wip

.dockerignore

@@ -19,7 +19,6 @@
 !trace/
 !vendor/postgres-v14/
 !vendor/postgres-v15/
-!vendor/postgres-v16/
 !workspace_hack/
 !neon_local/
 !scripts/ninstall.sh

.github/actionlint.yml

@@ -1,8 +0,0 @@
-self-hosted-runner:
-  labels:
-    - gen3
-    - large
-    - small
-    - us-east-2
-config-variables:
-  - SLACK_UPCOMING_RELEASE_CHANNEL_ID

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

@@ -76,8 +76,8 @@ runs:
           rm -f ${ALLURE_ZIP}
         fi
       env:
-        ALLURE_VERSION: 2.24.0
-        ALLURE_ZIP_SHA256: 60b1d6ce65d9ef24b23cf9c2c19fd736a123487c38e54759f1ed1a7a77353c90
+        ALLURE_VERSION: 2.23.1
+        ALLURE_ZIP_SHA256: 11141bfe727504b3fd80c0f9801eb317407fd0ac983ebb57e671f14bac4bcd86
 
     # Potentially we could have several running build for the same key (for example, for the main branch), so we use improvised lock for this
     - name: Acquire lock

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

@@ -70,9 +70,6 @@ runs:
         name: compatibility-snapshot-${{ inputs.build_type }}-pg${{ inputs.pg_version }}
         path: /tmp/compatibility_snapshot_pg${{ inputs.pg_version }}
         prefix: latest
-        # The lack of compatibility snapshot (for example, for the new Postgres version)
-        # shouldn't fail the whole job. Only relevant test should fail.
-        skip-if-does-not-exist: true
 
     - name: Checkout
       if: inputs.needs_postgres_source == 'true'

.github/workflows/actionlint.yml

@@ -1,31 +0,0 @@
-name: Lint GitHub Workflows
-
-on:
-  push:
-    branches:
-      - main
-      - release
-    paths:
-      - '.github/workflows/*.ya?ml'
-  pull_request:
-    paths:
-      - '.github/workflows/*.ya?ml'
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
-
-jobs:
-  actionlint:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-      - uses: reviewdog/action-actionlint@v1
-        env:
-          # SC2046 - Quote this to prevent word splitting. - https://www.shellcheck.net/wiki/SC2046
-          # SC2086 - Double quote to prevent globbing and word splitting. - https://www.shellcheck.net/wiki/SC2086
-          SHELLCHECK_OPTS: --exclude=SC2046,SC2086
-        with:
-          fail_on_error: true
-          filter_mode: nofilter
-          level: error

.github/workflows/approved-for-ci-run.yml

@@ -16,29 +16,20 @@ on:
       # Actual magic happens here:
       - labeled
 
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number }}
-
 env:
   GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   PR_NUMBER: ${{ github.event.pull_request.number }}
-  BRANCH: "ci-run/pr-${{ github.event.pull_request.number }}"
 
-# No permission for GITHUB_TOKEN by default; the **minimal required** set of permissions should be granted in each job.
-permissions: {}
+permissions: write-all
 
-defaults:
-  run:
-    shell: bash -euo pipefail {0}
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number }}
 
 jobs:
   remove-label:
     # Remove `approved-for-ci-run` label if the workflow is triggered by changes in a PR.
     # The PR should be reviewed and labelled manually again.
 
-    permissions:
-      pull-requests: write # For `gh pr edit`
-
     if: |
       contains(fromJSON('["opened", "synchronize", "reopened", "closed"]'), github.event.action) &&
       contains(github.event.pull_request.labels.*.name, 'approved-for-ci-run')
@@ -51,10 +42,6 @@ jobs:
   create-or-update-pr-for-ci-run:
     # Create local PR for an `approved-for-ci-run` labelled PR to run CI pipeline in it.
 
-    permissions:
-      pull-requests: write # for `gh pr edit`
-      # For `git push` and `gh pr create` we use CI_ACCESS_TOKEN
-
     if: |
       github.event.action == 'labeled' &&
       contains(github.event.pull_request.labels.*.name, 'approved-for-ci-run')
@@ -67,52 +54,25 @@ jobs:
       - uses: actions/checkout@v3
         with:
           ref: main
-          token: ${{ secrets.CI_ACCESS_TOKEN }}
 
       - run: gh pr checkout "${PR_NUMBER}"
 
-      - run: git checkout -b "${BRANCH}"
+      - run: git checkout -b "ci-run/pr-${PR_NUMBER}"
 
-      - run: git push --force origin "${BRANCH}"
+      - run: git push --force origin "ci-run/pr-${PR_NUMBER}"
 
       - name: Create a Pull Request for CI run (if required)
         env:
           GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
         run: |
-          cat << EOF > body.md
-            This Pull Request is created automatically to run the CI pipeline for #${PR_NUMBER}
+          HEAD="ci-run/pr-${PR_NUMBER}"
+          BODY="This Pull Request was create automatically to run CI pipeline for #${PR_NUMBER}.\n\nPlease do not alter or merge/close it.\n\nFeel free to comment the original PR."
 
-            Please do not alter or merge/close it.
-
-            Feel free to review/comment/discuss the original PR #${PR_NUMBER}.
-          EOF
-
-          ALREADY_CREATED="$(gh pr --repo ${GITHUB_REPOSITORY} list --head ${BRANCH} --base main --json number --jq '.[].number')"
+          ALREADY_CREATED=$(gh pr --repo "${GITHUB_REPOSITORY}" list --head "${HEAD}" --base "main" --json "number" --jq '.[].number')
           if [ -z "${ALREADY_CREATED}" ]; then
-            gh pr --repo "${GITHUB_REPOSITORY}" create --title "CI run for PR #${PR_NUMBER}" \
-                                                       --body-file "body.md" \
-                                                       --head "${BRANCH}" \
-                                                       --base "main" \
-                                                       --draft
-          fi
-
-  cleanup:
-    # Close PRs and delete branchs if the original PR is closed.
-
-    permissions:
-      contents: write # for `--delete-branch` flag in `gh pr close`
-      pull-requests: write # for `gh pr close`
-
-    if: |
-      github.event.action == 'closed' &&
-      github.event.pull_request.head.repo.full_name != github.repository
-
-    runs-on: ubuntu-latest
-
-    steps:
-      - name: Close PR and delete `ci-run/pr-${{ env.PR_NUMBER }}` branch
-        run: |
-          CLOSED="$(gh pr --repo ${GITHUB_REPOSITORY} list --head ${BRANCH} --json 'closed' --jq '.[].closed')"
-          if [ "${CLOSED}" == "false" ]; then
-            gh pr --repo "${GITHUB_REPOSITORY}" close "${BRANCH}" --delete-branch
+            gh pr --repo "${GITHUB_REPOSITORY}" create  --title "CI run for PR #${PR_NUMBER}" \
+                                                        --body "${BODY}" \
+                                                        --head "${HEAD}" \
+                                                        --base "main" \
+                                                        --draft
           fi

.github/workflows/benchmarking.yml

@@ -137,11 +137,11 @@ jobs:
         }'
 
         if [ "$(date +%A)" = "Saturday" ]; then
-          matrix=$(echo "$matrix" | jq '.include += [{ "platform": "rds-postgres", "db_size": "10gb"},
+          matrix=$(echo $matrix | jq '.include += [{ "platform": "rds-postgres", "db_size": "10gb"},
                                                    { "platform": "rds-aurora",   "db_size": "50gb"}]')
         fi
 
-        echo "matrix=$(echo "$matrix" | jq --compact-output '.')" >> $GITHUB_OUTPUT
+        echo "matrix=$(echo $matrix | jq --compact-output '.')" >> $GITHUB_OUTPUT
 
     - name: Generate matrix for OLAP benchmarks
       id: olap-compare-matrix
@@ -153,11 +153,11 @@ jobs:
         }'
 
         if [ "$(date +%A)" = "Saturday" ]; then
-          matrix=$(echo "$matrix" | jq '.include += [{ "platform": "rds-postgres" },
+          matrix=$(echo $matrix | jq '.include += [{ "platform": "rds-postgres" },
                                                    { "platform": "rds-aurora"   }]')
         fi
 
-        echo "matrix=$(echo "$matrix" | jq --compact-output '.')" >> $GITHUB_OUTPUT
+        echo "matrix=$(echo $matrix | jq --compact-output '.')" >> $GITHUB_OUTPUT
 
     - name: Generate matrix for TPC-H benchmarks
       id: tpch-compare-matrix
@@ -172,11 +172,11 @@ jobs:
         }'
 
         if [ "$(date +%A)" = "Saturday" ]; then
-          matrix=$(echo "$matrix" | jq '.include += [{ "platform": "rds-postgres", "scale": "10" },
+          matrix=$(echo $matrix | jq '.include += [{ "platform": "rds-postgres", "scale": "10" },
                                                    { "platform": "rds-aurora",   "scale": "10" }]')
         fi
 
-        echo "matrix=$(echo "$matrix" | jq --compact-output '.')" >> $GITHUB_OUTPUT
+        echo "matrix=$(echo $matrix | jq --compact-output '.')" >> $GITHUB_OUTPUT
 
   pgbench-compare:
     needs: [ generate-matrices ]
@@ -254,7 +254,7 @@ jobs:
         echo "connstr=${CONNSTR}" >> $GITHUB_OUTPUT
 
         QUERY="SELECT version();"
-        if [[ "${PLATFORM}" = "neon"* ]]; then
+        if [ "${PLATFORM}" = "neon"* ]; then
           QUERY="${QUERY} SHOW neon.tenant_id; SHOW neon.timeline_id;"
         fi
         psql ${CONNSTR} -c "${QUERY}"
@@ -383,7 +383,7 @@ jobs:
         echo "connstr=${CONNSTR}" >> $GITHUB_OUTPUT
 
         QUERY="SELECT version();"
-        if [[ "${PLATFORM}" = "neon"* ]]; then
+        if [ "${PLATFORM}" = "neon"* ]; then
           QUERY="${QUERY} SHOW neon.tenant_id; SHOW neon.timeline_id;"
         fi
         psql ${CONNSTR} -c "${QUERY}"
@@ -476,7 +476,7 @@ jobs:
             ;;
         esac
 
-        CONNSTR_SECRET_NAME="BENCHMARK_${ENV_PLATFORM}_S${TEST_OLAP_SCALE}_CONNSTR"
+        CONNSTR_SECRET_NAME="BENCHMARK_${ENV_PLATFORM}_S${SCALE}_CONNSTR"
         echo "CONNSTR_SECRET_NAME=${CONNSTR_SECRET_NAME}" >> $GITHUB_ENV
 
     - name: Set up Connection String
@@ -487,7 +487,7 @@ jobs:
         echo "connstr=${CONNSTR}" >> $GITHUB_OUTPUT
 
         QUERY="SELECT version();"
-        if [[ "${PLATFORM}" = "neon"* ]]; then
+        if [ "${PLATFORM}" = "neon"* ]; then
           QUERY="${QUERY} SHOW neon.tenant_id; SHOW neon.timeline_id;"
         fi
         psql ${CONNSTR} -c "${QUERY}"
@@ -577,7 +577,7 @@ jobs:
         echo "connstr=${CONNSTR}" >> $GITHUB_OUTPUT
 
         QUERY="SELECT version();"
-        if [[ "${PLATFORM}" = "neon"* ]]; then
+        if [ "${PLATFORM}" = "neon"* ]; then
           QUERY="${QUERY} SHOW neon.tenant_id; SHOW neon.timeline_id;"
         fi
         psql ${CONNSTR} -c "${QUERY}"

.github/workflows/build_and_test.yml

@@ -23,30 +23,7 @@ env:
   AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_KEY_DEV }}
 
 jobs:
-  check-permissions:
-    runs-on: ubuntu-latest
-
-    steps:
-    - name: Disallow PRs from forks
-      if: |
-        github.event_name == 'pull_request' &&
-        github.event.pull_request.head.repo.full_name != github.repository
-
-      run: |
-        if [ "${{ contains(fromJSON('["OWNER", "MEMBER", "COLLABORATOR"]'), github.event.pull_request.author_association) }}" = "true" ]; then
-          MESSAGE="Please create a PR from a branch of ${GITHUB_REPOSITORY} instead of a fork"
-        else
-          MESSAGE="The PR should be reviewed and labelled with 'approved-for-ci-run' to trigger a CI run"
-        fi
-
-        echo >&2 "We don't run CI for PRs from forks"
-        echo >&2 "${MESSAGE}"
-
-        exit 1
-
-
   tag:
-    needs: [ check-permissions ]
     runs-on: [ self-hosted, gen3, small ]
     container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/base:pinned
     outputs:
@@ -75,7 +52,6 @@ jobs:
         id: build-tag
 
   check-codestyle-python:
-    needs: [ check-permissions ]
     runs-on: [ self-hosted, gen3, small ]
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
@@ -108,7 +84,6 @@ jobs:
         run: poetry run mypy .
 
   check-codestyle-rust:
-    needs: [ check-permissions ]
     runs-on: [ self-hosted, gen3, large ]
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
@@ -175,7 +150,6 @@ jobs:
         run: cargo deny check
 
   build-neon:
-    needs: [ check-permissions ]
     runs-on: [ self-hosted, gen3, large ]
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
@@ -212,7 +186,7 @@ jobs:
           # Eventually it will be replaced by a regression test https://github.com/neondatabase/neon/pull/4603
 
           FAILED=false
-          for postgres in postgres-v14 postgres-v15 postgres-v16; do
+          for postgres in postgres-v14 postgres-v15; do
             expected=$(cat vendor/revisions.json | jq --raw-output '."'"${postgres}"'"')
             actual=$(git rev-parse "HEAD:vendor/${postgres}")
             if [ "${expected}" != "${actual}" ]; then
@@ -234,10 +208,6 @@ jobs:
         id: pg_v15_rev
         run: echo pg_rev=$(git rev-parse HEAD:vendor/postgres-v15) >> $GITHUB_OUTPUT
 
-      - name: Set pg 16 revision for caching
-        id: pg_v16_rev
-        run: echo pg_rev=$(git rev-parse HEAD:vendor/postgres-v16) >> $GITHUB_OUTPUT
-
       # Set some environment variables used by all the steps.
       #
       # CARGO_FLAGS is extra options to pass to "cargo build", "cargo test" etc.
@@ -258,12 +228,10 @@ jobs:
             cov_prefix=""
             CARGO_FLAGS="--locked --release"
           fi
-          {
-            echo "cov_prefix=${cov_prefix}"
-            echo "CARGO_FEATURES=${CARGO_FEATURES}"
-            echo "CARGO_FLAGS=${CARGO_FLAGS}"
-            echo "CARGO_HOME=${GITHUB_WORKSPACE}/.cargo"
-          } >> $GITHUB_ENV
+          echo "cov_prefix=${cov_prefix}" >> $GITHUB_ENV
+          echo "CARGO_FEATURES=${CARGO_FEATURES}" >> $GITHUB_ENV
+          echo "CARGO_FLAGS=${CARGO_FLAGS}" >> $GITHUB_ENV
+          echo "CARGO_HOME=${GITHUB_WORKSPACE}/.cargo" >> $GITHUB_ENV
 
       # Disabled for now
       # Don't include the ~/.cargo/registry/src directory. It contains just
@@ -298,13 +266,6 @@ jobs:
           path: pg_install/v15
           key: v1-${{ runner.os }}-${{ matrix.build_type }}-pg-${{ steps.pg_v15_rev.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
 
-      - name: Cache postgres v16 build
-        id: cache_pg_16
-        uses: actions/cache@v3
-        with:
-          path: pg_install/v16
-          key: v1-${{ runner.os }}-${{ matrix.build_type }}-pg-${{ steps.pg_v16_rev.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
-
       - name: Build postgres v14
         if: steps.cache_pg_14.outputs.cache-hit != 'true'
         run: mold -run make postgres-v14 -j$(nproc)
@@ -313,10 +274,6 @@ jobs:
         if: steps.cache_pg_15.outputs.cache-hit != 'true'
         run: mold -run make postgres-v15 -j$(nproc)
 
-      - name: Build postgres v16
-        if: steps.cache_pg_16.outputs.cache-hit != 'true'
-        run: mold -run make postgres-v16 -j$(nproc)
-
       - name: Build neon extensions
         run: mold -run make neon-pg-ext -j$(nproc)
 
@@ -390,17 +347,17 @@ jobs:
         uses: ./.github/actions/save-coverage-data
 
   regress-tests:
-    needs: [ check-permissions, build-neon ]
     runs-on: [ self-hosted, gen3, large ]
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
       # Default shared memory is 64mb
       options: --init --shm-size=512mb
+    needs: [ build-neon ]
     strategy:
       fail-fast: false
       matrix:
         build_type: [ debug, release ]
-        pg_version: [ v14, v15, v16 ]
+        pg_version: [ v14, v15 ]
     steps:
       - name: Checkout
         uses: actions/checkout@v3
@@ -428,12 +385,12 @@ jobs:
         uses: ./.github/actions/save-coverage-data
 
   benchmarks:
-    needs: [ check-permissions, build-neon ]
     runs-on: [ self-hosted, gen3, small ]
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
       # Default shared memory is 64mb
       options: --init --shm-size=512mb
+    needs: [ build-neon ]
     if: github.ref_name == 'main' || contains(github.event.pull_request.labels.*.name, 'run-benchmarks')
     strategy:
       fail-fast: false
@@ -460,13 +417,12 @@ jobs:
       # while coverage is currently collected for the debug ones
 
   create-test-report:
-    needs: [ check-permissions, regress-tests, coverage-report, benchmarks ]
-    if: ${{ !cancelled() && contains(fromJSON('["skipped", "success"]'), needs.check-permissions.result) }}
-
     runs-on: [ self-hosted, gen3, small ]
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
       options: --init
+    needs: [ regress-tests, coverage-report, benchmarks ]
+    if: ${{ !cancelled() }}
 
     steps:
       - uses: actions/checkout@v3
@@ -507,12 +463,11 @@ jobs:
             })
 
   coverage-report:
-    needs: [ check-permissions, regress-tests ]
-
     runs-on: [ self-hosted, gen3, small ]
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rust:pinned
       options: --init
+    needs: [ regress-tests ]
     strategy:
       fail-fast: false
       matrix:
@@ -627,11 +582,11 @@ jobs:
             })
 
   trigger-e2e-tests:
-    needs: [ check-permissions, promote-images, tag ]
     runs-on: [ self-hosted, gen3, small ]
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/base:pinned
       options: --init
+    needs: [ promote-images, tag ]
     steps:
       - name: Set PR's status to pending and request a remote CI test
         run: |
@@ -672,8 +627,8 @@ jobs:
             }"
 
   neon-image:
-    needs: [ check-permissions, tag ]
     runs-on: [ self-hosted, gen3, large ]
+    needs: [ tag ]
     container: gcr.io/kaniko-project/executor:v1.9.2-debug
     defaults:
       run:
@@ -720,7 +675,7 @@ jobs:
 
   compute-tools-image:
     runs-on: [ self-hosted, gen3, large ]
-    needs: [ check-permissions, tag ]
+    needs: [ tag ]
     container: gcr.io/kaniko-project/executor:v1.9.2-debug
     defaults:
       run:
@@ -765,17 +720,17 @@ jobs:
         run: rm -rf ~/.ecr
 
   compute-node-image:
-    needs: [ check-permissions, tag ]
     runs-on: [ self-hosted, gen3, large ]
     container:
       image: gcr.io/kaniko-project/executor:v1.9.2-debug
       # Workaround for "Resolving download.osgeo.org (download.osgeo.org)... failed: Temporary failure in name resolution.""
       # Should be prevented by https://github.com/neondatabase/neon/issues/4281
       options: --add-host=download.osgeo.org:140.211.15.30
+    needs: [ tag ]
     strategy:
       fail-fast: false
       matrix:
-        version: [ v14, v15, v16 ]
+        version: [ v14, v15 ]
     defaults:
       run:
         shell: sh -eu {0}
@@ -824,17 +779,17 @@ jobs:
         run: rm -rf ~/.ecr
 
   vm-compute-node-image:
-    needs: [ check-permissions, tag, compute-node-image ]
     runs-on: [ self-hosted, gen3, large ]
+    needs: [ tag, compute-node-image ]
     strategy:
       fail-fast: false
       matrix:
-        version: [ v14, v15, v16 ]
+        version: [ v14, v15 ]
     defaults:
       run:
         shell: sh -eu {0}
     env:
-      VM_BUILDER_VERSION: v0.17.12
+      VM_BUILDER_VERSION: v0.17.5
 
     steps:
       - name: Checkout
@@ -866,7 +821,7 @@ jobs:
           docker push 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-${{ matrix.version }}:${{needs.tag.outputs.build-tag}}
 
   test-images:
-    needs: [ check-permissions, tag, neon-image, compute-node-image, compute-tools-image ]
+    needs: [ tag, neon-image, compute-node-image, compute-tools-image ]
     runs-on: [ self-hosted, gen3, small ]
 
     steps:
@@ -909,8 +864,8 @@ jobs:
           docker compose -f ./docker-compose/docker-compose.yml down
 
   promote-images:
-    needs: [ check-permissions, tag, test-images, vm-compute-node-image ]
     runs-on: [ self-hosted, gen3, small ]
+    needs: [ tag, test-images, vm-compute-node-image ]
     container: golang:1.19-bullseye
     # Don't add if-condition here.
     # The job should always be run because we have dependant other jobs that shouldn't be skipped
@@ -930,7 +885,6 @@ jobs:
         run: |
           crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v14:${{needs.tag.outputs.build-tag}} vm-compute-node-v14
           crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v15:${{needs.tag.outputs.build-tag}} vm-compute-node-v15
-          crane pull 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v16:${{needs.tag.outputs.build-tag}} vm-compute-node-v16
 
       - name: Add latest tag to images
         if: |
@@ -943,8 +897,6 @@ jobs:
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v14:${{needs.tag.outputs.build-tag}} latest
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v15:${{needs.tag.outputs.build-tag}} latest
           crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v15:${{needs.tag.outputs.build-tag}} latest
-          crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v16:${{needs.tag.outputs.build-tag}} latest
-          crane tag 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v16:${{needs.tag.outputs.build-tag}} latest
 
       - name: Push images to production ECR
         if: |
@@ -957,8 +909,6 @@ jobs:
           crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v14:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v14:latest
           crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v15:${{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:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v15:latest
-          crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v16:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/compute-node-v16:latest
-          crane copy 369495373322.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v16:${{needs.tag.outputs.build-tag}} 093970136003.dkr.ecr.eu-central-1.amazonaws.com/vm-compute-node-v16:latest
 
       - name: Configure Docker Hub login
         run: |
@@ -970,7 +920,6 @@ jobs:
         run: |
           crane push vm-compute-node-v14 neondatabase/vm-compute-node-v14:${{needs.tag.outputs.build-tag}}
           crane push vm-compute-node-v15 neondatabase/vm-compute-node-v15:${{needs.tag.outputs.build-tag}}
-          crane push vm-compute-node-v16 neondatabase/vm-compute-node-v16:${{needs.tag.outputs.build-tag}}
 
       - name: Push latest tags to Docker Hub
         if: |
@@ -983,15 +932,13 @@ jobs:
           crane tag neondatabase/vm-compute-node-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:${{needs.tag.outputs.build-tag}} latest
-          crane tag neondatabase/compute-node-v16:${{needs.tag.outputs.build-tag}} latest
-          crane tag neondatabase/vm-compute-node-v16:${{needs.tag.outputs.build-tag}} latest
 
       - name: Cleanup ECR folder
         run: rm -rf ~/.ecr
 
   trigger-custom-extensions-build-and-wait:
-    needs: [ check-permissions, tag ]
     runs-on: ubuntu-latest
+    needs: [ tag ]
     steps:
       - name: Set PR's status to pending and request a remote CI test
         run: |
@@ -1034,7 +981,7 @@ jobs:
 
           last_status="" # a variable to carry the last status of the "build-and-upload-extensions" context
 
-          for ((i=0; i <= TIMEOUT; i+=INTERVAL)); do
+          for ((i=0; i <= $TIMEOUT; i+=$INTERVAL)); do
             sleep $INTERVAL
 
             # Get statuses for the latest commit in the PR / branch
@@ -1064,11 +1011,10 @@ jobs:
           exit 1
 
   deploy:
-    needs: [ check-permissions, promote-images, tag, regress-tests, trigger-custom-extensions-build-and-wait ]
-    if: ( github.ref_name == 'main' || github.ref_name == 'release' ) && github.event_name != 'workflow_dispatch'
-
     runs-on: [ self-hosted, gen3, small ]
     container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/ansible:latest
+    needs: [ promote-images, tag, regress-tests, trigger-custom-extensions-build-and-wait ]
+    if: ( github.ref_name == 'main' || github.ref_name == 'release' ) && github.event_name != 'workflow_dispatch'
     steps:
       - name: Fix git ownership
         run: |
@@ -1091,9 +1037,8 @@ jobs:
           GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
         run: |
           if [[ "$GITHUB_REF_NAME" == "main" ]]; then
-            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=false
+            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}}
           elif [[ "$GITHUB_REF_NAME" == "release" ]]; then
-            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=true
             gh workflow --repo neondatabase/aws run deploy-prod.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f disclamerAcknowledged=true
           else
             echo "GITHUB_REF_NAME (value '$GITHUB_REF_NAME') is not set to either 'main' or 'release'"
@@ -1107,35 +1052,20 @@ jobs:
           # Retry script for 5XX server errors: https://github.com/actions/github-script#retries
           retries: 5
           script: |
-            await github.rest.git.createRef({
+            github.rest.git.createRef({
               owner: context.repo.owner,
               repo: context.repo.repo,
               ref: "refs/tags/${{ needs.tag.outputs.build-tag }}",
               sha: context.sha,
             })
 
-      - name: Create GitHub release
-        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: |
-            await github.rest.repos.createRelease({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              tag_name: "${{ needs.tag.outputs.build-tag }}",
-              generate_release_notes: true,
-            })
-
   promote-compatibility-data:
-    needs: [ check-permissions, promote-images, tag, regress-tests ]
-    if: github.ref_name == 'release'
-
     runs-on: [ self-hosted, gen3, small ]
     container:
       image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/base:pinned
       options: --init
+    needs: [ promote-images, tag, regress-tests ]
+    if: github.ref_name == 'release' && github.event_name != 'workflow_dispatch'
     steps:
       - name: Promote compatibility snapshot for the release
         env:
@@ -1143,7 +1073,7 @@ jobs:
           PREFIX: artifacts/latest
         run: |
           # Update compatibility snapshot for the release
-          for pg_version in v14 v15 v16; do
+          for pg_version in v14 v15; do
             for build_type in debug release; do
               OLD_FILENAME=compatibility-snapshot-${build_type}-pg${pg_version}-${GITHUB_RUN_ID}.tar.zst
               NEW_FILENAME=compatibility-snapshot-${build_type}-pg${pg_version}.tar.zst

.github/workflows/neon_extra_builds.yml

@@ -38,7 +38,7 @@ jobs:
           fetch-depth: 1
 
       - name: Install macOS postgres dependencies
-        run: brew install flex bison openssl protobuf icu4c pkg-config
+        run: brew install flex bison openssl protobuf
 
       - name: Set pg 14 revision for caching
         id: pg_v14_rev
@@ -48,10 +48,6 @@ jobs:
         id: pg_v15_rev
         run: echo pg_rev=$(git rev-parse HEAD:vendor/postgres-v15) >> $GITHUB_OUTPUT
 
-      - name: Set pg 16 revision for caching
-        id: pg_v16_rev
-        run: echo pg_rev=$(git rev-parse HEAD:vendor/postgres-v16) >> $GITHUB_OUTPUT
-
       - name: Cache postgres v14 build
         id: cache_pg_14
         uses: actions/cache@v3
@@ -66,13 +62,6 @@ jobs:
           path: pg_install/v15
           key: v1-${{ runner.os }}-${{ env.BUILD_TYPE }}-pg-${{ steps.pg_v15_rev.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
 
-      - name: Cache postgres v16 build
-        id: cache_pg_16
-        uses: actions/cache@v3
-        with:
-          path: pg_install/v16
-          key: v1-${{ runner.os }}-${{ env.BUILD_TYPE }}-pg-${{ steps.pg_v16_rev.outputs.pg_rev }}-${{ hashFiles('Makefile') }}
-
       - name: Set extra env for macOS
         run: |
           echo 'LDFLAGS=-L/usr/local/opt/openssl@3/lib' >> $GITHUB_ENV
@@ -96,10 +85,6 @@ jobs:
         if: steps.cache_pg_15.outputs.cache-hit != 'true'
         run: make postgres-v15 -j$(nproc)
 
-      - name: Build postgres v16
-        if: steps.cache_pg_16.outputs.cache-hit != 'true'
-        run: make postgres-v16 -j$(nproc)
-
       - name: Build neon extensions
         run: make neon-pg-ext -j$(nproc)
 

.github/workflows/release-notify.yml

@@ -1,29 +0,0 @@
-name: Notify Slack channel about upcoming release
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.number }}
-  cancel-in-progress: true
-
-on:
-  pull_request:
-    branches:
-      - release
-    types:
-      # Default types that triggers a workflow:
-      # - https://docs.github.com/en/actions/using-workflows/events-that-trigger-workflows#pull_request
-      - opened
-      - synchronize
-      - reopened
-      # Additional types that we want to handle:
-      - closed
-
-jobs:
-  notify:
-    runs-on: [ ubuntu-latest ]
-
-    steps:
-      - uses: neondatabase/dev-actions/release-pr-notify@main
-        with:
-          slack-token: ${{ secrets.SLACK_BOT_TOKEN }}
-          slack-channel-id: ${{ vars.SLACK_UPCOMING_RELEASE_CHANNEL_ID || 'C05QQ9J1BRC' }} # if not set, then `#test-release-notifications`
-          github-token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/release.yml

@@ -7,14 +7,11 @@ on:
 
 jobs:
   create_release_branch:
-    runs-on: [ ubuntu-latest ]
-
-    permissions:
-      contents: write # for `git push`
+    runs-on: [ubuntu-latest]
 
     steps:
     - name: Check out code
-      uses: actions/checkout@v4
+      uses: actions/checkout@v3
       with:
         ref: main
 
@@ -29,16 +26,9 @@ jobs:
       run: git push origin releases/${{ steps.date.outputs.date }}
 
     - name: Create pull request into release
-      env:
-        GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
-      run: |
-        cat << EOF > body.md
-          ## Release ${{ steps.date.outputs.date }}
-
-          **Please merge this PR using 'Create a merge commit'!**
-        EOF
-
-        gh pr create --title "Release ${{ steps.date.outputs.date }}" \
-                     --body-file "body.md" \
-                     --head "releases/${{ steps.date.outputs.date }}" \
-                     --base "release"
+      uses: thomaseizinger/create-pull-request@e3972219c86a56550fb70708d96800d8e24ba862 # 1.3.0
+      with:
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        head: releases/${{ steps.date.outputs.date }}
+        base: release
+        title: Release ${{ steps.date.outputs.date }}

.gitmodules

@@ -6,7 +6,3 @@
 	path = vendor/postgres-v15
 	url = https://github.com/neondatabase/postgres.git
 	branch = REL_15_STABLE_neon
-[submodule "vendor/postgres-v16"]
-	path = vendor/postgres-v16
-	url = https://github.com/neondatabase/postgres.git
-	branch = REL_16_STABLE_neon

CODEOWNERS

@@ -5,7 +5,7 @@
 /libs/remote_storage/ @neondatabase/storage
 /libs/safekeeper_api/ @neondatabase/safekeepers
 /libs/vm_monitor/ @neondatabase/autoscaling @neondatabase/compute
-/pageserver/ @neondatabase/storage
+/pageserver/ @neondatabase/compute @neondatabase/storage
 /pgxn/ @neondatabase/compute
 /proxy/ @neondatabase/proxy
 /safekeeper/ @neondatabase/safekeepers

CONTRIBUTING.md

@@ -27,28 +27,3 @@ your patch's fault. Help to fix the root cause if something else has
 broken the CI, before pushing.
 
