flush_frozen_layer: assert we own the frozen layer when we remove its ephemeral file

TODO the assert is incorrect because get_reconstruct_data's traversal_path
can be holding references to inmemory and frozen layers.

.github/workflows/benchmarking.yml

@@ -180,8 +180,7 @@ jobs:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
       options: --init
 
-    # Increase timeout to 8h, default timeout is 6h
-    timeout-minutes: 480
+    timeout-minutes: 360 # 6h
 
     steps:
     - uses: actions/checkout@v3
@@ -322,6 +321,8 @@ jobs:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
       options: --init
 
+    timeout-minutes: 360 # 6h
+
     steps:
     - uses: actions/checkout@v3
 
@@ -413,6 +414,8 @@ jobs:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
       options: --init
 
+    timeout-minutes: 360 # 6h
+
     steps:
     - uses: actions/checkout@v3
 
@@ -498,6 +501,8 @@ jobs:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
       options: --init
 
+    timeout-minutes: 360 # 6h
+
     steps:
     - uses: actions/checkout@v3
 

.github/workflows/build_and_test.yml

@@ -623,6 +623,51 @@ jobs:
       - name: Cleanup ECR folder
         run: rm -rf ~/.ecr
 
+
+  neon-image-depot:
+    # For testing this will run side-by-side for a few merges.
+    # This action is not really optimized yet, but gets the job done
+    runs-on: [ self-hosted, gen3, large ]
+    needs: [ tag ]
+    container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/base:pinned
+    permissions:
+      contents: read
+      id-token: write
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v3
+        with:
+          submodules: true
+          fetch-depth: 0
+
+      - name: Setup go
+        uses: actions/setup-go@v3
+        with:
+          go-version: '1.19'
+
+      - name: Set up Depot CLI
+        uses: depot/setup-action@v1
+
+      - name: Install Crane & ECR helper
+        run: go install github.com/awslabs/amazon-ecr-credential-helper/ecr-login/cli/docker-credential-ecr-login@69c85dc22db6511932bbf119e1a0cc5c90c69a7f # v0.6.0
+
+      - name: Configure ECR login
+        run: |
+          mkdir /github/home/.docker/
+          echo "{\"credsStore\":\"ecr-login\"}" > /github/home/.docker/config.json
+
+      - name: Build and push
+        uses: depot/build-push-action@v1
+        with:
+          # if no depot.json file is at the root of your repo, you must specify the project id
+          project: nrdv0s4kcs
+          push: true
+          tags: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/neon:depot-${{needs.tag.outputs.build-tag}}
+          build-args: |
+            GIT_VERSION=${{ github.sha }}
+            REPOSITORY=369495373322.dkr.ecr.eu-central-1.amazonaws.com
+
   compute-tools-image:
     runs-on: [ self-hosted, gen3, large ]
     needs: [ tag ]
@@ -659,7 +704,6 @@ jobs:
                            --cache-repo 369495373322.dkr.ecr.eu-central-1.amazonaws.com/cache
                            --context .
                            --build-arg GIT_VERSION=${{ github.sha }}
-                           --build-arg BUILD_TAG=${{needs.tag.outputs.build-tag}}
                            --build-arg REPOSITORY=369495373322.dkr.ecr.eu-central-1.amazonaws.com
                            --dockerfile Dockerfile.compute-tools
                            --destination 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:${{needs.tag.outputs.build-tag}}
@@ -717,40 +761,10 @@ jobs:
                            --context .
                            --build-arg GIT_VERSION=${{ github.sha }}
                            --build-arg PG_VERSION=${{ matrix.version }}
-                           --build-arg BUILD_TAG=${{needs.tag.outputs.build-tag}}
                            --build-arg REPOSITORY=369495373322.dkr.ecr.eu-central-1.amazonaws.com
                            --dockerfile Dockerfile.compute-node
                            --destination 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-${{ matrix.version }}:${{needs.tag.outputs.build-tag}}
                            --destination neondatabase/compute-node-${{ matrix.version }}:${{needs.tag.outputs.build-tag}}
-                           --cleanup
-
-      # Due to a kaniko bug, we can't use cache for extensions image, thus it takes about the same amount of time as compute-node image to build (~10 min)
-      # During the transition period we need to have extensions in both places (in S3 and in compute-node image),
-      # so we won't build extension twice, but extract them from compute-node.
-      #
-      # For now we use extensions image only for new custom extensitons
-      - name: Kaniko build extensions only
-        run: |
-          # Kaniko is suposed to clean up after itself if --cleanup flag is set, but it doesn't.
-          # Despite some fixes were made in https://github.com/GoogleContainerTools/kaniko/pull/2504 (in kaniko v1.11.0),
-          # it still fails with error:
-          #   error building image: could not save file: copying file: symlink postgres /kaniko/1/usr/local/pgsql/bin/postmaster: file exists
-          #
-          # Ref https://github.com/GoogleContainerTools/kaniko/issues/1406
-          find /kaniko -maxdepth 1 -mindepth 1 -type d -regex "/kaniko/[0-9]*" -exec rm -rv {} \;
-
-          /kaniko/executor --reproducible --snapshot-mode=redo --skip-unused-stages --cache=true \
-                           --cache-repo 369495373322.dkr.ecr.eu-central-1.amazonaws.com/cache \
-                           --context . \
-                           --build-arg GIT_VERSION=${{ github.sha }} \
-                           --build-arg PG_VERSION=${{ matrix.version }} \
-                           --build-arg BUILD_TAG=${{needs.tag.outputs.build-tag}} \
-                           --build-arg REPOSITORY=369495373322.dkr.ecr.eu-central-1.amazonaws.com \
-                           --dockerfile Dockerfile.compute-node \
-                           --destination 369495373322.dkr.ecr.eu-central-1.amazonaws.com/extensions-${{ matrix.version }}:${{needs.tag.outputs.build-tag}} \
-                           --destination neondatabase/extensions-${{ matrix.version }}:${{needs.tag.outputs.build-tag}} \
-                           --cleanup \
-                           --target postgres-extensions
 
       # Cleanup script fails otherwise - rm: cannot remove '/nvme/actions-runner/_work/_temp/_github_home/.ecr': Permission denied
       - name: Cleanup ECR folder
@@ -767,7 +781,7 @@ jobs:
       run:
         shell: sh -eu {0}
     env:
-      VM_BUILDER_VERSION: v0.11.0
+      VM_BUILDER_VERSION: v0.8.0
 
     steps:
       - name: Checkout
@@ -869,10 +883,8 @@ jobs:
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:${{needs.tag.outputs.build-tag}} latest
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v14:${{needs.tag.outputs.build-tag}} latest
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v14:${{needs.tag.outputs.build-tag}} latest
-          crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/extensions-v14:${{needs.tag.outputs.build-tag}} latest
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v15:${{needs.tag.outputs.build-tag}} latest
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v15:${{needs.tag.outputs.build-tag}} latest
-          crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/extensions-v15:${{needs.tag.outputs.build-tag}} latest
 
       - name: Push images to production ECR
         if: |
@@ -883,10 +895,8 @@ jobs:
           crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/compute-tools:latest
           crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v14:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v14:latest
           crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v14:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v14:latest
-          crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/extensions-v14:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/extensions-v14:latest
           crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v15:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v15:latest
           crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v15:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v15:latest
-          crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/extensions-v15:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/extensions-v15:latest
 
       - name: Configure Docker Hub login
         run: |
@@ -908,89 +918,16 @@ jobs:
           crane tag neondatabase/compute-tools:${{needs.tag.outputs.build-tag}} latest
           crane tag neondatabase/compute-node-v14:${{needs.tag.outputs.build-tag}} latest
           crane tag neondatabase/vm-compute-node-v14:${{needs.tag.outputs.build-tag}} latest
-          crane tag neondatabase/extensions-v14:${{needs.tag.outputs.build-tag}} latest
           crane tag neondatabase/compute-node-v15:${{needs.tag.outputs.build-tag}} latest
           crane tag neondatabase/vm-compute-node-v15:${{needs.tag.outputs.build-tag}} latest
-          crane tag neondatabase/extensions-v15:${{needs.tag.outputs.build-tag}} latest
 
       - name: Cleanup ECR folder
         run: rm -rf ~/.ecr
 
-  upload-postgres-extensions-to-s3:
-    if: |
-      (github.ref_name == 'main' || github.ref_name == 'release') &&
-       github.event_name != 'workflow_dispatch'
-    runs-on: ${{ github.ref_name == 'release' && fromJSON('["self-hosted", "prod", "x64"]') || fromJSON('["self-hosted", "gen3", "small"]') }}
-    needs: [ tag, promote-images ]
-    strategy:
-      fail-fast: false
-      matrix:
-        version: [ v14, v15 ]
-
-    env:
-      # While on transition period we extract public extensions from compute-node image and custom extensions from extensions image.
-      # Later all the extensions will be moved to extensions image.
-      EXTENSIONS_IMAGE: ${{ github.ref_name == 'release' && '093970136003' || '369495373322'}}.dkr.ecr.eu-central-1.amazonaws.com/extensions-${{ matrix.version }}:latest
-      COMPUTE_NODE_IMAGE: ${{ github.ref_name == 'release' && '093970136003' || '369495373322'}}.dkr.ecr.eu-central-1.amazonaws.com/compute-node-${{ matrix.version }}:latest
-      AWS_ACCESS_KEY_ID: ${{ github.ref_name == 'release' && secrets.AWS_ACCESS_KEY_PROD || secrets.AWS_ACCESS_KEY_DEV }}
-      AWS_SECRET_ACCESS_KEY: ${{ github.ref_name == 'release' && secrets.AWS_SECRET_KEY_PROD || secrets.AWS_SECRET_KEY_DEV }}
-      S3_BUCKETS: |
-        ${{ github.ref_name == 'release' &&
-          'neon-prod-extensions-ap-southeast-1 neon-prod-extensions-eu-central-1 neon-prod-extensions-us-east-1 neon-prod-extensions-us-east-2 neon-prod-extensions-us-west-2' ||
-          'neon-dev-extensions-eu-central-1 neon-dev-extensions-eu-west-1 neon-dev-extensions-us-east-2' }}
-
-    steps:
-      - name: Pull postgres-extensions image
-        run: |
-          docker pull ${EXTENSIONS_IMAGE}
-          docker pull ${COMPUTE_NODE_IMAGE}
-
-      - name: Create postgres-extensions container
-        id: create-container
-        run: |
-          EID=$(docker create ${EXTENSIONS_IMAGE} true)
-          echo "EID=${EID}" >> $GITHUB_OUTPUT
-
-          CID=$(docker create ${COMPUTE_NODE_IMAGE} true)
-          echo "CID=${CID}" >> $GITHUB_OUTPUT
-
-      - name: Extract postgres-extensions from container
-        run: |
-          rm -rf ./extensions-to-upload ./custom-extensions # Just in case
-
-          # In compute image we have a bit different directory layout
-          mkdir -p extensions-to-upload/share
-          docker cp ${{ steps.create-container.outputs.CID }}:/usr/local/share/extension ./extensions-to-upload/share/extension
-          docker cp ${{ steps.create-container.outputs.CID }}:/usr/local/lib             ./extensions-to-upload/lib
-
-          # Delete Neon extensitons (they always present on compute-node image)
-          rm -rf ./extensions-to-upload/share/extension/neon*
-          rm -rf ./extensions-to-upload/lib/neon*
-
-          docker cp ${{ steps.create-container.outputs.EID }}:/extensions ./custom-extensions
-          for EXT_NAME in $(ls ./custom-extensions); do
-            mkdir -p ./extensions-to-upload/${EXT_NAME}/share
-
-            mv ./custom-extensions/${EXT_NAME}/share/extension ./extensions-to-upload/${EXT_NAME}/share/extension
-            mv ./custom-extensions/${EXT_NAME}/lib             ./extensions-to-upload/${EXT_NAME}/lib
-          done
-
-      - name: Upload postgres-extensions to S3
-        run: |
-          for BUCKET in $(echo ${S3_BUCKETS}); do
-            aws s3 cp --recursive --only-show-errors ./extensions-to-upload s3://${BUCKET}/${{ needs.tag.outputs.build-tag }}/${{ matrix.version }}
-          done
-
-      - name: Cleanup
-        if: ${{ always() && (steps.create-container.outputs.CID || steps.create-container.outputs.EID) }}
-        run: |
-          docker rm ${{ steps.create-container.outputs.CID }} || true
-          docker rm ${{ steps.create-container.outputs.EID }} || true
-
   deploy:
     runs-on: [ self-hosted, gen3, small ]
     container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/ansible:latest
-    needs: [ upload-postgres-extensions-to-s3, promote-images, tag, regress-tests ]
+    needs: [ promote-images, tag, regress-tests ]
     if: ( github.ref_name == 'main' || github.ref_name == 'release' ) && github.event_name != 'workflow_dispatch'
     steps:
       - name: Fix git ownership
@@ -1022,20 +959,6 @@ jobs:
             exit 1
           fi
 
-      - name: Create git tag
-        if: github.ref_name == 'release'
-        uses: actions/github-script@v6
-        with:
-          # Retry script for 5XX server errors: https://github.com/actions/github-script#retries
-          retries: 5
-          script: |
-            github.rest.git.createRef({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              ref: "refs/tags/${{ needs.tag.outputs.build-tag }}",
-              sha: context.sha,
-            })
-
   promote-compatibility-data:
     runs-on: [ self-hosted, gen3, small ]
     container:

.github/workflows/release.yml

@@ -3,7 +3,6 @@ name: Create Release Branch
 on:
   schedule:
     - cron: '0 10 * * 2'
-  workflow_dispatch:
 
 jobs:
   create_release_branch:

Cargo.lock

@@ -2349,9 +2349,9 @@ checksum = "0ab1bc2a289d34bd04a330323ac98a1b4bc82c9d9fcb1e66b63caa84da26b575"
 
 [[package]]
 name = "openssl"
-version = "0.10.55"
+version = "0.10.52"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "345df152bc43501c5eb9e4654ff05f794effb78d4efe3d53abc158baddc0703d"
+checksum = "01b8574602df80f7b85fdfc5392fa884a4e3b3f4f35402c070ab34c3d3f78d56"
 dependencies = [
  "bitflags",
  "cfg-if",
@@ -2381,9 +2381,9 @@ checksum = "ff011a302c396a5197692431fc1948019154afc178baf7d8e37367442a4601cf"
 
 [[package]]
 name = "openssl-sys"
-version = "0.9.90"
+version = "0.9.87"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "374533b0e45f3a7ced10fcaeccca020e66656bc03dac384f852e4e5a7a8104a6"
+checksum = "8e17f59264b2809d77ae94f0e1ebabc434773f370d6ca667bd223ea10e06cc7e"
 dependencies = [
  "cc",
  "libc",
@@ -2770,7 +2770,7 @@ dependencies = [
 [[package]]
 name = "postgres"
 version = "0.19.4"
-source = "git+https://github.com/neondatabase/rust-postgres.git?rev=1aaedab101b23f7612042850d8f2036810fa7c7f#1aaedab101b23f7612042850d8f2036810fa7c7f"
+source = "git+https://github.com/neondatabase/rust-postgres.git?rev=f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c#f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c"
 dependencies = [
  "bytes",
  "fallible-iterator",
@@ -2783,7 +2783,7 @@ dependencies = [
 [[package]]
 name = "postgres-native-tls"
 version = "0.5.0"
-source = "git+https://github.com/neondatabase/rust-postgres.git?rev=1aaedab101b23f7612042850d8f2036810fa7c7f#1aaedab101b23f7612042850d8f2036810fa7c7f"
+source = "git+https://github.com/neondatabase/rust-postgres.git?rev=f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c#f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c"
 dependencies = [
  "native-tls",
  "tokio",
@@ -2794,7 +2794,7 @@ dependencies = [
 [[package]]
 name = "postgres-protocol"
 version = "0.6.4"
-source = "git+https://github.com/neondatabase/rust-postgres.git?rev=1aaedab101b23f7612042850d8f2036810fa7c7f#1aaedab101b23f7612042850d8f2036810fa7c7f"
+source = "git+https://github.com/neondatabase/rust-postgres.git?rev=f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c#f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c"
 dependencies = [
  "base64 0.20.0",
  "byteorder",
@@ -2812,7 +2812,7 @@ dependencies = [
 [[package]]
 name = "postgres-types"
 version = "0.2.4"
-source = "git+https://github.com/neondatabase/rust-postgres.git?rev=1aaedab101b23f7612042850d8f2036810fa7c7f#1aaedab101b23f7612042850d8f2036810fa7c7f"
+source = "git+https://github.com/neondatabase/rust-postgres.git?rev=f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c#f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c"
 dependencies = [
  "bytes",
  "fallible-iterator",
@@ -4223,7 +4223,8 @@ dependencies = [
 [[package]]
 name = "tokio"
 version = "1.28.1"
-source = "git+https://github.com/problame/tokio.git?branch=problame/distinguish-core-and-worker-by-thread-name#d88791686cfc7fc7d010889ad7638d09646b3de7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0aa32867d44e6f2ce3385e89dceb990188b8bb0fb25b0cf576647a6f98ac5105"
 dependencies = [
  "autocfg",
  "bytes",
@@ -4250,7 +4251,8 @@ dependencies = [
 [[package]]
 name = "tokio-macros"
 version = "2.1.0"
-source = "git+https://github.com/problame/tokio.git?branch=problame/distinguish-core-and-worker-by-thread-name#d88791686cfc7fc7d010889ad7638d09646b3de7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "630bdcf245f78637c13ec01ffae6187cca34625e8c63150d424b59e55af2675e"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -4270,7 +4272,7 @@ dependencies = [
 [[package]]
 name = "tokio-postgres"
 version = "0.7.7"
-source = "git+https://github.com/neondatabase/rust-postgres.git?rev=1aaedab101b23f7612042850d8f2036810fa7c7f#1aaedab101b23f7612042850d8f2036810fa7c7f"
+source = "git+https://github.com/neondatabase/rust-postgres.git?rev=f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c#f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c"
 dependencies = [
  "async-trait",
  "byteorder",

Cargo.toml

@@ -140,11 +140,11 @@ 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", rev="1aaedab101b23f7612042850d8f2036810fa7c7f" }
-postgres-native-tls = { git = "https://github.com/neondatabase/rust-postgres.git", rev="1aaedab101b23f7612042850d8f2036810fa7c7f" }
-postgres-protocol = { git = "https://github.com/neondatabase/rust-postgres.git", rev="1aaedab101b23f7612042850d8f2036810fa7c7f" }
-postgres-types = { git = "https://github.com/neondatabase/rust-postgres.git", rev="1aaedab101b23f7612042850d8f2036810fa7c7f" }
-tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev="1aaedab101b23f7612042850d8f2036810fa7c7f" }
+postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev="f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c" }
+postgres-native-tls = { git = "https://github.com/neondatabase/rust-postgres.git", rev="f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c" }
+postgres-protocol = { git = "https://github.com/neondatabase/rust-postgres.git", rev="f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c" }
+postgres-types = { git = "https://github.com/neondatabase/rust-postgres.git", rev="f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c" }
+tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev="f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c" }
 tokio-tar = { git = "https://github.com/neondatabase/tokio-tar.git", rev="404df61437de0feef49ba2ccdbdd94eb8ad6e142" }
 
 ## Other git libraries
@@ -180,15 +180,13 @@ tonic-build = "0.9"
 
 # This is only needed for proxy's tests.
 # TODO: we should probably fork `tokio-postgres-rustls` instead.
-tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev="1aaedab101b23f7612042850d8f2036810fa7c7f" }
+tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev="f6ec31df3bcce89cb34f300f17c8a8c031c5ee8c" }
 
 # Changes the MAX_THREADS limit from 4096 to 32768.
 # This is a temporary workaround for using tracing from many threads in safekeepers code,
 # until async safekeepers patch is merged to the main.
 sharded-slab = { git = "https://github.com/neondatabase/sharded-slab.git", rev="98d16753ab01c61f0a028de44167307a00efea00" }
 
-tokio = { git = "https://github.com/problame/tokio.git", branch="problame/distinguish-core-and-worker-by-thread-name" }
-
 ################# Binary contents sections
 
 [profile.release]

Dockerfile.compute-node

@@ -2,7 +2,6 @@ ARG PG_VERSION
 ARG REPOSITORY=neondatabase
 ARG IMAGE=rust
 ARG TAG=pinned
-ARG BUILD_TAG
 
 #########################################################################################
 #
@@ -68,7 +67,7 @@ RUN apt update && \
 RUN 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 xvzf ../SFCGAL.tar.gz --strip-components=1 -C . && \
-    cmake -DCMAKE_BUILD_TYPE=Release . && make -j $(getconf _NPROCESSORS_ONLN) && \
+    cmake . && make -j $(getconf _NPROCESSORS_ONLN) && \
     DESTDIR=/sfcgal make install -j $(getconf _NPROCESSORS_ONLN) && \
     make clean && cp -R /sfcgal/* /
 
@@ -96,7 +95,7 @@ RUN wget https://github.com/pgRouting/pgrouting/archive/v3.4.2.tar.gz -O pgrouti
     mkdir pgrouting-src && cd pgrouting-src && tar xvzf ../pgrouting.tar.gz --strip-components=1 -C . && \
     mkdir build && \
     cd build && \
-    cmake -DCMAKE_BUILD_TYPE=Release .. && \
+    cmake .. && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
     make -j $(getconf _NPROCESSORS_ONLN) install && \
     echo 'trusted = true' >> /usr/local/pgsql/share/extension/pgrouting.control
@@ -356,7 +355,7 @@ RUN apt-get update && \
     wget https://github.com/timescale/timescaledb/archive/refs/tags/2.10.1.tar.gz -O timescaledb.tar.gz && \
     echo "6fca72a6ed0f6d32d2b3523951ede73dc5f9b0077b38450a029a5f411fdb8c73 timescaledb.tar.gz" | sha256sum --check && \
     mkdir timescaledb-src && cd timescaledb-src && tar xvzf ../timescaledb.tar.gz --strip-components=1 -C . && \
-    ./bootstrap -DSEND_TELEMETRY_DEFAULT:BOOL=OFF -DUSE_TELEMETRY:BOOL=OFF -DAPACHE_ONLY:BOOL=ON -DCMAKE_BUILD_TYPE=Release && \
+    ./bootstrap -DSEND_TELEMETRY_DEFAULT:BOOL=OFF -DUSE_TELEMETRY:BOOL=OFF -DAPACHE_ONLY:BOOL=ON && \
     cd build && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
     make install -j $(getconf _NPROCESSORS_ONLN) && \
@@ -411,7 +410,7 @@ RUN apt-get update && \
     mkdir kq_imcx-src && cd kq_imcx-src && tar xvzf ../kq_imcx.tar.gz --strip-components=1 -C . && \
     mkdir build && \
     cd build && \
-    cmake -DCMAKE_BUILD_TYPE=Release .. && \
+    cmake .. && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
     make -j $(getconf _NPROCESSORS_ONLN) install && \
     echo 'trusted = true' >> /usr/local/pgsql/share/extension/kq_imcx.control
@@ -433,108 +432,6 @@ RUN wget https://github.com/citusdata/pg_cron/archive/refs/tags/v1.5.2.tar.gz -O
     make -j $(getconf _NPROCESSORS_ONLN) install && \
     echo 'trusted = true' >> /usr/local/pgsql/share/extension/pg_cron.control
 
-#########################################################################################
-#
-# Layer "rdkit-pg-build"
-# compile rdkit extension
-#
-#########################################################################################
-FROM build-deps AS rdkit-pg-build
-COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
-
-RUN apt-get update && \
-    apt-get install -y \
-        cmake \
-        libboost-iostreams1.74-dev \
-        libboost-regex1.74-dev \
-        libboost-serialization1.74-dev \
-        libboost-system1.74-dev \
-        libeigen3-dev \
-        libfreetype6-dev
-
-ENV PATH "/usr/local/pgsql/bin/:/usr/local/pgsql/:$PATH"
-RUN wget https://github.com/rdkit/rdkit/archive/refs/tags/Release_2023_03_1.tar.gz -O rdkit.tar.gz && \
-    echo "db346afbd0ba52c843926a2a62f8a38c7b774ffab37eaf382d789a824f21996c rdkit.tar.gz" | sha256sum --check && \
-    mkdir rdkit-src && cd rdkit-src && tar xvzf ../rdkit.tar.gz --strip-components=1 -C . && \
-    cmake \
-        -D RDK_BUILD_CAIRO_SUPPORT=OFF \
-        -D RDK_BUILD_INCHI_SUPPORT=ON \
-        -D RDK_BUILD_AVALON_SUPPORT=ON \
-        -D RDK_BUILD_PYTHON_WRAPPERS=OFF \
-        -D RDK_BUILD_DESCRIPTORS3D=OFF \
-        -D RDK_BUILD_FREESASA_SUPPORT=OFF \
-        -D RDK_BUILD_COORDGEN_SUPPORT=ON \
-        -D RDK_BUILD_MOLINTERCHANGE_SUPPORT=OFF \
-        -D RDK_BUILD_YAEHMOP_SUPPORT=OFF \
-        -D RDK_BUILD_STRUCTCHECKER_SUPPORT=OFF \
-        -D RDK_USE_URF=OFF \
-        -D RDK_BUILD_PGSQL=ON \
-        -D RDK_PGSQL_STATIC=ON \
-        -D PostgreSQL_CONFIG=pg_config \
-        -D PostgreSQL_INCLUDE_DIR=`pg_config --includedir` \
-        -D PostgreSQL_TYPE_INCLUDE_DIR=`pg_config --includedir-server` \
-        -D PostgreSQL_LIBRARY_DIR=`pg_config --libdir` \
-        -D RDK_INSTALL_INTREE=OFF \
-        -D CMAKE_BUILD_TYPE=Release \
-        . && \
-    make -j $(getconf _NPROCESSORS_ONLN) && \
-    make -j $(getconf _NPROCESSORS_ONLN) install && \
-    echo 'trusted = true' >> /usr/local/pgsql/share/extension/rdkit.control
-
-#########################################################################################
-#
-# Layer "pg-uuidv7-pg-build"
-# compile pg_uuidv7 extension
-#
-#########################################################################################
-FROM build-deps AS pg-uuidv7-pg-build
-COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
-
-ENV PATH "/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/fboulnois/pg_uuidv7/archive/refs/tags/v1.0.1.tar.gz -O pg_uuidv7.tar.gz && \
-    echo "0d0759ab01b7fb23851ecffb0bce27822e1868a4a5819bfd276101c716637a7a pg_uuidv7.tar.gz" | sha256sum --check && \
-    mkdir pg_uuidv7-src && cd pg_uuidv7-src && tar xvzf ../pg_uuidv7.tar.gz --strip-components=1 -C . && \
-    make -j $(getconf _NPROCESSORS_ONLN) && \
-    make -j $(getconf _NPROCESSORS_ONLN) install && \
-    echo 'trusted = true' >> /usr/local/pgsql/share/extension/pg_uuidv7.control
-
-#########################################################################################
-#
-# Layer "pg-roaringbitmap-pg-build"
-# compile pg_roaringbitmap extension
-#
-#########################################################################################
-FROM build-deps AS pg-roaringbitmap-pg-build
-COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
-
-ENV PATH "/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/ChenHuajun/pg_roaringbitmap/archive/refs/tags/v0.5.4.tar.gz -O pg_roaringbitmap.tar.gz && \
-    echo "b75201efcb1c2d1b014ec4ae6a22769cc7a224e6e406a587f5784a37b6b5a2aa pg_roaringbitmap.tar.gz" | sha256sum --check && \
-    mkdir pg_roaringbitmap-src && cd pg_roaringbitmap-src && tar xvzf ../pg_roaringbitmap.tar.gz --strip-components=1 -C . && \
-    make -j $(getconf _NPROCESSORS_ONLN) && \
-    make -j $(getconf _NPROCESSORS_ONLN) install && \
-    echo 'trusted = true' >> /usr/local/pgsql/share/extension/roaringbitmap.control
-
-#########################################################################################
-#
-# Layer "pg-anon-pg-build"
-# compile anon extension
-#
-#########################################################################################
-FROM build-deps AS pg-anon-pg-build
-COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
-
-# Kaniko doesn't allow to do `${from#/usr/local/pgsql/}`, so we use `${from:17}` instead
-ENV PATH "/usr/local/pgsql/bin/:$PATH"
-RUN wget https://gitlab.com/dalibo/postgresql_anonymizer/-/archive/1.1.0/postgresql_anonymizer-1.1.0.tar.gz -O pg_anon.tar.gz && \
-    echo "08b09d2ff9b962f96c60db7e6f8e79cf7253eb8772516998fc35ece08633d3ad pg_anon.tar.gz" | sha256sum --check && \
-    mkdir pg_anon-src && cd pg_anon-src && tar xvzf ../pg_anon.tar.gz --strip-components=1 -C . && \
-    find /usr/local/pgsql -type f | sort  > /before.txt && \
-    make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
-    echo 'trusted = true' >> /usr/local/pgsql/share/extension/anon.control && \
-    find /usr/local/pgsql -type f | sort  > /after.txt && \
-    /bin/bash -c 'for from in $(comm -13 /before.txt /after.txt); do to=/extensions/anon/${from:17} && mkdir -p $(dirname ${to}) && cp -a ${from} ${to}; done'
-
 #########################################################################################
 #
 # Layer "rust extensions"
@@ -643,7 +540,6 @@ RUN wget https://github.com/pksunkara/pgx_ulid/archive/refs/tags/v0.1.0.tar.gz -
 #
 #########################################################################################
 FROM build-deps AS neon-pg-ext-build
-# Public extensions
 COPY --from=postgis-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=postgis-build /sfcgal/* /
 COPY --from=plv8-build /usr/local/pgsql/ /usr/local/pgsql/
@@ -668,9 +564,6 @@ COPY --from=pg-hint-plan-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=kq-imcx-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pg-cron-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pg-pgx-ulid-build /usr/local/pgsql/ /usr/local/pgsql/
-COPY --from=rdkit-pg-build /usr/local/pgsql/ /usr/local/pgsql/
-COPY --from=pg-uuidv7-pg-build /usr/local/pgsql/ /usr/local/pgsql/
-COPY --from=pg-roaringbitmap-pg-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY pgxn/ pgxn/
 
 RUN make -j $(getconf _NPROCESSORS_ONLN) \
@@ -692,9 +585,6 @@ RUN make -j $(getconf _NPROCESSORS_ONLN) \
 #
 #########################################################################################
 FROM $REPOSITORY/$IMAGE:$TAG AS compute-tools
-ARG BUILD_TAG
-ENV BUILD_TAG=$BUILD_TAG
-
 USER nonroot
 # Copy entire project to get Cargo.* files with proper dependencies for the whole project
 COPY --chown=nonroot . .
@@ -719,22 +609,6 @@ RUN rm -r /usr/local/pgsql/include
 # if they were to be used by other libraries.
 RUN rm /usr/local/pgsql/lib/lib*.a
 
-#########################################################################################
-#
-# Extenstion only
-#
-#########################################################################################
-FROM scratch AS postgres-extensions
-# After the transition this layer will include all extensitons.
-# As for now, it's only for new custom ones
-#
-# # Default extensions
-# COPY --from=postgres-cleanup-layer /usr/local/pgsql/share/extension /usr/local/pgsql/share/extension
-# COPY --from=postgres-cleanup-layer /usr/local/pgsql/lib             /usr/local/pgsql/lib
-# Custom extensions
-COPY --from=pg-anon-pg-build /extensions/anon/lib/ /extensions/anon/lib
-COPY --from=pg-anon-pg-build /extensions/anon/share/extension /extensions/anon/share/extension
-
 #########################################################################################
 #
 # Final layer
@@ -763,19 +637,14 @@ COPY --from=compute-tools --chown=postgres /home/nonroot/target/release-line-deb
 # libgeos, libgdal, libsfcgal1, libproj and libprotobuf-c1 for PostGIS
 # libxml2, libxslt1.1 for xml2
 # libzstd1 for zstd
-# libboost*, libfreetype6, and zlib1g for rdkit
 RUN apt update &&  \
     apt install --no-install-recommends -y \
         gdb \
+        locales \
         libicu67 \
         liblz4-1 \
         libreadline8 \
-        libboost-iostreams1.74.0 \
-        libboost-regex1.74.0 \
-        libboost-serialization1.74.0 \
-        libboost-system1.74.0 \
         libossp-uuid16 \
-        libfreetype6 \
         libgeos-c1v5 \
         libgdal28 \
         libproj19 \
@@ -785,9 +654,7 @@ RUN apt update &&  \
         libxslt1.1 \
         libzstd1 \
         libcurl4-openssl-dev \
-        locales \
-        procps \
-        zlib1g && \
+        procps && \
     rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/* && \
     localedef -i en_US -c -f UTF-8 -A /usr/share/locale/locale.alias en_US.UTF-8
 

Dockerfile.compute-tools

@@ -3,7 +3,6 @@
 ARG REPOSITORY=neondatabase
 ARG IMAGE=rust
 ARG TAG=pinned
-ARG BUILD_TAG
 
 FROM $REPOSITORY/$IMAGE:$TAG AS rust-build
 WORKDIR /home/nonroot
@@ -17,8 +16,6 @@ ENV CACHEPOT_S3_KEY_PREFIX=cachepot
 ARG CACHEPOT_BUCKET=neon-github-dev
 #ARG AWS_ACCESS_KEY_ID
 #ARG AWS_SECRET_ACCESS_KEY
-ARG BUILD_TAG
-ENV BUILD_TAG=$BUILD_TAG
 
 COPY . .
 

compute_tools/src/bin/compute_ctl.rs

@@ -54,15 +54,9 @@ use compute_tools::monitor::launch_monitor;
 use compute_tools::params::*;
 use compute_tools::spec::*;
 
-const BUILD_TAG_DEFAULT: &str = "local";
-
 fn main() -> Result<()> {
     init_tracing_and_logging(DEFAULT_LOG_LEVEL)?;
 
-    let build_tag = option_env!("BUILD_TAG").unwrap_or(BUILD_TAG_DEFAULT);
-
-    info!("build_tag: {build_tag}");
-
     let matches = cli().get_matches();
 
     let http_port = *matches

compute_tools/src/compute.rs

@@ -133,84 +133,6 @@ impl TryFrom<ComputeSpec> for ParsedSpec {
     }
 }
 
-/// Create special neon_superuser role, that's a slightly nerfed version of a real superuser
-/// that we give to customers
-fn create_neon_superuser(spec: &ComputeSpec, client: &mut Client) -> Result<()> {
-    let roles = spec
-        .cluster
-        .roles
-        .iter()
-        .map(|r| format!("'{}'", escape_literal(&r.name)))
-        .collect::<Vec<_>>();
-
-    let dbs = spec
-        .cluster
-        .databases
-        .iter()
-        .map(|db| format!("'{}'", escape_literal(&db.name)))
-        .collect::<Vec<_>>();
-
-    let roles_decl = if roles.is_empty() {
-        String::from("roles text[] := NULL;")
-    } else {
-        format!(
-            r#"
-               roles text[] := ARRAY(SELECT rolname
-                                     FROM pg_catalog.pg_roles
-                                     WHERE rolname IN ({}));"#,
-            roles.join(", ")
-        )
-    };
-
-    let database_decl = if dbs.is_empty() {
-        String::from("dbs text[] := NULL;")
-    } else {
-        format!(
-            r#"
-               dbs text[] := ARRAY(SELECT datname
-                                   FROM pg_catalog.pg_database
-                                   WHERE datname IN ({}));"#,
-            dbs.join(", ")
-        )
-    };
-
-    // ALL PRIVILEGES grants CREATE, CONNECT, and TEMPORARY on all databases
-    // (see https://www.postgresql.org/docs/current/ddl-priv.html)
-    let query = format!(
-        r#"
-            DO $$
-                DECLARE
-                    r text;
-                    {}
-                    {}
-                BEGIN
-                    IF NOT EXISTS (
-                        SELECT FROM pg_catalog.pg_roles WHERE rolname = 'neon_superuser')
-                    THEN
-                        CREATE ROLE neon_superuser CREATEDB CREATEROLE NOLOGIN IN ROLE pg_read_all_data, pg_write_all_data;
-                        IF array_length(roles, 1) IS NOT NULL THEN
-                            EXECUTE format('GRANT neon_superuser TO %s',
-                                           array_to_string(ARRAY(SELECT quote_ident(x) FROM unnest(roles) as x), ', '));
-                            FOREACH r IN ARRAY roles LOOP
-                                EXECUTE format('ALTER ROLE %s CREATEROLE CREATEDB', quote_ident(r));
-                            END LOOP;
-                        END IF;
-                        IF array_length(dbs, 1) IS NOT NULL THEN
-                            EXECUTE format('GRANT ALL PRIVILEGES ON DATABASE %s TO neon_superuser',
-                                           array_to_string(ARRAY(SELECT quote_ident(x) FROM unnest(dbs) as x), ', '));
-                        END IF;
-                    END IF;
-                END
-            $$;"#,
-        roles_decl, database_decl,
-    );
-    info!("Neon superuser created:\n{}", &query);
-    client
-        .simple_query(&query)
-        .map_err(|e| anyhow::anyhow!(e).context(query))?;
-    Ok(())
-}
-
 impl ComputeNode {
     pub fn set_status(&self, status: ComputeStatus) {
         let mut state = self.state.lock().unwrap();
@@ -235,7 +157,7 @@ impl ComputeNode {
 
     // Get basebackup from the libpq connection to pageserver using `connstr` and
     // unarchive it to `pgdata` directory overriding all its previous content.
-    #[instrument(skip_all, fields(%lsn))]
+    #[instrument(skip(self, compute_state))]
     fn get_basebackup(&self, compute_state: &ComputeState, lsn: Lsn) -> Result<()> {
         let spec = compute_state.pspec.as_ref().expect("spec must be set");
         let start_time = Utc::now();
@@ -277,7 +199,7 @@ impl ComputeNode {
 
     // Run `postgres` in a special mode with `--sync-safekeepers` argument
     // and return the reported LSN back to the caller.
-    #[instrument(skip_all)]
+    #[instrument(skip(self, storage_auth_token))]
     fn sync_safekeepers(&self, storage_auth_token: Option<String>) -> Result<Lsn> {
         let start_time = Utc::now();
 
@@ -322,7 +244,7 @@ impl ComputeNode {
 
     /// Do all the preparations like PGDATA directory creation, configuration,
     /// safekeepers sync, basebackup, etc.
-    #[instrument(skip_all)]
+    #[instrument(skip(self, compute_state))]
     pub fn prepare_pgdata(&self, compute_state: &ComputeState) -> Result<()> {
         let pspec = compute_state.pspec.as_ref().expect("spec must be set");
         let spec = &pspec.spec;
@@ -380,7 +302,7 @@ impl ComputeNode {
 
     /// Start Postgres as a child process and manage DBs/roles.
     /// After that this will hang waiting on the postmaster process to exit.
-    #[instrument(skip_all)]
+    #[instrument(skip(self))]
     pub fn start_postgres(
         &self,
         storage_auth_token: Option<String>,
@@ -404,7 +326,7 @@ impl ComputeNode {
     }
 
     /// Do initial configuration of the already started Postgres.
-    #[instrument(skip_all)]
+    #[instrument(skip(self, compute_state))]
     pub fn apply_config(&self, compute_state: &ComputeState) -> Result<()> {
         // If connection fails,
         // it may be the old node with `zenith_admin` superuser.
@@ -425,8 +347,6 @@ impl ComputeNode {
                     .map_err(|_| anyhow::anyhow!("invalid connstr"))?;
 
                 let mut client = Client::connect(zenith_admin_connstr.as_str(), NoTls)?;
-                // Disable forwarding so that users don't get a cloud_admin role
-                client.simple_query("SET neon.forward_ddl = false")?;
                 client.simple_query("CREATE USER cloud_admin WITH SUPERUSER")?;
                 client.simple_query("GRANT zenith_admin TO cloud_admin")?;
                 drop(client);
@@ -437,28 +357,31 @@ impl ComputeNode {
             Ok(client) => client,
         };
 
+        // Proceed with post-startup configuration. Note, that order of operations is important.
         // Disable DDL forwarding because control plane already knows about these roles/databases.
         client.simple_query("SET neon.forward_ddl = false")?;
-
-        // Proceed with post-startup configuration. Note, that order of operations is important.
         let spec = &compute_state.pspec.as_ref().expect("spec must be set").spec;
-        create_neon_superuser(spec, &mut client)?;
         handle_roles(spec, &mut client)?;
         handle_databases(spec, &mut client)?;
         handle_role_deletions(spec, self.connstr.as_str(), &mut client)?;
-        handle_grants(spec, self.connstr.as_str())?;
+        handle_grants(spec, self.connstr.as_str(), &mut client)?;
         handle_extensions(spec, &mut client)?;
 
         // 'Close' connection
         drop(client);
 
+        info!(
+            "finished configuration of compute for project {}",
+            spec.cluster.cluster_id.as_deref().unwrap_or("None")
+        );
+
         Ok(())
     }
 
     // We could've wrapped this around `pg_ctl reload`, but right now we don't use
     // `pg_ctl` for start / stop, so this just seems much easier to do as we already
     // have opened connection to Postgres and superuser access.
-    #[instrument(skip_all)]
+    #[instrument(skip(self, client))]
     fn pg_reload_conf(&self, client: &mut Client) -> Result<()> {
         client.simple_query("SELECT pg_reload_conf()")?;
         Ok(())
@@ -466,7 +389,7 @@ impl ComputeNode {
 
     /// Similar to `apply_config()`, but does a bit different sequence of operations,
     /// as it's used to reconfigure a previously started and configured Postgres node.
-    #[instrument(skip_all)]
+    #[instrument(skip(self))]
     pub fn reconfigure(&self) -> Result<()> {
         let spec = self.state.lock().unwrap().pspec.clone().unwrap().spec;
 
@@ -484,7 +407,7 @@ impl ComputeNode {
             handle_roles(&spec, &mut client)?;
             handle_databases(&spec, &mut client)?;
             handle_role_deletions(&spec, self.connstr.as_str(), &mut client)?;
-            handle_grants(&spec, self.connstr.as_str())?;
+            handle_grants(&spec, self.connstr.as_str(), &mut client)?;
             handle_extensions(&spec, &mut client)?;
         }
 
@@ -501,25 +424,25 @@ impl ComputeNode {
         Ok(())
     }
 
-    #[instrument(skip_all)]
+    #[instrument(skip(self))]
     pub fn start_compute(&self) -> Result<std::process::Child> {
         let compute_state = self.state.lock().unwrap().clone();
-        let pspec = compute_state.pspec.as_ref().expect("spec must be set");
+        let spec = compute_state.pspec.as_ref().expect("spec must be set");
         info!(
             "starting compute for project {}, operation {}, tenant {}, timeline {}",
-            pspec.spec.cluster.cluster_id.as_deref().unwrap_or("None"),
-            pspec.spec.operation_uuid.as_deref().unwrap_or("None"),
-            pspec.tenant_id,
-            pspec.timeline_id,
+            spec.spec.cluster.cluster_id.as_deref().unwrap_or("None"),
+            spec.spec.operation_uuid.as_deref().unwrap_or("None"),
+            spec.tenant_id,
+            spec.timeline_id,
         );
 
         self.prepare_pgdata(&compute_state)?;
 
         let start_time = Utc::now();
 
-        let pg = self.start_postgres(pspec.storage_auth_token.clone())?;
+        let pg = self.start_postgres(spec.storage_auth_token.clone())?;
 
-        if pspec.spec.mode == ComputeMode::Primary && !pspec.spec.skip_pg_catalog_updates {
+        if spec.spec.mode == ComputeMode::Primary {
             self.apply_config(&compute_state)?;
         }
 
@@ -539,11 +462,6 @@ impl ComputeNode {
         }
         self.set_status(ComputeStatus::Running);
 
-        info!(
-            "finished configuration of compute for project {}",
-            pspec.spec.cluster.cluster_id.as_deref().unwrap_or("None")
-        );
-
         Ok(pg)
     }
 

compute_tools/src/configurator.rs

@@ -8,7 +8,7 @@ use compute_api::responses::ComputeStatus;
 
 use crate::compute::ComputeNode;
 
-#[instrument(skip_all)]
+#[instrument(skip(compute))]
 fn configurator_main_loop(compute: &Arc<ComputeNode>) {
     info!("waiting for reconfiguration requests");
     loop {

compute_tools/src/pg_helpers.rs

@@ -17,7 +17,7 @@ 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
-pub fn escape_literal(s: &str) -> String {
+fn escape_literal(s: &str) -> String {
     s.replace('\'', "''").replace('\\', "\\\\")
 }
 
@@ -215,7 +215,7 @@ pub fn get_existing_dbs(client: &mut Client) -> Result<Vec<Database>> {
 /// Wait for Postgres to become ready to accept connections. It's ready to
 /// accept connections when the state-field in `pgdata/postmaster.pid` says
 /// 'ready'.
-#[instrument(skip_all, fields(pgdata = %pgdata.display()))]
+#[instrument(skip(pg))]
 pub fn wait_for_postgres(pg: &mut Child, pgdata: &Path) -> Result<()> {
     let pid_path = pgdata.join("postmaster.pid");
 

compute_tools/src/spec.rs

@@ -269,13 +269,17 @@ pub fn handle_roles(spec: &ComputeSpec, client: &mut Client) -> Result<()> {
                 xact.execute(query.as_str(), &[])?;
             }
             RoleAction::Create => {
-                let mut query: String = format!(
-                    "CREATE ROLE {} CREATEROLE CREATEDB IN ROLE neon_superuser",
-                    name.pg_quote()
-                );
+                let mut query: String = format!("CREATE ROLE {} ", name.pg_quote());
                 info!("role create query: '{}'", &query);
                 query.push_str(&role.to_pg_options());
                 xact.execute(query.as_str(), &[])?;
+
+                let grant_query = format!(
+                    "GRANT pg_read_all_data, pg_write_all_data TO {}",
+                    name.pg_quote()
+                );
+                xact.execute(grant_query.as_str(), &[])?;
+                info!("role grant query: '{}'", &grant_query);
             }
         }
 
@@ -472,11 +476,6 @@ pub fn handle_databases(spec: &ComputeSpec, client: &mut Client) -> Result<()> {
                 query.push_str(&db.to_pg_options());
                 let _guard = info_span!("executing", query).entered();
                 client.execute(query.as_str(), &[])?;
-                let grant_query: String = format!(
-                    "GRANT ALL PRIVILEGES ON DATABASE {} TO neon_superuser",
-                    name.pg_quote()
-                );
-                client.execute(grant_query.as_str(), &[])?;
             }
         };
 
@@ -496,9 +495,35 @@ pub fn handle_databases(spec: &ComputeSpec, client: &mut Client) -> Result<()> {
 /// Grant CREATE ON DATABASE to the database owner and do some other alters and grants
 /// to allow users creating trusted extensions and re-creating `public` schema, for example.
 #[instrument(skip_all)]
-pub fn handle_grants(spec: &ComputeSpec, connstr: &str) -> Result<()> {
+pub fn handle_grants(spec: &ComputeSpec, connstr: &str, client: &mut Client) -> Result<()> {
     info!("cluster spec grants:");
 
+    // We now have a separate `web_access` role to connect to the database
+    // via the web interface and proxy link auth. And also we grant a
+    // read / write all data privilege to every role. So also grant
+    // create to everyone.
+    // XXX: later we should stop messing with Postgres ACL in such horrible
+    // ways.
+    let roles = spec
+        .cluster
+        .roles
+        .iter()
+        .map(|r| r.name.pg_quote())
+        .collect::<Vec<_>>();
+
+    for db in &spec.cluster.databases {
+        let dbname = &db.name;
+
+        let query: String = format!(
+            "GRANT CREATE ON DATABASE {} TO {}",
+            dbname.pg_quote(),
+            roles.join(", ")
+        );
+        info!("grant query {}", &query);
+
+        client.execute(query.as_str(), &[])?;
+    }
+
     // Do some per-database access adjustments. We'd better do this at db creation time,
     // but CREATE DATABASE isn't transactional. So we cannot create db + do some grants
     // atomically.

control_plane/src/bin/neon_local.rs

@@ -308,8 +308,7 @@ fn handle_init(init_match: &ArgMatches) -> anyhow::Result<LocalEnv> {
 
     let mut env =
         LocalEnv::parse_config(&toml_file).context("Failed to create neon configuration")?;
-    let force = init_match.get_flag("force");
-    env.init(pg_version, force)
+    env.init(pg_version)
         .context("Failed to initialize neon repository")?;
 
     // Initialize pageserver, create initial tenant and timeline.
@@ -1014,13 +1013,6 @@ fn cli() -> Command {
         .help("If set, the node will be a hot replica on the specified timeline")
         .required(false);
 
