno);
                             flags = pg_constants::VISIBILITYMAP_ALL_FROZEN;
@@ -1204,7 +1375,7 @@ impl WalIngest {
             if rec.blkno % pg_constants::SLOTS_PER_FSM_PAGE != 0 {
                 // Tail of last remaining FSM page has to be zeroed.
                 // We are not precise here and instead of digging in FSM bitmap format just clear the whole page.
-                modification.put_rel_page_image_zero(rel, fsm_physical_page_no);
+                modification.put_rel_page_image_zero(rel, fsm_physical_page_no)?;
                 fsm_physical_page_no += 1;
             }
             let nblocks = get_relsize(modification, rel, ctx).await?;
@@ -1226,7 +1397,7 @@ impl WalIngest {
             if rec.blkno % pg_constants::VM_HEAPBLOCKS_PER_PAGE != 0 {
                 // Tail of last remaining vm page has to be zeroed.
                 // We are not precise here and instead of digging in VM bitmap format just clear the whole page.
-                modification.put_rel_page_image_zero(rel, vm_page_no);
+                modification.put_rel_page_image_zero(rel, vm_page_no)?;
                 vm_page_no += 1;
             }
             let nblocks = get_relsize(modification, rel, ctx).await?;
@@ -1242,12 +1413,17 @@ impl WalIngest {
     fn warn_on_ingest_lag(
         &mut self,
         conf: &crate::config::PageServerConf,
-        wal_timestmap: TimestampTz,
+        wal_timestamp: TimestampTz,
     ) {
         debug_assert_current_span_has_tenant_and_timeline_id();
         let now = SystemTime::now();
         let rate_limits = &mut self.warn_ingest_lag;
-        match try_from_pg_timestamp(wal_timestmap) {
+
+        let ts = enum_pgversion_dispatch!(&self.checkpoint, CheckPoint, _cp, {
+            pgv::xlog_utils::try_from_pg_timestamp(wal_timestamp)
+        });
+
+        match ts {
             Ok(ts) => {
                 match now.duration_since(ts) {
                     Ok(lag) => {
@@ -1257,7 +1433,7 @@ impl WalIngest {
                                 warn!(%rate_limit_stats, %lag, "ingesting record with timestamp lagging more than wait_lsn_timeout");
                             })
                         }
-                    },
+                    }
                     Err(e) => {
                         let delta_t = e.duration();
                         // determined by prod victoriametrics query: 1000 * (timestamp(node_time_seconds{neon_service="pageserver"}) - node_time_seconds)
@@ -1271,7 +1447,6 @@ impl WalIngest {
                         }
                     }
                 };
-
             }
             Err(error) => {
                 rate_limits.timestamp_invalid_msg_ratelimit.call2(|rate_limit_stats| {
@@ -1379,14 +1554,17 @@ impl WalIngest {
         // truncated, but a checkpoint record with the updated values isn't written until
         // later. In Neon, a server can start at any LSN, not just on a checkpoint record,
         // so we keep the oldestXid and oldestXidDB up-to-date.
-        self.checkpoint.oldestXid = xlrec.oldest_xid;
-        self.checkpoint.oldestXidDB = xlrec.oldest_xid_db;
+        enum_pgversion_dispatch!(&mut self.checkpoint, CheckPoint, cp, {
+            cp.oldestXid = xlrec.oldest_xid;
+            cp.oldestXidDB = xlrec.oldest_xid_db;
+        });
         self.checkpoint_modified = true;
 
         // TODO Treat AdvanceOldestClogXid() or write a comment why we don't need it
 
         let latest_page_number =
-            self.checkpoint.nextXid.value as u32 / pg_constants::CLOG_XACTS_PER_PAGE;
+            enum_pgversion_dispatch!(self.checkpoint, CheckPoint, cp, { cp.nextXid.value }) as u32
+                / pg_constants::CLOG_XACTS_PER_PAGE;
 
         // Now delete all segments containing pages between xlrec.pageno
         // and latest_page_number.
@@ -1394,7 +1572,9 @@ impl WalIngest {
         // First, make an important safety check:
         // the current endpoint page must not be eligible for removal.
         // See SimpleLruTruncate() in slru.c
-        if clogpage_precedes(latest_page_number, xlrec.pageno) {
+        if dispatch_pgversion!(modification.tline.pg_version, {
+            pgv::nonrelfile_utils::clogpage_precedes(latest_page_number, xlrec.pageno)
+        }) {
             info!("could not truncate directory pg_xact apparent wraparound");
             return Ok(());
         }
@@ -1411,7 +1591,12 @@ impl WalIngest {
             .await?
         {
             let segpage = segno * pg_constants::SLRU_PAGES_PER_SEGMENT;
-            if slru_may_delete_clogsegment(segpage, xlrec.pageno) {
+
+            let may_delete = dispatch_pgversion!(modification.tline.pg_version, {
+                pgv::nonrelfile_utils::slru_may_delete_clogsegment(segpage, xlrec.pageno)
+            });
+
+            if may_delete {
                 modification
                     .drop_slru_segment(SlruKind::Clog, segno, ctx)
                     .await?;
@@ -1530,14 +1715,23 @@ impl WalIngest {
         xlrec: &XlMultiXactTruncate,
         ctx: &RequestContext,
     ) -> Result<()> {
-        self.checkpoint.oldestMulti = xlrec.end_trunc_off;
-        self.checkpoint.oldestMultiDB = xlrec.oldest_multi_db;
+        let (maxsegment, startsegment, endsegment) =
+            enum_pgversion_dispatch!(&mut self.checkpoint, CheckPoint, cp, {
+                cp.oldestMulti = xlrec.end_trunc_off;
+                cp.oldestMultiDB = xlrec.oldest_multi_db;
+                let maxsegment: i32 = pgv::nonrelfile_utils::mx_offset_to_member_segment(
+                    pg_constants::MAX_MULTIXACT_OFFSET,
+                );
+                let startsegment: i32 =
+                    pgv::nonrelfile_utils::mx_offset_to_member_segment(xlrec.start_trunc_memb);
+                let endsegment: i32 =
+                    pgv::nonrelfile_utils::mx_offset_to_member_segment(xlrec.end_trunc_memb);
+                (maxsegment, startsegment, endsegment)
+            });
+
         self.checkpoint_modified = true;
 
         // PerformMembersTruncation
-        let maxsegment: i32 = mx_offset_to_member_segment(pg_constants::MAX_MULTIXACT_OFFSET);
-        let startsegment: i32 = mx_offset_to_member_segment(xlrec.start_trunc_memb);
-        let endsegment: i32 = mx_offset_to_member_segment(xlrec.end_trunc_memb);
         let mut segment: i32 = startsegment;
 
         // Delete all the segments except the last one. The last segment can still
@@ -1696,7 +1890,7 @@ impl WalIngest {
                     continue;
                 }
 
-                modification.put_rel_page_image_zero(rel, gap_blknum);
+                modification.put_rel_page_image_zero(rel, gap_blknum)?;
             }
         }
         Ok(())
@@ -1762,7 +1956,7 @@ impl WalIngest {
 
             // fill the gap with zeros
             for gap_blknum in old_nblocks..blknum {
-                modification.put_slru_page_image_zero(kind, segno, gap_blknum);
+                modification.put_slru_page_image_zero(kind, segno, gap_blknum)?;
             }
         }
         Ok(())
@@ -1811,11 +2005,23 @@ mod tests {
         // TODO
     }
 
-    static ZERO_CHECKPOINT: Bytes = Bytes::from_static(&[0u8; SIZEOF_CHECKPOINT]);
+    #[tokio::test]
+    async fn test_zeroed_checkpoint_decodes_correctly() -> Result<()> {
+        for i in 14..=16 {
+            dispatch_pgversion!(i, {
+                pgv::CheckPoint::decode(&pgv::ZERO_CHECKPOINT)?;
+            });
+        }
+
+        Ok(())
+    }
 
     async fn init_walingest_test(tline: &Timeline, ctx: &RequestContext) -> Result<WalIngest> {
         let mut m = tline.begin_modification(Lsn(0x10));
-        m.put_checkpoint(ZERO_CHECKPOINT.clone())?;
+        m.put_checkpoint(dispatch_pgversion!(
+            tline.pg_version,
+            pgv::ZERO_CHECKPOINT.clone()
+        ))?;
         m.put_relmap_file(0, 111, Bytes::from(""), ctx).await?; // dummy relmapper file
         m.commit(ctx).await?;
         let walingest = WalIngest::new(tline, Lsn(0x10), ctx).await?;