 *Happy Hacking!*
-
-# How to run a CI pipeline on Pull Requests from external contributors
-_An instruction for maintainers_
-
-## TL;DR:
-- Review the PR
-- If and only if it looks **safe** (i.e. it doesn't contain any malicious code which could expose secrets or harm the CI), then:
-    - Press the "Approve and run" button in GitHub UI
-    - Add the `approved-for-ci-run` label to the PR
-
-Repeat all steps after any change to the PR.
-- When the changes are ready to get merged — merge the original PR (not the internal one)
-
-## Longer version:
-
-GitHub Actions triggered by the `pull_request` event don't share repository secrets with the forks (for security reasons).
-So, passing the CI pipeline on Pull Requests from external contributors is impossible.
-
-We're using the following approach to make it work:
-- After the review, assign the `approved-for-ci-run` label to the PR if changes look safe
-- A GitHub Action will create an internal branch and a new PR with the same changes (for example, for a PR `#1234`, it'll be a branch `ci-run/pr-1234`)
-- Because the PR is created from the internal branch, it is able to access repository secrets (that's why it's crucial to make sure that the PR doesn't contain any malicious code that could expose our secrets or intentionally harm the CI)
-- The label gets removed automatically, so to run CI again with new changes, the label should be added again (after the review)
-
-For details see [`approved-for-ci-run.yml`](.github/workflows/approved-for-ci-run.yml)

Cargo.lock

@@ -636,7 +636,7 @@ dependencies = [
  "sha1",
  "sync_wrapper",
  "tokio",
- "tokio-tungstenite",
+ "tokio-tungstenite 0.20.0",
  "tower",
  "tower-layer",
  "tower-service",
@@ -798,22 +798,6 @@ dependencies = [
  "either",
 ]
 
-[[package]]
-name = "camino"
-version = "1.1.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c59e92b5a388f549b863a7bea62612c09f24c8393560709a54558a9abdfb3b9c"
-
-[[package]]
-name = "camino-tempfile"
-version = "1.0.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d2ab15a83d13f75dbd86f082bdefd160b628476ef58d3b900a0ef74e001bb097"
-dependencies = [
- "camino",
- "tempfile",
-]
-
 [[package]]
 name = "cast"
 version = "0.3.0"
@@ -1069,7 +1053,6 @@ name = "control_plane"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "camino",
  "clap",
  "comfy-table",
  "compute_api",
@@ -1797,9 +1780,18 @@ checksum = "95505c38b4572b2d910cecb0281560f54b440a19336cbbcb27bf6ce6adc6f5a8"
 
 [[package]]
 name = "hermit-abi"
-version = "0.3.3"
+version = "0.2.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d77f7ec81a6d05a3abb01ab6eb7590f6083d08449fe5a1c8b1e620283546ccb7"
+checksum = "ee512640fe35acbfb4bb779db6f0d80704c2cacfa2e39b601ef3e3f47d1ae4c7"
+dependencies = [
+ "libc",
+]
+
+[[package]]
+name = "hermit-abi"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fed44880c466736ef9a5c5b5facefb5ed0785676d0c02d612db14e54f0d84286"
 
 [[package]]
 name = "hex"
@@ -1949,15 +1941,15 @@ dependencies = [
 
 [[package]]
 name = "hyper-tungstenite"
-version = "0.11.1"
+version = "0.9.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7cc7dcb1ab67cd336f468a12491765672e61a3b6b148634dbfe2fe8acd3fe7d9"
+checksum = "880b8b1c98a5ec2a505c7c90db6d3f6f1f480af5655d9c5b55facc9382a5a5b5"
 dependencies = [
  "hyper",
- "pin-project-lite",
+ "pin-project",
  "tokio",
- "tokio-tungstenite",
- "tungstenite",
+ "tokio-tungstenite 0.18.0",
+ "tungstenite 0.18.0",
 ]
 
 [[package]]
@@ -2061,7 +2053,7 @@ version = "1.0.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "eae7b9aee968036d54dce06cebaefd919e4472e753296daccd6d344e3e2df0c2"
 dependencies = [
- "hermit-abi",
+ "hermit-abi 0.3.1",
  "libc",
  "windows-sys 0.48.0",
 ]
@@ -2078,7 +2070,7 @@ version = "0.4.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "adcf93614601c8129ddf72e2d5633df827ba6551541c6d8c59520a371475be1f"
 dependencies = [
- "hermit-abi",
+ "hermit-abi 0.3.1",
  "io-lifetimes",
  "rustix 0.37.19",
  "windows-sys 0.48.0",
@@ -2452,11 +2444,11 @@ dependencies = [
 
 [[package]]
 name = "num_cpus"
-version = "1.16.0"
+version = "1.15.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4161fcb6d602d4d2081af7c3a45852d875a03dd337a6bfdd6e06407b61342a43"
+checksum = "0fac9e2da13b5eb447a6ce3d392f23a29d8694bff781bf03a16cd9ac8697593b"
 dependencies = [
- "hermit-abi",
+ "hermit-abi 0.2.6",
  "libc",
 ]
 
@@ -2658,7 +2650,6 @@ version = "0.1.0"
 dependencies = [
  "anyhow",
  "bytes",
- "camino",
  "clap",
  "git-version",
  "pageserver",
@@ -2679,8 +2670,6 @@ dependencies = [
  "async-trait",
  "byteorder",
  "bytes",
- "camino",
- "camino-tempfile",
  "chrono",
  "clap",
  "close_fds",
@@ -2732,6 +2721,7 @@ dependencies = [
  "strum_macros",
  "svg_fmt",
  "sync_wrapper",
+ "tempfile",
  "tenant_size_model",
  "thiserror",
  "tokio",
@@ -2918,9 +2908,9 @@ dependencies = [
 
 [[package]]
 name = "pin-project-lite"
-version = "0.2.13"
+version = "0.2.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8afb450f006bf6385ca15ef45d71d2288452bc3683ce2e2cacc0d18e4be60b58"
+checksum = "e0a7ae3ac2f1173085d398531c705756c94a4c56843785df85a60c1a0afac116"
 
 [[package]]
 name = "pin-utils"
@@ -3256,7 +3246,6 @@ dependencies = [
  "reqwest-tracing",
  "routerify",
  "rstest",
- "rustc-hash",
  "rustls",
  "rustls-pemfile",
  "scopeguard",
@@ -3424,16 +3413,14 @@ dependencies = [
  "aws-sdk-s3",
  "aws-smithy-http",
  "aws-types",
- "camino",
- "camino-tempfile",
  "hyper",
  "metrics",
  "once_cell",
  "pin-project-lite",
- "rand",
  "scopeguard",
  "serde",
  "serde_json",
+ "tempfile",
  "test-context",
  "tokio",
  "tokio-util",
@@ -3785,8 +3772,6 @@ dependencies = [
  "async-trait",
  "byteorder",
  "bytes",
- "camino",
- "camino-tempfile",
  "chrono",
  "clap",
  "const_format",
@@ -3815,6 +3800,7 @@ dependencies = [
  "serde_with",
  "signal-hook",
  "storage_broker",
+ "tempfile",
  "thiserror",
  "tokio",
  "tokio-io-timeout",
@@ -4655,6 +4641,18 @@ dependencies = [
  "xattr",
 ]
 
+[[package]]
+name = "tokio-tungstenite"
+version = "0.18.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "54319c93411147bced34cb5609a80e0a8e44c5999c93903a81cd866630ec0bfd"
+dependencies = [
+ "futures-util",
+ "log",
+ "tokio",
+ "tungstenite 0.18.0",
+]
+
 [[package]]
 name = "tokio-tungstenite"
 version = "0.20.0"
@@ -4664,7 +4662,7 @@ dependencies = [
  "futures-util",
  "log",
  "tokio",
- "tungstenite",
+ "tungstenite 0.20.0",
 ]
 
 [[package]]
@@ -4979,9 +4977,28 @@ checksum = "3528ecfd12c466c6f163363caf2d02a71161dd5e1cc6ae7b34207ea2d42d81ed"
 
 [[package]]
 name = "tungstenite"
-version = "0.20.1"
+version = "0.18.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9e3dac10fd62eaf6617d3a904ae222845979aec67c615d1c842b4002c7666fb9"
+checksum = "30ee6ab729cd4cf0fd55218530c4522ed30b7b6081752839b68fcec8d0960788"
+dependencies = [
+ "base64 0.13.1",
+ "byteorder",
+ "bytes",
+ "http",
+ "httparse",
+ "log",
+ "rand",
+ "sha1",
+ "thiserror",
+ "url",
+ "utf-8",
+]
+
+[[package]]
+name = "tungstenite"
+version = "0.20.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e862a1c4128df0112ab625f55cd5c934bcb4312ba80b39ae4b4835a3fd58e649"
 dependencies = [
  "byteorder",
  "bytes",
@@ -5113,8 +5130,6 @@ dependencies = [
  "bincode",
  "byteorder",
  "bytes",
- "camino",
- "camino-tempfile",
  "chrono",
  "const_format",
  "criterion",
@@ -5140,6 +5155,7 @@ dependencies = [
  "signal-hook",
  "strum",
  "strum_macros",
+ "tempfile",
  "thiserror",
  "tokio",
  "tokio-stream",
@@ -5213,7 +5229,6 @@ name = "wal_craft"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "camino-tempfile",
  "clap",
  "env_logger",
  "log",
@@ -5221,6 +5236,7 @@ dependencies = [
  "postgres",
  "postgres_ffi",
  "regex",
+ "tempfile",
  "utils",
  "workspace_hack",
 ]
@@ -5632,7 +5648,6 @@ dependencies = [
  "tower",
  "tracing",
  "tracing-core",
- "tungstenite",
  "url",
  "uuid",
 ]

Cargo.toml

@@ -1,5 +1,4 @@
 [workspace]
-resolver = "2"
 members = [
     "compute_tools",
     "control_plane",
@@ -51,7 +50,6 @@ bindgen = "0.65"
 bstr = "1.0"
 byteorder = "1.4"
 bytes = "1.0"
-camino = "1.1.6"
 cfg-if = "1.0.0"
 chrono = { version = "0.4", default-features = false, features = ["clock"] }
 clap = { version = "4.0", features = ["derive"] }
@@ -79,7 +77,7 @@ hostname = "0.3.1"
 humantime = "2.1"
 humantime-serde = "1.1.1"
 hyper = "0.14"
-hyper-tungstenite = "0.11"
+hyper-tungstenite = "0.9"
 inotify = "0.10.2"
 itertools = "0.10"
 jsonwebtoken = "8"
@@ -108,7 +106,6 @@ reqwest-middleware = "0.2.0"
 reqwest-retry = "0.2.2"
 routerify = "3"
 rpds = "0.13"
-rustc-hash = "1.1.0"
 rustls = "0.21"
 rustls-pemfile = "1"
 rustls-split = "0.3"
@@ -188,7 +185,7 @@ workspace_hack = { version = "0.1", path = "./workspace_hack/" }
 criterion = "0.5.1"
 rcgen = "0.11"
 rstest = "0.18"
-camino-tempfile = "1.0.2"
+tempfile = "3.4"
 tonic-build = "0.9"
 
 [patch.crates-io]

Dockerfile

@@ -12,7 +12,6 @@ WORKDIR /home/nonroot
 
 COPY --chown=nonroot vendor/postgres-v14 vendor/postgres-v14
 COPY --chown=nonroot vendor/postgres-v15 vendor/postgres-v15
-COPY --chown=nonroot vendor/postgres-v16 vendor/postgres-v16
 COPY --chown=nonroot pgxn pgxn
 COPY --chown=nonroot Makefile Makefile
 COPY --chown=nonroot scripts/ninstall.sh scripts/ninstall.sh
@@ -40,7 +39,6 @@ ARG CACHEPOT_BUCKET=neon-github-dev
 
 COPY --from=pg-build /home/nonroot/pg_install/v14/include/postgresql/server pg_install/v14/include/postgresql/server
 COPY --from=pg-build /home/nonroot/pg_install/v15/include/postgresql/server pg_install/v15/include/postgresql/server
-COPY --from=pg-build /home/nonroot/pg_install/v16/include/postgresql/server pg_install/v16/include/postgresql/server
 COPY --chown=nonroot . .
 
 # Show build caching stats to check if it was used in the end.
@@ -67,7 +65,6 @@ RUN set -e \
     && apt install -y \
         libreadline-dev \
         libseccomp-dev \
-        libicu67 \
         openssl \
         ca-certificates \
     && rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/* \
@@ -84,7 +81,6 @@ COPY --from=build --chown=neon:neon /home/nonroot/target/release/neon_local
 
 COPY --from=pg-build /home/nonroot/pg_install/v14 /usr/local/v14/
 COPY --from=pg-build /home/nonroot/pg_install/v15 /usr/local/v15/
-COPY --from=pg-build /home/nonroot/pg_install/v16 /usr/local/v16/
 COPY --from=pg-build /home/nonroot/postgres_install.tar.gz /data/
 
 # By default, pageserver uses `.neon/` working directory in WORKDIR, so create one and fill it with the dummy config.

Dockerfile.compute-node

@@ -74,8 +74,8 @@ RUN wget https://gitlab.com/Oslandia/SFCGAL/-/archive/v1.3.10/SFCGAL-v1.3.10.tar
 
 ENV PATH "/usr/local/pgsql/bin:$PATH"
 
-RUN wget https://download.osgeo.org/postgis/source/postgis-3.3.3.tar.gz -O postgis.tar.gz && \
-    echo "74eb356e3f85f14233791013360881b6748f78081cc688ff9d6f0f673a762d13 postgis.tar.gz" | sha256sum --check && \
+RUN wget https://download.osgeo.org/postgis/source/postgis-3.3.2.tar.gz -O postgis.tar.gz && \
+    echo "9a2a219da005a1730a39d1959a1c7cec619b1efb009b65be80ffc25bad299068 postgis.tar.gz" | sha256sum --check && \
     mkdir postgis-src && cd postgis-src && tar xvzf ../postgis.tar.gz --strip-components=1 -C . && \
     find /usr/local/pgsql -type f | sed 's|^/usr/local/pgsql/||' > /before.txt &&\
     ./autogen.sh && \
@@ -124,21 +124,8 @@ COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 RUN apt update && \
     apt install -y ninja-build python3-dev libncurses5 binutils clang
 
-RUN case "${PG_VERSION}" in \
-      "v14" | "v15") \
-        export PLV8_VERSION=3.1.5 \
-        export PLV8_CHECKSUM=1e108d5df639e4c189e1c5bdfa2432a521c126ca89e7e5a969d46899ca7bf106 \
-        ;; \
-      "v16") \
-        export PLV8_VERSION=3.1.8 \
-        export PLV8_CHECKSUM=92b10c7db39afdae97ff748c9ec54713826af222c459084ad002571b79eb3f49 \
-        ;; \
-      *) \
-        echo "Export the valid PG_VERSION variable" && exit 1 \
-        ;; \
-    esac && \
-    wget https://github.com/plv8/plv8/archive/refs/tags/v${PLV8_VERSION}.tar.gz -O plv8.tar.gz && \
-    echo "${PLV8_CHECKSUM} plv8.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/plv8/plv8/archive/refs/tags/v3.1.5.tar.gz -O plv8.tar.gz && \
+    echo "1e108d5df639e4c189e1c5bdfa2432a521c126ca89e7e5a969d46899ca7bf106 plv8.tar.gz" | sha256sum --check && \
     mkdir plv8-src && cd plv8-src && tar xvzf ../plv8.tar.gz --strip-components=1 -C . && \
     export PATH="/usr/local/pgsql/bin:$PATH" && \
     make DOCKER=1 -j $(getconf _NPROCESSORS_ONLN) install && \
@@ -185,8 +172,8 @@ RUN wget https://github.com/uber/h3/archive/refs/tags/v4.1.0.tar.gz -O h3.tar.gz
     cp -R /h3/usr / && \
     rm -rf build
 
-RUN wget https://github.com/zachasme/h3-pg/archive/refs/tags/v4.1.3.tar.gz -O h3-pg.tar.gz && \
-    echo "5c17f09a820859ffe949f847bebf1be98511fb8f1bd86f94932512c00479e324 h3-pg.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/zachasme/h3-pg/archive/refs/tags/v4.1.2.tar.gz -O h3-pg.tar.gz && \
+    echo "c135aa45999b2ad1326d2537c1cadef96d52660838e4ca371706c08fdea1a956 h3-pg.tar.gz" | sha256sum --check && \
     mkdir h3-pg-src && cd h3-pg-src && tar xvzf ../h3-pg.tar.gz --strip-components=1 -C . && \
     export PATH="/usr/local/pgsql/bin:$PATH" && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
@@ -256,8 +243,8 @@ RUN wget https://github.com/michelp/pgjwt/archive/9742dab1b2f297ad3811120db7b214
 FROM build-deps AS hypopg-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/HypoPG/hypopg/archive/refs/tags/1.4.0.tar.gz -O hypopg.tar.gz && \
-    echo "0821011743083226fc9b813c1f2ef5897a91901b57b6bea85a78e466187c6819 hypopg.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/HypoPG/hypopg/archive/refs/tags/1.3.1.tar.gz -O hypopg.tar.gz && \
+    echo "e7f01ee0259dc1713f318a108f987663d60f3041948c2ada57a94b469565ca8e hypopg.tar.gz" | sha256sum --check && \
     mkdir hypopg-src && cd hypopg-src && tar xvzf ../hypopg.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -320,8 +307,8 @@ RUN wget https://github.com/theory/pgtap/archive/refs/tags/v1.2.0.tar.gz -O pgta
 FROM build-deps AS ip4r-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/RhodiumToad/ip4r/archive/refs/tags/2.4.2.tar.gz -O ip4r.tar.gz && \
-    echo "0f7b1f159974f49a47842a8ab6751aecca1ed1142b6d5e38d81b064b2ead1b4b ip4r.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/RhodiumToad/ip4r/archive/refs/tags/2.4.1.tar.gz -O ip4r.tar.gz && \
+    echo "78b9f0c1ae45c22182768fe892a32d533c82281035e10914111400bf6301c726 ip4r.tar.gz" | sha256sum --check && \
     mkdir ip4r-src && cd ip4r-src && tar xvzf ../ip4r.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -336,8 +323,8 @@ RUN wget https://github.com/RhodiumToad/ip4r/archive/refs/tags/2.4.2.tar.gz -O i
 FROM build-deps AS prefix-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/dimitri/prefix/archive/refs/tags/v1.2.10.tar.gz -O prefix.tar.gz && \
-    echo "4342f251432a5f6fb05b8597139d3ccde8dcf87e8ca1498e7ee931ca057a8575 prefix.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/dimitri/prefix/archive/refs/tags/v1.2.9.tar.gz -O prefix.tar.gz && \
+    echo "38d30a08d0241a8bbb8e1eb8f0152b385051665a8e621c8899e7c5068f8b511e prefix.tar.gz" | sha256sum --check && \
     mkdir prefix-src && cd prefix-src && tar xvzf ../prefix.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -352,8 +339,8 @@ RUN wget https://github.com/dimitri/prefix/archive/refs/tags/v1.2.10.tar.gz -O p
 FROM build-deps AS hll-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/citusdata/postgresql-hll/archive/refs/tags/v2.18.tar.gz -O hll.tar.gz && \
-    echo "e2f55a6f4c4ab95ee4f1b4a2b73280258c5136b161fe9d059559556079694f0e hll.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/citusdata/postgresql-hll/archive/refs/tags/v2.17.tar.gz -O hll.tar.gz && \
+    echo "9a18288e884f197196b0d29b9f178ba595b0dfc21fbf7a8699380e77fa04c1e9 hll.tar.gz" | sha256sum --check && \
     mkdir hll-src && cd hll-src && tar xvzf ../hll.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config && \
@@ -368,8 +355,8 @@ RUN wget https://github.com/citusdata/postgresql-hll/archive/refs/tags/v2.18.tar
 FROM build-deps AS plpgsql-check-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-RUN wget https://github.com/okbob/plpgsql_check/archive/refs/tags/v2.4.0.tar.gz -O plpgsql_check.tar.gz && \
-    echo "9ba58387a279b35a3bfa39ee611e5684e6cddb2ba046ddb2c5190b3bd2ca254a plpgsql_check.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/okbob/plpgsql_check/archive/refs/tags/v2.3.2.tar.gz -O plpgsql_check.tar.gz && \
+    echo "9d81167c4bbeb74eebf7d60147b21961506161addc2aee537f95ad8efeae427b plpgsql_check.tar.gz" | sha256sum --check && \
     mkdir plpgsql_check-src && cd plpgsql_check-src && tar xvzf ../plpgsql_check.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
     make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config USE_PGXS=1 && \
@@ -384,21 +371,12 @@ RUN wget https://github.com/okbob/plpgsql_check/archive/refs/tags/v2.4.0.tar.gz
 FROM build-deps AS timescaledb-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
-ARG PG_VERSION
 ENV PATH "/usr/local/pgsql/bin:$PATH"
 
-RUN case "${PG_VERSION}" in \
-      "v14" | "v15") \
-        export TIMESCALEDB_VERSION=2.10.1 \
-        export TIMESCALEDB_CHECKSUM=6fca72a6ed0f6d32d2b3523951ede73dc5f9b0077b38450a029a5f411fdb8c73 \
-        ;; \
-      *) \
-        echo "TimescaleDB not supported on this PostgreSQL version. See https://github.com/timescale/timescaledb/issues/5752" && exit 0;; \
-    esac && \
-    apt-get update && \
+RUN apt-get update && \
     apt-get install -y cmake && \
-    wget https://github.com/timescale/timescaledb/archive/refs/tags/${TIMESCALEDB_VERSION}.tar.gz -O timescaledb.tar.gz && \
-    echo "${TIMESCALEDB_CHECKSUM} timescaledb.tar.gz" | sha256sum --check && \
+    wget https://github.com/timescale/timescaledb/archive/refs/tags/2.10.1.tar.gz -O timescaledb.tar.gz && \
+    echo "6fca72a6ed0f6d32d2b3523951ede73dc5f9b0077b38450a029a5f411fdb8c73 timescaledb.tar.gz" | sha256sum --check && \
     mkdir timescaledb-src && cd timescaledb-src && tar xvzf ../timescaledb.tar.gz --strip-components=1 -C . && \
     ./bootstrap -DSEND_TELEMETRY_DEFAULT:BOOL=OFF -DUSE_TELEMETRY:BOOL=OFF -DAPACHE_ONLY:BOOL=ON -DCMAKE_BUILD_TYPE=Release && \
     cd build && \
@@ -427,10 +405,6 @@ RUN case "${PG_VERSION}" in \
         export PG_HINT_PLAN_VERSION=15_1_5_0 \
         export PG_HINT_PLAN_CHECKSUM=564cbbf4820973ffece63fbf76e3c0af62c4ab23543142c7caaa682bc48918be \
         ;; \
-      "v16") \
-        export PG_HINT_PLAN_VERSION=16_1_6_0 \
-        export PG_HINT_PLAN_CHECKSUM=fc85a9212e7d2819d4ae4ac75817481101833c3cfa9f0fe1f980984e12347d00 \
-        ;; \
       *) \
         echo "Export the valid PG_HINT_PLAN_VERSION variable" && exit 1 \
         ;; \
@@ -478,8 +452,8 @@ FROM build-deps AS pg-cron-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH "/usr/local/pgsql/bin/:$PATH"
-RUN wget https://github.com/citusdata/pg_cron/archive/refs/tags/v1.6.0.tar.gz -O pg_cron.tar.gz && \
-    echo "383a627867d730222c272bfd25cd5e151c578d73f696d32910c7db8c665cc7db pg_cron.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/citusdata/pg_cron/archive/refs/tags/v1.5.2.tar.gz -O pg_cron.tar.gz && \
+    echo "6f7f0980c03f1e2a6a747060e67bf4a303ca2a50e941e2c19daeed2b44dec744 pg_cron.tar.gz" | sha256sum --check && \
     mkdir pg_cron-src && cd pg_cron-src && tar xvzf ../pg_cron.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
     make -j $(getconf _NPROCESSORS_ONLN) install && \
@@ -505,8 +479,8 @@ RUN apt-get update && \
         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_3.tar.gz -O rdkit.tar.gz && \
-    echo "bdbf9a2e6988526bfeb8c56ce3cdfe2998d60ac289078e2215374288185e8c8d rdkit.tar.gz" | sha256sum --check && \
+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 \
@@ -577,19 +551,12 @@ FROM build-deps AS pg-embedding-pg-build
 COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
 
 ENV PATH "/usr/local/pgsql/bin/:$PATH"
-RUN case "${PG_VERSION}" in \
-      "v14" | "v15") \
-        export PG_EMBEDDING_VERSION=0.3.5 \
-        export PG_EMBEDDING_CHECKSUM=0e95b27b8b6196e2cf0a0c9ec143fe2219b82e54c5bb4ee064e76398cbe69ae9 \
-        ;; \
-      *) \
-        echo "pg_embedding not supported on this PostgreSQL version. Use pgvector instead." && exit 0;; \
-    esac && \
-    wget https://github.com/neondatabase/pg_embedding/archive/refs/tags/${PG_EMBEDDING_VERSION}.tar.gz -O pg_embedding.tar.gz && \
-    echo "${PG_EMBEDDING_CHECKSUM} pg_embedding.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/neondatabase/pg_embedding/archive/refs/tags/0.3.5.tar.gz -O pg_embedding.tar.gz && \
+    echo "0e95b27b8b6196e2cf0a0c9ec143fe2219b82e54c5bb4ee064e76398cbe69ae9 pg_embedding.tar.gz" | sha256sum --check && \
     mkdir pg_embedding-src && cd pg_embedding-src && tar xvzf ../pg_embedding.tar.gz --strip-components=1 -C . && \
     make -j $(getconf _NPROCESSORS_ONLN) && \
-    make -j $(getconf _NPROCESSORS_ONLN) install
+    make -j $(getconf _NPROCESSORS_ONLN) install && \
+    echo 'trusted = true' >> /usr/local/pgsql/share/extension/embedding.control
 
 #########################################################################################
 #
@@ -615,7 +582,7 @@ RUN wget https://gitlab.com/dalibo/postgresql_anonymizer/-/archive/1.1.0/postgre
 #########################################################################################
 #
 # Layer "rust extensions"
-# This layer is used to build `pgrx` deps
+# This layer is used to build `pgx` deps
 #
 #########################################################################################
 FROM build-deps AS rust-extensions-build
@@ -635,8 +602,8 @@ RUN curl -sSO https://static.rust-lang.org/rustup/dist/$(uname -m)-unknown-linux
     chmod +x rustup-init && \
     ./rustup-init -y --no-modify-path --profile minimal --default-toolchain stable && \
     rm rustup-init && \
-    cargo install --locked --version 0.10.2 cargo-pgrx && \
-    /bin/bash -c 'cargo pgrx init --pg${PG_VERSION:1}=/usr/local/pgsql/bin/pg_config'
+    cargo install --locked --version 0.7.3 cargo-pgx && \
+    /bin/bash -c 'cargo pgx init --pg${PG_VERSION:1}=/usr/local/pgsql/bin/pg_config'
 
 USER root
 
@@ -648,13 +615,14 @@ USER root
 #########################################################################################
 
 FROM rust-extensions-build AS pg-jsonschema-pg-build
-ARG PG_VERSION
 
-RUN wget https://github.com/supabase/pg_jsonschema/archive/refs/tags/v0.2.0.tar.gz -O pg_jsonschema.tar.gz && \
-    echo "9118fc508a6e231e7a39acaa6f066fcd79af17a5db757b47d2eefbe14f7794f0 pg_jsonschema.tar.gz" | sha256sum --check && \
+# caeab60d70b2fd3ae421ec66466a3abbb37b7ee6 made on 06/03/2023
+# there is no release tag yet, but we need it due to the superuser fix in the control file, switch to git tag after release >= 0.1.5
+RUN wget https://github.com/supabase/pg_jsonschema/archive/caeab60d70b2fd3ae421ec66466a3abbb37b7ee6.tar.gz -O pg_jsonschema.tar.gz && \
+    echo "54129ce2e7ee7a585648dbb4cef6d73f795d94fe72f248ac01119992518469a4 pg_jsonschema.tar.gz" | sha256sum --check && \
     mkdir pg_jsonschema-src && cd pg_jsonschema-src && tar xvzf ../pg_jsonschema.tar.gz --strip-components=1 -C . && \
-    sed -i 's/pgrx = "0.10.2"/pgrx = { version = "0.10.2", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
-    cargo pgrx install --release && \
+    sed -i 's/pgx = "0.7.1"/pgx = { version = "0.7.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
+    cargo pgx install --release && \
     echo "trusted = true" >> /usr/local/pgsql/share/extension/pg_jsonschema.control
 
 #########################################################################################
@@ -665,13 +633,17 @@ RUN wget https://github.com/supabase/pg_jsonschema/archive/refs/tags/v0.2.0.tar.
 #########################################################################################
 