-    let force_arg = Arg::new("force")
-        .value_parser(value_parser!(bool))
-        .long("force")
-        .action(ArgAction::SetTrue)
-        .help("Force initialization even if the repository is not empty")
-        .required(false);
-
     Command::new("Neon CLI")
         .arg_required_else_help(true)
         .version(GIT_VERSION)
@@ -1036,7 +1028,6 @@ fn cli() -> Command {
                         .value_name("config"),
                 )
                 .arg(pg_version_arg.clone())
-                .arg(force_arg)
         )
         .subcommand(
             Command::new("timeline")

control_plane/src/endpoint.rs

@@ -67,7 +67,6 @@ pub struct EndpointConf {
     pg_port: u16,
     http_port: u16,
     pg_version: u32,
-    skip_pg_catalog_updates: bool,
 }
 
 //
@@ -136,7 +135,6 @@ impl ComputeControlPlane {
             mode,
             tenant_id,
             pg_version,
-            skip_pg_catalog_updates: false,
         });
 
         ep.create_endpoint_dir()?;
@@ -150,7 +148,6 @@ impl ComputeControlPlane {
                 http_port,
                 pg_port,
                 pg_version,
-                skip_pg_catalog_updates: false,
             })?,
         )?;
         std::fs::write(
@@ -186,9 +183,6 @@ pub struct Endpoint {
     // the endpoint runs in.
     pub env: LocalEnv,
     pageserver: Arc<PageServerNode>,
-
-    // Optimizations
-    skip_pg_catalog_updates: bool,
 }
 
 impl Endpoint {
@@ -222,7 +216,6 @@ impl Endpoint {
             mode: conf.mode,
             tenant_id: conf.tenant_id,
             pg_version: conf.pg_version,
-            skip_pg_catalog_updates: conf.skip_pg_catalog_updates,
         })
     }
 
@@ -457,7 +450,6 @@ impl Endpoint {
 
         // Create spec file
         let spec = ComputeSpec {
-            skip_pg_catalog_updates: self.skip_pg_catalog_updates,
             format_version: 1.0,
             operation_uuid: None,
             cluster: Cluster {

control_plane/src/local_env.rs

@@ -364,7 +364,7 @@ impl LocalEnv {
     //
     // Initialize a new Neon repository
     //
-    pub fn init(&mut self, pg_version: u32, force: bool) -> anyhow::Result<()> {
+    pub fn init(&mut self, pg_version: u32) -> anyhow::Result<()> {
         // check if config already exists
         let base_path = &self.base_data_dir;
         ensure!(
@@ -372,29 +372,11 @@ impl LocalEnv {
             "repository base path is missing"
         );
 
-        if base_path.exists() {
-            if force {
-                println!("removing all contents of '{}'", base_path.display());
-                // instead of directly calling `remove_dir_all`, we keep the original dir but removing
-                // all contents inside. This helps if the developer symbol links another directory (i.e.,
-                // S3 local SSD) to the `.neon` base directory.
-                for entry in std::fs::read_dir(base_path)? {
-                    let entry = entry?;
-                    let path = entry.path();
-                    if path.is_dir() {
-                        fs::remove_dir_all(&path)?;
-                    } else {
-                        fs::remove_file(&path)?;
-                    }
-                }
-            } else {
-                bail!(
-                    "directory '{}' already exists. Perhaps already initialized? (Hint: use --force to remove all contents)",
-                    base_path.display()
-                );
-            }
-        }
-
+        ensure!(
+            !base_path.exists(),
+            "directory '{}' already exists. Perhaps already initialized?",
+            base_path.display()
+        );
         if !self.pg_bin_dir(pg_version)?.join("postgres").exists() {
             bail!(
                 "Can't find postgres binary at {}",
@@ -410,9 +392,7 @@ impl LocalEnv {
             }
         }
 
-        if !base_path.exists() {
-            fs::create_dir(base_path)?;
-        }
+        fs::create_dir(base_path)?;
 
         // Generate keypair for JWT.
         //

libs/compute_api/src/spec.rs

@@ -27,12 +27,6 @@ pub struct ComputeSpec {
     pub cluster: Cluster,
     pub delta_operations: Option<Vec<DeltaOp>>,
 
-    /// An optinal hint that can be passed to speed up startup time if we know
-    /// that no pg catalog mutations (like role creation, database creation,
-    /// extension creation) need to be done on the actual database to start.
-    #[serde(default)] // Default false
-    pub skip_pg_catalog_updates: bool,
-
     // Information needed to connect to the storage layer.
     //
     // `tenant_id`, `timeline_id` and `pageserver_connstring` are always needed.
@@ -148,14 +142,4 @@ mod tests {
         let file = File::open("tests/cluster_spec.json").unwrap();
         let _spec: ComputeSpec = serde_json::from_reader(file).unwrap();
     }
-
-    #[test]
-    fn parse_unknown_fields() {
-        // Forward compatibility test
-        let file = File::open("tests/cluster_spec.json").unwrap();
-        let mut json: serde_json::Value = serde_json::from_reader(file).unwrap();
-        let ob = json.as_object_mut().unwrap();
-        ob.insert("unknown_field_123123123".into(), "hello".into());
-        let _spec: ComputeSpec = serde_json::from_value(json).unwrap();
-    }
 }

libs/metrics/src/lib.rs

@@ -23,7 +23,6 @@ use prometheus::{Registry, Result};
 pub mod launch_timestamp;
 mod wrappers;
 pub use wrappers::{CountedReader, CountedWriter};
-pub mod metric_vec_duration;
 
 pub type UIntGauge = GenericGauge<AtomicU64>;
 pub type UIntGaugeVec = GenericGaugeVec<AtomicU64>;

libs/metrics/src/metric_vec_duration.rs

@@ -1,23 +0,0 @@
-//! Helpers for observing duration on HistogramVec / CounterVec / GaugeVec / MetricVec<T>.
-
-use std::{future::Future, time::Instant};
-
-pub trait DurationResultObserver {
-    fn observe_result<T, E>(&self, res: &Result<T, E>, duration: std::time::Duration);
-}
-
-pub async fn observe_async_block_duration_by_result<
-    T,
-    E,
-    F: Future<Output = Result<T, E>>,
-    O: DurationResultObserver,
->(
-    observer: &O,
-    block: F,
-) -> Result<T, E> {
-    let start = Instant::now();
-    let result = block.await;
-    let duration = start.elapsed();
-    observer.observe_result(&result, duration);
-    result
-}

libs/remote_storage/src/lib.rs

@@ -70,14 +70,6 @@ impl RemotePath {
     pub fn join(&self, segment: &Path) -> Self {
         Self(self.0.join(segment))
     }
-
-    pub fn get_path(&self) -> &PathBuf {
-        &self.0
-    }
-
-    pub fn extension(&self) -> Option<&str> {
-        self.0.extension()?.to_str()
-    }
 }
 
 /// Storage (potentially remote) API to manage its state.
@@ -94,19 +86,6 @@ pub trait RemoteStorage: Send + Sync + 'static {
         prefix: Option<&RemotePath>,
     ) -> Result<Vec<RemotePath>, DownloadError>;
 
-    /// Lists all files in directory "recursively"
-    /// (not really recursively, because AWS has a flat namespace)
-    /// Note: This is subtely different than list_prefixes,
-    /// because it is for listing files instead of listing
-    /// names sharing common prefixes.
-    /// For example,
-    /// list_files("foo/bar") = ["foo/bar/cat123.txt",
-    /// "foo/bar/cat567.txt", "foo/bar/dog123.txt", "foo/bar/dog456.txt"]
-    /// whereas,
-    /// list_prefixes("foo/bar/") = ["cat", "dog"]
-    /// See `test_real_s3.rs` for more details.
-    async fn list_files(&self, folder: Option<&RemotePath>) -> anyhow::Result<Vec<RemotePath>>;
-
     /// Streams the local file contents into remote into the remote storage entry.
     async fn upload(
         &self,
@@ -132,8 +111,6 @@ pub trait RemoteStorage: Send + Sync + 'static {
     ) -> Result<Download, DownloadError>;
 
     async fn delete(&self, path: &RemotePath) -> anyhow::Result<()>;
-
-    async fn delete_objects<'a>(&self, paths: &'a [RemotePath]) -> anyhow::Result<()>;
 }
 
 pub struct Download {
@@ -195,14 +172,6 @@ impl GenericRemoteStorage {
         }
     }
 
-    pub async fn list_files(&self, folder: Option<&RemotePath>) -> anyhow::Result<Vec<RemotePath>> {
-        match self {
-            Self::LocalFs(s) => s.list_files(folder).await,
-            Self::AwsS3(s) => s.list_files(folder).await,
-            Self::Unreliable(s) => s.list_files(folder).await,
-        }
-    }
-
     pub async fn upload(
         &self,
         from: impl io::AsyncRead + Unpin + Send + Sync + 'static,
@@ -254,14 +223,6 @@ impl GenericRemoteStorage {
             Self::Unreliable(s) => s.delete(path).await,
         }
     }
-
-    pub async fn delete_objects<'a>(&self, paths: &'a [RemotePath]) -> anyhow::Result<()> {
-        match self {
-            Self::LocalFs(s) => s.delete_objects(paths).await,
-            Self::AwsS3(s) => s.delete_objects(paths).await,
-            Self::Unreliable(s) => s.delete_objects(paths).await,
-        }
-    }
 }
 
 impl GenericRemoteStorage {

libs/remote_storage/src/local_fs.rs

@@ -48,14 +48,6 @@ impl LocalFs {
         Ok(Self { storage_root })
     }
 
-    // mirrors S3Bucket::s3_object_to_relative_path
-    fn local_file_to_relative_path(&self, key: PathBuf) -> RemotePath {
-        let relative_path = key
-            .strip_prefix(&self.storage_root)
-            .expect("relative path must contain storage_root as prefix");
-        RemotePath(relative_path.into())
-    }
-
     async fn read_storage_metadata(
         &self,
         file_path: &Path,
@@ -140,34 +132,6 @@ impl RemoteStorage for LocalFs {
         Ok(prefixes)
     }
 
-    // recursively lists all files in a directory,
-    // mirroring the `list_files` for `s3_bucket`
-    async fn list_files(&self, folder: Option<&RemotePath>) -> anyhow::Result<Vec<RemotePath>> {
-        let full_path = match folder {
-            Some(folder) => folder.with_base(&self.storage_root),
-            None => self.storage_root.clone(),
-        };
-        let mut files = vec![];
-        let mut directory_queue = vec![full_path.clone()];
-
-        while !directory_queue.is_empty() {
-            let cur_folder = directory_queue
-                .pop()
-                .expect("queue cannot be empty: we just checked");
-            let mut entries = fs::read_dir(cur_folder.clone()).await?;
-            while let Some(entry) = entries.next_entry().await? {
-                let file_name: PathBuf = entry.file_name().into();
-                let full_file_name = cur_folder.clone().join(&file_name);
-                let file_remote_path = self.local_file_to_relative_path(full_file_name.clone());
-                files.push(file_remote_path.clone());
-                if full_file_name.is_dir() {
-                    directory_queue.push(full_file_name);
-                }
-            }
-        }
-        Ok(files)
-    }
-
     async fn upload(
         &self,
         data: impl io::AsyncRead + Unpin + Send + Sync + 'static,
@@ -356,13 +320,6 @@ impl RemoteStorage for LocalFs {
             .await
             .map_err(|e| anyhow::anyhow!(e))?)
     }
-
-    async fn delete_objects<'a>(&self, paths: &'a [RemotePath]) -> anyhow::Result<()> {
-        for path in paths {
-            self.delete(path).await?
-        }
-        Ok(())
-    }
 }
 
 fn storage_metadata_path(original_path: &Path) -> PathBuf {

libs/remote_storage/src/s3_bucket.rs

@@ -17,7 +17,6 @@ use aws_sdk_s3::{
     error::SdkError,
     operation::get_object::GetObjectError,
     primitives::ByteStream,
-    types::{Delete, ObjectIdentifier},
     Client,
 };
 use aws_smithy_http::body::SdkBody;
@@ -34,8 +33,6 @@ use crate::{
     Download, DownloadError, RemotePath, RemoteStorage, S3Config, REMOTE_STORAGE_PREFIX_SEPARATOR,
 };
 
-const MAX_DELETE_OBJECTS_REQUEST_SIZE: usize = 1000;
-
 pub(super) mod metrics {
     use metrics::{register_int_counter_vec, IntCounterVec};
     use once_cell::sync::Lazy;
@@ -84,24 +81,12 @@ pub(super) mod metrics {
             .inc();
     }
 
-    pub fn inc_delete_objects(count: u64) {
-        S3_REQUESTS_COUNT
-            .with_label_values(&["delete_object"])
-            .inc_by(count);
-    }
-
     pub fn inc_delete_object_fail() {
         S3_REQUESTS_FAIL_COUNT
             .with_label_values(&["delete_object"])
             .inc();
     }
 
-    pub fn inc_delete_objects_fail(count: u64) {
-        S3_REQUESTS_FAIL_COUNT
-            .with_label_values(&["delete_object"])
-            .inc_by(count);
-    }
-
     pub fn inc_list_objects() {
         S3_REQUESTS_COUNT.with_label_values(&["list_objects"]).inc();
     }
@@ -347,51 +332,6 @@ impl RemoteStorage for S3Bucket {
         Ok(document_keys)
     }
 
-    /// See the doc for `RemoteStorage::list_files`
-    async fn list_files(&self, folder: Option<&RemotePath>) -> anyhow::Result<Vec<RemotePath>> {
-        let folder_name = folder
-            .map(|p| self.relative_path_to_s3_object(p))
-            .or_else(|| self.prefix_in_bucket.clone());
-
-        // AWS may need to break the response into several parts
-        let mut continuation_token = None;
-        let mut all_files = vec![];
-        loop {
-            let _guard = self
-                .concurrency_limiter
-                .acquire()
-                .await
-                .context("Concurrency limiter semaphore got closed during S3 list_files")?;
-            metrics::inc_list_objects();
-
-            let response = self
-                .client
-                .list_objects_v2()
-                .bucket(self.bucket_name.clone())
-                .set_prefix(folder_name.clone())
-                .set_continuation_token(continuation_token)
-                .set_max_keys(self.max_keys_per_list_response)
-                .send()
-                .await
-                .map_err(|e| {
-                    metrics::inc_list_objects_fail();
-                    e
-                })
-                .context("Failed to list files in S3 bucket")?;
-
-            for object in response.contents().unwrap_or_default() {
-                let object_path = object.key().expect("response does not contain a key");
-                let remote_path = self.s3_object_to_relative_path(object_path);
-                all_files.push(remote_path);
-            }
-            match response.next_continuation_token {
-                Some(new_token) => continuation_token = Some(new_token),
-                None => break,
-            }
-        }
-        Ok(all_files)
-    }
-
     async fn upload(
         &self,
         from: impl io::AsyncRead + Unpin + Send + Sync + 'static,
@@ -456,50 +396,6 @@ impl RemoteStorage for S3Bucket {
         })
         .await
     }
-    async fn delete_objects<'a>(&self, paths: &'a [RemotePath]) -> anyhow::Result<()> {
-        let _guard = self
-            .concurrency_limiter
-            .acquire()
-            .await
-            .context("Concurrency limiter semaphore got closed during S3 delete")?;
-
-        let mut delete_objects = Vec::with_capacity(paths.len());
-        for path in paths {
-            let obj_id = ObjectIdentifier::builder()
-                .set_key(Some(self.relative_path_to_s3_object(path)))
-                .build();
-            delete_objects.push(obj_id);
-        }
-
-        for chunk in delete_objects.chunks(MAX_DELETE_OBJECTS_REQUEST_SIZE) {
-            metrics::inc_delete_objects(chunk.len() as u64);
-
-            let resp = self
-                .client
-                .delete_objects()
-                .bucket(self.bucket_name.clone())
-                .delete(Delete::builder().set_objects(Some(chunk.to_vec())).build())
-                .send()
-                .await;
-
-            match resp {
-                Ok(resp) => {
-                    if let Some(errors) = resp.errors {
-                        metrics::inc_delete_objects_fail(errors.len() as u64);
-                        return Err(anyhow::format_err!(
-                            "Failed to delete {} objects",
-                            errors.len()
-                        ));
-                    }
-                }
-                Err(e) => {
-                    metrics::inc_delete_objects_fail(chunk.len() as u64);
-                    return Err(e.into());
-                }
-            }
-        }
-        Ok(())
-    }
 
     async fn delete(&self, path: &RemotePath) -> anyhow::Result<()> {
         let _guard = self

libs/remote_storage/src/simulate_failures.rs

@@ -24,7 +24,6 @@ enum RemoteOp {
     Upload(RemotePath),
     Download(RemotePath),
     Delete(RemotePath),
-    DeleteObjects(Vec<RemotePath>),
 }
 
 impl UnreliableWrapper {
@@ -83,11 +82,6 @@ impl RemoteStorage for UnreliableWrapper {
         self.inner.list_prefixes(prefix).await
     }
 
-    async fn list_files(&self, folder: Option<&RemotePath>) -> anyhow::Result<Vec<RemotePath>> {
-        self.attempt(RemoteOp::ListPrefixes(folder.cloned()))?;
-        self.inner.list_files(folder).await
-    }
-
     async fn upload(
         &self,
         data: impl tokio::io::AsyncRead + Unpin + Send + Sync + 'static,
@@ -125,21 +119,4 @@ impl RemoteStorage for UnreliableWrapper {
         self.attempt(RemoteOp::Delete(path.clone()))?;
         self.inner.delete(path).await
     }
-
-    async fn delete_objects<'a>(&self, paths: &'a [RemotePath]) -> anyhow::Result<()> {
-        self.attempt(RemoteOp::DeleteObjects(paths.to_vec()))?;
-        let mut error_counter = 0;
-        for path in paths {
-            if (self.delete(path).await).is_err() {
-                error_counter += 1;
-            }
-        }
-        if error_counter > 0 {
-            return Err(anyhow::anyhow!(
-                "failed to delete {} objects",
-                error_counter
-            ));
-        }
-        Ok(())
-    }
 }

libs/remote_storage/tests/test_real_s3.rs

@@ -88,58 +88,6 @@ async fn s3_pagination_should_work(ctx: &mut MaybeEnabledS3WithTestBlobs) -> any
     Ok(())
 }
 
-/// Tests that S3 client can list all files in a folder, even if the response comes paginated and requirees multiple S3 queries.
-/// Uses real S3 and requires [`ENABLE_REAL_S3_REMOTE_STORAGE_ENV_VAR_NAME`] and related S3 cred env vars specified. Test will skip real code and pass if env vars not set.
-/// See `s3_pagination_should_work` for more information.
-///
-/// First, create a set of S3 objects with keys `random_prefix/folder{j}/blob_{i}.txt` in [`upload_s3_data`]
-/// Then performs the following queries:
-///    1. `list_files(None)`. This should return all files `random_prefix/folder{j}/blob_{i}.txt`
-///    2. `list_files("folder1")`.  This  should return all files `random_prefix/folder1/blob_{i}.txt`
-#[test_context(MaybeEnabledS3WithSimpleTestBlobs)]
-#[tokio::test]
-async fn s3_list_files_works(ctx: &mut MaybeEnabledS3WithSimpleTestBlobs) -> anyhow::Result<()> {
-    let ctx = match ctx {
-        MaybeEnabledS3WithSimpleTestBlobs::Enabled(ctx) => ctx,
-        MaybeEnabledS3WithSimpleTestBlobs::Disabled => return Ok(()),
-        MaybeEnabledS3WithSimpleTestBlobs::UploadsFailed(e, _) => {
-            anyhow::bail!("S3 init failed: {e:?}")
-        }
-    };
-    let test_client = Arc::clone(&ctx.enabled.client);
-    let base_prefix =
-        RemotePath::new(Path::new("folder1")).context("common_prefix construction")?;
-    let root_files = test_client
-        .list_files(None)
-        .await
-        .context("client list root files failure")?
-        .into_iter()
-        .collect::<HashSet<_>>();
-    assert_eq!(
-        root_files,
-        ctx.remote_blobs.clone(),
-        "remote storage list_files on root mismatches with the uploads."
-    );
-    let nested_remote_files = test_client
-        .list_files(Some(&base_prefix))
-        .await
-        .context("client list nested files failure")?
-        .into_iter()
-        .collect::<HashSet<_>>();
-    let trim_remote_blobs: HashSet<_> = ctx
-        .remote_blobs
-        .iter()
-        .map(|x| x.get_path().to_str().expect("must be valid name"))
-        .filter(|x| x.starts_with("folder1"))
-        .map(|x| RemotePath::new(Path::new(x)).expect("must be valid name"))
-        .collect();
-    assert_eq!(
-        nested_remote_files, trim_remote_blobs,
-        "remote storage list_files on subdirrectory mismatches with the uploads."
-    );
-    Ok(())
-}
-
 #[test_context(MaybeEnabledS3)]
 #[tokio::test]
 async fn s3_delete_non_exising_works(ctx: &mut MaybeEnabledS3) -> anyhow::Result<()> {
@@ -159,52 +107,6 @@ async fn s3_delete_non_exising_works(ctx: &mut MaybeEnabledS3) -> anyhow::Result
     Ok(())
 }
 
-#[test_context(MaybeEnabledS3)]
-#[tokio::test]
-async fn s3_delete_objects_works(ctx: &mut MaybeEnabledS3) -> anyhow::Result<()> {
-    let ctx = match ctx {
-        MaybeEnabledS3::Enabled(ctx) => ctx,
-        MaybeEnabledS3::Disabled => return Ok(()),
-    };
-
-    let path1 = RemotePath::new(&PathBuf::from(format!("{}/path1", ctx.base_prefix,)))
-        .with_context(|| "RemotePath conversion")?;
-
-    let path2 = RemotePath::new(&PathBuf::from(format!("{}/path2", ctx.base_prefix,)))
-        .with_context(|| "RemotePath conversion")?;
-
-    let path3 = RemotePath::new(&PathBuf::from(format!("{}/path3", ctx.base_prefix,)))
-        .with_context(|| "RemotePath conversion")?;
-
-    let data1 = "remote blob data1".as_bytes();
-    let data1_len = data1.len();
-    let data2 = "remote blob data2".as_bytes();
-    let data2_len = data2.len();
-    let data3 = "remote blob data3".as_bytes();
-    let data3_len = data3.len();
-    ctx.client
-        .upload(std::io::Cursor::new(data1), data1_len, &path1, None)
-        .await?;
-
-    ctx.client
-        .upload(std::io::Cursor::new(data2), data2_len, &path2, None)
-        .await?;
-
-    ctx.client
-        .upload(std::io::Cursor::new(data3), data3_len, &path3, None)
-        .await?;
-
-    ctx.client.delete_objects(&[path1, path2]).await?;
-
-    let prefixes = ctx.client.list_prefixes(None).await?;
-
-    assert_eq!(prefixes.len(), 1);
-
-    ctx.client.delete_objects(&[path3]).await?;
-
-    Ok(())
-}
-
 fn ensure_logging_ready() {
     LOGGING_DONE.get_or_init(|| {
         utils::logging::init(
@@ -315,66 +217,6 @@ impl AsyncTestContext for MaybeEnabledS3WithTestBlobs {
     }
 }
 
-// NOTE: the setups for the list_prefixes test and the list_files test are very similar
-// However, they are not idential. The list_prefixes function is concerned with listing prefixes,
-// whereas the list_files function is concerned with listing files.
-// See `RemoteStorage::list_files` documentation for more details
-enum MaybeEnabledS3WithSimpleTestBlobs {
-    Enabled(S3WithSimpleTestBlobs),
-    Disabled,
-    UploadsFailed(anyhow::Error, S3WithSimpleTestBlobs),
-}
-struct S3WithSimpleTestBlobs {
-    enabled: EnabledS3,
-    remote_blobs: HashSet<RemotePath>,
-}
-
-#[async_trait::async_trait]
-impl AsyncTestContext for MaybeEnabledS3WithSimpleTestBlobs {
-    async fn setup() -> Self {
-        ensure_logging_ready();
-        if env::var(ENABLE_REAL_S3_REMOTE_STORAGE_ENV_VAR_NAME).is_err() {
-            info!(
-                "`{}` env variable is not set, skipping the test",
-                ENABLE_REAL_S3_REMOTE_STORAGE_ENV_VAR_NAME
-            );
-            return Self::Disabled;
-        }
-
-        let max_keys_in_list_response = 10;
-        let upload_tasks_count = 1 + (2 * usize::try_from(max_keys_in_list_response).unwrap());
-
-        let enabled = EnabledS3::setup(Some(max_keys_in_list_response)).await;
-
-        match upload_simple_s3_data(&enabled.client, upload_tasks_count).await {
-            ControlFlow::Continue(uploads) => {
-                info!("Remote objects created successfully");
-
-                Self::Enabled(S3WithSimpleTestBlobs {
-                    enabled,
-                    remote_blobs: uploads,
-                })
-            }
-            ControlFlow::Break(uploads) => Self::UploadsFailed(
-                anyhow::anyhow!("One or multiple blobs failed to upload to S3"),
-                S3WithSimpleTestBlobs {
-                    enabled,
-                    remote_blobs: uploads,
-                },
-            ),
-        }
-    }
-
-    async fn teardown(self) {
-        match self {
-            Self::Disabled => {}
-            Self::Enabled(ctx) | Self::UploadsFailed(_, ctx) => {
-                cleanup(&ctx.enabled.client, ctx.remote_blobs).await;
-            }
-        }
-    }
-}
-
 fn create_s3_client(
     max_keys_per_list_response: Option<i32>,
 ) -> anyhow::Result<Arc<GenericRemoteStorage>> {
@@ -385,7 +227,7 @@ fn create_s3_client(
     let random_prefix_part = std::time::SystemTime::now()
         .duration_since(UNIX_EPOCH)
         .context("random s3 test prefix part calculation")?
-        .as_nanos();
+        .as_millis();
     let remote_storage_config = RemoteStorageConfig {
         max_concurrent_syncs: NonZeroUsize::new(100).unwrap(),
         max_sync_errors: NonZeroU32::new(5).unwrap(),
@@ -491,52 +333,3 @@ async fn cleanup(client: &Arc<GenericRemoteStorage>, objects_to_delete: HashSet<
         }
     }
 }
-
-// Uploads files `folder{j}/blob{i}.txt`. See test description for more details.
-async fn upload_simple_s3_data(
-    client: &Arc<GenericRemoteStorage>,
-    upload_tasks_count: usize,
-) -> ControlFlow<HashSet<RemotePath>, HashSet<RemotePath>> {
-    info!("Creating {upload_tasks_count} S3 files");
-    let mut upload_tasks = JoinSet::new();
-    for i in 1..upload_tasks_count + 1 {
-        let task_client = Arc::clone(client);
-        upload_tasks.spawn(async move {
-            let blob_path = PathBuf::from(format!("folder{}/blob_{}.txt", i / 7, i));
-            let blob_path = RemotePath::new(&blob_path)
-                .with_context(|| format!("{blob_path:?} to RemotePath conversion"))?;
-            debug!("Creating remote item {i} at path {blob_path:?}");
-
-            let data = format!("remote blob data {i}").into_bytes();
-            let data_len = data.len();
-            task_client
-                .upload(std::io::Cursor::new(data), data_len, &blob_path, None)
-                .await?;
-
-            Ok::<_, anyhow::Error>(blob_path)
-        });
-    }
-
-    let mut upload_tasks_failed = false;
-    let mut uploaded_blobs = HashSet::with_capacity(upload_tasks_count);
-    while let Some(task_run_result) = upload_tasks.join_next().await {
-        match task_run_result
-            .context("task join failed")
-            .and_then(|task_result| task_result.context("upload task failed"))
-        {
-            Ok(upload_path) => {
-                uploaded_blobs.insert(upload_path);
-            }
-            Err(e) => {
-                error!("Upload task failed: {e:?}");
-                upload_tasks_failed = true;
-            }
-        }
-    }
-
-    if upload_tasks_failed {
-        ControlFlow::Break(uploaded_blobs)
-    } else {
-        ControlFlow::Continue(uploaded_blobs)
-    }
-}

libs/utils/src/http/endpoint.rs

@@ -1,18 +1,19 @@
 use crate::auth::{Claims, JwtAuth};
 use crate::http::error::{api_error_handler, route_error_handler, ApiError};
-use anyhow::Context;
+use anyhow::{anyhow, Context};
 use hyper::header::{HeaderName, AUTHORIZATION};
 use hyper::http::HeaderValue;
 use hyper::Method;
-use hyper::{header::CONTENT_TYPE, Body, Request, Response};
+use hyper::{header::CONTENT_TYPE, Body, Request, Response, Server};
 use metrics::{register_int_counter, Encoder, IntCounter, TextEncoder};
 use once_cell::sync::Lazy;
 use routerify::ext::RequestExt;
-use routerify::{Middleware, RequestInfo, Router, RouterBuilder};
+use routerify::{Middleware, RequestInfo, Router, RouterBuilder, RouterService};
 use tokio::task::JoinError;
 use tracing::{self, debug, info, info_span, warn, Instrument};
 
 use std::future::Future;
+use std::net::TcpListener;
 use std::str::FromStr;
 
 static SERVE_METRICS_COUNT: Lazy<IntCounter> = Lazy::new(|| {
@@ -347,6 +348,40 @@ pub fn check_permission_with(
     }
 }
 
+///
+/// Start listening for HTTP requests on given socket.
+///
+/// 'shutdown_future' can be used to stop. If the Future becomes
+/// ready, we stop listening for new requests, and the function returns.
+///
+pub fn serve_thread_main<S>(
+    router_builder: RouterBuilder<hyper::Body, ApiError>,
+    listener: TcpListener,
+    shutdown_future: S,
+) -> anyhow::Result<()>
+where
+    S: Future<Output = ()> + Send + Sync,
+{
+    info!("Starting an HTTP endpoint at {}", listener.local_addr()?);
+
+    // Create a Service from the router above to handle incoming requests.
+    let service = RouterService::new(router_builder.build().map_err(|err| anyhow!(err))?).unwrap();
+
+    // Enter a single-threaded tokio runtime bound to the current thread
+    let runtime = tokio::runtime::Builder::new_current_thread()
+        .enable_all()
+        .build()?;
+
+    let _guard = runtime.enter();
+
+    let server = Server::from_tcp(listener)?
+        .serve(service)
+        .with_graceful_shutdown(shutdown_future);
+
+    runtime.block_on(server)?;
+
+    Ok(())
+}
 #[cfg(test)]
 mod tests {
     use super::*;

libs/utils/src/http/error.rs

@@ -1,6 +1,5 @@
 use hyper::{header, Body, Response, StatusCode};
 use serde::{Deserialize, Serialize};
-use std::error::Error as StdError;
 use thiserror::Error;
 use tracing::error;
 
@@ -16,7 +15,7 @@ pub enum ApiError {
     Unauthorized(String),
 
     #[error("NotFound: {0}")]
-    NotFound(Box<dyn StdError + Send + Sync + 'static>),
+    NotFound(anyhow::Error),
 
     #[error("Conflict: {0}")]
     Conflict(String),

pageserver/benches/bench_layer_map.rs

@@ -1,23 +1,22 @@
 use pageserver::keyspace::{KeyPartitioning, KeySpace};
 use pageserver::repository::Key;
 use pageserver::tenant::layer_map::LayerMap;
-use pageserver::tenant::storage_layer::{tests::LayerDescriptor, Layer, LayerFileName};
-use pageserver::tenant::storage_layer::{PersistentLayer, PersistentLayerDesc};
+use pageserver::tenant::storage_layer::{Layer, LayerDescriptor, LayerFileName};
 use rand::prelude::{SeedableRng, SliceRandom, StdRng};
 use std::cmp::{max, min};
 use std::fs::File;
 use std::io::{BufRead, BufReader};
 use std::path::PathBuf;
 use std::str::FromStr;
+use std::sync::Arc;
 use std::time::Instant;
-use utils::id::{TenantId, TimelineId};
 
 use utils::lsn::Lsn;
 
 use criterion::{black_box, criterion_group, criterion_main, Criterion};
 
-fn build_layer_map(filename_dump: PathBuf) -> LayerMap {
-    let mut layer_map = LayerMap::default();
+fn build_layer_map(filename_dump: PathBuf) -> LayerMap<LayerDescriptor> {
+    let mut layer_map = LayerMap::<LayerDescriptor>::default();
 
     let mut min_lsn = Lsn(u64::MAX);
     let mut max_lsn = Lsn(0);
@@ -34,7 +33,7 @@ fn build_layer_map(filename_dump: PathBuf) -> LayerMap {
         min_lsn = min(min_lsn, lsn_range.start);
         max_lsn = max(max_lsn, Lsn(lsn_range.end.0 - 1));
 
-        updates.insert_historic(layer.layer_desc().clone());
+        updates.insert_historic(layer.get_persistent_layer_desc(), Arc::new(layer));
     }
 
     println!("min: {min_lsn}, max: {max_lsn}");
@@ -44,7 +43,7 @@ fn build_layer_map(filename_dump: PathBuf) -> LayerMap {
 }
 
 /// Construct a layer map query pattern for benchmarks
-fn uniform_query_pattern(layer_map: &LayerMap) -> Vec<(Key, Lsn)> {
+fn uniform_query_pattern(layer_map: &LayerMap<LayerDescriptor>) -> Vec<(Key, Lsn)> {
     // For each image layer we query one of the pages contained, at LSN right
     // before the image layer was created. This gives us a somewhat uniform
     // coverage of both the lsn and key space because image layers have
@@ -70,7 +69,7 @@ fn uniform_query_pattern(layer_map: &LayerMap) -> Vec<(Key, Lsn)> {
 
 // Construct a partitioning for testing get_difficulty map when we
 // don't have an exact result of `collect_keyspace` to work with.
-fn uniform_key_partitioning(layer_map: &LayerMap, _lsn: Lsn) -> KeyPartitioning {
+fn uniform_key_partitioning(layer_map: &LayerMap<LayerDescriptor>, _lsn: Lsn) -> KeyPartitioning {
     let mut parts = Vec::new();
 
     // We add a partition boundary at the start of each image layer,
@@ -210,15 +209,13 @@ fn bench_sequential(c: &mut Criterion) {
     for i in 0..100_000 {
         let i32 = (i as u32) % 100;
         let zero = Key::from_hex("000000000000000000000000000000000000").unwrap();
-        let layer = LayerDescriptor::from(PersistentLayerDesc::new_img(
-            TenantId::generate(),
-            TimelineId::generate(),
-            zero.add(10 * i32)..zero.add(10 * i32 + 1),
-            Lsn(i),
-            false,
-            0,
-        ));
-        updates.insert_historic(layer.layer_desc().clone());
+        let layer = LayerDescriptor {
+            key: zero.add(10 * i32)..zero.add(10 * i32 + 1),
+            lsn: Lsn(i)..Lsn(i + 1),
+            is_incremental: false,
+            short_id: format!("Layer {}", i),
+        };
+        updates.insert_historic(layer.get_persistent_layer_desc(), Arc::new(layer));
     }
     updates.flush();
     println!("Finished layer map init in {:?}", now.elapsed());

pageserver/src/bin/pageserver.rs

@@ -495,50 +495,50 @@ fn start_pageserver(
                 Ok(())
             },
         );
-    }
 
-    if let Some(metric_collection_endpoint) = &conf.metric_collection_endpoint {
-        let background_jobs_barrier = background_jobs_barrier;
-        let metrics_ctx = RequestContext::todo_child(
-            TaskKind::MetricsCollection,
-            // This task itself shouldn't download anything.
-            // The actual size calculation does need downloads, and
-            // creates a child context with the right DownloadBehavior.
-            DownloadBehavior::Error,
-        );
-        task_mgr::spawn(
-            crate::BACKGROUND_RUNTIME.handle(),
-            TaskKind::MetricsCollection,
-            None,
-            None,
-            "consumption metrics collection",
-            true,
-            async move {
-                // first wait until background jobs are cleared to launch.
-                //
-                // this is because we only process active tenants and timelines, and the
-                // Timeline::get_current_logical_size will spawn the logical size calculation,
-                // which will not be rate-limited.
-                let cancel = task_mgr::shutdown_token();
+        if let Some(metric_collection_endpoint) = &conf.metric_collection_endpoint {
+            let background_jobs_barrier = background_jobs_barrier;
+            let metrics_ctx = RequestContext::todo_child(
+                TaskKind::MetricsCollection,
+                // This task itself shouldn't download anything.
+                // The actual size calculation does need downloads, and
+                // creates a child context with the right DownloadBehavior.
+                DownloadBehavior::Error,
+            );
+            task_mgr::spawn(
+                MGMT_REQUEST_RUNTIME.handle(),
+                TaskKind::MetricsCollection,
+                None,
+                None,
+                "consumption metrics collection",
+                true,
+                async move {
+                    // first wait until background jobs are cleared to launch.
+                    //
+                    // this is because we only process active tenants and timelines, and the
+                    // Timeline::get_current_logical_size will spawn the logical size calculation,
+                    // which will not be rate-limited.
+                    let cancel = task_mgr::shutdown_token();
 
-                tokio::select! {
-                    _ = cancel.cancelled() => { return Ok(()); },
-                    _ = background_jobs_barrier.wait() => {}
-                };
+                    tokio::select! {
+                        _ = cancel.cancelled() => { return Ok(()); },
+                        _ = background_jobs_barrier.wait() => {}
+                    };
 
-                pageserver::consumption_metrics::collect_metrics(
-                    metric_collection_endpoint,
-                    conf.metric_collection_interval,
-                    conf.cached_metric_collection_interval,
-                    conf.synthetic_size_calculation_interval,
-                    conf.id,
-                    metrics_ctx,
-                )
-                .instrument(info_span!("metrics_collection"))
-                .await?;
-                Ok(())
-            },
-        );
+                    pageserver::consumption_metrics::collect_metrics(
+                        metric_collection_endpoint,
+                        conf.metric_collection_interval,
+                        conf.cached_metric_collection_interval,
+                        conf.synthetic_size_calculation_interval,
+                        conf.id,
+                        metrics_ctx,
+                    )
+                    .instrument(info_span!("metrics_collection"))
+                    .await?;
+                    Ok(())
+                },
+            );
+        }
     }
 
     // Spawn a task to listen for libpq connections. It will spawn further tasks

pageserver/src/config.rs

@@ -96,12 +96,12 @@ pub mod defaults {
 
 #background_task_maximum_delay = '{DEFAULT_BACKGROUND_TASK_MAXIMUM_DELAY}'
 
-[tenant_config]
+# [tenant_config]
 #checkpoint_distance = {DEFAULT_CHECKPOINT_DISTANCE} # in bytes
 #checkpoint_timeout = {DEFAULT_CHECKPOINT_TIMEOUT}
 #compaction_target_size = {DEFAULT_COMPACTION_TARGET_SIZE} # in bytes
 #compaction_period = '{DEFAULT_COMPACTION_PERIOD}'
-#compaction_threshold = {DEFAULT_COMPACTION_THRESHOLD}
+#compaction_threshold = '{DEFAULT_COMPACTION_THRESHOLD}'
 
 #gc_period = '{DEFAULT_GC_PERIOD}'
 #gc_horizon = {DEFAULT_GC_HORIZON}
@@ -111,8 +111,7 @@ pub mod defaults {
 #min_resident_size_override = .. # in bytes
 #evictions_low_residence_duration_metric_threshold = '{DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD}'
 #gc_feedback = false
-
-[remote_storage]
+# [remote_storage]
 
 "###
     );

pageserver/src/disk_usage_eviction_task.rs

@@ -516,7 +516,7 @@ async fn collect_eviction_candidates(
             if !tl.is_active() {
                 continue;
             }
-            let info = tl.get_local_layers_for_disk_usage_eviction().await;
+            let info = tl.get_local_layers_for_disk_usage_eviction();
             debug!(tenant_id=%tl.tenant_id, timeline_id=%tl.timeline_id, "timeline resident layers count: {}", info.resident_layers.len());
             tenant_candidates.extend(
                 info.resident_layers

pageserver/src/http/openapi_spec.yml

@@ -186,8 +186,10 @@ paths:
               schema:
                 $ref: "#/components/schemas/Error"
     delete:
-      description: "Attempts to delete specified timeline. 500 and 409 errors should be retried"
+      description: "Attempts to delete specified timeline. On 500 errors should be retried"
       responses:
+        "200":
+          description: Ok
         "400":
           description: Error when no tenant id found in path or no timeline id
           content:
@@ -212,12 +214,6 @@ paths:
             application/json:
               schema:
                 $ref: "#/components/schemas/NotFoundError"
-        "409":
-          description: Deletion is already in progress, continue polling
-          content:
-            application/json:
-              schema:
-                $ref: "#/components/schemas/ConflictError"
         "412":
           description: Tenant is missing, or timeline has children
           content:

pageserver/src/http/routes.rs

@@ -23,6 +23,7 @@ use super::models::{
     TimelineCreateRequest, TimelineGcRequest, TimelineInfo,
 };
 use crate::context::{DownloadBehavior, RequestContext};
+use crate::disk_usage_eviction_task;
 use crate::metrics::{StorageTimeOperation, STORAGE_TIME_GLOBAL};
 use crate::pgdatadir_mapping::LsnForTimestamp;
 use crate::task_mgr::TaskKind;
@@ -34,7 +35,6 @@ use crate::tenant::size::ModelInputs;
 use crate::tenant::storage_layer::LayerAccessStatsReset;
 use crate::tenant::{LogicalSizeCalculationCause, PageReconstructError, Timeline};
 use crate::{config::PageServerConf, tenant::mgr};
-use crate::{disk_usage_eviction_task, tenant};
 use utils::{
     auth::JwtAuth,
     http::{
@@ -142,7 +142,7 @@ impl From<TenantMapInsertError> for ApiError {
 impl From<TenantStateError> for ApiError {
     fn from(tse: TenantStateError) -> ApiError {
         match tse {
-            TenantStateError::NotFound(tid) => ApiError::NotFound(anyhow!("tenant {}", tid).into()),
+            TenantStateError::NotFound(tid) => ApiError::NotFound(anyhow!("tenant {}", tid)),
             _ => ApiError::InternalServerError(anyhow::Error::new(tse)),
         }
     }
@@ -151,7 +151,7 @@ impl From<TenantStateError> for ApiError {
 impl From<GetTenantError> for ApiError {
     fn from(tse: GetTenantError) -> ApiError {
         match tse {
-            GetTenantError::NotFound(tid) => ApiError::NotFound(anyhow!("tenant {}", tid).into()),
+            GetTenantError::NotFound(tid) => ApiError::NotFound(anyhow!("tenant {}", tid)),
             e @ GetTenantError::NotActive(_) => {
                 // Why is this not `ApiError::NotFound`?
                 // Because we must be careful to never return 404 for a tenant if it does
@@ -169,7 +169,7 @@ impl From<SetNewTenantConfigError> for ApiError {
     fn from(e: SetNewTenantConfigError) -> ApiError {
         match e {
             SetNewTenantConfigError::GetTenant(tid) => {
-                ApiError::NotFound(anyhow!("tenant {}", tid).into())
+                ApiError::NotFound(anyhow!("tenant {}", tid))
             }
             e @ SetNewTenantConfigError::Persist(_) => {
                 ApiError::InternalServerError(anyhow::Error::new(e))
@@ -182,12 +182,11 @@ impl From<crate::tenant::DeleteTimelineError> for ApiError {
     fn from(value: crate::tenant::DeleteTimelineError) -> Self {
         use crate::tenant::DeleteTimelineError::*;
         match value {
-            NotFound => ApiError::NotFound(anyhow::anyhow!("timeline not found").into()),
+            NotFound => ApiError::NotFound(anyhow::anyhow!("timeline not found")),
             HasChildren(children) => ApiError::PreconditionFailed(
                 format!("Cannot delete timeline which has child timelines: {children:?}")
                     .into_boxed_str(),
             ),
-            a @ AlreadyInProgress => ApiError::Conflict(a.to_string()),
             Other(e) => ApiError::InternalServerError(e),
         }
     }
@@ -216,7 +215,7 @@ async fn build_timeline_info(
 ) -> anyhow::Result<TimelineInfo> {
     crate::tenant::debug_assert_current_span_has_tenant_and_timeline_id();
 
-    let mut info = build_timeline_info_common(timeline, ctx).await?;
+    let mut info = build_timeline_info_common(timeline, ctx)?;
     if include_non_incremental_logical_size {
         // XXX we should be using spawn_ondemand_logical_size_calculation here.
         // Otherwise, if someone deletes the timeline / detaches the tenant while
@@ -234,7 +233,7 @@ async fn build_timeline_info(
     Ok(info)
 }
 
-async fn build_timeline_info_common(
+fn build_timeline_info_common(
     timeline: &Arc<Timeline>,
     ctx: &RequestContext,
 ) -> anyhow::Result<TimelineInfo> {
@@ -265,7 +264,7 @@ async fn build_timeline_info_common(
             None
         }
     };
-    let current_physical_size = Some(timeline.layer_size_sum().await);
+    let current_physical_size = Some(timeline.layer_size_sum());
     let state = timeline.current_state();
     let remote_consistent_lsn = timeline.get_remote_consistent_lsn().unwrap_or(Lsn(0));
 
@@ -328,17 +327,14 @@ async fn timeline_create_handler(
             &ctx,
         )
         .await {
-            Ok(new_timeline) => {
+            Ok(Some(new_timeline)) => {
                 // Created. Construct a TimelineInfo for it.
                 let timeline_info = build_timeline_info_common(&new_timeline, &ctx)
-                    .await
                     .map_err(ApiError::InternalServerError)?;
                 json_response(StatusCode::CREATED, timeline_info)
             }
-            Err(tenant::CreateTimelineError::AlreadyExists) => {
-                json_response(StatusCode::CONFLICT, ())
-            }
-            Err(tenant::CreateTimelineError::Other(err)) => Err(ApiError::InternalServerError(err)),
+            Ok(None) => json_response(StatusCode::CONFLICT, ()), // timeline already exists
+            Err(err) => Err(ApiError::InternalServerError(err)),
         }
     }
     .instrument(info_span!("timeline_create", tenant = %tenant_id, timeline_id = %new_timeline_id, lsn=?request_data.ancestor_start_lsn, pg_version=?request_data.pg_version))
@@ -400,7 +396,7 @@ async fn timeline_detail_handler(
 
         let timeline = tenant
             .get_timeline(timeline_id, false)
-            .map_err(|e| ApiError::NotFound(e.into()))?;
+            .map_err(ApiError::NotFound)?;
 
         let timeline_info = build_timeline_info(
             &timeline,
@@ -595,7 +591,7 @@ async fn tenant_status(
         // Calculate total physical size of all timelines
         let mut current_physical_size = 0;
         for timeline in tenant.list_timelines().iter() {
-            current_physical_size += timeline.layer_size_sum().await;
+            current_physical_size += timeline.layer_size_sum();
         }
 
         let state = tenant.current_state();
@@ -705,7 +701,7 @@ async fn layer_map_info_handler(
     check_permission(&request, Some(tenant_id))?;
 
     let timeline = active_timeline_of_active_tenant(tenant_id, timeline_id).await?;
-    let layer_map_info = timeline.layer_map_info(reset).await;
+    let layer_map_info = timeline.layer_map_info(reset);
 
     json_response(StatusCode::OK, layer_map_info)
 }
@@ -1064,7 +1060,7 @@ async fn timeline_download_remote_layers_handler_get(
     let info = timeline
         .get_download_all_remote_layers_task_info()
         .context("task never started since last pageserver process start")
-        .map_err(|e| ApiError::NotFound(e.into()))?;
+        .map_err(ApiError::NotFound)?;
     json_response(StatusCode::OK, info)
 }
 
@@ -1075,7 +1071,7 @@ async fn active_timeline_of_active_tenant(
     let tenant = mgr::get_tenant(tenant_id, true).await?;
     tenant
         .get_timeline(timeline_id, true)
-        .map_err(|e| ApiError::NotFound(e.into()))
+        .map_err(ApiError::NotFound)
 }
 
 async fn always_panic_handler(
@@ -1131,6 +1127,8 @@ async fn disk_usage_eviction_run(
         freed_bytes: 0,
     };
 
+    use crate::task_mgr::MGMT_REQUEST_RUNTIME;
+
     let (tx, rx) = tokio::sync::oneshot::channel();
 
     let state = get_state(&r);
@@ -1148,7 +1146,7 @@ async fn disk_usage_eviction_run(
     let _g = cancel.drop_guard();
 
     crate::task_mgr::spawn(
-        crate::task_mgr::BACKGROUND_RUNTIME.handle(),
+        MGMT_REQUEST_RUNTIME.handle(),
         TaskKind::DiskUsageEviction,
         None,
         None,

pageserver/src/import_datadir.rs

@@ -75,12 +75,12 @@ pub async fn import_timeline_from_postgres_datadir(
             {
                 pg_control = Some(control_file);
             }
-            modification.flush().await?;
+            modification.flush()?;
         }
     }
 