pageserver/src/walrecord.rs

@@ -174,6 +174,7 @@ impl DecodedWALRecord {
                 }
                 15 => info == postgres_ffi::v15::bindings::XLOG_DBASE_CREATE_FILE_COPY,
                 16 => info == postgres_ffi::v16::bindings::XLOG_DBASE_CREATE_FILE_COPY,
+                17 => info == postgres_ffi::v17::bindings::XLOG_DBASE_CREATE_FILE_COPY,
                 _ => {
                     panic!("Unsupported postgres version {pg_version}")
                 }
@@ -341,16 +342,47 @@ pub mod v14 {
             }
         }
     }
+
+    #[repr(C)]
+    #[derive(Debug)]
+    pub struct XlParameterChange {
+        pub max_connections: i32,
+        pub max_worker_processes: i32,
+        pub max_wal_senders: i32,
+        pub max_prepared_xacts: i32,
+        pub max_locks_per_xact: i32,
+        pub wal_level: i32,
+        pub wal_log_hints: bool,
+        pub track_commit_timestamp: bool,
+        pub _padding: [u8; 2],
+    }
+
+    impl XlParameterChange {
+        pub fn decode(buf: &mut Bytes) -> XlParameterChange {
+            XlParameterChange {
+                max_connections: buf.get_i32_le(),
+                max_worker_processes: buf.get_i32_le(),
+                max_wal_senders: buf.get_i32_le(),
+                max_prepared_xacts: buf.get_i32_le(),
+                max_locks_per_xact: buf.get_i32_le(),
+                wal_level: buf.get_i32_le(),
+                wal_log_hints: buf.get_u8() != 0,
+                track_commit_timestamp: buf.get_u8() != 0,
+                _padding: [buf.get_u8(), buf.get_u8()],
+            }
+        }
+    }
 }
 
 pub mod v15 {
     pub use super::v14::{
         XlHeapDelete, XlHeapInsert, XlHeapLock, XlHeapLockUpdated, XlHeapMultiInsert, XlHeapUpdate,
+        XlParameterChange,
     };
 }
 
 pub mod v16 {
-    pub use super::v14::{XlHeapInsert, XlHeapLockUpdated, XlHeapMultiInsert};
+    pub use super::v14::{XlHeapInsert, XlHeapLockUpdated, XlHeapMultiInsert, XlParameterChange};
     use bytes::{Buf, Bytes};
     use postgres_ffi::{OffsetNumber, TransactionId};
 
@@ -529,6 +561,37 @@ pub mod v16 {
     }
 }
 
+pub mod v17 {
+    pub use super::v14::XlHeapLockUpdated;
+    use bytes::{Buf, Bytes};
+    pub use postgres_ffi::{TimeLineID, TimestampTz};
+
+    pub use super::v16::rm_neon;
+    pub use super::v16::{
+        XlHeapDelete, XlHeapInsert, XlHeapLock, XlHeapMultiInsert, XlHeapUpdate, XlParameterChange,
+    };
+
+    #[repr(C)]
+    #[derive(Debug)]
+    pub struct XlEndOfRecovery {
+        pub end_time: TimestampTz,
+        pub this_time_line_id: TimeLineID,
+        pub prev_time_line_id: TimeLineID,
+        pub wal_level: i32,
+    }
+
+    impl XlEndOfRecovery {
+        pub fn decode(buf: &mut Bytes) -> XlEndOfRecovery {
+            XlEndOfRecovery {
+                end_time: buf.get_i64_le(),
+                this_time_line_id: buf.get_u32_le(),
+                prev_time_line_id: buf.get_u32_le(),
+                wal_level: buf.get_i32_le(),
+            }
+        }
+    }
+}
+
 #[repr(C)]
 #[derive(Debug)]
 pub struct XlSmgrCreate {
@@ -746,9 +809,13 @@ pub struct XlClogTruncate {
 }
 
 impl XlClogTruncate {
-    pub fn decode(buf: &mut Bytes) -> XlClogTruncate {
+    pub fn decode(buf: &mut Bytes, pg_version: u32) -> XlClogTruncate {
         XlClogTruncate {
-            pageno: buf.get_u32_le(),
+            pageno: if pg_version < 17 {
+                buf.get_u32_le()
+            } else {
+                buf.get_u64_le() as u32
+            },
             oldest_xid: buf.get_u32_le(),
             oldest_xid_db: buf.get_u32_le(),
         }

pageserver/src/walredo.rs

@@ -43,13 +43,12 @@ use utils::lsn::Lsn;
 use utils::sync::gate::GateError;
 use utils::sync::heavier_once_cell;
 
+/// The real implementation that uses a Postgres process to
+/// perform WAL replay.
 ///
-/// This is the real implementation that uses a Postgres process to
-/// perform WAL replay. Only one thread can use the process at a time,
-/// that is controlled by the Mutex. In the future, we might want to
-/// launch a pool of processes to allow concurrent replay of multiple
-/// records.
-///
+/// Only one thread can use the process at a time, that is controlled by the
+/// Mutex. In the future, we might want to launch a pool of processes to allow
+/// concurrent replay of multiple records.
 pub struct PostgresRedoManager {
     tenant_shard_id: TenantShardId,
     conf: &'static PageServerConf,

pgxn/neon/bitmap.h

@@ -0,0 +1,12 @@
+#ifndef NEON_BITMAP_H
+#define NEON_BITMAP_H
+
+/*
+ * Utilities for manipulating bits8* as bitmaps.
+ */
+
+#define BITMAP_ISSET(bm, bit) ((bm)[(bit) >> 3] & (1 << ((bit) & 7)))
+#define BITMAP_SET(bm, bit) (bm)[(bit) >> 3] |= (1 << ((bit) & 7))
+#define BITMAP_CLR(bm, bit) (bm)[(bit) >> 3] &= ~(1 << ((bit) & 7))
+
+#endif //NEON_BITMAP_H

pgxn/neon/file_cache.c

@@ -27,6 +27,7 @@
 #include "pagestore_client.h"
 #include "common/hashfn.h"
 #include "pgstat.h"
+#include "port/pg_iovec.h"
 #include "postmaster/bgworker.h"
 #include RELFILEINFO_HDR
 #include "storage/buf_internals.h"
@@ -40,6 +41,7 @@
 #include "utils/guc.h"
 
 #include "hll.h"
+#include "bitmap.h"
 
 #define CriticalAssert(cond) do if (!(cond)) elog(PANIC, "Assertion %s failed at %s:%d: ", #cond, __FILE__, __LINE__); while (0)
 
@@ -469,6 +471,99 @@ lfc_cache_contains(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno)
 	return found;
 }
 
+/*
+ * Check if page is present in the cache.
+ * Returns true if page is found in local cache.
+ */
+int
+lfc_cache_containsv(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
+					int nblocks, bits8 *bitmap)
+{
+	BufferTag	tag;
+	FileCacheEntry *entry;
+	uint32		chunk_offs;
+	int			found = 0;
+	uint32		hash;
+	int			i = 0;
+
+	if (lfc_maybe_disabled())	/* fast exit if file cache is disabled */
+		return 0;
+
+	CopyNRelFileInfoToBufTag(tag, rinfo);
+	tag.forkNum = forkNum;
+
+	CriticalAssert(BufTagGetRelNumber(&tag) != InvalidRelFileNumber);
+
+	tag.blockNum = (blkno + i) & ~(BLOCKS_PER_CHUNK - 1);
+	hash = get_hash_value(lfc_hash, &tag);
+	chunk_offs = (blkno + i) & (BLOCKS_PER_CHUNK - 1);
+
+	LWLockAcquire(lfc_lock, LW_SHARED);
+
+	while (true)
+	{
+		int		this_chunk = Min(nblocks, BLOCKS_PER_CHUNK - chunk_offs);
+		if (LFC_ENABLED())
+		{
+			entry = hash_search_with_hash_value(lfc_hash, &tag, hash, HASH_FIND, NULL);
+
+			if (entry != NULL)
+			{
+				for (; chunk_offs < BLOCKS_PER_CHUNK && i < nblocks; chunk_offs++, i++)
+				{
+					if ((entry->bitmap[chunk_offs >> 5] & 
+						(1 << (chunk_offs & 31))) != 0)
+					{
+						BITMAP_SET(bitmap, i);
+						found++;
+					}
+				}
+			}
+			else
+			{
+				i += this_chunk;
+			}
+		}
+		else
+		{
+			return found;
+		}
+
+		/*
+		 * Break out of the iteration before doing expensive stuff for
+		 * a next iteration
+		 */
+		if (i + 1 >= nblocks)
+			break;
+
+		/*
+		 * Prepare for the next iteration. We don't unlock here, as that'd
+		 * probably be more expensive than the gains it'd get us.
+		 */
+		tag.blockNum = (blkno + i) & ~(BLOCKS_PER_CHUNK - 1);
+		hash = get_hash_value(lfc_hash, &tag);
+		chunk_offs = (blkno + i) & (BLOCKS_PER_CHUNK - 1);
+	}
+
+	LWLockRelease(lfc_lock);
+
+#if USE_ASSERT_CHECKING
+	do {
+		int count = 0;
+
+		for (int j = 0; j < nblocks; j++)
+		{
+			if (BITMAP_ISSET(bitmap, j))
+				count++;
+		}
+
+		Assert(count == found);
+	} while (false);
+#endif
+
+	return found;
+}
+
 /*
  * Evict a page (if present) from the local file cache
  */
@@ -548,91 +643,171 @@ lfc_evict(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno)
 }
 