 FROM rust-extensions-build AS pg-graphql-pg-build
-ARG PG_VERSION
 
-RUN wget https://github.com/supabase/pg_graphql/archive/refs/tags/v1.4.0.tar.gz -O pg_graphql.tar.gz && \
-    echo "bd8dc7230282b3efa9ae5baf053a54151ed0e66881c7c53750e2d0c765776edc pg_graphql.tar.gz" | sha256sum --check && \
+# b4988843647450a153439be367168ed09971af85 made on 22/02/2023 (from remove-pgx-contrib-spiext branch)
+# Currently pgx version bump to >= 0.7.2  causes "call to unsafe function" compliation errors in
+# pgx-contrib-spiext. There is a branch that removes that dependency, so use it. It is on the
+# same 1.1 version we've used before.
+RUN wget https://github.com/yrashk/pg_graphql/archive/b4988843647450a153439be367168ed09971af85.tar.gz -O pg_graphql.tar.gz && \
+    echo "0c7b0e746441b2ec24187d0e03555faf935c2159e2839bddd14df6dafbc8c9bd pg_graphql.tar.gz" | sha256sum --check && \
     mkdir pg_graphql-src && cd pg_graphql-src && tar xvzf ../pg_graphql.tar.gz --strip-components=1 -C . && \
-    sed -i 's/pgrx = "=0.10.2"/pgrx = { version = "0.10.2", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
-    cargo pgrx install --release && \
+    sed -i 's/pgx = "~0.7.1"/pgx = { version = "0.7.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
+    sed -i 's/pgx-tests = "~0.7.1"/pgx-tests = "0.7.3"/g' Cargo.toml && \
+    cargo pgx install --release && \
     # it's needed to enable extension because it uses untrusted C language
     sed -i 's/superuser = false/superuser = true/g' /usr/local/pgsql/share/extension/pg_graphql.control && \
     echo "trusted = true" >> /usr/local/pgsql/share/extension/pg_graphql.control
@@ -684,13 +656,12 @@ RUN wget https://github.com/supabase/pg_graphql/archive/refs/tags/v1.4.0.tar.gz
 #########################################################################################
 
 FROM rust-extensions-build AS pg-tiktoken-pg-build
-ARG PG_VERSION
 
-# 26806147b17b60763039c6a6878884c41a262318 made on 26/09/2023
-RUN wget https://github.com/kelvich/pg_tiktoken/archive/26806147b17b60763039c6a6878884c41a262318.tar.gz -O pg_tiktoken.tar.gz && \
-    echo "e64e55aaa38c259512d3e27c572da22c4637418cf124caba904cd50944e5004e pg_tiktoken.tar.gz" | sha256sum --check && \
+# 801f84f08c6881c8aa30f405fafbf00eec386a72 made on 10/03/2023
+RUN wget https://github.com/kelvich/pg_tiktoken/archive/801f84f08c6881c8aa30f405fafbf00eec386a72.tar.gz -O pg_tiktoken.tar.gz && \
+    echo "52f60ac800993a49aa8c609961842b611b6b1949717b69ce2ec9117117e16e4a pg_tiktoken.tar.gz" | sha256sum --check && \
     mkdir pg_tiktoken-src && cd pg_tiktoken-src && tar xvzf ../pg_tiktoken.tar.gz --strip-components=1 -C . && \
-    cargo pgrx install --release && \
+    cargo pgx install --release && \
     echo "trusted = true" >> /usr/local/pgsql/share/extension/pg_tiktoken.control
 
 #########################################################################################
@@ -701,17 +672,12 @@ RUN wget https://github.com/kelvich/pg_tiktoken/archive/26806147b17b60763039c6a6
 #########################################################################################
 
 FROM rust-extensions-build AS pg-pgx-ulid-build
-ARG PG_VERSION
 
-RUN wget https://github.com/pksunkara/pgx_ulid/archive/refs/tags/v0.1.3.tar.gz -O pgx_ulid.tar.gz && \
-    echo "ee5db82945d2d9f2d15597a80cf32de9dca67b897f605beb830561705f12683c pgx_ulid.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/pksunkara/pgx_ulid/archive/refs/tags/v0.1.0.tar.gz -O pgx_ulid.tar.gz && \
+    echo "908b7358e6f846e87db508ae5349fb56a88ee6305519074b12f3d5b0ff09f791 pgx_ulid.tar.gz" | sha256sum --check && \
     mkdir pgx_ulid-src && cd pgx_ulid-src && tar xvzf ../pgx_ulid.tar.gz --strip-components=1 -C . && \
-    echo "******************* Apply a patch for Postgres 16 support; delete in the next release ******************" && \
-    wget https://github.com/pksunkara/pgx_ulid/commit/f84954cf63fc8c80d964ac970d9eceed3c791196.patch && \
-    patch -p1 < f84954cf63fc8c80d964ac970d9eceed3c791196.patch && \
-    echo "********************************************************************************************************" && \
-    sed -i 's/pgrx       = "=0.10.2"/pgrx = { version = "=0.10.2", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
-    cargo pgrx install --release && \
+    sed -i 's/pgx        = "=0.7.3"/pgx = { version = "0.7.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
+    cargo pgx install --release && \
     echo "trusted = true" >> /usr/local/pgsql/share/extension/ulid.control
 
 #########################################################################################
@@ -760,20 +726,6 @@ RUN make -j $(getconf _NPROCESSORS_ONLN) \
         PG_CONFIG=/usr/local/pgsql/bin/pg_config \
         -C pgxn/neon_utils \
         -s install && \
-    make -j $(getconf _NPROCESSORS_ONLN) \
-        PG_CONFIG=/usr/local/pgsql/bin/pg_config \
-        -C pgxn/neon_rmgr \
-        -s install && \
-    case "${PG_VERSION}" in \
-        "v14" | "v15") \
-        ;; \
-        "v16") \
-            echo "Skipping HNSW for PostgreSQL 16" && exit 0 \
-        ;; \
-        *) \
-            echo "unexpected PostgreSQL version" && exit 1 \
-        ;; \
-        esac && \
     make -j $(getconf _NPROCESSORS_ONLN) \
         PG_CONFIG=/usr/local/pgsql/bin/pg_config \
         -C pgxn/hnsw \

Makefile

@@ -29,7 +29,6 @@ else ifeq ($(UNAME_S),Darwin)
 	# It can be configured with OPENSSL_PREFIX variable
 	OPENSSL_PREFIX ?= $(shell brew --prefix openssl@3)
 	PG_CONFIGURE_OPTS += --with-includes=$(OPENSSL_PREFIX)/include --with-libraries=$(OPENSSL_PREFIX)/lib
-	PG_CONFIGURE_OPTS += PKG_CONFIG_PATH=$(shell brew --prefix icu4c)/lib/pkgconfig
 	# macOS already has bison and flex in the system, but they are old and result in postgres-v14 target failure
 	# brew formulae are keg-only and not symlinked into HOMEBREW_PREFIX, force their usage
 	EXTRA_PATH_OVERRIDES += $(shell brew --prefix bison)/bin/:$(shell brew --prefix flex)/bin/:
@@ -84,8 +83,6 @@ $(POSTGRES_INSTALL_DIR)/build/%/config.status:
 # I'm not sure why it wouldn't work, but this is the only place (apart from
 # the "build-all-versions" entry points) where direct mention of PostgreSQL
 # versions is used.
-.PHONY: postgres-configure-v16
-postgres-configure-v16: $(POSTGRES_INSTALL_DIR)/build/v16/config.status
 .PHONY: postgres-configure-v15
 postgres-configure-v15: $(POSTGRES_INSTALL_DIR)/build/v15/config.status
 .PHONY: postgres-configure-v14
@@ -121,10 +118,6 @@ postgres-clean-%:
 	$(MAKE) -C $(POSTGRES_INSTALL_DIR)/build/$*/contrib/pageinspect clean
 	$(MAKE) -C $(POSTGRES_INSTALL_DIR)/build/$*/src/interfaces/libpq clean
 
-.PHONY: postgres-check-%
-postgres-check-%: postgres-%
-	$(MAKE) -C $(POSTGRES_INSTALL_DIR)/build/$* MAKELEVEL=0 check
-
 .PHONY: neon-pg-ext-%
 neon-pg-ext-%: postgres-%
 	+@echo "Compiling neon $*"
@@ -137,11 +130,6 @@ neon-pg-ext-%: postgres-%
 	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
 		-C $(POSTGRES_INSTALL_DIR)/build/neon-walredo-$* \
 		-f $(ROOT_PROJECT_DIR)/pgxn/neon_walredo/Makefile install
-	+@echo "Compiling neon_rmgr $*"
-	mkdir -p $(POSTGRES_INSTALL_DIR)/build/neon-rmgr-$*
-	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
-		-C $(POSTGRES_INSTALL_DIR)/build/neon-rmgr-$* \
-		-f $(ROOT_PROJECT_DIR)/pgxn/neon_rmgr/Makefile install
 	+@echo "Compiling neon_test_utils $*"
 	mkdir -p $(POSTGRES_INSTALL_DIR)/build/neon-test-utils-$*
 	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
@@ -152,6 +140,11 @@ neon-pg-ext-%: postgres-%
 	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
 		-C $(POSTGRES_INSTALL_DIR)/build/neon-utils-$* \
 		-f $(ROOT_PROJECT_DIR)/pgxn/neon_utils/Makefile install
+	+@echo "Compiling hnsw $*"
+	mkdir -p $(POSTGRES_INSTALL_DIR)/build/hnsw-$*
+	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
+		-C $(POSTGRES_INSTALL_DIR)/build/hnsw-$* \
+		-f $(ROOT_PROJECT_DIR)/pgxn/hnsw/Makefile install
 
 .PHONY: neon-pg-ext-clean-%
 neon-pg-ext-clean-%:
@@ -167,43 +160,35 @@ neon-pg-ext-clean-%:
 	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config \
 	-C $(POSTGRES_INSTALL_DIR)/build/neon-utils-$* \
 	-f $(ROOT_PROJECT_DIR)/pgxn/neon_utils/Makefile clean
+	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config \
+	-C $(POSTGRES_INSTALL_DIR)/build/hnsw-$* \
+	-f $(ROOT_PROJECT_DIR)/pgxn/hnsw/Makefile clean
 
 .PHONY: neon-pg-ext
 neon-pg-ext: \
 	neon-pg-ext-v14 \
-	neon-pg-ext-v15 \
-	neon-pg-ext-v16
+	neon-pg-ext-v15
 
 .PHONY: neon-pg-ext-clean
 neon-pg-ext-clean: \
 	neon-pg-ext-clean-v14 \
-	neon-pg-ext-clean-v15 \
-	neon-pg-ext-clean-v16
+	neon-pg-ext-clean-v15
 
 # shorthand to build all Postgres versions
 .PHONY: postgres
 postgres: \
 	postgres-v14 \
-	postgres-v15 \
-	postgres-v16
+	postgres-v15
 
 .PHONY: postgres-headers
 postgres-headers: \
 	postgres-headers-v14 \
-	postgres-headers-v15 \
-	postgres-headers-v16
+	postgres-headers-v15
 
 .PHONY: postgres-clean
 postgres-clean: \
 	postgres-clean-v14 \
-	postgres-clean-v15 \
-	postgres-clean-v16
-
-.PHONY: postgres-check
-postgres-check: \
-	postgres-check-v14 \
-	postgres-check-v15 \
-	postgres-check-v16
+	postgres-clean-v15
 
 # This doesn't remove the effects of 'configure'.
 .PHONY: clean

README.md

@@ -29,18 +29,18 @@ See developer documentation in [SUMMARY.md](/docs/SUMMARY.md) for more informati
 ```bash
 apt install build-essential libtool libreadline-dev zlib1g-dev flex bison libseccomp-dev \
 libssl-dev clang pkg-config libpq-dev cmake postgresql-client protobuf-compiler \
-libcurl4-openssl-dev openssl python-poetry lsof libicu-dev
+libcurl4-openssl-dev openssl python-poetry
 ```
 * On Fedora, these packages are needed:
 ```bash
 dnf install flex bison readline-devel zlib-devel openssl-devel \
   libseccomp-devel perl clang cmake postgresql postgresql-contrib protobuf-compiler \
-  protobuf-devel libcurl-devel openssl poetry lsof libicu-devel
+  protobuf-devel libcurl-devel openssl poetry
 ```
 * On Arch based systems, these packages are needed:
 ```bash
 pacman -S base-devel readline zlib libseccomp openssl clang \
-postgresql-libs cmake postgresql protobuf curl lsof
+postgresql-libs cmake postgresql protobuf curl
 ```
 