     // We're done importing all the data files.
-    modification.commit().await?;
+    modification.commit()?;
 
     // We expect the Postgres server to be shut down cleanly.
     let pg_control = pg_control.context("pg_control file not found")?;
@@ -148,17 +148,17 @@ async fn import_rel(
     // because there is no guarantee about the order in which we are processing segments.
     // ignore "relation already exists" error
     //
-    // FIXME: Keep track of which relations we've already created?
+    // FIXME: use proper error type for this, instead of parsing the error message.
+    // Or better yet, keep track of which relations we've already created
     // https://github.com/neondatabase/neon/issues/3309
     if let Err(e) = modification
         .put_rel_creation(rel, nblocks as u32, ctx)
         .await
     {
-        match e {
-            RelationError::AlreadyExists => {
-                debug!("Relation {} already exist. We must be extending it.", rel)
-            }
-            _ => return Err(e.into()),
+        if e.to_string().contains("already exists") {
+            debug!("relation {} already exists. we must be extending it", rel);
+        } else {
+            return Err(e);
         }
     }
 
@@ -359,7 +359,7 @@ pub async fn import_basebackup_from_tar(
                     // We found the pg_control file.
                     pg_control = Some(res);
                 }
-                modification.flush().await?;
+                modification.flush()?;
             }
             tokio_tar::EntryType::Directory => {
                 debug!("directory {:?}", file_path);
@@ -377,7 +377,7 @@ pub async fn import_basebackup_from_tar(
     // sanity check: ensure that pg_control is loaded
     let _pg_control = pg_control.context("pg_control file not found")?;
 
-    modification.commit().await?;
+    modification.commit()?;
     Ok(())
 }
 
@@ -594,7 +594,7 @@ async fn import_file(
         // zenith.signal is not necessarily the last file, that we handle
         // but it is ok to call `finish_write()`, because final `modification.commit()`
         // will update lsn once more to the final one.
-        let writer = modification.tline.writer().await;
+        let writer = modification.tline.writer();
         writer.finish_write(prev_lsn);
 
         debug!("imported zenith signal {}", prev_lsn);

pageserver/src/metrics.rs

@@ -1,4 +1,4 @@
-use metrics::metric_vec_duration::DurationResultObserver;
+use metrics::core::{AtomicU64, GenericCounter};
 use metrics::{
     register_counter_vec, register_histogram, register_histogram_vec, register_int_counter,
     register_int_counter_vec, register_int_gauge, register_int_gauge_vec, register_uint_gauge_vec,
@@ -95,19 +95,21 @@ static READ_NUM_FS_LAYERS: Lazy<HistogramVec> = Lazy::new(|| {
 });
 
 // Metrics collected on operations on the storage repository.
-pub static RECONSTRUCT_TIME: Lazy<Histogram> = Lazy::new(|| {
-    register_histogram!(
+static RECONSTRUCT_TIME: Lazy<HistogramVec> = Lazy::new(|| {
+    register_histogram_vec!(
         "pageserver_getpage_reconstruct_seconds",
-        "Time spent in reconstruct_value (reconstruct a page from deltas)",
+        "Time spent in reconstruct_value",
+        &["tenant_id", "timeline_id"],
         CRITICAL_OP_BUCKETS.into(),
     )
     .expect("failed to define a metric")
 });
 
-pub static MATERIALIZED_PAGE_CACHE_HIT_DIRECT: Lazy<IntCounter> = Lazy::new(|| {
-    register_int_counter!(
+static MATERIALIZED_PAGE_CACHE_HIT_DIRECT: Lazy<IntCounterVec> = Lazy::new(|| {
+    register_int_counter_vec!(
         "pageserver_materialized_cache_hits_direct_total",
         "Number of cache hits from materialized page cache without redo",
+        &["tenant_id", "timeline_id"]
     )
     .expect("failed to define a metric")
 });
@@ -122,74 +124,15 @@ static GET_RECONSTRUCT_DATA_TIME: Lazy<HistogramVec> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
-pub static MATERIALIZED_PAGE_CACHE_HIT: Lazy<IntCounter> = Lazy::new(|| {
-    register_int_counter!(
+static MATERIALIZED_PAGE_CACHE_HIT: Lazy<IntCounterVec> = Lazy::new(|| {
+    register_int_counter_vec!(
         "pageserver_materialized_cache_hits_total",
         "Number of cache hits from materialized page cache",
+        &["tenant_id", "timeline_id"]
     )
     .expect("failed to define a metric")
 });
 
-pub static PAGE_CACHE_READ_ACCESSES: Lazy<IntCounterVec> = Lazy::new(|| {
-    register_int_counter_vec!(
-        "pageserver_page_cache_read_accesses_total",
-        "Number of read accesses to the page cache",
-        &["key_kind"]
-    )
-    .expect("failed to define a metric")
-});
-
-pub static PAGE_CACHE_READ_ACCESSES_MATERIALIZED_PAGE: Lazy<IntCounter> = Lazy::new(|| {
-    PAGE_CACHE_READ_ACCESSES
-        .get_metric_with_label_values(&["materialized_page"])
-        .unwrap()
-});
-
-pub static PAGE_CACHE_READ_ACCESSES_EPHEMERAL: Lazy<IntCounter> = Lazy::new(|| {
-    PAGE_CACHE_READ_ACCESSES
-        .get_metric_with_label_values(&["ephemeral"])
-        .unwrap()
-});
-
-pub static PAGE_CACHE_READ_ACCESSES_IMMUTABLE: Lazy<IntCounter> = Lazy::new(|| {
-    PAGE_CACHE_READ_ACCESSES
-        .get_metric_with_label_values(&["immutable"])
-        .unwrap()
-});
-
-pub static PAGE_CACHE_READ_HITS: Lazy<IntCounterVec> = Lazy::new(|| {
-    register_int_counter_vec!(
-        "pageserver_page_cache_read_hits_total",
-        "Number of read accesses to the page cache that hit",
-        &["key_kind", "hit_kind"]
-    )
-    .expect("failed to define a metric")
-});
-
-pub static PAGE_CACHE_READ_HITS_EPHEMERAL: Lazy<IntCounter> = Lazy::new(|| {
-    PAGE_CACHE_READ_HITS
-        .get_metric_with_label_values(&["ephemeral", "-"])
-        .unwrap()
-});
-
-pub static PAGE_CACHE_READ_HITS_IMMUTABLE: Lazy<IntCounter> = Lazy::new(|| {
-    PAGE_CACHE_READ_HITS
-        .get_metric_with_label_values(&["immutable", "-"])
-        .unwrap()
-});
-
-pub static PAGE_CACHE_READ_HITS_MATERIALIZED_PAGE_EXACT: Lazy<IntCounter> = Lazy::new(|| {
-    PAGE_CACHE_READ_HITS
-        .get_metric_with_label_values(&["materialized_page", "exact"])
-        .unwrap()
-});
-
-pub static PAGE_CACHE_READ_HITS_MATERIALIZED_PAGE_OLDER_LSN: Lazy<IntCounter> = Lazy::new(|| {
-    PAGE_CACHE_READ_HITS
-        .get_metric_with_label_values(&["materialized_page", "older_lsn"])
-        .unwrap()
-});
-
 static WAIT_LSN_TIME: Lazy<HistogramVec> = Lazy::new(|| {
     register_histogram_vec!(
         "pageserver_wait_lsn_seconds",
@@ -485,27 +428,6 @@ pub static SMGR_QUERY_TIME: Lazy<HistogramVec> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
-pub struct BasebackupQueryTime(HistogramVec);
-pub static BASEBACKUP_QUERY_TIME: Lazy<BasebackupQueryTime> = Lazy::new(|| {
-    BasebackupQueryTime({
-        register_histogram_vec!(
-            "pageserver_basebackup_query_seconds",
-            "Histogram of basebackup queries durations, by result type",
-            &["result"],
-            CRITICAL_OP_BUCKETS.into(),
-        )
-        .expect("failed to define a metric")
-    })
-});
-
-impl DurationResultObserver for BasebackupQueryTime {
-    fn observe_result<T, E>(&self, res: &Result<T, E>, duration: std::time::Duration) {
-        let label_value = if res.is_ok() { "ok" } else { "error" };
-        let metric = self.0.get_metric_with_label_values(&[label_value]).unwrap();
-        metric.observe(duration.as_secs_f64());
-    }
-}
-
 pub static LIVE_CONNECTIONS_COUNT: Lazy<IntGaugeVec> = Lazy::new(|| {
     register_int_gauge_vec!(
         "pageserver_live_connections",
@@ -674,79 +596,6 @@ pub static WALRECEIVER_CANDIDATES_ADDED: Lazy<IntCounter> =
 pub static WALRECEIVER_CANDIDATES_REMOVED: Lazy<IntCounter> =
     Lazy::new(|| WALRECEIVER_CANDIDATES_EVENTS.with_label_values(&["remove"]));
 
-pub static LAYER_GET_VALUE_RECONSTRUCT_DATA_SPAWN_BLOCKING_STARTED_COUNT: Lazy<IntCounter> =
-    Lazy::new(|| {
-        register_int_counter!(
-            "pageserver_layer_get_value_reconstruct_data_spawn_blocking_started_count",
-            "Number of spawn_blocking calls made in Layer::get_value_reconstruct_data"
-        )
-        .expect("failed to define a metric")
-    });
-
-pub static LAYER_GET_VALUE_RECONSTRUCT_DATA_SPAWN_BLOCKING_ACTIVE_GAUGE: Lazy<IntGauge> =
-    Lazy::new(|| {
-        register_int_gauge!(
-            "pageserver_layer_get_value_reconstruct_data_spawn_blocking_active_gauge",
-            "Number of spawn_blocking calls active in Layer::get_value_reconstruct_data"
-        )
-        .expect("failed to define a metric")
-    });
-
-pub static LAYER_GET_VALUE_RECONSTRUCT_DATA_SPAWN_BLOCKING_QUEUE_DELAY: Lazy<Histogram> = Lazy::new(
-    || {
-        register_histogram!(
-            "pageserver_layer_get_value_reconstruct_data_spawn_blocking_queue_delay_seconds",
-            "Time a Layer::get_value_reconstruct_data call spends in spawn_blocking queue until the first line of blockign code runs inside spawn_blocking",
-            vec![
-            0.000_005, 0.000_010, 0.000_025, 0.000_050, 0.000_100, 0.000_250, 0.000_500, 0.001_000,
-            0.002_500, 0.005_000, 0.010_000, 0.025_000, 0.050_000, 0.100_000, 0.250_000, 0.500_000,
-            1.000_000, 2.000_000, 5.000_000, 10.000_000, 25.000_000, 50.000_000, 100.000_000,
-            ],
-        )
-        .expect("failed to define a metric")
-    },
-);
-
-pub static LAYER_GET_VALUE_RECONSTRUCT_DATA_COMPLETION_TIME: Lazy<HistogramVec> = Lazy::new(|| {
-    register_histogram_vec!(
-        "pageserver_layer_get_value_reconstruct_data_completion_time_seconds",
-        "Time a Layer::get_value_reconstruct_data call takes to complete",
-        &["result"],
-        vec![
-            0.000_005,
-            0.000_010,
-            0.000_025,
-            0.000_050,
-            0.000_100,
-            0.000_250,
-            0.000_500,
-            0.001_000,
-            0.002_500,
-            0.005_000,
-            0.010_000,
-            0.025_000,
-            0.050_000,
-            0.100_000,
-            0.250_000,
-            0.500_000,
-            1.000_000,
-            2.000_000,
-            5.000_000,
-            10.000_000,
-            25.000_000,
-            50.000_000,
-            100.000_000,
-        ]
-    )
-    .expect("failed to define a metric")
-});
-
-pub static LAYER_GET_VALUE_RECONSTRUCT_DATA_COMPLETION_TIME_OK: Lazy<Histogram> =
-    Lazy::new(|| LAYER_GET_VALUE_RECONSTRUCT_DATA_COMPLETION_TIME.with_label_values(&["ok"]));
-
-pub static LAYER_GET_VALUE_RECONSTRUCT_DATA_COMPLETION_TIME_ERROR: Lazy<Histogram> =
-    Lazy::new(|| LAYER_GET_VALUE_RECONSTRUCT_DATA_COMPLETION_TIME.with_label_values(&["error"]));
-
 // Metrics collected on WAL redo operations
 //
 // We collect the time spent in actual WAL redo ('redo'), and time waiting
@@ -903,7 +752,10 @@ impl StorageTimeMetrics {
 pub struct TimelineMetrics {
     tenant_id: String,
     timeline_id: String,
+    pub reconstruct_time_histo: Histogram,
     pub get_reconstruct_data_time_histo: Histogram,
+    pub materialized_page_cache_hit_counter: GenericCounter<AtomicU64>,
+    pub materialized_page_cache_hit_upon_request_counter: GenericCounter<AtomicU64>,
     pub flush_time_histo: StorageTimeMetrics,
     pub compact_time_histo: StorageTimeMetrics,
     pub create_images_time_histo: StorageTimeMetrics,
@@ -931,9 +783,15 @@ impl TimelineMetrics {
     ) -> Self {
         let tenant_id = tenant_id.to_string();
         let timeline_id = timeline_id.to_string();
+        let reconstruct_time_histo = RECONSTRUCT_TIME
+            .get_metric_with_label_values(&[&tenant_id, &timeline_id])
+            .unwrap();
         let get_reconstruct_data_time_histo = GET_RECONSTRUCT_DATA_TIME
             .get_metric_with_label_values(&[&tenant_id, &timeline_id])
             .unwrap();
+        let materialized_page_cache_hit_counter = MATERIALIZED_PAGE_CACHE_HIT
+            .get_metric_with_label_values(&[&tenant_id, &timeline_id])
+            .unwrap();
         let flush_time_histo =
             StorageTimeMetrics::new(StorageTimeOperation::LayerFlush, &tenant_id, &timeline_id);
         let compact_time_histo =
@@ -975,13 +833,19 @@ impl TimelineMetrics {
         let read_num_fs_layers = READ_NUM_FS_LAYERS
             .get_metric_with_label_values(&[&tenant_id, &timeline_id])
             .unwrap();
+        let materialized_page_cache_hit_upon_request_counter = MATERIALIZED_PAGE_CACHE_HIT_DIRECT
+            .get_metric_with_label_values(&[&tenant_id, &timeline_id])
+            .unwrap();
         let evictions_with_low_residence_duration =
             evictions_with_low_residence_duration_builder.build(&tenant_id, &timeline_id);
 
         TimelineMetrics {
             tenant_id,
             timeline_id,
+            reconstruct_time_histo,
             get_reconstruct_data_time_histo,
+            materialized_page_cache_hit_counter,
+            materialized_page_cache_hit_upon_request_counter,
             flush_time_histo,
             compact_time_histo,
             create_images_time_histo,
@@ -1008,7 +872,10 @@ impl Drop for TimelineMetrics {
     fn drop(&mut self) {
         let tenant_id = &self.tenant_id;
         let timeline_id = &self.timeline_id;
+        let _ = RECONSTRUCT_TIME.remove_label_values(&[tenant_id, timeline_id]);
         let _ = GET_RECONSTRUCT_DATA_TIME.remove_label_values(&[tenant_id, timeline_id]);
+        let _ = MATERIALIZED_PAGE_CACHE_HIT.remove_label_values(&[tenant_id, timeline_id]);
+        let _ = MATERIALIZED_PAGE_CACHE_HIT_DIRECT.remove_label_values(&[tenant_id, timeline_id]);
         let _ = LAST_RECORD_LSN.remove_label_values(&[tenant_id, timeline_id]);
         let _ = WAIT_LSN_TIME.remove_label_values(&[tenant_id, timeline_id]);
         let _ = RESIDENT_PHYSICAL_SIZE.remove_label_values(&[tenant_id, timeline_id]);
@@ -1452,8 +1319,4 @@ pub fn preinitialize_metrics() {
 
     // Same as above for this metric, but, it's a Vec-type metric for which we don't know all the labels.
     BACKGROUND_LOOP_PERIOD_OVERRUN_COUNT.reset();
-
-    // Python tests need these.
-    MATERIALIZED_PAGE_CACHE_HIT_DIRECT.get();
-    MATERIALIZED_PAGE_CACHE_HIT.get();
 }

pageserver/src/page_cache.rs

@@ -313,8 +313,6 @@ impl PageCache {
         key: &Key,
         lsn: Lsn,
     ) -> Option<(Lsn, PageReadGuard)> {
-        crate::metrics::PAGE_CACHE_READ_ACCESSES_MATERIALIZED_PAGE.inc();
-
         let mut cache_key = CacheKey::MaterializedPage {
             hash_key: MaterializedPageHashKey {
                 tenant_id,
@@ -325,17 +323,8 @@ impl PageCache {
         };
 
         if let Some(guard) = self.try_lock_for_read(&mut cache_key) {
-            if let CacheKey::MaterializedPage {
-                hash_key: _,
-                lsn: available_lsn,
-            } = cache_key
-            {
-                if available_lsn == lsn {
-                    crate::metrics::PAGE_CACHE_READ_HITS_MATERIALIZED_PAGE_EXACT.inc();
-                } else {
-                    crate::metrics::PAGE_CACHE_READ_HITS_MATERIALIZED_PAGE_OLDER_LSN.inc();
-                }
-                Some((available_lsn, guard))
+            if let CacheKey::MaterializedPage { hash_key: _, lsn } = cache_key {
+                Some((lsn, guard))
             } else {
                 panic!("unexpected key type in slot");
             }
@@ -510,31 +499,11 @@ impl PageCache {
     /// ```
     ///
     fn lock_for_read(&self, cache_key: &mut CacheKey) -> anyhow::Result<ReadBufResult> {
-        let (read_access, hit) = match cache_key {
-            CacheKey::MaterializedPage { .. } => {
-                unreachable!("Materialized pages use lookup_materialized_page")
-            }
-            CacheKey::EphemeralPage { .. } => (
-                &crate::metrics::PAGE_CACHE_READ_ACCESSES_EPHEMERAL,
-                &crate::metrics::PAGE_CACHE_READ_HITS_EPHEMERAL,
-            ),
-            CacheKey::ImmutableFilePage { .. } => (
-                &crate::metrics::PAGE_CACHE_READ_ACCESSES_IMMUTABLE,
-                &crate::metrics::PAGE_CACHE_READ_HITS_IMMUTABLE,
-            ),
-        };
-        read_access.inc();
-
-        let mut is_first_iteration = true;
         loop {
             // First check if the key already exists in the cache.
             if let Some(read_guard) = self.try_lock_for_read(cache_key) {
-                if is_first_iteration {
-                    hit.inc();
-                }
                 return Ok(ReadBufResult::Found(read_guard));
             }
-            is_first_iteration = false;
 
             // Not found. Find a victim buffer
             let (slot_idx, mut inner) =
@@ -830,8 +799,12 @@ impl PageCache {
                             // a different victim. But if the problem persists, the page cache
                             // could fill up with dirty pages that we cannot evict, and we will
                             // loop retrying the writebacks indefinitely.
-                            error!("writeback of buffer {:?} failed: {}", old_key, err);
-                            continue;
+                            if cfg!(test) {
+                                anyhow::bail!("writeback of buffer {:?} failed: {}", old_key, err);
+                            } else {
+                                error!("writeback of buffer {:?} failed: {}", old_key, err);
+                                continue;
+                            }
                         }
                     }
 

pageserver/src/page_service.rs

@@ -390,9 +390,7 @@ impl PageServerHandler {
         };
 
         // Check that the timeline exists
-        let timeline = tenant
-            .get_timeline(timeline_id, true)
-            .map_err(|e| anyhow::anyhow!(e))?;
+        let timeline = tenant.get_timeline(timeline_id, true)?;
 
         // switch client to COPYBOTH
         pgb.write_message_noflush(&BeMessage::CopyBothResponse)?;
@@ -904,7 +902,7 @@ where
 
             self.check_permission(Some(tenant_id))?;
 
-            let lsn = if params.len() >= 3 {
+            let lsn = if params.len() == 3 {
                 Some(
                     Lsn::from_str(params[2])
                         .with_context(|| format!("Failed to parse Lsn from {}", params[2]))?,
@@ -913,24 +911,10 @@ where
                 None
             };
 
-            metrics::metric_vec_duration::observe_async_block_duration_by_result(
-                &*crate::metrics::BASEBACKUP_QUERY_TIME,
-                async move {
-                    self.handle_basebackup_request(
-                        pgb,
-                        tenant_id,
-                        timeline_id,
-                        lsn,
-                        None,
-                        false,
-                        ctx,
-                    )
-                    .await?;
-                    pgb.write_message_noflush(&BeMessage::CommandComplete(b"SELECT 1"))?;
-                    anyhow::Ok(())
-                },
-            )
-            .await?;
+            // Check that the timeline exists
+            self.handle_basebackup_request(pgb, tenant_id, timeline_id, lsn, None, false, ctx)
+                .await?;
+            pgb.write_message_noflush(&BeMessage::CommandComplete(b"SELECT 1"))?;
         }
         // return pair of prev_lsn and last_lsn
         else if query_string.starts_with("get_last_record_rlsn ") {
@@ -1246,6 +1230,6 @@ async fn get_active_tenant_timeline(
         .map_err(GetActiveTimelineError::Tenant)?;
     let timeline = tenant
         .get_timeline(timeline_id, true)
-        .map_err(|e| GetActiveTimelineError::Timeline(anyhow::anyhow!(e)))?;
+        .map_err(GetActiveTimelineError::Timeline)?;
     Ok(timeline)
 }

pageserver/src/pgdatadir_mapping.rs

@@ -43,16 +43,6 @@ pub enum CalculateLogicalSizeError {
     Other(#[from] anyhow::Error),
 }
 
-#[derive(Debug, thiserror::Error)]
-pub enum RelationError {
-    #[error("Relation Already Exists")]
-    AlreadyExists,
-    #[error("invalid relnode")]
-    InvalidRelnode,
-    #[error(transparent)]
-    Other(#[from] anyhow::Error),
-}
-
 ///
 /// This impl provides all the functionality to store PostgreSQL relations, SLRUs,
 /// and other special kinds of files, in a versioned key-value store. The
@@ -111,9 +101,9 @@ impl Timeline {
         ctx: &RequestContext,
     ) -> Result<Bytes, PageReconstructError> {
         if tag.relnode == 0 {
-            return Err(PageReconstructError::Other(
-                RelationError::InvalidRelnode.into(),
-            ));
+            return Err(PageReconstructError::Other(anyhow::anyhow!(
+                "invalid relnode"
+            )));
         }
 
         let nblocks = self.get_rel_size(tag, lsn, latest, ctx).await?;
@@ -158,9 +148,9 @@ impl Timeline {
         ctx: &RequestContext,
     ) -> Result<BlockNumber, PageReconstructError> {
         if tag.relnode == 0 {
-            return Err(PageReconstructError::Other(
-                RelationError::InvalidRelnode.into(),
-            ));
+            return Err(PageReconstructError::Other(anyhow::anyhow!(
+                "invalid relnode"
+            )));
         }
 
         if let Some(nblocks) = self.get_cached_rel_size(&tag, lsn) {
@@ -203,9 +193,9 @@ impl Timeline {
         ctx: &RequestContext,
     ) -> Result<bool, PageReconstructError> {
         if tag.relnode == 0 {
-            return Err(PageReconstructError::Other(
-                RelationError::InvalidRelnode.into(),
-            ));
+            return Err(PageReconstructError::Other(anyhow::anyhow!(
+                "invalid relnode"
+            )));
         }
 
         // first try to lookup relation in cache
@@ -709,20 +699,6 @@ impl<'a> DatadirModification<'a> {
         Ok(())
     }
 
-    #[cfg(test)]
-    pub fn init_empty_test_timeline(&mut self) -> anyhow::Result<()> {
-        self.init_empty()?;
-        self.put_control_file(bytes::Bytes::from_static(
-            b"control_file contents do not matter",
-        ))
-        .context("put_control_file")?;
-        self.put_checkpoint(bytes::Bytes::from_static(
-            b"checkpoint_file contents do not matter",
-        ))
-        .context("put_checkpoint_file")?;
-        Ok(())
-    }
-
     /// Put a new page version that can be constructed from a WAL record
     ///
     /// NOTE: this will *not* implicitly extend the relation, if the page is beyond the
@@ -734,7 +710,7 @@ impl<'a> DatadirModification<'a> {
         blknum: BlockNumber,
         rec: NeonWalRecord,
     ) -> anyhow::Result<()> {
-        anyhow::ensure!(rel.relnode != 0, RelationError::InvalidRelnode);
+        anyhow::ensure!(rel.relnode != 0, "invalid relnode");
         self.put(rel_block_to_key(rel, blknum), Value::WalRecord(rec));
         Ok(())
     }
@@ -761,7 +737,7 @@ impl<'a> DatadirModification<'a> {
         blknum: BlockNumber,
         img: Bytes,
     ) -> anyhow::Result<()> {
-        anyhow::ensure!(rel.relnode != 0, RelationError::InvalidRelnode);
+        anyhow::ensure!(rel.relnode != 0, "invalid relnode");
         self.put(rel_block_to_key(rel, blknum), Value::Image(img));
         Ok(())
     }
@@ -885,38 +861,32 @@ impl<'a> DatadirModification<'a> {
         rel: RelTag,
         nblocks: BlockNumber,
         ctx: &RequestContext,
-    ) -> Result<(), RelationError> {
-        if rel.relnode == 0 {
-            return Err(RelationError::InvalidRelnode);
-        }
+    ) -> anyhow::Result<()> {
+        anyhow::ensure!(rel.relnode != 0, "invalid relnode");
         // It's possible that this is the first rel for this db in this
         // tablespace.  Create the reldir entry for it if so.
-        let mut dbdir = DbDirectory::des(&self.get(DBDIR_KEY, ctx).await.context("read db")?)
-            .context("deserialize db")?;
+        let mut dbdir = DbDirectory::des(&self.get(DBDIR_KEY, ctx).await?)?;
         let rel_dir_key = rel_dir_to_key(rel.spcnode, rel.dbnode);
         let mut rel_dir = if dbdir.dbdirs.get(&(rel.spcnode, rel.dbnode)).is_none() {
             // Didn't exist. Update dbdir
             dbdir.dbdirs.insert((rel.spcnode, rel.dbnode), false);
-            let buf = DbDirectory::ser(&dbdir).context("serialize db")?;
+            let buf = DbDirectory::ser(&dbdir)?;
             self.put(DBDIR_KEY, Value::Image(buf.into()));
 
             // and create the RelDirectory
             RelDirectory::default()
         } else {
             // reldir already exists, fetch it
-            RelDirectory::des(&self.get(rel_dir_key, ctx).await.context("read db")?)
-                .context("deserialize db")?
+            RelDirectory::des(&self.get(rel_dir_key, ctx).await?)?
         };
 
         // Add the new relation to the rel directory entry, and write it back
         if !rel_dir.rels.insert((rel.relnode, rel.forknum)) {
-            return Err(RelationError::AlreadyExists);
+            anyhow::bail!("rel {rel} already exists");
         }
         self.put(
             rel_dir_key,
-            Value::Image(Bytes::from(
-                RelDirectory::ser(&rel_dir).context("serialize")?,
-            )),
+            Value::Image(Bytes::from(RelDirectory::ser(&rel_dir)?)),
         );
 
         // Put size
@@ -941,7 +911,7 @@ impl<'a> DatadirModification<'a> {
         nblocks: BlockNumber,
         ctx: &RequestContext,
     ) -> anyhow::Result<()> {
-        anyhow::ensure!(rel.relnode != 0, RelationError::InvalidRelnode);
+        anyhow::ensure!(rel.relnode != 0, "invalid relnode");
         let last_lsn = self.tline.get_last_record_lsn();
         if self.tline.get_rel_exists(rel, last_lsn, true, ctx).await? {
             let size_key = rel_size_to_key(rel);
@@ -972,7 +942,7 @@ impl<'a> DatadirModification<'a> {
         nblocks: BlockNumber,
         ctx: &RequestContext,
     ) -> anyhow::Result<()> {
-        anyhow::ensure!(rel.relnode != 0, RelationError::InvalidRelnode);
+        anyhow::ensure!(rel.relnode != 0, "invalid relnode");
 
         // Put size
         let size_key = rel_size_to_key(rel);
@@ -993,7 +963,7 @@ impl<'a> DatadirModification<'a> {
 
     /// Drop a relation.
     pub async fn put_rel_drop(&mut self, rel: RelTag, ctx: &RequestContext) -> anyhow::Result<()> {
-        anyhow::ensure!(rel.relnode != 0, RelationError::InvalidRelnode);
+        anyhow::ensure!(rel.relnode != 0, "invalid relnode");
 
         // Remove it from the directory entry
         let dir_key = rel_dir_to_key(rel.spcnode, rel.dbnode);
@@ -1138,7 +1108,7 @@ impl<'a> DatadirModification<'a> {
     /// retains all the metadata, but data pages are flushed. That's again OK
     /// for bulk import, where you are just loading data pages and won't try to
     /// modify the same pages twice.
-    pub async fn flush(&mut self) -> anyhow::Result<()> {
+    pub fn flush(&mut self) -> anyhow::Result<()> {
         // Unless we have accumulated a decent amount of changes, it's not worth it
         // to scan through the pending_updates list.
         let pending_nblocks = self.pending_nblocks;
@@ -1146,20 +1116,19 @@ impl<'a> DatadirModification<'a> {
             return Ok(());
         }
 
-        let writer = self.tline.writer().await;
+        let writer = self.tline.writer();
 
         // Flush relation and  SLRU data blocks, keep metadata.
-        let mut retained_pending_updates = HashMap::new();
-        for (key, value) in self.pending_updates.drain() {
-            if is_rel_block_key(key) || is_slru_block_key(key) {
-                // This bails out on first error without modifying pending_updates.
-                // That's Ok, cf this function's doc comment.
-                writer.put(key, self.lsn, &value).await?;
+        let mut result: anyhow::Result<()> = Ok(());
+        self.pending_updates.retain(|&key, value| {
+            if result.is_ok() && (is_rel_block_key(key) || is_slru_block_key(key)) {
+                result = writer.put(key, self.lsn, value);
+                false
             } else {
-                retained_pending_updates.insert(key, value);
+                true
             }
-        }
-        self.pending_updates.extend(retained_pending_updates);
+        });
+        result?;
 
         if pending_nblocks != 0 {
             writer.update_current_logical_size(pending_nblocks * i64::from(BLCKSZ));
@@ -1174,17 +1143,17 @@ impl<'a> DatadirModification<'a> {
     /// underlying timeline.
     /// All the modifications in this atomic update are stamped by the specified LSN.
     ///
-    pub async fn commit(&mut self) -> anyhow::Result<()> {
-        let writer = self.tline.writer().await;
+    pub fn commit(&mut self) -> anyhow::Result<()> {
+        let writer = self.tline.writer();
         let lsn = self.lsn;
         let pending_nblocks = self.pending_nblocks;
         self.pending_nblocks = 0;
 
         for (key, value) in self.pending_updates.drain() {
-            writer.put(key, lsn, &value).await?;
+            writer.put(key, lsn, &value)?;
         }
         for key_range in self.pending_deletions.drain(..) {
-            writer.delete(key_range, lsn).await?;
+            writer.delete(key_range, lsn)?;
         }
 
         writer.finish_write(lsn);
@@ -1624,6 +1593,17 @@ fn is_slru_block_key(key: Key) -> bool {
         && key.field6 != 0xffffffff // and not SlruSegSize
 }
 
+#[cfg(test)]
+pub fn create_test_timeline(
+    tenant: &crate::tenant::Tenant,
+    timeline_id: utils::id::TimelineId,
+    pg_version: u32,
+    ctx: &RequestContext,
+) -> anyhow::Result<std::sync::Arc<Timeline>> {
+    let tline = tenant.create_test_timeline(timeline_id, Lsn(8), pg_version, ctx)?;
+    Ok(tline)
+}
+
 #[allow(clippy::bool_assert_comparison)]
 #[cfg(test)]
 mod tests {

pageserver/src/task_mgr.rs

@@ -102,33 +102,11 @@ use crate::shutdown_pageserver;
 // It's also good to avoid hogging all threads that would be needed to process
 // other operations, if the upload tasks e.g. get blocked on locks. It shouldn't
 // happen, but still.
-
-static PAGESERVER_TOKIO_MAX_BLOCKING_THREADS_OVERRIDE: Lazy<Option<usize>> = Lazy::new(|| {
-    let env_var: String = match std::env::var("PAGESERVER_TOKIO_MAX_BLOCKING_THREADS") {
-        Ok(v) => v,
-        Err(std::env::VarError::NotPresent) => {
-            debug!("env var PAGESERVER_TOKIO_MAX_BLOCKING_THREADS not set, using default");
-            return None;
-        }
-        Err(std::env::VarError::NotUnicode(_)) => {
-            panic!("env var PAGESERVER_TOKIO_MAX_BLOCKING_THREADS is not valid UTF-8");
-        }
-    };
-    let pool_size = match env_var.parse() {
-        Ok(v) => v,
-        Err(e) => {
-            panic!("Failed to parse PAGESERVER_TOKIO_MAX_BLOCKING_THREADS: {e:?}");
-        }
-    };
-    eprintln!("using spawn_blocking pool size override from env var PAGESERVER_TOKIO_MAX_BLOCKING_THREADS: {pool_size:?}");
-    Some(pool_size)
-});
-
+//
 pub static COMPUTE_REQUEST_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
     tokio::runtime::Builder::new_multi_thread()
         .thread_name("compute request worker")
         .enable_all()
-        .max_blocking_threads((*PAGESERVER_TOKIO_MAX_BLOCKING_THREADS_OVERRIDE).unwrap_or(512))
         .build()
         .expect("Failed to create compute request runtime")
 });
@@ -137,7 +115,6 @@ pub static MGMT_REQUEST_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
     tokio::runtime::Builder::new_multi_thread()
         .thread_name("mgmt request worker")
         .enable_all()
-        .max_blocking_threads((*PAGESERVER_TOKIO_MAX_BLOCKING_THREADS_OVERRIDE).unwrap_or(512))
         .build()
         .expect("Failed to create mgmt request runtime")
 });
@@ -146,7 +123,6 @@ pub static WALRECEIVER_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
     tokio::runtime::Builder::new_multi_thread()
         .thread_name("walreceiver worker")
         .enable_all()
-        .max_blocking_threads((*PAGESERVER_TOKIO_MAX_BLOCKING_THREADS_OVERRIDE).unwrap_or(512))
         .build()
         .expect("Failed to create walreceiver runtime")
 });
@@ -155,7 +131,6 @@ pub static BACKGROUND_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
     tokio::runtime::Builder::new_multi_thread()
         .thread_name("background op worker")
         .enable_all()
-        .max_blocking_threads((*PAGESERVER_TOKIO_MAX_BLOCKING_THREADS_OVERRIDE).unwrap_or(512))
         .build()
         .expect("Failed to create background op runtime")
 });
@@ -531,17 +506,17 @@ pub async fn shutdown_tasks(
                     warn!(name = task.name, tenant_id = ?tenant_id, timeline_id = ?timeline_id, kind = ?task_kind, "stopping left-over");
                 }
             }
-            let join_handle = tokio::select! {
+            let completed = tokio::select! {
                 biased;
-                _ = &mut join_handle => { None },
+                _ = &mut join_handle => { true },
                 _ = tokio::time::sleep(std::time::Duration::from_secs(1)) => {
                     // allow some time to elapse before logging to cut down the number of log
                     // lines.
                     info!("waiting for {} to shut down", task.name);
-                    Some(join_handle)
+                    false
                 }
             };
-            if let Some(join_handle) = join_handle {
+            if !completed {
                 // we never handled this return value, but:
                 // - we don't deschedule which would lead to is_cancelled
                 // - panics are already logged (is_panicked)

pageserver/src/tenant/config.rs

@@ -38,8 +38,8 @@ pub mod defaults {
     pub const DEFAULT_GC_PERIOD: &str = "1 hr";
     pub const DEFAULT_IMAGE_CREATION_THRESHOLD: usize = 3;
     pub const DEFAULT_PITR_INTERVAL: &str = "7 days";
-    pub const DEFAULT_WALRECEIVER_CONNECT_TIMEOUT: &str = "10 seconds";
-    pub const DEFAULT_WALRECEIVER_LAGGING_WAL_TIMEOUT: &str = "10 seconds";
+    pub const DEFAULT_WALRECEIVER_CONNECT_TIMEOUT: &str = "2 seconds";
+    pub const DEFAULT_WALRECEIVER_LAGGING_WAL_TIMEOUT: &str = "3 seconds";
     pub const DEFAULT_MAX_WALRECEIVER_LSN_WAL_LAG: u64 = 10 * 1024 * 1024;
     pub const DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD: &str = "24 hour";
 }

pageserver/src/tenant/ephemeral_file.rs

@@ -39,7 +39,7 @@ pub struct EphemeralFile {
     file_id: u64,
     _tenant_id: TenantId,
     _timeline_id: TimelineId,
-    file: Arc<VirtualFile>,
+    file: Option<Arc<VirtualFile>>,
 
     pub size: u64,
 }
@@ -52,7 +52,10 @@ impl EphemeralFile {
     ) -> Result<EphemeralFile, io::Error> {
         let mut l = EPHEMERAL_FILES.write().unwrap();
         let file_id = l.next_file_id;
-        l.next_file_id += 1;
+        l.next_file_id = l
+            .next_file_id
+            .checked_add(1)
+            .expect("next_file_id is u64, expecting it to not overflow");
 
         let filename = conf
             .timeline_path(&timeline_id, &tenant_id)
@@ -60,16 +63,30 @@ impl EphemeralFile {
 
         let file = VirtualFile::open_with_options(
             &filename,
-            OpenOptions::new().read(true).write(true).create(true),
+            OpenOptions::new()
+                .read(true)
+                .write(true)
+                // The next_file_id doesn't overlfow, so technically, `create_new` is not needed.
+                // But it's cheap, so why not.
+                .create_new(true),
         )?;
         let file_rc = Arc::new(file);
         l.files.insert(file_id, file_rc.clone());
 
+        #[cfg(debug_assertions)]
+        debug!(
+            "created ephemeral file {}\n{}",
+            filename.display(),
+            std::backtrace::Backtrace::force_capture()
+        );
+        #[cfg(not(debug_assertions))]
+        debug!("created ephemeral file {}", filename.display());
+
         Ok(EphemeralFile {
             file_id,
             _tenant_id: tenant_id,
             _timeline_id: timeline_id,
-            file: file_rc,
+            file: Some(file_rc),
             size: 0,
         })
     }
@@ -79,6 +96,8 @@ impl EphemeralFile {
         while off < PAGE_SZ {
             let n = self
                 .file
+                .as_ref()
+                .unwrap()
                 .read_at(&mut buf[off..], blkno as u64 * PAGE_SZ as u64 + off as u64)?;
 
             if n == 0 {
@@ -261,17 +280,43 @@ impl Drop for EphemeralFile {
         cache.drop_buffers_for_ephemeral(self.file_id);
 
         // remove entry from the hash map
-        EPHEMERAL_FILES.write().unwrap().files.remove(&self.file_id);
+        let virtual_file = EPHEMERAL_FILES
+            .write()
+            .unwrap()
+            .files
+            .remove(&self.file_id)
+            .unwrap();
+
+        // remove file from self
+        let self_file = self.file.take().unwrap();
+
+        assert_eq!(
+            Arc::as_ptr(&virtual_file) as *const (),
+            Arc::as_ptr(&self_file) as *const ()
+        );
+        drop(self_file);
+
+        // XXX once we upgrade to Rust 1.70, use Arc::into_inner.
+        // It does the following checks atomically.
+        assert_eq!(Arc::weak_count(&virtual_file), 0);
+        let virtual_file = Arc::try_unwrap(virtual_file).expect(
+            "we are being dropped and EPHEMERAL_FILES is the only other place where we put the Arc",
+        );
 
         // unlink the file
-        let res = std::fs::remove_file(&self.file.path);
-        if let Err(e) = res {
-            warn!(
-                "could not remove ephemeral file '{}': {}",
-                self.file.path.display(),
-                e
-            );
-        }
+        // TODO: we should be able to unwrap here, but, timeline delete and tenant detach do
+        //       std::fs::remove_dir_all without dropping all InMemoryLayer => EphemeralFile
+        //       of the tenant => need to fix that first.
+        match virtual_file.remove() {
+            Ok(()) => (),
+            Err((virtual_file, e)) => {
+                warn!(
+                    "could not remove ephemeral file '{}': {}",
+                    virtual_file.path.display(),
+                    e
+                );
+            }
+        };
     }
 }
 

pageserver/src/tenant/layer_map.rs

@@ -51,23 +51,25 @@ use crate::keyspace::KeyPartitioning;
 use crate::repository::Key;
 use crate::tenant::storage_layer::InMemoryLayer;
 use crate::tenant::storage_layer::Layer;
+use anyhow::Context;
 use anyhow::Result;
+use std::collections::HashMap;
 use std::collections::VecDeque;
 use std::ops::Range;
 use std::sync::Arc;
 use utils::lsn::Lsn;
 
 use historic_layer_coverage::BufferedHistoricLayerCoverage;
-pub use historic_layer_coverage::LayerKey;
+pub use historic_layer_coverage::Replacement;
 
 use super::storage_layer::range_eq;
 use super::storage_layer::PersistentLayerDesc;
+use super::storage_layer::PersistentLayerKey;
 
 ///
 /// LayerMap tracks what layers exist on a timeline.
 ///
-#[derive(Default)]
-pub struct LayerMap {
+pub struct LayerMap<L: ?Sized> {
     //
     // 'open_layer' holds the current InMemoryLayer that is accepting new
     // records. If it is None, 'next_open_layer_at' will be set instead, indicating
@@ -93,6 +95,24 @@ pub struct LayerMap {
     /// L0 layers have key range Key::MIN..Key::MAX, and locating them using R-Tree search is very inefficient.
     /// So L0 layers are held in l0_delta_layers vector, in addition to the R-tree.
     l0_delta_layers: Vec<Arc<PersistentLayerDesc>>,
+
+    /// Mapping from persistent layer key to the actual layer object. Currently, it stores delta, image, and
+    /// remote layers. In future refactors, this will be eventually moved out of LayerMap into Timeline, and
+    /// RemoteLayer will be removed.
+    mapping: HashMap<PersistentLayerKey, Arc<L>>,
+}
+
+impl<L: ?Sized> Default for LayerMap<L> {
+    fn default() -> Self {
+        Self {
+            open_layer: None,
+            next_open_layer_at: None,
+            frozen_layers: VecDeque::default(),
+            l0_delta_layers: Vec::default(),
+            historic: BufferedHistoricLayerCoverage::default(),
+            mapping: HashMap::default(),
+        }
+    }
 }
 
 /// The primary update API for the layer map.
@@ -100,21 +120,24 @@ pub struct LayerMap {
 /// Batching historic layer insertions and removals is good for
 /// performance and this struct helps us do that correctly.
 #[must_use]
-pub struct BatchedUpdates<'a> {
+pub struct BatchedUpdates<'a, L: ?Sized + Layer> {
     // While we hold this exclusive reference to the layer map the type checker
     // will prevent us from accidentally reading any unflushed updates.
-    layer_map: &'a mut LayerMap,
+    layer_map: &'a mut LayerMap<L>,
 }
 
 /// Provide ability to batch more updates while hiding the read
 /// API so we don't accidentally read without flushing.
-impl BatchedUpdates<'_> {
+impl<L> BatchedUpdates<'_, L>
+where
+    L: ?Sized + Layer,
+{
     ///
     /// Insert an on-disk layer.
     ///
     // TODO remove the `layer` argument when `mapping` is refactored out of `LayerMap`
-    pub fn insert_historic(&mut self, layer_desc: PersistentLayerDesc) {
-        self.layer_map.insert_historic_noflush(layer_desc)
+    pub fn insert_historic(&mut self, layer_desc: PersistentLayerDesc, layer: Arc<L>) {
+        self.layer_map.insert_historic_noflush(layer_desc, layer)
     }
 
     ///
@@ -122,8 +145,31 @@ impl BatchedUpdates<'_> {
     ///
     /// This should be called when the corresponding file on disk has been deleted.
     ///
-    pub fn remove_historic(&mut self, layer_desc: PersistentLayerDesc) {
-        self.layer_map.remove_historic_noflush(layer_desc)
+    pub fn remove_historic(&mut self, layer_desc: PersistentLayerDesc, layer: Arc<L>) {
+        self.layer_map.remove_historic_noflush(layer_desc, layer)
+    }
+
+    /// Replaces existing layer iff it is the `expected`.
+    ///
+    /// If the expected layer has been removed it will not be inserted by this function.
+    ///
+    /// Returned `Replacement` describes succeeding in replacement or the reason why it could not
+    /// be done.
+    ///
+    /// TODO replacement can be done without buffering and rebuilding layer map updates.
+    ///      One way to do that is to add a layer of indirection for returned values, so
+    ///      that we can replace values only by updating a hashmap.
+    pub fn replace_historic(
+        &mut self,
+        expected_desc: PersistentLayerDesc,
+        expected: &Arc<L>,
+        new_desc: PersistentLayerDesc,
+        new: Arc<L>,
+    ) -> anyhow::Result<Replacement<Arc<L>>> {
+        fail::fail_point!("layermap-replace-notfound", |_| Ok(Replacement::NotFound));
+
+        self.layer_map
+            .replace_historic_noflush(expected_desc, expected, new_desc, new)
     }
 
     // We will flush on drop anyway, but this method makes it
@@ -139,19 +185,25 @@ impl BatchedUpdates<'_> {
 // than panic later or read without flushing.
 //
 // TODO maybe warn if flush hasn't explicitly been called
-impl Drop for BatchedUpdates<'_> {
+impl<L> Drop for BatchedUpdates<'_, L>
+where
+    L: ?Sized + Layer,
+{
     fn drop(&mut self) {
         self.layer_map.flush_updates();
     }
 }
 
 /// Return value of LayerMap::search
-pub struct SearchResult {
-    pub layer: Arc<PersistentLayerDesc>,
+pub struct SearchResult<L: ?Sized> {
+    pub layer: Arc<L>,
     pub lsn_floor: Lsn,
 }
 
-impl LayerMap {
+impl<L> LayerMap<L>
+where
+    L: ?Sized + Layer,
+{
     ///
     /// Find the latest layer (by lsn.end) that covers the given
     /// 'key', with lsn.start < 'end_lsn'.
@@ -183,7 +235,7 @@ impl LayerMap {
     /// NOTE: This only searches the 'historic' layers, *not* the
     /// 'open' and 'frozen' layers!
     ///
-    pub fn search(&self, key: Key, end_lsn: Lsn) -> Option<SearchResult> {
+    pub fn search(&self, key: Key, end_lsn: Lsn) -> Option<SearchResult<L>> {
         let version = self.historic.get().unwrap().get_version(end_lsn.0 - 1)?;
         let latest_delta = version.delta_coverage.query(key.to_i128());
         let latest_image = version.image_coverage.query(key.to_i128());
@@ -192,6 +244,7 @@ impl LayerMap {
             (None, None) => None,
             (None, Some(image)) => {
                 let lsn_floor = image.get_lsn_range().start;
+                let image = self.get_layer_from_mapping(&image.key()).clone();
                 Some(SearchResult {
                     layer: image,
                     lsn_floor,
@@ -199,6 +252,7 @@ impl LayerMap {
             }
             (Some(delta), None) => {
                 let lsn_floor = delta.get_lsn_range().start;
+                let delta = self.get_layer_from_mapping(&delta.key()).clone();
                 Some(SearchResult {
                     layer: delta,
                     lsn_floor,
@@ -209,6 +263,7 @@ impl LayerMap {
                 let image_is_newer = image.get_lsn_range().end >= delta.get_lsn_range().end;
                 let image_exact_match = img_lsn + 1 == end_lsn;
                 if image_is_newer || image_exact_match {
+                    let image = self.get_layer_from_mapping(&image.key()).clone();
                     Some(SearchResult {
                         layer: image,
                         lsn_floor: img_lsn,
@@ -216,6 +271,7 @@ impl LayerMap {
                 } else {
                     let lsn_floor =
                         std::cmp::max(delta.get_lsn_range().start, image.get_lsn_range().start + 1);
+                    let delta = self.get_layer_from_mapping(&delta.key()).clone();
                     Some(SearchResult {
                         layer: delta,
                         lsn_floor,
@@ -226,7 +282,7 @@ impl LayerMap {
     }
 