 /*
- * Try to read page from local cache.
- * Returns true if page is found in local cache.
- * In case of error local file cache is disabled (lfc->limit is set to zero).
+ * Try to read pages from local cache.
+ * Returns the number of pages read from the local cache, and sets bits in
+ * 'read' for the pages which were read. This may scribble over buffers not
+ * marked in 'read', so be careful with operation ordering.
+ *
+ * In case of error local file cache is disabled (lfc->limit is set to zero),
+ * and -1 is returned. Note that 'read' and the buffers may be touched and in
+ * an otherwise invalid state.
+ *
+ * If the mask argument is supplied, bits will be set at the offsets of pages
+ * that were present and read from the LFC.
  */
-bool
-lfc_read(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
-		 char *buffer)
+int
+lfc_readv_select(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
+				 void **buffers, BlockNumber nblocks, bits8 *mask)
 {
 	BufferTag	tag;
 	FileCacheEntry *entry;
 	ssize_t		rc;
-	int			chunk_offs = blkno & (BLOCKS_PER_CHUNK - 1);
 	bool		result = true;
 	uint32		hash;
 	uint64		generation;
 	uint32		entry_offset;
+	int			blocks_read = 0;
+	int			buf_offset = 0;
 
 	if (lfc_maybe_disabled())	/* fast exit if file cache is disabled */
-		return false;
+		return 0;
 
 	if (!lfc_ensure_opened())
-		return false;
+		return 0;
 
 	CopyNRelFileInfoToBufTag(tag, rinfo);
 	tag.forkNum = forkNum;
-	tag.blockNum = blkno & ~(BLOCKS_PER_CHUNK - 1);
 
 	CriticalAssert(BufTagGetRelNumber(&tag) != InvalidRelFileNumber);
-	hash = get_hash_value(lfc_hash, &tag);
 
-	LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
-
-	if (!LFC_ENABLED())
+	/* 
+	 * For every chunk that has blocks we're interested in, we
+	 * 1. get the chunk header
+	 * 2. Check if the chunk actually has the blocks we're interested in
+	 * 3. Read the blocks we're looking for (in one preadv), assuming they exist
+	 * 4. Update the statistics for the read call.
+	 *
+	 * If there is an error, we do an early return.
+	 */
+	while (nblocks > 0)
 	{
+		struct iovec iov[PG_IOV_MAX];
+		int		chunk_offs = blkno & (BLOCKS_PER_CHUNK - 1);
+		int		blocks_in_chunk = Min(nblocks, BLOCKS_PER_CHUNK - (blkno % BLOCKS_PER_CHUNK));
+		int		iteration_hits = 0;
+		int		iteration_misses = 0;
+		Assert(blocks_in_chunk > 0);
+
+		for (int i = 0; i < blocks_in_chunk; i++)
+		{
+			iov[i].iov_base = buffers[buf_offset + i];
+			iov[i].iov_len = BLCKSZ;
+		}
+
+		tag.blockNum = blkno - chunk_offs;
+		hash = get_hash_value(lfc_hash, &tag);
+
+		LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
+
+		/* We can return the blocks we've read before LFC got disabled;
+		 * assuming we read any. */
+		if (!LFC_ENABLED())
+		{
+			LWLockRelease(lfc_lock);
+			return blocks_read;
+		}
+
+		entry = hash_search_with_hash_value(lfc_hash, &tag, hash, HASH_FIND, NULL);
+
+		/* Approximate working set for the blocks assumed in this entry */
+		for (int i = 0; i < blocks_in_chunk; i++)
+		{
+			tag.blockNum = blkno + i;
+			addSHLL(&lfc_ctl->wss_estimation, hash_bytes((uint8_t const*)&tag, sizeof(tag)));
+		}
+
+		if (entry == NULL)
+		{
+			/* Pages are not cached */
+			lfc_ctl->misses += blocks_in_chunk;
+			pgBufferUsage.file_cache.misses += blocks_in_chunk;
+			LWLockRelease(lfc_lock);
+
+			buf_offset += blocks_in_chunk;
+			nblocks -= blocks_in_chunk;
+			blkno += blocks_in_chunk;
+
+			continue;
+		}
+
+		/* Unlink entry from LRU list to pin it for the duration of IO operation */
+		if (entry->access_count++ == 0)
+			dlist_delete(&entry->list_node);
+
+		generation = lfc_ctl->generation;
+		entry_offset = entry->offset;
+
 		LWLockRelease(lfc_lock);
-		return false;
-	}
 
-	entry = hash_search_with_hash_value(lfc_hash, &tag, hash, HASH_FIND, NULL);
+		for (int i = 0; i < blocks_in_chunk; i++)
+		{
+			/*
+			 * If the page is valid, we consider it "read".
+			 * All other pages will be fetched separately by the next cache
+			 */
+			if (entry->bitmap[(chunk_offs + i) / 32] & (1 << ((chunk_offs + i) % 32)))
+			{
+				BITMAP_SET(mask, buf_offset + i);
+				iteration_hits++;
+			}
+			else
+				iteration_misses++;
+		}
 
-	/* Approximate working set */
-	tag.blockNum = blkno;
-	addSHLL(&lfc_ctl->wss_estimation, hash_bytes((uint8_t const*)&tag, sizeof(tag)));
+		Assert(iteration_hits + iteration_misses > 0);
+
+		if (iteration_hits != 0)
+		{
+			rc = preadv(lfc_desc, iov, blocks_in_chunk,
+						((off_t) entry_offset * BLOCKS_PER_CHUNK + chunk_offs) * BLCKSZ);
+
+			if (rc != (BLCKSZ * blocks_in_chunk))
+			{
+				lfc_disable("read");
+				return -1;
+			}
+		}
+
+		/* Place entry to the head of LRU list */
+		LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
+
+		if (lfc_ctl->generation == generation)
+		{
+			CriticalAssert(LFC_ENABLED());
+			lfc_ctl->hits += iteration_hits;
+			lfc_ctl->misses += iteration_misses;
+			pgBufferUsage.file_cache.hits += iteration_hits;
+			pgBufferUsage.file_cache.misses += iteration_misses;
+			CriticalAssert(entry->access_count > 0);
+			if (--entry->access_count == 0)
+				dlist_push_tail(&lfc_ctl->lru, &entry->list_node);
+		}
+		else
+		{
+			/* generation mismatch, assume error condition */
+			LWLockRelease(lfc_lock);
+			return -1;
+		}
 
-	if (entry == NULL || (entry->bitmap[chunk_offs >> 5] & (1 << (chunk_offs & 31))) == 0)
-	{
-		/* Page is not cached */
-		lfc_ctl->misses += 1;
-		pgBufferUsage.file_cache.misses += 1;
 		LWLockRelease(lfc_lock);
-		return false;
-	}
-	/* Unlink entry from LRU list to pin it for the duration of IO operation */
-	if (entry->access_count++ == 0)
-		dlist_delete(&entry->list_node);
-	generation = lfc_ctl->generation;
-	entry_offset = entry->offset;
 
-	LWLockRelease(lfc_lock);
-
-	rc = pread(lfc_desc, buffer, BLCKSZ, ((off_t) entry_offset * BLOCKS_PER_CHUNK + chunk_offs) * BLCKSZ);
-	if (rc != BLCKSZ)
-	{
-		lfc_disable("read");
-		return false;
+		buf_offset += blocks_in_chunk;
+		nblocks -= blocks_in_chunk;
+		blkno += blocks_in_chunk;
+		blocks_read += iteration_hits;
 	}
 
-	/* Place entry to the head of LRU list */
-	LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
-
-	if (lfc_ctl->generation == generation)
-	{
-		CriticalAssert(LFC_ENABLED());
-		lfc_ctl->hits += 1;
-		pgBufferUsage.file_cache.hits += 1;
-		CriticalAssert(entry->access_count > 0);
-		if (--entry->access_count == 0)
-			dlist_push_tail(&lfc_ctl->lru, &entry->list_node);
-	}
-	else
-		result = false;
-
-	LWLockRelease(lfc_lock);
-
-	return result;
+	return blocks_read;
 }
 
 /*
@@ -640,20 +815,17 @@ lfc_read(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
  * If cache is full then evict some other page.
  */
 void
-#if PG_MAJORVERSION_NUM < 16
-lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno, char *buffer)
-#else
-lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno, const void *buffer)
-#endif
+lfc_writev(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
+		   const void *const *buffers, BlockNumber nblocks)
 {
 	BufferTag	tag;
 	FileCacheEntry *entry;
 	ssize_t		rc;
 	bool		found;
-	int			chunk_offs = blkno & (BLOCKS_PER_CHUNK - 1);
 	uint32		hash;
 	uint64		generation;
 	uint32		entry_offset;
+	int			buf_offset = 0;
 
 	if (lfc_maybe_disabled())	/* fast exit if file cache is disabled */
 		return;
@@ -661,110 +833,142 @@ lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno, const void
 	if (!lfc_ensure_opened())
 		return;
 