 Building Neon requires 3.15+ version of `protoc` (protobuf-compiler). If your distribution provides an older version, you can install a newer version from [here](https://github.com/protocolbuffers/protobuf/releases).
@@ -55,7 +55,7 @@ curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
 1. Install XCode and dependencies
 ```
 xcode-select --install
-brew install protobuf openssl flex bison icu4c pkg-config
+brew install protobuf openssl flex bison
 
 # add openssl to PATH, required for ed25519 keys generation in neon_local
 echo 'export PATH="$(brew --prefix openssl)/bin:$PATH"' >> ~/.zshrc

clippy.toml

@@ -1,5 +0,0 @@
-disallowed-methods = [
-    "tokio::task::block_in_place",
-    # Allow this for now, to deny it later once we stop using Handle::block_on completely
-    # "tokio::runtime::Handle::block_on",
-]

compute_tools/src/checker.rs

@@ -1,7 +1,7 @@
 use anyhow::{anyhow, Ok, Result};
 use postgres::Client;
 use tokio_postgres::NoTls;
-use tracing::{error, instrument, warn};
+use tracing::{error, instrument};
 
 use crate::compute::ComputeNode;
 
@@ -55,24 +55,13 @@ pub async fn check_writability(compute: &ComputeNode) -> Result<()> {
         ON CONFLICT (id) DO UPDATE
          SET updated_at = now();";
 
-    match client.simple_query(query).await {
-        Result::Ok(result) => {
-            if result.len() != 1 {
-                return Err(anyhow::anyhow!(
-                    "expected 1 query results, but got {}",
-                    result.len()
-                ));
-            }
-        }
-        Err(err) => {
-            if let Some(state) = err.code() {
-                if state == &tokio_postgres::error::SqlState::DISK_FULL {
-                    warn!("Tenant disk is full");
-                    return Ok(());
-                }
-            }
-            return Err(err.into());
-        }
+    let result = client.simple_query(query).await?;
+
+    if result.len() != 1 {
+        return Err(anyhow::format_err!(
+            "expected 1 query result, but got {}",
+            result.len()
+        ));
     }
 
     Ok(())

compute_tools/src/compute.rs

@@ -1039,7 +1039,7 @@ LIMIT 100",
         let remote_extensions = spec
             .remote_extensions
             .as_ref()
-            .ok_or(anyhow::anyhow!("Remote extensions are not configured"))?;
+            .ok_or(anyhow::anyhow!("Remote extensions are not configured",))?;
 
         info!("parse shared_preload_libraries from spec.cluster.settings");
         let mut libs_vec = Vec::new();

compute_tools/src/config.rs

@@ -46,6 +46,8 @@ pub fn write_postgres_conf(
         writeln!(file, "{}", conf)?;
     }
 
+    write!(file, "{}", &spec.cluster.settings.as_pg_settings())?;
+
     // Add options for connecting to storage
     writeln!(file, "# Neon storage settings")?;
     if let Some(s) = &spec.pageserver_connstring {

compute_tools/src/extension_server.rs

@@ -74,7 +74,6 @@ More specifically, here is an example ext_index.json
 use anyhow::Context;
 use anyhow::{self, Result};
 use compute_api::spec::RemoteExtSpec;
-use regex::Regex;
 use remote_storage::*;
 use serde_json;
 use std::io::Read;
@@ -107,71 +106,16 @@ fn get_pg_config(argument: &str, pgbin: &str) -> String {
 
 pub fn get_pg_version(pgbin: &str) -> String {
     // pg_config --version returns a (platform specific) human readable string
-    // such as "PostgreSQL 15.4". We parse this to v14/v15/v16 etc.
+    // such as "PostgreSQL 15.4". We parse this to v14/v15
     let human_version = get_pg_config("--version", pgbin);
-    return parse_pg_version(&human_version).to_string();
-}
-
-fn parse_pg_version(human_version: &str) -> &str {
-    // Normal releases have version strings like "PostgreSQL 15.4". But there
-    // are also pre-release versions like "PostgreSQL 17devel" or "PostgreSQL
-    // 16beta2" or "PostgreSQL 17rc1". And with the --with-extra-version
-    // configure option, you can tack any string to the version number,
-    // e.g. "PostgreSQL 15.4foobar".
-    match Regex::new(r"^PostgreSQL (?<major>\d+).+")
-        .unwrap()
-        .captures(human_version)
-    {
-        Some(captures) if captures.len() == 2 => match &captures["major"] {
-            "14" => return "v14",
-            "15" => return "v15",
-            "16" => return "v16",
-            _ => {}
-        },
-        _ => {}
+    if human_version.contains("15") {
+        return "v15".to_string();
+    } else if human_version.contains("14") {
+        return "v14".to_string();
     }
     panic!("Unsuported postgres version {human_version}");
 }
 
-#[cfg(test)]
-mod tests {
-    use super::parse_pg_version;
-
-    #[test]
-    fn test_parse_pg_version() {
-        assert_eq!(parse_pg_version("PostgreSQL 15.4"), "v15");
-        assert_eq!(parse_pg_version("PostgreSQL 15.14"), "v15");
-        assert_eq!(
-            parse_pg_version("PostgreSQL 15.4 (Ubuntu 15.4-0ubuntu0.23.04.1)"),
-            "v15"
-        );
-
-        assert_eq!(parse_pg_version("PostgreSQL 14.15"), "v14");
-        assert_eq!(parse_pg_version("PostgreSQL 14.0"), "v14");
-        assert_eq!(
-            parse_pg_version("PostgreSQL 14.9 (Debian 14.9-1.pgdg120+1"),
-            "v14"
-        );
-
-        assert_eq!(parse_pg_version("PostgreSQL 16devel"), "v16");
-        assert_eq!(parse_pg_version("PostgreSQL 16beta1"), "v16");
-        assert_eq!(parse_pg_version("PostgreSQL 16rc2"), "v16");
-        assert_eq!(parse_pg_version("PostgreSQL 16extra"), "v16");
-    }
-
-    #[test]
-    #[should_panic]
-    fn test_parse_pg_unsupported_version() {
-        parse_pg_version("PostgreSQL 13.14");
-    }
-
-    #[test]
-    #[should_panic]
-    fn test_parse_pg_incorrect_version_format() {
-        parse_pg_version("PostgreSQL 14");
-    }
-}
-
 // download the archive for a given extension,
 // unzip it, and place files in the appropriate locations (share/lib)
 pub async fn download_extension(

compute_tools/src/monitor.rs

@@ -1,5 +1,5 @@
 use std::sync::Arc;
-use std::{thread, time::Duration};
+use std::{thread, time};
 
 use chrono::{DateTime, Utc};
 use postgres::{Client, NoTls};
@@ -7,7 +7,7 @@ use tracing::{debug, info};
 
 use crate::compute::ComputeNode;
 
-const MONITOR_CHECK_INTERVAL: Duration = Duration::from_millis(500);
+const MONITOR_CHECK_INTERVAL: u64 = 500; // milliseconds
 
 // Spin in a loop and figure out the last activity time in the Postgres.
 // Then update it in the shared state. This function never errors out.
@@ -17,12 +17,13 @@ fn watch_compute_activity(compute: &ComputeNode) {
     let connstr = compute.connstr.as_str();
     // Define `client` outside of the loop to reuse existing connection if it's active.
     let mut client = Client::connect(connstr, NoTls);
+    let timeout = time::Duration::from_millis(MONITOR_CHECK_INTERVAL);
 
     info!("watching Postgres activity at {}", connstr);
 
     loop {
         // Should be outside of the write lock to allow others to read while we sleep.
-        thread::sleep(MONITOR_CHECK_INTERVAL);
+        thread::sleep(timeout);
 
         match &mut client {
             Ok(cli) => {

control_plane/Cargo.toml

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

control_plane/simple.conf

@@ -1,7 +1,6 @@
 # Minimal neon environment with one safekeeper. This is equivalent to the built-in
 # defaults that you get with no --config
-[[pageservers]]
-id=1
+[pageserver]
 listen_pg_addr = '127.0.0.1:64000'
 listen_http_addr = '127.0.0.1:9898'
 pg_auth_type = 'Trust'

control_plane/src/attachment_service.rs

@@ -1,6 +1,5 @@
 use crate::{background_process, local_env::LocalEnv};
 use anyhow::anyhow;
-use camino::Utf8PathBuf;
 use serde::{Deserialize, Serialize};
 use serde_with::{serde_as, DisplayFromStr};
 use std::{path::PathBuf, process::Child};
@@ -33,7 +32,7 @@ impl AttachmentService {
 
         // Makes no sense to construct this if pageservers aren't going to use it: assume
         // pageservers have control plane API set
-        let listen_url = env.control_plane_api.clone().unwrap();
+        let listen_url = env.pageserver.control_plane_api.clone().unwrap();
 
         let listen = format!(
             "{}:{}",
@@ -48,9 +47,8 @@ impl AttachmentService {
         }
     }
 
-    fn pid_file(&self) -> Utf8PathBuf {
-        Utf8PathBuf::from_path_buf(self.env.base_data_dir.join("attachment_service.pid"))
-            .expect("non-Unicode path")
+    fn pid_file(&self) -> PathBuf {
+        self.env.base_data_dir.join("attachment_service.pid")
     }
 
     pub fn start(&self) -> anyhow::Result<Child> {
@@ -82,6 +80,7 @@ impl AttachmentService {
 
         let url = self
             .env
+            .pageserver
             .control_plane_api
             .clone()
             .unwrap()

control_plane/src/background_process.rs

@@ -16,13 +16,12 @@ use std::ffi::OsStr;
 use std::io::Write;
 use std::os::unix::prelude::AsRawFd;
 use std::os::unix::process::CommandExt;
-use std::path::Path;
+use std::path::{Path, PathBuf};
 use std::process::{Child, Command};
 use std::time::Duration;
 use std::{fs, io, thread};
 
 use anyhow::Context;
-use camino::{Utf8Path, Utf8PathBuf};
 use nix::errno::Errno;
 use nix::fcntl::{FcntlArg, FdFlag};
 use nix::sys::signal::{kill, Signal};
@@ -46,9 +45,9 @@ const NOTICE_AFTER_RETRIES: u64 = 50;
 /// it itself.
 pub enum InitialPidFile<'t> {
     /// Create a pidfile, to allow future CLI invocations to manipulate the process.
-    Create(&'t Utf8Path),
+    Create(&'t Path),
     /// The process will create the pidfile itself, need to wait for that event.
-    Expect(&'t Utf8Path),
+    Expect(&'t Path),
 }
 
 /// Start a background child process using the parameters given.
@@ -138,11 +137,7 @@ where
 }
 
 /// Stops the process, using the pid file given. Returns Ok also if the process is already not running.
-pub fn stop_process(
-    immediate: bool,
-    process_name: &str,
-    pid_file: &Utf8Path,
-) -> anyhow::Result<()> {
+pub fn stop_process(immediate: bool, process_name: &str, pid_file: &Path) -> anyhow::Result<()> {
     let pid = match pid_file::read(pid_file)
         .with_context(|| format!("read pid_file {pid_file:?}"))?
     {
@@ -257,9 +252,9 @@ fn fill_aws_secrets_vars(mut cmd: &mut Command) -> &mut Command {
 ///    will remain held until the cmd exits.
 fn pre_exec_create_pidfile<P>(cmd: &mut Command, path: P) -> &mut Command
 where
-    P: Into<Utf8PathBuf>,
+    P: Into<PathBuf>,
 {
-    let path: Utf8PathBuf = path.into();
+    let path: PathBuf = path.into();
     // SAFETY
     // pre_exec is marked unsafe because it runs between fork and exec.
     // Why is that dangerous in various ways?
@@ -316,7 +311,7 @@ where
 
 fn process_started<F>(
     pid: Pid,
-    pid_file_to_check: Option<&Utf8Path>,
+    pid_file_to_check: Option<&Path>,
     status_check: &F,
 ) -> anyhow::Result<bool>
 where

control_plane/src/bin/attachment_service.rs

@@ -223,7 +223,6 @@ async fn handle_attach_hook(mut req: Request<Body>) -> Result<Response<Body>, Ap
     if attach_req.pageserver_id.is_some() {
         tenant_state.generation += 1;
     }
-    tenant_state.pageserver = attach_req.pageserver_id;
     let generation = tenant_state.generation;
 
     locked.save().await.map_err(ApiError::InternalServerError)?;

control_plane/src/bin/neon_local.rs

@@ -50,17 +50,16 @@ fn default_conf() -> String {
     format!(
         r#"
 # Default built-in configuration, defined in main.rs
-control_plane_api = '{DEFAULT_PAGESERVER_CONTROL_PLANE_API}'
-
 [broker]
 listen_addr = '{DEFAULT_BROKER_ADDR}'
 
-[[pageservers]]
+[pageserver]
 id = {DEFAULT_PAGESERVER_ID}
 listen_pg_addr = '{DEFAULT_PAGESERVER_PG_ADDR}'
 listen_http_addr = '{DEFAULT_PAGESERVER_HTTP_ADDR}'
 pg_auth_type = '{trust_auth}'
 http_auth_type = '{trust_auth}'
+control_plane_api = '{DEFAULT_PAGESERVER_CONTROL_PLANE_API}'
 
 [[safekeepers]]
 id = {DEFAULT_SAFEKEEPER_ID}
@@ -259,7 +258,7 @@ fn get_timeline_infos(
     env: &local_env::LocalEnv,
     tenant_id: &TenantId,
 ) -> Result<HashMap<TimelineId, TimelineInfo>> {
-    Ok(get_default_pageserver(env)
+    Ok(PageServerNode::from_env(env)
         .timeline_list(tenant_id)?
         .into_iter()
         .map(|timeline_info| (timeline_info.timeline_id, timeline_info))
@@ -320,30 +319,17 @@ fn handle_init(init_match: &ArgMatches) -> anyhow::Result<LocalEnv> {
         .context("Failed to initialize neon repository")?;
 
     // Initialize pageserver, create initial tenant and timeline.
-    for ps_conf in &env.pageservers {
-        PageServerNode::from_env(&env, ps_conf)
-            .initialize(&pageserver_config_overrides(init_match))
-            .unwrap_or_else(|e| {
-                eprintln!("pageserver init failed: {e:?}");
-                exit(1);
-            });
-    }
+    let pageserver = PageServerNode::from_env(&env);
+    pageserver
+        .initialize(&pageserver_config_overrides(init_match))
+        .unwrap_or_else(|e| {
+            eprintln!("pageserver init failed: {e:?}");
+            exit(1);
+        });
 
     Ok(env)
 }
 
-/// The default pageserver is the one where CLI tenant/timeline operations are sent by default.
-/// For typical interactive use, one would just run with a single pageserver.  Scenarios with
-/// tenant/timeline placement across multiple pageservers are managed by python test code rather
-/// than this CLI.
-fn get_default_pageserver(env: &local_env::LocalEnv) -> PageServerNode {
-    let ps_conf = env
-        .pageservers
-        .first()
-        .expect("Config is validated to contain at least one pageserver");
-    PageServerNode::from_env(env, ps_conf)
-}
-
 fn pageserver_config_overrides(init_match: &ArgMatches) -> Vec<&str> {
     init_match
         .get_many::<String>("pageserver-config-override")
@@ -354,7 +340,7 @@ fn pageserver_config_overrides(init_match: &ArgMatches) -> Vec<&str> {
 }
 
 fn handle_tenant(tenant_match: &ArgMatches, env: &mut local_env::LocalEnv) -> anyhow::Result<()> {
-    let pageserver = get_default_pageserver(env);
+    let pageserver = PageServerNode::from_env(env);
     match tenant_match.subcommand() {
         Some(("list", _)) => {
             for t in pageserver.tenant_list()? {
@@ -370,11 +356,11 @@ fn handle_tenant(tenant_match: &ArgMatches, env: &mut local_env::LocalEnv) -> an
             // If tenant ID was not specified, generate one
             let tenant_id = parse_tenant_id(create_match)?.unwrap_or_else(TenantId::generate);
 
-            let generation = if env.control_plane_api.is_some() {
+            let generation = if env.pageserver.control_plane_api.is_some() {
                 // We must register the tenant with the attachment service, so
                 // that when the pageserver restarts, it will be re-attached.
                 let attachment_service = AttachmentService::from_env(env);
-                attachment_service.attach_hook(tenant_id, pageserver.conf.id)?
+                attachment_service.attach_hook(tenant_id, env.pageserver.id)?
             } else {
                 None
             };
@@ -439,7 +425,7 @@ fn handle_tenant(tenant_match: &ArgMatches, env: &mut local_env::LocalEnv) -> an
 }
 
 fn handle_timeline(timeline_match: &ArgMatches, env: &mut local_env::LocalEnv) -> Result<()> {
-    let pageserver = get_default_pageserver(env);
+    let pageserver = PageServerNode::from_env(env);
 
     match timeline_match.subcommand() {
         Some(("list", list_match)) => {
@@ -516,7 +502,6 @@ fn handle_timeline(timeline_match: &ArgMatches, env: &mut local_env::LocalEnv) -
                 None,
                 pg_version,
                 ComputeMode::Primary,
-                DEFAULT_PAGESERVER_ID,
             )?;
             println!("Done");
         }
@@ -570,6 +555,7 @@ fn handle_endpoint(ep_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<(
         Some(ep_subcommand_data) => ep_subcommand_data,
         None => bail!("no endpoint subcommand provided"),
     };
+
     let mut cplane = ComputeControlPlane::load(env.clone())?;
 
     // All subcommands take an optional --tenant-id option
@@ -666,13 +652,6 @@ fn handle_endpoint(ep_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<(
                 .copied()
                 .unwrap_or(false);
 
-            let pageserver_id =
-                if let Some(id_str) = sub_args.get_one::<String>("endpoint-pageserver-id") {
-                    NodeId(id_str.parse().context("while parsing pageserver id")?)
-                } else {
-                    DEFAULT_PAGESERVER_ID
-                };
-
             let mode = match (lsn, hot_standby) {
                 (Some(lsn), false) => ComputeMode::Static(lsn),
                 (None, true) => ComputeMode::Replica,
@@ -688,7 +667,6 @@ fn handle_endpoint(ep_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<(
                 http_port,
                 pg_version,
                 mode,
-                pageserver_id,
             )?;
         }
         "start" => {
@@ -698,13 +676,6 @@ fn handle_endpoint(ep_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<(
                 .get_one::<String>("endpoint_id")
                 .ok_or_else(|| anyhow!("No endpoint ID was provided to start"))?;
 
-            let pageserver_id =
-                if let Some(id_str) = sub_args.get_one::<String>("endpoint-pageserver-id") {
-                    NodeId(id_str.parse().context("while parsing pageserver id")?)
-                } else {
-                    DEFAULT_PAGESERVER_ID
-                };
-
             let remote_ext_config = sub_args.get_one::<String>("remote-ext-config");
 
             // If --safekeepers argument is given, use only the listed safekeeper nodes.
@@ -724,8 +695,7 @@ fn handle_endpoint(ep_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<(
 
             let endpoint = cplane.endpoints.get(endpoint_id.as_str());
 
-            let ps_conf = env.get_pageserver_conf(pageserver_id)?;
-            let auth_token = if matches!(ps_conf.pg_auth_type, AuthType::NeonJWT) {
+            let auth_token = if matches!(env.pageserver.pg_auth_type, AuthType::NeonJWT) {
                 let claims = Claims::new(Some(tenant_id), Scope::Tenant);
 
                 Some(env.generate_auth_token(&claims)?)
@@ -792,7 +762,6 @@ fn handle_endpoint(ep_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<(
                     http_port,
                     pg_version,
                     mode,
-                    pageserver_id,
                 )?;
                 ep.start(&auth_token, safekeepers, remote_ext_config)?;
             }
@@ -817,64 +786,48 @@ fn handle_endpoint(ep_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<(
 }
 
 fn handle_pageserver(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
-    fn get_pageserver(env: &local_env::LocalEnv, args: &ArgMatches) -> Result<PageServerNode> {
-        let node_id = if let Some(id_str) = args.get_one::<String>("pageserver-id") {
-            NodeId(id_str.parse().context("while parsing pageserver id")?)
-        } else {
-            DEFAULT_PAGESERVER_ID
-        };
-
-        Ok(PageServerNode::from_env(
-            env,
-            env.get_pageserver_conf(node_id)?,
-        ))
-    }
+    let pageserver = PageServerNode::from_env(env);
 
     match sub_match.subcommand() {
-        Some(("start", subcommand_args)) => {
-            if let Err(e) = get_pageserver(env, subcommand_args)?
-                .start(&pageserver_config_overrides(subcommand_args))
-            {
+        Some(("start", start_match)) => {
+            if let Err(e) = pageserver.start(&pageserver_config_overrides(start_match)) {
                 eprintln!("pageserver start failed: {e}");
                 exit(1);
             }
         }
 
-        Some(("stop", subcommand_args)) => {
-            let immediate = subcommand_args
+        Some(("stop", stop_match)) => {
+            let immediate = stop_match
                 .get_one::<String>("stop-mode")
                 .map(|s| s.as_str())
                 == Some("immediate");
 
-            if let Err(e) = get_pageserver(env, subcommand_args)?.stop(immediate) {
+            if let Err(e) = pageserver.stop(immediate) {
                 eprintln!("pageserver stop failed: {}", e);
                 exit(1);
             }
         }
 
-        Some(("restart", subcommand_args)) => {
-            let pageserver = get_pageserver(env, subcommand_args)?;
+        Some(("restart", restart_match)) => {
             //TODO what shutdown strategy should we use here?
             if let Err(e) = pageserver.stop(false) {
                 eprintln!("pageserver stop failed: {}", e);
                 exit(1);
             }
 
-            if let Err(e) = pageserver.start(&pageserver_config_overrides(subcommand_args)) {
+            if let Err(e) = pageserver.start(&pageserver_config_overrides(restart_match)) {
                 eprintln!("pageserver start failed: {e}");
                 exit(1);
             }
         }
 
-        Some(("status", subcommand_args)) => {
-            match get_pageserver(env, subcommand_args)?.check_status() {
-                Ok(_) => println!("Page server is up and running"),
-                Err(err) => {
-                    eprintln!("Page server is not available: {}", err);
-                    exit(1);
-                }
+        Some(("status", _)) => match PageServerNode::from_env(env).check_status() {
+            Ok(_) => println!("Page server is up and running"),
+            Err(err) => {
+                eprintln!("Page server is not available: {}", err);
+                exit(1);
             }
-        }
+        },
 
         Some((sub_name, _)) => bail!("Unexpected pageserver subcommand '{}'", sub_name),
         None => bail!("no pageserver subcommand provided"),
@@ -990,7 +943,7 @@ fn handle_start_all(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> anyhow
     broker::start_broker_process(env)?;
 
     // Only start the attachment service if the pageserver is configured to need it
-    if env.control_plane_api.is_some() {
+    if env.pageserver.control_plane_api.is_some() {
         let attachment_service = AttachmentService::from_env(env);
         if let Err(e) = attachment_service.start() {
             eprintln!("attachment_service start failed: {:#}", e);
@@ -999,13 +952,11 @@ fn handle_start_all(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> anyhow
         }
     }
 
-    for ps_conf in &env.pageservers {
-        let pageserver = PageServerNode::from_env(env, ps_conf);
-        if let Err(e) = pageserver.start(&pageserver_config_overrides(sub_match)) {
-            eprintln!("pageserver {} start failed: {:#}", ps_conf.id, e);
-            try_stop_all(env, true);
-            exit(1);
-        }
+    let pageserver = PageServerNode::from_env(env);
+    if let Err(e) = pageserver.start(&pageserver_config_overrides(sub_match)) {
+        eprintln!("pageserver {} start failed: {:#}", env.pageserver.id, e);
+        try_stop_all(env, true);
+        exit(1);
     }
 
     for node in env.safekeepers.iter() {
@@ -1029,6 +980,8 @@ fn handle_stop_all(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<
 }
 
 fn try_stop_all(env: &local_env::LocalEnv, immediate: bool) {
+    let pageserver = PageServerNode::from_env(env);
+
     // Stop all endpoints
     match ComputeControlPlane::load(env.clone()) {
         Ok(cplane) => {
@@ -1043,11 +996,8 @@ fn try_stop_all(env: &local_env::LocalEnv, immediate: bool) {
         }
     }
 
-    for ps_conf in &env.pageservers {
-        let pageserver = PageServerNode::from_env(env, ps_conf);
-        if let Err(e) = pageserver.stop(immediate) {
-            eprintln!("pageserver {} stop failed: {:#}", ps_conf.id, e);
-        }
+    if let Err(e) = pageserver.stop(immediate) {
+        eprintln!("pageserver {} stop failed: {:#}", env.pageserver.id, e);
     }
 
     for node in env.safekeepers.iter() {
@@ -1061,7 +1011,7 @@ fn try_stop_all(env: &local_env::LocalEnv, immediate: bool) {
         eprintln!("neon broker stop failed: {e:#}");
     }
 
-    if env.control_plane_api.is_some() {
+    if env.pageserver.control_plane_api.is_some() {
         let attachment_service = AttachmentService::from_env(env);
         if let Err(e) = attachment_service.stop(immediate) {
             eprintln!("attachment service stop failed: {e:#}");
@@ -1081,16 +1031,6 @@ fn cli() -> Command {
 
     let safekeeper_id_arg = Arg::new("id").help("safekeeper id").required(false);
 
-    // --id, when using a pageserver command
-    let pageserver_id_arg = Arg::new("pageserver-id")
-        .long("id")
-        .help("pageserver id")
-        .required(false);
-    // --pageserver-id when using a non-pageserver command
-    let endpoint_pageserver_id_arg = Arg::new("endpoint-pageserver-id")
-        .long("pageserver-id")
-        .required(false);
-
     let safekeeper_extra_opt_arg = Arg::new("safekeeper-extra-opt")
         .short('e')
         .long("safekeeper-extra-opt")
@@ -1255,16 +1195,10 @@ fn cli() -> Command {
                 .arg_required_else_help(true)
                 .about("Manage pageserver")
                 .subcommand(Command::new("status"))
-                .arg(pageserver_id_arg.clone())
-                .subcommand(Command::new("start").about("Start local pageserver")
-                .arg(pageserver_id_arg.clone())
-                .arg(pageserver_config_args.clone()))
+                .subcommand(Command::new("start").about("Start local pageserver").arg(pageserver_config_args.clone()))
                 .subcommand(Command::new("stop").about("Stop local pageserver")
-                .arg(pageserver_id_arg.clone())
                             .arg(stop_mode_arg.clone()))
-                .subcommand(Command::new("restart").about("Restart local pageserver")
-                .arg(pageserver_id_arg.clone())
-                .arg(pageserver_config_args.clone()))
+                .subcommand(Command::new("restart").about("Restart local pageserver").arg(pageserver_config_args.clone()))
         )
         .subcommand(
             Command::new("attachment_service")
@@ -1308,7 +1242,6 @@ fn cli() -> Command {
                     .arg(lsn_arg.clone())
                     .arg(pg_port_arg.clone())
                     .arg(http_port_arg.clone())
-                    .arg(endpoint_pageserver_id_arg.clone())
                     .arg(
                         Arg::new("config-only")
                             .help("Don't do basebackup, create endpoint directory with only config files")
@@ -1326,7 +1259,6 @@ fn cli() -> Command {
                     .arg(lsn_arg)
                     .arg(pg_port_arg)
                     .arg(http_port_arg)
-                    .arg(endpoint_pageserver_id_arg.clone())
                     .arg(pg_version_arg)
                     .arg(hot_standby_arg)
                     .arg(safekeepers_arg)

control_plane/src/broker.rs

@@ -7,7 +7,7 @@
 //! ```
 use anyhow::Context;
 
-use camino::Utf8PathBuf;
+use std::path::PathBuf;
 
 use crate::{background_process, local_env};
 
@@ -30,7 +30,7 @@ pub fn start_broker_process(env: &local_env::LocalEnv) -> anyhow::Result<()> {
         || {
             let url = broker.client_url();
             let status_url = url.join("status").with_context(|| {
-                format!("Failed to append /status path to broker endpoint {url}")
+                format!("Failed to append /status path to broker endpoint {url}",)
             })?;
             let request = client
                 .get(status_url)
@@ -50,7 +50,6 @@ pub fn stop_broker_process(env: &local_env::LocalEnv) -> anyhow::Result<()> {
     background_process::stop_process(true, "storage_broker", &storage_broker_pid_file_path(env))
 }
 
-fn storage_broker_pid_file_path(env: &local_env::LocalEnv) -> Utf8PathBuf {
-    Utf8PathBuf::from_path_buf(env.base_data_dir.join("storage_broker.pid"))
-        .expect("non-Unicode path")
+fn storage_broker_pid_file_path(env: &local_env::LocalEnv) -> PathBuf {
+    env.base_data_dir.join("storage_broker.pid")
 }

control_plane/src/endpoint.rs

@@ -70,7 +70,6 @@ pub struct EndpointConf {
     http_port: u16,
     pg_version: u32,
     skip_pg_catalog_updates: bool,
-    pageserver_id: NodeId,
 }
 
 //
@@ -83,16 +82,19 @@ pub struct ComputeControlPlane {
     pub endpoints: BTreeMap<String, Arc<Endpoint>>,
 
     env: LocalEnv,
+    pageserver: Arc<PageServerNode>,
 }
 
 impl ComputeControlPlane {
     // Load current endpoints from the endpoints/ subdirectories
     pub fn load(env: LocalEnv) -> Result<ComputeControlPlane> {
+        let pageserver = Arc::new(PageServerNode::from_env(&env));
+
         let mut endpoints = BTreeMap::default();
         for endpoint_dir in std::fs::read_dir(env.endpoints_path())
             .with_context(|| format!("failed to list {}", env.endpoints_path().display()))?
         {
-            let ep = Endpoint::from_dir_entry(endpoint_dir?, &env)?;
+            let ep = Endpoint::from_dir_entry(endpoint_dir?, &env, &pageserver)?;
             endpoints.insert(ep.endpoint_id.clone(), Arc::new(ep));
         }
 
@@ -100,6 +102,7 @@ impl ComputeControlPlane {
             base_port: 55431,
             endpoints,
             env,
+            pageserver,
         })
     }
 
@@ -122,18 +125,15 @@ impl ComputeControlPlane {
         http_port: Option<u16>,
         pg_version: u32,
         mode: ComputeMode,
-        pageserver_id: NodeId,
     ) -> Result<Arc<Endpoint>> {
         let pg_port = pg_port.unwrap_or_else(|| self.get_port());
         let http_port = http_port.unwrap_or_else(|| self.get_port() + 1);
-        let pageserver =
-            PageServerNode::from_env(&self.env, self.env.get_pageserver_conf(pageserver_id)?);
         let ep = Arc::new(Endpoint {
             endpoint_id: endpoint_id.to_owned(),
             pg_address: SocketAddr::new("127.0.0.1".parse().unwrap(), pg_port),
             http_address: SocketAddr::new("127.0.0.1".parse().unwrap(), http_port),
             env: self.env.clone(),
-            pageserver,
+            pageserver: Arc::clone(&self.pageserver),
             timeline_id,
             mode,
             tenant_id,
@@ -159,7 +159,6 @@ impl ComputeControlPlane {
                 pg_port,
                 pg_version,
                 skip_pg_catalog_updates: true,
-                pageserver_id,
             })?,
         )?;
         std::fs::write(
@@ -194,14 +193,18 @@ pub struct Endpoint {
     // These are not part of the endpoint as such, but the environment
     // the endpoint runs in.
     pub env: LocalEnv,
-    pageserver: PageServerNode,
+    pageserver: Arc<PageServerNode>,
 
     // Optimizations
     skip_pg_catalog_updates: bool,
 }
 
 impl Endpoint {
-    fn from_dir_entry(entry: std::fs::DirEntry, env: &LocalEnv) -> Result<Endpoint> {
+    fn from_dir_entry(
+        entry: std::fs::DirEntry,
+        env: &LocalEnv,
+        pageserver: &Arc<PageServerNode>,
+    ) -> Result<Endpoint> {
         if !entry.file_type()?.is_dir() {
             anyhow::bail!(
                 "Endpoint::from_dir_entry failed: '{}' is not a directory",
@@ -217,15 +220,12 @@ impl Endpoint {
         let conf: EndpointConf =
             serde_json::from_slice(&std::fs::read(entry.path().join("endpoint.json"))?)?;
 
-        let pageserver =
-            PageServerNode::from_env(env, env.get_pageserver_conf(conf.pageserver_id)?);
-
         Ok(Endpoint {
             pg_address: SocketAddr::new("127.0.0.1".parse().unwrap(), conf.pg_port),
             http_address: SocketAddr::new("127.0.0.1".parse().unwrap(), conf.http_port),
             endpoint_id,
             env: env.clone(),
-            pageserver,
+            pageserver: Arc::clone(pageserver),
             timeline_id: conf.timeline_id,
             mode: conf.mode,
             tenant_id: conf.tenant_id,

control_plane/src/local_env.rs

@@ -68,17 +68,11 @@ pub struct LocalEnv {
 
     pub broker: NeonBroker,
 
-    /// This Vec must always contain at least one pageserver
-    pub pageservers: Vec<PageServerConf>,
+    pub pageserver: PageServerConf,
 
     #[serde(default)]
     pub safekeepers: Vec<SafekeeperConf>,
 
-    // Control plane location: if None, we will not run attachment_service.  If set, this will
-    // be propagated into each pageserver's configuration.
-    #[serde(default)]
-    pub control_plane_api: Option<Url>,
-
     /// Keep human-readable aliases in memory (and persist them to config), to hide ZId hex strings from the user.
     #[serde(default)]
     // A `HashMap<String, HashMap<TenantId, TimelineId>>` would be more appropriate here,
@@ -124,6 +118,9 @@ pub struct PageServerConf {
     // auth type used for the PG and HTTP ports
     pub pg_auth_type: AuthType,
     pub http_auth_type: AuthType,
+
+    // Control plane location
+    pub control_plane_api: Option<Url>,
 }
 
 impl Default for PageServerConf {
@@ -134,6 +131,7 @@ impl Default for PageServerConf {
             listen_http_addr: String::new(),
             pg_auth_type: AuthType::Trust,
             http_auth_type: AuthType::Trust,
+            control_plane_api: None,
         }
     }
 }
@@ -182,18 +180,26 @@ impl LocalEnv {
     pub fn pg_distrib_dir(&self, pg_version: u32) -> anyhow::Result<PathBuf> {
         let path = self.pg_distrib_dir.clone();
 
-        #[allow(clippy::manual_range_patterns)]
         match pg_version {
-            14 | 15 | 16 => Ok(path.join(format!("v{pg_version}"))),
+            14 => Ok(path.join(format!("v{pg_version}"))),
+            15 => Ok(path.join(format!("v{pg_version}"))),