     /// Start a batch of updates, applied on drop
-    pub fn batch_update(&mut self) -> BatchedUpdates<'_> {
+    pub fn batch_update(&mut self) -> BatchedUpdates<'_, L> {
         BatchedUpdates { layer_map: self }
     }
 
@@ -236,32 +292,48 @@ impl LayerMap {
     /// Helper function for BatchedUpdates::insert_historic
     ///
     /// TODO(chi): remove L generic so that we do not need to pass layer object.
-    pub(self) fn insert_historic_noflush(&mut self, layer_desc: PersistentLayerDesc) {
+    pub(self) fn insert_historic_noflush(
+        &mut self,
+        layer_desc: PersistentLayerDesc,
+        layer: Arc<L>,
+    ) {
+        self.mapping.insert(layer_desc.key(), layer.clone());
+
         // TODO: See #3869, resulting #4088, attempted fix and repro #4094
 
-        if Self::is_l0(&layer_desc) {
+        if Self::is_l0(&layer) {
             self.l0_delta_layers.push(layer_desc.clone().into());
         }
 
         self.historic.insert(
-            historic_layer_coverage::LayerKey::from(&layer_desc),
+            historic_layer_coverage::LayerKey::from(&*layer),
             layer_desc.into(),
         );
     }
 
+    fn get_layer_from_mapping(&self, key: &PersistentLayerKey) -> &Arc<L> {
+        let layer = self
+            .mapping
+            .get(key)
+            .with_context(|| format!("{key:?}"))
+            .expect("inconsistent layer mapping");
+        layer
+    }
+
     ///
     /// Remove an on-disk layer from the map.
     ///
     /// Helper function for BatchedUpdates::remove_historic
     ///
-    pub fn remove_historic_noflush(&mut self, layer_desc: PersistentLayerDesc) {
+    pub fn remove_historic_noflush(&mut self, layer_desc: PersistentLayerDesc, layer: Arc<L>) {
         self.historic
-            .remove(historic_layer_coverage::LayerKey::from(&layer_desc));
-        let layer_key = layer_desc.key();
-        if Self::is_l0(&layer_desc) {
+            .remove(historic_layer_coverage::LayerKey::from(&*layer));
+        if Self::is_l0(&layer) {
             let len_before = self.l0_delta_layers.len();
             let mut l0_delta_layers = std::mem::take(&mut self.l0_delta_layers);
-            l0_delta_layers.retain(|other| other.key() != layer_key);
+            l0_delta_layers.retain(|other| {
+                !Self::compare_arced_layers(self.get_layer_from_mapping(&other.key()), &layer)
+            });
             self.l0_delta_layers = l0_delta_layers;
             // this assertion is related to use of Arc::ptr_eq in Self::compare_arced_layers,
             // there's a chance that the comparison fails at runtime due to it comparing (pointer,
@@ -272,6 +344,69 @@ impl LayerMap {
                 "failed to locate removed historic layer from l0_delta_layers"
             );
         }
+        self.mapping.remove(&layer_desc.key());
+    }
+
+    pub(self) fn replace_historic_noflush(
+        &mut self,
+        expected_desc: PersistentLayerDesc,
+        expected: &Arc<L>,
+        new_desc: PersistentLayerDesc,
+        new: Arc<L>,
+    ) -> anyhow::Result<Replacement<Arc<L>>> {
+        let key = historic_layer_coverage::LayerKey::from(&**expected);
+        let other = historic_layer_coverage::LayerKey::from(&*new);
+
+        let expected_l0 = Self::is_l0(expected);
+        let new_l0 = Self::is_l0(&new);
+
+        anyhow::ensure!(
+            key == other,
+            "expected and new must have equal LayerKeys: {key:?} != {other:?}"
+        );
+
+        anyhow::ensure!(
+            expected_l0 == new_l0,
+            "expected and new must both be l0 deltas or neither should be: {expected_l0} != {new_l0}"
+        );
+
+        let l0_index = if expected_l0 {
+            // find the index in case replace worked, we need to replace that as well
+            let pos = self.l0_delta_layers.iter().position(|slot| {
+                Self::compare_arced_layers(self.get_layer_from_mapping(&slot.key()), expected)
+            });
+
+            if pos.is_none() {
+                return Ok(Replacement::NotFound);
+            }
+            pos
+        } else {
+            None
+        };
+
+        let new_desc = Arc::new(new_desc);
+        let replaced = self.historic.replace(&key, new_desc.clone(), |existing| {
+            **existing == expected_desc
+        });
+
+        if let Replacement::Replaced { .. } = &replaced {
+            self.mapping.remove(&expected_desc.key());
+            self.mapping.insert(new_desc.key(), new);
+            if let Some(index) = l0_index {
+                self.l0_delta_layers[index] = new_desc;
+            }
+        }
+
+        let replaced = match replaced {
+            Replacement::Replaced { in_buffered } => Replacement::Replaced { in_buffered },
+            Replacement::NotFound => Replacement::NotFound,
+            Replacement::RemovalBuffered => Replacement::RemovalBuffered,
+            Replacement::Unexpected(x) => {
+                Replacement::Unexpected(self.get_layer_from_mapping(&x.key()).clone())
+            }
+        };
+
+        Ok(replaced)
     }
 
     /// Helper function for BatchedUpdates::drop.
@@ -319,8 +454,10 @@ impl LayerMap {
         Ok(true)
     }
 
-    pub fn iter_historic_layers(&self) -> impl '_ + Iterator<Item = Arc<PersistentLayerDesc>> {
-        self.historic.iter()
+    pub fn iter_historic_layers(&self) -> impl '_ + Iterator<Item = Arc<L>> {
+        self.historic
+            .iter()
+            .map(|x| self.get_layer_from_mapping(&x.key()).clone())
     }
 
     ///
@@ -335,7 +472,7 @@ impl LayerMap {
         &self,
         key_range: &Range<Key>,
         lsn: Lsn,
-    ) -> Result<Vec<(Range<Key>, Option<Arc<PersistentLayerDesc>>)>> {
+    ) -> Result<Vec<(Range<Key>, Option<Arc<L>>)>> {
         let version = match self.historic.get().unwrap().get_version(lsn.0) {
             Some(v) => v,
             None => return Ok(vec![]),
@@ -345,26 +482,36 @@ impl LayerMap {
         let end = key_range.end.to_i128();
 
         // Initialize loop variables
-        let mut coverage: Vec<(Range<Key>, Option<Arc<PersistentLayerDesc>>)> = vec![];
+        let mut coverage: Vec<(Range<Key>, Option<Arc<L>>)> = vec![];
         let mut current_key = start;
         let mut current_val = version.image_coverage.query(start);
 
         // Loop through the change events and push intervals
         for (change_key, change_val) in version.image_coverage.range(start..end) {
             let kr = Key::from_i128(current_key)..Key::from_i128(change_key);
-            coverage.push((kr, current_val.take()));
+            coverage.push((
+                kr,
+                current_val
+                    .take()
+                    .map(|l| self.get_layer_from_mapping(&l.key()).clone()),
+            ));
             current_key = change_key;
             current_val = change_val.clone();
         }
 
         // Add the final interval
         let kr = Key::from_i128(current_key)..Key::from_i128(end);
-        coverage.push((kr, current_val.take()));
+        coverage.push((
+            kr,
+            current_val
+                .take()
+                .map(|l| self.get_layer_from_mapping(&l.key()).clone()),
+        ));
 
         Ok(coverage)
     }
 
-    pub fn is_l0(layer: &PersistentLayerDesc) -> bool {
+    pub fn is_l0(layer: &L) -> bool {
         range_eq(&layer.get_key_range(), &(Key::MIN..Key::MAX))
     }
 
@@ -390,7 +537,7 @@ impl LayerMap {
     /// TODO The optimal number should probably be slightly higher than 1, but to
     ///      implement that we need to plumb a lot more context into this function
     ///      than just the current partition_range.
-    pub fn is_reimage_worthy(layer: &PersistentLayerDesc, partition_range: &Range<Key>) -> bool {
+    pub fn is_reimage_worthy(layer: &L, partition_range: &Range<Key>) -> bool {
         // Case 1
         if !Self::is_l0(layer) {
             return true;
@@ -448,7 +595,9 @@ impl LayerMap {
                     let kr = Key::from_i128(current_key)..Key::from_i128(change_key);
                     let lr = lsn.start..val.get_lsn_range().start;
                     if !kr.is_empty() {
-                        let base_count = Self::is_reimage_worthy(&val, key) as usize;
+                        let base_count =
+                            Self::is_reimage_worthy(self.get_layer_from_mapping(&val.key()), key)
+                                as usize;
                         let new_limit = limit.map(|l| l - base_count);
                         let max_stacked_deltas_underneath =
                             self.count_deltas(&kr, &lr, new_limit)?;
@@ -471,7 +620,9 @@ impl LayerMap {
                 let lr = lsn.start..val.get_lsn_range().start;
 
                 if !kr.is_empty() {
-                    let base_count = Self::is_reimage_worthy(&val, key) as usize;
+                    let base_count =
+                        Self::is_reimage_worthy(self.get_layer_from_mapping(&val.key()), key)
+                            as usize;
                     let new_limit = limit.map(|l| l - base_count);
                     let max_stacked_deltas_underneath = self.count_deltas(&kr, &lr, new_limit)?;
                     max_stacked_deltas = std::cmp::max(
@@ -621,8 +772,12 @@ impl LayerMap {
     }
 
     /// Return all L0 delta layers
-    pub fn get_level0_deltas(&self) -> Result<Vec<Arc<PersistentLayerDesc>>> {
-        Ok(self.l0_delta_layers.to_vec())
+    pub fn get_level0_deltas(&self) -> Result<Vec<Arc<L>>> {
+        Ok(self
+            .l0_delta_layers
+            .iter()
+            .map(|x| self.get_layer_from_mapping(&x.key()).clone())
+            .collect())
     }
 
     /// debugging function to print out the contents of the layer map
@@ -647,51 +802,72 @@ impl LayerMap {
         println!("End dump LayerMap");
         Ok(())
     }
+
+    /// Similar to `Arc::ptr_eq`, but only compares the object pointers, not vtables.
+    ///
+    /// Returns `true` if the two `Arc` point to the same layer, false otherwise.
+    #[inline(always)]
+    pub fn compare_arced_layers(left: &Arc<L>, right: &Arc<L>) -> bool {
+        // "dyn Trait" objects are "fat pointers" in that they have two components:
+        // - pointer to the object
+        // - pointer to the vtable
+        //
+        // rust does not provide a guarantee that these vtables are unique, but however
+        // `Arc::ptr_eq` as of writing (at least up to 1.67) uses a comparison where both the
+        // pointer and the vtable need to be equal.
+        //
+        // See: https://github.com/rust-lang/rust/issues/103763
+        //
+        // A future version of rust will most likely use this form below, where we cast each
+        // pointer into a pointer to unit, which drops the inaccessible vtable pointer, making it
+        // not affect the comparison.
+        //
+        // See: https://github.com/rust-lang/rust/pull/106450
+        let left = Arc::as_ptr(left) as *const ();
+        let right = Arc::as_ptr(right) as *const ();
+
+        left == right
+    }
 }
 
 #[cfg(test)]
 mod tests {
-    use super::LayerMap;
-    use crate::tenant::storage_layer::{tests::LayerDescriptor, LayerFileName};
+    use super::{LayerMap, Replacement};
+    use crate::tenant::storage_layer::{Layer, LayerDescriptor, LayerFileName};
     use std::str::FromStr;
     use std::sync::Arc;
 
     mod l0_delta_layers_updated {
 
-        use crate::tenant::{
-            storage_layer::{PersistentLayer, PersistentLayerDesc},
-            timeline::LayerFileManager,
-        };
-
         use super::*;
 
         #[test]
         fn for_full_range_delta() {
             // l0_delta_layers are used by compaction, and should observe all buffered updates
             l0_delta_layers_updated_scenario(
-                 "000000000000000000000000000000000000-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF__0000000053423C21-0000000053424D69",
-                 true
-             )
+                "000000000000000000000000000000000000-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF__0000000053423C21-0000000053424D69",
+                true
+            )
         }
 
         #[test]
         fn for_non_full_range_delta() {
             // has minimal uncovered areas compared to l0_delta_layers_updated_on_insert_replace_remove_for_full_range_delta
             l0_delta_layers_updated_scenario(
-                 "000000000000000000000000000000000001-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFE__0000000053423C21-0000000053424D69",
-                 // because not full range
-                 false
-             )
+                "000000000000000000000000000000000001-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFE__0000000053423C21-0000000053424D69",
+                // because not full range
+                false
+            )
         }
 
         #[test]
         fn for_image() {
             l0_delta_layers_updated_scenario(
-                 "000000000000000000000000000000000000-000000000000000000000000000000010000__0000000053424D69",
-                 // code only checks if it is a full range layer, doesn't care about images, which must
-                 // mean we should in practice never have full range images
-                 false
-             )
+                "000000000000000000000000000000000000-000000000000000000000000000000010000__0000000053424D69",
+                // code only checks if it is a full range layer, doesn't care about images, which must
+                // mean we should in practice never have full range images
+                false
+            )
         }
 
         #[test]
@@ -707,16 +883,16 @@ mod tests {
             let not_found = Arc::new(layer.clone());
             let new_version = Arc::new(layer);
 
-            // after the immutable storage state refactor, the replace operation
-            // will not use layer map any more. We keep it here for consistency in test cases
-            // and can remove it in the future.
-            let _map = LayerMap::default();
+            let mut map = LayerMap::default();
 
-            let mut mapping = LayerFileManager::new();
+            let res = map.batch_update().replace_historic(
+                not_found.get_persistent_layer_desc(),
+                &not_found,
+                new_version.get_persistent_layer_desc(),
+                new_version,
+            );
 
-            mapping
-                .replace_and_verify(not_found, new_version)
-                .unwrap_err();
+            assert!(matches!(res, Ok(Replacement::NotFound)), "{res:?}");
         }
 
         fn l0_delta_layers_updated_scenario(layer_name: &str, expected_l0: bool) {
@@ -727,44 +903,49 @@ mod tests {
             let downloaded = Arc::new(skeleton);
 
             let mut map = LayerMap::default();
-            let mut mapping = LayerFileManager::new();
 
             // two disjoint Arcs in different lifecycle phases. even if it seems they must be the
             // same layer, we use LayerMap::compare_arced_layers as the identity of layers.
-            assert_eq!(remote.layer_desc(), downloaded.layer_desc());
+            assert!(!LayerMap::compare_arced_layers(&remote, &downloaded));
 
             let expected_in_counts = (1, usize::from(expected_l0));
 
             map.batch_update()
-                .insert_historic(remote.layer_desc().clone());
-            mapping.insert(remote.clone());
-            assert_eq!(
-                count_layer_in(&map, remote.layer_desc()),
-                expected_in_counts
-            );
+                .insert_historic(remote.get_persistent_layer_desc(), remote.clone());
+            assert_eq!(count_layer_in(&map, &remote), expected_in_counts);
 
-            mapping
-                .replace_and_verify(remote, downloaded.clone())
+            let replaced = map
+                .batch_update()
+                .replace_historic(
+                    remote.get_persistent_layer_desc(),
+                    &remote,
+                    downloaded.get_persistent_layer_desc(),
+                    downloaded.clone(),
+                )
                 .expect("name derived attributes are the same");
-            assert_eq!(
-                count_layer_in(&map, downloaded.layer_desc()),
-                expected_in_counts
+            assert!(
+                matches!(replaced, Replacement::Replaced { .. }),
+                "{replaced:?}"
             );
+            assert_eq!(count_layer_in(&map, &downloaded), expected_in_counts);
 
             map.batch_update()
-                .remove_historic(downloaded.layer_desc().clone());
-            assert_eq!(count_layer_in(&map, downloaded.layer_desc()), (0, 0));
+                .remove_historic(downloaded.get_persistent_layer_desc(), downloaded.clone());
+            assert_eq!(count_layer_in(&map, &downloaded), (0, 0));
         }
 
-        fn count_layer_in(map: &LayerMap, layer: &PersistentLayerDesc) -> (usize, usize) {
+        fn count_layer_in<L: Layer + ?Sized>(map: &LayerMap<L>, layer: &Arc<L>) -> (usize, usize) {
             let historic = map
                 .iter_historic_layers()
-                .filter(|x| x.key() == layer.key())
+                .filter(|x| LayerMap::compare_arced_layers(x, layer))
                 .count();
             let l0s = map
                 .get_level0_deltas()
                 .expect("why does this return a result");
-            let l0 = l0s.iter().filter(|x| x.key() == layer.key()).count();
+            let l0 = l0s
+                .iter()
+                .filter(|x| LayerMap::compare_arced_layers(x, layer))
+                .count();
 
             (historic, l0)
         }

pageserver/src/tenant/layer_map/historic_layer_coverage.rs

@@ -3,8 +3,6 @@ use std::ops::Range;
 
 use tracing::info;
 
-use crate::tenant::storage_layer::PersistentLayerDesc;
-
 use super::layer_coverage::LayerCoverageTuple;
 
 /// Layers in this module are identified and indexed by this data.
@@ -43,8 +41,8 @@ impl Ord for LayerKey {
     }
 }
 
-impl From<&PersistentLayerDesc> for LayerKey {
-    fn from(layer: &PersistentLayerDesc) -> Self {
+impl<'a, L: crate::tenant::storage_layer::Layer + ?Sized> From<&'a L> for LayerKey {
+    fn from(layer: &'a L) -> Self {
         let kr = layer.get_key_range();
         let lr = layer.get_lsn_range();
         LayerKey {
@@ -456,6 +454,59 @@ impl<Value: Clone> BufferedHistoricLayerCoverage<Value> {
         self.buffer.insert(layer_key, None);
     }
 
+    /// Replaces a previous layer with a new layer value.
+    ///
+    /// The replacement is conditional on:
+    /// - there is an existing `LayerKey` record
+    /// - there is no buffered removal for the given `LayerKey`
+    /// - the given closure returns true for the current `Value`
+    ///
+    /// The closure is used to compare the latest value (buffered insert, or existing layer)
+    /// against some expectation. This allows to use `Arc::ptr_eq` or similar which would be
+    /// inaccessible via `PartialEq` trait.
+    ///
+    /// Returns a `Replacement` value describing the outcome; only the case of
+    /// `Replacement::Replaced` modifies the map and requires a rebuild.
+    pub fn replace<F>(
+        &mut self,
+        layer_key: &LayerKey,
+        new: Value,
+        check_expected: F,
+    ) -> Replacement<Value>
+    where
+        F: FnOnce(&Value) -> bool,
+    {
+        let (slot, in_buffered) = match self.buffer.get(layer_key) {
+            Some(inner @ Some(_)) => {
+                // we compare against the buffered version, because there will be a later
+                // rebuild before querying
+                (inner.as_ref(), true)
+            }
+            Some(None) => {
+                // buffer has removal for this key; it will not be equivalent by any check_expected.
+                return Replacement::RemovalBuffered;
+            }
+            None => {
+                // no pending modification for the key, check layers
+                (self.layers.get(layer_key), false)
+            }
+        };
+
+        match slot {
+            Some(existing) if !check_expected(existing) => {
+                // unfortunate clone here, but otherwise the nll borrowck grows the region of
+                // 'a to cover the whole function, and we could not mutate in the other
+                // Some(existing) branch
+                Replacement::Unexpected(existing.clone())
+            }
+            None => Replacement::NotFound,
+            Some(_existing) => {
+                self.insert(layer_key.to_owned(), new);
+                Replacement::Replaced { in_buffered }
+            }
+        }
+    }
+
     pub fn rebuild(&mut self) {
         // Find the first LSN that needs to be rebuilt
         let rebuild_since: u64 = match self.buffer.iter().next() {
@@ -524,6 +575,22 @@ impl<Value: Clone> BufferedHistoricLayerCoverage<Value> {
     }
 }
 
+/// Outcome of the replace operation.
+#[derive(Debug)]
+pub enum Replacement<Value> {
+    /// Previous value was replaced with the new value.
+    Replaced {
+        /// Replacement happened for a scheduled insert.
+        in_buffered: bool,
+    },
+    /// Key was not found buffered updates or existing layers.
+    NotFound,
+    /// Key has been scheduled for removal, it was not replaced.
+    RemovalBuffered,
+    /// Previous value was rejected by the closure.
+    Unexpected(Value),
+}
+
 #[test]
 fn test_retroactive_regression_1() {
     let mut map = BufferedHistoricLayerCoverage::new();
@@ -632,3 +699,139 @@ fn test_retroactive_simple() {
         assert_eq!(version.image_coverage.query(8), Some("Image 4".to_string()));
     }
 }
+
+#[test]
+fn test_retroactive_replacement() {
+    let mut map = BufferedHistoricLayerCoverage::new();
+
+    let keys = [
+        LayerKey {
+            key: 0..5,
+            lsn: 100..101,
+            is_image: true,
+        },
+        LayerKey {
+            key: 3..9,
+            lsn: 110..111,
+            is_image: true,
+        },
+        LayerKey {
+            key: 4..6,
+            lsn: 120..121,
+            is_image: true,
+        },
+    ];
+
+    let layers = [
+        "Image 1".to_string(),
+        "Image 2".to_string(),
+        "Image 3".to_string(),
+    ];
+
+    for (key, layer) in keys.iter().zip(layers.iter()) {
+        map.insert(key.to_owned(), layer.to_owned());
+    }
+
+    // rebuild is not necessary here, because replace works for both buffered updates and existing
+    // layers.
+
+    for (key, orig_layer) in keys.iter().zip(layers.iter()) {
+        let replacement = format!("Remote {orig_layer}");
+
+        // evict
+        let ret = map.replace(key, replacement.clone(), |l| l == orig_layer);
+        assert!(
+            matches!(ret, Replacement::Replaced { .. }),
+            "replace {orig_layer}: {ret:?}"
+        );
+        map.rebuild();
+
+        let at = key.lsn.end + 1;
+
+        let version = map.get().expect("rebuilt").get_version(at).unwrap();
+        assert_eq!(
+            version.image_coverage.query(4).as_deref(),
+            Some(replacement.as_str()),
+            "query for 4 at version {at} after eviction",
+        );
+
+        // download
+        let ret = map.replace(key, orig_layer.clone(), |l| l == &replacement);
+        assert!(
+            matches!(ret, Replacement::Replaced { .. }),
+            "replace {orig_layer} back: {ret:?}"
+        );
+        map.rebuild();
+        let version = map.get().expect("rebuilt").get_version(at).unwrap();
+        assert_eq!(
+            version.image_coverage.query(4).as_deref(),
+            Some(orig_layer.as_str()),
+            "query for 4 at version {at} after download",
+        );
+    }
+}
+
+#[test]
+fn missing_key_is_not_inserted_with_replace() {
+    let mut map = BufferedHistoricLayerCoverage::new();
+    let key = LayerKey {
+        key: 0..5,
+        lsn: 100..101,
+        is_image: true,
+    };
+
+    let ret = map.replace(&key, "should not replace", |_| true);
+    assert!(matches!(ret, Replacement::NotFound), "{ret:?}");
+    map.rebuild();
+    assert!(map
+        .get()
+        .expect("no changes to rebuild")
+        .get_version(102)
+        .is_none());
+}
+
+#[test]
+fn replacing_buffered_insert_and_remove() {
+    let mut map = BufferedHistoricLayerCoverage::new();
+    let key = LayerKey {
+        key: 0..5,
+        lsn: 100..101,
+        is_image: true,
+    };
+
+    map.insert(key.clone(), "Image 1");
+    let ret = map.replace(&key, "Remote Image 1", |&l| l == "Image 1");
+    assert!(
+        matches!(ret, Replacement::Replaced { in_buffered: true }),
+        "{ret:?}"
+    );
+    map.rebuild();
+
+    assert_eq!(
+        map.get()
+            .expect("rebuilt")
+            .get_version(102)
+            .unwrap()
+            .image_coverage
+            .query(4),
+        Some("Remote Image 1")
+    );
+
+    map.remove(key.clone());
+    let ret = map.replace(&key, "should not replace", |_| true);
+    assert!(
+        matches!(ret, Replacement::RemovalBuffered),
+        "cannot replace after scheduled remove: {ret:?}"
+    );
+
+    map.rebuild();
+
+    let ret = map.replace(&key, "should not replace", |_| true);
+    assert!(
+        matches!(ret, Replacement::NotFound),
+        "cannot replace after remove + rebuild: {ret:?}"
+    );
+
+    let at_version = map.get().expect("rebuilt").get_version(102);
+    assert!(at_version.is_none());
+}

pageserver/src/tenant/mgr.rs

@@ -675,7 +675,7 @@ pub async fn immediate_gc(
         .get(&tenant_id)
         .map(Arc::clone)
         .with_context(|| format!("tenant {tenant_id}"))
-        .map_err(|e| ApiError::NotFound(e.into()))?;
+        .map_err(ApiError::NotFound)?;
 
     let gc_horizon = gc_req.gc_horizon.unwrap_or_else(|| tenant.get_gc_horizon());
     // Use tenant's pitr setting
@@ -724,11 +724,11 @@ pub async fn immediate_compact(
         .get(&tenant_id)
         .map(Arc::clone)
         .with_context(|| format!("tenant {tenant_id}"))
-        .map_err(|e| ApiError::NotFound(e.into()))?;
+        .map_err(ApiError::NotFound)?;
 
     let timeline = tenant
         .get_timeline(timeline_id, true)
-        .map_err(|e| ApiError::NotFound(e.into()))?;
+        .map_err(ApiError::NotFound)?;
 
     // Run in task_mgr to avoid race with tenant_detach operation
     let ctx = ctx.detached_child(TaskKind::Compaction, DownloadBehavior::Download);

pageserver/src/tenant/remote_timeline_client.rs

@@ -753,18 +753,22 @@ impl RemoteTimelineClient {
 
         // Have a failpoint that can use the `pause` failpoint action.
         // We don't want to block the executor thread, hence, spawn_blocking + await.
-        if cfg!(feature = "testing") {
-            tokio::task::spawn_blocking({
-                let current = tracing::Span::current();
-                move || {
-                    let _entered = current.entered();
-                    tracing::info!("at failpoint persist_deleted_index_part");
-                    fail::fail_point!("persist_deleted_index_part");
-                }
-            })
-            .await
-            .expect("spawn_blocking");
-        }
+        #[cfg(feature = "testing")]
+        tokio::task::spawn_blocking({
+            let current = tracing::Span::current();
+            move || {
+                let _entered = current.entered();
+                tracing::info!(
+                    "at failpoint persist_index_part_with_deleted_flag_after_set_before_upload_pause"
+                );
+                fail::fail_point!(
+                    "persist_index_part_with_deleted_flag_after_set_before_upload_pause"
+                );
+            }
+        })
+        .await
+        .expect("spawn_blocking");
+
         upload::upload_index_part(
             self.conf,
             &self.storage_impl,
@@ -862,8 +866,10 @@ impl RemoteTimelineClient {
                 "Found {} files not bound to index_file.json, proceeding with their deletion",
                 remaining.len()
             );
-            warn!("About to remove {} files", remaining.len());
-            self.storage_impl.delete_objects(&remaining).await?;
+            for file in remaining {
+                warn!("Removing {}", file.object_name().unwrap_or_default());
+                self.storage_impl.delete(&file).await?;
+            }
         }
 
         let index_file_path = timeline_storage_path.join(Path::new(IndexPart::FILE_NAME));
@@ -1365,7 +1371,7 @@ mod tests {
     struct TestSetup {
         runtime: &'static tokio::runtime::Runtime,
         entered_runtime: EnterGuard<'static>,
-        harness: TenantHarness,
+        harness: TenantHarness<'static>,
         tenant: Arc<Tenant>,
         tenant_ctx: RequestContext,
         remote_fs_dir: PathBuf,
@@ -1386,12 +1392,7 @@ mod tests {
             let harness = TenantHarness::create(test_name)?;
             let (tenant, ctx) = runtime.block_on(harness.load());
             // create an empty timeline directory
-            let _ = runtime.block_on(tenant.create_test_timeline(
-                TIMELINE_ID,
-                Lsn(8),
-                DEFAULT_PG_VERSION,
-                &ctx,
-            ))?;
+            let _ = tenant.create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)?;
 
             let remote_fs_dir = harness.conf.workdir.join("remote_fs");
             std::fs::create_dir_all(remote_fs_dir)?;

pageserver/src/tenant/storage_layer.rs

@@ -12,7 +12,7 @@ use crate::context::RequestContext;
 use crate::repository::{Key, Value};
 use crate::task_mgr::TaskKind;
 use crate::walrecord::NeonWalRecord;
-use anyhow::{Context, Result};
+use anyhow::Result;
 use bytes::Bytes;
 use enum_map::EnumMap;
 use enumset::EnumSet;
@@ -24,7 +24,7 @@ use pageserver_api::models::{
 use std::ops::Range;
 use std::path::PathBuf;
 use std::sync::{Arc, Mutex};
-use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};
+use std::time::{Duration, SystemTime, UNIX_EPOCH};
 use tracing::warn;
 use utils::history_buffer::HistoryBufferWithDropCounter;
 use utils::rate_limit::RateLimit;
@@ -176,10 +176,13 @@ impl LayerAccessStats {
     /// Create an empty stats object and record a [`LayerLoad`] event with the given residence status.
     ///
     /// See [`record_residence_event`] for why you need to do this while holding the layer map lock.
-    pub(crate) fn for_loading_layer(
-        layer_map_lock_held_witness: &BatchedUpdates<'_>,
+    pub(crate) fn for_loading_layer<L>(
+        layer_map_lock_held_witness: &BatchedUpdates<'_, L>,
         status: LayerResidenceStatus,
-    ) -> Self {
+    ) -> Self
+    where
+        L: ?Sized + Layer,
+    {
         let new = LayerAccessStats(Mutex::new(LayerAccessStatsLocked::default()));
         new.record_residence_event(
             layer_map_lock_held_witness,
@@ -194,11 +197,14 @@ impl LayerAccessStats {
     /// The `new_status` is not recorded in `self`.
     ///
     /// See [`record_residence_event`] for why you need to do this while holding the layer map lock.
-    pub(crate) fn clone_for_residence_change(
+    pub(crate) fn clone_for_residence_change<L>(
         &self,
-        layer_map_lock_held_witness: &BatchedUpdates<'_>,
+        layer_map_lock_held_witness: &BatchedUpdates<'_, L>,
         new_status: LayerResidenceStatus,
-    ) -> LayerAccessStats {
+    ) -> LayerAccessStats
+    where
+        L: ?Sized + Layer,
+    {
         let clone = {
             let inner = self.0.lock().unwrap();
             inner.clone()
@@ -226,12 +232,14 @@ impl LayerAccessStats {
     /// - Compact: Grab layer map lock, add the new L1 to layer map and remove the L0s, release layer map lock.
     /// - Eviction: observes the new L1 layer whose only activity timestamp is the LayerCreate event.
     ///
-    pub(crate) fn record_residence_event(
+    pub(crate) fn record_residence_event<L>(
         &self,
-        _layer_map_lock_held_witness: &BatchedUpdates<'_>,
+        _layer_map_lock_held_witness: &BatchedUpdates<'_, L>,
         status: LayerResidenceStatus,
         reason: LayerResidenceEventReason,
-    ) {
+    ) where
+        L: ?Sized + Layer,
+    {
         let mut locked = self.0.lock().unwrap();
         locked.iter_mut().for_each(|inner| {
             inner
@@ -335,8 +343,7 @@ impl LayerAccessStats {
 /// All layers should implement a minimal `std::fmt::Debug` without tenant or
 /// timeline names, because those are known in the context of which the layers
 /// are used in (timeline).
-#[async_trait::async_trait]
-pub trait Layer: std::fmt::Debug + Send + Sync + 'static {
+pub trait Layer: std::fmt::Debug + Send + Sync {
     /// Range of keys that this layer covers
     fn get_key_range(&self) -> Range<Key>;
 
@@ -366,74 +373,13 @@ pub trait Layer: std::fmt::Debug + Send + Sync + 'static {
     /// is available. If this returns ValueReconstructResult::Continue, look up
     /// the predecessor layer and call again with the same 'reconstruct_data' to
     /// collect more data.
-    fn get_value_reconstruct_data_blocking(
+    fn get_value_reconstruct_data(
         &self,
         key: Key,
         lsn_range: Range<Lsn>,
-        reconstruct_data: ValueReconstructState,
-        ctx: RequestContext,
-    ) -> Result<(ValueReconstructState, ValueReconstructResult)>;
-
-    /// CANCEL SAFETY: if the returned future is dropped,
-    /// the wrapped closure still run to completion and the return value discarded.
-    /// For the case of get_value_reconstruct_data, we expect the closure to not
-    /// have any side effects, as it only attempts to read a layer (and stuff like
-    /// page cache isn't considered a real side effect).
-    /// But, ...
-    /// TRACING:
-    /// If the returned future is cancelled, the spawn_blocking span can outlive
-    /// the caller's span.
-    /// So, technically, we should be using `parent: None` and `follows_from: current`
-    /// instead. However, in practice, the advantage of maintaining the span stack
-    /// in logs outweighs the disadvantage of having a dangling span in a case that
-    /// is not expected to happen because in pageserver we generally don't drop pending futures.
-    async fn get_value_reconstruct_data(
-        self: Arc<Self>,
-        key: Key,
-        lsn_range: Range<Lsn>,
-        reconstruct_data: ValueReconstructState,
-        ctx: RequestContext,
-    ) -> Result<(ValueReconstructState, ValueReconstructResult)> {
-        let span = tracing::info_span!("get_value_reconstruct_data_spawn_blocking");
-        static USE_SPAWN_BLOCKING: Lazy<bool> = Lazy::new(|| {
-            let val = std::env::var("PAGESERVER_LAYER_GET_RECONSTRUCT_DATA_USE_SPAWN_BLOCKING")
-                .map(|s| s == "1")
-                .unwrap_or(false);
-            tracing::info!("PAGESERVER_LAYER_GET_RECONSTRUCT_DATA_USE_SPAWN_BLOCKING={val}");
-            val
-        });
-        let use_spawn_blocking = *USE_SPAWN_BLOCKING;
-        let start = Instant::now();
-        let res = if !use_spawn_blocking {
-            anyhow::Ok(self.get_value_reconstruct_data_blocking(
-                key,
-                lsn_range,
-                reconstruct_data,
-                ctx,
-            ))
-        } else {
-            crate::metrics::LAYER_GET_VALUE_RECONSTRUCT_DATA_SPAWN_BLOCKING_STARTED_COUNT.inc();
-            crate::metrics::LAYER_GET_VALUE_RECONSTRUCT_DATA_SPAWN_BLOCKING_ACTIVE_GAUGE.inc();
-            let res = tokio::task::spawn_blocking(move || {
-                crate::metrics::LAYER_GET_VALUE_RECONSTRUCT_DATA_SPAWN_BLOCKING_QUEUE_DELAY
-                    .observe(start.elapsed().as_secs_f64());
-                let _enter = span.enter();
-                self.get_value_reconstruct_data_blocking(key, lsn_range, reconstruct_data, ctx)
-            })
-            .await
-            .context("spawn_blocking");
-            crate::metrics::LAYER_GET_VALUE_RECONSTRUCT_DATA_SPAWN_BLOCKING_ACTIVE_GAUGE.dec();
-            res
-        };
-        let histo = match &res {
-            Ok(Ok(_)) => &crate::metrics::LAYER_GET_VALUE_RECONSTRUCT_DATA_COMPLETION_TIME_OK,
-            Ok(Err(_)) | Err(_) => {
-                &crate::metrics::LAYER_GET_VALUE_RECONSTRUCT_DATA_COMPLETION_TIME_ERROR
-            }
-        };
-        histo.observe(start.elapsed().as_secs_f64());
-        res?
-    }
+        reconstruct_data: &mut ValueReconstructState,
+        ctx: &RequestContext,
+    ) -> Result<ValueReconstructResult>;
 
     /// A short ID string that uniquely identifies the given layer within a [`LayerMap`].
     fn short_id(&self) -> String;
@@ -443,10 +389,10 @@ pub trait Layer: std::fmt::Debug + Send + Sync + 'static {
 }
 
 /// Returned by [`Layer::iter`]
-pub type LayerIter<'i> = Box<dyn Iterator<Item = Result<(Key, Lsn, Value)>> + 'i + Send>;
+pub type LayerIter<'i> = Box<dyn Iterator<Item = Result<(Key, Lsn, Value)>> + 'i>;
 
 /// Returned by [`Layer::key_iter`]
-pub type LayerKeyIter<'i> = Box<dyn Iterator<Item = (Key, Lsn, u64)> + 'i + Send>;
+pub type LayerKeyIter<'i> = Box<dyn Iterator<Item = (Key, Lsn, u64)> + 'i>;
 
 /// A Layer contains all data in a "rectangle" consisting of a range of keys and
 /// range of LSNs.
@@ -527,126 +473,94 @@ pub fn downcast_remote_layer(
     }
 }
 
-pub mod tests {
-    use super::*;
+/// Holds metadata about a layer without any content. Used mostly for testing.
+///
+/// To use filenames as fixtures, parse them as [`LayerFileName`] then convert from that to a
+/// LayerDescriptor.
+#[derive(Clone, Debug)]
+pub struct LayerDescriptor {
+    pub key: Range<Key>,
+    pub lsn: Range<Lsn>,
+    pub is_incremental: bool,
+    pub short_id: String,
+}
 
-    /// Holds metadata about a layer without any content. Used mostly for testing.
-    ///
-    /// To use filenames as fixtures, parse them as [`LayerFileName`] then convert from that to a
-    /// LayerDescriptor.
-    #[derive(Clone, Debug)]
-    pub struct LayerDescriptor {
-        base: PersistentLayerDesc,
+impl LayerDescriptor {
+    /// `LayerDescriptor` is only used for testing purpose so it does not matter whether it is image / delta,
+    /// and the tenant / timeline id does not matter.
+    pub fn get_persistent_layer_desc(&self) -> PersistentLayerDesc {
+        PersistentLayerDesc::new_delta(
+            TenantId::from_array([0; 16]),
+            TimelineId::from_array([0; 16]),
+            self.key.clone(),
+            self.lsn.clone(),
+            233,
+        )
+    }
+}
+
+impl Layer for LayerDescriptor {
+    fn get_key_range(&self) -> Range<Key> {
+        self.key.clone()
     }
 
-    impl From<PersistentLayerDesc> for LayerDescriptor {
-        fn from(base: PersistentLayerDesc) -> Self {
-            Self { base }
-        }
+    fn get_lsn_range(&self) -> Range<Lsn> {
+        self.lsn.clone()
     }
 
-    #[async_trait::async_trait]
-    impl Layer for LayerDescriptor {
-        fn dump(&self, _verbose: bool, _ctx: &RequestContext) -> Result<()> {
-            todo!()
-        }
-
-        /// Boilerplate to implement the Layer trait, always use layer_desc for persistent layers.
-        fn get_key_range(&self) -> Range<Key> {
-            self.layer_desc().key_range.clone()
-        }
-
-        /// Boilerplate to implement the Layer trait, always use layer_desc for persistent layers.
-        fn get_lsn_range(&self) -> Range<Lsn> {
-            self.layer_desc().lsn_range.clone()
-        }
-
-        fn get_value_reconstruct_data_blocking(
-            &self,
-            _key: Key,
-            _lsn_range: Range<Lsn>,
-            _reconstruct_data: ValueReconstructState,
-            _ctx: RequestContext,
-        ) -> Result<(ValueReconstructState, ValueReconstructResult)> {
-            todo!("This method shouldn't be part of the Layer trait")
-        }
-
-        /// Boilerplate to implement the Layer trait, always use layer_desc for persistent layers.
-        fn is_incremental(&self) -> bool {
-            self.layer_desc().is_incremental
-        }
-
-        /// Boilerplate to implement the Layer trait, always use layer_desc for persistent layers.
-        fn short_id(&self) -> String {
-            self.layer_desc().short_id()
-        }
+    fn is_incremental(&self) -> bool {
+        self.is_incremental
     }
 
-    impl PersistentLayer for LayerDescriptor {
-        fn layer_desc(&self) -> &PersistentLayerDesc {
-            &self.base
-        }
-
-        fn local_path(&self) -> Option<PathBuf> {
-            unimplemented!()
-        }
-
-        fn iter(&self, _: &RequestContext) -> Result<LayerIter<'_>> {
-            unimplemented!()
-        }
-
-        fn key_iter(&self, _: &RequestContext) -> Result<LayerKeyIter<'_>> {
-            unimplemented!()
-        }
-
-        fn delete_resident_layer_file(&self) -> Result<()> {
-            unimplemented!()
-        }
-
-        fn info(&self, _: LayerAccessStatsReset) -> HistoricLayerInfo {
-            unimplemented!()
-        }
-
-        fn access_stats(&self) -> &LayerAccessStats {
-            unimplemented!()
-        }
+    fn get_value_reconstruct_data(
+        &self,
+        _key: Key,
+        _lsn_range: Range<Lsn>,
+        _reconstruct_data: &mut ValueReconstructState,
+        _ctx: &RequestContext,
+    ) -> Result<ValueReconstructResult> {
+        todo!("This method shouldn't be part of the Layer trait")
     }
 
-    impl From<DeltaFileName> for LayerDescriptor {
-        fn from(value: DeltaFileName) -> Self {
-            LayerDescriptor {
-                base: PersistentLayerDesc::new_delta(
-                    TenantId::from_array([0; 16]),
-                    TimelineId::from_array([0; 16]),
-                    value.key_range,
-                    value.lsn_range,
-                    233,
-                ),
-            }
-        }
+    fn short_id(&self) -> String {
+        self.short_id.clone()
     }
 
-    impl From<ImageFileName> for LayerDescriptor {
-        fn from(value: ImageFileName) -> Self {
-            LayerDescriptor {
-                base: PersistentLayerDesc::new_img(
-                    TenantId::from_array([0; 16]),
-                    TimelineId::from_array([0; 16]),
-                    value.key_range,
-                    value.lsn,
-                    false,
-                    233,
-                ),
-            }
+    fn dump(&self, _verbose: bool, _ctx: &RequestContext) -> Result<()> {
+        todo!()
+    }
+}
+
+impl From<DeltaFileName> for LayerDescriptor {
+    fn from(value: DeltaFileName) -> Self {
+        let short_id = value.to_string();
+        LayerDescriptor {
+            key: value.key_range,
+            lsn: value.lsn_range,
+            is_incremental: true,
+            short_id,
         }
     }
+}
 
-    impl From<LayerFileName> for LayerDescriptor {
-        fn from(value: LayerFileName) -> Self {
-            match value {
-                LayerFileName::Delta(d) => Self::from(d),
-                LayerFileName::Image(i) => Self::from(i),
-            }
+impl From<ImageFileName> for LayerDescriptor {
+    fn from(value: ImageFileName) -> Self {
+        let short_id = value.to_string();
+        let lsn = value.lsn_as_range();
+        LayerDescriptor {
+            key: value.key_range,
+            lsn,
+            is_incremental: false,
+            short_id,
+        }
+    }
+}
+
+impl From<LayerFileName> for LayerDescriptor {
+    fn from(value: LayerFileName) -> Self {
+        match value {
+            LayerFileName::Delta(d) => Self::from(d),
+            LayerFileName::Image(i) => Self::from(i),
         }
     }
 }

pageserver/src/tenant/storage_layer/delta_layer.rs

@@ -37,7 +37,6 @@ use crate::virtual_file::VirtualFile;
 use crate::{walrecord, TEMP_FILE_SUFFIX};
 use crate::{DELTA_FILE_MAGIC, STORAGE_FORMAT_VERSION};
 use anyhow::{bail, ensure, Context, Result};
-use once_cell::sync::OnceCell;
 use pageserver_api::models::{HistoricLayerInfo, LayerAccessKind};
 use rand::{distributions::Alphanumeric, Rng};
 use serde::{Deserialize, Serialize};
@@ -47,6 +46,7 @@ use std::io::{Seek, SeekFrom};
 use std::ops::Range;
 use std::os::unix::fs::FileExt;
 use std::path::{Path, PathBuf};
+use std::sync::{RwLock, RwLockReadGuard, RwLockWriteGuard};
 use tracing::*;
 
 use utils::{
@@ -184,7 +184,7 @@ pub struct DeltaLayer {
 
     access_stats: LayerAccessStats,
 
-    inner: OnceCell<DeltaLayerInner>,
+    inner: RwLock<DeltaLayerInner>,
 }
 
 impl std::fmt::Debug for DeltaLayer {
@@ -201,24 +201,27 @@ impl std::fmt::Debug for DeltaLayer {
 }
 
 pub struct DeltaLayerInner {
+    /// If false, the fields below have not been loaded into memory yet.
+    loaded: bool,
+
     // values copied from summary
     index_start_blk: u32,
     index_root_blk: u32,
 
-    /// Reader object for reading blocks from the file.
-    file: FileBlockReader<VirtualFile>,
+    /// Reader object for reading blocks from the file. (None if not loaded yet)
+    file: Option<FileBlockReader<VirtualFile>>,
 }
 
 impl std::fmt::Debug for DeltaLayerInner {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
         f.debug_struct("DeltaLayerInner")
+            .field("loaded", &self.loaded)
             .field("index_start_blk", &self.index_start_blk)
             .field("index_root_blk", &self.index_root_blk)
             .finish()
     }
 }
 
-#[async_trait::async_trait]
 impl Layer for DeltaLayer {
     /// debugging function to print out the contents of the layer
     fn dump(&self, verbose: bool, ctx: &RequestContext) -> Result<()> {
@@ -243,7 +246,7 @@ impl Layer for DeltaLayer {
             inner.index_start_blk, inner.index_root_blk
         );
 
-        let file = &inner.file;
+        let file = inner.file.as_ref().unwrap();
         let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
             inner.index_start_blk,
             inner.index_root_blk,
@@ -295,13 +298,13 @@ impl Layer for DeltaLayer {
         Ok(())
     }
 
-    fn get_value_reconstruct_data_blocking(
+    fn get_value_reconstruct_data(
         &self,
         key: Key,
         lsn_range: Range<Lsn>,
-        mut reconstruct_state: ValueReconstructState,
-        ctx: RequestContext,
-    ) -> anyhow::Result<(ValueReconstructState, ValueReconstructResult)> {
+        reconstruct_state: &mut ValueReconstructState,
+        ctx: &RequestContext,
+    ) -> anyhow::Result<ValueReconstructResult> {
         ensure!(lsn_range.start >= self.desc.lsn_range.start);
         let mut need_image = true;
 
@@ -309,10 +312,10 @@ impl Layer for DeltaLayer {
 
         {
             // Open the file and lock the metadata in memory
-            let inner = self.load(LayerAccessKind::GetValueReconstructData, &ctx)?;
+            let inner = self.load(LayerAccessKind::GetValueReconstructData, ctx)?;
 
             // Scan the page versions backwards, starting from `lsn`.
-            let file = &inner.file;
+            let file = inner.file.as_ref().unwrap();
             let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
                 inner.index_start_blk,
                 inner.index_root_blk,
@@ -375,9 +378,9 @@ impl Layer for DeltaLayer {
         // If an older page image is needed to reconstruct the page, let the
         // caller know.
         if need_image {
-            Ok((reconstruct_state, ValueReconstructResult::Continue))
+            Ok(ValueReconstructResult::Continue)
         } else {
-            Ok((reconstruct_state, ValueReconstructResult::Complete))
+            Ok(ValueReconstructResult::Complete)
         }
     }
 
@@ -497,22 +500,51 @@ impl DeltaLayer {
     /// Open the underlying file and read the metadata into memory, if it's
     /// not loaded already.
     ///
-    fn load(&self, access_kind: LayerAccessKind, ctx: &RequestContext) -> Result<&DeltaLayerInner> {
+    fn load(
+        &self,
+        access_kind: LayerAccessKind,
+        ctx: &RequestContext,
+    ) -> Result<RwLockReadGuard<DeltaLayerInner>> {
         self.access_stats
             .record_access(access_kind, ctx.task_kind());
-        // Quick exit if already loaded
-        self.inner
-            .get_or_try_init(|| self.load_inner())
-            .with_context(|| format!("Failed to load delta layer {}", self.path().display()))
+        loop {
+            // Quick exit if already loaded
+            let inner = self.inner.read().unwrap();
+            if inner.loaded {
+                return Ok(inner);
+            }
+
+            // Need to open the file and load the metadata. Upgrade our lock to
+            // a write lock. (Or rather, release and re-lock in write mode.)
+            drop(inner);
+            let inner = self.inner.write().unwrap();
+            if !inner.loaded {
+                self.load_inner(inner).with_context(|| {
+                    format!("Failed to load delta layer {}", self.path().display())
+                })?;
+            } else {
+                // Another thread loaded it while we were not holding the lock.
+            }
+
+            // We now have the file open and loaded. There's no function to do
+            // that in the std library RwLock, so we have to release and re-lock
+            // in read mode. (To be precise, the lock guard was moved in the
+            // above call to `load_inner`, so it's already been released). And
+            // while we do that, another thread could unload again, so we have
+            // to re-check and retry if that happens.
+        }
     }
 