-	tag.forkNum = forkNum;
-	tag.blockNum = blkno & ~(BLOCKS_PER_CHUNK - 1);
 	CopyNRelFileInfoToBufTag(tag, rinfo);
+	tag.forkNum = forkNum;
 
 	CriticalAssert(BufTagGetRelNumber(&tag) != InvalidRelFileNumber);
-	hash = get_hash_value(lfc_hash, &tag);
 
-	LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
-
-	if (!LFC_ENABLED())
+	/* 
+	 * For every chunk that has blocks we're interested in, we
+	 * 1. get the chunk header
+	 * 2. Check if the chunk actually has the blocks we're interested in
+	 * 3. Read the blocks we're looking for (in one preadv), assuming they exist
+	 * 4. Update the statistics for the read call.
+	 *
+	 * If there is an error, we do an early return.
+	 */
+	while (nblocks > 0)
 	{
-		LWLockRelease(lfc_lock);
-		return;
-	}
+		struct iovec iov[PG_IOV_MAX];
+		int		chunk_offs = blkno & (BLOCKS_PER_CHUNK - 1);
+		int		blocks_in_chunk = Min(nblocks, BLOCKS_PER_CHUNK - (blkno % BLOCKS_PER_CHUNK));
+		Assert(blocks_in_chunk > 0);
 
-	entry = hash_search_with_hash_value(lfc_hash, &tag, hash, HASH_ENTER, &found);
-
-	if (found)
-	{
-		/*
-		 * Unlink entry from LRU list to pin it for the duration of IO
-		 * operation
-		 */
-		if (entry->access_count++ == 0)
-			dlist_delete(&entry->list_node);
-	}
-	else
-	{
-		/*
-		 * We have two choices if all cache pages are pinned (i.e. used in IO
-		 * operations):
-		 *
-		 * 1) Wait until some of this operation is completed and pages is
-		 * unpinned.
-		 *
-		 * 2) Allocate one more chunk, so that specified cache size is more
-		 * recommendation than hard limit.
-		 *
-		 * As far as probability of such event (that all pages are pinned) is
-		 * considered to be very very small: there are should be very large
-		 * number of concurrent IO operations and them are limited by
-		 * max_connections, we prefer not to complicate code and use second
-		 * approach.
-		 */
-		if (lfc_ctl->used >= lfc_ctl->limit && !dlist_is_empty(&lfc_ctl->lru))
+		for (int i = 0; i < blocks_in_chunk; i++)
 		{
-			/* Cache overflow: evict least recently used chunk */
-			FileCacheEntry *victim = dlist_container(FileCacheEntry, list_node, dlist_pop_head_node(&lfc_ctl->lru));
-
-			CriticalAssert(victim->access_count == 0);
-			entry->offset = victim->offset; /* grab victim's chunk */
-			hash_search_with_hash_value(lfc_hash, &victim->key, victim->hash, HASH_REMOVE, NULL);
-			neon_log(DEBUG2, "Swap file cache page");
+			iov[i].iov_base = unconstify(void *, buffers[buf_offset + i]);
+			iov[i].iov_len = BLCKSZ;
 		}
-		else if (!dlist_is_empty(&lfc_ctl->holes))
+
+		tag.blockNum = blkno & ~(BLOCKS_PER_CHUNK - 1);
+		hash = get_hash_value(lfc_hash, &tag);
+
+		LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
+
+		if (!LFC_ENABLED())
 		{
-			/* We can reuse a hole that was left behind when the LFC was shrunk previously */
-			FileCacheEntry *hole = dlist_container(FileCacheEntry, list_node, dlist_pop_head_node(&lfc_ctl->holes));
-			uint32		offset = hole->offset;
-			bool		found;
+			LWLockRelease(lfc_lock);
+			return;
+		}
 
-			hash_search_with_hash_value(lfc_hash, &hole->key, hole->hash, HASH_REMOVE, &found);
-			CriticalAssert(found);
+		entry = hash_search_with_hash_value(lfc_hash, &tag, hash, HASH_ENTER, &found);
 
-			lfc_ctl->used += 1;
-			entry->offset = offset;	/* reuse the hole */
+		if (found)
+		{
+			/*
+			 * Unlink entry from LRU list to pin it for the duration of IO
+			 * operation
+			 */
+			if (entry->access_count++ == 0)
+				dlist_delete(&entry->list_node);
 		}
 		else
 		{
-			lfc_ctl->used += 1;
-			entry->offset = lfc_ctl->size++;	/* allocate new chunk at end
-												 * of file */
-		}
-		entry->access_count = 1;
-		entry->hash = hash;
-		memset(entry->bitmap, 0, sizeof entry->bitmap);
-	}
-
-	generation = lfc_ctl->generation;
-	entry_offset = entry->offset;
-	lfc_ctl->writes += 1;
-	LWLockRelease(lfc_lock);
-
-	rc = pwrite(lfc_desc, buffer, BLCKSZ, ((off_t) entry_offset * BLOCKS_PER_CHUNK + chunk_offs) * BLCKSZ);
-	if (rc != BLCKSZ)
-	{
-		lfc_disable("write");
-	}
-	else
-	{
-		LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
-
-		if (lfc_ctl->generation == generation)
-		{
-			CriticalAssert(LFC_ENABLED());
-			/* Place entry to the head of LRU list */
-			CriticalAssert(entry->access_count > 0);
-			if (--entry->access_count == 0)
-				dlist_push_tail(&lfc_ctl->lru, &entry->list_node);
-
-			entry->bitmap[chunk_offs >> 5] |= (1 << (chunk_offs & 31));
+			/*
+			 * We have two choices if all cache pages are pinned (i.e. used in IO
+			 * operations):
+			 *
+			 * 1) Wait until some of this operation is completed and pages is
+			 * unpinned.
+			 *
+			 * 2) Allocate one more chunk, so that specified cache size is more
+			 * recommendation than hard limit.
+			 *
+			 * As far as probability of such event (that all pages are pinned) is
+			 * considered to be very very small: there are should be very large
+			 * number of concurrent IO operations and them are limited by
+			 * max_connections, we prefer not to complicate code and use second
+			 * approach.
+			 */
+			if (lfc_ctl->used >= lfc_ctl->limit && !dlist_is_empty(&lfc_ctl->lru))
+			{
+				/* Cache overflow: evict least recently used chunk */
+				FileCacheEntry *victim = dlist_container(FileCacheEntry, list_node, dlist_pop_head_node(&lfc_ctl->lru));
+	
+				CriticalAssert(victim->access_count == 0);
+				entry->offset = victim->offset; /* grab victim's chunk */
+				hash_search_with_hash_value(lfc_hash, &victim->key, victim->hash, HASH_REMOVE, NULL);
+				neon_log(DEBUG2, "Swap file cache page");
+			}
+			else if (!dlist_is_empty(&lfc_ctl->holes))
+			{
+				/* We can reuse a hole that was left behind when the LFC was shrunk previously */
+				FileCacheEntry *hole = dlist_container(FileCacheEntry, list_node, dlist_pop_head_node(&lfc_ctl->holes));
+				uint32		offset = hole->offset;
+				bool		found;
+	
+				hash_search_with_hash_value(lfc_hash, &hole->key, hole->hash, HASH_REMOVE, &found);
+				CriticalAssert(found);
+	
+				lfc_ctl->used += 1;
+				entry->offset = offset;	/* reuse the hole */
+			}
+			else
+			{
+				lfc_ctl->used += 1;
+				entry->offset = lfc_ctl->size++;	/* allocate new chunk at end
+													 * of file */
+			}
+			entry->access_count = 1;
+			entry->hash = hash;
+			memset(entry->bitmap, 0, sizeof entry->bitmap);
 		}
 
+		generation = lfc_ctl->generation;
+		entry_offset = entry->offset;
+		lfc_ctl->writes += blocks_in_chunk;
 		LWLockRelease(lfc_lock);
+
+		rc = pwritev(lfc_desc, iov, blocks_in_chunk,
+					 ((off_t) entry_offset * BLOCKS_PER_CHUNK + chunk_offs) * BLCKSZ);
+		if (rc != BLCKSZ * blocks_in_chunk)
+		{
+			lfc_disable("write");
+		}
+		else
+		{
+			LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
+
+			if (lfc_ctl->generation == generation)
+			{
+				CriticalAssert(LFC_ENABLED());
+				/* Place entry to the head of LRU list */
+				CriticalAssert(entry->access_count > 0);
+				if (--entry->access_count == 0)
+					dlist_push_tail(&lfc_ctl->lru, &entry->list_node);
+
+				for (int i = 0; i < blocks_in_chunk; i++)
+				{
+					entry->bitmap[(chunk_offs + i) >> 5] |=
+						(1 << ((chunk_offs + i) & 31));
+				}
+			}
+
+			LWLockRelease(lfc_lock);
+		}
+		blkno += blocks_in_chunk;
+		buf_offset += blocks_in_chunk;
+		nblocks -= blocks_in_chunk;
 	}
 }
 

pgxn/neon/libpagestore.c

@@ -537,7 +537,11 @@ pageserver_connect(shardno_t shard_no, int elevel)
 		/* No more polling needed; connection succeeded */
 		shard->last_connect_time = GetCurrentTimestamp();
 