             _ => bail!("Unsupported postgres version: {}", pg_version),
         }
     }
 
     pub fn pg_bin_dir(&self, pg_version: u32) -> anyhow::Result<PathBuf> {
-        Ok(self.pg_distrib_dir(pg_version)?.join("bin"))
+        match pg_version {
+            14 => Ok(self.pg_distrib_dir(pg_version)?.join("bin")),
+            15 => Ok(self.pg_distrib_dir(pg_version)?.join("bin")),
+            _ => bail!("Unsupported postgres version: {}", pg_version),
+        }
     }
     pub fn pg_lib_dir(&self, pg_version: u32) -> anyhow::Result<PathBuf> {
-        Ok(self.pg_distrib_dir(pg_version)?.join("lib"))
+        match pg_version {
+            14 => Ok(self.pg_distrib_dir(pg_version)?.join("lib")),
+            15 => Ok(self.pg_distrib_dir(pg_version)?.join("lib")),
+            _ => bail!("Unsupported postgres version: {}", pg_version),
+        }
     }
 
     pub fn pageserver_bin(&self) -> PathBuf {
@@ -216,23 +222,15 @@ impl LocalEnv {
         self.base_data_dir.join("endpoints")
     }
 
-    pub fn pageserver_data_dir(&self, pageserver_id: NodeId) -> PathBuf {
-        self.base_data_dir
-            .join(format!("pageserver_{pageserver_id}"))
+    // TODO: move pageserver files into ./pageserver
+    pub fn pageserver_data_dir(&self) -> PathBuf {
+        self.base_data_dir.clone()
     }
 
     pub fn safekeeper_data_dir(&self, data_dir_name: &str) -> PathBuf {
         self.base_data_dir.join("safekeepers").join(data_dir_name)
     }
 
-    pub fn get_pageserver_conf(&self, id: NodeId) -> anyhow::Result<&PageServerConf> {
-        if let Some(conf) = self.pageservers.iter().find(|node| node.id == id) {
-            Ok(conf)
-        } else {
-            bail!("could not find pageserver {id}")
-        }
-    }
-
     pub fn register_branch_mapping(
         &mut self,
         branch_name: String,
@@ -309,10 +307,6 @@ impl LocalEnv {
             env.neon_distrib_dir = env::current_exe()?.parent().unwrap().to_owned();
         }
 
-        if env.pageservers.is_empty() {
-            anyhow::bail!("Configuration must contain at least one pageserver");
-        }
-
         env.base_data_dir = base_path();
 
         Ok(env)
@@ -345,7 +339,7 @@ impl LocalEnv {
         // We read that in, in `create_config`, and fill any missing defaults. Then it's saved
         // to .neon/config. TODO: We lose any formatting and comments along the way, which is
         // a bit sad.
-        let mut conf_content = r#"# This file describes a local deployment of the page server
+        let mut conf_content = r#"# This file describes a locale deployment of the page server
 # and safekeeeper node. It is read by the 'neon_local' command-line
 # utility.
 "#
@@ -475,9 +469,9 @@ impl LocalEnv {
     }
 
     fn auth_keys_needed(&self) -> bool {
-        self.pageservers.iter().any(|ps| {
-            ps.pg_auth_type == AuthType::NeonJWT || ps.http_auth_type == AuthType::NeonJWT
-        }) || self.safekeepers.iter().any(|sk| sk.auth_enabled)
+        self.pageserver.pg_auth_type == AuthType::NeonJWT
+            || self.pageserver.http_auth_type == AuthType::NeonJWT
+            || self.safekeepers.iter().any(|sk| sk.auth_enabled)
     }
 }
 

control_plane/src/pageserver.rs

@@ -14,7 +14,6 @@ use std::process::{Child, Command};
 use std::{io, result};
 
 use anyhow::{bail, Context};
-use camino::Utf8PathBuf;
 use pageserver_api::models::{self, TenantInfo, TimelineInfo};
 use postgres_backend::AuthType;
 use postgres_connection::{parse_host_port, PgConnectionConfig};
@@ -28,7 +27,6 @@ use utils::{
     lsn::Lsn,
 };
 
-use crate::local_env::PageServerConf;
 use crate::{background_process, local_env::LocalEnv};
 
 #[derive(Error, Debug)]
@@ -78,40 +76,43 @@ impl ResponseErrorMessageExt for Response {
 #[derive(Debug)]
 pub struct PageServerNode {
     pub pg_connection_config: PgConnectionConfig,
-    pub conf: PageServerConf,
     pub env: LocalEnv,
     pub http_client: Client,
     pub http_base_url: String,
 }
 
 impl PageServerNode {
-    pub fn from_env(env: &LocalEnv, conf: &PageServerConf) -> PageServerNode {
-        let (host, port) =
-            parse_host_port(&conf.listen_pg_addr).expect("Unable to parse listen_pg_addr");
+    pub fn from_env(env: &LocalEnv) -> PageServerNode {
+        let (host, port) = parse_host_port(&env.pageserver.listen_pg_addr)
+            .expect("Unable to parse listen_pg_addr");
         let port = port.unwrap_or(5432);
         Self {
             pg_connection_config: PgConnectionConfig::new_host_port(host, port),
-            conf: conf.clone(),
             env: env.clone(),
             http_client: Client::new(),
-            http_base_url: format!("http://{}/v1", conf.listen_http_addr),
+            http_base_url: format!("http://{}/v1", env.pageserver.listen_http_addr),
         }
     }
 
     // pageserver conf overrides defined by neon_local configuration.
     fn neon_local_overrides(&self) -> Vec<String> {
-        let id = format!("id={}", self.conf.id);
+        let id = format!("id={}", self.env.pageserver.id);
         // FIXME: the paths should be shell-escaped to handle paths with spaces, quotas etc.
         let pg_distrib_dir_param = format!(
             "pg_distrib_dir='{}'",
             self.env.pg_distrib_dir_raw().display()
         );
 
-        let http_auth_type_param = format!("http_auth_type='{}'", self.conf.http_auth_type);
-        let listen_http_addr_param = format!("listen_http_addr='{}'", self.conf.listen_http_addr);
+        let http_auth_type_param =
+            format!("http_auth_type='{}'", self.env.pageserver.http_auth_type);
+        let listen_http_addr_param = format!(
+            "listen_http_addr='{}'",
+            self.env.pageserver.listen_http_addr
+        );
 
-        let pg_auth_type_param = format!("pg_auth_type='{}'", self.conf.pg_auth_type);
-        let listen_pg_addr_param = format!("listen_pg_addr='{}'", self.conf.listen_pg_addr);
+        let pg_auth_type_param = format!("pg_auth_type='{}'", self.env.pageserver.pg_auth_type);
+        let listen_pg_addr_param =
+            format!("listen_pg_addr='{}'", self.env.pageserver.listen_pg_addr);
 
         let broker_endpoint_param = format!("broker_endpoint='{}'", self.env.broker.client_url());
 
@@ -125,18 +126,17 @@ impl PageServerNode {
             broker_endpoint_param,
         ];
 
-        if let Some(control_plane_api) = &self.env.control_plane_api {
+        if let Some(control_plane_api) = &self.env.pageserver.control_plane_api {
             overrides.push(format!(
                 "control_plane_api='{}'",
                 control_plane_api.as_str()
             ));
         }
 
-        if self.conf.http_auth_type != AuthType::Trust || self.conf.pg_auth_type != AuthType::Trust
+        if self.env.pageserver.http_auth_type != AuthType::Trust
+            || self.env.pageserver.pg_auth_type != AuthType::Trust
         {
-            // Keys are generated in the toplevel repo dir, pageservers' workdirs
-            // are one level below that, so refer to keys with ../
-            overrides.push("auth_validation_public_key_path='../auth_public_key.pem'".to_owned());
+            overrides.push("auth_validation_public_key_path='auth_public_key.pem'".to_owned());
         }
         overrides
     }
@@ -144,20 +144,23 @@ impl PageServerNode {
     /// Initializes a pageserver node by creating its config with the overrides provided.
     pub fn initialize(&self, config_overrides: &[&str]) -> anyhow::Result<()> {
         // First, run `pageserver --init` and wait for it to write a config into FS and exit.
-        self.pageserver_init(config_overrides)
-            .with_context(|| format!("Failed to run init for pageserver node {}", self.conf.id))
+        self.pageserver_init(config_overrides).with_context(|| {
+            format!(
+                "Failed to run init for pageserver node {}",
+                self.env.pageserver.id,
+            )
+        })
     }
 
     pub fn repo_path(&self) -> PathBuf {
-        self.env.pageserver_data_dir(self.conf.id)
+        self.env.pageserver_data_dir()
     }
 
     /// The pid file is created by the pageserver process, with its pid stored inside.
     /// Other pageservers cannot lock the same file and overwrite it for as long as the current
     /// pageserver runs. (Unless someone removes the file manually; never do that!)
-    fn pid_file(&self) -> Utf8PathBuf {
-        Utf8PathBuf::from_path_buf(self.repo_path().join("pageserver.pid"))
-            .expect("non-Unicode path")
+    fn pid_file(&self) -> PathBuf {
+        self.repo_path().join("pageserver.pid")
     }
 
     pub fn start(&self, config_overrides: &[&str]) -> anyhow::Result<Child> {
@@ -166,7 +169,7 @@ impl PageServerNode {
 
     fn pageserver_init(&self, config_overrides: &[&str]) -> anyhow::Result<()> {
         let datadir = self.repo_path();
-        let node_id = self.conf.id;
+        let node_id = self.env.pageserver.id;
         println!(
             "Initializing pageserver node {} at '{}' in {:?}",
             node_id,
@@ -175,10 +178,6 @@ impl PageServerNode {
         );
         io::stdout().flush()?;
 
-        if !datadir.exists() {
-            std::fs::create_dir(&datadir)?;
-        }
-
         let datadir_path_str = datadir.to_str().with_context(|| {
             format!("Cannot start pageserver node {node_id} in path that has no string representation: {datadir:?}")
         })?;
@@ -209,7 +208,7 @@ impl PageServerNode {
         let datadir = self.repo_path();
         print!(
             "Starting pageserver node {} at '{}' in {:?}",
-            self.conf.id,
+            self.env.pageserver.id,
             self.pg_connection_config.raw_address(),
             datadir
         );
@@ -218,7 +217,7 @@ impl PageServerNode {
         let datadir_path_str = datadir.to_str().with_context(|| {
             format!(
                 "Cannot start pageserver node {} in path that has no string representation: {:?}",
-                self.conf.id, datadir,
+                self.env.pageserver.id, datadir,
             )
         })?;
         let mut args = self.pageserver_basic_args(config_overrides, datadir_path_str);
@@ -262,7 +261,7 @@ impl PageServerNode {
         // FIXME: why is this tied to pageserver's auth type? Whether or not the safekeeper
         // needs a token, and how to generate that token, seems independent to whether
         // the pageserver requires a token in incoming requests.
-        Ok(if self.conf.http_auth_type != AuthType::Trust {
+        Ok(if self.env.pageserver.http_auth_type != AuthType::Trust {
             // Generate a token to connect from the pageserver to a safekeeper
             let token = self
                 .env
@@ -287,7 +286,7 @@ impl PageServerNode {
 
     pub fn page_server_psql_client(&self) -> anyhow::Result<postgres::Client> {
         let mut config = self.pg_connection_config.clone();
-        if self.conf.pg_auth_type == AuthType::NeonJWT {
+        if self.env.pageserver.pg_auth_type == AuthType::NeonJWT {
             let token = self
                 .env
                 .generate_auth_token(&Claims::new(None, Scope::PageServerApi))?;
@@ -298,7 +297,7 @@ impl PageServerNode {
 
     fn http_request<U: IntoUrl>(&self, method: Method, url: U) -> anyhow::Result<RequestBuilder> {
         let mut builder = self.http_client.request(method, url);
-        if self.conf.http_auth_type == AuthType::NeonJWT {
+        if self.env.pageserver.http_auth_type == AuthType::NeonJWT {
             let token = self
                 .env
                 .generate_auth_token(&Claims::new(None, Scope::PageServerApi))?;

control_plane/src/safekeeper.rs

@@ -11,7 +11,6 @@ use std::process::Child;
 use std::{io, result};
 
 use anyhow::Context;
-use camino::Utf8PathBuf;
 use postgres_connection::PgConnectionConfig;
 use reqwest::blocking::{Client, RequestBuilder, Response};
 use reqwest::{IntoUrl, Method};
@@ -98,9 +97,8 @@ impl SafekeeperNode {
         SafekeeperNode::datadir_path_by_id(&self.env, self.id)
     }
 
-    pub fn pid_file(&self) -> Utf8PathBuf {
-        Utf8PathBuf::from_path_buf(self.datadir_path().join("safekeeper.pid"))
-            .expect("non-Unicode path")
+    pub fn pid_file(&self) -> PathBuf {
+        self.datadir_path().join("safekeeper.pid")
     }
 
     pub fn start(&self, extra_opts: Vec<String>) -> anyhow::Result<Child> {

deny.toml

@@ -23,7 +23,7 @@ vulnerability = "deny"
 unmaintained = "warn"
 yanked = "warn"
 notice = "warn"
-ignore = []
+ignore = ["RUSTSEC-2023-0052"]
 
 # This section is considered when running `cargo deny check licenses`
 # More documentation for the licenses section can be found here:

docker-compose/docker_compose_test.sh

@@ -30,7 +30,7 @@ cleanup() {
 echo "clean up containers if exists"
 cleanup
 
-for pg_version in 14 15 16; do
+for pg_version in 14 15; do
     echo "start containers (pg_version=$pg_version)."
     PG_VERSION=$pg_version docker compose -f $COMPOSE_FILE up --build -d
 

docs/rfcs/028-pageserver-migration.md

@@ -1,599 +0,0 @@
-# Seamless tenant migration
-
-- Author: john@neon.tech
-- Created on 2023-08-11
-- Implemented on ..
-
-## Summary
-
-The preceding [generation numbers RFC](025-generation-numbers.md) may be thought of as "making tenant
-migration safe". Following that,
-this RFC is about how those migrations are to be done:
-
-1. Seamlessly (without interruption to client availability)
-2. Quickly (enabling faster operations)
-3. Efficiently (minimizing I/O and $ cost)
-
-These points are in priority order: if we have to sacrifice
-efficiency to make a migration seamless for clients, we will
-do so, etc.
-
-This is accomplished by introducing two high level changes:
-
-- A dual-attached state for tenants, used in a control-plane-orchestrated
-  migration procedure that preserves availability during a migration.
-- Warm secondary locations for tenants, where on-disk content is primed
-  for a fast migration of the tenant from its current attachment to this
-  secondary location.
-
-## Motivation
-
-Migrating tenants between pageservers is essential to operating a service
-at scale, in several contexts:
-
-1. Responding to a pageserver node failure by migrating tenants to other pageservers
-2. Balancing load and capacity across pageservers, for example when a user expands their
-   database and they need to migrate to a pageserver with more capacity.
-3. Restarting pageservers for upgrades and maintenance
-
-The current situation steps for migration are:
-
-- detach from old node; skip if old node is dead; (the [skip part is still WIP](https://github.com/neondatabase/cloud/issues/5426)).
-- attach to new node
-- re-configure endpoints to use the new node
-
-Once [generation numbers](025-generation-numbers.md) are implemented,
-the detach step is no longer critical for correctness. So, we can
-
-- attach to a new node,
-- re-configure endpoints to use the new node, and then
-- detach from the old node.
-
-However, this still does not meet our seamless/fast/efficient goals:
-
-- Not fast: The new node will have to download potentially large amounts
-  of data from S3, which may take many minutes.
-- Not seamless: If we attach to a new pageserver before detaching an old one,
-  the new one might delete some objects that interrupt availability of reads on the old one.
-- Not efficient: the old pageserver will continue uploading
-  S3 content during the migration that will never be read.
-
-The user expectations for availability are:
-
-- For planned maintenance, there should be zero availability
-  gap. This expectation is fulfilled by this RFC.
-- For unplanned changes (e.g. node failures), there should be
-  minimal availability gap. This RFC provides the _mechanism_
-  to fail over quickly, but does not provide the failure _detection_
-  nor failover _policy_.
-
-## Non Goals
-
-- Defining service tiers with different storage strategies: the same
-  level of HA & overhead will apply to all tenants. This doesn't rule out
-  adding such tiers in future.
-- Enabling pageserver failover in the absence of a control plane: the control
-  plane will remain the source of truth for what should be attached where.
-- Totally avoiding availability gaps on unplanned migrations during
-  a failure (we expect a small, bounded window of
-  read unavailability of very recent LSNs)
-- Workload balancing: this RFC defines the mechanism for moving tenants
-  around, not the higher level logic for deciding who goes where.
-- Defining all possible configuration flows for tenants: the migration process
-  defined in this RFC demonstrates the sufficiency of the pageserver API, but
-  is not the only kind of configuration change the control plane will ever do.
-  The APIs defined here should let the control plane move tenants around in
-  whatever way is needed while preserving data safety and read availability.
-
-## Impacted components
-
-Pageserver, control plane
-
-## Terminology
-
-- **Attachment**: a tenant is _attached_ to a pageserver if it has
-  been issued a generation number, and is running an instance of
-  the `Tenant` type, ingesting the WAL, and available to serve
-  page reads.
-- **Location**: locations are a superset of attachments. A location
-  is a combination of a tenant and a pageserver. We may _attach_ at a _location_.
-
-- **Secondary location**: a location which is not currently attached.
-- **Warm secondary location**: a location which is not currently attached, but is endeavoring to maintain a warm local cache of layers. We avoid calling this a _warm standby_ to avoid confusion with similar postgres features.
-
-## Implementation (high level)
-
-### Warm secondary locations
-
-To enable faster migrations, we will identify at least one _secondary location_
-for each tenant. This secondary location will keep a warm cache of layers
-for the tenant, so that if it is later attached, it can catch up with the
-latest LSN quickly: rather than downloading everything, it only has to replay
-the recent part of the WAL to advance from the remote_consistent_offset to the
-most recent LSN in the WAL.
-
-The control plane is responsible for selecting secondary locations, and
-calling into pageservers to configure tenants into a secondary mode at this
-new location, as well as attaching the tenant in its existing primary location.
-
-The attached pageserver for a tenant will publish a [layer heatmap](#layer-heatmap)
-to advise secondaries of which layers should be downloaded.
-
-### Location modes
-
-Currently, we consider a tenant to be in one of two states on a pageserver:
-
-- Attached: active `Tenant` object, and layers on local disk
-- Detached: no layers on local disk, no runtime state.
-
-We will extend this with finer-grained modes, whose purpose will become
-clear in later sections:
-
-- **AttachedSingle**: equivalent the existing attached state.
-- **AttachedMulti**: like AttachedSingle, holds an up to date generation, but
-  does not do deletions.
-- **AttachedStale**: like AttachedSingle, holds a stale generation,
-  do not do any remote storage operations.
-- **Secondary**: keep local state on disk, periodically update from S3.
-- **Detached**: equivalent to existing detached state.
-
-To control these finer grained states, a new pageserver API endpoint will be added.
-
-### Cutover procedure
-
-Define old location and new location as "Node A" and "Node B". Consider
-the case where both nodes are available, and Node B was previously configured
-as a secondary location for the tenant we are migrating.
-
-The cutover procedure is orchestrated by the control plane, calling into
-the pageservers' APIs:
-
-1. Call to Node A requesting it to flush to S3 and enter AttachedStale state
-2. Increment generation, and call to Node B requesting it to enter AttachedMulti
-   state with the new generation.
-3. Call to Node B, requesting it to download the latest hot layers from remote storage,
-   according to the latest heatmap flushed by Node A.
-4. Wait for Node B's WAL ingestion to catch up with node A's
-5. Update endpoints to use node B instead of node A
-6. Call to node B requesting it to enter state AttachedSingle.
-7. Call to node A requesting it to enter state Secondary
-
-The following table summarizes how the state of the system advances:
-
-|     Step      |     Node A     |     Node B     | Node used by endpoints |
-| :-----------: | :------------: | :------------: | :--------------------: |
-| 1 (_initial_) | AttachedSingle |   Secondary    |           A            |
-|       2       | AttachedStale  | AttachedMulti  |           A            |
-|       3       | AttachedStale  | AttachedMulti  |           A            |
-|       4       | AttachedStale  | AttachedMulti  |           A            |
-| 5 (_cutover_) | AttachedStale  | AttachedMulti  |           B            |
-|       6       | AttachedStale  | AttachedSingle |           B            |
-|  7 (_final_)  |   Secondary    | AttachedSingle |           B            |
-
-The procedure described for a clean handover from a live node to a secondary
-is also used for failure cases and for migrations to a location that is not
-configured as a secondary, by simply skipping irrelevant steps, as described in
-the following sections.
-
-#### Migration from an unresponsive node
-
-If node A is unavailable, then all calls into
-node A are skipped and we don't wait for B to catch up before
-switching updating the endpoints to use B.
-
-#### Migration to a location that is not a secondary
-
-If node B is initially in Detached state, the procedure is identical. Since Node B
-is coming from a Detached state rather than Secondary, the download of layers and
-catch up with WAL will take much longer.
-
-We might do this if:
-
-- Attached and secondary locations are both critically low on disk, and we need
-  to migrate to a third node with more resources available.
-- We are migrating a tenant which does not use secondary locations to save on cost.
-
-#### Permanent migration away from a node
-
-In the final step of the migration, we generally request the original node to enter a Secondary
-state. This is typical if we are doing a planned migration during maintenance, or to
-balance CPU/network load away from a node.
-
-One might also want to permanently migrate away: this can be done by simply removing the secondary
-location after the migration is complete, or as an optimization by substituting the Detached state
-for the Secondary state in the final step.
-
-#### Cutover diagram
-
-```mermaid
-sequenceDiagram
-participant CP as Control plane
-participant A as Node A
-participant B as Node B
-participant E as Endpoint
-
-CP->>A: PUT Flush & go to AttachedStale
-note right of A: A continues to ingest WAL
-CP->>B: PUT AttachedMulti
-CP->>B: PUT Download layers from latest heatmap
-note right of B: B downloads from S3
-loop Poll until download complete
-CP->>B: GET download status
-end
-activate B
-note right of B: B ingests WAL
-loop Poll until catch up
-CP->>B: GET visible WAL
-CP->>A: GET visible WAL
-end
-deactivate B
-CP->>E: Configure to use Node B
-E->>B: Connect for reads
-CP->>B: PUT AttachedSingle
-CP->>A: PUT Secondary
-```
-
-#### Cutover from an unavailable pageserver
-
-This case is far simpler: we may skip straight to our intended
-end state.
-
-```mermaid
-sequenceDiagram
-participant A as Node A
-participant CP as Control plane
-participant B as Node B
-participant E as Endpoint
-
-note right of A: Node A offline
-activate A
-CP->>B: PUT AttachedSingle
-CP->>E: Configure to use Node B
-E->>B: Connect for reads
-deactivate A
-```
-
-## Implementation (detail)
-
-### Purpose of AttachedMulti, AttachedStale
-
-#### AttachedMulti
-
-Ordinarily, an attached pageserver whose generation is the latest may delete
-layers at will (e.g. during compaction). If a previous generation pageserver
-is also still attached, and in use by endpoints, then this layer deletion could
-lead to a loss of availability for the endpoint when reading from the previous
-generation pageserver.
-
-The _AttachedMulti_ state simply disables deletions. These will be enqueued
-in `RemoteTimelineClient` until the control plane transitions the
-node into AttachedSingle, which unblocks deletions.  Other remote storage operations
-such as uploads are not blocked.
-
-AttachedMulti is not required for data safety, only to preserve availability
-on pageservers running with stale generations.
-
-A node enters AttachedMulti only when explicitly asked to by the control plane. It should
-only remain in this state for the duration of a migration.
-
-If a control plane bug leaves
-the node in AttachedMulti for a long time, then we must avoid unbounded memory use from enqueued
-deletions. This may be accomplished simply, by dropping enqueued deletions when some modest
-threshold of delayed deletions (e.g. 10k layers per tenant) is reached. As with all deletions,
-it is safe to skip them, and the leaked objects will be eventually cleaned up by scrub or
-by timeline deletion.
-
-During AttachedMulti, the Tenant is free to drop layers from local disk in response to
-disk pressure: only the deletion of remote layers is blocked.
-
-#### AttachedStale
-
-Currently, a pageserver with a stale generation number will continue to
-upload layers, but be prevented from completing deletions. This is safe, but inefficient: layers uploaded by this stale generation
-will not be read back by future generations of pageservers.
-
-The _AttachedStale_ state disables S3 uploads. The stale pageserver
-will continue to ingest the WAL and write layers to local disk, but not to
-do any uploads to S3.
-
-A node may enter AttachedStale in two ways:
-
-- Explicitly, when control plane calls into the node at the start of a migration.
-- Implicitly, when the node tries to validate some deletions and discovers
-  that its generation is stale.
-
-The AttachedStale state also disables sending consumption metrics from
-that location: it is interpreted as an indication that some other pageserver
-is already attached or is about to be attached, and that new pageserver will
-be responsible for sending consumption metrics.
-
-#### Disk Pressure & AttachedStale
-
-Over long periods of time, a tenant location in AttachedStale will accumulate data
-on local disk, as it cannot evict any layers written since it entered the
-AttachStale state. We rely on the control plane to revert the location to
-Secondary or Detached at the end of a migration.
-
-This scenario is particularly noteworthy when evacuating all tenants on a pageserver:
-since _all_ the attached tenants will go into AttachedStale, we will be doing no
-uploads at all, therefore ingested data will cause disk usage to increase continuously.
-Under nominal conditions, the available disk space on pageservers should be sufficient
-to complete the evacuation before this becomes a problem, but we must also handle
-the case where we hit a low disk situation while in this state.
-
-The concept of disk pressure already exists in the pageserver: the `disk_usage_eviction_task`
-touches each Tenant when it determines that a low-disk condition requires
-some layer eviction. Having selected layers for eviction, the eviction
-task calls `Timeline::evict_layers`.
-
-**Safety**: If evict_layers is called while in AttachedStale state, and some of the to-be-evicted
-layers are not yet uploaded to S3, then the block on uploads will be lifted. This
-will result in leaking some objects once a migration is complete, but will enable
-the node to manage its disk space properly: if a node is left with some tenants
-in AttachedStale indefinitely due to a network partition or control plane bug,
-these tenants will not cause a full disk condition.
-
-### Warm secondary updates
-
-#### Layer heatmap
-
-The secondary location's job is to serve reads **with the same quality of service as the original location
-was serving them around the time of a migration**. This does not mean the secondary
-location needs the whole set of layers: inactive layers that might soon
-be evicted on the attached pageserver need not be downloaded by the
-secondary. A totally idle tenant only needs to maintain enough on-disk
-state to enable a fast cold start (i.e. the most recent image layers are
-typically sufficient).
-
-To enable this, we introduce the concept of a _layer heatmap_, which
-acts as an advisory input to secondary locations to decide which
-layers to download from S3.
-
-#### Attached pageserver
-
-The attached pageserver, if in state AttachedSingle, periodically
-uploads a serialized heat map to S3. It may skip this if there
-is no change since the last time it uploaded (e.g. if the tenant
-is totally idle).
-
-Additionally, when the tenant is flushed to remote storage prior to a migration
-(the first step in [cutover procedure](#cutover-procedure)), 
-the heatmap is written out. This enables a future attached pageserver
-to get an up to date view when deciding which layers to download.
-
-#### Secondary location behavior
-
-Secondary warm locations run a simple loop, implemented separately from
-the main `Tenant` type, which represents attached tenants:
-
-- Download the layer heatmap
-- Select any "hot enough" layers to download, if there is sufficient
-  free disk space.
-- Download layers, if they were not previously evicted (see below)
-- Download the latest index_part.json
-- Check if any layers currently on disk are no longer referenced by
-  IndexPart & delete them
-
-Note that the heatmap is only advisory: if a secondary location has plenty
-of disk space, it may choose to retain layers that aren't referenced
-by the heatmap, as long as they are still referenced by the IndexPart. Conversely,
-if a node is very low on disk space, it might opt to raise the heat threshold required
-to both downloading a layer, until more disk space is available.
-
-#### Secondary locations & disk pressure
-
-Secondary locations are subject to eviction on disk pressure, just as
-attached locations are.  For eviction purposes, the access time of a
-layer in a secondary location will be the access time given in the heatmap,
-rather than the literal time at which the local layer file was accessed.
-
-The heatmap will indicate which layers are in local storage on the attached
-location.  The secondary will always attempt to get back to having that
-set of layers on disk, but to avoid flapping, it will remember the access
-time of the layer it was most recently asked to evict, and layers whose
-access time is below that will not be re-downloaded.
-
-The resulting behavior is that after a layer is evicted from a secondary
-location, it is only re-downloaded once the attached pageserver accesses
-the layer and uploads a heatmap reflecting that access time.  On a pageserver
-restart, the secondary location will attempt to download all layers in
-the heatmap again, if they are not on local disk.
-
-This behavior will be slightly different when secondary locations are
-used for "low energy tenants", but that is beyond the scope of this RFC.
-
-### Location configuration API
-
-Currently, the `/tenant/<tenant_id>/config` API defines various
-tunables like compaction settings, which apply to the tenant irrespective
-of which pageserver it is running on.
-
-A new "location config" structure will be introduced, which defines
-configuration which is per-tenant, but local to a particular pageserver,
-such as the attachment mode and whether it is a secondary.
-
-The pageserver will expose a new per-tenant API for setting
-the state: `/tenant/<tenant_id>/location/config`.
-
-Body content:
-
-```
-{
-  state: 'enum{Detached, Secondary, AttachedSingle, AttachedMulti, AttachedStale}',
-  generation: Option<u32>,
-  configuration: `Option<TenantConfig>`
-  flush: bool
-}
-```
-
-Existing `/attach` and `/detach` endpoint will have the same
-behavior as calling `/location/config` with `AttachedSingle` and `Detached`
-states respectively. These endpoints will be deprecated and later
-removed.
-
-The generation attribute is mandatory for entering `AttachedSingle` or
-`AttachedMulti`.
-
-The configuration attribute is mandatory when entering any state other
-than `Detached`. This configuration is the same as the body for
-the existing `/tenant/<tenant_id>/config` endpoint.
-
-The `flush` argument indicates whether the pageservers should flush
-to S3 before proceeding: this only has any effect if the node is
-currently in AttachedSingle or AttachedMulti. This is used
-during the first phase of migration, when transitioning the
-old pageserver to AttachedSingle.
-
-The `/re-attach` API response will be extended to include a `state` as
-well as a `generation`, enabling the pageserver to enter the
-correct state for each tenant on startup.
-
-### Database schema for locations
-
-A new table `ProjectLocation`:
-
-- pageserver_id: int
-- tenant_id: TenantId
-- generation: Option<int>
-- state: `enum(Secondary, AttachedSingle, AttachedMulti)`
-
-Notes:
-
-- It is legacy for a Project to have zero `ProjectLocation`s
-- The `pageserver` column in `Project` now means "to which pageserver should
-  endpoints connect", rather than simply which pageserver is attached.
-- The `generation` column in `Project` remains, and is incremented and used
-  to set the generation of `ProjectLocation` rows when they are set into
-  an attached state.
-- The `Detached` state is implicitly represented as the absence of
-  a `ProjectLocation`.
-
-### Executing migrations
-
-Migrations will be implemented as Go functions, within the