-    fn load_inner(&self) -> Result<DeltaLayerInner> {
+    fn load_inner(&self, mut inner: RwLockWriteGuard<DeltaLayerInner>) -> Result<()> {
         let path = self.path();
 
-        let file = VirtualFile::open(&path)
-            .with_context(|| format!("Failed to open file '{}'", path.display()))?;
-        let file = FileBlockReader::new(file);
-
+        // Open the file if it's not open already.
+        if inner.file.is_none() {
+            let file = VirtualFile::open(&path)
+                .with_context(|| format!("Failed to open file '{}'", path.display()))?;
+            inner.file = Some(FileBlockReader::new(file));
+        }
+        let file = inner.file.as_mut().unwrap();
         let summary_blk = file.read_blk(0)?;
         let actual_summary = Summary::des_prefix(summary_blk.as_ref())?;
 
@@ -539,13 +571,13 @@ impl DeltaLayer {
             }
         }
 
+        inner.index_start_blk = actual_summary.index_start_blk;
+        inner.index_root_blk = actual_summary.index_root_blk;
+
         debug!("loaded from {}", &path.display());
 
-        Ok(DeltaLayerInner {
-            file,
-            index_start_blk: actual_summary.index_start_blk,
-            index_root_blk: actual_summary.index_root_blk,
-        })
+        inner.loaded = true;
+        Ok(())
     }
 
     /// Create a DeltaLayer struct representing an existing file on disk.
@@ -567,7 +599,12 @@ impl DeltaLayer {
                 file_size,
             ),
             access_stats,
-            inner: once_cell::sync::OnceCell::new(),
+            inner: RwLock::new(DeltaLayerInner {
+                loaded: false,
+                file: None,
+                index_start_blk: 0,
+                index_root_blk: 0,
+            }),
         }
     }
 
@@ -594,7 +631,12 @@ impl DeltaLayer {
                 metadata.len(),
             ),
             access_stats: LayerAccessStats::empty_will_record_residence_event_later(),
-            inner: once_cell::sync::OnceCell::new(),
+            inner: RwLock::new(DeltaLayerInner {
+                loaded: false,
+                file: None,
+                index_start_blk: 0,
+                index_root_blk: 0,
+            }),
         })
     }
 
@@ -758,7 +800,12 @@ impl DeltaLayerWriterInner {
                 metadata.len(),
             ),
             access_stats: LayerAccessStats::empty_will_record_residence_event_later(),
-            inner: once_cell::sync::OnceCell::new(),
+            inner: RwLock::new(DeltaLayerInner {
+                loaded: false,
+                file: None,
+                index_start_blk,
+                index_root_blk,
+            }),
         };
 
         // fsync the file
@@ -870,7 +917,7 @@ impl Drop for DeltaLayerWriter {
     fn drop(&mut self) {
         if let Some(inner) = self.inner.take() {
             match inner.blob_writer.into_inner().into_inner() {
-                Ok(vfile) => vfile.remove(),
+                Ok(vfile) => vfile.remove().unwrap(),
                 Err(err) => warn!(
                     "error while flushing buffer of image layer temporary file: {}",
                     err
@@ -893,13 +940,13 @@ struct DeltaValueIter<'a> {
     reader: BlockCursor<Adapter<'a>>,
 }
 
-struct Adapter<'a>(&'a DeltaLayerInner);
+struct Adapter<'a>(RwLockReadGuard<'a, DeltaLayerInner>);
 
 impl<'a> BlockReader for Adapter<'a> {
     type BlockLease = PageReadGuard<'static>;
 
     fn read_blk(&self, blknum: u32) -> Result<Self::BlockLease, std::io::Error> {
-        self.0.file.read_blk(blknum)
+        self.0.file.as_ref().unwrap().read_blk(blknum)
     }
 }
 
@@ -912,8 +959,8 @@ impl<'a> Iterator for DeltaValueIter<'a> {
 }
 
 impl<'a> DeltaValueIter<'a> {
-    fn new(inner: &'a DeltaLayerInner) -> Result<Self> {
-        let file = &inner.file;
+    fn new(inner: RwLockReadGuard<'a, DeltaLayerInner>) -> Result<Self> {
+        let file = inner.file.as_ref().unwrap();
         let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
             inner.index_start_blk,
             inner.index_root_blk,
@@ -986,8 +1033,8 @@ impl Iterator for DeltaKeyIter {
 }
 
 impl<'a> DeltaKeyIter {
-    fn new(inner: &'a DeltaLayerInner) -> Result<Self> {
-        let file = &inner.file;
+    fn new(inner: RwLockReadGuard<'a, DeltaLayerInner>) -> Result<Self> {
+        let file = inner.file.as_ref().unwrap();
         let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
             inner.index_start_blk,
             inner.index_root_blk,
@@ -1027,21 +1074,3 @@ impl<'a> DeltaKeyIter {
         Ok(iter)
     }
 }
-
-#[cfg(test)]
-mod test {
-    use super::DeltaKeyIter;
-    use super::DeltaLayer;
-    use super::DeltaValueIter;
-
-    // We will soon need the iters to be send in the compaction code.
-    // Cf https://github.com/neondatabase/neon/pull/4462#issuecomment-1587398883
-    // Cf https://github.com/neondatabase/neon/issues/4471
-    #[test]
-    fn is_send() {
-        fn assert_send<T: Send>() {}
-        assert_send::<DeltaLayer>();
-        assert_send::<DeltaValueIter>();
-        assert_send::<DeltaKeyIter>();
-    }
-}

pageserver/src/tenant/storage_layer/image_layer.rs

@@ -149,7 +149,6 @@ impl std::fmt::Debug for ImageLayerInner {
     }
 }
 
-#[async_trait::async_trait]
 impl Layer for ImageLayer {
     /// debugging function to print out the contents of the layer
     fn dump(&self, verbose: bool, ctx: &RequestContext) -> Result<()> {
@@ -182,18 +181,18 @@ impl Layer for ImageLayer {
     }
 
     /// Look up given page in the file
-    fn get_value_reconstruct_data_blocking(
+    fn get_value_reconstruct_data(
         &self,
         key: Key,
         lsn_range: Range<Lsn>,
-        mut reconstruct_state: ValueReconstructState,
-        ctx: RequestContext,
-    ) -> anyhow::Result<(ValueReconstructState, ValueReconstructResult)> {
+        reconstruct_state: &mut ValueReconstructState,
+        ctx: &RequestContext,
+    ) -> anyhow::Result<ValueReconstructResult> {
         assert!(self.desc.key_range.contains(&key));
         assert!(lsn_range.start >= self.lsn);
         assert!(lsn_range.end >= self.lsn);
 
-        let inner = self.load(LayerAccessKind::GetValueReconstructData, &ctx)?;
+        let inner = self.load(LayerAccessKind::GetValueReconstructData, ctx)?;
 
         let file = inner.file.as_ref().unwrap();
         let tree_reader = DiskBtreeReader::new(inner.index_start_blk, inner.index_root_blk, file);
@@ -211,9 +210,9 @@ impl Layer for ImageLayer {
             let value = Bytes::from(blob);
 
             reconstruct_state.img = Some((self.lsn, value));
-            Ok((reconstruct_state, ValueReconstructResult::Complete))
+            Ok(ValueReconstructResult::Complete)
         } else {
-            Ok((reconstruct_state, ValueReconstructResult::Missing))
+            Ok(ValueReconstructResult::Missing)
         }
     }
 
@@ -710,7 +709,7 @@ impl ImageLayerWriter {
 impl Drop for ImageLayerWriter {
     fn drop(&mut self) {
         if let Some(inner) = self.inner.take() {
-            inner.blob_writer.into_inner().remove();
+            inner.blob_writer.into_inner().remove().unwrap();
         }
     }
 }

pageserver/src/tenant/storage_layer/inmemory_layer.rs

@@ -110,7 +110,6 @@ impl InMemoryLayer {
     }
 }
 
-#[async_trait::async_trait]
 impl Layer for InMemoryLayer {
     fn get_key_range(&self) -> Range<Key> {
         Key::MIN..Key::MAX
@@ -191,13 +190,13 @@ impl Layer for InMemoryLayer {
     }
 
     /// Look up given value in the layer.
-    fn get_value_reconstruct_data_blocking(
+    fn get_value_reconstruct_data(
         &self,
         key: Key,
         lsn_range: Range<Lsn>,
-        mut reconstruct_state: ValueReconstructState,
-        _ctx: RequestContext,
-    ) -> anyhow::Result<(ValueReconstructState, ValueReconstructResult)> {
+        reconstruct_state: &mut ValueReconstructState,
+        _ctx: &RequestContext,
+    ) -> anyhow::Result<ValueReconstructResult> {
         ensure!(lsn_range.start >= self.start_lsn);
         let mut need_image = true;
 
@@ -214,7 +213,7 @@ impl Layer for InMemoryLayer {
                 match value {
                     Value::Image(img) => {
                         reconstruct_state.img = Some((*entry_lsn, img));
-                        return Ok((reconstruct_state, ValueReconstructResult::Complete));
+                        return Ok(ValueReconstructResult::Complete);
                     }
                     Value::WalRecord(rec) => {
                         let will_init = rec.will_init();
@@ -234,9 +233,9 @@ impl Layer for InMemoryLayer {
         // If an older page image is needed to reconstruct the page, let the
         // caller know.
         if need_image {
-            Ok((reconstruct_state, ValueReconstructResult::Continue))
+            Ok(ValueReconstructResult::Continue)
         } else {
-            Ok((reconstruct_state, ValueReconstructResult::Complete))
+            Ok(ValueReconstructResult::Complete)
         }
     }
 }
@@ -305,7 +304,7 @@ impl InMemoryLayer {
         Ok(())
     }
 
-    pub async fn put_tombstone(&self, _key_range: Range<Key>, _lsn: Lsn) -> Result<()> {
+    pub fn put_tombstone(&self, _key_range: Range<Key>, _lsn: Lsn) -> Result<()> {
         // TODO: Currently, we just leak the storage for any deleted keys
 
         Ok(())

pageserver/src/tenant/storage_layer/remote_layer.rs

@@ -6,7 +6,7 @@ use crate::context::RequestContext;
 use crate::repository::Key;
 use crate::tenant::layer_map::BatchedUpdates;
 use crate::tenant::remote_timeline_client::index::LayerFileMetadata;
-use crate::tenant::storage_layer::{Layer, ValueReconstructState};
+use crate::tenant::storage_layer::{Layer, ValueReconstructResult, ValueReconstructState};
 use anyhow::{bail, Result};
 use pageserver_api::models::HistoricLayerInfo;
 use std::ops::Range;
@@ -21,7 +21,7 @@ use utils::{
 use super::filename::{DeltaFileName, ImageFileName};
 use super::{
     DeltaLayer, ImageLayer, LayerAccessStats, LayerAccessStatsReset, LayerIter, LayerKeyIter,
-    LayerResidenceStatus, PersistentLayer, PersistentLayerDesc, ValueReconstructResult,
+    LayerResidenceStatus, PersistentLayer, PersistentLayerDesc,
 };
 
 /// RemoteLayer is a not yet downloaded [`ImageLayer`] or
@@ -63,15 +63,14 @@ impl std::fmt::Debug for RemoteLayer {
     }
 }
 
-#[async_trait::async_trait]
 impl Layer for RemoteLayer {
-    fn get_value_reconstruct_data_blocking(
+    fn get_value_reconstruct_data(
         &self,
         _key: Key,
         _lsn_range: Range<Lsn>,
-        _reconstruct_state: ValueReconstructState,
-        _ctx: RequestContext,
-    ) -> Result<(ValueReconstructState, ValueReconstructResult)> {
+        _reconstruct_state: &mut ValueReconstructState,
+        _ctx: &RequestContext,
+    ) -> Result<ValueReconstructResult> {
         bail!(
             "layer {} needs to be downloaded",
             self.filename().file_name()
@@ -219,12 +218,15 @@ impl RemoteLayer {
     }
 
     /// Create a Layer struct representing this layer, after it has been downloaded.
-    pub fn create_downloaded_layer(
+    pub fn create_downloaded_layer<L>(
         &self,
-        layer_map_lock_held_witness: &BatchedUpdates<'_>,
+        layer_map_lock_held_witness: &BatchedUpdates<'_, L>,
         conf: &'static PageServerConf,
         file_size: u64,
-    ) -> Arc<dyn PersistentLayer> {
+    ) -> Arc<dyn PersistentLayer>
+    where
+        L: ?Sized + Layer,
+    {
         if self.desc.is_delta {
             let fname = self.desc.delta_file_name();
             Arc::new(DeltaLayer::new(

pageserver/src/tenant/timeline/eviction_task.rs

@@ -197,11 +197,9 @@ impl Timeline {
         // We don't want to hold the layer map lock during eviction.
         // So, we just need to deal with this.
         let candidates: Vec<Arc<dyn PersistentLayer>> = {
-            let guard = self.layers.read().await;
-            let (layers, mapping) = &*guard;
+            let layers = self.layers.read().unwrap();
             let mut candidates = Vec::new();
             for hist_layer in layers.iter_historic_layers() {
-                let hist_layer = mapping.get_from_desc(&hist_layer);
                 if hist_layer.is_remote_layer() {
                     continue;
                 }

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

@@ -1321,11 +1321,10 @@ mod tests {
 
     const DUMMY_SAFEKEEPER_HOST: &str = "safekeeper_connstr";
 
-    async fn dummy_state(harness: &TenantHarness) -> ConnectionManagerState {
+    async fn dummy_state(harness: &TenantHarness<'_>) -> ConnectionManagerState {
         let (tenant, ctx) = harness.load().await;
         let timeline = tenant
             .create_test_timeline(TIMELINE_ID, Lsn(0x8), crate::DEFAULT_PG_VERSION, &ctx)
-            .await
             .expect("Failed to create an empty timeline for dummy wal connection manager");
 
         ConnectionManagerState {

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

@@ -304,15 +304,12 @@ pub(super) async fn handle_walreceiver_connection(
             }
         }
 
-        timeline
-            .check_checkpoint_distance()
-            .await
-            .with_context(|| {
-                format!(
-                    "Failed to check checkpoint distance for timeline {}",
-                    timeline.timeline_id
-                )
-            })?;
+        timeline.check_checkpoint_distance().with_context(|| {
+            format!(
+                "Failed to check checkpoint distance for timeline {}",
+                timeline.timeline_id
+            )
+        })?;
 
         if let Some(last_lsn) = status_update {
             let timeline_remote_consistent_lsn =

pageserver/src/virtual_file.rs

@@ -324,16 +324,8 @@ impl VirtualFile {
         Ok(result)
     }
 
-    pub fn remove(self) {
-        let path = self.path.clone();
-        drop(self);
-        std::fs::remove_file(path).expect("failed to remove the virtual file");
-    }
-}
-
-impl Drop for VirtualFile {
-    /// If a VirtualFile is dropped, close the underlying file if it was open.
-    fn drop(&mut self) {
+    /// Idempotently close the file descriptor we might have or have not open for this VirtualFile.
+    pub fn close(&mut self) {
         let handle = self.handle.get_mut().unwrap();
 
         // We could check with a read-lock first, to avoid waiting on an
@@ -351,6 +343,26 @@ impl Drop for VirtualFile {
                 .observe_closure_duration(|| slot_guard.file.take());
         }
     }
+
+    /// Caller can retry if we return an `Err`.
+    #[allow(clippy::result_large_err)]
+    pub fn remove(mut self) -> Result<(), (Self, std::io::Error)> {
+        // close our fd before unlink system call, so that the unlink actually performs the removal
+        self.close();
+        // Try to remove file on disk.
+        // If it fails, we idempotently closed the fd, but the caller can choose to retry.
+        match std::fs::remove_file(&self.path) {
+            Ok(()) => Ok(()),
+            Err(e) => Err((self, e)),
+        }
+    }
+}
+
+impl Drop for VirtualFile {
+    /// If a VirtualFile is dropped, close the underlying file if it was open.
+    fn drop(&mut self) {
+        self.close();
+    }
 }
 
 impl Read for VirtualFile {

pageserver/src/walingest.rs

@@ -25,7 +25,7 @@ use postgres_ffi::v14::nonrelfile_utils::clogpage_precedes;
 use postgres_ffi::v14::nonrelfile_utils::slru_may_delete_clogsegment;
 use postgres_ffi::{fsm_logical_to_physical, page_is_new, page_set_lsn};
 
-use anyhow::{Context, Result};
+use anyhow::Result;
 use bytes::{Buf, Bytes, BytesMut};
 use tracing::*;
 
@@ -333,7 +333,7 @@ impl<'a> WalIngest<'a> {
 
         // Now that this record has been fully handled, including updating the
         // checkpoint data, let the repository know that it is up-to-date to this LSN
-        modification.commit().await?;
+        modification.commit()?;
 
         Ok(())
     }
@@ -1082,10 +1082,7 @@ impl<'a> WalIngest<'a> {
             .await?
         {
             // create it with 0 size initially, the logic below will extend it
-            modification
-                .put_rel_creation(rel, 0, ctx)
-                .await
-                .context("Relation Error")?;
+            modification.put_rel_creation(rel, 0, ctx).await?;
             0
         } else {
             self.timeline.get_rel_size(rel, last_lsn, true, ctx).await?
@@ -1174,6 +1171,7 @@ impl<'a> WalIngest<'a> {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use crate::pgdatadir_mapping::create_test_timeline;
     use crate::tenant::harness::*;
     use crate::tenant::Timeline;
     use postgres_ffi::v14::xlog_utils::SIZEOF_CHECKPOINT;
@@ -1202,7 +1200,7 @@ mod tests {
         let mut m = tline.begin_modification(Lsn(0x10));
         m.put_checkpoint(ZERO_CHECKPOINT.clone())?;
         m.put_relmap_file(0, 111, Bytes::from(""), ctx).await?; // dummy relmapper file
-        m.commit().await?;
+        m.commit()?;
         let walingest = WalIngest::new(tline, Lsn(0x10), ctx).await?;
 
         Ok(walingest)
@@ -1210,10 +1208,9 @@ mod tests {
 
     #[tokio::test]
     async fn test_relsize() -> Result<()> {
-        let (tenant, ctx) = TenantHarness::create("test_relsize")?.load().await;
-        let tline = tenant
-            .create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)
-            .await?;
+        let harness = TenantHarness::create("test_relsize")?;
+        let (tenant, ctx) = harness.load().await;
+        let tline = create_test_timeline(&tenant, TIMELINE_ID, DEFAULT_PG_VERSION, &ctx)?;
         let mut walingest = init_walingest_test(&tline, &ctx).await?;
 
         let mut m = tline.begin_modification(Lsn(0x20));
@@ -1221,22 +1218,22 @@ mod tests {
         walingest
             .put_rel_page_image(&mut m, TESTREL_A, 0, TEST_IMG("foo blk 0 at 2"), &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
         let mut m = tline.begin_modification(Lsn(0x30));
         walingest
             .put_rel_page_image(&mut m, TESTREL_A, 0, TEST_IMG("foo blk 0 at 3"), &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
         let mut m = tline.begin_modification(Lsn(0x40));
         walingest
             .put_rel_page_image(&mut m, TESTREL_A, 1, TEST_IMG("foo blk 1 at 4"), &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
         let mut m = tline.begin_modification(Lsn(0x50));
         walingest
             .put_rel_page_image(&mut m, TESTREL_A, 2, TEST_IMG("foo blk 2 at 5"), &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
 
         assert_current_logical_size(&tline, Lsn(0x50));
 
@@ -1322,7 +1319,7 @@ mod tests {
         walingest
             .put_rel_truncation(&mut m, TESTREL_A, 2, &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
         assert_current_logical_size(&tline, Lsn(0x60));
 
         // Check reported size and contents after truncation
@@ -1364,7 +1361,7 @@ mod tests {
         walingest
             .put_rel_truncation(&mut m, TESTREL_A, 0, &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
         assert_eq!(
             tline
                 .get_rel_size(TESTREL_A, Lsn(0x68), false, &ctx)
@@ -1377,7 +1374,7 @@ mod tests {
         walingest
             .put_rel_page_image(&mut m, TESTREL_A, 1, TEST_IMG("foo blk 1"), &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
         assert_eq!(
             tline
                 .get_rel_size(TESTREL_A, Lsn(0x70), false, &ctx)
@@ -1402,7 +1399,7 @@ mod tests {
         walingest
             .put_rel_page_image(&mut m, TESTREL_A, 1500, TEST_IMG("foo blk 1500"), &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
         assert_eq!(
             tline
                 .get_rel_size(TESTREL_A, Lsn(0x80), false, &ctx)
@@ -1431,17 +1428,16 @@ mod tests {
     // and then created it again within the same layer.
     #[tokio::test]
     async fn test_drop_extend() -> Result<()> {
-        let (tenant, ctx) = TenantHarness::create("test_drop_extend")?.load().await;
-        let tline = tenant
-            .create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)
-            .await?;
+        let harness = TenantHarness::create("test_drop_extend")?;
+        let (tenant, ctx) = harness.load().await;
+        let tline = create_test_timeline(&tenant, TIMELINE_ID, DEFAULT_PG_VERSION, &ctx)?;
         let mut walingest = init_walingest_test(&tline, &ctx).await?;
 
         let mut m = tline.begin_modification(Lsn(0x20));
         walingest
             .put_rel_page_image(&mut m, TESTREL_A, 0, TEST_IMG("foo blk 0 at 2"), &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
 
         // Check that rel exists and size is correct
         assert_eq!(
@@ -1460,7 +1456,7 @@ mod tests {
         // Drop rel
         let mut m = tline.begin_modification(Lsn(0x30));
         walingest.put_rel_drop(&mut m, TESTREL_A, &ctx).await?;
-        m.commit().await?;
+        m.commit()?;
 
         // Check that rel is not visible anymore
         assert_eq!(
@@ -1478,7 +1474,7 @@ mod tests {
         walingest
             .put_rel_page_image(&mut m, TESTREL_A, 0, TEST_IMG("foo blk 0 at 4"), &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
 
         // Check that rel exists and size is correct
         assert_eq!(
@@ -1502,10 +1498,9 @@ mod tests {
     // and then extended it again within the same layer.
     #[tokio::test]
     async fn test_truncate_extend() -> Result<()> {
-        let (tenant, ctx) = TenantHarness::create("test_truncate_extend")?.load().await;
-        let tline = tenant
-            .create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)
-            .await?;
+        let harness = TenantHarness::create("test_truncate_extend")?;
+        let (tenant, ctx) = harness.load().await;
+        let tline = create_test_timeline(&tenant, TIMELINE_ID, DEFAULT_PG_VERSION, &ctx)?;
         let mut walingest = init_walingest_test(&tline, &ctx).await?;
 
         // Create a 20 MB relation (the size is arbitrary)
@@ -1517,7 +1512,7 @@ mod tests {
                 .put_rel_page_image(&mut m, TESTREL_A, blkno, TEST_IMG(&data), &ctx)
                 .await?;
         }
-        m.commit().await?;
+        m.commit()?;
 
         // The relation was created at LSN 20, not visible at LSN 1 yet.
         assert_eq!(
@@ -1562,7 +1557,7 @@ mod tests {
         walingest
             .put_rel_truncation(&mut m, TESTREL_A, 1, &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
 
         // Check reported size and contents after truncation
         assert_eq!(
@@ -1611,7 +1606,7 @@ mod tests {
                 .put_rel_page_image(&mut m, TESTREL_A, blkno, TEST_IMG(&data), &ctx)
                 .await?;
         }
-        m.commit().await?;
+        m.commit()?;
 
         assert_eq!(
             tline
@@ -1644,10 +1639,9 @@ mod tests {
     /// split into multiple 1 GB segments in Postgres.
     #[tokio::test]
     async fn test_large_rel() -> Result<()> {
-        let (tenant, ctx) = TenantHarness::create("test_large_rel")?.load().await;
-        let tline = tenant
-            .create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)
-            .await?;
+        let harness = TenantHarness::create("test_large_rel")?;
+        let (tenant, ctx) = harness.load().await;
+        let tline = create_test_timeline(&tenant, TIMELINE_ID, DEFAULT_PG_VERSION, &ctx)?;
         let mut walingest = init_walingest_test(&tline, &ctx).await?;
 
         let mut lsn = 0x10;
@@ -1658,7 +1652,7 @@ mod tests {
             walingest
                 .put_rel_page_image(&mut m, TESTREL_A, blknum as BlockNumber, img, &ctx)
                 .await?;
-            m.commit().await?;
+            m.commit()?;
         }
 
         assert_current_logical_size(&tline, Lsn(lsn));
@@ -1674,7 +1668,7 @@ mod tests {
         walingest
             .put_rel_truncation(&mut m, TESTREL_A, RELSEG_SIZE, &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
         assert_eq!(
             tline.get_rel_size(TESTREL_A, Lsn(lsn), false, &ctx).await?,
             RELSEG_SIZE
@@ -1687,7 +1681,7 @@ mod tests {
         walingest
             .put_rel_truncation(&mut m, TESTREL_A, RELSEG_SIZE - 1, &ctx)
             .await?;
-        m.commit().await?;
+        m.commit()?;
         assert_eq!(
             tline.get_rel_size(TESTREL_A, Lsn(lsn), false, &ctx).await?,
             RELSEG_SIZE - 1
@@ -1703,7 +1697,7 @@ mod tests {
             walingest
                 .put_rel_truncation(&mut m, TESTREL_A, size as BlockNumber, &ctx)
                 .await?;
-            m.commit().await?;
+            m.commit()?;
             assert_eq!(
                 tline.get_rel_size(TESTREL_A, Lsn(lsn), false, &ctx).await?,
                 size as BlockNumber

pgxn/neon/pagestore_smgr.c

@@ -2675,6 +2675,7 @@ bool
 neon_redo_read_buffer_filter(XLogReaderState *record, uint8 block_id)
 {
 	XLogRecPtr	end_recptr = record->EndRecPtr;
+	XLogRecPtr	prev_end_recptr = record->ReadRecPtr - 1;
 	RelFileNode	rnode;
 	ForkNumber	forknum;
 	BlockNumber	blkno;
@@ -2718,15 +2719,16 @@ neon_redo_read_buffer_filter(XLogReaderState *record, uint8 block_id)
 
 	no_redo_needed = buffer < 0;
 
-	/* In both cases st lwlsn past this WAL record */
-	SetLastWrittenLSNForBlock(end_recptr, rnode, forknum, blkno);
-
-	/* we don't have the buffer in memory, update lwLsn past this record,
-	 * also evict page fro file cache
-	 */
+	/* we don't have the buffer in memory, update lwLsn past this record */
 	if (no_redo_needed)
+	{
+		SetLastWrittenLSNForBlock(end_recptr, rnode, forknum, blkno);
 		lfc_evict(rnode, forknum, blkno);
-
+	}
+	else
+	{
+		SetLastWrittenLSNForBlock(prev_end_recptr, rnode, forknum, blkno);
+	}
 
 	LWLockRelease(partitionLock);
 
@@ -2734,10 +2736,7 @@ neon_redo_read_buffer_filter(XLogReaderState *record, uint8 block_id)
 	if (get_cached_relsize(rnode, forknum, &relsize))
 	{
 		if (relsize < blkno + 1)
-		{
 			update_cached_relsize(rnode, forknum, blkno + 1);
-			SetLastWrittenLSNForRelation(end_recptr, rnode, forknum);
-		}
 	}
 	else
 	{
@@ -2769,7 +2768,6 @@ neon_redo_read_buffer_filter(XLogReaderState *record, uint8 block_id)
 		Assert(nbresponse->n_blocks > blkno);
 
 		set_cached_relsize(rnode, forknum, nbresponse->n_blocks);
-		SetLastWrittenLSNForRelation(end_recptr, rnode, forknum);
 
 		elog(SmgrTrace, "Set length to %d", nbresponse->n_blocks);
 	}

pgxn/neon/walproposer.c

@@ -257,7 +257,7 @@ nwp_register_gucs(void)
 							"Walproposer reconnects to offline safekeepers once in this interval.",
 							NULL,
 							&wal_acceptor_reconnect_timeout,
-							1000, 0, INT_MAX,	/* default, min, max */
+							5000, 0, INT_MAX,	/* default, min, max */
 							PGC_SIGHUP, /* context */
 							GUC_UNIT_MS,	/* flags */
 							NULL, NULL, NULL);

poetry.lock

@@ -1,4 +1,4 @@
-# This file is automatically @generated by Poetry and should not be changed by hand.
+# This file is automatically @generated by Poetry 1.4.2 and should not be changed by hand.
 
 [[package]]
 name = "aiohttp"
@@ -855,31 +855,35 @@ files = [
 
 [[package]]
 name = "cryptography"
-version = "41.0.0"
+version = "39.0.1"
 description = "cryptography is a package which provides cryptographic recipes and primitives to Python developers."
 category = "main"
 optional = false
-python-versions = ">=3.7"
+python-versions = ">=3.6"
 files = [
-    {file = "cryptography-41.0.0-cp37-abi3-macosx_10_12_universal2.whl", hash = "sha256:3c5ef25d060c80d6d9f7f9892e1d41bb1c79b78ce74805b8cb4aa373cb7d5ec8"},
-    {file = "cryptography-41.0.0-cp37-abi3-macosx_10_12_x86_64.whl", hash = "sha256:8362565b3835ceacf4dc8f3b56471a2289cf51ac80946f9087e66dc283a810e0"},
-    {file = "cryptography-41.0.0-cp37-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:3680248309d340fda9611498a5319b0193a8dbdb73586a1acf8109d06f25b92d"},
-    {file = "cryptography-41.0.0-cp37-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:84a165379cb9d411d58ed739e4af3396e544eac190805a54ba2e0322feb55c46"},
-    {file = "cryptography-41.0.0-cp37-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:4ab14d567f7bbe7f1cdff1c53d5324ed4d3fc8bd17c481b395db224fb405c237"},
-    {file = "cryptography-41.0.0-cp37-abi3-manylinux_2_28_x86_64.whl", hash = "sha256:9f65e842cb02550fac96536edb1d17f24c0a338fd84eaf582be25926e993dde4"},
-    {file = "cryptography-41.0.0-cp37-abi3-musllinux_1_1_aarch64.whl", hash = "sha256:b7f2f5c525a642cecad24ee8670443ba27ac1fab81bba4cc24c7b6b41f2d0c75"},
-    {file = "cryptography-41.0.0-cp37-abi3-musllinux_1_1_x86_64.whl", hash = "sha256:7d92f0248d38faa411d17f4107fc0bce0c42cae0b0ba5415505df72d751bf62d"},
-    {file = "cryptography-41.0.0-cp37-abi3-win32.whl", hash = "sha256:34d405ea69a8b34566ba3dfb0521379b210ea5d560fafedf9f800a9a94a41928"},
-    {file = "cryptography-41.0.0-cp37-abi3-win_amd64.whl", hash = "sha256:344c6de9f8bda3c425b3a41b319522ba3208551b70c2ae00099c205f0d9fd3be"},
-    {file = "cryptography-41.0.0-pp38-pypy38_pp73-macosx_10_12_x86_64.whl", hash = "sha256:88ff107f211ea696455ea8d911389f6d2b276aabf3231bf72c8853d22db755c5"},
-    {file = "cryptography-41.0.0-pp38-pypy38_pp73-manylinux_2_28_aarch64.whl", hash = "sha256:b846d59a8d5a9ba87e2c3d757ca019fa576793e8758174d3868aecb88d6fc8eb"},
-    {file = "cryptography-41.0.0-pp38-pypy38_pp73-manylinux_2_28_x86_64.whl", hash = "sha256:f5d0bf9b252f30a31664b6f64432b4730bb7038339bd18b1fafe129cfc2be9be"},
-    {file = "cryptography-41.0.0-pp38-pypy38_pp73-win_amd64.whl", hash = "sha256:5c1f7293c31ebc72163a9a0df246f890d65f66b4a40d9ec80081969ba8c78cc9"},
-    {file = "cryptography-41.0.0-pp39-pypy39_pp73-macosx_10_12_x86_64.whl", hash = "sha256:bf8fc66012ca857d62f6a347007e166ed59c0bc150cefa49f28376ebe7d992a2"},
-    {file = "cryptography-41.0.0-pp39-pypy39_pp73-manylinux_2_28_aarch64.whl", hash = "sha256:a4fc68d1c5b951cfb72dfd54702afdbbf0fb7acdc9b7dc4301bbf2225a27714d"},
-    {file = "cryptography-41.0.0-pp39-pypy39_pp73-manylinux_2_28_x86_64.whl", hash = "sha256:14754bcdae909d66ff24b7b5f166d69340ccc6cb15731670435efd5719294895"},
-    {file = "cryptography-41.0.0-pp39-pypy39_pp73-win_amd64.whl", hash = "sha256:0ddaee209d1cf1f180f1efa338a68c4621154de0afaef92b89486f5f96047c55"},
-    {file = "cryptography-41.0.0.tar.gz", hash = "sha256:6b71f64beeea341c9b4f963b48ee3b62d62d57ba93eb120e1196b31dc1025e78"},
+    {file = "cryptography-39.0.1-cp36-abi3-macosx_10_12_universal2.whl", hash = "sha256:6687ef6d0a6497e2b58e7c5b852b53f62142cfa7cd1555795758934da363a965"},
+    {file = "cryptography-39.0.1-cp36-abi3-macosx_10_12_x86_64.whl", hash = "sha256:706843b48f9a3f9b9911979761c91541e3d90db1ca905fd63fee540a217698bc"},
+    {file = "cryptography-39.0.1-cp36-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.manylinux_2_24_aarch64.whl", hash = "sha256:5d2d8b87a490bfcd407ed9d49093793d0f75198a35e6eb1a923ce1ee86c62b41"},
+    {file = "cryptography-39.0.1-cp36-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:83e17b26de248c33f3acffb922748151d71827d6021d98c70e6c1a25ddd78505"},
+    {file = "cryptography-39.0.1-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:e124352fd3db36a9d4a21c1aa27fd5d051e621845cb87fb851c08f4f75ce8be6"},
+    {file = "cryptography-39.0.1-cp36-abi3-manylinux_2_24_x86_64.whl", hash = "sha256:5aa67414fcdfa22cf052e640cb5ddc461924a045cacf325cd164e65312d99502"},
+    {file = "cryptography-39.0.1-cp36-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:35f7c7d015d474f4011e859e93e789c87d21f6f4880ebdc29896a60403328f1f"},
+    {file = "cryptography-39.0.1-cp36-abi3-manylinux_2_28_x86_64.whl", hash = "sha256:f24077a3b5298a5a06a8e0536e3ea9ec60e4c7ac486755e5fb6e6ea9b3500106"},
+    {file = "cryptography-39.0.1-cp36-abi3-musllinux_1_1_aarch64.whl", hash = "sha256:f0c64d1bd842ca2633e74a1a28033d139368ad959872533b1bab8c80e8240a0c"},
+    {file = "cryptography-39.0.1-cp36-abi3-musllinux_1_1_x86_64.whl", hash = "sha256:0f8da300b5c8af9f98111ffd512910bc792b4c77392a9523624680f7956a99d4"},
+    {file = "cryptography-39.0.1-cp36-abi3-win32.whl", hash = "sha256:fe913f20024eb2cb2f323e42a64bdf2911bb9738a15dba7d3cce48151034e3a8"},
+    {file = "cryptography-39.0.1-cp36-abi3-win_amd64.whl", hash = "sha256:ced4e447ae29ca194449a3f1ce132ded8fcab06971ef5f618605aacaa612beac"},
+    {file = "cryptography-39.0.1-pp38-pypy38_pp73-macosx_10_12_x86_64.whl", hash = "sha256:807ce09d4434881ca3a7594733669bd834f5b2c6d5c7e36f8c00f691887042ad"},
+    {file = "cryptography-39.0.1-pp38-pypy38_pp73-manylinux_2_28_aarch64.whl", hash = "sha256:c5caeb8188c24888c90b5108a441c106f7faa4c4c075a2bcae438c6e8ca73cef"},
+    {file = "cryptography-39.0.1-pp38-pypy38_pp73-manylinux_2_28_x86_64.whl", hash = "sha256:4789d1e3e257965e960232345002262ede4d094d1a19f4d3b52e48d4d8f3b885"},
+    {file = "cryptography-39.0.1-pp38-pypy38_pp73-win_amd64.whl", hash = "sha256:96f1157a7c08b5b189b16b47bc9db2332269d6680a196341bf30046330d15388"},
+    {file = "cryptography-39.0.1-pp39-pypy39_pp73-macosx_10_12_x86_64.whl", hash = "sha256:e422abdec8b5fa8462aa016786680720d78bdce7a30c652b7fadf83a4ba35336"},
+    {file = "cryptography-39.0.1-pp39-pypy39_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.manylinux_2_24_aarch64.whl", hash = "sha256:b0afd054cd42f3d213bf82c629efb1ee5f22eba35bf0eec88ea9ea7304f511a2"},
+    {file = "cryptography-39.0.1-pp39-pypy39_pp73-manylinux_2_24_x86_64.whl", hash = "sha256:6f8ba7f0328b79f08bdacc3e4e66fb4d7aab0c3584e0bd41328dce5262e26b2e"},
+    {file = "cryptography-39.0.1-pp39-pypy39_pp73-manylinux_2_28_aarch64.whl", hash = "sha256:ef8b72fa70b348724ff1218267e7f7375b8de4e8194d1636ee60510aae104cd0"},
+    {file = "cryptography-39.0.1-pp39-pypy39_pp73-manylinux_2_28_x86_64.whl", hash = "sha256:aec5a6c9864be7df2240c382740fcf3b96928c46604eaa7f3091f58b878c0bb6"},
+    {file = "cryptography-39.0.1-pp39-pypy39_pp73-win_amd64.whl", hash = "sha256:fdd188c8a6ef8769f148f88f859884507b954cc64db6b52f66ef199bb9ad660a"},
+    {file = "cryptography-39.0.1.tar.gz", hash = "sha256:d1f6198ee6d9148405e49887803907fe8962a23e6c6f83ea7d98f1c0de375695"},
 ]
 
 [package.dependencies]
@@ -888,12 +892,12 @@ cffi = ">=1.12"
 [package.extras]
 docs = ["sphinx (>=5.3.0)", "sphinx-rtd-theme (>=1.1.1)"]
 docstest = ["pyenchant (>=1.6.11)", "sphinxcontrib-spelling (>=4.0.1)", "twine (>=1.12.0)"]
-nox = ["nox"]
-pep8test = ["black", "check-sdist", "mypy", "ruff"]
-sdist = ["build"]
+pep8test = ["black", "check-manifest", "mypy", "ruff", "types-pytz", "types-requests"]
+sdist = ["setuptools-rust (>=0.11.4)"]
 ssh = ["bcrypt (>=3.1.5)"]
-test = ["pretend", "pytest (>=6.2.0)", "pytest-benchmark", "pytest-cov", "pytest-xdist"]
+test = ["hypothesis (>=1.11.4,!=3.79.2)", "iso8601", "pretend", "pytest (>=6.2.0)", "pytest-benchmark", "pytest-cov", "pytest-shard (>=0.1.2)", "pytest-subtests", "pytest-xdist", "pytz"]
 test-randomorder = ["pytest-randomly"]
+tox = ["tox"]
 
 [[package]]
 name = "docker"

proxy/src/http/sql_over_http.rs

@@ -1,6 +1,5 @@
 use futures::pin_mut;
 use futures::StreamExt;
-use futures::TryFutureExt;
 use hyper::body::HttpBody;
 use hyper::http::HeaderName;
 use hyper::http::HeaderValue;
@@ -12,13 +11,8 @@ use serde_json::Value;
 use tokio_postgres::types::Kind;
 use tokio_postgres::types::Type;
 use tokio_postgres::Row;
-use tracing::error;
-use tracing::info;
-use tracing::instrument;
 use url::Url;
 
-use crate::proxy::invalidate_cache;
-use crate::proxy::NUM_RETRIES_WAKE_COMPUTE;
 use crate::{auth, config::ProxyConfig, console};
 
 #[derive(serde::Deserialize)]
@@ -96,17 +90,10 @@ fn json_array_to_pg_array(value: &Value) -> Result<Option<String>, serde_json::E
     }
 }
 
-struct ConnInfo {
-    username: String,
-    dbname: String,
-    hostname: String,
-    password: String,
-}
-
 fn get_conn_info(
     headers: &HeaderMap,
     sni_hostname: Option<String>,
-) -> Result<ConnInfo, anyhow::Error> {
+) -> Result<(String, String, String, String), anyhow::Error> {
     let connection_string = headers
         .get("Neon-Connection-String")
         .ok_or(anyhow::anyhow!("missing connection string"))?
@@ -159,12 +146,12 @@ fn get_conn_info(
         }
     }
 
-    Ok(ConnInfo {
-        username: username.to_owned(),
-        dbname: dbname.to_owned(),
-        hostname: hostname.to_owned(),
-        password: password.to_owned(),
-    })
+    Ok((
+        username.to_owned(),
+        dbname.to_owned(),
+        hostname.to_owned(),
+        password.to_owned(),
+    ))
 }
 
 // TODO: return different http error codes
@@ -177,10 +164,10 @@ pub async fn handle(
     // Determine the destination and connection params
     //
     let headers = request.headers();
-    let conn_info = get_conn_info(headers, sni_hostname)?;
+    let (username, dbname, hostname, password) = get_conn_info(headers, sni_hostname)?;
     let credential_params = StartupMessageParams::new([
-        ("user", &conn_info.username),
-        ("database", &conn_info.dbname),
+        ("user", &username),
+        ("database", &dbname),
         ("application_name", APP_NAME),
     ]);
 
@@ -199,20 +186,21 @@ pub async fn handle(
     let creds = config
         .auth_backend
         .as_ref()
-        .map(|_| {
-            auth::ClientCredentials::parse(
-                &credential_params,
-                Some(&conn_info.hostname),
-                common_names,
-            )
-        })
+        .map(|_| auth::ClientCredentials::parse(&credential_params, Some(&hostname), common_names))
         .transpose()?;
     let extra = console::ConsoleReqExtra {
         session_id: uuid::Uuid::new_v4(),
         application_name: Some(APP_NAME),
     };
-
-    let mut node_info = creds.wake_compute(&extra).await?.expect("msg");
+    let node = creds.wake_compute(&extra).await?.expect("msg");
+    let conf = node.value.config;
+    let port = *conf.get_ports().first().expect("no port");
+    let host = match conf.get_hosts().first().expect("no host") {
+        tokio_postgres::config::Host::Tcp(host) => host,
+        tokio_postgres::config::Host::Unix(_) => {
+            return Err(anyhow::anyhow!("unix socket is not supported"));
+        }
+    };
 
     let request_content_length = match request.body().size_hint().upper() {
         Some(v) => v,
@@ -232,10 +220,28 @@ pub async fn handle(
     let QueryData { query, params } = serde_json::from_slice(&body)?;
     let query_params = json_to_pg_text(params)?;
 
+    //
+    // Connenct to the destination
+    //
+    let (client, connection) = tokio_postgres::Config::new()
+        .host(host)
+        .port(port)
+        .user(&username)
+        .password(&password)
+        .dbname(&dbname)
+        .max_backend_message_size(MAX_RESPONSE_SIZE)
+        .connect(tokio_postgres::NoTls)
+        .await?;
+
+    tokio::spawn(async move {
+        if let Err(e) = connection.await {
+            eprintln!("connection error: {}", e);
+        }
+    });
+
     //
     // Now execute the query and return the result
     //
-    let client = connect_to_compute(&mut node_info, &extra, &creds, &conn_info).await?;
     let row_stream = client.query_raw_txt(query, query_params).await?;
 
     // Manually drain the stream into a vector to leave row_stream hanging
@@ -274,11 +280,6 @@ pub async fn handle(
                 json!({
                     "name": Value::String(c.name().to_owned()),
                     "dataTypeID": Value::Number(c.type_().oid().into()),
-                    "tableID": c.table_oid(),
-                    "columnID": c.column_id(),
-                    "dataTypeSize": c.type_size(),
-                    "dataTypeModifier": c.type_modifier(),
-                    "format": "text",
                 })
             })
             .collect::<Vec<_>>()
@@ -302,70 +303,6 @@ pub async fn handle(
     }))
 }
 
-/// This function is a copy of `connect_to_compute` from `src/proxy.rs` with
-/// the difference that it uses `tokio_postgres` for the connection.
-#[instrument(skip_all)]
-async fn connect_to_compute(
-    node_info: &mut console::CachedNodeInfo,
-    extra: &console::ConsoleReqExtra<'_>,
-    creds: &auth::BackendType<'_, auth::ClientCredentials<'_>>,
-    conn_info: &ConnInfo,
-) -> anyhow::Result<tokio_postgres::Client> {
-    let mut num_retries: usize = NUM_RETRIES_WAKE_COMPUTE;
-
-    loop {
-        match connect_to_compute_once(node_info, conn_info).await {
-            Err(e) if num_retries > 0 => {
-                info!("compute node's state has changed; requesting a wake-up");
-                match creds.wake_compute(extra).await? {
-                    // Update `node_info` and try one more time.
-                    Some(new) => {
-                        *node_info = new;
-                    }
-                    // Link auth doesn't work that way, so we just exit.
-                    None => return Err(e),
-                }
-            }
-            other => return other,
-        }
-
-        num_retries -= 1;
-        info!("retrying after wake-up ({num_retries} attempts left)");
-    }
-}
-
-async fn connect_to_compute_once(
-    node_info: &console::CachedNodeInfo,
-    conn_info: &ConnInfo,
-) -> anyhow::Result<tokio_postgres::Client> {
-    let mut config = (*node_info.config).clone();
-
-    let (client, connection) = config
-        .user(&conn_info.username)
-        .password(&conn_info.password)
-        .dbname(&conn_info.dbname)
-        .max_backend_message_size(MAX_RESPONSE_SIZE)
-        .connect(tokio_postgres::NoTls)
-        .inspect_err(|e: &tokio_postgres::Error| {
-            error!(
-                "failed to connect to compute node hosts={:?} ports={:?}: {}",
-                node_info.config.get_hosts(),
-                node_info.config.get_ports(),
-                e
-            );
-            invalidate_cache(node_info)
-        })
-        .await?;
-
-    tokio::spawn(async move {
-        if let Err(e) = connection.await {
-            error!("connection error: {}", e);
-        }
-    });
-
-    Ok(client)
-}
-
 //
 // Convert postgres row with text-encoded values to JSON object
 //

proxy/src/http/websocket.rs

@@ -26,6 +26,7 @@ use tls_listener::TlsListener;
 use tokio::{
     io::{self, AsyncBufRead, AsyncRead, AsyncWrite, ReadBuf},
     net::TcpListener,
+    select,
 };
 use tokio_util::sync::CancellationToken;
 use tracing::{error, info, info_span, warn, Instrument};
@@ -192,9 +193,14 @@ async fn ws_handler(
     // TODO: that deserves a refactor as now this function also handles http json client besides websockets.
     // Right now I don't want to blow up sql-over-http patch with file renames and do that as a follow up instead.
     } else if request.uri().path() == "/sql" && request.method() == Method::POST {
-        let result = sql_over_http::handle(config, request, sni_hostname)
-            .instrument(info_span!("sql-over-http"))
-            .await;
+        let result = select! {
+            _ = tokio::time::sleep(std::time::Duration::from_secs(10)) => {
+                Err(anyhow::anyhow!("Query timed out"))
+            }
+            response = sql_over_http::handle(config, request, sni_hostname) => {
+                response
+            }
+        };
         let status_code = match result {
             Ok(_) => StatusCode::OK,
             Err(_) => StatusCode::BAD_REQUEST,

proxy/src/proxy.rs

@@ -22,7 +22,7 @@ use tracing::{error, info, warn};
 use utils::measured_stream::MeasuredStream;
 
 /// Number of times we should retry the `/proxy_wake_compute` http request.
-pub const NUM_RETRIES_WAKE_COMPUTE: usize = 1;
+const NUM_RETRIES_WAKE_COMPUTE: usize = 1;
 
 const ERR_INSECURE_CONNECTION: &str = "connection is insecure (try using `sslmode=require`)";
 const ERR_PROTO_VIOLATION: &str = "protocol violation";
@@ -283,35 +283,34 @@ async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
     }
 }
 