+#if PG_MAJORVERSION_NUM >= 17
+		shard->wes_read = CreateWaitEventSet(NULL, 3);
+#else
 		shard->wes_read = CreateWaitEventSet(TopMemoryContext, 3);
+#endif
 		AddWaitEventToSet(shard->wes_read, WL_LATCH_SET, PGINVALID_SOCKET,
 						  MyLatch, NULL);
 		AddWaitEventToSet(shard->wes_read, WL_EXIT_ON_PM_DEATH, PGINVALID_SOCKET,

pgxn/neon/neon_pgversioncompat.h

@@ -6,7 +6,11 @@
 #ifndef NEON_PGVERSIONCOMPAT_H
 #define NEON_PGVERSIONCOMPAT_H
 
+#if PG_MAJORVERSION_NUM < 17
 #define NRelFileInfoBackendIsTemp(rinfo) (rinfo.backend != InvalidBackendId)
+#else
+#define NRelFileInfoBackendIsTemp(rinfo) (rinfo.backend != INVALID_PROC_NUMBER)
+#endif
 
 #define RelFileInfoEquals(a, b) ( \
 	NInfoGetSpcOid(a) == NInfoGetSpcOid(b) && \
@@ -50,7 +54,7 @@
 #define CopyNRelFileInfoToBufTag(tag, rinfo) \
 	do { \
 		(tag).rnode = (rinfo); \
-	} while (false);
+	} while (false)
 
 #define BufTagGetNRelFileInfo(tag) tag.rnode
 
@@ -98,7 +102,7 @@
 		(tag).spcOid = (rinfo).spcOid; \
 		(tag).dbOid = (rinfo).dbOid; \
 		(tag).relNumber = (rinfo).relNumber; \
-	} while (false);
+	} while (false)
 
 #define BufTagGetNRelFileInfo(tag) \
 	((RelFileLocator) { \
@@ -113,4 +117,10 @@
 #define DropRelationAllLocalBuffers DropRelationAllLocalBuffers
 #endif
 
+#if PG_MAJORVERSION_NUM < 17
+#define ProcNumber BackendId
+#define INVALID_PROC_NUMBER InvalidBackendId
+#define AmAutoVacuumWorkerProcess() (IsAutoVacuumWorkerProcess())
+#endif
+
 #endif							/* NEON_PGVERSIONCOMPAT_H */

pgxn/neon/pagestore_client.h

@@ -6,8 +6,6 @@
  * Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * contrib/neon/pagestore_client.h
- *
  *-------------------------------------------------------------------------
  */
 #ifndef pageserver_h
@@ -187,7 +185,7 @@ extern char *nm_to_string(NeonMessage *msg);
  * API
  */
 
-typedef unsigned shardno_t;
+typedef uint16 shardno_t;
 
 typedef struct
 {
@@ -211,7 +209,7 @@ extern int  neon_protocol_version;
 
 extern shardno_t get_shard_number(BufferTag* tag);
 
-extern const f_smgr *smgr_neon(BackendId backend, NRelFileInfo rinfo);
+extern const f_smgr *smgr_neon(ProcNumber backend, NRelFileInfo rinfo);
 extern void smgr_init_neon(void);
 extern void readahead_buffer_resize(int newsize, void *extra);
 
@@ -233,8 +231,13 @@ extern void neon_zeroextend(SMgrRelation reln, ForkNumber forknum,
 							BlockNumber blocknum, int nbuffers, bool skipFsync);
 #endif
 
+#if PG_MAJORVERSION_NUM >=17
+extern bool neon_prefetch(SMgrRelation reln, ForkNumber forknum,
+						  BlockNumber blocknum, int nblocks);
+#else
 extern bool neon_prefetch(SMgrRelation reln, ForkNumber forknum,
 						  BlockNumber blocknum);
+#endif
 
 /*
  * LSN values associated with each request to the pageserver
@@ -269,19 +272,11 @@ typedef struct
 } neon_request_lsns;
 
 #if PG_MAJORVERSION_NUM < 16
-extern void neon_read(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
-					  char *buffer);
 extern PGDLLEXPORT void neon_read_at_lsn(NRelFileInfo rnode, ForkNumber forkNum, BlockNumber blkno,
 										 neon_request_lsns request_lsns, char *buffer);
-extern void neon_write(SMgrRelation reln, ForkNumber forknum,
-					   BlockNumber blocknum, char *buffer, bool skipFsync);
 #else
-extern void neon_read(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
-					  void *buffer);
 extern PGDLLEXPORT void neon_read_at_lsn(NRelFileInfo rnode, ForkNumber forkNum, BlockNumber blkno,
 										 neon_request_lsns request_lsns, void *buffer);
-extern void neon_write(SMgrRelation reln, ForkNumber forknum,
-					   BlockNumber blocknum, const void *buffer, bool skipFsync);
 #endif
 extern void neon_writeback(SMgrRelation reln, ForkNumber forknum,
 						   BlockNumber blocknum, BlockNumber nblocks);
@@ -299,17 +294,34 @@ extern void update_cached_relsize(NRelFileInfo rinfo, ForkNumber forknum, BlockN
 extern void forget_cached_relsize(NRelFileInfo rinfo, ForkNumber forknum);
 
 /* functions for local file cache */
-#if PG_MAJORVERSION_NUM < 16
-extern void lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
-					  char *buffer);
-#else
-extern void lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
-					  const void *buffer);
-#endif
-extern bool lfc_read(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno, char *buffer);
-extern bool lfc_cache_contains(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno);
+extern void lfc_writev(NRelFileInfo rinfo, ForkNumber forkNum,
+					   BlockNumber blkno, const void *const *buffers,
+					   BlockNumber nblocks);
+/* returns number of blocks read, with one bit set in *read for each  */
+extern int lfc_readv_select(NRelFileInfo rinfo, ForkNumber forkNum,
+							BlockNumber blkno, void **buffers,
+							BlockNumber nblocks, bits8 *mask);
+
+extern bool lfc_cache_contains(NRelFileInfo rinfo, ForkNumber forkNum,
+							   BlockNumber blkno);
+extern int lfc_cache_containsv(NRelFileInfo rinfo, ForkNumber forkNum,
+							   BlockNumber blkno, int nblocks, bits8 *bitmap);
 extern void lfc_evict(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno);
 extern void lfc_init(void);
 
+static inline bool
+lfc_read(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
+		 void *buffer)
+{
+	bits8		rv = 0;
+	return lfc_readv_select(rinfo, forkNum, blkno, &buffer, 1, &rv) == 1;
+}
+
+static inline void
+lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
+		  const void *buffer)
+{
+	return lfc_writev(rinfo, forkNum, blkno, &buffer, 1);
+}
 
 #endif

pgxn/neon/walproposer.c

@@ -1038,9 +1038,12 @@ DetermineEpochStartLsn(WalProposer *wp)
 		if (SkipXLogPageHeader(wp, wp->propEpochStartLsn) != wp->api.get_redo_start_lsn(wp))
 		{
 			/*
-			 * However, allow to proceed if previously elected leader was me;
-			 * plain restart of walproposer not intervened by concurrent
-			 * compute (who could generate WAL) is ok.
+			 * However, allow to proceed if last_log_term on the node which gave
+			 * the highest vote (i.e. point where we are going to start writing)
+			 * actually had been won by me; plain restart of walproposer not
+			 * intervened by concurrent compute which wrote WAL is ok.
+			 *
+			 * This avoids compute crash after manual term_bump.
 			 */
 			if (!((dth->n_entries >= 1) && (dth->entries[dth->n_entries - 1].term ==
 											pg_atomic_read_u64(&walprop_shared->mineLastElectedTerm))))
@@ -1442,12 +1445,17 @@ RecvAppendResponses(Safekeeper *sk)
 		if (sk->appendResponse.term > wp->propTerm)
 		{
 			/*
-			 * Another compute with higher term is running. Panic to restart
-			 * PG as we likely need to retake basebackup. However, don't dump
-			 * core as this is kinda expected scenario.
+			 *
+			 * Term has changed to higher one, probably another compute is
+			 * running. If this is the case we could PANIC as well because
+			 * likely it inserted some data and our basebackup is unsuitable
+			 * anymore. However, we also bump term manually (term_bump endpoint)
+			 * on safekeepers for migration purposes, in this case we do want
+			 * compute to stay alive. So restart walproposer with FATAL instead
+			 * of panicking; if basebackup is spoiled next election will notice
+			 * this.
 			 */
-			disable_core_dump();
-			wp_log(PANIC, "WAL acceptor %s:%s with term " INT64_FORMAT " rejected our request, our term " INT64_FORMAT ", meaning another compute is running at the same time, and it conflicts with us",
+			wp_log(FATAL, "WAL acceptor %s:%s with term " INT64_FORMAT " rejected our request, our term " INT64_FORMAT ", meaning another compute is running at the same time, and it conflicts with us",
 				   sk->host, sk->port,
 				   sk->appendResponse.term, wp->propTerm);
 		}

pgxn/neon/walproposer_pg.c

@@ -81,6 +81,7 @@ static void nwp_register_gucs(void);
 static void assign_neon_safekeepers(const char *newval, void *extra);
 static void nwp_prepare_shmem(void);
 static uint64 backpressure_lag_impl(void);
+static uint64 startup_backpressure_wrap(void);
 static bool backpressure_throttling_impl(void);
 static void walprop_register_bgworker(void);
 