-existing `Operation` framework in the control plane. These
-operations are persistent, such that they will always keep
-trying until completion: this property is important to avoid
-leaving garbage behind on pageservers, such as AttachedStale
-locations.
-
-### Recovery from failures during migration
-
-During migration, the control plane may encounter failures of either
-the original or new pageserver, or both:
-
-- If the original fails, skip past waiting for the new pageserver
-  to catch up, and put it into AttachedSingle immediately.
-- If the new node fails, put the old pageserver into Secondary
-  and then back into AttachedSingle (this has the effect of
-  retaining on-disk state and granting it a fresh generation number).
-- If both nodes fail, keep trying until one of them is available
-  again.
-
-### Control plane -> Pageserver reconciliation
-
-A migration may be done while the old node is unavailable,
-in which case the old node may still be running in an AttachedStale
-state.
-
-In this case, it is undesirable to have the migration `Operation`
-stay alive until the old node eventually comes back online
-and can be cleaned up. To handle this, the control plane
-should run a background reconciliation process to compare
-a pageserver's attachments with the database, and clean up
-any that shouldn't be there any more.
-
-Note that there will be no work to do if the old node was really
-offline, as during startup it will call into `/re-attach` and
-be updated that way. The reconciliation will only be needed
-if the node was unavailable but still running.
-
-## Alternatives considered
-
-### Only enabling secondary locations for tenants on a higher service tier
-
-This will make sense in future, especially for tiny databases that may be
-downloaded from S3 in milliseconds when needed.
-
-However, it is not wise to do it immediately, because pageservers contain
-a mixture of higher and lower tier workloads. If we had 1 tenant with
-a secondary location and 9 without, then those other 9 tenants will do
-a lot of I/O as they try to recover from S3, which may degrade the
-service of the tenant which had a secondary location.
-
-Until we segregate tenant on different service tiers on different pageserver
-nodes, or implement & test QoS to ensure that tenants with secondaries are
-not harmed by tenants without, we should use the same failover approach
-for all the tenants.
-
-### Hot secondary locations (continuous WAL replay)
-
-Instead of secondary locations populating their caches from S3, we could
-have them consume the WAL from safekeepers. The downsides of this would be:
-
-- Double load on safekeepers, which are a less scalable service than S3
-- Secondary locations' on-disk state would end up subtly different to
-  the remote state, which would make synchronizing with S3 more complex/expensive
-  when going into attached state.
-
-The downside of only updating secondary locations from S3 is that we will
-have a delay during migration from replaying the LSN range between what's
-in S3 and what's in the pageserver. This range will be very small on
-planned migrations, as we have the old pageserver flush to S3 immediately
-before attaching the new pageserver. On unplanned migrations (old pageserver
-is unavailable), the range of LSNs to replay is bounded by the flush frequency
-on the old pageserver. However, the migration doesn't have to wait for the
-replay: it's just that not-yet-replayed LSNs will be unavailable for read
-until the new pageserver catches up.
-
-We expect that pageserver reads of the most recent LSNs will be relatively
-rare, as for an active endpoint those pages will usually still be in the postgres
-page cache: this leads us to prefer synchronizing from S3 on secondary
-locations, rather than consuming the WAL from safekeepers.
-
-### Cold secondary locations
-
-It is not functionally necessary to keep warm caches on secondary locations at all. However, if we do not, then
-we would experience a de-facto availability loss in unplanned migrations, as reads to the new node would take an extremely long time (many seconds, perhaps minutes).
-
-Warm caches on secondary locations are necessary to meet
-our availability goals.
-
-### Pageserver-granularity failover
-
-Instead of migrating tenants individually, we could have entire spare nodes,
-and on a node death, move all its work to one of these spares.
-
-This approach is avoided for several reasons:
-
-- we would still need fine-grained tenant migration for other
-  purposes such as balancing load
-- by sharing the spare capacity over many peers rather than one spare node,
-  these peers may use the capacity for other purposes, until it is needed
-  to handle migrated tenants. e.g. for keeping a deeper cache of their
-  attached tenants.
-
-### Readonly during migration
-
-We could simplify migrations by making both previous and new nodes go into a
-readonly state, then flush remote content from the previous node, then activate
-attachment on the secondary node.
-
-The downside to this approach is a potentially large gap in readability of
-recent LSNs while loading data onto the new node. To avoid this, it is worthwhile
-to incur the extra cost of double-replaying the WAL onto old and new nodes' local
-storage during a migration.
-
-### Peer-to-peer pageserver communication
-
-Rather than uploading the heatmap to S3, attached pageservers could make it
-available to peers.
-
-Currently, pageservers have no peer to peer communication, so adding this
-for heatmaps would incur significant overhead in deployment and configuration
-of the service, and ensuring that when a new pageserver is deployed, other
-pageservers are updated to be aware of it.
-
-As well as simplifying implementation, putting heatmaps in S3 will be useful
-for future analytics purposes -- gathering aggregated statistics on activity
-pattersn across many tenants may be done directly from data in S3.

libs/consumption_metrics/src/lib.rs

@@ -3,9 +3,9 @@
 //!
 use chrono::{DateTime, Utc};
 use rand::Rng;
-use serde::{Deserialize, Serialize};
+use serde::Serialize;
 
-#[derive(Serialize, serde::Deserialize, Debug, Clone, Copy, Eq, PartialEq, Ord, PartialOrd)]
+#[derive(Serialize, Debug, Clone, Copy, Eq, PartialEq, Ord, PartialOrd)]
 #[serde(tag = "type")]
 pub enum EventType {
     #[serde(rename = "absolute")]
@@ -27,8 +27,7 @@ impl EventType {
     }
 
     pub fn incremental_timerange(&self) -> Option<std::ops::Range<&DateTime<Utc>>> {
-        // these can most likely be thought of as Range or RangeFull, at least pageserver creates
-        // incremental ranges where the stop and next start are equal.
+        // these can most likely be thought of as Range or RangeFull
         use EventType::*;
         match self {
             Incremental {
@@ -42,25 +41,15 @@ impl EventType {
     pub fn is_incremental(&self) -> bool {
         matches!(self, EventType::Incremental { .. })
     }
-
-    /// Returns the absolute time, or for incremental ranges, the stop time.
-    pub fn recorded_at(&self) -> &DateTime<Utc> {
-        use EventType::*;
-
-        match self {
-            Absolute { time } => time,
-            Incremental { stop_time, .. } => stop_time,
-        }
-    }
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, Eq, PartialEq, Ord, PartialOrd)]
-pub struct Event<Extra, Metric> {
+#[derive(Serialize, Debug, Clone, Eq, PartialEq, Ord, PartialOrd)]
+pub struct Event<Extra> {
     #[serde(flatten)]
     #[serde(rename = "type")]
     pub kind: EventType,
 
-    pub metric: Metric,
+    pub metric: &'static str,
     pub idempotency_key: String,
     pub value: u64,
 
@@ -69,45 +58,19 @@ pub struct Event<Extra, Metric> {
 }
 
 pub fn idempotency_key(node_id: &str) -> String {
-    IdempotencyKey::generate(node_id).to_string()
-}
-
-/// Downstream users will use these to detect upload retries.
-pub struct IdempotencyKey<'a> {
-    now: chrono::DateTime<Utc>,
-    node_id: &'a str,
-    nonce: u16,
-}
-
-impl std::fmt::Display for IdempotencyKey<'_> {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        write!(f, "{}-{}-{:04}", self.now, self.node_id, self.nonce)
-    }
-}
-
-impl<'a> IdempotencyKey<'a> {
-    pub fn generate(node_id: &'a str) -> Self {
-        IdempotencyKey {
-            now: Utc::now(),
-            node_id,
-            nonce: rand::thread_rng().gen_range(0..=9999),
-        }
-    }
-
-    pub fn for_tests(now: DateTime<Utc>, node_id: &'a str, nonce: u16) -> Self {
-        IdempotencyKey {
-            now,
-            node_id,
-            nonce,
-        }
-    }
+    format!(
+        "{}-{}-{:04}",
+        Utc::now(),
+        node_id,
+        rand::thread_rng().gen_range(0..=9999)
+    )
 }
 
 pub const CHUNK_SIZE: usize = 1000;
 
 // Just a wrapper around a slice of events
 // to serialize it as `{"events" : [ ] }
-#[derive(serde::Serialize, serde::Deserialize)]
+#[derive(serde::Serialize)]
 pub struct EventChunk<'a, T: Clone> {
     pub events: std::borrow::Cow<'a, [T]>,
 }

libs/pageserver_api/src/models.rs

@@ -363,15 +363,8 @@ pub struct TimelineInfo {
     pub latest_gc_cutoff_lsn: Lsn,
     #[serde_as(as = "DisplayFromStr")]
     pub disk_consistent_lsn: Lsn,
-
-    /// The LSN that we have succesfully uploaded to remote storage
     #[serde_as(as = "DisplayFromStr")]
     pub remote_consistent_lsn: Lsn,
-
-    /// The LSN that we are advertizing to safekeepers
-    #[serde_as(as = "DisplayFromStr")]
-    pub remote_consistent_lsn_visible: Lsn,
-
     pub current_logical_size: Option<u64>, // is None when timeline is Unloaded
     /// Sum of the size of all layer files.
     /// If a layer is present in both local FS and S3, it counts only once.
@@ -388,8 +381,6 @@ pub struct TimelineInfo {
     pub pg_version: u32,
 
     pub state: TimelineState,
-
-    pub walreceiver_status: String,
 }
 
 #[derive(Debug, Clone, Serialize)]

libs/postgres_ffi/README.md

@@ -10,11 +10,9 @@ should be auto-generated too, but that's a TODO.
 The PostgreSQL on-disk file format is not portable across different
 CPU architectures and operating systems. It is also subject to change
 in each major PostgreSQL version. Currently, this module supports
-PostgreSQL v14, v15 and v16: bindings and code that depends on them are
-version-specific.
-This code is organized in modules `postgres_ffi::v14`, `postgres_ffi::v15` and
-`postgres_ffi::v16`. Version independent code is explicitly exported into
-shared `postgres_ffi`.
+PostgreSQL v14 and v15: bindings and code that depends on them are version-specific.
+This code is organized in modules: `postgres_ffi::v14` and `postgres_ffi::v15`
+Version independend code is explicitly exported into shared `postgres_ffi`.
 
 
 TODO: Currently, there is also some code that deals with WAL records

libs/postgres_ffi/build.rs

@@ -56,7 +56,7 @@ fn main() -> anyhow::Result<()> {
         PathBuf::from("pg_install")
     };
 
-    for pg_version in &["v14", "v15", "v16"] {
+    for pg_version in &["v14", "v15"] {
         let mut pg_install_dir_versioned = pg_install_dir.join(pg_version);
         if pg_install_dir_versioned.is_relative() {
             let cwd = env::current_dir().context("Failed to get current_dir")?;
@@ -125,7 +125,6 @@ fn main() -> anyhow::Result<()> {
             .allowlist_var("PG_CONTROLFILEDATA_OFFSETOF_CRC")
             .allowlist_type("PageHeaderData")
             .allowlist_type("DBState")
-            .allowlist_type("RelMapFile")
             // Because structs are used for serialization, tell bindgen to emit
             // explicit padding fields.
             .explicit_padding(true)

libs/postgres_ffi/src/lib.rs

@@ -51,59 +51,11 @@ macro_rules! for_all_postgres_versions {
     ($macro:tt) => {
         $macro!(v14);
         $macro!(v15);
-        $macro!(v16);
     };
 }
 
 for_all_postgres_versions! { postgres_ffi }
 
-/// dispatch_pgversion
-///
-/// Run a code block in a context where the postgres_ffi bindings for a
-/// specific (supported) PostgreSQL version are `use`-ed in scope under the pgv
-/// identifier.
-/// If the provided pg_version is not supported, we panic!(), unless the
-/// optional third argument was provided (in which case that code will provide
-/// the default handling instead).
-///
-/// Use like
-///
-/// dispatch_pgversion!(my_pgversion, { pgv::constants::XLOG_DBASE_CREATE })
-/// dispatch_pgversion!(my_pgversion, pgv::constants::XLOG_DBASE_CREATE)
-///
-/// Other uses are for macro-internal purposes only and strictly unsupported.
-///
-#[macro_export]
-macro_rules! dispatch_pgversion {
-    ($version:expr, $code:expr) => {
-        dispatch_pgversion!($version, $code, panic!("Unknown PostgreSQL version {}", $version))
-    };
-    ($version:expr, $code:expr, $invalid_pgver_handling:expr) => {
-        dispatch_pgversion!(
-            $version => $code,
-            default = $invalid_pgver_handling,
-            pgversions = [
-                14 : v14,
-                15 : v15,
-                16 : v16,
-            ]
-        )
-    };
-    ($pgversion:expr => $code:expr,
-     default = $default:expr,
-     pgversions = [$($sv:literal : $vsv:ident),+ $(,)?]) => {
-        match ($pgversion) {
-            $($sv => {
-                use $crate::$vsv as pgv;
-                $code
-            },)+
-            _ => {
-                $default
-            }
-        }
-    };
-}
-
 pub mod pg_constants;
 pub mod relfile_utils;
 
@@ -138,7 +90,13 @@ pub use v14::xlog_utils::XLogFileName;
 pub use v14::bindings::DBState_DB_SHUTDOWNED;
 
 pub fn bkpimage_is_compressed(bimg_info: u8, version: u32) -> anyhow::Result<bool> {
-    dispatch_pgversion!(version, Ok(pgv::bindings::bkpimg_is_compressed(bimg_info)))
+    match version {
+        14 => Ok(bimg_info & v14::bindings::BKPIMAGE_IS_COMPRESSED != 0),
+        15 => Ok(bimg_info & v15::bindings::BKPIMAGE_COMPRESS_PGLZ != 0
+            || bimg_info & v15::bindings::BKPIMAGE_COMPRESS_LZ4 != 0
+            || bimg_info & v15::bindings::BKPIMAGE_COMPRESS_ZSTD != 0),
+        _ => anyhow::bail!("Unknown version {}", version),
+    }
 }
 
 pub fn generate_wal_segment(
@@ -149,11 +107,11 @@ pub fn generate_wal_segment(
 ) -> Result<Bytes, SerializeError> {
     assert_eq!(segno, lsn.segment_number(WAL_SEGMENT_SIZE));
 
-    dispatch_pgversion!(
-        pg_version,
-        pgv::xlog_utils::generate_wal_segment(segno, system_id, lsn),
-        Err(SerializeError::BadInput)
-    )
+    match pg_version {
+        14 => v14::xlog_utils::generate_wal_segment(segno, system_id, lsn),
+        15 => v15::xlog_utils::generate_wal_segment(segno, system_id, lsn),
+        _ => Err(SerializeError::BadInput),
+    }
 }
 
 pub fn generate_pg_control(
@@ -162,11 +120,11 @@ pub fn generate_pg_control(
     lsn: Lsn,
     pg_version: u32,
 ) -> anyhow::Result<(Bytes, u64)> {
-    dispatch_pgversion!(
-        pg_version,
-        pgv::xlog_utils::generate_pg_control(pg_control_bytes, checkpoint_bytes, lsn),
-        anyhow::bail!("Unknown version {}", pg_version)
-    )
+    match pg_version {
+        14 => v14::xlog_utils::generate_pg_control(pg_control_bytes, checkpoint_bytes, lsn),
+        15 => v15::xlog_utils::generate_pg_control(pg_control_bytes, checkpoint_bytes, lsn),
+        _ => anyhow::bail!("Unknown version {}", pg_version),
+    }
 }
 
 // PG timeline is always 1, changing it doesn't have any useful meaning in Neon.
@@ -238,6 +196,8 @@ pub fn fsm_logical_to_physical(addr: BlockNumber) -> BlockNumber {
 }
 
 pub mod waldecoder {
+
+    use crate::{v14, v15};
     use bytes::{Buf, Bytes, BytesMut};
     use std::num::NonZeroU32;
     use thiserror::Error;
@@ -288,17 +248,22 @@ pub mod waldecoder {
         }
 
         pub fn poll_decode(&mut self) -> Result<Option<(Lsn, Bytes)>, WalDecodeError> {
-            dispatch_pgversion!(
-                self.pg_version,
-                {
-                    use pgv::waldecoder_handler::WalStreamDecoderHandler;
+            match self.pg_version {
+                // This is a trick to support both versions simultaneously.
+                // See WalStreamDecoderHandler comments.
+                14 => {
+                    use self::v14::waldecoder_handler::WalStreamDecoderHandler;
                     self.poll_decode_internal()
-                },
-                Err(WalDecodeError {
+                }
+                15 => {
+                    use self::v15::waldecoder_handler::WalStreamDecoderHandler;
+                    self.poll_decode_internal()
+                }
+                _ => Err(WalDecodeError {
                     msg: format!("Unknown version {}", self.pg_version),
                     lsn: self.lsn,
-                })
-            )
+                }),
+            }
         }
     }
 }

libs/postgres_ffi/src/pg_constants.rs

@@ -137,12 +137,9 @@ pub const XLOG_HEAP_INSERT: u8 = 0x00;
 pub const XLOG_HEAP_DELETE: u8 = 0x10;
 pub const XLOG_HEAP_UPDATE: u8 = 0x20;
 pub const XLOG_HEAP_HOT_UPDATE: u8 = 0x40;
-pub const XLOG_HEAP_LOCK: u8 = 0x60;
 pub const XLOG_HEAP_INIT_PAGE: u8 = 0x80;
 pub const XLOG_HEAP2_VISIBLE: u8 = 0x40;
 pub const XLOG_HEAP2_MULTI_INSERT: u8 = 0x50;
-pub const XLOG_HEAP2_LOCK_UPDATED: u8 = 0x60;
-pub const XLH_LOCK_ALL_FROZEN_CLEARED: u8 = 0x01;
 pub const XLH_INSERT_ALL_FROZEN_SET: u8 = (1 << 5) as u8;
 pub const XLH_INSERT_ALL_VISIBLE_CLEARED: u8 = (1 << 0) as u8;
 pub const XLH_UPDATE_OLD_ALL_VISIBLE_CLEARED: u8 = (1 << 0) as u8;
@@ -166,20 +163,6 @@ pub const RM_HEAP2_ID: u8 = 9;
 pub const RM_HEAP_ID: u8 = 10;
 pub const RM_LOGICALMSG_ID: u8 = 21;
 
-// from neon_rmgr.h
-pub const RM_NEON_ID: u8 = 134;
-
-pub const XLOG_NEON_HEAP_INIT_PAGE: u8 = 0x80;
-
-pub const XLOG_NEON_HEAP_INSERT: u8 = 0x00;
-pub const XLOG_NEON_HEAP_DELETE: u8 = 0x10;
-pub const XLOG_NEON_HEAP_UPDATE: u8 = 0x20;
-pub const XLOG_NEON_HEAP_HOT_UPDATE: u8 = 0x30;
-pub const XLOG_NEON_HEAP_LOCK: u8 = 0x40;
-pub const XLOG_NEON_HEAP_MULTI_INSERT: u8 = 0x50;
-
-pub const XLOG_NEON_HEAP_VISIBLE: u8 = 0x40;
-
 // from xlogreader.h
 pub const XLR_INFO_MASK: u8 = 0x0F;
 pub const XLR_RMGR_INFO_MASK: u8 = 0xF0;

libs/postgres_ffi/src/pg_constants_v14.rs

@@ -3,8 +3,3 @@ pub const XLOG_DBASE_DROP: u8 = 0x10;
 
 pub const BKPIMAGE_IS_COMPRESSED: u8 = 0x02; /* page image is compressed */
 pub const BKPIMAGE_APPLY: u8 = 0x04; /* page image should be restored during replay */
-pub const SIZEOF_RELMAPFILE: usize = 512; /* sizeof(RelMapFile) in relmapper.c */
-
-pub fn bkpimg_is_compressed(bimg_info: u8) -> bool {
-    (bimg_info & BKPIMAGE_IS_COMPRESSED) != 0
-}

libs/postgres_ffi/src/pg_constants_v15.rs

@@ -1,18 +1,10 @@
 pub const XACT_XINFO_HAS_DROPPED_STATS: u32 = 1u32 << 8;
 
 pub const XLOG_DBASE_CREATE_FILE_COPY: u8 = 0x00;
-pub const XLOG_DBASE_CREATE_WAL_LOG: u8 = 0x10;
+pub const XLOG_DBASE_CREATE_WAL_LOG: u8 = 0x00;
 pub const XLOG_DBASE_DROP: u8 = 0x20;
 
 pub const BKPIMAGE_APPLY: u8 = 0x02; /* page image should be restored during replay */
 pub const BKPIMAGE_COMPRESS_PGLZ: u8 = 0x04; /* page image is compressed */
 pub const BKPIMAGE_COMPRESS_LZ4: u8 = 0x08; /* page image is compressed */
 pub const BKPIMAGE_COMPRESS_ZSTD: u8 = 0x10; /* page image is compressed */
-
-pub const SIZEOF_RELMAPFILE: usize = 512; /* sizeof(RelMapFile) in relmapper.c */
-
-pub fn bkpimg_is_compressed(bimg_info: u8) -> bool {
-    const ANY_COMPRESS_FLAG: u8 = BKPIMAGE_COMPRESS_PGLZ | BKPIMAGE_COMPRESS_LZ4 | BKPIMAGE_COMPRESS_ZSTD;
-
-    (bimg_info & ANY_COMPRESS_FLAG) != 0
-}

libs/postgres_ffi/src/pg_constants_v16.rs

@@ -1,18 +0,0 @@
-pub const XACT_XINFO_HAS_DROPPED_STATS: u32 = 1u32 << 8;
-
-pub const XLOG_DBASE_CREATE_FILE_COPY: u8 = 0x00;
-pub const XLOG_DBASE_CREATE_WAL_LOG: u8 = 0x10;
-pub const XLOG_DBASE_DROP: u8 = 0x20;
-
-pub const BKPIMAGE_APPLY: u8 = 0x02; /* page image should be restored during replay */
-pub const BKPIMAGE_COMPRESS_PGLZ: u8 = 0x04; /* page image is compressed */
-pub const BKPIMAGE_COMPRESS_LZ4: u8 = 0x08; /* page image is compressed */
-pub const BKPIMAGE_COMPRESS_ZSTD: u8 = 0x10; /* page image is compressed */
-
-pub const SIZEOF_RELMAPFILE: usize = 524; /* sizeof(RelMapFile) in relmapper.c */
-
-pub fn bkpimg_is_compressed(bimg_info: u8) -> bool {
-    const ANY_COMPRESS_FLAG: u8 = BKPIMAGE_COMPRESS_PGLZ | BKPIMAGE_COMPRESS_LZ4 | BKPIMAGE_COMPRESS_ZSTD;
-
-    (bimg_info & ANY_COMPRESS_FLAG) != 0
-}

libs/postgres_ffi/wal_craft/Cargo.toml

@@ -12,7 +12,7 @@ log.workspace = true
 once_cell.workspace = true
 postgres.workspace = true
 postgres_ffi.workspace = true
-camino-tempfile.workspace = true
+tempfile.workspace = true
 
 workspace_hack.workspace = true
 

libs/postgres_ffi/wal_craft/src/lib.rs

@@ -1,5 +1,4 @@
 use anyhow::{bail, ensure};
-use camino_tempfile::{tempdir, Utf8TempDir};
 use log::*;
 use postgres::types::PgLsn;
 use postgres::Client;
@@ -9,6 +8,7 @@ use std::cmp::Ordering;
 use std::path::{Path, PathBuf};
 use std::process::Command;
 use std::time::{Duration, Instant};
+use tempfile::{tempdir, TempDir};
 
 macro_rules! xlog_utils_test {
     ($version:ident) => {
@@ -33,7 +33,7 @@ pub struct Conf {
 
 pub struct PostgresServer {
     process: std::process::Child,
-    _unix_socket_dir: Utf8TempDir,
+    _unix_socket_dir: TempDir,
     client_config: postgres::Config,
 }
 
@@ -49,9 +49,9 @@ impl Conf {
     pub fn pg_distrib_dir(&self) -> anyhow::Result<PathBuf> {
         let path = self.pg_distrib_dir.clone();
 
-        #[allow(clippy::manual_range_patterns)]
         match self.pg_version {
-            14 | 15 | 16 => Ok(path.join(format!("v{}", self.pg_version))),
+            14 => Ok(path.join(format!("v{}", self.pg_version))),
+            15 => Ok(path.join(format!("v{}", self.pg_version))),
             _ => bail!("Unsupported postgres version: {}", self.pg_version),
         }
     }
@@ -250,18 +250,11 @@ fn craft_internal<C: postgres::GenericClient>(
     let (mut intermediate_lsns, last_lsn) = f(client, initial_lsn)?;
     let last_lsn = match last_lsn {
         None => client.pg_current_wal_insert_lsn()?,
-        Some(last_lsn) => {
-            let insert_lsn = client.pg_current_wal_insert_lsn()?;
-            match last_lsn.cmp(&insert_lsn) {
-                Ordering::Less => bail!(
-                    "Some records were inserted after the crafted WAL: {} vs {}",
-                    last_lsn,
-                    insert_lsn
-                ),
-                Ordering::Equal => last_lsn,
-                Ordering::Greater => bail!("Reported LSN is greater than insert_lsn"),
-            }
-        }
+        Some(last_lsn) => match last_lsn.cmp(&client.pg_current_wal_insert_lsn()?) {
+            Ordering::Less => bail!("Some records were inserted after the crafted WAL"),
+            Ordering::Equal => last_lsn,
+            Ordering::Greater => bail!("Reported LSN is greater than insert_lsn"),
+        },
     };
     if !intermediate_lsns.starts_with(&[initial_lsn]) {
         intermediate_lsns.insert(0, initial_lsn);
@@ -370,9 +363,8 @@ impl Crafter for LastWalRecordXlogSwitchEndsOnPageBoundary {
         );
         ensure!(
             u64::from(after_xlog_switch) as usize % XLOG_BLCKSZ == XLOG_SIZE_OF_XLOG_SHORT_PHD,
-            "XLOG_SWITCH message ended not on page boundary: {}, offset = {}",
-            after_xlog_switch,
-            u64::from(after_xlog_switch) as usize % XLOG_BLCKSZ
+            "XLOG_SWITCH message ended not on page boundary: {}",
+            after_xlog_switch
         );
         Ok((vec![before_xlog_switch, after_xlog_switch], next_segment))
     }

libs/pq_proto/src/lib.rs

@@ -959,7 +959,7 @@ mod tests {
         let make_params = |options| StartupMessageParams::new([("options", options)]);
 
         let params = StartupMessageParams::new([]);
-        assert!(params.options_escaped().is_none());
+        assert!(matches!(params.options_escaped(), None));
 
         let params = make_params("");
         assert!(split_options(&params).is_empty());

libs/remote_storage/Cargo.toml

@@ -13,7 +13,6 @@ aws-types.workspace = true
 aws-config.workspace = true
 aws-sdk-s3.workspace = true
 aws-credential-types.workspace = true
-camino.workspace = true
 hyper = { workspace = true, features = ["stream"] }
 serde.workspace = true
 serde_json.workspace = true
@@ -28,6 +27,5 @@ pin-project-lite.workspace = true
 workspace_hack.workspace = true
 
 [dev-dependencies]
-camino-tempfile.workspace = true
+tempfile.workspace = true
 test-context.workspace = true
-rand.workspace = true

libs/remote_storage/src/lib.rs

@@ -13,14 +13,13 @@ use std::{
     collections::HashMap,
     fmt::Debug,
     num::{NonZeroU32, NonZeroUsize},
+    path::{Path, PathBuf},
     pin::Pin,
     sync::Arc,
 };
 
 use anyhow::{bail, Context};
-use camino::{Utf8Path, Utf8PathBuf};
 
-use serde::{Deserialize, Serialize};
 use tokio::io;
 use toml_edit::Item;
 use tracing::info;
@@ -43,44 +42,22 @@ pub const DEFAULT_REMOTE_STORAGE_S3_CONCURRENCY_LIMIT: usize = 100;
 /// <https://docs.aws.amazon.com/AmazonS3/latest/API/API_ListObjectsV2.html#API_ListObjectsV2_RequestSyntax>
 pub const DEFAULT_MAX_KEYS_PER_LIST_RESPONSE: Option<i32> = None;
 
-/// As defined in S3 docs
-pub const MAX_KEYS_PER_DELETE: usize = 1000;
-
 const REMOTE_STORAGE_PREFIX_SEPARATOR: char = '/';
 
 /// Path on the remote storage, relative to some inner prefix.
 /// The prefix is an implementation detail, that allows representing local paths
 /// as the remote ones, stripping the local storage prefix away.
 #[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
-pub struct RemotePath(Utf8PathBuf);
-
-impl Serialize for RemotePath {
-    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
-    where
-        S: serde::Serializer,
-    {
-        serializer.collect_str(self)
-    }
-}
-
-impl<'de> Deserialize<'de> for RemotePath {
-    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
-    where
-        D: serde::Deserializer<'de>,
-    {
-        let str = String::deserialize(deserializer)?;
-        Ok(Self(Utf8PathBuf::from(&str)))
-    }
-}
+pub struct RemotePath(PathBuf);
 
 impl std::fmt::Display for RemotePath {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        std::fmt::Display::fmt(&self.0, f)
+        write!(f, "{}", self.0.display())
     }
 }
 
 impl RemotePath {
-    pub fn new(relative_path: &Utf8Path) -> anyhow::Result<Self> {
+    pub fn new(relative_path: &Path) -> anyhow::Result<Self> {
         anyhow::ensure!(
             relative_path.is_relative(),
             "Path {relative_path:?} is not relative"
@@ -89,31 +66,27 @@ impl RemotePath {
     }
 
     pub fn from_string(relative_path: &str) -> anyhow::Result<Self> {
-        Self::new(Utf8Path::new(relative_path))
+        Self::new(Path::new(relative_path))
     }
 
-    pub fn with_base(&self, base_path: &Utf8Path) -> Utf8PathBuf {
+    pub fn with_base(&self, base_path: &Path) -> PathBuf {
         base_path.join(&self.0)
     }
 
     pub fn object_name(&self) -> Option<&str> {
-        self.0.file_name()
+        self.0.file_name().and_then(|os_str| os_str.to_str())
     }
 
-    pub fn join(&self, segment: &Utf8Path) -> Self {
+    pub fn join(&self, segment: &Path) -> Self {
         Self(self.0.join(segment))
     }
 
-    pub fn get_path(&self) -> &Utf8PathBuf {
+    pub fn get_path(&self) -> &PathBuf {
         &self.0
     }
 
     pub fn extension(&self) -> Option<&str> {
-        self.0.extension()
-    }
-
-    pub fn strip_prefix(&self, p: &RemotePath) -> Result<&Utf8Path, std::path::StripPrefixError> {
-        self.0.strip_prefix(&p.0)
+        self.0.extension()?.to_str()
     }
 }
 
@@ -311,7 +284,7 @@ impl GenericRemoteStorage {
     pub fn from_config(storage_config: &RemoteStorageConfig) -> anyhow::Result<Self> {
         Ok(match &storage_config.storage {
             RemoteStorageKind::LocalFs(root) => {
-                info!("Using fs root '{root}' as a remote storage");
+                info!("Using fs root '{}' as a remote storage", root.display());
                 Self::LocalFs(LocalFs::new(root.clone())?)
             }
             RemoteStorageKind::AwsS3(s3_config) => {
@@ -379,7 +352,7 @@ pub struct RemoteStorageConfig {
 pub enum RemoteStorageKind {
     /// Storage based on local file system.
     /// Specify a root folder to place all stored files into.
-    LocalFs(Utf8PathBuf),
+    LocalFs(PathBuf),
     /// AWS S3 based storage, storing all files in the S3 bucket
     /// specified by the config
     AwsS3(S3Config),
@@ -474,7 +447,7 @@ impl RemoteStorageConfig {
                 concurrency_limit,
                 max_keys_per_list_response,
             }),
-            (Some(local_path), None, None) => RemoteStorageKind::LocalFs(Utf8PathBuf::from(
+            (Some(local_path), None, None) => RemoteStorageKind::LocalFs(PathBuf::from(
                 parse_toml_string("local_path", local_path)?,
             )),
             (Some(_), Some(_), _) => bail!("local_path and bucket_name are mutually exclusive"),
@@ -519,23 +492,23 @@ mod tests {
 
     #[test]
     fn test_object_name() {
-        let k = RemotePath::new(Utf8Path::new("a/b/c")).unwrap();
+        let k = RemotePath::new(Path::new("a/b/c")).unwrap();
         assert_eq!(k.object_name(), Some("c"));
 
-        let k = RemotePath::new(Utf8Path::new("a/b/c/")).unwrap();
+        let k = RemotePath::new(Path::new("a/b/c/")).unwrap();
         assert_eq!(k.object_name(), Some("c"));
 
-        let k = RemotePath::new(Utf8Path::new("a/")).unwrap();
+        let k = RemotePath::new(Path::new("a/")).unwrap();
         assert_eq!(k.object_name(), Some("a"));
 
         // XXX is it impossible to have an empty key?
-        let k = RemotePath::new(Utf8Path::new("")).unwrap();
+        let k = RemotePath::new(Path::new("")).unwrap();
         assert_eq!(k.object_name(), None);
     }
 
     #[test]
     fn rempte_path_cannot_be_created_from_absolute_ones() {
-        let err = RemotePath::new(Utf8Path::new("/")).expect_err("Should fail on absolute paths");
+        let err = RemotePath::new(Path::new("/")).expect_err("Should fail on absolute paths");
         assert_eq!(err.to_string(), "Path \"/\" is not relative");
     }
 }

libs/remote_storage/src/local_fs.rs

@@ -4,10 +4,15 @@
 //! This storage used in tests, but can also be used in cases when a certain persistent
 //! volume is mounted to the local FS.
 
-use std::{borrow::Cow, future::Future, io::ErrorKind, pin::Pin};
+use std::{
+    borrow::Cow,
+    future::Future,
+    io::ErrorKind,
+    path::{Path, PathBuf},
+    pin::Pin,
+};
 
 use anyhow::{bail, ensure, Context};
-use camino::{Utf8Path, Utf8PathBuf};
 use tokio::{
     fs,
     io::{self, AsyncReadExt, AsyncSeekExt, AsyncWriteExt},
@@ -23,20 +28,20 @@ const LOCAL_FS_TEMP_FILE_SUFFIX: &str = "___temp";
 
 #[derive(Debug, Clone)]
 pub struct LocalFs {
-    storage_root: Utf8PathBuf,
+    storage_root: PathBuf,
 }
 
 impl LocalFs {
     /// Attempts to create local FS storage, along with its root directory.
     /// Storage root will be created (if does not exist) and transformed into an absolute path (if passed as relative).
-    pub fn new(mut storage_root: Utf8PathBuf) -> anyhow::Result<Self> {
+    pub fn new(mut storage_root: PathBuf) -> anyhow::Result<Self> {
         if !storage_root.exists() {
             std::fs::create_dir_all(&storage_root).with_context(|| {
                 format!("Failed to create all directories in the given root path {storage_root:?}")
             })?;
         }
         if !storage_root.is_absolute() {
-            storage_root = storage_root.canonicalize_utf8().with_context(|| {
+            storage_root = storage_root.canonicalize().with_context(|| {
                 format!("Failed to represent path {storage_root:?} as an absolute path")
             })?;
         }
@@ -45,7 +50,7 @@ impl LocalFs {
     }
 
     // mirrors S3Bucket::s3_object_to_relative_path
-    fn local_file_to_relative_path(&self, key: Utf8PathBuf) -> RemotePath {
+    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");
@@ -54,18 +59,22 @@ impl LocalFs {
 
     async fn read_storage_metadata(
         &self,
-        file_path: &Utf8Path,
+        file_path: &Path,
     ) -> anyhow::Result<Option<StorageMetadata>> {
         let metadata_path = storage_metadata_path(file_path);
         if metadata_path.exists() && metadata_path.is_file() {
             let metadata_string = fs::read_to_string(&metadata_path).await.with_context(|| {
-                format!("Failed to read metadata from the local storage at '{metadata_path}'")
+                format!(
+                    "Failed to read metadata from the local storage at '{}'",
+                    metadata_path.display()
+                )
             })?;
 
             serde_json::from_str(&metadata_string)
                 .with_context(|| {
                     format!(
-                        "Failed to deserialize metadata from the local storage at '{metadata_path}'",
+                        "Failed to deserialize metadata from the local storage at '{}'",
+                        metadata_path.display()
                     )
                 })
                 .map(|metadata| Some(StorageMetadata(metadata)))
@@ -162,21 +171,25 @@ impl RemoteStorage for LocalFs {
             }
         }
 
-        // Note that Utf8PathBuf starts_with only considers full path segments, but
+        // Note that PathBuf starts_with only considers full path segments, but
         // object prefixes are arbitrary strings, so we need the strings for doing
         // starts_with later.
-        let prefix = full_path.as_str();