-/// If we couldn't connect, a cached connection info might be to blame
-/// (e.g. the compute node's address might've changed at the wrong time).
-/// Invalidate the cache entry (if any) to prevent subsequent errors.
-#[tracing::instrument(name = "invalidate_cache", skip_all)]
-pub fn invalidate_cache(node_info: &console::CachedNodeInfo) {
-    let is_cached = node_info.cached();
-    if is_cached {
-        warn!("invalidating stalled compute node info cache entry");
-        node_info.invalidate();
-    }
-
-    let label = match is_cached {
-        true => "compute_cached",
-        false => "compute_uncached",
-    };
-    NUM_CONNECTION_FAILURES.with_label_values(&[label]).inc();
-}
-
 /// Try to connect to the compute node once.
 #[tracing::instrument(name = "connect_once", skip_all)]
 async fn connect_to_compute_once(
     node_info: &console::CachedNodeInfo,
 ) -> Result<PostgresConnection, compute::ConnectionError> {
+    // If we couldn't connect, a cached connection info might be to blame
+    // (e.g. the compute node's address might've changed at the wrong time).
+    // Invalidate the cache entry (if any) to prevent subsequent errors.
+    let invalidate_cache = |_: &compute::ConnectionError| {
+        let is_cached = node_info.cached();
+        if is_cached {
+            warn!("invalidating stalled compute node info cache entry");
+            node_info.invalidate();
+        }
+
+        let label = match is_cached {
+            true => "compute_cached",
+            false => "compute_uncached",
+        };
+        NUM_CONNECTION_FAILURES.with_label_values(&[label]).inc();
+    };
+
     let allow_self_signed_compute = node_info.allow_self_signed_compute;
 
     node_info
         .config
         .connect(allow_self_signed_compute)
-        .inspect_err(|_: &compute::ConnectionError| invalidate_cache(node_info))
+        .inspect_err(invalidate_cache)
         .await
 }
 

rust-toolchain.toml

@@ -1,5 +1,5 @@
 [toolchain]
-channel = "1.70.0"
+channel = "1.68.2"
 profile = "default"
 # The default profile includes rustc, rust-std, cargo, rust-docs, rustfmt and clippy.
 # https://rust-lang.github.io/rustup/concepts/profiles.html

safekeeper/src/bin/safekeeper.rs

@@ -3,19 +3,15 @@
 //
 use anyhow::{bail, Context, Result};
 use clap::Parser;
-use futures::future::BoxFuture;
-use futures::stream::FuturesUnordered;
-use futures::{FutureExt, StreamExt};
 use remote_storage::RemoteStorageConfig;
-use tokio::runtime::Handle;
-use tokio::signal::unix::{signal, SignalKind};
-use tokio::task::JoinError;
 use toml_edit::Document;
+use utils::signals::ShutdownSignals;
 
 use std::fs::{self, File};
 use std::io::{ErrorKind, Write};
 use std::path::{Path, PathBuf};
 use std::sync::Arc;
+use std::thread;
 use std::time::Duration;
 use storage_broker::Uri;
 use tokio::sync::mpsc;
@@ -24,21 +20,22 @@ use tracing::*;
 use utils::pid_file;
 
 use metrics::set_build_info_metric;
+use safekeeper::broker;
+use safekeeper::control_file;
 use safekeeper::defaults::{
     DEFAULT_HEARTBEAT_TIMEOUT, DEFAULT_HTTP_LISTEN_ADDR, DEFAULT_MAX_OFFLOADER_LAG_BYTES,
     DEFAULT_PG_LISTEN_ADDR,
 };
+use safekeeper::http;
+use safekeeper::remove_wal;
+use safekeeper::wal_backup;
 use safekeeper::wal_service;
 use safekeeper::GlobalTimelines;
 use safekeeper::SafeKeeperConf;
-use safekeeper::{broker, WAL_SERVICE_RUNTIME};
-use safekeeper::{control_file, BROKER_RUNTIME};
-use safekeeper::{http, WAL_REMOVER_RUNTIME};
-use safekeeper::{remove_wal, WAL_BACKUP_RUNTIME};
-use safekeeper::{wal_backup, HTTP_RUNTIME};
 use storage_broker::DEFAULT_ENDPOINT;
 use utils::auth::JwtAuth;
 use utils::{
+    http::endpoint,
     id::NodeId,
     logging::{self, LogFormat},
     project_git_version,
@@ -107,6 +104,10 @@ struct Args {
     /// Safekeeper won't be elected for WAL offloading if it is lagging for more than this value in bytes
     #[arg(long, default_value_t = DEFAULT_MAX_OFFLOADER_LAG_BYTES)]
     max_offloader_lag: u64,
+    /// Number of threads for wal backup runtime, by default number of cores
+    /// available to the system.
+    #[arg(long)]
+    wal_backup_threads: Option<usize>,
     /// Number of max parallel WAL segments to be offloaded to remote storage.
     #[arg(long, default_value = "5")]
     wal_backup_parallel_jobs: usize,
@@ -120,14 +121,9 @@ struct Args {
     /// Format for logging, either 'plain' or 'json'.
     #[arg(long, default_value = "plain")]
     log_format: String,
-    /// Run everything in single threaded current thread runtime, might be
-    /// useful for debugging.
-    #[arg(long)]
-    current_thread_runtime: bool,
 }
 
-#[tokio::main(flavor = "current_thread")]
-async fn main() -> anyhow::Result<()> {
+fn main() -> anyhow::Result<()> {
     let args = Args::parse();
 
     if let Some(addr) = args.dump_control_file {
@@ -187,10 +183,10 @@ async fn main() -> anyhow::Result<()> {
         heartbeat_timeout: args.heartbeat_timeout,
         remote_storage: args.remote_storage,
         max_offloader_lag_bytes: args.max_offloader_lag,
+        backup_runtime_threads: args.wal_backup_threads,
         wal_backup_enabled: !args.disable_wal_backup,
         backup_parallel_jobs: args.wal_backup_parallel_jobs,
         auth,
-        current_thread_runtime: args.current_thread_runtime,
     };
 
     // initialize sentry if SENTRY_DSN is provided
@@ -198,14 +194,10 @@ async fn main() -> anyhow::Result<()> {
         Some(GIT_VERSION.into()),
         &[("node_id", &conf.my_id.to_string())],
     );
-    start_safekeeper(conf).await
+    start_safekeeper(conf)
 }
 
-/// Result of joining any of main tasks: upper error means task failed to
-/// complete, e.g. panicked, inner is error produced by task itself.
-type JoinTaskRes = Result<anyhow::Result<()>, JoinError>;
-
-async fn start_safekeeper(conf: SafeKeeperConf) -> Result<()> {
+fn start_safekeeper(conf: SafeKeeperConf) -> Result<()> {
     // Prevent running multiple safekeepers on the same directory
     let lock_file_path = conf.workdir.join(PID_FILE_NAME);
     let lock_file =
@@ -216,18 +208,14 @@ async fn start_safekeeper(conf: SafeKeeperConf) -> Result<()> {
     // we need to release the lock file only when the current process is gone
     std::mem::forget(lock_file);
 
-    info!("starting safekeeper WAL service on {}", conf.listen_pg_addr);
-    let pg_listener = tcp_listener::bind(conf.listen_pg_addr.clone()).map_err(|e| {
-        error!("failed to bind to address {}: {}", conf.listen_pg_addr, e);
+    let http_listener = tcp_listener::bind(conf.listen_http_addr.clone()).map_err(|e| {
+        error!("failed to bind to address {}: {}", conf.listen_http_addr, e);
         e
     })?;
 
-    info!(
-        "starting safekeeper HTTP service on {}",
-        conf.listen_http_addr
-    );
-    let http_listener = tcp_listener::bind(conf.listen_http_addr.clone()).map_err(|e| {
-        error!("failed to bind to address {}: {}", conf.listen_http_addr, e);
+    info!("starting safekeeper on {}", conf.listen_pg_addr);
+    let pg_listener = tcp_listener::bind(conf.listen_pg_addr.clone()).map_err(|e| {
+        error!("failed to bind to address {}: {}", conf.listen_pg_addr, e);
         e
     })?;
 
@@ -236,88 +224,71 @@ async fn start_safekeeper(conf: SafeKeeperConf) -> Result<()> {
     let timeline_collector = safekeeper::metrics::TimelineCollector::new();
     metrics::register_internal(Box::new(timeline_collector))?;
 
+    let mut threads = vec![];
     let (wal_backup_launcher_tx, wal_backup_launcher_rx) = mpsc::channel(100);
 
     // Load all timelines from disk to memory.
     GlobalTimelines::init(conf.clone(), wal_backup_launcher_tx)?;
 
-    // Keep handles to main tasks to die if any of them disappears.
-    let mut tasks_handles: FuturesUnordered<BoxFuture<(String, JoinTaskRes)>> =
-        FuturesUnordered::new();
+    let conf_ = conf.clone();
+    threads.push(
+        thread::Builder::new()
+            .name("http_endpoint_thread".into())
+            .spawn(|| {
+                let router = http::make_router(conf_);
+                endpoint::serve_thread_main(
+                    router,
+                    http_listener,
+                    std::future::pending(), // never shut down
+                )
+                .unwrap();
+            })?,
+    );
+
+    let conf_cloned = conf.clone();
+    let safekeeper_thread = thread::Builder::new()
+        .name("WAL service thread".into())
+        .spawn(|| wal_service::thread_main(conf_cloned, pg_listener))
+        .unwrap();
+
+    threads.push(safekeeper_thread);
 
     let conf_ = conf.clone();
-    // Run everything in current thread rt, if asked.
-    if conf.current_thread_runtime {
-        info!("running in current thread runtime");
-    }
-    let current_thread_rt = conf
-        .current_thread_runtime
-        .then(|| Handle::try_current().expect("no runtime in main"));
-    let wal_service_handle = current_thread_rt
-        .as_ref()
-        .unwrap_or_else(|| WAL_SERVICE_RUNTIME.handle())
-        .spawn(wal_service::task_main(conf_, pg_listener))
-        // wrap with task name for error reporting
-        .map(|res| ("WAL service main".to_owned(), res));
-    tasks_handles.push(Box::pin(wal_service_handle));
+    threads.push(
+        thread::Builder::new()
+            .name("broker thread".into())
+            .spawn(|| {
+                broker::thread_main(conf_);
+            })?,
+    );
 
     let conf_ = conf.clone();
-    let http_handle = current_thread_rt
-        .as_ref()
-        .unwrap_or_else(|| HTTP_RUNTIME.handle())
-        .spawn(http::task_main(conf_, http_listener))
-        .map(|res| ("HTTP service main".to_owned(), res));
-    tasks_handles.push(Box::pin(http_handle));
+    threads.push(
+        thread::Builder::new()
+            .name("WAL removal thread".into())
+            .spawn(|| {
+                remove_wal::thread_main(conf_);
+            })?,
+    );
 
-    let conf_ = conf.clone();
-    let broker_task_handle = current_thread_rt
-        .as_ref()
-        .unwrap_or_else(|| BROKER_RUNTIME.handle())
-        .spawn(broker::task_main(conf_).instrument(info_span!("broker")))
-        .map(|res| ("broker main".to_owned(), res));
-    tasks_handles.push(Box::pin(broker_task_handle));
-
-    let conf_ = conf.clone();
-    let wal_remover_handle = current_thread_rt
-        .as_ref()
-        .unwrap_or_else(|| WAL_REMOVER_RUNTIME.handle())
-        .spawn(remove_wal::task_main(conf_))
-        .map(|res| ("WAL remover".to_owned(), res));
-    tasks_handles.push(Box::pin(wal_remover_handle));
-
-    let conf_ = conf.clone();
-    let wal_backup_handle = current_thread_rt
-        .as_ref()
-        .unwrap_or_else(|| WAL_BACKUP_RUNTIME.handle())
-        .spawn(wal_backup::wal_backup_launcher_task_main(
-            conf_,
-            wal_backup_launcher_rx,
-        ))
-        .map(|res| ("WAL backup launcher".to_owned(), res));
-    tasks_handles.push(Box::pin(wal_backup_handle));
+    threads.push(
+        thread::Builder::new()
+            .name("WAL backup launcher thread".into())
+            .spawn(move || {
+                wal_backup::wal_backup_launcher_thread_main(conf, wal_backup_launcher_rx);
+            })?,
+    );
 
     set_build_info_metric(GIT_VERSION);
+    // TODO: put more thoughts into handling of failed threads
+    // We should catch & die if they are in trouble.
 
-    // TODO: update tokio-stream, convert to real async Stream with
-    // SignalStream, map it to obtain missing signal name, combine streams into
-    // single stream we can easily sit on.
-    let mut sigquit_stream = signal(SignalKind::quit())?;
-    let mut sigint_stream = signal(SignalKind::interrupt())?;
-    let mut sigterm_stream = signal(SignalKind::terminate())?;
-
-    tokio::select! {
-        Some((task_name, res)) = tasks_handles.next()=> {
-            error!("{} task failed: {:?}, exiting", task_name, res);
-            std::process::exit(1);
-        }
-        // On any shutdown signal, log receival and exit. Additionally, handling
-        // SIGQUIT prevents coredump.
-        _ = sigquit_stream.recv() => info!("received SIGQUIT, terminating"),
-        _ = sigint_stream.recv() => info!("received SIGINT, terminating"),
-        _ = sigterm_stream.recv() => info!("received SIGTERM, terminating")
-
-    };
-    std::process::exit(0);
+    // On any shutdown signal, log receival and exit. Additionally, handling
+    // SIGQUIT prevents coredump.
+    ShutdownSignals::handle(|signal| {
+        info!("received {}, terminating", signal.name());
+        std::process::exit(0);
+    })
 }
 
 /// Determine safekeeper id.

safekeeper/src/broker.rs

@@ -8,7 +8,7 @@ use anyhow::Error;
 use anyhow::Result;
 
 use storage_broker::parse_proto_ttid;
-
+use storage_broker::proto::broker_service_client::BrokerServiceClient;
 use storage_broker::proto::subscribe_safekeeper_info_request::SubscriptionKey as ProtoSubscriptionKey;
 use storage_broker::proto::SubscribeSafekeeperInfoRequest;
 use storage_broker::Request;
@@ -16,7 +16,7 @@ use storage_broker::Request;
 use std::time::Duration;
 use std::time::Instant;
 use tokio::task::JoinHandle;
-use tokio::time::sleep;
+use tokio::{runtime, time::sleep};
 use tracing::*;
 
 use crate::metrics::BROKER_ITERATION_TIMELINES;
@@ -29,10 +29,23 @@ use crate::SafeKeeperConf;
 const RETRY_INTERVAL_MSEC: u64 = 1000;
 const PUSH_INTERVAL_MSEC: u64 = 1000;
 
+pub fn thread_main(conf: SafeKeeperConf) {
+    let runtime = runtime::Builder::new_current_thread()
+        .enable_all()
+        .build()
+        .unwrap();
+
+    let _enter = info_span!("broker").entered();
+    info!("started, broker endpoint {:?}", conf.broker_endpoint);
+
+    runtime.block_on(async {
+        main_loop(conf).await;
+    });
+}
+
 /// Push once in a while data about all active timelines to the broker.
 async fn push_loop(conf: SafeKeeperConf) -> anyhow::Result<()> {
-    let mut client =
-        storage_broker::connect(conf.broker_endpoint.clone(), conf.broker_keepalive_interval)?;
+    let mut client = BrokerServiceClient::connect(conf.broker_endpoint.clone()).await?;
     let push_interval = Duration::from_millis(PUSH_INTERVAL_MSEC);
 
     let outbound = async_stream::stream! {
@@ -42,27 +55,20 @@ async fn push_loop(conf: SafeKeeperConf) -> anyhow::Result<()> {
             // sensitive and there is no risk of deadlock as we don't await while
             // lock is held.
             let now = Instant::now();
-            let all_tlis = GlobalTimelines::get_all();
-            let mut n_pushed_tlis = 0;
-            for tli in &all_tlis {
-                // filtering alternative futures::stream::iter(all_tlis)
-                //   .filter(|tli| {let tli = tli.clone(); async move { tli.is_active().await}}).collect::<Vec<_>>().await;
-                // doesn't look better, and I'm not sure how to do that without collect.
-                if !tli.is_active().await {
-                    continue;
-                }
-                let sk_info = tli.get_safekeeper_info(&conf).await;
+            let mut active_tlis = GlobalTimelines::get_all();
+            active_tlis.retain(|tli| tli.is_active());
+            for tli in &active_tlis {
+                let sk_info = tli.get_safekeeper_info(&conf);
                 yield sk_info;
                 BROKER_PUSHED_UPDATES.inc();
-                n_pushed_tlis += 1;
             }
             let elapsed = now.elapsed();
 
             BROKER_PUSH_ALL_UPDATES_SECONDS.observe(elapsed.as_secs_f64());
-            BROKER_ITERATION_TIMELINES.observe(n_pushed_tlis as f64);
+            BROKER_ITERATION_TIMELINES.observe(active_tlis.len() as f64);
 
             if elapsed > push_interval / 2 {
-                info!("broker push is too long, pushed {} timeline updates to broker in {:?}", n_pushed_tlis, elapsed);
+                info!("broker push is too long, pushed {} timeline updates to broker in {:?}", active_tlis.len(), elapsed);
             }
 
             sleep(push_interval).await;
@@ -119,13 +125,10 @@ async fn pull_loop(conf: SafeKeeperConf) -> Result<()> {
     bail!("end of stream");
 }
 
-pub async fn task_main(conf: SafeKeeperConf) -> anyhow::Result<()> {
-    info!("started, broker endpoint {:?}", conf.broker_endpoint);
-
+async fn main_loop(conf: SafeKeeperConf) {
     let mut ticker = tokio::time::interval(Duration::from_millis(RETRY_INTERVAL_MSEC));
     let mut push_handle: Option<JoinHandle<Result<(), Error>>> = None;
     let mut pull_handle: Option<JoinHandle<Result<(), Error>>> = None;
-
     // Selecting on JoinHandles requires some squats; is there a better way to
     // reap tasks individually?
 

safekeeper/src/control_file.rs

@@ -2,10 +2,9 @@
 
 use anyhow::{bail, ensure, Context, Result};
 use byteorder::{LittleEndian, ReadBytesExt, WriteBytesExt};
-use tokio::fs::{self, File};
-use tokio::io::AsyncWriteExt;
 
-use std::io::Read;
+use std::fs::{self, File, OpenOptions};
+use std::io::{Read, Write};
 use std::ops::Deref;
 use std::path::{Path, PathBuf};
 use std::time::Instant;
@@ -27,10 +26,9 @@ pub const CHECKSUM_SIZE: usize = std::mem::size_of::<u32>();
 
 /// Storage should keep actual state inside of it. It should implement Deref
 /// trait to access state fields and have persist method for updating that state.
-#[async_trait::async_trait]
 pub trait Storage: Deref<Target = SafeKeeperState> {
     /// Persist safekeeper state on disk and update internal state.
-    async fn persist(&mut self, s: &SafeKeeperState) -> Result<()>;
+    fn persist(&mut self, s: &SafeKeeperState) -> Result<()>;
 
     /// Timestamp of last persist.
     fn last_persist_at(&self) -> Instant;
@@ -84,7 +82,7 @@ impl FileStorage {
     /// Check the magic/version in the on-disk data and deserialize it, if possible.
     fn deser_sk_state(buf: &mut &[u8]) -> Result<SafeKeeperState> {
         // Read the version independent part
-        let magic = ReadBytesExt::read_u32::<LittleEndian>(buf)?;
+        let magic = buf.read_u32::<LittleEndian>()?;
         if magic != SK_MAGIC {
             bail!(
                 "bad control file magic: {:X}, expected {:X}",
@@ -92,7 +90,7 @@ impl FileStorage {
                 SK_MAGIC
             );
         }
-        let version = ReadBytesExt::read_u32::<LittleEndian>(buf)?;
+        let version = buf.read_u32::<LittleEndian>()?;
         if version == SK_FORMAT_VERSION {
             let res = SafeKeeperState::des(buf)?;
             return Ok(res);
@@ -112,7 +110,7 @@ impl FileStorage {
 
     /// Read in the control file.
     pub fn load_control_file<P: AsRef<Path>>(control_file_path: P) -> Result<SafeKeeperState> {
-        let mut control_file = std::fs::OpenOptions::new()
+        let mut control_file = OpenOptions::new()
             .read(true)
             .write(true)
             .open(&control_file_path)
@@ -161,31 +159,30 @@ impl Deref for FileStorage {
     }
 }
 
-#[async_trait::async_trait]
 impl Storage for FileStorage {
     /// persists state durably to underlying storage
     /// for description see https://lwn.net/Articles/457667/
-    async fn persist(&mut self, s: &SafeKeeperState) -> Result<()> {
+    fn persist(&mut self, s: &SafeKeeperState) -> Result<()> {
         let _timer = PERSIST_CONTROL_FILE_SECONDS.start_timer();
 
         // write data to safekeeper.control.partial
         let control_partial_path = self.timeline_dir.join(CONTROL_FILE_NAME_PARTIAL);
-        let mut control_partial = File::create(&control_partial_path).await.with_context(|| {
+        let mut control_partial = File::create(&control_partial_path).with_context(|| {
             format!(
                 "failed to create partial control file at: {}",
                 &control_partial_path.display()
             )
         })?;
         let mut buf: Vec<u8> = Vec::new();
-        WriteBytesExt::write_u32::<LittleEndian>(&mut buf, SK_MAGIC)?;
-        WriteBytesExt::write_u32::<LittleEndian>(&mut buf, SK_FORMAT_VERSION)?;
+        buf.write_u32::<LittleEndian>(SK_MAGIC)?;
+        buf.write_u32::<LittleEndian>(SK_FORMAT_VERSION)?;
         s.ser_into(&mut buf)?;
 
         // calculate checksum before resize
         let checksum = crc32c::crc32c(&buf);
         buf.extend_from_slice(&checksum.to_le_bytes());
 
-        control_partial.write_all(&buf).await.with_context(|| {
+        control_partial.write_all(&buf).with_context(|| {
             format!(
                 "failed to write safekeeper state into control file at: {}",
                 control_partial_path.display()
@@ -194,7 +191,7 @@ impl Storage for FileStorage {
 
         // fsync the file
         if !self.conf.no_sync {
-            control_partial.sync_all().await.with_context(|| {
+            control_partial.sync_all().with_context(|| {
                 format!(
                     "failed to sync partial control file at {}",
                     control_partial_path.display()
@@ -205,22 +202,21 @@ impl Storage for FileStorage {
         let control_path = self.timeline_dir.join(CONTROL_FILE_NAME);
 
         // rename should be atomic
-        fs::rename(&control_partial_path, &control_path).await?;
+        fs::rename(&control_partial_path, &control_path)?;
         // this sync is not required by any standard but postgres does this (see durable_rename)
         if !self.conf.no_sync {
-            let new_f = File::open(&control_path).await?;
-            new_f.sync_all().await.with_context(|| {
-                format!(
-                    "failed to sync control file at: {}",
-                    &control_path.display()
-                )
-            })?;
+            File::open(&control_path)
+                .and_then(|f| f.sync_all())
+                .with_context(|| {
+                    format!(
+                        "failed to sync control file at: {}",
+                        &control_path.display()
+                    )
+                })?;
 
             // fsync the directory (linux specific)
-            let tli_dir = File::open(&self.timeline_dir).await?;
-            tli_dir
-                .sync_all()
-                .await
+            File::open(&self.timeline_dir)
+                .and_then(|f| f.sync_all())
                 .context("failed to sync control file directory")?;
         }
 
@@ -240,6 +236,7 @@ mod test {
     use super::*;
     use crate::{safekeeper::SafeKeeperState, SafeKeeperConf};
     use anyhow::Result;
+    use std::fs;
     use utils::{id::TenantTimelineId, lsn::Lsn};
 
     fn stub_conf() -> SafeKeeperConf {
@@ -250,75 +247,59 @@ mod test {
         }
     }
 
-    async fn load_from_control_file(
+    fn load_from_control_file(
         conf: &SafeKeeperConf,
         ttid: &TenantTimelineId,
     ) -> Result<(FileStorage, SafeKeeperState)> {
-        fs::create_dir_all(conf.timeline_dir(ttid))
-            .await
-            .expect("failed to create timeline dir");
+        fs::create_dir_all(conf.timeline_dir(ttid)).expect("failed to create timeline dir");
         Ok((
             FileStorage::restore_new(ttid, conf)?,
             FileStorage::load_control_file_conf(conf, ttid)?,
         ))
     }
 
-    async fn create(
+    fn create(
         conf: &SafeKeeperConf,
         ttid: &TenantTimelineId,
     ) -> Result<(FileStorage, SafeKeeperState)> {
-        fs::create_dir_all(conf.timeline_dir(ttid))
-            .await
-            .expect("failed to create timeline dir");
+        fs::create_dir_all(conf.timeline_dir(ttid)).expect("failed to create timeline dir");
         let state = SafeKeeperState::empty();
         let storage = FileStorage::create_new(ttid, conf, state.clone())?;
         Ok((storage, state))
     }
 
-    #[tokio::test]
-    async fn test_read_write_safekeeper_state() {
+    #[test]
+    fn test_read_write_safekeeper_state() {
         let conf = stub_conf();
         let ttid = TenantTimelineId::generate();
         {
-            let (mut storage, mut state) =
-                create(&conf, &ttid).await.expect("failed to create state");
+            let (mut storage, mut state) = create(&conf, &ttid).expect("failed to create state");
             // change something
             state.commit_lsn = Lsn(42);
-            storage
-                .persist(&state)
-                .await
-                .expect("failed to persist state");
+            storage.persist(&state).expect("failed to persist state");
         }
 
-        let (_, state) = load_from_control_file(&conf, &ttid)
-            .await
-            .expect("failed to read state");
+        let (_, state) = load_from_control_file(&conf, &ttid).expect("failed to read state");
         assert_eq!(state.commit_lsn, Lsn(42));
     }
 
-    #[tokio::test]
-    async fn test_safekeeper_state_checksum_mismatch() {
+    #[test]
+    fn test_safekeeper_state_checksum_mismatch() {
         let conf = stub_conf();
         let ttid = TenantTimelineId::generate();
         {
-            let (mut storage, mut state) =
-                create(&conf, &ttid).await.expect("failed to read state");
+            let (mut storage, mut state) = create(&conf, &ttid).expect("failed to read state");
 
             // change something
             state.commit_lsn = Lsn(42);
-            storage
-                .persist(&state)
-                .await
-                .expect("failed to persist state");
+            storage.persist(&state).expect("failed to persist state");
         }
         let control_path = conf.timeline_dir(&ttid).join(CONTROL_FILE_NAME);
-        let mut data = fs::read(&control_path).await.unwrap();
+        let mut data = fs::read(&control_path).unwrap();
         data[0] += 1; // change the first byte of the file to fail checksum validation
-        fs::write(&control_path, &data)
-            .await
-            .expect("failed to write control file");
+        fs::write(&control_path, &data).expect("failed to write control file");
 
-        match load_from_control_file(&conf, &ttid).await {
+        match load_from_control_file(&conf, &ttid) {
             Err(err) => assert!(err
                 .to_string()
                 .contains("safekeeper control file checksum mismatch")),

safekeeper/src/debug_dump.rs

@@ -121,7 +121,7 @@ pub struct FileInfo {
 }
 
 /// Build debug dump response, using the provided [`Args`] filters.
-pub async fn build(args: Args) -> Result<Response> {
+pub fn build(args: Args) -> Result<Response> {
     let start_time = Utc::now();
     let timelines_count = GlobalTimelines::timelines_count();
 
@@ -155,7 +155,7 @@ pub async fn build(args: Args) -> Result<Response> {
         }
 
         let control_file = if args.dump_control_file {
-            let mut state = tli.get_state().await.1;
+            let mut state = tli.get_state().1;
             if !args.dump_term_history {
                 state.acceptor_state.term_history = TermHistory(vec![]);
             }
@@ -165,7 +165,7 @@ pub async fn build(args: Args) -> Result<Response> {
         };
 
         let memory = if args.dump_memory {
-            Some(tli.memory_dump().await)
+            Some(tli.memory_dump())
         } else {
             None
         };

safekeeper/src/handler.rs

@@ -256,14 +256,14 @@ impl SafekeeperPostgresHandler {
 
         let lsn = if self.is_walproposer_recovery() {
             // walproposer should get all local WAL until flush_lsn
-            tli.get_flush_lsn().await
+            tli.get_flush_lsn()
         } else {
             // other clients shouldn't get any uncommitted WAL
-            tli.get_state().await.0.commit_lsn
+            tli.get_state().0.commit_lsn
         }
         .to_string();
 
-        let sysid = tli.get_state().await.1.server.system_id.to_string();
+        let sysid = tli.get_state().1.server.system_id.to_string();
         let lsn_bytes = lsn.as_bytes();
         let tli = PG_TLI.to_string();
         let tli_bytes = tli.as_bytes();

safekeeper/src/http/mod.rs

@@ -2,18 +2,3 @@ pub mod routes;
 pub use routes::make_router;
 
 pub use safekeeper_api::models;
-
-use crate::SafeKeeperConf;
-
-pub async fn task_main(
-    conf: SafeKeeperConf,
-    http_listener: std::net::TcpListener,
-) -> anyhow::Result<()> {
-    let router = make_router(conf)
-        .build()
-        .map_err(|err| anyhow::anyhow!(err))?;
-    let service = utils::http::RouterService::new(router).unwrap();
-    let server = hyper::Server::from_tcp(http_listener)?;
-    server.serve(service).await?;
-    Ok(()) // unreachable
-}

safekeeper/src/http/routes.rs

@@ -13,7 +13,7 @@ use storage_broker::proto::SafekeeperTimelineInfo;
 use storage_broker::proto::TenantTimelineId as ProtoTenantTimelineId;
 use tokio::fs::File;
 use tokio::io::AsyncReadExt;
-use utils::http::endpoint::request_span;
+use tokio::task::JoinError;
 
 use crate::safekeeper::ServerInfo;
 use crate::safekeeper::Term;
@@ -116,8 +116,8 @@ async fn timeline_status_handler(request: Request<Body>) -> Result<Response<Body
     check_permission(&request, Some(ttid.tenant_id))?;
 
     let tli = GlobalTimelines::get(ttid).map_err(ApiError::from)?;
-    let (inmem, state) = tli.get_state().await;
-    let flush_lsn = tli.get_flush_lsn().await;
+    let (inmem, state) = tli.get_state();
+    let flush_lsn = tli.get_flush_lsn();
 
     let epoch = state.acceptor_state.get_epoch(flush_lsn);
     let term_history = state
@@ -232,11 +232,13 @@ async fn timeline_delete_force_handler(
     );
     check_permission(&request, Some(ttid.tenant_id))?;
     ensure_no_body(&mut request).await?;
-    // FIXME: `delete_force` can fail from both internal errors and bad requests. Add better
-    // error handling here when we're able to.
-    let resp = GlobalTimelines::delete_force(&ttid)
-        .await
-        .map_err(ApiError::InternalServerError)?;
+    let resp = tokio::task::spawn_blocking(move || {
+        // FIXME: `delete_force` can fail from both internal errors and bad requests. Add better
+        // error handling here when we're able to.
+        GlobalTimelines::delete_force(&ttid).map_err(ApiError::InternalServerError)
+    })
+    .await
+    .map_err(|e: JoinError| ApiError::InternalServerError(e.into()))??;
     json_response(StatusCode::OK, resp)
 }
 
@@ -248,11 +250,14 @@ async fn tenant_delete_force_handler(
     let tenant_id = parse_request_param(&request, "tenant_id")?;
     check_permission(&request, Some(tenant_id))?;
     ensure_no_body(&mut request).await?;
-    // FIXME: `delete_force_all_for_tenant` can return an error for multiple different reasons;
-    // Using an `InternalServerError` should be fixed when the types support it
-    let delete_info = GlobalTimelines::delete_force_all_for_tenant(&tenant_id)
-        .await
-        .map_err(ApiError::InternalServerError)?;
+    let delete_info = tokio::task::spawn_blocking(move || {
+        // FIXME: `delete_force_all_for_tenant` can return an error for multiple different reasons;
+        // Using an `InternalServerError` should be fixed when the types support it
+        GlobalTimelines::delete_force_all_for_tenant(&tenant_id)
+            .map_err(ApiError::InternalServerError)
+    })
+    .await
+    .map_err(|e: JoinError| ApiError::InternalServerError(e.into()))??;
     json_response(
         StatusCode::OK,
         delete_info
@@ -348,9 +353,11 @@ async fn dump_debug_handler(mut request: Request<Body>) -> Result<Response<Body>
         timeline_id,
     };
 
-    let resp = debug_dump::build(args)
-        .await
-        .map_err(ApiError::InternalServerError)?;
+    let resp = tokio::task::spawn_blocking(move || {
+        debug_dump::build(args).map_err(ApiError::InternalServerError)
+    })
+    .await
+    .map_err(|e: JoinError| ApiError::InternalServerError(e.into()))??;
 
     // TODO: use streaming response
     json_response(StatusCode::OK, resp)
@@ -379,32 +386,29 @@ pub fn make_router(conf: SafeKeeperConf) -> RouterBuilder<hyper::Body, ApiError>
     router
         .data(Arc::new(conf))
         .data(auth)
-        .get("/v1/status", |r| request_span(r, status_handler))
+        .get("/v1/status", status_handler)
         // Will be used in the future instead of implicit timeline creation
-        .post("/v1/tenant/timeline", |r| {
-            request_span(r, timeline_create_handler)
-        })
-        .get("/v1/tenant/:tenant_id/timeline/:timeline_id", |r| {
-            request_span(r, timeline_status_handler)
-        })
-        .delete("/v1/tenant/:tenant_id/timeline/:timeline_id", |r| {
-            request_span(r, timeline_delete_force_handler)
-        })
-        .delete("/v1/tenant/:tenant_id", |r| {
-            request_span(r, tenant_delete_force_handler)
-        })
-        .post("/v1/pull_timeline", |r| {
-            request_span(r, timeline_pull_handler)
-        })
+        .post("/v1/tenant/timeline", timeline_create_handler)
+        .get(
+            "/v1/tenant/:tenant_id/timeline/:timeline_id",
+            timeline_status_handler,
+        )
+        .delete(
+            "/v1/tenant/:tenant_id/timeline/:timeline_id",
+            timeline_delete_force_handler,
+        )
+        .delete("/v1/tenant/:tenant_id", tenant_delete_force_handler)
+        .post("/v1/pull_timeline", timeline_pull_handler)
         .get(
             "/v1/tenant/:tenant_id/timeline/:timeline_id/file/:filename",
-            |r| request_span(r, timeline_files_handler),
+            timeline_files_handler,
         )
         // for tests
-        .post("/v1/record_safekeeper_info/:tenant_id/:timeline_id", |r| {
-            request_span(r, record_safekeeper_info)
-        })
-        .get("/v1/debug_dump", |r| request_span(r, dump_debug_handler))
+        .post(
+            "/v1/record_safekeeper_info/:tenant_id/:timeline_id",
+            record_safekeeper_info,
+        )
+        .get("/v1/debug_dump", dump_debug_handler)
 }
 
 #[cfg(test)]

safekeeper/src/json_ctrl.rs

@@ -73,12 +73,12 @@ pub async fn handle_json_ctrl<IO: AsyncRead + AsyncWrite + Unpin>(
 
     // if send_proposer_elected is true, we need to update local history
     if append_request.send_proposer_elected {
-        send_proposer_elected(&tli, append_request.term, append_request.epoch_start_lsn).await?;
+        send_proposer_elected(&tli, append_request.term, append_request.epoch_start_lsn)?;
     }
 
-    let inserted_wal = append_logical_message(&tli, append_request).await?;
+    let inserted_wal = append_logical_message(&tli, append_request)?;
     let response = AppendResult {
-        state: tli.get_state().await.1,
+        state: tli.get_state().1,
         inserted_wal,
     };
     let response_data = serde_json::to_vec(&response)
@@ -114,9 +114,9 @@ async fn prepare_safekeeper(
     .await
 }
 
-async fn send_proposer_elected(tli: &Arc<Timeline>, term: Term, lsn: Lsn) -> anyhow::Result<()> {
+fn send_proposer_elected(tli: &Arc<Timeline>, term: Term, lsn: Lsn) -> anyhow::Result<()> {
     // add new term to existing history
-    let history = tli.get_state().await.1.acceptor_state.term_history;
+    let history = tli.get_state().1.acceptor_state.term_history;
     let history = history.up_to(lsn.checked_sub(1u64).unwrap());
     let mut history_entries = history.0;
     history_entries.push(TermSwitchEntry { term, lsn });
@@ -129,7 +129,7 @@ async fn send_proposer_elected(tli: &Arc<Timeline>, term: Term, lsn: Lsn) -> any
         timeline_start_lsn: lsn,
     });
 
-    tli.process_msg(&proposer_elected_request).await?;
+    tli.process_msg(&proposer_elected_request)?;
     Ok(())
 }
 
@@ -142,12 +142,12 @@ pub struct InsertedWAL {
 
 /// Extend local WAL with new LogicalMessage record. To do that,
 /// create AppendRequest with new WAL and pass it to safekeeper.
-pub async fn append_logical_message(
+pub fn append_logical_message(
     tli: &Arc<Timeline>,
     msg: &AppendLogicalMessage,
 ) -> anyhow::Result<InsertedWAL> {
     let wal_data = encode_logical_message(&msg.lm_prefix, &msg.lm_message);
-    let sk_state = tli.get_state().await.1;
+    let sk_state = tli.get_state().1;
 
     let begin_lsn = msg.begin_lsn;
     let end_lsn = begin_lsn + wal_data.len() as u64;
@@ -171,7 +171,7 @@ pub async fn append_logical_message(
         wal_data: Bytes::from(wal_data),
     });
 
-    let response = tli.process_msg(&append_request).await?;
+    let response = tli.process_msg(&append_request)?;
 
     let append_response = match response {
         Some(AcceptorProposerMessage::AppendResponse(resp)) => resp,

safekeeper/src/lib.rs

@@ -1,6 +1,4 @@
-use once_cell::sync::Lazy;
 use remote_storage::RemoteStorageConfig;
-use tokio::runtime::Runtime;
 
 use std::path::PathBuf;
 use std::time::Duration;
@@ -38,6 +36,7 @@ pub mod defaults {
         DEFAULT_PG_LISTEN_PORT,
     };
 
+    pub const DEFAULT_WAL_BACKUP_RUNTIME_THREADS: usize = 8;
     pub const DEFAULT_HEARTBEAT_TIMEOUT: &str = "5000ms";
     pub const DEFAULT_MAX_OFFLOADER_LAG_BYTES: u64 = 128 * (1 << 20);
 }
@@ -61,10 +60,10 @@ pub struct SafeKeeperConf {
     pub heartbeat_timeout: Duration,
     pub remote_storage: Option<RemoteStorageConfig>,
     pub max_offloader_lag_bytes: u64,
+    pub backup_runtime_threads: Option<usize>,
     pub backup_parallel_jobs: usize,
     pub wal_backup_enabled: bool,
     pub auth: Option<Arc<JwtAuth>>,
-    pub current_thread_runtime: bool,
 }
 
 impl SafeKeeperConf {
@@ -93,64 +92,12 @@ impl SafeKeeperConf {
                 .parse()
                 .expect("failed to parse default broker endpoint"),
             broker_keepalive_interval: Duration::from_secs(5),
+            backup_runtime_threads: None,
             wal_backup_enabled: true,
             backup_parallel_jobs: 1,
             auth: None,
             heartbeat_timeout: Duration::new(5, 0),
             max_offloader_lag_bytes: defaults::DEFAULT_MAX_OFFLOADER_LAG_BYTES,
-            current_thread_runtime: false,
         }
     }
 }
-
-// Tokio runtimes.
-pub static WAL_SERVICE_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
-    tokio::runtime::Builder::new_multi_thread()
-        .thread_name("WAL service worker")
-        .enable_all()
-        .build()
-        .expect("Failed to create WAL service runtime")
-});
-
-pub static HTTP_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
-    tokio::runtime::Builder::new_multi_thread()
-        .thread_name("HTTP worker")
-        .enable_all()
-        .build()
-        .expect("Failed to create WAL service runtime")
-});
-
-pub static BROKER_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
-    tokio::runtime::Builder::new_multi_thread()
-        .thread_name("broker worker")
-        .worker_threads(2) // there are only 2 tasks, having more threads doesn't make sense
-        .enable_all()
-        .build()
-        .expect("Failed to create broker runtime")
-});
-
-pub static WAL_REMOVER_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
-    tokio::runtime::Builder::new_multi_thread()
-        .thread_name("WAL remover")
-        .worker_threads(1)
-        .enable_all()
-        .build()
-        .expect("Failed to create broker runtime")
-});
-
-pub static WAL_BACKUP_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
-    tokio::runtime::Builder::new_multi_thread()
-        .thread_name("WAL backup worker")
-        .enable_all()
-        .build()
-        .expect("Failed to create WAL backup runtime")
-});
-
-pub static METRICS_SHIFTER_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
-    tokio::runtime::Builder::new_multi_thread()
-        .thread_name("metric shifter")
-        .worker_threads(1)
-        .enable_all()
-        .build()
-        .expect("Failed to create broker runtime")
-});

safekeeper/src/metrics.rs

@@ -7,7 +7,6 @@ use std::{
 
 use ::metrics::{register_histogram, GaugeVec, Histogram, IntGauge, DISK_WRITE_SECONDS_BUCKETS};
 use anyhow::Result;
-use futures::Future;
 use metrics::{
     core::{AtomicU64, Collector, Desc, GenericCounter, GenericGaugeVec, Opts},
     proto::MetricFamily,
@@ -293,17 +292,14 @@ impl WalStorageMetrics {
     }
 }
 
-/// Accepts async function that returns empty anyhow result, and returns the duration of its execution.
-pub async fn time_io_closure<E: Into<anyhow::Error>>(
-    closure: impl Future<Output = Result<(), E>>,
-) -> Result<f64> {
+/// Accepts a closure that returns a result, and returns the duration of the closure.
+pub fn time_io_closure(closure: impl FnOnce() -> Result<()>) -> Result<f64> {
     let start = std::time::Instant::now();
-    closure.await.map_err(|e| e.into())?;
+    closure()?;
     Ok(start.elapsed().as_secs_f64())
 }
 
 /// Metrics for a single timeline.
-#[derive(Clone)]
 pub struct FullTimelineInfo {
     pub ttid: TenantTimelineId,
     pub ps_feedback: PageserverFeedback,
@@ -579,19 +575,13 @@ impl Collector for TimelineCollector {
         let timelines = GlobalTimelines::get_all();
         let timelines_count = timelines.len();
 
-        // Prometheus Collector is sync, and data is stored under async lock. To
-        // bridge the gap with a crutch, collect data in spawned thread with
-        // local tokio runtime.
-        let infos = std::thread::spawn(|| {
-            let rt = tokio::runtime::Builder::new_current_thread()
-                .build()
-                .expect("failed to create rt");
-            rt.block_on(collect_timeline_metrics())
-        })
-        .join()
-        .expect("collect_timeline_metrics thread panicked");
+        for arc_tli in timelines {
+            let tli = arc_tli.info_for_metrics();
+            if tli.is_none() {
+                continue;
+            }
+            let tli = tli.unwrap();
 
-        for tli in &infos {
             let tenant_id = tli.ttid.tenant_id.to_string();
             let timeline_id = tli.ttid.timeline_id.to_string();
             let labels = &[tenant_id.as_str(), timeline_id.as_str()];
@@ -692,15 +682,3 @@ impl Collector for TimelineCollector {
         mfs
     }
 }
-
-async fn collect_timeline_metrics() -> Vec<FullTimelineInfo> {
-    let mut res = vec![];
-    let timelines = GlobalTimelines::get_all();
-
-    for tli in timelines {
-        if let Some(info) = tli.info_for_metrics().await {
-            res.push(info);
-        }
-    }
-    res
-}

safekeeper/src/pull_timeline.rs

@@ -231,7 +231,7 @@ async fn pull_timeline(status: TimelineStatus, host: String) -> Result<Response>
     info!(
         "Loaded timeline {}, flush_lsn={}",
         ttid,
-        tli.get_flush_lsn().await
+        tli.get_flush_lsn()
     );
 
     Ok(Response {

safekeeper/src/receive_wal.rs

@@ -18,14 +18,15 @@ use postgres_backend::QueryError;
 use pq_proto::BeMessage;
 use std::net::SocketAddr;
 use std::sync::Arc;
+use std::thread;
+use std::thread::JoinHandle;
 use tokio::io::AsyncRead;
 use tokio::io::AsyncWrite;
 use tokio::sync::mpsc::channel;
 use tokio::sync::mpsc::error::TryRecvError;
 use tokio::sync::mpsc::Receiver;
 use tokio::sync::mpsc::Sender;
-use tokio::task;
-use tokio::task::JoinHandle;
+use tokio::task::spawn_blocking;
 use tokio::time::Duration;
 use tokio::time::Instant;
 use tracing::*;
@@ -96,7 +97,7 @@ impl SafekeeperPostgresHandler {
                 Err(res.expect_err("no error with WalAcceptor not spawn"))
             }
             Some(handle) => {
-                let wal_acceptor_res = handle.await;
+                let wal_acceptor_res = handle.join();
 
                 // If there was any network error, return it.
                 res?;
@@ -106,7 +107,7 @@ impl SafekeeperPostgresHandler {
                     Ok(Ok(_)) => Ok(()), // can't happen currently; would be if we add graceful termination
                     Ok(Err(e)) => Err(CopyStreamHandlerEnd::Other(e.context("WAL acceptor"))),
                     Err(_) => Err(CopyStreamHandlerEnd::Other(anyhow!(
-                        "WalAcceptor task panicked",
+                        "WalAcceptor thread panicked",
                     ))),
                 }
             }
@@ -153,12 +154,10 @@ impl<'a, IO: AsyncRead + AsyncWrite + Unpin> NetworkReader<'a, IO> {
             }
         };
 
-        *self.acceptor_handle = Some(WalAcceptor::spawn(
-            tli.clone(),
-            msg_rx,
-            reply_tx,
-            self.conn_id,
-        ));
+        *self.acceptor_handle = Some(
+            WalAcceptor::spawn(tli.clone(), msg_rx, reply_tx, self.conn_id)
+                .context("spawn WalAcceptor thread")?,
+        );
 
         // Forward all messages to WalAcceptor
         read_network_loop(self.pgb_reader, msg_tx, next_msg).await
@@ -227,19 +226,28 @@ impl WalAcceptor {
         msg_rx: Receiver<ProposerAcceptorMessage>,
         reply_tx: Sender<AcceptorProposerMessage>,
         conn_id: ConnectionId,
-    ) -> JoinHandle<anyhow::Result<()>> {
-        task::spawn(async move {
-            let mut wa = WalAcceptor {
-                tli,
-                msg_rx,
-                reply_tx,
-            };
+    ) -> anyhow::Result<JoinHandle<anyhow::Result<()>>> {
+        let thread_name = format!("WAL acceptor {}", tli.ttid);
+        thread::Builder::new()
+            .name(thread_name)
+            .spawn(move || -> anyhow::Result<()> {
+                let mut wa = WalAcceptor {
+                    tli,
+                    msg_rx,
+                    reply_tx,
+                };
 
-            let span_ttid = wa.tli.ttid; // satisfy borrow checker
-            wa.run()
-                .instrument(info_span!("WAL acceptor", cid = %conn_id, ttid = %span_ttid))
-                .await
-        })
+                let runtime = tokio::runtime::Builder::new_current_thread()
+                    .enable_all()
+                    .build()?;
+
+                let span_ttid = wa.tli.ttid; // satisfy borrow checker
+                runtime.block_on(
+                    wa.run()
+                        .instrument(info_span!("WAL acceptor", cid = %conn_id, ttid = %span_ttid)),
+                )
+            })
+            .map_err(anyhow::Error::from)
     }
 
     /// The main loop. Returns Ok(()) if either msg_rx or reply_tx got closed;
@@ -273,7 +281,7 @@ impl WalAcceptor {
                 while let ProposerAcceptorMessage::AppendRequest(append_request) = next_msg {
                     let noflush_msg = ProposerAcceptorMessage::NoFlushAppendRequest(append_request);
 
-                    if let Some(reply) = self.tli.process_msg(&noflush_msg).await? {
+                    if let Some(reply) = self.tli.process_msg(&noflush_msg)? {
                         if self.reply_tx.send(reply).await.is_err() {
                             return Ok(()); // chan closed, streaming terminated
                         }
@@ -292,12 +300,10 @@ impl WalAcceptor {
                 }
 
                 // flush all written WAL to the disk
-                self.tli
-                    .process_msg(&ProposerAcceptorMessage::FlushWAL)
-                    .await?
+                self.tli.process_msg(&ProposerAcceptorMessage::FlushWAL)?
             } else {
                 // process message other than AppendRequest
-                self.tli.process_msg(&next_msg).await?
+                self.tli.process_msg(&next_msg)?
             };
 
             if let Some(reply) = reply_msg {
@@ -320,8 +326,8 @@ impl Drop for ComputeConnectionGuard {
         let tli = self.timeline.clone();
         // tokio forbids to call blocking_send inside the runtime, and see
         // comments in on_compute_disconnect why we call blocking_send.
-        tokio::spawn(async move {
-            if let Err(e) = tli.on_compute_disconnect().await {
+        spawn_blocking(move || {
+            if let Err(e) = tli.on_compute_disconnect() {
                 error!("failed to unregister compute connection: {}", e);
             }
         });

safekeeper/src/remove_wal.rs

@@ -1,36 +1,29 @@
 //! Thread removing old WAL.
 