@@ -90,7 +91,7 @@ static void walprop_pg_init_bgworker(void);
 static TimestampTz walprop_pg_get_current_timestamp(WalProposer *wp);
 static void walprop_pg_load_libpqwalreceiver(void);
 
-static process_interrupts_callback_t PrevProcessInterruptsCallback;
+static process_interrupts_callback_t PrevProcessInterruptsCallback = NULL;
 static shmem_startup_hook_type prev_shmem_startup_hook_type;
 #if PG_VERSION_NUM >= 150000
 static shmem_request_hook_type prev_shmem_request_hook = NULL;
@@ -178,7 +179,7 @@ pg_init_walproposer(void)
 
 	nwp_prepare_shmem();
 
-	delay_backend_us = &backpressure_lag_impl;
+	delay_backend_us = &startup_backpressure_wrap;
 	PrevProcessInterruptsCallback = ProcessInterruptsCallback;
 	ProcessInterruptsCallback = backpressure_throttling_impl;
 
@@ -352,6 +353,22 @@ backpressure_lag_impl(void)
 	return 0;
 }
 
+/*
+ * We don't apply backpressure when we're the postmaster, or the startup
+ * process, because in postmaster we can't apply backpressure, and in
+ * the startup process we can't afford to slow down.
+ */
+static uint64
+startup_backpressure_wrap(void)
+{
+	if (AmStartupProcess() || !IsUnderPostmaster)
+		return 0;
+
+	delay_backend_us = &backpressure_lag_impl;
+
+	return backpressure_lag_impl();
+}
+
 /*
  * WalproposerShmemSize --- report amount of shared memory space needed
  */
@@ -401,12 +418,13 @@ WalproposerShmemInit_SyncSafekeeper(void)
 static bool
 backpressure_throttling_impl(void)
 {
-	int64		lag;
+	uint64		lag;
 	TimestampTz start,
 				stop;
-	bool		retry = PrevProcessInterruptsCallback
-		? PrevProcessInterruptsCallback()
-		: false;
+	bool		retry = false;
+
+	if (PointerIsValid(PrevProcessInterruptsCallback))
+		retry = PrevProcessInterruptsCallback();
 
 	/*
 	 * Don't throttle read only transactions or wal sender. Do throttle CREATE
@@ -602,7 +620,12 @@ walprop_pg_init_walsender(void)
 	/* Create replication slot for WAL proposer if not exists */
 	if (SearchNamedReplicationSlot(WAL_PROPOSER_SLOT_NAME, false) == NULL)
 	{
+#if PG_MAJORVERSION_NUM >= 17
+		ReplicationSlotCreate(WAL_PROPOSER_SLOT_NAME, false, RS_PERSISTENT,
+							  false, false, false);
+#else
 		ReplicationSlotCreate(WAL_PROPOSER_SLOT_NAME, false, RS_PERSISTENT, false);
+#endif
 		ReplicationSlotReserveWal();
 		/* Write this slot to disk */
 		ReplicationSlotMarkDirty();
@@ -1509,7 +1532,11 @@ walprop_pg_init_event_set(WalProposer *wp)
 		wpg_log(FATAL, "double-initialization of event set");
 
 	/* for each sk, we have socket plus potentially socket for neon walreader */
+#if PG_MAJORVERSION_NUM >= 17
+	waitEvents = CreateWaitEventSet(NULL, 2 + 2 * wp->n_safekeepers);
+#else
 	waitEvents = CreateWaitEventSet(TopMemoryContext, 2 + 2 * wp->n_safekeepers);
+#endif
 	AddWaitEventToSet(waitEvents, WL_LATCH_SET, PGINVALID_SOCKET,
 					  MyLatch, NULL);
 	AddWaitEventToSet(waitEvents, WL_EXIT_ON_PM_DEATH, PGINVALID_SOCKET,

pgxn/neon_rmgr/neon_rmgr_decode.c

@@ -1,6 +1,7 @@
 #include "postgres.h"
 
 #if PG_MAJORVERSION_NUM >= 16
+
 #include "access/heapam_xlog.h"
 #include "access/neon_xlog.h"
 #include "replication/decode.h"
@@ -9,6 +10,10 @@
 
 #include "neon_rmgr.h"
 
+#endif /* PG >= 16 */
+
+#if PG_MAJORVERSION_NUM == 16
+
 /* individual record(group)'s handlers */
 static void DecodeNeonInsert(LogicalDecodingContext *ctx, XLogRecordBuffer *buf);
 static void DecodeNeonUpdate(LogicalDecodingContext *ctx, XLogRecordBuffer *buf);
@@ -399,6 +404,398 @@ DecodeXLogTuple(char *data, Size len, ReorderBufferTupleBuf *tuple)
 	header->t_infomask2 = xlhdr.t_infomask2;
 	header->t_hoff = xlhdr.t_hoff;
 }
+#endif
+
+#if PG_MAJORVERSION_NUM == 17
+
+/* individual record(group)'s handlers */
+static void DecodeNeonInsert(LogicalDecodingContext *ctx, XLogRecordBuffer *buf);
+static void DecodeNeonUpdate(LogicalDecodingContext *ctx, XLogRecordBuffer *buf);
+static void DecodeNeonDelete(LogicalDecodingContext *ctx, XLogRecordBuffer *buf);
+static void DecodeNeonMultiInsert(LogicalDecodingContext *ctx, XLogRecordBuffer *buf);
+
+/* common function to decode tuples */
+static void DecodeXLogTuple(char *data, Size len, HeapTuple tuple);
 