+        let prefix = full_path.to_string_lossy();
 
         let mut files = vec![];
-        let mut directory_queue = vec![initial_dir];
+        let mut directory_queue = vec![initial_dir.clone()];
         while let Some(cur_folder) = directory_queue.pop() {
-            let mut entries = cur_folder.read_dir_utf8()?;
-            while let Some(Ok(entry)) = entries.next() {
-                let file_name = entry.file_name();
-                let full_file_name = cur_folder.join(file_name);
-                if full_file_name.as_str().starts_with(prefix) {
+            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);
+                if full_file_name
+                    .to_str()
+                    .map(|s| s.starts_with(prefix.as_ref()))
+                    .unwrap_or(false)
+                {
                     let file_remote_path = self.local_file_to_relative_path(full_file_name.clone());
-                    files.push(file_remote_path);
+                    files.push(file_remote_path.clone());
                     if full_file_name.is_dir() {
                         directory_queue.push(full_file_name);
                     }
@@ -217,7 +230,10 @@ impl RemoteStorage for LocalFs {
                 .open(&temp_file_path)
                 .await
                 .with_context(|| {
-                    format!("Failed to open target fs destination at '{target_file_path}'")
+                    format!(
+                        "Failed to open target fs destination at '{}'",
+                        target_file_path.display()
+                    )
                 })?,
         );
 
@@ -228,7 +244,8 @@ impl RemoteStorage for LocalFs {
             .await
             .with_context(|| {
                 format!(
-                    "Failed to upload file (write temp) to the local storage at '{temp_file_path}'",
+                    "Failed to upload file (write temp) to the local storage at '{}'",
+                    temp_file_path.display()
                 )
             })?;
 
@@ -245,7 +262,8 @@ impl RemoteStorage for LocalFs {
 
         destination.flush().await.with_context(|| {
             format!(
-                "Failed to upload (flush temp) file to the local storage at '{temp_file_path}'",
+                "Failed to upload (flush temp) file to the local storage at '{}'",
+                temp_file_path.display()
             )
         })?;
 
@@ -253,7 +271,8 @@ impl RemoteStorage for LocalFs {
             .await
             .with_context(|| {
                 format!(
-                    "Failed to upload (rename) file to the local storage at '{target_file_path}'",
+                    "Failed to upload (rename) file to the local storage at '{}'",
+                    target_file_path.display()
                 )
             })?;
 
@@ -267,7 +286,8 @@ impl RemoteStorage for LocalFs {
             .await
             .with_context(|| {
                 format!(
-                    "Failed to write metadata to the local storage at '{storage_metadata_path}'",
+                    "Failed to write metadata to the local storage at '{}'",
+                    storage_metadata_path.display()
                 )
             })?;
         }
@@ -373,16 +393,16 @@ impl RemoteStorage for LocalFs {
     }
 }
 
-fn storage_metadata_path(original_path: &Utf8Path) -> Utf8PathBuf {
+fn storage_metadata_path(original_path: &Path) -> PathBuf {
     path_with_suffix_extension(original_path, "metadata")
 }
 
 fn get_all_files<'a, P>(
     directory_path: P,
     recursive: bool,
-) -> Pin<Box<dyn Future<Output = anyhow::Result<Vec<Utf8PathBuf>>> + Send + Sync + 'a>>
+) -> Pin<Box<dyn Future<Output = anyhow::Result<Vec<PathBuf>>> + Send + Sync + 'a>>
 where
-    P: AsRef<Utf8Path> + Send + Sync + 'a,
+    P: AsRef<Path> + Send + Sync + 'a,
 {
     Box::pin(async move {
         let directory_path = directory_path.as_ref();
@@ -392,13 +412,7 @@ where
                 let mut dir_contents = fs::read_dir(directory_path).await?;
                 while let Some(dir_entry) = dir_contents.next_entry().await? {
                     let file_type = dir_entry.file_type().await?;
-                    let entry_path =
-                        Utf8PathBuf::from_path_buf(dir_entry.path()).map_err(|pb| {
-                            anyhow::Error::msg(format!(
-                                "non-Unicode path: {}",
-                                pb.to_string_lossy()
-                            ))
-                        })?;
+                    let entry_path = dir_entry.path();
                     if file_type.is_symlink() {
                         debug!("{entry_path:?} is a symlink, skipping")
                     } else if file_type.is_dir() {
@@ -421,10 +435,13 @@ where
     })
 }
 
-async fn create_target_directory(target_file_path: &Utf8Path) -> anyhow::Result<()> {
+async fn create_target_directory(target_file_path: &Path) -> anyhow::Result<()> {
     let target_dir = match target_file_path.parent() {
         Some(parent_dir) => parent_dir,
-        None => bail!("File path '{target_file_path}' has no parent directory"),
+        None => bail!(
+            "File path '{}' has no parent directory",
+            target_file_path.display()
+        ),
     };
     if !target_dir.exists() {
         fs::create_dir_all(target_dir).await?;
@@ -432,9 +449,13 @@ async fn create_target_directory(target_file_path: &Utf8Path) -> anyhow::Result<
     Ok(())
 }
 
-fn file_exists(file_path: &Utf8Path) -> anyhow::Result<bool> {
+fn file_exists(file_path: &Path) -> anyhow::Result<bool> {
     if file_path.exists() {
-        ensure!(file_path.is_file(), "file path '{file_path}' is not a file");
+        ensure!(
+            file_path.is_file(),
+            "file path '{}' is not a file",
+            file_path.display()
+        );
         Ok(true)
     } else {
         Ok(false)
@@ -445,13 +466,13 @@ fn file_exists(file_path: &Utf8Path) -> anyhow::Result<bool> {
 mod fs_tests {
     use super::*;
 
-    use camino_tempfile::tempdir;
     use std::{collections::HashMap, io::Write};
+    use tempfile::tempdir;
 
     async fn read_and_assert_remote_file_contents(
         storage: &LocalFs,
         #[allow(clippy::ptr_arg)]
-        // have to use &Utf8PathBuf due to `storage.local_path` parameter requirements
+        // have to use &PathBuf due to `storage.local_path` parameter requirements
         remote_storage_path: &RemotePath,
         expected_metadata: Option<&StorageMetadata>,
     ) -> anyhow::Result<String> {
@@ -498,7 +519,7 @@ mod fs_tests {
     async fn upload_file_negatives() -> anyhow::Result<()> {
         let storage = create_storage()?;
 
-        let id = RemotePath::new(Utf8Path::new("dummy"))?;
+        let id = RemotePath::new(Path::new("dummy"))?;
         let content = std::io::Cursor::new(b"12345");
 
         // Check that you get an error if the size parameter doesn't match the actual
@@ -523,8 +544,7 @@ mod fs_tests {
     }
 
     fn create_storage() -> anyhow::Result<LocalFs> {
-        let storage_root = tempdir()?.path().to_path_buf();
-        LocalFs::new(storage_root)
+        LocalFs::new(tempdir()?.path().to_owned())
     }
 
     #[tokio::test]
@@ -541,7 +561,7 @@ mod fs_tests {
         );
 
         let non_existing_path = "somewhere/else";
-        match storage.download(&RemotePath::new(Utf8Path::new(non_existing_path))?).await {
+        match storage.download(&RemotePath::new(Path::new(non_existing_path))?).await {
             Err(DownloadError::NotFound) => {} // Should get NotFound for non existing keys
             other => panic!("Should get a NotFound error when downloading non-existing storage files, but got: {other:?}"),
         }
@@ -755,7 +775,7 @@ mod fs_tests {
     }
 
     async fn create_file_for_upload(
-        path: &Utf8Path,
+        path: &Path,
         contents: &str,
     ) -> anyhow::Result<(io::BufReader<fs::File>, usize)> {
         std::fs::create_dir_all(path.parent().unwrap())?;

libs/remote_storage/src/s3_bucket.rs

@@ -33,10 +33,11 @@ use tracing::debug;
 
 use super::StorageMetadata;
 use crate::{
-    Download, DownloadError, RemotePath, RemoteStorage, S3Config, MAX_KEYS_PER_DELETE,
-    REMOTE_STORAGE_PREFIX_SEPARATOR,
+    Download, DownloadError, RemotePath, RemoteStorage, S3Config, REMOTE_STORAGE_PREFIX_SEPARATOR,
 };
 
+const MAX_DELETE_OBJECTS_REQUEST_SIZE: usize = 1000;
+
 pub(super) mod metrics;
 
 use self::metrics::{AttemptOutcome, RequestKind};
@@ -47,47 +48,10 @@ pub struct S3Bucket {
     bucket_name: String,
     prefix_in_bucket: Option<String>,
     max_keys_per_list_response: Option<i32>,
-    concurrency_limiter: ConcurrencyLimiter,
-}
-
-struct ConcurrencyLimiter {
     // Every request to S3 can be throttled or cancelled, if a certain number of requests per second is exceeded.
     // Same goes to IAM, which is queried before every S3 request, if enabled. IAM has even lower RPS threshold.
     // The helps to ensure we don't exceed the thresholds.
-    write: Arc<Semaphore>,
-    read: Arc<Semaphore>,
-}
-
-impl ConcurrencyLimiter {
-    fn for_kind(&self, kind: RequestKind) -> &Arc<Semaphore> {
-        match kind {
-            RequestKind::Get => &self.read,
-            RequestKind::Put => &self.write,
-            RequestKind::List => &self.read,
-            RequestKind::Delete => &self.write,
-        }
-    }
-
-    async fn acquire(
-        &self,
-        kind: RequestKind,
-    ) -> Result<tokio::sync::SemaphorePermit<'_>, tokio::sync::AcquireError> {
-        self.for_kind(kind).acquire().await
-    }
-
-    async fn acquire_owned(
-        &self,
-        kind: RequestKind,
-    ) -> Result<tokio::sync::OwnedSemaphorePermit, tokio::sync::AcquireError> {
-        Arc::clone(self.for_kind(kind)).acquire_owned().await
-    }
-
-    fn new(limit: usize) -> ConcurrencyLimiter {
-        Self {
-            read: Arc::new(Semaphore::new(limit)),
-            write: Arc::new(Semaphore::new(limit)),
-        }
-    }
+    concurrency_limiter: Arc<Semaphore>,
 }
 
 #[derive(Default)]
@@ -154,7 +118,7 @@ impl S3Bucket {
             bucket_name: aws_config.bucket_name.clone(),
             max_keys_per_list_response: aws_config.max_keys_per_list_response,
             prefix_in_bucket,
-            concurrency_limiter: ConcurrencyLimiter::new(aws_config.concurrency_limit.get()),
+            concurrency_limiter: Arc::new(Semaphore::new(aws_config.concurrency_limit.get())),
         })
     }
 
@@ -180,11 +144,12 @@ impl S3Bucket {
         assert_eq!(std::path::MAIN_SEPARATOR, REMOTE_STORAGE_PREFIX_SEPARATOR);
         let path_string = path
             .get_path()
-            .as_str()
-            .trim_end_matches(REMOTE_STORAGE_PREFIX_SEPARATOR);
+            .to_string_lossy()
+            .trim_end_matches(REMOTE_STORAGE_PREFIX_SEPARATOR)
+            .to_string();
         match &self.prefix_in_bucket {
-            Some(prefix) => prefix.clone() + "/" + path_string,
-            None => path_string.to_string(),
+            Some(prefix) => prefix.clone() + "/" + &path_string,
+            None => path_string,
         }
     }
 
@@ -192,7 +157,7 @@ impl S3Bucket {
         let started_at = start_counting_cancelled_wait(kind);
         let permit = self
             .concurrency_limiter
-            .acquire(kind)
+            .acquire()
             .await
             .expect("semaphore is never closed");
 
@@ -208,7 +173,8 @@ impl S3Bucket {
         let started_at = start_counting_cancelled_wait(kind);
         let permit = self
             .concurrency_limiter
-            .acquire_owned(kind)
+            .clone()
+            .acquire_owned()
             .await
             .expect("semaphore is never closed");
 
@@ -534,7 +500,7 @@ impl RemoteStorage for S3Bucket {
             delete_objects.push(obj_id);
         }
 
-        for chunk in delete_objects.chunks(MAX_KEYS_PER_DELETE) {
+        for chunk in delete_objects.chunks(MAX_DELETE_OBJECTS_REQUEST_SIZE) {
             let started_at = start_measuring_requests(kind);
 
             let resp = self
@@ -600,17 +566,17 @@ fn start_measuring_requests(
 
 #[cfg(test)]
 mod tests {
-    use camino::Utf8Path;
     use std::num::NonZeroUsize;
+    use std::path::Path;
 
     use crate::{RemotePath, S3Bucket, S3Config};
 
     #[test]
     fn relative_path() {
-        let all_paths = ["", "some/path", "some/path/"];
+        let all_paths = vec!["", "some/path", "some/path/"];
         let all_paths: Vec<RemotePath> = all_paths
             .iter()
-            .map(|x| RemotePath::new(Utf8Path::new(x)).expect("bad path"))
+            .map(|x| RemotePath::new(Path::new(x)).expect("bad path"))
             .collect();
         let prefixes = [
             None,

libs/remote_storage/tests/test_real_s3.rs

@@ -2,12 +2,11 @@ use std::collections::HashSet;
 use std::env;
 use std::num::{NonZeroU32, NonZeroUsize};
 use std::ops::ControlFlow;
-use std::path::PathBuf;
+use std::path::{Path, PathBuf};
 use std::sync::Arc;
 use std::time::UNIX_EPOCH;
 
 use anyhow::Context;
-use camino::Utf8Path;
 use once_cell::sync::OnceCell;
 use remote_storage::{
     GenericRemoteStorage, RemotePath, RemoteStorageConfig, RemoteStorageKind, S3Config,
@@ -56,7 +55,7 @@ async fn s3_pagination_should_work(ctx: &mut MaybeEnabledS3WithTestBlobs) -> any
     let test_client = Arc::clone(&ctx.enabled.client);
     let expected_remote_prefixes = ctx.remote_prefixes.clone();
 
-    let base_prefix = RemotePath::new(Utf8Path::new(ctx.enabled.base_prefix))
+    let base_prefix = RemotePath::new(Path::new(ctx.enabled.base_prefix))
         .context("common_prefix construction")?;
     let root_remote_prefixes = test_client
         .list_prefixes(None)
@@ -109,7 +108,7 @@ async fn s3_list_files_works(ctx: &mut MaybeEnabledS3WithSimpleTestBlobs) -> any
     };
     let test_client = Arc::clone(&ctx.enabled.client);
     let base_prefix =
-        RemotePath::new(Utf8Path::new("folder1")).context("common_prefix construction")?;
+        RemotePath::new(Path::new("folder1")).context("common_prefix construction")?;
     let root_files = test_client
         .list_files(None)
         .await
@@ -130,9 +129,9 @@ async fn s3_list_files_works(ctx: &mut MaybeEnabledS3WithSimpleTestBlobs) -> any
     let trim_remote_blobs: HashSet<_> = ctx
         .remote_blobs
         .iter()
-        .map(|x| x.get_path())
+        .map(|x| x.get_path().to_str().expect("must be valid name"))
         .filter(|x| x.starts_with("folder1"))
-        .map(|x| RemotePath::new(x).expect("must be valid path"))
+        .map(|x| RemotePath::new(Path::new(x)).expect("must be valid name"))
         .collect();
     assert_eq!(
         nested_remote_files, trim_remote_blobs,
@@ -149,9 +148,10 @@ async fn s3_delete_non_exising_works(ctx: &mut MaybeEnabledS3) -> anyhow::Result
         MaybeEnabledS3::Disabled => return Ok(()),
     };
 
-    let path = RemotePath::new(Utf8Path::new(
-        format!("{}/for_sure_there_is_nothing_there_really", ctx.base_prefix).as_str(),
-    ))
+    let path = RemotePath::new(&PathBuf::from(format!(
+        "{}/for_sure_there_is_nothing_there_really",
+        ctx.base_prefix,
+    )))
     .with_context(|| "RemotePath conversion")?;
 
     ctx.client.delete(&path).await.expect("should succeed");
@@ -167,13 +167,13 @@ async fn s3_delete_objects_works(ctx: &mut MaybeEnabledS3) -> anyhow::Result<()>
         MaybeEnabledS3::Disabled => return Ok(()),
     };
 
-    let path1 = RemotePath::new(Utf8Path::new(format!("{}/path1", ctx.base_prefix).as_str()))
+    let path1 = RemotePath::new(&PathBuf::from(format!("{}/path1", ctx.base_prefix,)))
         .with_context(|| "RemotePath conversion")?;
 
-    let path2 = RemotePath::new(Utf8Path::new(format!("{}/path2", ctx.base_prefix).as_str()))
+    let path2 = RemotePath::new(&PathBuf::from(format!("{}/path2", ctx.base_prefix,)))
         .with_context(|| "RemotePath conversion")?;
 
-    let path3 = RemotePath::new(Utf8Path::new(format!("{}/path3", ctx.base_prefix).as_str()))
+    let path3 = RemotePath::new(&PathBuf::from(format!("{}/path3", ctx.base_prefix,)))
         .with_context(|| "RemotePath conversion")?;
 
     let data1 = "remote blob data1".as_bytes();
@@ -378,30 +378,21 @@ impl AsyncTestContext for MaybeEnabledS3WithSimpleTestBlobs {
 fn create_s3_client(
     max_keys_per_list_response: Option<i32>,
 ) -> anyhow::Result<Arc<GenericRemoteStorage>> {
-    use rand::Rng;
-
     let remote_storage_s3_bucket = env::var("REMOTE_STORAGE_S3_BUCKET")
         .context("`REMOTE_STORAGE_S3_BUCKET` env var is not set, but real S3 tests are enabled")?;
     let remote_storage_s3_region = env::var("REMOTE_STORAGE_S3_REGION")
         .context("`REMOTE_STORAGE_S3_REGION` env var is not set, but real S3 tests are enabled")?;
-
-    // due to how time works, we've had test runners use the same nanos as bucket prefixes.
-    // millis is just a debugging aid for easier finding the prefix later.
-    let millis = std::time::SystemTime::now()
+    let random_prefix_part = std::time::SystemTime::now()
         .duration_since(UNIX_EPOCH)
         .context("random s3 test prefix part calculation")?
-        .as_millis();
-
-    // because nanos can be the same for two threads so can millis, add randomness
-    let random = rand::thread_rng().gen::<u32>();
-
+        .as_nanos();
     let remote_storage_config = RemoteStorageConfig {
         max_concurrent_syncs: NonZeroUsize::new(100).unwrap(),
         max_sync_errors: NonZeroU32::new(5).unwrap(),
         storage: RemoteStorageKind::AwsS3(S3Config {
             bucket_name: remote_storage_s3_bucket,
             bucket_region: remote_storage_s3_region,
-            prefix_in_bucket: Some(format!("test_{millis}_{random:08x}/")),
+            prefix_in_bucket: Some(format!("pagination_should_work_test_{random_prefix_part}/")),
             endpoint: None,
             concurrency_limit: NonZeroUsize::new(100).unwrap(),
             max_keys_per_list_response,
@@ -427,10 +418,10 @@ async fn upload_s3_data(
     for i in 1..upload_tasks_count + 1 {
         let task_client = Arc::clone(client);
         upload_tasks.spawn(async move {
-            let prefix = format!("{base_prefix_str}/sub_prefix_{i}/");
-            let blob_prefix = RemotePath::new(Utf8Path::new(&prefix))
+            let prefix = PathBuf::from(format!("{base_prefix_str}/sub_prefix_{i}/"));
+            let blob_prefix = RemotePath::new(&prefix)
                 .with_context(|| format!("{prefix:?} to RemotePath conversion"))?;
-            let blob_path = blob_prefix.join(Utf8Path::new(&format!("blob_{i}")));
+            let blob_path = blob_prefix.join(Path::new(&format!("blob_{i}")));
             debug!("Creating remote item {i} at path {blob_path:?}");
 
             let data = format!("remote blob data {i}").into_bytes();
@@ -512,10 +503,8 @@ async fn upload_simple_s3_data(
         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(
-                Utf8Path::from_path(blob_path.as_path()).expect("must be valid blob path"),
-            )
-            .with_context(|| format!("{blob_path:?} to RemotePath conversion"))?;
+            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();

libs/utils/Cargo.toml

@@ -10,7 +10,6 @@ async-trait.workspace = true
 anyhow.workspace = true
 bincode.workspace = true
 bytes.workspace = true
-camino.workspace = true
 chrono.workspace = true
 heapless.workspace = true
 hex = { workspace = true, features = ["serde"] }
@@ -54,7 +53,7 @@ byteorder.workspace = true
 bytes.workspace = true
 criterion.workspace = true
 hex-literal.workspace = true
-camino-tempfile.workspace = true
+tempfile.workspace = true
 
 [[bench]]
 name = "benchmarks"

libs/utils/scripts/restore_from_wal.sh

@@ -9,12 +9,11 @@ PORT=$4
 SYSID=$(od -A n -j 24 -N 8 -t d8 "$WAL_PATH"/000000010000000000000002* | cut -c 3-)
 rm -fr "$DATA_DIR"
 env -i LD_LIBRARY_PATH="$PG_BIN"/../lib "$PG_BIN"/initdb -E utf8 -U cloud_admin -D "$DATA_DIR" --sysid="$SYSID"
-echo "port=$PORT" >> "$DATA_DIR"/postgresql.conf
-echo "shared_preload_libraries='\$libdir/neon_rmgr.so'" >> "$DATA_DIR"/postgresql.conf
+echo port="$PORT" >> "$DATA_DIR"/postgresql.conf
 REDO_POS=0x$("$PG_BIN"/pg_controldata -D "$DATA_DIR" | grep -F "REDO location"| cut -c 42-)
 declare -i WAL_SIZE=$REDO_POS+114
-"$PG_BIN"/pg_ctl -D "$DATA_DIR" -l "$DATA_DIR/logfile.log" start
-"$PG_BIN"/pg_ctl -D "$DATA_DIR" -l "$DATA_DIR/logfile.log" stop -m immediate
+"$PG_BIN"/pg_ctl -D "$DATA_DIR" -l logfile start
+"$PG_BIN"/pg_ctl -D "$DATA_DIR" -l logfile stop -m immediate
 cp "$DATA_DIR"/pg_wal/000000010000000000000001 .
 cp "$WAL_PATH"/* "$DATA_DIR"/pg_wal/
 for partial in "$DATA_DIR"/pg_wal/*.partial ; do mv "$partial" "${partial%.partial}" ; done

libs/utils/src/auth.rs

@@ -2,9 +2,9 @@
 
 use serde;
 use std::fs;
+use std::path::Path;
 
 use anyhow::Result;
-use camino::Utf8Path;
 use jsonwebtoken::{
     decode, encode, Algorithm, DecodingKey, EncodingKey, Header, TokenData, Validation,
 };
@@ -65,7 +65,7 @@ impl JwtAuth {
         }
     }
 
-    pub fn from_key_path(key_path: &Utf8Path) -> Result<Self> {
+    pub fn from_key_path(key_path: &Path) -> Result<Self> {
         let public_key = fs::read(key_path)?;
         Ok(Self::new(DecodingKey::from_ed_pem(&public_key)?))
     }

libs/utils/src/crashsafe.rs

@@ -1,14 +1,14 @@
 use std::{
     borrow::Cow,
+    ffi::OsStr,
     fs::{self, File},
     io,
+    path::{Path, PathBuf},
 };
 
-use camino::{Utf8Path, Utf8PathBuf};
-
 /// Similar to [`std::fs::create_dir`], except we fsync the
 /// created directory and its parent.
-pub fn create_dir(path: impl AsRef<Utf8Path>) -> io::Result<()> {
+pub fn create_dir(path: impl AsRef<Path>) -> io::Result<()> {
     let path = path.as_ref();
 
     fs::create_dir(path)?;
@@ -18,7 +18,7 @@ pub fn create_dir(path: impl AsRef<Utf8Path>) -> io::Result<()> {
 
 /// Similar to [`std::fs::create_dir_all`], except we fsync all
 /// newly created directories and the pre-existing parent.
-pub fn create_dir_all(path: impl AsRef<Utf8Path>) -> io::Result<()> {
+pub fn create_dir_all(path: impl AsRef<Path>) -> io::Result<()> {
     let mut path = path.as_ref();
 
     let mut dirs_to_create = Vec::new();
@@ -30,7 +30,7 @@ pub fn create_dir_all(path: impl AsRef<Utf8Path>) -> io::Result<()> {
             Ok(_) => {
                 return Err(io::Error::new(
                     io::ErrorKind::AlreadyExists,
-                    format!("non-directory found in path: {path}"),
+                    format!("non-directory found in path: {}", path.display()),
                 ));
             }
             Err(ref e) if e.kind() == io::ErrorKind::NotFound => {}
@@ -44,7 +44,7 @@ pub fn create_dir_all(path: impl AsRef<Utf8Path>) -> io::Result<()> {
             None => {
                 return Err(io::Error::new(
                     io::ErrorKind::InvalidInput,
-                    format!("can't find parent of path '{path}'"),
+                    format!("can't find parent of path '{}'", path.display()).as_str(),
                 ));
             }
         }
@@ -70,18 +70,21 @@ pub fn create_dir_all(path: impl AsRef<Utf8Path>) -> io::Result<()> {
 
 /// Adds a suffix to the file(directory) name, either appending the suffix to the end of its extension,
 /// or if there's no extension, creates one and puts a suffix there.
-pub fn path_with_suffix_extension(
-    original_path: impl AsRef<Utf8Path>,
-    suffix: &str,
-) -> Utf8PathBuf {
-    let new_extension = match original_path.as_ref().extension() {
+pub fn path_with_suffix_extension(original_path: impl AsRef<Path>, suffix: &str) -> PathBuf {
+    let new_extension = match original_path
+        .as_ref()
+        .extension()
+        .map(OsStr::to_string_lossy)
+    {
         Some(extension) => Cow::Owned(format!("{extension}.{suffix}")),
         None => Cow::Borrowed(suffix),
     };
-    original_path.as_ref().with_extension(new_extension)
+    original_path
+        .as_ref()
+        .with_extension(new_extension.as_ref())
 }
 
-pub fn fsync_file_and_parent(file_path: &Utf8Path) -> io::Result<()> {
+pub fn fsync_file_and_parent(file_path: &Path) -> io::Result<()> {
     let parent = file_path.parent().ok_or_else(|| {
         io::Error::new(
             io::ErrorKind::Other,
@@ -94,7 +97,7 @@ pub fn fsync_file_and_parent(file_path: &Utf8Path) -> io::Result<()> {
     Ok(())
 }
 
-pub fn fsync(path: &Utf8Path) -> io::Result<()> {
+pub fn fsync(path: &Path) -> io::Result<()> {
     File::open(path)
         .map_err(|e| io::Error::new(e.kind(), format!("Failed to open the file {path:?}: {e}")))
         .and_then(|file| {
@@ -108,18 +111,19 @@ pub fn fsync(path: &Utf8Path) -> io::Result<()> {
         .map_err(|e| io::Error::new(e.kind(), format!("Failed to fsync file {path:?}: {e}")))
 }
 
-pub async fn fsync_async(path: impl AsRef<Utf8Path>) -> Result<(), std::io::Error> {
-    tokio::fs::File::open(path.as_ref()).await?.sync_all().await
+pub async fn fsync_async(path: impl AsRef<std::path::Path>) -> Result<(), std::io::Error> {
+    tokio::fs::File::open(path).await?.sync_all().await
 }
 
 #[cfg(test)]
 mod tests {
+    use tempfile::tempdir;
 
     use super::*;
 
     #[test]
     fn test_create_dir_fsyncd() {
-        let dir = camino_tempfile::tempdir().unwrap();
+        let dir = tempdir().unwrap();
 
         let existing_dir_path = dir.path();
         let err = create_dir(existing_dir_path).unwrap_err();
@@ -135,7 +139,7 @@ mod tests {
 
     #[test]
     fn test_create_dir_all_fsyncd() {
-        let dir = camino_tempfile::tempdir().unwrap();
+        let dir = tempdir().unwrap();
 
         let existing_dir_path = dir.path();
         create_dir_all(existing_dir_path).unwrap();
@@ -162,29 +166,29 @@ mod tests {
 
     #[test]
     fn test_path_with_suffix_extension() {
-        let p = Utf8PathBuf::from("/foo/bar");
+        let p = PathBuf::from("/foo/bar");
         assert_eq!(
-            &path_with_suffix_extension(p, "temp").to_string(),
+            &path_with_suffix_extension(p, "temp").to_string_lossy(),
             "/foo/bar.temp"
         );
-        let p = Utf8PathBuf::from("/foo/bar");
+        let p = PathBuf::from("/foo/bar");
         assert_eq!(
-            &path_with_suffix_extension(p, "temp.temp").to_string(),
+            &path_with_suffix_extension(p, "temp.temp").to_string_lossy(),
             "/foo/bar.temp.temp"
         );
-        let p = Utf8PathBuf::from("/foo/bar.baz");
+        let p = PathBuf::from("/foo/bar.baz");
         assert_eq!(
-            &path_with_suffix_extension(p, "temp.temp").to_string(),
+            &path_with_suffix_extension(p, "temp.temp").to_string_lossy(),
             "/foo/bar.baz.temp.temp"
         );
-        let p = Utf8PathBuf::from("/foo/bar.baz");
+        let p = PathBuf::from("/foo/bar.baz");
         assert_eq!(
-            &path_with_suffix_extension(p, ".temp").to_string(),
+            &path_with_suffix_extension(p, ".temp").to_string_lossy(),
             "/foo/bar.baz..temp"
         );
-        let p = Utf8PathBuf::from("/foo/bar/dir/");
+        let p = PathBuf::from("/foo/bar/dir/");
         assert_eq!(
-            &path_with_suffix_extension(p, ".temp").to_string(),
+            &path_with_suffix_extension(p, ".temp").to_string_lossy(),
             "/foo/bar/dir..temp"
         );
     }

libs/utils/src/fs_ext.rs

@@ -55,6 +55,8 @@ where
 
 #[cfg(test)]
 mod test {
+    use std::path::PathBuf;
+
     use crate::fs_ext::{is_directory_empty, list_dir};
 
     use super::ignore_absent_files;
@@ -63,7 +65,7 @@ mod test {
     fn is_empty_dir() {
         use super::PathExt;
 
-        let dir = camino_tempfile::tempdir().unwrap();
+        let dir = tempfile::tempdir().unwrap();
         let dir_path = dir.path();
 
         // test positive case
@@ -73,7 +75,7 @@ mod test {
         );
 
         // invoke on a file to ensure it returns an error
-        let file_path = dir_path.join("testfile");
+        let file_path: PathBuf = dir_path.join("testfile");
         let f = std::fs::File::create(&file_path).unwrap();
         drop(f);
         assert!(file_path.is_empty_dir().is_err());
@@ -85,7 +87,7 @@ mod test {
 
     #[tokio::test]
     async fn is_empty_dir_async() {
-        let dir = camino_tempfile::tempdir().unwrap();
+        let dir = tempfile::tempdir().unwrap();
         let dir_path = dir.path();
 
         // test positive case
@@ -95,7 +97,7 @@ mod test {
         );
 
         // invoke on a file to ensure it returns an error
-        let file_path = dir_path.join("testfile");
+        let file_path: PathBuf = dir_path.join("testfile");
         let f = std::fs::File::create(&file_path).unwrap();
         drop(f);
         assert!(is_directory_empty(&file_path).await.is_err());
@@ -107,9 +109,10 @@ mod test {
 
     #[test]
     fn ignore_absent_files_works() {
-        let dir = camino_tempfile::tempdir().unwrap();
+        let dir = tempfile::tempdir().unwrap();
+        let dir_path = dir.path();
 
-        let file_path = dir.path().join("testfile");
+        let file_path: PathBuf = dir_path.join("testfile");
 
         ignore_absent_files(|| std::fs::remove_file(&file_path)).expect("should execute normally");
 
@@ -123,17 +126,17 @@ mod test {
 
     #[tokio::test]
     async fn list_dir_works() {
-        let dir = camino_tempfile::tempdir().unwrap();
+        let dir = tempfile::tempdir().unwrap();
         let dir_path = dir.path();
 
         assert!(list_dir(dir_path).await.unwrap().is_empty());
 
-        let file_path = dir_path.join("testfile");
+        let file_path: PathBuf = dir_path.join("testfile");
         let _ = std::fs::File::create(&file_path).unwrap();
 
         assert_eq!(&list_dir(dir_path).await.unwrap(), &["testfile"]);
 
-        let another_dir_path = dir_path.join("testdir");
+        let another_dir_path: PathBuf = dir_path.join("testdir");
         std::fs::create_dir(another_dir_path).unwrap();
 
         let expected = &["testdir", "testfile"];

libs/utils/src/generation.rs

@@ -89,22 +89,6 @@ impl Generation {
             Self::Broken => panic!("Attempted to use a broken generation"),
         }
     }
-
-    pub fn next(&self) -> Generation {
-        match self {
-            Self::Valid(n) => Self::Valid(*n + 1),
-            Self::None => Self::Valid(1),
-            Self::Broken => panic!("Attempted to use a broken generation"),
-        }
-    }
-
-    pub fn into(self) -> Option<u32> {
-        if let Self::Valid(v) = self {
-            Some(v)
-        } else {
-            None
-        }
-    }
 }
 
 impl Serialize for Generation {

libs/utils/src/http/error.rs

@@ -24,12 +24,6 @@ pub enum ApiError {
     #[error("Precondition failed: {0}")]
     PreconditionFailed(Box<str>),
 
-    #[error("Resource temporarily unavailable: {0}")]
-    ResourceUnavailable(String),
-
-    #[error("Shutting down")]
-    ShuttingDown,
-
     #[error(transparent)]
     InternalServerError(anyhow::Error),
 }
@@ -58,14 +52,6 @@ impl ApiError {
                 self.to_string(),
                 StatusCode::PRECONDITION_FAILED,
             ),
-            ApiError::ShuttingDown => HttpErrorBody::response_from_msg_and_status(
-                "Shutting down".to_string(),
-                StatusCode::SERVICE_UNAVAILABLE,
-            ),
-            ApiError::ResourceUnavailable(err) => HttpErrorBody::response_from_msg_and_status(
-                err.to_string(),
-                StatusCode::SERVICE_UNAVAILABLE,
-            ),
             ApiError::InternalServerError(err) => HttpErrorBody::response_from_msg_and_status(
                 err.to_string(),
                 StatusCode::INTERNAL_SERVER_ERROR,

libs/utils/src/id.rs

@@ -1,3 +1,4 @@
+use std::ffi::OsStr;
 use std::{fmt, str::FromStr};
 
 use anyhow::Context;
@@ -214,11 +215,12 @@ pub struct TimelineId(Id);
 
 id_newtype!(TimelineId);
 
-impl TryFrom<Option<&str>> for TimelineId {
+impl TryFrom<Option<&OsStr>> for TimelineId {
     type Error = anyhow::Error;
 
-    fn try_from(value: Option<&str>) -> Result<Self, Self::Error> {
+    fn try_from(value: Option<&OsStr>) -> Result<Self, Self::Error> {
         value
+            .and_then(OsStr::to_str)
             .unwrap_or_default()
             .parse::<TimelineId>()
             .with_context(|| format!("Could not parse timeline id from {:?}", value))

libs/utils/src/lock_file.rs

@@ -11,10 +11,10 @@ use std::{
     io::{Read, Write},
     ops::Deref,
     os::unix::prelude::AsRawFd,
+    path::{Path, PathBuf},
 };
 
 use anyhow::Context;
-use camino::{Utf8Path, Utf8PathBuf};
 use nix::{errno::Errno::EAGAIN, fcntl};
 
 use crate::crashsafe;
@@ -23,7 +23,7 @@ use crate::crashsafe;
 /// Returned by [`create_exclusive`].
 #[must_use]
 pub struct UnwrittenLockFile {
-    path: Utf8PathBuf,
+    path: PathBuf,
     file: fs::File,
 }
 
@@ -60,7 +60,7 @@ impl UnwrittenLockFile {
 ///
 /// It is not an error if the file already exists.
 /// It is an error if the file is already locked.
-pub fn create_exclusive(lock_file_path: &Utf8Path) -> anyhow::Result<UnwrittenLockFile> {
+pub fn create_exclusive(lock_file_path: &Path) -> anyhow::Result<UnwrittenLockFile> {
     let lock_file = fs::OpenOptions::new()
         .create(true) // O_CREAT
         .write(true)
@@ -101,7 +101,7 @@ pub enum LockFileRead {
 /// Open & try to lock the lock file at the given `path`, returning a [handle][`LockFileRead`] to
 /// inspect its content. It is not an `Err(...)` if the file does not exist or is already locked.