-use std::time::Duration;
+use std::{thread, time::Duration};
 
-use tokio::time::sleep;
 use tracing::*;
 
 use crate::{GlobalTimelines, SafeKeeperConf};
 
-pub async fn task_main(conf: SafeKeeperConf) -> anyhow::Result<()> {
+pub fn thread_main(conf: SafeKeeperConf) {
     let wal_removal_interval = Duration::from_millis(5000);
     loop {
         let tlis = GlobalTimelines::get_all();
         for tli in &tlis {
-            if !tli.is_active().await {
+            if !tli.is_active() {
                 continue;
             }
             let ttid = tli.ttid;
-            if let Err(e) = tli
-                .maybe_persist_control_file()
-                .instrument(info_span!("", tenant = %ttid.tenant_id, timeline = %ttid.timeline_id))
-                .await
-            {
+            let _enter =
+                info_span!("", tenant = %ttid.tenant_id, timeline = %ttid.timeline_id).entered();
+            if let Err(e) = tli.maybe_pesist_control_file() {
                 warn!("failed to persist control file: {e}");
             }
-            if let Err(e) = tli
-                .remove_old_wal(conf.wal_backup_enabled)
-                .instrument(info_span!("", tenant = %ttid.tenant_id, timeline = %ttid.timeline_id))
-                .await
-            {
-                error!("failed to remove WAL: {}", e);
+            if let Err(e) = tli.remove_old_wal(conf.wal_backup_enabled) {
+                warn!("failed to remove WAL: {}", e);
             }
         }
-        sleep(wal_removal_interval).await;
+        thread::sleep(wal_removal_interval)
     }
 }

safekeeper/src/safekeeper.rs

@@ -568,27 +568,25 @@ where
 
     /// Process message from proposer and possibly form reply. Concurrent
     /// callers must exclude each other.
-    pub async fn process_msg(
+    pub fn process_msg(
         &mut self,
         msg: &ProposerAcceptorMessage,
     ) -> Result<Option<AcceptorProposerMessage>> {
         match msg {
-            ProposerAcceptorMessage::Greeting(msg) => self.handle_greeting(msg).await,
-            ProposerAcceptorMessage::VoteRequest(msg) => self.handle_vote_request(msg).await,
-            ProposerAcceptorMessage::Elected(msg) => self.handle_elected(msg).await,
-            ProposerAcceptorMessage::AppendRequest(msg) => {
-                self.handle_append_request(msg, true).await
-            }
+            ProposerAcceptorMessage::Greeting(msg) => self.handle_greeting(msg),
+            ProposerAcceptorMessage::VoteRequest(msg) => self.handle_vote_request(msg),
+            ProposerAcceptorMessage::Elected(msg) => self.handle_elected(msg),
+            ProposerAcceptorMessage::AppendRequest(msg) => self.handle_append_request(msg, true),
             ProposerAcceptorMessage::NoFlushAppendRequest(msg) => {
-                self.handle_append_request(msg, false).await
+                self.handle_append_request(msg, false)
             }
-            ProposerAcceptorMessage::FlushWAL => self.handle_flush().await,
+            ProposerAcceptorMessage::FlushWAL => self.handle_flush(),
         }
     }
 
     /// Handle initial message from proposer: check its sanity and send my
     /// current term.
-    async fn handle_greeting(
+    fn handle_greeting(
         &mut self,
         msg: &ProposerGreeting,
     ) -> Result<Option<AcceptorProposerMessage>> {
@@ -651,7 +649,7 @@ where
             if msg.pg_version != UNKNOWN_SERVER_VERSION {
                 state.server.pg_version = msg.pg_version;
             }
-            self.state.persist(&state).await?;
+            self.state.persist(&state)?;
         }
 
         info!(
@@ -666,7 +664,7 @@ where
     }
 
     /// Give vote for the given term, if we haven't done that previously.
-    async fn handle_vote_request(
+    fn handle_vote_request(
         &mut self,
         msg: &VoteRequest,
     ) -> Result<Option<AcceptorProposerMessage>> {
@@ -680,7 +678,7 @@ where
         // handle_elected instead. Currently not a big deal, as proposer is the
         // only source of WAL; with peer2peer recovery it would be more
         // important.
-        self.wal_store.flush_wal().await?;
+        self.wal_store.flush_wal()?;
         // initialize with refusal
         let mut resp = VoteResponse {
             term: self.state.acceptor_state.term,
@@ -694,7 +692,7 @@ where
             let mut state = self.state.clone();
             state.acceptor_state.term = msg.term;
             // persist vote before sending it out
-            self.state.persist(&state).await?;
+            self.state.persist(&state)?;
 
             resp.term = self.state.acceptor_state.term;
             resp.vote_given = true as u64;
@@ -717,15 +715,12 @@ where
         ar
     }
 
-    async fn handle_elected(
-        &mut self,
-        msg: &ProposerElected,
-    ) -> Result<Option<AcceptorProposerMessage>> {
+    fn handle_elected(&mut self, msg: &ProposerElected) -> Result<Option<AcceptorProposerMessage>> {
         info!("received ProposerElected {:?}", msg);
         if self.state.acceptor_state.term < msg.term {
             let mut state = self.state.clone();
             state.acceptor_state.term = msg.term;
-            self.state.persist(&state).await?;
+            self.state.persist(&state)?;
         }
 
         // If our term is higher, ignore the message (next feedback will inform the compute)
@@ -755,7 +750,7 @@ where
         // intersection of our history and history from msg
 
         // truncate wal, update the LSNs
-        self.wal_store.truncate_wal(msg.start_streaming_at).await?;
+        self.wal_store.truncate_wal(msg.start_streaming_at)?;
 
         // and now adopt term history from proposer
         {
@@ -789,7 +784,7 @@ where
             self.inmem.backup_lsn = max(self.inmem.backup_lsn, state.timeline_start_lsn);
 
             state.acceptor_state.term_history = msg.term_history.clone();
-            self.persist_control_file(state).await?;
+            self.persist_control_file(state)?;
         }
 
         info!("start receiving WAL since {:?}", msg.start_streaming_at);
@@ -801,7 +796,7 @@ where
     ///
     /// Note: it is assumed that 'WAL we have is from the right term' check has
     /// already been done outside.
-    async fn update_commit_lsn(&mut self, mut candidate: Lsn) -> Result<()> {
+    fn update_commit_lsn(&mut self, mut candidate: Lsn) -> Result<()> {
         // Both peers and walproposer communicate this value, we might already
         // have a fresher (higher) version.
         candidate = max(candidate, self.inmem.commit_lsn);
@@ -823,32 +818,29 @@ where
         // that we receive new epoch_start_lsn, and we still need to sync
         // control file in this case.
         if commit_lsn == self.epoch_start_lsn && self.state.commit_lsn != commit_lsn {
-            self.persist_control_file(self.state.clone()).await?;
+            self.persist_control_file(self.state.clone())?;
         }
 
         Ok(())
     }
 
     /// Persist control file to disk, called only after timeline creation (bootstrap).
-    pub async fn persist(&mut self) -> Result<()> {
-        self.persist_control_file(self.state.clone()).await
+    pub fn persist(&mut self) -> Result<()> {
+        self.persist_control_file(self.state.clone())
     }
 
     /// Persist in-memory state to the disk, taking other data from state.
-    async fn persist_control_file(&mut self, mut state: SafeKeeperState) -> Result<()> {
+    fn persist_control_file(&mut self, mut state: SafeKeeperState) -> Result<()> {
         state.commit_lsn = self.inmem.commit_lsn;
         state.backup_lsn = self.inmem.backup_lsn;
         state.peer_horizon_lsn = self.inmem.peer_horizon_lsn;
         state.proposer_uuid = self.inmem.proposer_uuid;
-        self.state.persist(&state).await
+        self.state.persist(&state)
     }
 
     /// Persist control file if there is something to save and enough time
     /// passed after the last save.
-    pub async fn maybe_persist_control_file(
-        &mut self,
-        inmem_remote_consistent_lsn: Lsn,
-    ) -> Result<()> {
+    pub fn maybe_persist_control_file(&mut self, inmem_remote_consistent_lsn: Lsn) -> Result<()> {
         const CF_SAVE_INTERVAL: Duration = Duration::from_secs(300);
         if self.state.last_persist_at().elapsed() < CF_SAVE_INTERVAL {
             return Ok(());
@@ -860,7 +852,7 @@ where
         if need_persist {
             let mut state = self.state.clone();
             state.remote_consistent_lsn = inmem_remote_consistent_lsn;
-            self.persist_control_file(state).await?;
+            self.persist_control_file(state)?;
             trace!("saved control file: {CF_SAVE_INTERVAL:?} passed");
         }
         Ok(())
@@ -868,7 +860,7 @@ where
 
     /// Handle request to append WAL.
     #[allow(clippy::comparison_chain)]
-    async fn handle_append_request(
+    fn handle_append_request(
         &mut self,
         msg: &AppendRequest,
         require_flush: bool,
@@ -891,19 +883,17 @@ where
 
         // do the job
         if !msg.wal_data.is_empty() {
-            self.wal_store
-                .write_wal(msg.h.begin_lsn, &msg.wal_data)
-                .await?;
+            self.wal_store.write_wal(msg.h.begin_lsn, &msg.wal_data)?;
         }
 
         // flush wal to the disk, if required
         if require_flush {
-            self.wal_store.flush_wal().await?;
+            self.wal_store.flush_wal()?;
         }
 
         // Update commit_lsn.
         if msg.h.commit_lsn != Lsn(0) {
-            self.update_commit_lsn(msg.h.commit_lsn).await?;
+            self.update_commit_lsn(msg.h.commit_lsn)?;
         }
         // Value calculated by walproposer can always lag:
         // - safekeepers can forget inmem value and send to proposer lower
@@ -919,7 +909,7 @@ where
         if self.state.peer_horizon_lsn + (self.state.server.wal_seg_size as u64)
             < self.inmem.peer_horizon_lsn
         {
-            self.persist_control_file(self.state.clone()).await?;
+            self.persist_control_file(self.state.clone())?;
         }
 
         trace!(
@@ -941,15 +931,15 @@ where
     }
 
     /// Flush WAL to disk. Return AppendResponse with latest LSNs.
-    async fn handle_flush(&mut self) -> Result<Option<AcceptorProposerMessage>> {
-        self.wal_store.flush_wal().await?;
+    fn handle_flush(&mut self) -> Result<Option<AcceptorProposerMessage>> {
+        self.wal_store.flush_wal()?;
         Ok(Some(AcceptorProposerMessage::AppendResponse(
             self.append_response(),
         )))
     }
 
     /// Update timeline state with peer safekeeper data.
-    pub async fn record_safekeeper_info(&mut self, sk_info: &SafekeeperTimelineInfo) -> Result<()> {
+    pub fn record_safekeeper_info(&mut self, sk_info: &SafekeeperTimelineInfo) -> Result<()> {
         let mut sync_control_file = false;
 
         if (Lsn(sk_info.commit_lsn) != Lsn::INVALID) && (sk_info.last_log_term != INVALID_TERM) {
@@ -957,7 +947,7 @@ where
             // commit_lsn if our history matches (is part of) history of advanced
             // commit_lsn provider.
             if sk_info.last_log_term == self.get_epoch() {
-                self.update_commit_lsn(Lsn(sk_info.commit_lsn)).await?;
+                self.update_commit_lsn(Lsn(sk_info.commit_lsn))?;
             }
         }
 
@@ -983,7 +973,7 @@ where
             // Note: we could make remote_consistent_lsn update in cf common by
             // storing Arc to walsenders in Safekeeper.
             state.remote_consistent_lsn = new_remote_consistent_lsn;
-            self.persist_control_file(state).await?;
+            self.persist_control_file(state)?;
         }
         Ok(())
     }
@@ -1007,7 +997,6 @@ where
 
 #[cfg(test)]
 mod tests {
-    use futures::future::BoxFuture;
     use postgres_ffi::WAL_SEGMENT_SIZE;
 
     use super::*;
@@ -1019,9 +1008,8 @@ mod tests {
         persisted_state: SafeKeeperState,
     }
 
-    #[async_trait::async_trait]
     impl control_file::Storage for InMemoryState {
-        async fn persist(&mut self, s: &SafeKeeperState) -> Result<()> {
+        fn persist(&mut self, s: &SafeKeeperState) -> Result<()> {
             self.persisted_state = s.clone();
             Ok(())
         }
@@ -1051,28 +1039,27 @@ mod tests {
         lsn: Lsn,
     }
 
-    #[async_trait::async_trait]
     impl wal_storage::Storage for DummyWalStore {
         fn flush_lsn(&self) -> Lsn {
             self.lsn
         }
 
-        async fn write_wal(&mut self, startpos: Lsn, buf: &[u8]) -> Result<()> {
+        fn write_wal(&mut self, startpos: Lsn, buf: &[u8]) -> Result<()> {
             self.lsn = startpos + buf.len() as u64;
             Ok(())
         }
 
-        async fn truncate_wal(&mut self, end_pos: Lsn) -> Result<()> {
+        fn truncate_wal(&mut self, end_pos: Lsn) -> Result<()> {
             self.lsn = end_pos;
             Ok(())
         }
 
-        async fn flush_wal(&mut self) -> Result<()> {
+        fn flush_wal(&mut self) -> Result<()> {
             Ok(())
         }
 
-        fn remove_up_to(&self, _segno_up_to: XLogSegNo) -> BoxFuture<'static, anyhow::Result<()>> {
-            Box::pin(async { Ok(()) })
+        fn remove_up_to(&self) -> Box<dyn Fn(XLogSegNo) -> Result<()>> {
+            Box::new(move |_segno_up_to: XLogSegNo| Ok(()))
         }
 
         fn get_metrics(&self) -> crate::metrics::WalStorageMetrics {
@@ -1080,8 +1067,8 @@ mod tests {
         }
     }
 
-    #[tokio::test]
-    async fn test_voting() {
+    #[test]
+    fn test_voting() {
         let storage = InMemoryState {
             persisted_state: test_sk_state(),
         };
@@ -1090,7 +1077,7 @@ mod tests {
 
         // check voting for 1 is ok
         let vote_request = ProposerAcceptorMessage::VoteRequest(VoteRequest { term: 1 });
-        let mut vote_resp = sk.process_msg(&vote_request).await;
+        let mut vote_resp = sk.process_msg(&vote_request);
         match vote_resp.unwrap() {
             Some(AcceptorProposerMessage::VoteResponse(resp)) => assert!(resp.vote_given != 0),
             r => panic!("unexpected response: {:?}", r),
@@ -1105,15 +1092,15 @@ mod tests {
         sk = SafeKeeper::new(storage, sk.wal_store, NodeId(0)).unwrap();
 
         // and ensure voting second time for 1 is not ok
-        vote_resp = sk.process_msg(&vote_request).await;
+        vote_resp = sk.process_msg(&vote_request);
         match vote_resp.unwrap() {
             Some(AcceptorProposerMessage::VoteResponse(resp)) => assert!(resp.vote_given == 0),
             r => panic!("unexpected response: {:?}", r),
         }
     }
 
-    #[tokio::test]
-    async fn test_epoch_switch() {
+    #[test]
+    fn test_epoch_switch() {
         let storage = InMemoryState {
             persisted_state: test_sk_state(),
         };
@@ -1145,13 +1132,10 @@ mod tests {
             timeline_start_lsn: Lsn(0),
         };
         sk.process_msg(&ProposerAcceptorMessage::Elected(pem))
-            .await
             .unwrap();
 
         // check that AppendRequest before epochStartLsn doesn't switch epoch
-        let resp = sk
-            .process_msg(&ProposerAcceptorMessage::AppendRequest(append_request))
-            .await;
+        let resp = sk.process_msg(&ProposerAcceptorMessage::AppendRequest(append_request));
         assert!(resp.is_ok());
         assert_eq!(sk.get_epoch(), 0);
 
@@ -1162,11 +1146,9 @@ mod tests {
             h: ar_hdr,
             wal_data: Bytes::from_static(b"b"),
         };
-        let resp = sk
-            .process_msg(&ProposerAcceptorMessage::AppendRequest(append_request))
-            .await;
+        let resp = sk.process_msg(&ProposerAcceptorMessage::AppendRequest(append_request));
         assert!(resp.is_ok());
-        sk.wal_store.truncate_wal(Lsn(3)).await.unwrap(); // imitate the complete record at 3 %)
+        sk.wal_store.truncate_wal(Lsn(3)).unwrap(); // imitate the complete record at 3 %)
         assert_eq!(sk.get_epoch(), 1);
     }
 }

safekeeper/src/send_wal.rs

@@ -396,7 +396,7 @@ impl SafekeeperPostgresHandler {
         // on this safekeeper itself. That's ok as (old) proposer will never be
         // able to commit such WAL.
         let stop_pos: Option<Lsn> = if self.is_walproposer_recovery() {
-            let wal_end = tli.get_flush_lsn().await;
+            let wal_end = tli.get_flush_lsn();
             Some(wal_end)
         } else {
             None
@@ -418,7 +418,7 @@ impl SafekeeperPostgresHandler {
         // switch to copy
         pgb.write_message(&BeMessage::CopyBothResponse).await?;
 
-        let (_, persisted_state) = tli.get_state().await;
+        let (_, persisted_state) = tli.get_state();
         let wal_reader = WalReader::new(
             self.conf.workdir.clone(),
             self.conf.timeline_dir(&tli.ttid),
@@ -562,7 +562,7 @@ impl<IO: AsyncRead + AsyncWrite + Unpin> WalSender<'_, IO> {
                 .walsenders
                 .get_ws_remote_consistent_lsn(self.ws_guard.id)
             {
-                if self.tli.should_walsender_stop(remote_consistent_lsn).await {
+                if self.tli.should_walsender_stop(remote_consistent_lsn) {
                     // Terminate if there is nothing more to send.
                     return Err(CopyStreamHandlerEnd::ServerInitiated(format!(
                         "ending streaming to {:?} at {}, receiver is caughtup and there is no computes",

safekeeper/src/timeline.rs

@@ -2,13 +2,12 @@
 //! to glue together SafeKeeper and all other background services.
 
 use anyhow::{anyhow, bail, Result};
+use parking_lot::{Mutex, MutexGuard};
 use postgres_ffi::XLogSegNo;
-use tokio::fs;
 
 use std::cmp::max;
 use std::path::PathBuf;
 use std::sync::Arc;
-use tokio::sync::{Mutex, MutexGuard};
 use tokio::{
     sync::{mpsc::Sender, watch},
     time::Instant,
@@ -266,7 +265,7 @@ impl From<TimelineError> for ApiError {
     fn from(te: TimelineError) -> ApiError {
         match te {
             TimelineError::NotFound(ttid) => {
-                ApiError::NotFound(anyhow!("timeline {} not found", ttid).into())
+                ApiError::NotFound(anyhow!("timeline {} not found", ttid))
             }
             _ => ApiError::InternalServerError(anyhow!("{}", te)),
         }
@@ -287,9 +286,8 @@ pub struct Timeline {
     commit_lsn_watch_tx: watch::Sender<Lsn>,
     commit_lsn_watch_rx: watch::Receiver<Lsn>,
 
-    /// Safekeeper and other state, that should remain consistent and
-    /// synchronized with the disk. This is tokio mutex as we write WAL to disk
-    /// while holding it, ensuring that consensus checks are in order.
+    /// Safekeeper and other state, that should remain consistent and synchronized
+    /// with the disk.
     mutex: Mutex<SharedState>,
     walsenders: Arc<WalSenders>,
 
@@ -363,8 +361,8 @@ impl Timeline {
     ///
     /// Bootstrap is transactional, so if it fails, created files will be deleted,
     /// and state on disk should remain unchanged.
-    pub async fn bootstrap(&self, shared_state: &mut MutexGuard<'_, SharedState>) -> Result<()> {
-        match fs::metadata(&self.timeline_dir).await {
+    pub fn bootstrap(&self, shared_state: &mut MutexGuard<SharedState>) -> Result<()> {
+        match std::fs::metadata(&self.timeline_dir) {
             Ok(_) => {
                 // Timeline directory exists on disk, we should leave state unchanged
                 // and return error.
@@ -377,51 +375,53 @@ impl Timeline {
         }
 
         // Create timeline directory.
-        fs::create_dir_all(&self.timeline_dir).await?;
+        std::fs::create_dir_all(&self.timeline_dir)?;
 
         // Write timeline to disk and TODO: start background tasks.
-        if let Err(e) = shared_state.sk.persist().await {
-            // Bootstrap failed, cancel timeline and remove timeline directory.
-            self.cancel(shared_state);
+        match || -> Result<()> {
+            shared_state.sk.persist()?;
+            // TODO: add more initialization steps here
+            self.update_status(shared_state);
+            Ok(())
+        }() {
+            Ok(_) => Ok(()),
+            Err(e) => {
+                // Bootstrap failed, cancel timeline and remove timeline directory.
+                self.cancel(shared_state);
 
-            if let Err(fs_err) = fs::remove_dir_all(&self.timeline_dir).await {
-                warn!(
-                    "failed to remove timeline {} directory after bootstrap failure: {}",
-                    self.ttid, fs_err
-                );
+                if let Err(fs_err) = std::fs::remove_dir_all(&self.timeline_dir) {
+                    warn!(
+                        "failed to remove timeline {} directory after bootstrap failure: {}",
+                        self.ttid, fs_err
+                    );
+                }
+
+                Err(e)
             }
-
-            return Err(e);
         }
-
-        // TODO: add more initialization steps here
-        self.update_status(shared_state);
-        Ok(())
     }
 
     /// Delete timeline from disk completely, by removing timeline directory. Background
     /// timeline activities will stop eventually.
-    pub async fn delete_from_disk(
+    pub fn delete_from_disk(
         &self,
-        shared_state: &mut MutexGuard<'_, SharedState>,
+        shared_state: &mut MutexGuard<SharedState>,
     ) -> Result<(bool, bool)> {
         let was_active = shared_state.active;
         self.cancel(shared_state);
-        let dir_existed = delete_dir(&self.timeline_dir).await?;
+        let dir_existed = delete_dir(&self.timeline_dir)?;
         Ok((dir_existed, was_active))
     }
 
     /// Cancel timeline to prevent further usage. Background tasks will stop
     /// eventually after receiving cancellation signal.
-    ///
-    /// Note that we can't notify backup launcher here while holding
-    /// shared_state lock, as this is a potential deadlock: caller is
-    /// responsible for that. Generally we should probably make WAL backup tasks
-    /// to shut down on their own, checking once in a while whether it is the
-    /// time.
-    fn cancel(&self, shared_state: &mut MutexGuard<'_, SharedState>) {
+    fn cancel(&self, shared_state: &mut MutexGuard<SharedState>) {
         info!("timeline {} is cancelled", self.ttid);
         let _ = self.cancellation_tx.send(true);
+        let res = self.wal_backup_launcher_tx.blocking_send(self.ttid);
+        if let Err(e) = res {
+            error!("Failed to send stop signal to wal_backup_launcher: {}", e);
+        }
         // Close associated FDs. Nobody will be able to touch timeline data once
         // it is cancelled, so WAL storage won't be opened again.
         shared_state.sk.wal_store.close();
@@ -433,8 +433,8 @@ impl Timeline {
     }
 
     /// Take a writing mutual exclusive lock on timeline shared_state.
-    pub async fn write_shared_state(&self) -> MutexGuard<SharedState> {
-        self.mutex.lock().await
+    pub fn write_shared_state(&self) -> MutexGuard<SharedState> {
+        self.mutex.lock()
     }
 
     fn update_status(&self, shared_state: &mut SharedState) -> bool {
@@ -450,7 +450,7 @@ impl Timeline {
 
         let is_wal_backup_action_pending: bool;
         {
-            let mut shared_state = self.write_shared_state().await;
+            let mut shared_state = self.write_shared_state();
             shared_state.num_computes += 1;
             is_wal_backup_action_pending = self.update_status(&mut shared_state);
         }
@@ -464,17 +464,22 @@ impl Timeline {
 
     /// De-register compute connection, shutting down timeline activity if
     /// pageserver doesn't need catchup.
-    pub async fn on_compute_disconnect(&self) -> Result<()> {
+    pub fn on_compute_disconnect(&self) -> Result<()> {
         let is_wal_backup_action_pending: bool;
         {
-            let mut shared_state = self.write_shared_state().await;
+            let mut shared_state = self.write_shared_state();
             shared_state.num_computes -= 1;
             is_wal_backup_action_pending = self.update_status(&mut shared_state);
         }
         // Wake up wal backup launcher, if it is time to stop the offloading.
         if is_wal_backup_action_pending {
             // Can fail only if channel to a static thread got closed, which is not normal at all.
-            self.wal_backup_launcher_tx.send(self.ttid).await?;
+            //
+            // Note: this is blocking_send because on_compute_disconnect is called in Drop, there is
+            // no async Drop and we use current thread runtimes. With current thread rt spawning
+            // task in drop impl is racy, as thread along with runtime might finish before the task.
+            // This should be switched send.await when/if we go to full async.
+            self.wal_backup_launcher_tx.blocking_send(self.ttid)?;
         }
         Ok(())
     }
@@ -484,11 +489,11 @@ impl Timeline {
     /// computes. While there might be nothing to stream already, we learn about
     /// remote_consistent_lsn update through replication feedback, and we want
     /// to stop pushing to the broker if pageserver is fully caughtup.
-    pub async fn should_walsender_stop(&self, reported_remote_consistent_lsn: Lsn) -> bool {
+    pub fn should_walsender_stop(&self, reported_remote_consistent_lsn: Lsn) -> bool {
         if self.is_cancelled() {
             return true;
         }
-        let shared_state = self.write_shared_state().await;
+        let shared_state = self.write_shared_state();
         if shared_state.num_computes == 0 {
             return shared_state.sk.inmem.commit_lsn == Lsn(0) || // no data at all yet
             reported_remote_consistent_lsn >= shared_state.sk.inmem.commit_lsn;
@@ -498,12 +503,12 @@ impl Timeline {
 
     /// Returns whether s3 offloading is required and sets current status as
     /// matching it.
-    pub async fn wal_backup_attend(&self) -> bool {
+    pub fn wal_backup_attend(&self) -> bool {
         if self.is_cancelled() {
             return false;
         }
 
-        self.write_shared_state().await.wal_backup_attend()
+        self.write_shared_state().wal_backup_attend()
     }
 
     /// Returns commit_lsn watch channel.
@@ -512,7 +517,7 @@ impl Timeline {
     }
 
     /// Pass arrived message to the safekeeper.
-    pub async fn process_msg(
+    pub fn process_msg(
         &self,
         msg: &ProposerAcceptorMessage,
     ) -> Result<Option<AcceptorProposerMessage>> {
@@ -523,8 +528,8 @@ impl Timeline {
         let mut rmsg: Option<AcceptorProposerMessage>;
         let commit_lsn: Lsn;
         {
-            let mut shared_state = self.write_shared_state().await;
-            rmsg = shared_state.sk.process_msg(msg).await?;
+            let mut shared_state = self.write_shared_state();
+            rmsg = shared_state.sk.process_msg(msg)?;
 
             // if this is AppendResponse, fill in proper pageserver and hot
             // standby feedback.
@@ -541,37 +546,37 @@ impl Timeline {
     }
 
     /// Returns wal_seg_size.
-    pub async fn get_wal_seg_size(&self) -> usize {
-        self.write_shared_state().await.get_wal_seg_size()
+    pub fn get_wal_seg_size(&self) -> usize {
+        self.write_shared_state().get_wal_seg_size()
     }
 
     /// Returns true only if the timeline is loaded and active.
-    pub async fn is_active(&self) -> bool {
+    pub fn is_active(&self) -> bool {
         if self.is_cancelled() {
             return false;
         }
 
-        self.write_shared_state().await.active
+        self.write_shared_state().active
     }
 
     /// Returns state of the timeline.
-    pub async fn get_state(&self) -> (SafekeeperMemState, SafeKeeperState) {
-        let state = self.write_shared_state().await;
+    pub fn get_state(&self) -> (SafekeeperMemState, SafeKeeperState) {
+        let state = self.write_shared_state();
         (state.sk.inmem.clone(), state.sk.state.clone())
     }
 
     /// Returns latest backup_lsn.
-    pub async fn get_wal_backup_lsn(&self) -> Lsn {
-        self.write_shared_state().await.sk.inmem.backup_lsn
+    pub fn get_wal_backup_lsn(&self) -> Lsn {
+        self.write_shared_state().sk.inmem.backup_lsn
     }
 
     /// Sets backup_lsn to the given value.
-    pub async fn set_wal_backup_lsn(&self, backup_lsn: Lsn) -> Result<()> {
+    pub fn set_wal_backup_lsn(&self, backup_lsn: Lsn) -> Result<()> {
         if self.is_cancelled() {
             bail!(TimelineError::Cancelled(self.ttid));
         }
 
-        let mut state = self.write_shared_state().await;
+        let mut state = self.write_shared_state();
         state.sk.inmem.backup_lsn = max(state.sk.inmem.backup_lsn, backup_lsn);
         // we should check whether to shut down offloader, but this will be done
         // soon by peer communication anyway.
@@ -579,8 +584,8 @@ impl Timeline {
     }
 
     /// Get safekeeper info for broadcasting to broker and other peers.
-    pub async fn get_safekeeper_info(&self, conf: &SafeKeeperConf) -> SafekeeperTimelineInfo {
-        let shared_state = self.write_shared_state().await;
+    pub fn get_safekeeper_info(&self, conf: &SafeKeeperConf) -> SafekeeperTimelineInfo {
+        let shared_state = self.write_shared_state();
         shared_state.get_safekeeper_info(
             &self.ttid,
             conf,
@@ -599,8 +604,8 @@ impl Timeline {
         let is_wal_backup_action_pending: bool;
         let commit_lsn: Lsn;
         {
-            let mut shared_state = self.write_shared_state().await;
-            shared_state.sk.record_safekeeper_info(&sk_info).await?;
+            let mut shared_state = self.write_shared_state();
+            shared_state.sk.record_safekeeper_info(&sk_info)?;
             let peer_info = PeerInfo::from_sk_info(&sk_info, Instant::now());
             shared_state.peers_info.upsert(&peer_info);
             is_wal_backup_action_pending = self.update_status(&mut shared_state);
@@ -617,8 +622,8 @@ impl Timeline {
     /// Get our latest view of alive peers status on the timeline.
     /// We pass our own info through the broker as well, so when we don't have connection
     /// to the broker returned vec is empty.
-    pub async fn get_peers(&self, conf: &SafeKeeperConf) -> Vec<PeerInfo> {
-        let shared_state = self.write_shared_state().await;
+    pub fn get_peers(&self, conf: &SafeKeeperConf) -> Vec<PeerInfo> {
+        let shared_state = self.write_shared_state();
         let now = Instant::now();
         shared_state
             .peers_info
@@ -635,34 +640,34 @@ impl Timeline {
     }
 
     /// Returns flush_lsn.
-    pub async fn get_flush_lsn(&self) -> Lsn {
-        self.write_shared_state().await.sk.wal_store.flush_lsn()
+    pub fn get_flush_lsn(&self) -> Lsn {
+        self.write_shared_state().sk.wal_store.flush_lsn()
     }
 
     /// Delete WAL segments from disk that are no longer needed. This is determined
     /// based on pageserver's remote_consistent_lsn and local backup_lsn/peer_lsn.
-    pub async fn remove_old_wal(&self, wal_backup_enabled: bool) -> Result<()> {
+    pub fn remove_old_wal(&self, wal_backup_enabled: bool) -> Result<()> {
         if self.is_cancelled() {
             bail!(TimelineError::Cancelled(self.ttid));
         }
 
         let horizon_segno: XLogSegNo;
-        let remover = {
-            let shared_state = self.write_shared_state().await;
+        let remover: Box<dyn Fn(u64) -> Result<(), anyhow::Error>>;
+        {
+            let shared_state = self.write_shared_state();
             horizon_segno = shared_state.sk.get_horizon_segno(wal_backup_enabled);
+            remover = shared_state.sk.wal_store.remove_up_to();
             if horizon_segno <= 1 || horizon_segno <= shared_state.last_removed_segno {
-                return Ok(()); // nothing to do
+                return Ok(());
             }
-            let remover = shared_state.sk.wal_store.remove_up_to(horizon_segno - 1);
             // release the lock before removing
-            remover
-        };
+        }
 
         // delete old WAL files
-        remover.await?;
+        remover(horizon_segno - 1)?;
 
         // update last_removed_segno
-        let mut shared_state = self.write_shared_state().await;
+        let mut shared_state = self.write_shared_state();
         shared_state.last_removed_segno = horizon_segno;
         Ok(())
     }
@@ -671,24 +676,22 @@ impl Timeline {
     /// passed after the last save. This helps to keep remote_consistent_lsn up
     /// to date so that storage nodes restart doesn't cause many pageserver ->
     /// safekeeper reconnections.
-    pub async fn maybe_persist_control_file(&self) -> Result<()> {
+    pub fn maybe_pesist_control_file(&self) -> Result<()> {
         let remote_consistent_lsn = self.walsenders.get_remote_consistent_lsn();
         self.write_shared_state()
-            .await
             .sk
             .maybe_persist_control_file(remote_consistent_lsn)
-            .await
     }
 
-    /// Gather timeline data for metrics. If the timeline is not active, returns
-    /// None, we do not collect these.
-    pub async fn info_for_metrics(&self) -> Option<FullTimelineInfo> {
+    /// Returns full timeline info, required for the metrics. If the timeline is
+    /// not active, returns None instead.
+    pub fn info_for_metrics(&self) -> Option<FullTimelineInfo> {
         if self.is_cancelled() {
             return None;
         }
 
         let ps_feedback = self.walsenders.get_ps_feedback();
-        let state = self.write_shared_state().await;
+        let state = self.write_shared_state();
         if state.active {
             Some(FullTimelineInfo {
                 ttid: self.ttid,
@@ -710,8 +713,8 @@ impl Timeline {
     }
 
     /// Returns in-memory timeline state to build a full debug dump.
-    pub async fn memory_dump(&self) -> debug_dump::Memory {
-        let state = self.write_shared_state().await;
+    pub fn memory_dump(&self) -> debug_dump::Memory {
+        let state = self.write_shared_state();
 
         let (write_lsn, write_record_lsn, flush_lsn, file_open) =
             state.sk.wal_store.internal_state();
@@ -735,8 +738,8 @@ impl Timeline {
 }
 
 /// Deletes directory and it's contents. Returns false if directory does not exist.
-async fn delete_dir(path: &PathBuf) -> Result<bool> {
-    match fs::remove_dir_all(path).await {
+fn delete_dir(path: &PathBuf) -> Result<bool> {
+    match std::fs::remove_dir_all(path) {
         Ok(_) => Ok(true),
         Err(e) if e.kind() == std::io::ErrorKind::NotFound => Ok(false),
         Err(e) => Err(e.into()),

safekeeper/src/timelines_global_map.rs

@@ -113,17 +113,9 @@ impl GlobalTimelines {
         Ok(())
     }
 
-    /// Loads all timelines for the given tenant to memory. Returns fs::read_dir
-    /// errors if any.
-    ///
-    /// Note: This function (and all reading/loading below) is sync because
-    /// timelines are loaded while holding GlobalTimelinesState lock. Which is
-    /// fine as this is called only from single threaded main runtime on boot,
-    /// but clippy complains anyway, and suppressing that isn't trivial as async
-    /// is the keyword, ha. That only other user is pull_timeline.rs for which
-    /// being blocked is not that bad, and we can do spawn_blocking.
+    /// Loads all timelines for the given tenant to memory. Returns fs::read_dir errors if any.
     fn load_tenant_timelines(
-        state: &mut MutexGuard<'_, GlobalTimelinesState>,
+        state: &mut MutexGuard<GlobalTimelinesState>,
         tenant_id: TenantId,
     ) -> Result<()> {
         let timelines_dir = state.get_conf().tenant_dir(&tenant_id);
@@ -228,7 +220,7 @@ impl GlobalTimelines {
         // Take a lock and finish the initialization holding this mutex. No other threads
         // can interfere with creation after we will insert timeline into the map.
         {
-            let mut shared_state = timeline.write_shared_state().await;
+            let mut shared_state = timeline.write_shared_state();
 
             // We can get a race condition here in case of concurrent create calls, but only
             // in theory. create() will return valid timeline on the next try.
@@ -240,7 +232,7 @@ impl GlobalTimelines {
             // Write the new timeline to the disk and start background workers.
             // Bootstrap is transactional, so if it fails, the timeline will be deleted,
             // and the state on disk should remain unchanged.
-            if let Err(e) = timeline.bootstrap(&mut shared_state).await {
+            if let Err(e) = timeline.bootstrap(&mut shared_state) {
                 // Note: the most likely reason for bootstrap failure is that the timeline
                 // directory already exists on disk. This happens when timeline is corrupted
                 // and wasn't loaded from disk on startup because of that. We want to preserve
@@ -302,16 +294,15 @@ impl GlobalTimelines {
     }
 
     /// Cancels timeline, then deletes the corresponding data directory.
-    pub async fn delete_force(ttid: &TenantTimelineId) -> Result<TimelineDeleteForceResult> {
+    pub fn delete_force(ttid: &TenantTimelineId) -> Result<TimelineDeleteForceResult> {
         let tli_res = TIMELINES_STATE.lock().unwrap().get(ttid);
         match tli_res {
             Ok(timeline) => {
                 // Take a lock and finish the deletion holding this mutex.
-                let mut shared_state = timeline.write_shared_state().await;
+                let mut shared_state = timeline.write_shared_state();
 
                 info!("deleting timeline {}", ttid);
-                let (dir_existed, was_active) =
-                    timeline.delete_from_disk(&mut shared_state).await?;
+                let (dir_existed, was_active) = timeline.delete_from_disk(&mut shared_state)?;
 
                 // Remove timeline from the map.
                 // FIXME: re-enable it once we fix the issue with recreation of deleted timelines
@@ -344,7 +335,7 @@ impl GlobalTimelines {
     /// the tenant had, `true` if a timeline was active. There may be a race if new timelines are
     /// created simultaneously. In that case the function will return error and the caller should
     /// retry tenant deletion again later.
-    pub async fn delete_force_all_for_tenant(
+    pub fn delete_force_all_for_tenant(
         tenant_id: &TenantId,
     ) -> Result<HashMap<TenantTimelineId, TimelineDeleteForceResult>> {
         info!("deleting all timelines for tenant {}", tenant_id);
@@ -354,7 +345,7 @@ impl GlobalTimelines {
 
         let mut deleted = HashMap::new();
         for tli in &to_delete {
-            match Self::delete_force(&tli.ttid).await {
+            match Self::delete_force(&tli.ttid) {
                 Ok(result) => {
                     deleted.insert(tli.ttid, result);
                 }

safekeeper/src/wal_backup.rs

@@ -17,6 +17,7 @@ use postgres_ffi::XLogFileName;
 use postgres_ffi::{XLogSegNo, PG_TLI};
 use remote_storage::{GenericRemoteStorage, RemotePath};
 use tokio::fs::File;
+use tokio::runtime::Builder;
 
 use tokio::select;
 use tokio::sync::mpsc::{self, Receiver, Sender};
@@ -35,16 +36,30 @@ use once_cell::sync::OnceCell;
 const UPLOAD_FAILURE_RETRY_MIN_MS: u64 = 10;
 const UPLOAD_FAILURE_RETRY_MAX_MS: u64 = 5000;
 
+pub fn wal_backup_launcher_thread_main(
+    conf: SafeKeeperConf,
+    wal_backup_launcher_rx: Receiver<TenantTimelineId>,
+) {
+    let mut builder = Builder::new_multi_thread();
+    if let Some(num_threads) = conf.backup_runtime_threads {
+        builder.worker_threads(num_threads);
+    }
+    let rt = builder
+        .enable_all()
+        .build()
+        .expect("failed to create wal backup runtime");
+
+    rt.block_on(async {
+        wal_backup_launcher_main_loop(conf, wal_backup_launcher_rx).await;
+    });
+}
+
 /// Check whether wal backup is required for timeline. If yes, mark that launcher is
 /// aware of current status and return the timeline.
-async fn is_wal_backup_required(ttid: TenantTimelineId) -> Option<Arc<Timeline>> {
-    match GlobalTimelines::get(ttid).ok() {
-        Some(tli) => {
-            tli.wal_backup_attend().await;
-            Some(tli)
-        }
-        None => None,
-    }
+fn is_wal_backup_required(ttid: TenantTimelineId) -> Option<Arc<Timeline>> {
+    GlobalTimelines::get(ttid)
+        .ok()
+        .filter(|tli| tli.wal_backup_attend())
 }
 
 struct WalBackupTaskHandle {
@@ -128,8 +143,8 @@ async fn update_task(
     ttid: TenantTimelineId,
     entry: &mut WalBackupTimelineEntry,
 ) {
-    let alive_peers = entry.timeline.get_peers(conf).await;
-    let wal_backup_lsn = entry.timeline.get_wal_backup_lsn().await;
+    let alive_peers = entry.timeline.get_peers(conf);
+    let wal_backup_lsn = entry.timeline.get_wal_backup_lsn();
     let (offloader, election_dbg_str) =
         determine_offloader(&alive_peers, wal_backup_lsn, ttid, conf);
     let elected_me = Some(conf.my_id) == offloader;
@@ -168,10 +183,10 @@ const CHECK_TASKS_INTERVAL_MSEC: u64 = 1000;
 /// Sits on wal_backup_launcher_rx and starts/stops per timeline wal backup
 /// tasks. Having this in separate task simplifies locking, allows to reap
 /// panics and separate elections from offloading itself.
-pub async fn wal_backup_launcher_task_main(
+async fn wal_backup_launcher_main_loop(
     conf: SafeKeeperConf,
     mut wal_backup_launcher_rx: Receiver<TenantTimelineId>,
-) -> anyhow::Result<()> {
+) {
     info!(
         "WAL backup launcher started, remote config {:?}",
         conf.remote_storage
@@ -199,7 +214,7 @@ pub async fn wal_backup_launcher_task_main(
                 if conf.remote_storage.is_none() || !conf.wal_backup_enabled {
                     continue; /* just drain the channel and do nothing */
                 }
-                let timeline = is_wal_backup_required(ttid).await;
+                let timeline = is_wal_backup_required(ttid);
                 // do we need to do anything at all?
                 if timeline.is_some() != tasks.contains_key(&ttid) {
                     if let Some(timeline) = timeline {
@@ -254,7 +269,7 @@ async fn backup_task_main(
     let tli = res.unwrap();
 
     let mut wb = WalBackupTask {
-        wal_seg_size: tli.get_wal_seg_size().await,
+        wal_seg_size: tli.get_wal_seg_size(),
         commit_lsn_watch_rx: tli.get_commit_lsn_watch_rx(),
         timeline: tli,
         timeline_dir,
@@ -311,7 +326,7 @@ impl WalBackupTask {
                 continue; /* nothing to do, common case as we wake up on every commit_lsn bump */
             }
             // Perhaps peers advanced the position, check shmem value.
-            backup_lsn = self.timeline.get_wal_backup_lsn().await;
+            backup_lsn = self.timeline.get_wal_backup_lsn();
             if backup_lsn.segment_number(self.wal_seg_size)
                 >= commit_lsn.segment_number(self.wal_seg_size)
             {
@@ -387,7 +402,6 @@ pub async fn backup_lsn_range(
                 let new_backup_lsn = segment.end_lsn;
                 timeline
                     .set_wal_backup_lsn(new_backup_lsn)
-                    .await
                     .context("setting wal_backup_lsn")?;
                 *backup_lsn = new_backup_lsn;
             } else {

safekeeper/src/wal_service.rs

@@ -4,7 +4,7 @@
 //!
 use anyhow::{Context, Result};
 use postgres_backend::QueryError;
-use std::{future, time::Duration};
+use std::{future, thread, time::Duration};
 use tokio::net::TcpStream;
 use tokio_io_timeout::TimeoutReader;
 use tracing::*;
@@ -16,82 +16,104 @@ use crate::SafeKeeperConf;
 use postgres_backend::{AuthType, PostgresBackend};
 
 /// Accept incoming TCP connections and spawn them into a background thread.
-pub async fn task_main(
-    conf: SafeKeeperConf,
-    pg_listener: std::net::TcpListener,
-) -> anyhow::Result<()> {
-    // Tokio's from_std won't do this for us, per its comment.
-    pg_listener.set_nonblocking(true)?;
+pub fn thread_main(conf: SafeKeeperConf, pg_listener: std::net::TcpListener) {
+    let runtime = tokio::runtime::Builder::new_current_thread()
+        .enable_all()
+        .build()
+        .context("create runtime")
+        // todo catch error in main thread
+        .expect("failed to create runtime");
 
-    let listener = tokio::net::TcpListener::from_std(pg_listener)?;
-    let mut connection_count: ConnectionCount = 0;
+    runtime
+        .block_on(async move {
+            // Tokio's from_std won't do this for us, per its comment.
+            pg_listener.set_nonblocking(true)?;
+            let listener = tokio::net::TcpListener::from_std(pg_listener)?;
+            let mut connection_count: ConnectionCount = 0;
 
-    loop {
-        let (socket, peer_addr) = listener.accept().await.context("accept")?;
-        debug!("accepted connection from {}", peer_addr);
-        let conf = conf.clone();
-        let conn_id = issue_connection_id(&mut connection_count);
+            loop {
+                match listener.accept().await {
+                    Ok((socket, peer_addr)) => {
+                        debug!("accepted connection from {}", peer_addr);
+                        let conf = conf.clone();
+                        let conn_id = issue_connection_id(&mut connection_count);
 
-        tokio::spawn(async move {
-            if let Err(err) = handle_socket(socket, conf, conn_id)
-                .instrument(info_span!("", cid = %conn_id))
-                .await
-            {
-                error!("connection handler exited: {}", err);
+                        let _ = thread::Builder::new()
+                            .name("WAL service thread".into())
+                            .spawn(move || {
+                                if let Err(err) = handle_socket(socket, conf, conn_id) {
+                                    error!("connection handler exited: {}", err);
+                                }
+                            })
+                            .unwrap();
+                    }
+                    Err(e) => error!("Failed to accept connection: {}", e),
+                }
             }
-        });
-    }
+            #[allow(unreachable_code)] // hint compiler the closure return type
+            Ok::<(), anyhow::Error>(())
+        })
+        .expect("listener failed")
 }
 
-/// This is run by `task_main` above, inside a background thread.
+/// This is run by `thread_main` above, inside a background thread.
 ///
-async fn handle_socket(
+fn handle_socket(
     socket: TcpStream,
     conf: SafeKeeperConf,
     conn_id: ConnectionId,
 ) -> Result<(), QueryError> {
+    let _enter = info_span!("", cid = %conn_id).entered();
+
+    let runtime = tokio::runtime::Builder::new_current_thread()
+        .enable_all()
+        .build()?;
+
     socket.set_nodelay(true)?;
     let peer_addr = socket.peer_addr()?;
 
-    // Set timeout on reading from the socket. It prevents hanged up connection
-    // if client suddenly disappears. Note that TCP_KEEPALIVE is not enabled by
-    // default, and tokio doesn't provide ability to set it out of the box.
-    let mut socket = TimeoutReader::new(socket);
-    let wal_service_timeout = Duration::from_secs(60 * 10);
-    socket.set_timeout(Some(wal_service_timeout));
-    // pin! is here because TimeoutReader (due to storing sleep future inside)
-    // is not Unpin, and all pgbackend/framed/tokio dependencies require stream
-    // to be Unpin. Which is reasonable, as indeed something like TimeoutReader
-    // shouldn't be moved.
-    tokio::pin!(socket);
+    // TimeoutReader wants async runtime during creation.
+    runtime.block_on(async move {
+        // Set timeout on reading from the socket. It prevents hanged up connection
+        // if client suddenly disappears. Note that TCP_KEEPALIVE is not enabled by
+        // default, and tokio doesn't provide ability to set it out of the box.
+        let mut socket = TimeoutReader::new(socket);
+        let wal_service_timeout = Duration::from_secs(60 * 10);
+        socket.set_timeout(Some(wal_service_timeout));
+        // pin! is here because TimeoutReader (due to storing sleep future inside)
+        // is not Unpin, and all pgbackend/framed/tokio dependencies require stream
+        // to be Unpin. Which is reasonable, as indeed something like TimeoutReader
+        // shouldn't be moved.
+        tokio::pin!(socket);
 