 
-#endif
+void
+neon_rm_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
+{
+	uint8		info = XLogRecGetInfo(buf->record) & XLOG_NEON_OPMASK;
+	TransactionId xid = XLogRecGetXid(buf->record);
+	SnapBuild  *builder = ctx->snapshot_builder;
+
+	ReorderBufferProcessXid(ctx->reorder, xid, buf->origptr);
+
+	/*
+	 * If we don't have snapshot or we are just fast-forwarding, there is no
+	 * point in decoding data changes.
+	 */
+	if (SnapBuildCurrentState(builder) < SNAPBUILD_FULL_SNAPSHOT ||
+		ctx->fast_forward)
+		return;
+
+	switch (info)
+	{
+		case XLOG_NEON_HEAP_INSERT:
+			if (SnapBuildProcessChange(builder, xid, buf->origptr))
+				DecodeNeonInsert(ctx, buf);
+			break;
+		case XLOG_NEON_HEAP_DELETE:
+			if (SnapBuildProcessChange(builder, xid, buf->origptr))
+				DecodeNeonDelete(ctx, buf);
+			break;
+		case XLOG_NEON_HEAP_UPDATE:
+		case XLOG_NEON_HEAP_HOT_UPDATE:
+			if (SnapBuildProcessChange(builder, xid, buf->origptr))
+				DecodeNeonUpdate(ctx, buf);
+			break;
+		case XLOG_NEON_HEAP_LOCK:
+			break;
+		case XLOG_NEON_HEAP_MULTI_INSERT:
+			if (SnapBuildProcessChange(builder, xid, buf->origptr))
+				DecodeNeonMultiInsert(ctx, buf);
+			break;
+		default:
+			elog(ERROR, "unexpected RM_HEAP_ID record type: %u", info);
+			break;
+	}
+}
+
+static inline bool
+FilterByOrigin(LogicalDecodingContext *ctx, RepOriginId origin_id)
+{
+	if (ctx->callbacks.filter_by_origin_cb == NULL)
+		return false;
+
+	return filter_by_origin_cb_wrapper(ctx, origin_id);
+}
+
+/*
+ * Parse XLOG_HEAP_INSERT (not MULTI_INSERT!) records into tuplebufs.
+ *
+ * Deletes can contain the new tuple.
+ */
+static void
+DecodeNeonInsert(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
+{
+	Size		datalen;
+	char	   *tupledata;
+	Size		tuplelen;
+	XLogReaderState *r = buf->record;
+	xl_neon_heap_insert *xlrec;
+	ReorderBufferChange *change;
+	RelFileLocator target_locator;
+
+	xlrec = (xl_neon_heap_insert *) XLogRecGetData(r);
+
+	/*
+	 * Ignore insert records without new tuples (this does happen when
+	 * raw_heap_insert marks the TOAST record as HEAP_INSERT_NO_LOGICAL).
+	 */
+	if (!(xlrec->flags & XLH_INSERT_CONTAINS_NEW_TUPLE))
+		return;
+
+	/* only interested in our database */
+	XLogRecGetBlockTag(r, 0, &target_locator, NULL, NULL);
+	if (target_locator.dbOid != ctx->slot->data.database)
+		return;
+
+	/* output plugin doesn't look for this origin, no need to queue */
+	if (FilterByOrigin(ctx, XLogRecGetOrigin(r)))
+		return;
+
+	change = ReorderBufferGetChange(ctx->reorder);
+	if (!(xlrec->flags & XLH_INSERT_IS_SPECULATIVE))
+		change->action = REORDER_BUFFER_CHANGE_INSERT;
+	else
+		change->action = REORDER_BUFFER_CHANGE_INTERNAL_SPEC_INSERT;
+	change->origin_id = XLogRecGetOrigin(r);
+
+	memcpy(&change->data.tp.rlocator, &target_locator, sizeof(RelFileLocator));
+
+	tupledata = XLogRecGetBlockData(r, 0, &datalen);
+	tuplelen = datalen - SizeOfHeapHeader;
+
+	change->data.tp.newtuple =
+		ReorderBufferGetTupleBuf(ctx->reorder, tuplelen);
+
+	DecodeXLogTuple(tupledata, datalen, change->data.tp.newtuple);
+
+	change->data.tp.clear_toast_afterwards = true;
+
+	ReorderBufferQueueChange(ctx->reorder, XLogRecGetXid(r), buf->origptr,
+							 change,
+							 xlrec->flags & XLH_INSERT_ON_TOAST_RELATION);
+}
+
+/*
+ * Parse XLOG_HEAP_DELETE from wal into proper tuplebufs.
+ *
+ * Deletes can possibly contain the old primary key.
+ */
+static void
+DecodeNeonDelete(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
+{
+	XLogReaderState *r = buf->record;
+	xl_neon_heap_delete *xlrec;
+	ReorderBufferChange *change;
+	RelFileLocator target_locator;
+
+	xlrec = (xl_neon_heap_delete *) XLogRecGetData(r);
+
+	/* only interested in our database */
+	XLogRecGetBlockTag(r, 0, &target_locator, NULL, NULL);
+	if (target_locator.dbOid != ctx->slot->data.database)
+		return;
+
+	/* output plugin doesn't look for this origin, no need to queue */
+	if (FilterByOrigin(ctx, XLogRecGetOrigin(r)))
+		return;
+
+	change = ReorderBufferGetChange(ctx->reorder);
+
+	if (xlrec->flags & XLH_DELETE_IS_SUPER)
+		change->action = REORDER_BUFFER_CHANGE_INTERNAL_SPEC_ABORT;
+	else
+		change->action = REORDER_BUFFER_CHANGE_DELETE;
+
+	change->origin_id = XLogRecGetOrigin(r);
+
+	memcpy(&change->data.tp.rlocator, &target_locator, sizeof(RelFileLocator));
+
+	/* old primary key stored */
+	if (xlrec->flags & XLH_DELETE_CONTAINS_OLD)
+	{
+		Size		datalen = XLogRecGetDataLen(r) - SizeOfNeonHeapHeader;
+		Size		tuplelen = datalen - SizeOfNeonHeapHeader;
+
+		Assert(XLogRecGetDataLen(r) > (SizeOfNeonHeapDelete + SizeOfNeonHeapHeader));
+
+		change->data.tp.oldtuple =
+			ReorderBufferGetTupleBuf(ctx->reorder, tuplelen);
+
+		DecodeXLogTuple((char *) xlrec + SizeOfNeonHeapDelete,
+						datalen, change->data.tp.oldtuple);
+	}
+
+	change->data.tp.clear_toast_afterwards = true;
+
+	ReorderBufferQueueChange(ctx->reorder, XLogRecGetXid(r), buf->origptr,
+							 change, false);
+}
+
+/*
+ * Parse XLOG_HEAP_UPDATE and XLOG_HEAP_HOT_UPDATE, which have the same layout
+ * in the record, from wal into proper tuplebufs.
+ *
+ * Updates can possibly contain a new tuple and the old primary key.
+ */
+static void
+DecodeNeonUpdate(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
+{
+	XLogReaderState *r = buf->record;
+	xl_neon_heap_update *xlrec;
+	ReorderBufferChange *change;
+	char	   *data;
+	RelFileLocator target_locator;
+
+	xlrec = (xl_neon_heap_update *) XLogRecGetData(r);
+
+	/* only interested in our database */
+	XLogRecGetBlockTag(r, 0, &target_locator, NULL, NULL);
+	if (target_locator.dbOid != ctx->slot->data.database)
+		return;
+
+	/* output plugin doesn't look for this origin, no need to queue */
+	if (FilterByOrigin(ctx, XLogRecGetOrigin(r)))
+		return;
+
+	change = ReorderBufferGetChange(ctx->reorder);
+	change->action = REORDER_BUFFER_CHANGE_UPDATE;
+	change->origin_id = XLogRecGetOrigin(r);
+	memcpy(&change->data.tp.rlocator, &target_locator, sizeof(RelFileLocator));
+
+	if (xlrec->flags & XLH_UPDATE_CONTAINS_NEW_TUPLE)
+	{
+		Size		datalen;
+		Size		tuplelen;
+
+		data = XLogRecGetBlockData(r, 0, &datalen);
+
+		tuplelen = datalen - SizeOfNeonHeapHeader;
+
+		change->data.tp.newtuple =
+			ReorderBufferGetTupleBuf(ctx->reorder, tuplelen);
+
+		DecodeXLogTuple(data, datalen, change->data.tp.newtuple);
+	}
+
+	if (xlrec->flags & XLH_UPDATE_CONTAINS_OLD)
+	{
+		Size		datalen;
+		Size		tuplelen;
+
+		/* caution, remaining data in record is not aligned */
+		data = XLogRecGetData(r) + SizeOfNeonHeapUpdate;
+		datalen = XLogRecGetDataLen(r) - SizeOfNeonHeapUpdate;
+		tuplelen = datalen - SizeOfNeonHeapHeader;
+
+		change->data.tp.oldtuple =
+			ReorderBufferGetTupleBuf(ctx->reorder, tuplelen);
+
+		DecodeXLogTuple(data, datalen, change->data.tp.oldtuple);
+	}
+
+	change->data.tp.clear_toast_afterwards = true;
+
+	ReorderBufferQueueChange(ctx->reorder, XLogRecGetXid(r), buf->origptr,
+							 change, false);
+}
+
+/*
+ * Decode XLOG_HEAP2_MULTI_INSERT_insert record into multiple tuplebufs.
+ *
+ * Currently MULTI_INSERT will always contain the full tuples.
+ */
+static void
+DecodeNeonMultiInsert(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
+{
+	XLogReaderState *r = buf->record;
+	xl_neon_heap_multi_insert *xlrec;
+	int			i;
+	char	   *data;
+	char	   *tupledata;
+	Size		tuplelen;
+	RelFileLocator rlocator;
+
+	xlrec = (xl_neon_heap_multi_insert *) XLogRecGetData(r);
+
+	/*
+	 * Ignore insert records without new tuples.  This happens when a
+	 * multi_insert is done on a catalog or on a non-persistent relation.
+	 */
+	if (!(xlrec->flags & XLH_INSERT_CONTAINS_NEW_TUPLE))
+		return;
+
+	/* only interested in our database */
+	XLogRecGetBlockTag(r, 0, &rlocator, NULL, NULL);
+	if (rlocator.dbOid != ctx->slot->data.database)
+		return;
+
+	/* output plugin doesn't look for this origin, no need to queue */
+	if (FilterByOrigin(ctx, XLogRecGetOrigin(r)))
+		return;
+
+	/*
+	 * We know that this multi_insert isn't for a catalog, so the block should
+	 * always have data even if a full-page write of it is taken.
+	 */
+	tupledata = XLogRecGetBlockData(r, 0, &tuplelen);
+	Assert(tupledata != NULL);
+
+	data = tupledata;
+	for (i = 0; i < xlrec->ntuples; i++)
+	{
+		ReorderBufferChange *change;
+		xl_neon_multi_insert_tuple *xlhdr;
+		int			datalen;
+		HeapTuple	tuple;
+		HeapTupleHeader header;
+
+		change = ReorderBufferGetChange(ctx->reorder);
+		change->action = REORDER_BUFFER_CHANGE_INSERT;
+		change->origin_id = XLogRecGetOrigin(r);
+
+		memcpy(&change->data.tp.rlocator, &rlocator, sizeof(RelFileLocator));
+
+		xlhdr = (xl_neon_multi_insert_tuple *) SHORTALIGN(data);
+		data = ((char *) xlhdr) + SizeOfNeonMultiInsertTuple;
+		datalen = xlhdr->datalen;
+
+		change->data.tp.newtuple =
+			ReorderBufferGetTupleBuf(ctx->reorder, datalen);
+
+		tuple = change->data.tp.newtuple;
+		header = tuple->t_data;
+
+		/* not a disk based tuple */
+		ItemPointerSetInvalid(&tuple->t_self);
+
+		/*
+		 * We can only figure this out after reassembling the transactions.
+		 */
+		tuple->t_tableOid = InvalidOid;
+
+		tuple->t_len = datalen + SizeofHeapTupleHeader;
+
+		memset(header, 0, SizeofHeapTupleHeader);
+
+		memcpy((char *) tuple->t_data + SizeofHeapTupleHeader,
+			   (char *) data,
+			   datalen);
+		header->t_infomask = xlhdr->t_infomask;
+		header->t_infomask2 = xlhdr->t_infomask2;
+		header->t_hoff = xlhdr->t_hoff;
+
+		/*
+		 * Reset toast reassembly state only after the last row in the last
+		 * xl_multi_insert_tuple record emitted by one heap_multi_insert()
+		 * call.
+		 */
+		if (xlrec->flags & XLH_INSERT_LAST_IN_MULTI &&
+			(i + 1) == xlrec->ntuples)
+			change->data.tp.clear_toast_afterwards = true;
+		else
+			change->data.tp.clear_toast_afterwards = false;
+
+		ReorderBufferQueueChange(ctx->reorder, XLogRecGetXid(r),
+								 buf->origptr, change, false);
+
+		/* move to the next xl_neon_multi_insert_tuple entry */
+		data += datalen;
+	}
+	Assert(data == tupledata + tuplelen);
+}
+
+/*
+ * Read a HeapTuple as WAL logged by heap_insert, heap_update and heap_delete
+ * (but not by heap_multi_insert) into a tuplebuf.
+ *
+ * The size 'len' and the pointer 'data' in the record need to be
+ * computed outside as they are record specific.
+ */
+static void
+DecodeXLogTuple(char *data, Size len, HeapTuple tuple)
+{
+	xl_neon_heap_header xlhdr;
+	int			datalen = len - SizeOfNeonHeapHeader;
+	HeapTupleHeader header;
+
+	Assert(datalen >= 0);
+
+	tuple->t_len = datalen + SizeofHeapTupleHeader;
+	header = tuple->t_data;
+
+	/* not a disk based tuple */
+	ItemPointerSetInvalid(&tuple->t_self);
+
+	/* we can only figure this out after reassembling the transactions */
+	tuple->t_tableOid = InvalidOid;
+
+	/* data is not stored aligned, copy to aligned storage */
+	memcpy((char *) &xlhdr,
+		   data,
+		   SizeOfNeonHeapHeader);
+
+	memset(header, 0, SizeofHeapTupleHeader);
+
+	memcpy(((char *) tuple->t_data) + SizeofHeapTupleHeader,
+		   data + SizeOfNeonHeapHeader,
+		   datalen);
+
+	header->t_infomask = xlhdr.t_infomask;
+	header->t_infomask2 = xlhdr.t_infomask2;
+	header->t_hoff = xlhdr.t_hoff;
+}
+#endif