-    let traffic_metrics = TrafficMetrics::new();
-    if let Some(current_az) = conf.availability_zone.as_deref() {
-        traffic_metrics.set_sk_az(current_az);
-    }
+        let traffic_metrics = TrafficMetrics::new();
+        if let Some(current_az) = conf.availability_zone.as_deref() {
+            traffic_metrics.set_sk_az(current_az);
+        }
 
-    let socket = MeasuredStream::new(
-        socket,
-        |cnt| {
-            traffic_metrics.observe_read(cnt);
-        },
-        |cnt| {
-            traffic_metrics.observe_write(cnt);
-        },
-    );
+        let socket = MeasuredStream::new(
+            socket,
+            |cnt| {
+                traffic_metrics.observe_read(cnt);
+            },
+            |cnt| {
+                traffic_metrics.observe_write(cnt);
+            },
+        );
 
-    let auth_type = match conf.auth {
-        None => AuthType::Trust,
-        Some(_) => AuthType::NeonJWT,
-    };
-    let mut conn_handler =
-        SafekeeperPostgresHandler::new(conf, conn_id, Some(traffic_metrics.clone()));
-    let pgbackend = PostgresBackend::new_from_io(socket, peer_addr, auth_type, None)?;
-    // libpq protocol between safekeeper and walproposer / pageserver
-    // We don't use shutdown.
-    pgbackend
-        .run(&mut conn_handler, future::pending::<()>)
-        .await
+        let auth_type = match conf.auth {
+            None => AuthType::Trust,
+            Some(_) => AuthType::NeonJWT,
+        };
+        let mut conn_handler =
+            SafekeeperPostgresHandler::new(conf, conn_id, Some(traffic_metrics.clone()));
+        let pgbackend = PostgresBackend::new_from_io(socket, peer_addr, auth_type, None)?;
+        // libpq protocol between safekeeper and walproposer / pageserver
+        // We don't use shutdown.
+        pgbackend
+            .run(&mut conn_handler, future::pending::<()>)
+            .await
+    })
 }
 
 /// Unique WAL service connection ids are logged in spans for observability.

safekeeper/src/wal_storage.rs

@@ -8,47 +8,54 @@
 //! Note that last file has `.partial` suffix, that's different from postgres.
 
 use anyhow::{bail, Context, Result};
-use bytes::Bytes;
-use futures::future::BoxFuture;
+use remote_storage::RemotePath;
+
+use std::io::{self, Seek, SeekFrom};
+use std::pin::Pin;
+use tokio::io::AsyncRead;
+
 use postgres_ffi::v14::xlog_utils::{IsPartialXLogFileName, IsXLogFileName, XLogFromFileName};
 use postgres_ffi::{XLogSegNo, PG_TLI};
-use remote_storage::RemotePath;
 use std::cmp::{max, min};
-use std::io::{self, SeekFrom};
+
+use bytes::Bytes;
+use std::fs::{self, remove_file, File, OpenOptions};
+use std::io::Write;
 use std::path::{Path, PathBuf};
-use std::pin::Pin;
-use tokio::fs::{self, remove_file, File, OpenOptions};
-use tokio::io::{AsyncRead, AsyncWriteExt};
-use tokio::io::{AsyncReadExt, AsyncSeekExt};
+
 use tracing::*;
 
+use utils::{id::TenantTimelineId, lsn::Lsn};
+
 use crate::metrics::{time_io_closure, WalStorageMetrics, REMOVED_WAL_SEGMENTS};
 use crate::safekeeper::SafeKeeperState;
+
 use crate::wal_backup::read_object;
 use crate::SafeKeeperConf;
-use postgres_ffi::waldecoder::WalStreamDecoder;
 use postgres_ffi::XLogFileName;
 use postgres_ffi::XLOG_BLCKSZ;
-use pq_proto::SystemId;
-use utils::{id::TenantTimelineId, lsn::Lsn};
 
-#[async_trait::async_trait]
+use postgres_ffi::waldecoder::WalStreamDecoder;
+
+use pq_proto::SystemId;
+use tokio::io::{AsyncReadExt, AsyncSeekExt};
+
 pub trait Storage {
     /// LSN of last durably stored WAL record.
     fn flush_lsn(&self) -> Lsn;
 
     /// Write piece of WAL from buf to disk, but not necessarily sync it.
-    async fn write_wal(&mut self, startpos: Lsn, buf: &[u8]) -> Result<()>;
+    fn write_wal(&mut self, startpos: Lsn, buf: &[u8]) -> Result<()>;
 
     /// Truncate WAL at specified LSN, which must be the end of WAL record.
-    async fn truncate_wal(&mut self, end_pos: Lsn) -> Result<()>;
+    fn truncate_wal(&mut self, end_pos: Lsn) -> Result<()>;
 
     /// Durably store WAL on disk, up to the last written WAL record.
-    async fn flush_wal(&mut self) -> Result<()>;
+    fn flush_wal(&mut self) -> Result<()>;
 
-    /// Remove all segments <= given segno. Returns function doing that as we
-    /// want to perform it without timeline lock.
-    fn remove_up_to(&self, segno_up_to: XLogSegNo) -> BoxFuture<'static, anyhow::Result<()>>;
+    /// Remove all segments <= given segno. Returns closure as we want to do
+    /// that without timeline lock.
+    fn remove_up_to(&self) -> Box<dyn Fn(XLogSegNo) -> Result<()>>;
 
     /// Release resources associated with the storage -- technically, close FDs.
     /// Currently we don't remove timelines until restart (#3146), so need to
@@ -98,22 +105,6 @@ pub struct PhysicalStorage {
     /// - points to write_lsn, so no seek is needed for writing
     /// - doesn't point to the end of the segment
     file: Option<File>,
-
-    /// When false, we have just initialized storage using the LSN from find_end_of_wal().
-    /// In this case, [`write_lsn`] can be less than actually written WAL on disk. In particular,
-    /// there can be a case with unexpected .partial file.
-    ///
-    /// Imagine the following:
-    /// - 000000010000000000000001
-    ///   - it was fully written, but the last record is split between 2 segments
-    ///   - after restart, find_end_of_wal() returned 0/1FFFFF0, which is in the end of this segment
-    ///   - write_lsn, write_record_lsn and flush_record_lsn were initialized to 0/1FFFFF0
-    /// - 000000010000000000000002.partial
-    ///   - it has only 1 byte written, which is not enough to make a full WAL record
-    ///
-    /// Partial segment 002 has no WAL records, and it will be removed by the next truncate_wal().
-    /// This flag will be set to true after the first truncate_wal() call.
-    is_truncated_after_restart: bool,
 }
 
 impl PhysicalStorage {
@@ -173,7 +164,6 @@ impl PhysicalStorage {
             flush_record_lsn: flush_lsn,
             decoder: WalStreamDecoder::new(write_lsn, state.server.pg_version / 10000),
             file: None,
-            is_truncated_after_restart: false,
         })
     }
 
@@ -188,37 +178,33 @@ impl PhysicalStorage {
     }
 
     /// Call fdatasync if config requires so.
-    async fn fdatasync_file(&mut self, file: &mut File) -> Result<()> {
+    fn fdatasync_file(&mut self, file: &mut File) -> Result<()> {
         if !self.conf.no_sync {
             self.metrics
-                .observe_flush_seconds(time_io_closure(file.sync_data()).await?);
+                .observe_flush_seconds(time_io_closure(|| Ok(file.sync_data()?))?);
         }
         Ok(())
     }
 
     /// Call fsync if config requires so.
-    async fn fsync_file(&mut self, file: &mut File) -> Result<()> {
+    fn fsync_file(&mut self, file: &mut File) -> Result<()> {
         if !self.conf.no_sync {
             self.metrics
-                .observe_flush_seconds(time_io_closure(file.sync_all()).await?);
+                .observe_flush_seconds(time_io_closure(|| Ok(file.sync_all()?))?);
         }
         Ok(())
     }
 
     /// Open or create WAL segment file. Caller must call seek to the wanted position.
     /// Returns `file` and `is_partial`.
-    async fn open_or_create(&mut self, segno: XLogSegNo) -> Result<(File, bool)> {
+    fn open_or_create(&mut self, segno: XLogSegNo) -> Result<(File, bool)> {
         let (wal_file_path, wal_file_partial_path) =
             wal_file_paths(&self.timeline_dir, segno, self.wal_seg_size)?;
 
         // Try to open already completed segment
-        if let Ok(file) = OpenOptions::new().write(true).open(&wal_file_path).await {
+        if let Ok(file) = OpenOptions::new().write(true).open(&wal_file_path) {
             Ok((file, false))
-        } else if let Ok(file) = OpenOptions::new()
-            .write(true)
-            .open(&wal_file_partial_path)
-            .await
-        {
+        } else if let Ok(file) = OpenOptions::new().write(true).open(&wal_file_partial_path) {
             // Try to open existing partial file
             Ok((file, true))
         } else {
@@ -227,36 +213,35 @@ impl PhysicalStorage {
                 .create(true)
                 .write(true)
                 .open(&wal_file_partial_path)
-                .await
                 .with_context(|| format!("Failed to open log file {:?}", &wal_file_path))?;
 
-            write_zeroes(&mut file, self.wal_seg_size).await?;
-            self.fsync_file(&mut file).await?;
+            write_zeroes(&mut file, self.wal_seg_size)?;
+            self.fsync_file(&mut file)?;
             Ok((file, true))
         }
     }
 
     /// Write WAL bytes, which are known to be located in a single WAL segment.
-    async fn write_in_segment(&mut self, segno: u64, xlogoff: usize, buf: &[u8]) -> Result<()> {
+    fn write_in_segment(&mut self, segno: u64, xlogoff: usize, buf: &[u8]) -> Result<()> {
         let mut file = if let Some(file) = self.file.take() {
             file
         } else {
-            let (mut file, is_partial) = self.open_or_create(segno).await?;
+            let (mut file, is_partial) = self.open_or_create(segno)?;
             assert!(is_partial, "unexpected write into non-partial segment file");
-            file.seek(SeekFrom::Start(xlogoff as u64)).await?;
+            file.seek(SeekFrom::Start(xlogoff as u64))?;
             file
         };
 
-        file.write_all(buf).await?;
+        file.write_all(buf)?;
 
         if xlogoff + buf.len() == self.wal_seg_size {
             // If we reached the end of a WAL segment, flush and close it.
-            self.fdatasync_file(&mut file).await?;
+            self.fdatasync_file(&mut file)?;
 
             // Rename partial file to completed file
             let (wal_file_path, wal_file_partial_path) =
                 wal_file_paths(&self.timeline_dir, segno, self.wal_seg_size)?;
-            fs::rename(wal_file_partial_path, wal_file_path).await?;
+            fs::rename(wal_file_partial_path, wal_file_path)?;
         } else {
             // otherwise, file can be reused later
             self.file = Some(file);
@@ -270,11 +255,11 @@ impl PhysicalStorage {
     /// be flushed separately later.
     ///
     /// Updates `write_lsn`.
-    async fn write_exact(&mut self, pos: Lsn, mut buf: &[u8]) -> Result<()> {
+    fn write_exact(&mut self, pos: Lsn, mut buf: &[u8]) -> Result<()> {
         if self.write_lsn != pos {
             // need to flush the file before discarding it
             if let Some(mut file) = self.file.take() {
-                self.fdatasync_file(&mut file).await?;
+                self.fdatasync_file(&mut file)?;
             }
 
             self.write_lsn = pos;
@@ -292,8 +277,7 @@ impl PhysicalStorage {
                 buf.len()
             };
 
-            self.write_in_segment(segno, xlogoff, &buf[..bytes_write])
-                .await?;
+            self.write_in_segment(segno, xlogoff, &buf[..bytes_write])?;
             self.write_lsn += bytes_write as u64;
             buf = &buf[bytes_write..];
         }
@@ -302,7 +286,6 @@ impl PhysicalStorage {
     }
 }
 
-#[async_trait::async_trait]
 impl Storage for PhysicalStorage {
     /// flush_lsn returns LSN of last durably stored WAL record.
     fn flush_lsn(&self) -> Lsn {
@@ -310,7 +293,7 @@ impl Storage for PhysicalStorage {
     }
 
     /// Write WAL to disk.
-    async fn write_wal(&mut self, startpos: Lsn, buf: &[u8]) -> Result<()> {
+    fn write_wal(&mut self, startpos: Lsn, buf: &[u8]) -> Result<()> {
         // Disallow any non-sequential writes, which can result in gaps or overwrites.
         // If we need to move the pointer, use truncate_wal() instead.
         if self.write_lsn > startpos {
@@ -328,7 +311,7 @@ impl Storage for PhysicalStorage {
             );
         }
 
-        let write_seconds = time_io_closure(self.write_exact(startpos, buf)).await?;
+        let write_seconds = time_io_closure(|| self.write_exact(startpos, buf))?;
         // WAL is written, updating write metrics
         self.metrics.observe_write_seconds(write_seconds);
         self.metrics.observe_write_bytes(buf.len());
@@ -357,14 +340,14 @@ impl Storage for PhysicalStorage {
         Ok(())
     }
 
-    async fn flush_wal(&mut self) -> Result<()> {
+    fn flush_wal(&mut self) -> Result<()> {
         if self.flush_record_lsn == self.write_record_lsn {
             // no need to do extra flush
             return Ok(());
         }
 
         if let Some(mut unflushed_file) = self.file.take() {
-            self.fdatasync_file(&mut unflushed_file).await?;
+            self.fdatasync_file(&mut unflushed_file)?;
             self.file = Some(unflushed_file);
         } else {
             // We have unflushed data (write_lsn != flush_lsn), but no file.
@@ -386,7 +369,7 @@ impl Storage for PhysicalStorage {
 
     /// Truncate written WAL by removing all WAL segments after the given LSN.
     /// end_pos must point to the end of the WAL record.
-    async fn truncate_wal(&mut self, end_pos: Lsn) -> Result<()> {
+    fn truncate_wal(&mut self, end_pos: Lsn) -> Result<()> {
         // Streaming must not create a hole, so truncate cannot be called on non-written lsn
         if self.write_lsn != Lsn(0) && end_pos > self.write_lsn {
             bail!(
@@ -398,51 +381,47 @@ impl Storage for PhysicalStorage {
 
         // Quick exit if nothing to do to avoid writing up to 16 MiB of zeros on
         // disk (this happens on each connect).
-        if self.is_truncated_after_restart
-            && end_pos == self.write_lsn
-            && end_pos == self.flush_record_lsn
-        {
+        if end_pos == self.write_lsn {
             return Ok(());
         }
 
         // Close previously opened file, if any
         if let Some(mut unflushed_file) = self.file.take() {
-            self.fdatasync_file(&mut unflushed_file).await?;
+            self.fdatasync_file(&mut unflushed_file)?;
         }
 
         let xlogoff = end_pos.segment_offset(self.wal_seg_size);
         let segno = end_pos.segment_number(self.wal_seg_size);
 
         // Remove all segments after the given LSN.
-        remove_segments_from_disk(&self.timeline_dir, self.wal_seg_size, |x| x > segno).await?;
+        remove_segments_from_disk(&self.timeline_dir, self.wal_seg_size, |x| x > segno)?;
 
-        let (mut file, is_partial) = self.open_or_create(segno).await?;
+        let (mut file, is_partial) = self.open_or_create(segno)?;
 
         // Fill end with zeroes
-        file.seek(SeekFrom::Start(xlogoff as u64)).await?;
-        write_zeroes(&mut file, self.wal_seg_size - xlogoff).await?;
-        self.fdatasync_file(&mut file).await?;
+        file.seek(SeekFrom::Start(xlogoff as u64))?;
+        write_zeroes(&mut file, self.wal_seg_size - xlogoff)?;
+        self.fdatasync_file(&mut file)?;
 
         if !is_partial {
             // Make segment partial once again
             let (wal_file_path, wal_file_partial_path) =
                 wal_file_paths(&self.timeline_dir, segno, self.wal_seg_size)?;
-            fs::rename(wal_file_path, wal_file_partial_path).await?;
+            fs::rename(wal_file_path, wal_file_partial_path)?;
         }
 
         // Update LSNs
         self.write_lsn = end_pos;
         self.write_record_lsn = end_pos;
         self.flush_record_lsn = end_pos;
-        self.is_truncated_after_restart = true;
         Ok(())
     }
 
-    fn remove_up_to(&self, segno_up_to: XLogSegNo) -> BoxFuture<'static, anyhow::Result<()>> {
+    fn remove_up_to(&self) -> Box<dyn Fn(XLogSegNo) -> Result<()>> {
         let timeline_dir = self.timeline_dir.clone();
         let wal_seg_size = self.wal_seg_size;
-        Box::pin(async move {
-            remove_segments_from_disk(&timeline_dir, wal_seg_size, |x| x <= segno_up_to).await
+        Box::new(move |segno_up_to: XLogSegNo| {
+            remove_segments_from_disk(&timeline_dir, wal_seg_size, |x| x <= segno_up_to)
         })
     }
 
@@ -457,7 +436,7 @@ impl Storage for PhysicalStorage {
 }
 
 /// Remove all WAL segments in timeline_dir that match the given predicate.
-async fn remove_segments_from_disk(
+fn remove_segments_from_disk(
     timeline_dir: &Path,
     wal_seg_size: usize,
     remove_predicate: impl Fn(XLogSegNo) -> bool,
@@ -466,8 +445,8 @@ async fn remove_segments_from_disk(
     let mut min_removed = u64::MAX;
     let mut max_removed = u64::MIN;
 
-    let mut entries = fs::read_dir(timeline_dir).await?;
-    while let Some(entry) = entries.next_entry().await? {
+    for entry in fs::read_dir(timeline_dir)? {
+        let entry = entry?;
         let entry_path = entry.path();
         let fname = entry_path.file_name().unwrap();
 
@@ -478,7 +457,7 @@ async fn remove_segments_from_disk(
             }
             let (segno, _) = XLogFromFileName(fname_str, wal_seg_size);
             if remove_predicate(segno) {
-                remove_file(entry_path).await?;
+                remove_file(entry_path)?;
                 n_removed += 1;
                 min_removed = min(min_removed, segno);
                 max_removed = max(max_removed, segno);
@@ -710,12 +689,12 @@ impl WalReader {
 const ZERO_BLOCK: &[u8] = &[0u8; XLOG_BLCKSZ];
 
 /// Helper for filling file with zeroes.
-async fn write_zeroes(file: &mut File, mut count: usize) -> Result<()> {
+fn write_zeroes(file: &mut File, mut count: usize) -> Result<()> {
     while count >= XLOG_BLCKSZ {
-        file.write_all(ZERO_BLOCK).await?;
+        file.write_all(ZERO_BLOCK)?;
         count -= XLOG_BLCKSZ;
     }
-    file.write_all(&ZERO_BLOCK[0..count]).await?;
+    file.write_all(&ZERO_BLOCK[0..count])?;
     Ok(())
 }
 

scripts/comment-test-report.js

@@ -1,5 +1,3 @@
-#! /usr/bin/env node
-
 //
 // The script parses Allure reports and posts a comment with a summary of the test results to the PR or to the latest commit in the branch.
 //
@@ -21,7 +19,7 @@
 //       })
 //
 
-// Equivalent of Python's defaultdict.
+// Analog of Python's defaultdict.
 //
 // const dm = new DefaultMap(() => new DefaultMap(() => []))
 // dm["firstKey"]["secondKey"].push("value")
@@ -34,7 +32,34 @@ class DefaultMap extends Map {
     }
 }
 
-const parseReportJson = async ({ reportJsonUrl, fetch }) => {
+module.exports = async ({ github, context, fetch, report }) => {
+    // Marker to find the comment in the subsequent runs
+    const startMarker = `<!--AUTOMATIC COMMENT START #${context.payload.number}-->`
+    // If we run the script in the PR or in the branch (main/release/...)
+    const isPullRequest = !!context.payload.pull_request
+    // Latest commit in PR or in the branch
+    const commitSha = isPullRequest ? context.payload.pull_request.head.sha : context.sha
+    // Let users know that the comment is updated automatically
+    const autoupdateNotice = `<div align="right"><sub>The comment gets automatically updated with the latest test results<br>${commitSha} at ${new Date().toISOString()} :recycle:</sub></div>`
+    // GitHub bot id taken from (https://api.github.com/users/github-actions[bot])
+    const githubActionsBotId = 41898282
+    // Commend body itself
+    let commentBody = `${startMarker}\n`
+
+    // Common parameters for GitHub API requests
+    const ownerRepoParams = {
+        owner: context.repo.owner,
+        repo: context.repo.repo,
+    }
+
+    const {reportUrl, reportJsonUrl} = report
+
+    if (!reportUrl || !reportJsonUrl) {
+        commentBody += `#### No tests were run or test report is not available\n`
+        commentBody += autoupdateNotice
+        return
+    }
+
     const suites = await (await fetch(reportJsonUrl)).json()
 
     // Allure distinguishes "failed" (with an assertion error) and "broken" (with any other error) tests.
@@ -58,7 +83,7 @@ const parseReportJson = async ({ reportJsonUrl, fetch }) => {
                 let buildType, pgVersion
                 const match = test.name.match(/[\[-](?<buildType>debug|release)-pg(?<pgVersion>\d+)[-\]]/)?.groups
                 if (match) {
-                    ({ buildType, pgVersion } = match)
+                    ({buildType, pgVersion} = match)
                 } else {
                     // It's ok, we embed BUILD_TYPE and Postgres Version into the test name only for regress suite and do not for other suites (like performance).
                     console.info(`Cannot get BUILD_TYPE and Postgres Version from test name: "${test.name}", defaulting to "release" and "14"`)
@@ -98,68 +123,37 @@ const parseReportJson = async ({ reportJsonUrl, fetch }) => {
         }
     }
 
-    return {
-        failedTests,
-        failedTestsCount,
-        passedTests,
-        passedTestsCount,
-        skippedTests,
-        skippedTestsCount,
-        flakyTests,
-        flakyTestsCount,
-        retriedTests,
-        pgVersions,
-    }
-}
-
-const reportSummary = async (params) => {
-    const {
-        failedTests,
-        failedTestsCount,
-        passedTests,
-        passedTestsCount,
-        skippedTests,
-        skippedTestsCount,
-        flakyTests,
-        flakyTestsCount,
-        retriedTests,
-        pgVersions,
-        reportUrl,
-    } = params
-
-    let summary = ""
-
     const totalTestsCount = failedTestsCount + passedTestsCount + skippedTestsCount
-    summary += `### ${totalTestsCount} tests run: ${passedTestsCount} passed, ${failedTestsCount} failed, ${skippedTestsCount} skipped ([full report](${reportUrl}))\n___\n`
+    commentBody += `### ${totalTestsCount} tests run: ${passedTestsCount} passed, ${failedTestsCount} failed, ${skippedTestsCount} skipped ([full report](${reportUrl}))\n___\n`
 
     // Print test resuls from the newest to the oldest Postgres version for release and debug builds.
     for (const pgVersion of Array.from(pgVersions).sort().reverse()) {
         if (Object.keys(failedTests[pgVersion]).length > 0) {
-            summary += `#### Failures on Posgres ${pgVersion}\n\n`
+            commentBody += `#### Failures on Posgres ${pgVersion}\n\n`
             for (const [testName, tests] of Object.entries(failedTests[pgVersion])) {
                 const links = []
                 for (const test of tests) {
                     const allureLink = `${reportUrl}#suites/${test.parentUid}/${test.uid}`
                     links.push(`[${test.buildType}](${allureLink})`)
                 }
-                summary += `- \`${testName}\`: ${links.join(", ")}\n`
+                commentBody += `- \`${testName}\`: ${links.join(", ")}\n`
             }
 
             const testsToRerun = Object.values(failedTests[pgVersion]).map(x => x[0].name)
             const command = `DEFAULT_PG_VERSION=${pgVersion} scripts/pytest -k "${testsToRerun.join(" or ")}"`
 
-            summary += "```\n"
-            summary += `# Run failed on Postgres ${pgVersion} tests locally:\n`
-            summary += `${command}\n`
-            summary += "```\n"
+            commentBody += "```\n"
+            commentBody += `# Run failed on Postgres ${pgVersion} tests locally:\n`
+            commentBody += `${command}\n`
+            commentBody += "```\n"
         }
     }
 
     if (flakyTestsCount > 0) {
-        summary += `<details>\n<summary>Flaky tests (${flakyTestsCount})</summary>\n\n`
+        commentBody += `<details>\n<summary>Flaky tests (${flakyTestsCount})</summary>\n\n`
         for (const pgVersion of Array.from(pgVersions).sort().reverse()) {
             if (Object.keys(flakyTests[pgVersion]).length > 0) {
-                summary += `#### Postgres ${pgVersion}\n\n`
+                commentBody += `#### Postgres ${pgVersion}\n\n`
                 for (const [testName, tests] of Object.entries(flakyTests[pgVersion])) {
                     const links = []
                     for (const test of tests) {
@@ -167,57 +161,11 @@ const reportSummary = async (params) => {
                         const status = test.status === "passed" ? ":white_check_mark:" : ":x:"
                         links.push(`[${status} ${test.buildType}](${allureLink})`)
                     }
-                    summary += `- \`${testName}\`: ${links.join(", ")}\n`
+                    commentBody += `- \`${testName}\`: ${links.join(", ")}\n`
                 }
             }
         }
-        summary += "\n</details>\n"
-    }
-
-    return summary
-}
-
-module.exports = async ({ github, context, fetch, report }) => {
-    // Marker to find the comment in the subsequent runs
-    const startMarker = `<!--AUTOMATIC COMMENT START #${context.payload.number}-->`
-    // If we run the script in the PR or in the branch (main/release/...)
-    const isPullRequest = !!context.payload.pull_request
-    // Latest commit in PR or in the branch
-    const commitSha = isPullRequest ? context.payload.pull_request.head.sha : context.sha
-    // Let users know that the comment is updated automatically
-    const autoupdateNotice = `<div align="right"><sub>The comment gets automatically updated with the latest test results<br>${commitSha} at ${new Date().toISOString()} :recycle:</sub></div>`
-    // GitHub bot id taken from (https://api.github.com/users/github-actions[bot])
-    const githubActionsBotId = 41898282
-    // Commend body itself
-    let commentBody = `${startMarker}\n`
-
-    // Common parameters for GitHub API requests
-    const ownerRepoParams = {
-        owner: context.repo.owner,
-        repo: context.repo.repo,
-    }
-
-    const {reportUrl, reportJsonUrl} = report
-
-    if (!reportUrl || !reportJsonUrl) {
-        commentBody += `#### No tests were run or test report is not available\n`
-        commentBody += autoupdateNotice
-        return
-    }
-
-    try {
-        const parsed = await parseReportJson({ reportJsonUrl, fetch })
-        commentBody += await reportSummary({ ...parsed, reportUrl })
-    } catch (error) {
-        commentBody += `### [full report](${reportUrl})\n___\n`
-        commentBody += `#### Failed to create a summary for the test run: \n`
-        commentBody += "```\n"
-        commentBody += `${error.stack}\n`
-        commentBody += "```\n"
-        commentBody += "\nTo reproduce and debug the error locally run:\n"
-        commentBody += "```\n"
-        commentBody += `scripts/comment-test-report.js ${reportJsonUrl}`
-        commentBody += "\n```\n"
+        commentBody += "\n</details>\n"
     }
 
     commentBody += autoupdateNotice
@@ -259,60 +207,3 @@ module.exports = async ({ github, context, fetch, report }) => {
         })
     }
 }
-
-// Equivalent of Python's `if __name__ == "__main__":`
-// https://nodejs.org/docs/latest/api/modules.html#accessing-the-main-module
-if (require.main === module) {
-    // Poor man's argument parsing: we expect the third argument is a JSON URL (0: node binary, 1: this script, 2: JSON url)
-    if (process.argv.length !== 3) {
-        console.error(`Unexpected number of arguments\nUsage: node ${process.argv[1]} <jsonUrl>`)
-        process.exit(1)
-    }
-    const jsonUrl = process.argv[2]
-
-    try {
-        new URL(jsonUrl)
-    } catch (error) {
-        console.error(`Invalid URL: ${jsonUrl}\nUsage: node ${process.argv[1]} <jsonUrl>`)
-        process.exit(1)
-    }
-
-    const htmlUrl = jsonUrl.replace("/data/suites.json", "/index.html")
-
-    const githubMock = {
-        rest: {
-            issues: {
-                createComment: console.log,
-                listComments: async () => ({ data: [] }),
-                updateComment: console.log
-            },
-            repos: {
-                createCommitComment: console.log,
-                listCommentsForCommit: async () => ({ data: [] }),
-                updateCommitComment: console.log
-            }
-        }
-    }
-
-    const contextMock = {
-        repo: {
-            owner: 'testOwner',
-            repo: 'testRepo'
-        },
-        payload: {
-            number: 42,
-            pull_request: null,
-        },
-        sha: '0000000000000000000000000000000000000000',
-    }
-
-    module.exports({
-        github: githubMock,
-        context: contextMock,
-        fetch: fetch,
-        report: {
-            reportUrl: htmlUrl,
-            reportJsonUrl: jsonUrl,
-        }
-    })
-}

scripts/ingest_perf_test_result.py

@@ -1,14 +1,12 @@
 #!/usr/bin/env python3
 import argparse
 import json
-import logging
 import os
 import sys
 from contextlib import contextmanager
 from datetime import datetime
 from pathlib import Path
 
-import backoff
 import psycopg2
 import psycopg2.extras
 
@@ -37,20 +35,9 @@ def get_connection_cursor():
     connstr = os.getenv("DATABASE_URL")
     if not connstr:
         err("DATABASE_URL environment variable is not set")
-
-    @backoff.on_exception(backoff.expo, psycopg2.OperationalError, max_time=150)
-    def connect(connstr):
-        conn = psycopg2.connect(connstr, connect_timeout=30)
-        conn.autocommit = True
-        return conn
-
-    conn = connect(connstr)
-    try:
+    with psycopg2.connect(connstr, connect_timeout=30) as conn:
         with conn.cursor() as cur:
             yield cur
-    finally:
-        if conn is not None:
-            conn.close()
 
 
 def create_table(cur):
@@ -128,7 +115,6 @@ def main():
     parser.add_argument(
         "--ingest",
         type=Path,
-        required=True,
         help="Path to perf test result file, or directory with perf test result files",
     )
     parser.add_argument("--initdb", action="store_true", help="Initialuze database")
@@ -154,5 +140,4 @@ def main():
 
 
 if __name__ == "__main__":
-    logging.getLogger("backoff").addHandler(logging.StreamHandler())
     main()

scripts/ingest_regress_test_result.py

@@ -1,13 +1,11 @@
 #!/usr/bin/env python3
 import argparse
-import logging
 import os
 import re
 import sys
 from contextlib import contextmanager
 from pathlib import Path
 
-import backoff
 import psycopg2
 
 CREATE_TABLE = """
@@ -31,20 +29,9 @@ def get_connection_cursor():
     connstr = os.getenv("DATABASE_URL")
     if not connstr:
         err("DATABASE_URL environment variable is not set")
-
-    @backoff.on_exception(backoff.expo, psycopg2.OperationalError, max_time=150)
-    def connect(connstr):
-        conn = psycopg2.connect(connstr, connect_timeout=30)
-        conn.autocommit = True
-        return conn
-
-    conn = connect(connstr)
-    try:
+    with psycopg2.connect(connstr, connect_timeout=30) as conn:
         with conn.cursor() as cur:
             yield cur
-    finally:
-        if conn is not None:
-            conn.close()
 
 
 def create_table(cur):
@@ -114,5 +101,4 @@ def main():
 
 
 if __name__ == "__main__":
-    logging.getLogger("backoff").addHandler(logging.StreamHandler())
     main()

storage_broker/src/lib.rs

@@ -32,7 +32,6 @@ pub const DEFAULT_LISTEN_ADDR: &str = "127.0.0.1:50051";
 pub const DEFAULT_ENDPOINT: &str = const_format::formatcp!("http://{DEFAULT_LISTEN_ADDR}");
 
 pub const DEFAULT_KEEPALIVE_INTERVAL: &str = "5000 ms";
-pub const DEFAULT_CONNECT_TIMEOUT: Duration = Duration::from_millis(5000);
 
 // BrokerServiceClient charged with tonic provided Channel transport; helps to
 // avoid depending on tonic directly in user crates.
@@ -59,8 +58,7 @@ where
     }
     tonic_endpoint = tonic_endpoint
         .http2_keep_alive_interval(keepalive_interval)
-        .keep_alive_while_idle(true)
-        .connect_timeout(DEFAULT_CONNECT_TIMEOUT);
+        .keep_alive_while_idle(true);
     //  keep_alive_timeout is 20s by default on both client and server side
     let channel = tonic_endpoint.connect_lazy();
     Ok(BrokerClientChannel::new(channel))

test_runner/fixtures/flaky.py

@@ -1,6 +1,6 @@
 import json
 from pathlib import Path
-from typing import Any, List, MutableMapping, cast
+from typing import List
 
 import pytest
 from _pytest.config import Config
@@ -56,15 +56,3 @@ def pytest_collection_modifyitems(config: Config, items: List[pytest.Item]):
             # Rerun 3 times = 1 original run + 2 reruns
             log.info(f"Marking {item.nodeid} as flaky. It will be rerun up to 3 times")
             item.add_marker(pytest.mark.flaky(reruns=2))
-
-            # pytest-rerunfailures is not compatible with pytest-timeout (timeout is not set for reruns),
-            #   we can workaround it by setting `timeout_func_only` to True[1].
-            # Unfortunately, setting `timeout_func_only = True` globally in pytest.ini is broken[2],
-            #   but we still can do it using pytest marker.
-            #
-            # - [1] https://github.com/pytest-dev/pytest-rerunfailures/issues/99
-            # - [2] https://github.com/pytest-dev/pytest-timeout/issues/142
-            timeout_marker = item.get_closest_marker("timeout")
-            if timeout_marker is not None:
-                kwargs = cast(MutableMapping[str, Any], timeout_marker.kwargs)
-                kwargs["func_only"] = True

test_runner/fixtures/metrics.py

@@ -57,17 +57,14 @@ PAGESERVER_GLOBAL_METRICS: Tuple[str, ...] = (
     "libmetrics_launch_timestamp",
     "libmetrics_build_info",
     "libmetrics_tracing_event_count_total",
-    "pageserver_materialized_cache_hits_total",
-    "pageserver_materialized_cache_hits_direct_total",
-    "pageserver_getpage_reconstruct_seconds_bucket",
-    "pageserver_getpage_reconstruct_seconds_count",
-    "pageserver_getpage_reconstruct_seconds_sum",
-    *[f"pageserver_basebackup_query_seconds_{x}" for x in ["bucket", "count", "sum"]],
 )
 
 PAGESERVER_PER_TENANT_METRICS: Tuple[str, ...] = (
     "pageserver_current_logical_size",
     "pageserver_resident_physical_size",
+    "pageserver_getpage_reconstruct_seconds_bucket",
+    "pageserver_getpage_reconstruct_seconds_count",
+    "pageserver_getpage_reconstruct_seconds_sum",
     "pageserver_getpage_get_reconstruct_data_seconds_bucket",
     "pageserver_getpage_get_reconstruct_data_seconds_count",
     "pageserver_getpage_get_reconstruct_data_seconds_sum",
@@ -76,6 +73,8 @@ PAGESERVER_PER_TENANT_METRICS: Tuple[str, ...] = (
     "pageserver_io_operations_seconds_count",
     "pageserver_io_operations_seconds_sum",
     "pageserver_last_record_lsn",
+    "pageserver_materialized_cache_hits_total",
+    "pageserver_materialized_cache_hits_direct_total",
     "pageserver_read_num_fs_layers_bucket",
     "pageserver_read_num_fs_layers_count",
     "pageserver_read_num_fs_layers_sum",

test_runner/fixtures/neon_fixtures.py

@@ -1631,8 +1631,6 @@ class NeonPageserver(PgProtocol):
             r".*ERROR.*ancestor timeline \S+ is being stopped",
             # this is expected given our collaborative shutdown approach for the UploadQueue
             ".*Compaction failed, retrying in .*: queue is in state Stopped.*",
-            # Pageserver timeline deletion should be polled until it gets 404, so ignore it globally
-            ".*Error processing HTTP request: NotFound: Timeline .* was not found",
         ]
 
     def start(
@@ -2415,17 +2413,6 @@ class Endpoint(PgProtocol):
 
         return self
 
-    def respec(self, **kwargs):
-        """Update the endpoint.json file used by control_plane."""
-        # Read config
-        config_path = os.path.join(self.endpoint_path(), "endpoint.json")
-        with open(config_path, "r") as f:
-            data_dict = json.load(f)
-
-        # Write it back updated
-        with open(config_path, "w") as file:
-            json.dump(dict(data_dict, **kwargs), file, indent=4)
-
     def stop(self) -> "Endpoint":
         """
         Stop the Postgres instance if it's running.

test_runner/fixtures/pageserver/http.py

@@ -342,11 +342,6 @@ class PageserverHttpClient(requests.Session):
         return res_json
 
     def timeline_delete(self, tenant_id: TenantId, timeline_id: TimelineId, **kwargs):
-        """
-        Note that deletion is not instant, it is scheduled and performed mostly in the background.
-        So if you need to wait for it to complete use `timeline_delete_wait_completed`.
-        For longer description consult with pageserver openapi spec.
-        """
         res = self.delete(
             f"http://localhost:{self.port}/v1/tenant/{tenant_id}/timeline/{timeline_id}", **kwargs
         )

test_runner/fixtures/pageserver/utils.py

@@ -193,30 +193,19 @@ def wait_for_upload_queue_empty(
         time.sleep(0.2)
 
 
-def wait_timeline_detail_404(
-    pageserver_http: PageserverHttpClient, tenant_id: TenantId, timeline_id: TimelineId
-):
-    last_exc = None
-    for _ in range(2):
-        time.sleep(0.250)
-        try:
-            data = pageserver_http.timeline_detail(tenant_id, timeline_id)
-            log.error(f"detail {data}")
-        except PageserverApiException as e:
-            log.debug(e)
-            if e.status_code == 404:
-                return
-
-            last_exc = e
-
-    raise last_exc or RuntimeError(f"Timeline wasnt deleted in time, state: {data['state']}")
-
-
-def timeline_delete_wait_completed(
+def assert_timeline_detail_404(
     pageserver_http: PageserverHttpClient,
     tenant_id: TenantId,
     timeline_id: TimelineId,
-    **delete_args,
 ):
-    pageserver_http.timeline_delete(tenant_id=tenant_id, timeline_id=timeline_id, **delete_args)
-    wait_timeline_detail_404(pageserver_http, tenant_id, timeline_id)
+    """Asserts that timeline_detail returns 404, or dumps the detail."""
+    try:
+        data = pageserver_http.timeline_detail(tenant_id, timeline_id)
+        log.error(f"detail {data}")
+    except PageserverApiException as e:
+        log.error(e)
+        if e.status_code == 404:
+            return
+        else:
+            raise
+    raise Exception("detail succeeded (it should return 404)")

test_runner/performance/test_startup.py

@@ -32,18 +32,13 @@ def test_startup_simple(neon_env_builder: NeonEnvBuilder, zenbenchmark: NeonBenc
 
     env.neon_cli.create_branch("test_startup")
 
-    endpoint = None
-
     # We do two iterations so we can see if the second startup is faster. It should
     # be because the compute node should already be configured with roles, databases,
     # extensions, etc from the first run.
     for i in range(2):
         # Start
         with zenbenchmark.record_duration(f"{i}_start_and_select"):
-            if endpoint:
-                endpoint.start()
-            else:
-                endpoint = env.endpoints.create_start("test_startup")
+            endpoint = env.endpoints.create_start("test_startup")
             endpoint.safe_psql("select 1;")
 
         # Get metrics
@@ -62,9 +57,6 @@ def test_startup_simple(neon_env_builder: NeonEnvBuilder, zenbenchmark: NeonBenc
         # Stop so we can restart
         endpoint.stop()
 
-        # Imitate optimizations that console would do for the second start
-        endpoint.respec(skip_pg_catalog_updates=True)
-
 
 # This test sometimes runs for longer than the global 5 minute timeout.
 @pytest.mark.timeout(600)

test_runner/pg_clients/python/pg8000/requirements.txt

@@ -1,2 +1,2 @@
-pg8000==1.29.8
+pg8000==1.29.4
 scramp>=1.4.3

test_runner/pg_clients/rust/tokio-postgres/Cargo.lock

@@ -396,9 +396,9 @@ checksum = "b7e5500299e16ebb147ae15a00a942af264cf3688f47923b8fc2cd5858f23ad3"
 
 [[package]]
 name = "openssl"
-version = "0.10.55"
+version = "0.10.52"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "345df152bc43501c5eb9e4654ff05f794effb78d4efe3d53abc158baddc0703d"
+checksum = "01b8574602df80f7b85fdfc5392fa884a4e3b3f4f35402c070ab34c3d3f78d56"
 dependencies = [
  "bitflags",
  "cfg-if",
@@ -428,9 +428,9 @@ checksum = "ff011a302c396a5197692431fc1948019154afc178baf7d8e37367442a4601cf"
 
 [[package]]
 name = "openssl-sys"
-version = "0.9.90"
+version = "0.9.87"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "374533b0e45f3a7ced10fcaeccca020e66656bc03dac384f852e4e5a7a8104a6"
+checksum = "8e17f59264b2809d77ae94f0e1ebabc434773f370d6ca667bd223ea10e06cc7e"
 dependencies = [
  "cc",
  "libc",

test_runner/pg_clients/rust/tokio-postgres/Dockerfile

@@ -1,4 +1,4 @@
-FROM rust:1.70
+FROM rust:1.69
 WORKDIR /source
 
 COPY . .

test_runner/pg_clients/swift/PostgresNIOExample/Package.resolved

@@ -5,8 +5,8 @@
       "kind" : "remoteSourceControl",
       "location" : "https://github.com/vapor/postgres-nio.git",
       "state" : {
-        "revision" : "061a0836d7c1887e04a975d1d2eaa2ef5fd7dfab",
-        "version" : "1.16.0"
+        "revision" : "dbf9c2eb596df39cba8ff3f74d74b2e6a31bd937",
+        "version" : "1.14.1"
       }
     },
     {
@@ -59,8 +59,8 @@
       "kind" : "remoteSourceControl",
       "location" : "https://github.com/apple/swift-nio.git",
       "state" : {
-        "revision" : "6213ba7a06febe8fef60563a4a7d26a4085783cf",
-        "version" : "2.54.0"
+        "revision" : "d1690f85419fdac8d54e350fb6d2ab9fd95afd75",
+        "version" : "2.51.1"
       }
     },
     {

test_runner/pg_clients/swift/PostgresNIOExample/Package.swift

@@ -4,7 +4,7 @@ import PackageDescription
 let package = Package(
     name: "PostgresNIOExample",
     dependencies: [
-        .package(url: "https://github.com/vapor/postgres-nio.git", from: "1.16.0")
+        .package(url: "https://github.com/vapor/postgres-nio.git", from: "1.14.1")
     ],
     targets: [
         .executableTarget(

test_runner/pg_clients/typescript/postgresql-client/package-lock.json

@@ -5,7 +5,23 @@
   "packages": {
     "": {
       "dependencies": {
-        "postgresql-client": "2.5.9"
+        "postgresql-client": "2.5.5"
+      }
+    },
+    "node_modules/debug": {
+      "version": "4.3.4",
+      "resolved": "https://registry.npmjs.org/debug/-/debug-4.3.4.tgz",
+      "integrity": "sha512-PRWFHuSU3eDtQJPvnNY7Jcket1j0t5OuOsFzPPzsekD52Zl8qUfFIPEiswXqIvHWGVHOgX+7G/vCNNhehwxfkQ==",
+      "dependencies": {
+        "ms": "2.1.2"
+      },
+      "engines": {
+        "node": ">=6.0"
+      },
+      "peerDependenciesMeta": {
+        "supports-color": {
+          "optional": true
+        }
       }
     },
     "node_modules/doublylinked": {
@@ -25,6 +41,11 @@
         "putil-promisify": "^1.8.6"
       }
     },
+    "node_modules/ms": {
+      "version": "2.1.2",
+      "resolved": "https://registry.npmjs.org/ms/-/ms-2.1.2.tgz",
+      "integrity": "sha512-sGkPx+VjMtmA6MX27oA4FBFELFCZZ4S4XqeGOXCv68tT+jb3vk/RyaKWP0PTKyWtmLSM0b+adUTEvbs1PEaH2w=="
+    },
     "node_modules/obuf": {
       "version": "1.1.2",
       "resolved": "https://registry.npmjs.org/obuf/-/obuf-1.1.2.tgz",
@@ -42,28 +63,30 @@
       }
     },
     "node_modules/postgresql-client": {
-      "version": "2.5.9",
-      "resolved": "https://registry.npmjs.org/postgresql-client/-/postgresql-client-2.5.9.tgz",
-      "integrity": "sha512-s+kgTN6TfWLzehEyxw4Im4odnxVRCbZ0DEJzWS6SLowPAmB2m1/DOiOvZC0+ZVoi5AfbGE6SBqFxKguSyVAXZg==",
+      "version": "2.5.5",
+      "resolved": "https://registry.npmjs.org/postgresql-client/-/postgresql-client-2.5.5.tgz",
+      "integrity": "sha512-2Mu3i+6NQ9cnkoZNd0XeSZo9WoUpuWf4ZSiCCoDWSj82T93py2/SKXZ1aUaP8mVaU0oKpyyGe0IwLYZ1VHShnA==",
       "dependencies": {
+        "debug": "^4.3.4",
         "doublylinked": "^2.5.2",
         "lightning-pool": "^4.2.1",
         "postgres-bytea": "^3.0.0",
-        "power-tasks": "^1.7.0",
+        "power-tasks": "^1.6.4",
         "putil-merge": "^3.10.3",
         "putil-promisify": "^1.10.0",
         "putil-varhelpers": "^1.6.5"
       },
       "engines": {
-        "node": ">=16.0",
+        "node": ">=14.0",
         "npm": ">=7.0.0"
       }
     },
     "node_modules/power-tasks": {
-      "version": "1.7.0",
-      "resolved": "https://registry.npmjs.org/power-tasks/-/power-tasks-1.7.0.tgz",
-      "integrity": "sha512-rndZXCDxhuIDjPUJJvQwBDHaYagCkjvbPF/NA+omh/Ef4rAI9KtnvdA0k98dyiGpn1zXOpc6c2c0JWzg/xAhJg==",
+      "version": "1.6.4",
+      "resolved": "https://registry.npmjs.org/power-tasks/-/power-tasks-1.6.4.tgz",
+      "integrity": "sha512-LX8GGgEIP1N7jsZqlqZ275e6f1Ehq97APCEGj8uVO0NoEoB+77QUX12BFv3LmlNKfq4fIuNSPiHhyHFjqn2gfA==",
       "dependencies": {
+        "debug": "^4.3.4",
         "doublylinked": "^2.5.2",
         "strict-typed-events": "^2.3.1"
       },
@@ -109,9 +132,9 @@
       }
     },
     "node_modules/ts-gems": {
-      "version": "2.4.0",
-      "resolved": "https://registry.npmjs.org/ts-gems/-/ts-gems-2.4.0.tgz",
-      "integrity": "sha512-SdugYAXoWvbqrxLodIObzxhEKacDxh5LfAJIiIkiH7q5thvuuCzdmkdTVQYf7uEDrEpPhfx4tokDMamdO3be9A=="
+      "version": "2.3.0",
+      "resolved": "https://registry.npmjs.org/ts-gems/-/ts-gems-2.3.0.tgz",
+      "integrity": "sha512-bUvrwrzlct7vfaNvtgMhynDf6lAki/kTtrNsIGhX6l7GJGK3s6b8Ro7dazOLXabV0m2jyShBzDQ8X1+h/C2Cug=="
     }
   }
 }

test_runner/pg_clients/typescript/postgresql-client/package.json

@@ -1,6 +1,6 @@
 {
   "type": "module",
   "dependencies": {
-    "postgresql-client": "2.5.9"
+    "postgresql-client": "2.5.5"
   }
 }

test_runner/pg_clients/typescript/serverless-driver/Dockerfile

@@ -1,4 +1,4 @@
-FROM node:20
+FROM node:18
 WORKDIR /source
 
 COPY . .