pgxn/neon_walredo/inmem_smgr.c

@@ -68,8 +68,13 @@ static void inmem_close(SMgrRelation reln, ForkNumber forknum);
 static void inmem_create(SMgrRelation reln, ForkNumber forknum, bool isRedo);
 static bool inmem_exists(SMgrRelation reln, ForkNumber forknum);
 static void inmem_unlink(NRelFileInfoBackend rinfo, ForkNumber forknum, bool isRedo);
+#if PG_MAJORVERSION_NUM >= 17
+static bool inmem_prefetch(SMgrRelation reln, ForkNumber forknum,
+						   BlockNumber blocknum, int nblocks);
+#else
 static bool inmem_prefetch(SMgrRelation reln, ForkNumber forknum,
 						   BlockNumber blocknum);
+#endif
 #if PG_MAJORVERSION_NUM < 16
 static void inmem_extend(SMgrRelation reln, ForkNumber forknum,
 						 BlockNumber blocknum, char *buffer, bool skipFsync);
@@ -93,7 +98,9 @@ static BlockNumber inmem_nblocks(SMgrRelation reln, ForkNumber forknum);
 static void inmem_truncate(SMgrRelation reln, ForkNumber forknum,
 						   BlockNumber nblocks);
 static void inmem_immedsync(SMgrRelation reln, ForkNumber forknum);
-
+#if PG_MAJORVERSION_NUM >= 17
+static void inmem_registersync(SMgrRelation reln, ForkNumber forknum);
+#endif
 
 /*
  *	inmem_init() -- Initialize private state
@@ -190,6 +197,14 @@ inmem_close(SMgrRelation reln, ForkNumber forknum)
 {
 }
 
+#if PG_MAJORVERSION_NUM >= 17
+static bool
+inmem_prefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+			   int nblocks)
+{
+	return true;
+}
+#else
 /*
  *	inmem_prefetch() -- Initiate asynchronous read of the specified block of a relation
  */
@@ -198,6 +213,7 @@ inmem_prefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
 {
 	return true;
 }
+#endif
 
 /*
  * inmem_writeback() -- Tell the kernel to write pages back to storage.
@@ -211,11 +227,13 @@ inmem_writeback(SMgrRelation reln, ForkNumber forknum,
 /*
  *	inmem_read() -- Read the specified block from a relation.
  */
+#if PG_MAJORVERSION_NUM < 16
 static void
 inmem_read(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
-#if PG_MAJORVERSION_NUM < 16
 		   char *buffer)
 #else
+static void
+inmem_read(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
 		   void *buffer)
 #endif
 {
@@ -228,6 +246,18 @@ inmem_read(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
 		memcpy(buffer, page_body[pg], BLCKSZ);
 }
 
+#if PG_MAJORVERSION_NUM >= 17
+static void
+inmem_readv(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
+			void **buffers, BlockNumber nblocks)
+{
+	for (int i = 0; i < nblocks; i++)
+	{
+		inmem_read(reln, forknum, blkno, buffers[i]);
+	}
+}
+#endif
+
 /*
  *	inmem_write() -- Write the supplied block at the appropriate location.
  *
@@ -280,6 +310,18 @@ inmem_write(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
 	memcpy(page_body[pg], buffer, BLCKSZ);
 }
 
+#if PG_MAJORVERSION_NUM >= 17
+static void
+inmem_writev(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
+			 const void **buffers, BlockNumber nblocks, bool skipFsync)
+{
+	for (int i = 0; i < nblocks; i++)
+	{
+		inmem_write(reln, forknum, blkno, buffers[i], skipFsync);
+	}
+}
+#endif
+
 /*
  *	inmem_nblocks() -- Get the number of blocks stored in a relation.
  */
@@ -315,6 +357,13 @@ inmem_immedsync(SMgrRelation reln, ForkNumber forknum)
 {
 }
 
+#if PG_MAJORVERSION_NUM >= 17
+static void
+inmem_registersync(SMgrRelation reln, ForkNumber forknum)
+{
+}
+#endif
+
 static const struct f_smgr inmem_smgr =
 {
 	.smgr_init = inmem_init,
@@ -328,23 +377,39 @@ static const struct f_smgr inmem_smgr =
 #if PG_MAJORVERSION_NUM >= 16
 	.smgr_zeroextend = inmem_zeroextend,
 #endif
+#if PG_MAJORVERSION_NUM >= 17
+	.smgr_prefetch = inmem_prefetch,
+	.smgr_readv = inmem_readv,
+	.smgr_writev = inmem_writev,
+#else
 	.smgr_prefetch = inmem_prefetch,
 	.smgr_read = inmem_read,
 	.smgr_write = inmem_write,
+#endif
 	.smgr_writeback = inmem_writeback,
 	.smgr_nblocks = inmem_nblocks,
 	.smgr_truncate = inmem_truncate,
 	.smgr_immedsync = inmem_immedsync,
+
+#if PG_MAJORVERSION_NUM >= 17
+	.smgr_registersync = inmem_registersync,
+#endif
+
+	.smgr_start_unlogged_build = NULL,
+	.smgr_finish_unlogged_build_phase_1 = NULL,
+	.smgr_end_unlogged_build = NULL,
+	.smgr_read_slru_segment = NULL,
 };
 
 const f_smgr *
-smgr_inmem(BackendId backend, NRelFileInfo rinfo)
+smgr_inmem(ProcNumber backend, NRelFileInfo rinfo)
 {
 	Assert(InRecovery);
-	if (backend != InvalidBackendId)
-		return smgr_standard(backend, rinfo);
-	else
-		return &inmem_smgr;
+	// // What does this code do?
+	// if (backend != INVALID_PROC_NUMBER)
+	// 	return smgr_standard(backend, rinfo);
+	// else
+	return &inmem_smgr;
 }
 
 void