 /// Check the [`LockFileRead`] variants for details.
-pub fn read_and_hold_lock_file(path: &Utf8Path) -> anyhow::Result<LockFileRead> {
+pub fn read_and_hold_lock_file(path: &Path) -> anyhow::Result<LockFileRead> {
     let res = fs::OpenOptions::new().read(true).open(path);
     let mut lock_file = match res {
         Ok(f) => f,

libs/utils/src/logging.rs

@@ -216,30 +216,6 @@ impl std::fmt::Debug for PrettyLocation<'_, '_> {
     }
 }
 
-/// When you will store a secret but want to make sure it won't
-/// be accidentally logged, wrap it in a SecretString, whose Debug
-/// implementation does not expose the contents.
-#[derive(Clone, Eq, PartialEq)]
-pub struct SecretString(String);
-
-impl SecretString {
-    pub fn get_contents(&self) -> &str {
-        self.0.as_str()
-    }
-}
-
-impl From<String> for SecretString {
-    fn from(s: String) -> Self {
-        Self(s)
-    }
-}
-
-impl std::fmt::Debug for SecretString {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        write!(f, "[SECRET]")
-    }
-}
-
 #[cfg(test)]
 mod tests {
     use metrics::{core::Opts, IntCounterVec};

libs/utils/src/lsn.rs

@@ -1,9 +1,9 @@
 #![warn(missing_docs)]
 
-use camino::Utf8Path;
 use serde::{Deserialize, Serialize};
 use std::fmt;
 use std::ops::{Add, AddAssign};
+use std::path::Path;
 use std::str::FromStr;
 use std::sync::atomic::{AtomicU64, Ordering};
 
@@ -44,9 +44,11 @@ impl Lsn {
     /// Parse an LSN from a filename in the form `0000000000000000`
     pub fn from_filename<F>(filename: F) -> Result<Self, LsnParseError>
     where
-        F: AsRef<Utf8Path>,
+        F: AsRef<Path>,
     {
-        Lsn::from_hex(filename.as_ref().as_str())
+        let filename: &Path = filename.as_ref();
+        let filename = filename.to_str().ok_or(LsnParseError)?;
+        Lsn::from_hex(filename)
     }
 
     /// Parse an LSN from a string in the form `0000000000000000`

libs/utils/src/pid_file.rs

@@ -49,10 +49,9 @@
 //! At this point, `B` and `C` are running, which is hazardous.
 //! Morale of the story: don't unlink pidfiles, ever.
 
-use std::ops::Deref;
+use std::{ops::Deref, path::Path};
 
 use anyhow::Context;
-use camino::Utf8Path;
 use nix::unistd::Pid;
 
 use crate::lock_file::{self, LockFileRead};
@@ -85,7 +84,7 @@ impl Deref for PidFileGuard {
 /// The claim ends as soon as the returned guard object is dropped.
 /// To maintain the claim for the remaining lifetime of the current process,
 /// use [`std::mem::forget`] or similar.
-pub fn claim_for_current_process(path: &Utf8Path) -> anyhow::Result<PidFileGuard> {
+pub fn claim_for_current_process(path: &Path) -> anyhow::Result<PidFileGuard> {
     let unwritten_lock_file = lock_file::create_exclusive(path).context("lock file")?;
     // if any of the next steps fail, we drop the file descriptor and thereby release the lock
     let guard = unwritten_lock_file
@@ -133,7 +132,7 @@ pub enum PidFileRead {
 ///
 /// On success, this function returns a [`PidFileRead`].
 /// Check its docs for a description of the meaning of its different variants.
-pub fn read(pidfile: &Utf8Path) -> anyhow::Result<PidFileRead> {
+pub fn read(pidfile: &Path) -> anyhow::Result<PidFileRead> {
     let res = lock_file::read_and_hold_lock_file(pidfile).context("read and hold pid file")?;
     let ret = match res {
         LockFileRead::NotExist => PidFileRead::NotExist,

libs/vm_monitor/src/cgroup.rs

@@ -315,8 +315,12 @@ impl CgroupWatcher {
     where
         E: Stream<Item = Sequenced<u64>>,
     {
+        // There are several actions might do when receiving a `memory.high`,
+        // such as freezing the cgroup, or increasing its `memory.high`. We don't
+        // want to do these things too often (because postgres needs to run, and
+        // we only have so much memory). These timers serve as rate limits for this.
         let mut wait_to_freeze = pin!(tokio::time::sleep(Duration::ZERO));
-        let mut last_memory_high_increase_at: Option<Instant> = None;
+        let mut wait_to_increase_memory_high = pin!(tokio::time::sleep(Duration::ZERO));
         let mut events = pin!(events);
 
         // Are we waiting to be upscaled? Could be true if we request upscale due
@@ -328,8 +332,6 @@ impl CgroupWatcher {
                 upscale = upscales.recv() => {
                     let Sequenced { seqnum, data } = upscale
                         .context("failed to listen on upscale notification channel")?;
-                    waiting_on_upscale = false;
-                    last_memory_high_increase_at = None;
                     self.last_upscale_seqnum.store(seqnum, Ordering::Release);
                     info!(cpu = data.cpu, mem_bytes = data.mem, "received upscale");
                 }
@@ -394,17 +396,12 @@ impl CgroupWatcher {
                             .send(())
                             .await
                             .context("failed to request upscale")?;
-                        waiting_on_upscale = true;
                         continue;
                     }
 
                     // Shoot, we can't freeze or and we're still waiting on upscale,
                     // increase memory.high to reduce throttling
-                    let can_increase_memory_high = match last_memory_high_increase_at {
-                        None => true,
-                        Some(t) => t.elapsed() > self.config.memory_high_increase_every,
-                    };
-                    if can_increase_memory_high {
+                    if wait_to_increase_memory_high.is_elapsed() {
                         info!(
                             "received memory.high event, \
                             but too soon to refreeze and already requested upscale \
@@ -431,20 +428,21 @@ impl CgroupWatcher {
                             .context("failed to request upscale")?;
 
                         let memory_high =
-                            self.get_memory_high_bytes().context("failed to get memory.high")?;
+                            self.get_high_bytes().context("failed to get memory.high")?;
                         let new_high = memory_high + self.config.memory_high_increase_by_bytes;
                         info!(
                             current_high_bytes = memory_high,
                             new_high_bytes = new_high,
                             "updating memory.high"
                         );
-                        self.set_memory_high_bytes(new_high)
+                        self.set_high_bytes(new_high)
                             .context("failed to set memory.high")?;
-                        last_memory_high_increase_at = Some(Instant::now());
-                        continue;
+                        wait_to_increase_memory_high
+                            .as_mut()
+                            .reset(Instant::now() + self.config.memory_high_increase_every)
                     }
 
-                    info!("received memory.high event, but can't do anything");
+                    // we can't do anything
                 }
             };
         }
@@ -556,6 +554,21 @@ impl CgroupWatcher {
     }
 }
 
+/// Represents a set of limits we apply to a cgroup to control memory usage.
+///
+/// Setting these values also affects the thresholds for receiving usage alerts.
+#[derive(Debug)]
+pub struct MemoryLimits {
+    high: u64,
+    max: u64,
+}
+
+impl MemoryLimits {
+    pub fn new(high: u64, max: u64) -> Self {
+        Self { max, high }
+    }
+}
+
 // Methods for manipulating the actual cgroup
 impl CgroupWatcher {
     /// Get a handle on the freezer subsystem.
@@ -616,29 +629,55 @@ impl CgroupWatcher {
     }
 
     /// Set cgroup memory.high threshold.
-    pub fn set_memory_high_bytes(&self, bytes: u64) -> anyhow::Result<()> {
-        self.set_memory_high_internal(MaxValue::Value(u64::min(bytes, i64::MAX as u64) as i64))
-    }
-
-    /// Set the cgroup's memory.high to 'max', disabling it.
-    pub fn unset_memory_high(&self) -> anyhow::Result<()> {
-        self.set_memory_high_internal(MaxValue::Max)
-    }
-
-    fn set_memory_high_internal(&self, value: MaxValue) -> anyhow::Result<()> {
+    pub fn set_high_bytes(&self, bytes: u64) -> anyhow::Result<()> {
         self.memory()
             .context("failed to get memory subsystem")?
             .set_mem(cgroups_rs::memory::SetMemory {
                 low: None,
-                high: Some(value),
+                high: Some(MaxValue::Value(u64::min(bytes, i64::MAX as u64) as i64)),
                 min: None,
                 max: None,
             })
-            .map_err(anyhow::Error::from)
+            .context("failed to set memory.high")
+    }
+
+    /// Set cgroup memory.high and memory.max.
+    pub fn set_limits(&self, limits: &MemoryLimits) -> anyhow::Result<()> {
+        info!(
+            limits.high,
+            limits.max,
+            path = self.path(),
+            "writing new memory limits",
+        );
+        self.memory()
+            .context("failed to get memory subsystem while setting memory limits")?
+            .set_mem(cgroups_rs::memory::SetMemory {
+                min: None,
+                low: None,
+                high: Some(MaxValue::Value(
+                    u64::min(limits.high, i64::MAX as u64) as i64
+                )),
+                max: Some(MaxValue::Value(u64::min(limits.max, i64::MAX as u64) as i64)),
+            })
+            .context("failed to set memory limits")
+    }
+
+    /// Given some amount of available memory, set the desired cgroup memory limits
+    pub fn set_memory_limits(&mut self, available_memory: u64) -> anyhow::Result<()> {
+        let new_high = self.config.calculate_memory_high_value(available_memory);
+        let limits = MemoryLimits::new(new_high, available_memory);
+        info!(
+            path = self.path(),
+            memory = ?limits,
+            "setting cgroup memory",
+        );
+        self.set_limits(&limits)
+            .context("failed to set cgroup memory limits")?;
+        Ok(())
     }
 
     /// Get memory.high threshold.
-    pub fn get_memory_high_bytes(&self) -> anyhow::Result<u64> {
+    pub fn get_high_bytes(&self) -> anyhow::Result<u64> {
         let high = self
             .memory()
             .context("failed to get memory subsystem while getting memory statistics")?

libs/vm_monitor/src/lib.rs

@@ -178,17 +178,14 @@ pub async fn ws_handler(
 
 /// Starts the monitor. If startup fails or the monitor exits, an error will
 /// be logged and our internal state will be reset to allow for new connections.
-#[tracing::instrument(skip_all)]
+#[tracing::instrument(skip_all, fields(?args))]
 async fn start_monitor(
     ws: WebSocket,
     args: &Args,
     kill: broadcast::Receiver<()>,
     token: CancellationToken,
 ) {
-    info!(
-        ?args,
-        "accepted new websocket connection -> starting monitor"
-    );
+    info!("accepted new websocket connection -> starting monitor");
     let timeout = Duration::from_secs(4);
     let monitor = tokio::time::timeout(
         timeout,

libs/vm_monitor/src/runner.rs

@@ -4,9 +4,8 @@
 //! This is the "Monitor" part of the monitor binary and is the main entrypoint for
 //! all functionality.
 
-use std::fmt::Debug;
 use std::sync::Arc;
-use std::time::{Duration, Instant};
+use std::{fmt::Debug, mem};
 
 use anyhow::{bail, Context};
 use axum::extract::ws::{Message, WebSocket};
@@ -16,7 +15,7 @@ use tokio::sync::mpsc;
 use tokio_util::sync::CancellationToken;
 use tracing::{error, info, warn};
 
-use crate::cgroup::{CgroupWatcher, Sequenced};
+use crate::cgroup::{CgroupWatcher, MemoryLimits, Sequenced};
 use crate::dispatcher::Dispatcher;
 use crate::filecache::{FileCacheConfig, FileCacheState};
 use crate::protocol::{InboundMsg, InboundMsgKind, OutboundMsg, OutboundMsgKind, Resources};
@@ -37,8 +36,6 @@ pub struct Runner {
     /// by us vs the autoscaler-agent.
     counter: usize,
 
-    last_upscale_request_at: Option<Instant>,
-
     /// A signal to kill the main thread produced by `self.run()`. This is triggered
     /// when the server receives a new connection. When the thread receives the
     /// signal off this channel, it will gracefully shutdown.
@@ -102,47 +99,9 @@ impl Runner {
             cgroup: None,
             dispatcher,
             counter: 1, // NB: must be odd, see the comment about the field for more.
-            last_upscale_request_at: None,
             kill,
         };
 
-        // If we have both the cgroup and file cache integrations enabled, it's possible for
-        // temporary failures to result in cgroup throttling (from memory.high), that in turn makes
-        // it near-impossible to connect to the file cache (because it times out). Unfortunately,
-        // we *do* still want to determine the file cache size before setting the cgroup's
-        // memory.high, so it's not as simple as just swapping the order.
-        //
-        // Instead, the resolution here is that on vm-monitor startup (note: happens on each
-        // connection from autoscaler-agent, possibly multiple times per compute_ctl lifecycle), we
-        // temporarily unset memory.high, to allow any existing throttling to dissipate. It's a bit
-        // of a hacky solution, but helps with reliability.
-        if let Some(name) = &args.cgroup {
-            // Best not to set up cgroup stuff more than once, so we'll initialize cgroup state
-            // now, and then set limits later.
-            info!("initializing cgroup");
-
-            let (cgroup, cgroup_event_stream) = CgroupWatcher::new(name.clone(), requesting_send)
-                .context("failed to create cgroup manager")?;
-
-            info!("temporarily unsetting memory.high");
-
-            // Temporarily un-set cgroup memory.high; see above.
-            cgroup
-                .unset_memory_high()
-                .context("failed to unset memory.high")?;
-
-            let cgroup = Arc::new(cgroup);
-
-            let cgroup_clone = Arc::clone(&cgroup);
-            spawn_with_cancel(
-                token.clone(),
-                |_| error!("cgroup watcher terminated"),
-                async move { cgroup_clone.watch(notified_recv, cgroup_event_stream).await },
-            );
-
-            state.cgroup = Some(cgroup);
-        }
-
         let mut file_cache_reserved_bytes = 0;
         let mem = get_total_system_memory();
 
@@ -156,7 +115,7 @@ impl Runner {
                 false => FileCacheConfig::default_in_memory(),
             };
 
-            let mut file_cache = FileCacheState::new(connstr, config, token)
+            let mut file_cache = FileCacheState::new(connstr, config, token.clone())
                 .await
                 .context("failed to create file cache")?;
 
@@ -189,15 +148,35 @@ impl Runner {
             state.filecache = Some(file_cache);
         }
 
-        if let Some(cgroup) = &state.cgroup {
-            let available = mem - file_cache_reserved_bytes;
-            let value = cgroup.config.calculate_memory_high_value(available);
+        if let Some(name) = &args.cgroup {
+            let (mut cgroup, cgroup_event_stream) =
+                CgroupWatcher::new(name.clone(), requesting_send)
+                    .context("failed to create cgroup manager")?;
 
-            info!(value, "setting memory.high");
+            let available = mem - file_cache_reserved_bytes;
 
             cgroup
-                .set_memory_high_bytes(value)
-                .context("failed to set cgroup memory.high")?;
+                .set_memory_limits(available)
+                .context("failed to set cgroup memory limits")?;
+
+            let cgroup = Arc::new(cgroup);
+
+            // Some might call this . . . cgroup v2
+            let cgroup_clone = Arc::clone(&cgroup);
+
+            spawn_with_cancel(token, |_| error!("cgroup watcher terminated"), async move {
+                cgroup_clone.watch(notified_recv, cgroup_event_stream).await
+            });
+
+            state.cgroup = Some(cgroup);
+        } else {
+            // *NOTE*: We need to forget the sender so that its drop impl does not get ran.
+            // This allows us to poll it in `Monitor::run` regardless of whether we
+            // are managing a cgroup or not. If we don't forget it, all receives will
+            // immediately return an error because the sender is droped and it will
+            // claim all select! statements, effectively turning `Monitor::run` into
+            // `loop { fail to receive }`.
+            mem::forget(requesting_send);
         }
 
         Ok(state)
@@ -274,11 +253,15 @@ impl Runner {
                 new_cgroup_mem_high = cgroup.config.calculate_memory_high_value(available_memory);
             }
 
-            // new_cgroup_mem_high is initialized to 0 but it is guaranteed to not be here
-            // since it is properly initialized in the previous cgroup if let block
+            let limits = MemoryLimits::new(
+                // new_cgroup_mem_high is initialized to 0 but it is guarancontextd to not be here
+                // since it is properly initialized in the previous cgroup if let block
+                new_cgroup_mem_high,
+                available_memory,
+            );
             cgroup
-                .set_memory_high_bytes(new_cgroup_mem_high)
-                .context("failed to set cgroup memory.high")?;
+                .set_limits(&limits)
+                .context("failed to set cgroup memory limits")?;
 
             let message = format!(
                 "set cgroup memory.high to {} MiB, of new max {} MiB",
@@ -341,9 +324,10 @@ impl Runner {
                 name = cgroup.path(),
                 "updating cgroup memory.high",
             );
+            let limits = MemoryLimits::new(new_cgroup_mem_high, available_memory);
             cgroup
-                .set_memory_high_bytes(new_cgroup_mem_high)
-                .context("failed to set cgroup memory.high")?;
+                .set_limits(&limits)
+                .context("failed to set file cache size")?;
         }
 
         Ok(())
@@ -409,24 +393,10 @@ impl Runner {
                     }
                 }
                 // we need to propagate an upscale request
-                request = self.dispatcher.request_upscale_events.recv(), if self.cgroup.is_some() => {
+                request = self.dispatcher.request_upscale_events.recv() => {
                     if request.is_none() {
                         bail!("failed to listen for upscale event from cgroup")
                     }
-
-                    // If it's been less than 1 second since the last time we requested upscaling,
-                    // ignore the event, to avoid spamming the agent (otherwise, this can happen
-                    // ~1k times per second).
-                    if let Some(t) = self.last_upscale_request_at {
-                        let elapsed = t.elapsed();
-                        if elapsed < Duration::from_secs(1) {
-                            info!(elapsed_millis = elapsed.as_millis(), "cgroup asked for upscale but too soon to forward the request, ignoring");
-                            continue;
-                        }
-                    }
-
-                    self.last_upscale_request_at = Some(Instant::now());
-
                     info!("cgroup asking for upscale; forwarding request");
                     self.counter += 2; // Increment, preserving parity (i.e. keep the
                                        // counter odd). See the field comment for more.

pageserver/Cargo.toml

@@ -17,8 +17,6 @@ async-stream.workspace = true
 async-trait.workspace = true
 byteorder.workspace = true
 bytes.workspace = true
-camino.workspace = true
-camino-tempfile.workspace = true
 chrono = { workspace = true, features = ["serde"] }
 clap = { workspace = true, features = ["string"] }
 close_fds.workspace = true
@@ -86,6 +84,7 @@ strum_macros.workspace = true
 [dev-dependencies]
 criterion.workspace = true
 hex-literal.workspace = true
+tempfile.workspace = true
 tokio = { workspace = true, features = ["process", "sync", "fs", "rt", "io-util", "time", "test-util"] }
 
 [[bench]]

pageserver/benches/bench_walredo.rs

@@ -25,7 +25,7 @@ fn redo_scenarios(c: &mut Criterion) {
     // input to the stderr.
     // utils::logging::init(utils::logging::LogFormat::Plain).unwrap();
 
-    let repo_dir = camino_tempfile::tempdir_in(env!("CARGO_TARGET_TMPDIR")).unwrap();
+    let repo_dir = tempfile::tempdir_in(env!("CARGO_TARGET_TMPDIR")).unwrap();
 
     let conf = PageServerConf::dummy_conf(repo_dir.path().to_path_buf());
     let conf = Box::leak(Box::new(conf));

pageserver/ctl/Cargo.toml

@@ -9,7 +9,6 @@ license.workspace = true
 [dependencies]
 anyhow.workspace = true
 bytes.workspace = true
-camino.workspace = true
 clap = { workspace = true, features = ["string"] }
 git-version.workspace = true
 pageserver = { path = ".." }

pageserver/ctl/src/layer_map_analyzer.rs

@@ -3,14 +3,11 @@
 //! Currently it only analyzes holes, which are regions within the layer range that the layer contains no updates for. In the future it might do more analysis (maybe key quantiles?) but it should never return sensitive data.
 
 use anyhow::Result;
-use camino::{Utf8Path, Utf8PathBuf};
-use pageserver::context::{DownloadBehavior, RequestContext};
-use pageserver::task_mgr::TaskKind;
 use pageserver::tenant::{TENANTS_SEGMENT_NAME, TIMELINES_SEGMENT_NAME};
 use std::cmp::Ordering;
 use std::collections::BinaryHeap;
 use std::ops::Range;
-use std::{fs, str};
+use std::{fs, path::Path, str};
 
 use pageserver::page_cache::PAGE_SZ;
 use pageserver::repository::{Key, KEY_SIZE};
@@ -99,9 +96,9 @@ pub(crate) fn parse_filename(name: &str) -> Option<LayerFile> {
 }
 
 // Finds the max_holes largest holes, ignoring any that are smaller than MIN_HOLE_LENGTH"
-async fn get_holes(path: &Utf8Path, max_holes: usize, ctx: &RequestContext) -> Result<Vec<Hole>> {
+async fn get_holes(path: &Path, max_holes: usize) -> Result<Vec<Hole>> {
     let file = FileBlockReader::new(VirtualFile::open(path).await?);
-    let summary_blk = file.read_blk(0, ctx).await?;
+    let summary_blk = file.read_blk(0).await?;
     let actual_summary = Summary::des_prefix(summary_blk.as_ref())?;
     let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
         actual_summary.index_start_blk,
@@ -128,7 +125,6 @@ async fn get_holes(path: &Utf8Path, max_holes: usize, ctx: &RequestContext) -> R
                 prev_key = Some(curr.next());
                 true
             },
-            ctx,
         )
         .await?;
     let mut holes = heap.into_vec();
@@ -139,7 +135,6 @@ async fn get_holes(path: &Utf8Path, max_holes: usize, ctx: &RequestContext) -> R
 pub(crate) async fn main(cmd: &AnalyzeLayerMapCmd) -> Result<()> {
     let storage_path = &cmd.path;
     let max_holes = cmd.max_holes.unwrap_or(DEFAULT_MAX_HOLES);
-    let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);
 
     // Initialize virtual_file (file desriptor cache) and page cache which are needed to access layer persistent B-Tree.
     pageserver::virtual_file::init(10);
@@ -168,9 +163,7 @@ pub(crate) async fn main(cmd: &AnalyzeLayerMapCmd) -> Result<()> {
                     parse_filename(&layer.file_name().into_string().unwrap())
                 {
                     if layer_file.is_delta {
-                        let layer_path =
-                            Utf8PathBuf::from_path_buf(layer.path()).expect("non-Unicode path");
-                        layer_file.holes = get_holes(&layer_path, max_holes, &ctx).await?;
+                        layer_file.holes = get_holes(&layer.path(), max_holes).await?;
                         n_deltas += 1;
                     }
                     layers.push(layer_file);

pageserver/ctl/src/layers.rs

@@ -1,10 +1,7 @@
 use std::path::{Path, PathBuf};
 
 use anyhow::Result;
-use camino::Utf8Path;
 use clap::Subcommand;
-use pageserver::context::{DownloadBehavior, RequestContext};
-use pageserver::task_mgr::TaskKind;
 use pageserver::tenant::block_io::BlockCursor;
 use pageserver::tenant::disk_btree::DiskBtreeReader;
 use pageserver::tenant::storage_layer::delta_layer::{BlobRef, Summary};
@@ -47,12 +44,12 @@ pub(crate) enum LayerCmd {
     },
 }
 
-async fn read_delta_file(path: impl AsRef<Path>, ctx: &RequestContext) -> Result<()> {
-    let path = Utf8Path::from_path(path.as_ref()).expect("non-Unicode path");
+async fn read_delta_file(path: impl AsRef<Path>) -> Result<()> {
+    let path = path.as_ref();
     virtual_file::init(10);
     page_cache::init(100);
     let file = FileBlockReader::new(VirtualFile::open(path).await?);
-    let summary_blk = file.read_blk(0, ctx).await?;
+    let summary_blk = file.read_blk(0).await?;
     let actual_summary = Summary::des_prefix(summary_blk.as_ref())?;
     let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
         actual_summary.index_start_blk,
@@ -70,12 +67,11 @@ async fn read_delta_file(path: impl AsRef<Path>, ctx: &RequestContext) -> Result
                 all.push((curr, BlobRef(value_offset)));
                 true
             },
-            ctx,
         )
         .await?;
     let cursor = BlockCursor::new_fileblockreader(&file);
     for (k, v) in all {
-        let value = cursor.read_blob(v.pos(), ctx).await?;
+        let value = cursor.read_blob(v.pos()).await?;
         println!("key:{} value_len:{}", k, value.len());
     }
     // TODO(chi): special handling for last key?
@@ -83,7 +79,6 @@ async fn read_delta_file(path: impl AsRef<Path>, ctx: &RequestContext) -> Result
 }
 
 pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
-    let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);
     match cmd {
         LayerCmd::List { path } => {
             for tenant in fs::read_dir(path.join(TENANTS_SEGMENT_NAME))? {
@@ -158,7 +153,7 @@ pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
                         );
 
                         if layer_file.is_delta {
-                            read_delta_file(layer.path(), &ctx).await?;
+                            read_delta_file(layer.path()).await?;
                         } else {
                             anyhow::bail!("not supported yet :(");
                         }

pageserver/ctl/src/main.rs

@@ -8,7 +8,6 @@ mod draw_timeline_dir;
 mod layer_map_analyzer;
 mod layers;
 
-use camino::{Utf8Path, Utf8PathBuf};
 use clap::{Parser, Subcommand};
 use layers::LayerCmd;
 use pageserver::{
@@ -19,6 +18,7 @@ use pageserver::{
     virtual_file,
 };
 use postgres_ffi::ControlFileData;
+use std::path::{Path, PathBuf};
 use utils::{lsn::Lsn, project_git_version};
 
 project_git_version!(GIT_VERSION);
@@ -49,7 +49,7 @@ enum Commands {
 #[derive(Parser)]
 struct MetadataCmd {
     /// Input metadata file path
-    metadata_path: Utf8PathBuf,
+    metadata_path: PathBuf,
     /// Replace disk consistent Lsn
     disk_consistent_lsn: Option<Lsn>,
     /// Replace previous record Lsn
@@ -61,13 +61,13 @@ struct MetadataCmd {
 #[derive(Parser)]
 struct PrintLayerFileCmd {
     /// Pageserver data path
-    path: Utf8PathBuf,
+    path: PathBuf,
 }
 
 #[derive(Parser)]
 struct AnalyzeLayerMapCmd {
     /// Pageserver data path
-    path: Utf8PathBuf,
+    path: PathBuf,
     /// Max holes
     max_holes: Option<usize>,
 }
@@ -102,7 +102,7 @@ async fn main() -> anyhow::Result<()> {
     Ok(())
 }
 
-fn read_pg_control_file(control_file_path: &Utf8Path) -> anyhow::Result<()> {
+fn read_pg_control_file(control_file_path: &Path) -> anyhow::Result<()> {
     let control_file = ControlFileData::decode(&std::fs::read(control_file_path)?)?;
     println!("{control_file:?}");
     let control_file_initdb = Lsn(control_file.checkPoint);
@@ -114,7 +114,7 @@ fn read_pg_control_file(control_file_path: &Utf8Path) -> anyhow::Result<()> {
     Ok(())
 }
 
-async fn print_layerfile(path: &Utf8Path) -> anyhow::Result<()> {
+async fn print_layerfile(path: &Path) -> anyhow::Result<()> {
     // Basic initialization of things that don't change after startup
     virtual_file::init(10);
     page_cache::init(100);

pageserver/src/basebackup.rs

@@ -25,7 +25,6 @@ use crate::context::RequestContext;
 use crate::tenant::Timeline;
 use pageserver_api::reltag::{RelTag, SlruKind};
 
-use postgres_ffi::dispatch_pgversion;
 use postgres_ffi::pg_constants::{DEFAULTTABLESPACE_OID, GLOBALTABLESPACE_OID};
 use postgres_ffi::pg_constants::{PGDATA_SPECIAL_FILES, PGDATA_SUBDIRS, PG_HBA};
 use postgres_ffi::relfile_utils::{INIT_FORKNUM, MAIN_FORKNUM};
@@ -324,25 +323,14 @@ where
                 .timeline
                 .get_relmap_file(spcnode, dbnode, self.lsn, self.ctx)
                 .await?;
-
-            ensure!(
-                img.len()
-                    == dispatch_pgversion!(
-                        self.timeline.pg_version,
-                        pgv::bindings::SIZEOF_RELMAPFILE
-                    )
-            );
-
+            ensure!(img.len() == 512);
             Some(img)
         } else {
             None
         };
 
         if spcnode == GLOBALTABLESPACE_OID {
-            let pg_version_str = match self.timeline.pg_version {
-                14 | 15 => self.timeline.pg_version.to_string(),
-                ver => format!("{ver}\x0A"),
-            };
+            let pg_version_str = self.timeline.pg_version.to_string();
             let header = new_tar_header("PG_VERSION", pg_version_str.len() as u64)?;
             self.ar.append(&header, pg_version_str.as_bytes()).await?;
 
@@ -386,10 +374,7 @@ where
             if let Some(img) = relmap_img {
                 let dst_path = format!("base/{}/PG_VERSION", dbnode);
 
-                let pg_version_str = match self.timeline.pg_version {
-                    14 | 15 => self.timeline.pg_version.to_string(),
-                    ver => format!("{ver}\x0A"),
-                };
+                let pg_version_str = self.timeline.pg_version.to_string();
                 let header = new_tar_header(&dst_path, pg_version_str.len() as u64)?;
                 self.ar.append(&header, pg_version_str.as_bytes()).await?;
 

pageserver/src/bin/pageserver.rs

@@ -2,14 +2,12 @@
 
 use std::env::{var, VarError};
 use std::sync::Arc;
-use std::{env, ops::ControlFlow, str::FromStr};
+use std::{env, ops::ControlFlow, path::Path, str::FromStr};
 
 use anyhow::{anyhow, Context};
-use camino::Utf8Path;
 use clap::{Arg, ArgAction, Command};
 
 use metrics::launch_timestamp::{set_launch_timestamp_metric, LaunchTimestamp};
-use pageserver::control_plane_client::ControlPlaneClient;
 use pageserver::disk_usage_eviction_task::{self, launch_disk_usage_global_eviction_task};
 use pageserver::metrics::{STARTUP_DURATION, STARTUP_IS_LOADING};
 use pageserver::task_mgr::WALRECEIVER_RUNTIME;
@@ -22,7 +20,6 @@ use metrics::set_build_info_metric;
 use pageserver::{
     config::{defaults::*, PageServerConf},
     context::{DownloadBehavior, RequestContext},
-    deletion_queue::DeletionQueue,
     http, page_cache, page_service, task_mgr,
     task_mgr::TaskKind,
     task_mgr::{BACKGROUND_RUNTIME, COMPUTE_REQUEST_RUNTIME, MGMT_REQUEST_RUNTIME},
@@ -66,17 +63,21 @@ fn main() -> anyhow::Result<()> {
 
     let workdir = arg_matches
         .get_one::<String>("workdir")
-        .map(Utf8Path::new)
-        .unwrap_or_else(|| Utf8Path::new(".neon"));
+        .map(Path::new)
+        .unwrap_or_else(|| Path::new(".neon"));
     let workdir = workdir
-        .canonicalize_utf8()
-        .with_context(|| format!("Error opening workdir '{workdir}'"))?;
+        .canonicalize()
+        .with_context(|| format!("Error opening workdir '{}'", workdir.display()))?;
 
     let cfg_file_path = workdir.join("pageserver.toml");
 
     // Set CWD to workdir for non-daemon modes
-    env::set_current_dir(&workdir)
-        .with_context(|| format!("Failed to set application's current dir to '{workdir}'"))?;
+    env::set_current_dir(&workdir).with_context(|| {
+        format!(
+            "Failed to set application's current dir to '{}'",
+            workdir.display()
+        )
+    })?;
 
     let conf = match initialize_config(&cfg_file_path, arg_matches, &workdir)? {
         ControlFlow::Continue(conf) => conf,
@@ -112,8 +113,12 @@ fn main() -> anyhow::Result<()> {
 
     let tenants_path = conf.tenants_path();
     if !tenants_path.exists() {
-        utils::crashsafe::create_dir_all(conf.tenants_path())
-            .with_context(|| format!("Failed to create tenants root dir at '{tenants_path}'"))?;
+        utils::crashsafe::create_dir_all(conf.tenants_path()).with_context(|| {
+            format!(
+                "Failed to create tenants root dir at '{}'",
+                tenants_path.display()
+            )
+        })?;
     }
 
     // Initialize up failpoints support
@@ -130,9 +135,9 @@ fn main() -> anyhow::Result<()> {
 }
 
 fn initialize_config(
-    cfg_file_path: &Utf8Path,
+    cfg_file_path: &Path,
     arg_matches: clap::ArgMatches,
-    workdir: &Utf8Path,
+    workdir: &Path,
 ) -> anyhow::Result<ControlFlow<(), &'static PageServerConf>> {
     let init = arg_matches.get_flag("init");
     let update_config = init || arg_matches.get_flag("update-config");
@@ -140,22 +145,33 @@ fn initialize_config(
     let (mut toml, config_file_exists) = if cfg_file_path.is_file() {
         if init {
             anyhow::bail!(
-                "Config file '{cfg_file_path}' already exists, cannot init it, use --update-config to update it",
+                "Config file '{}' already exists, cannot init it, use --update-config to update it",
+                cfg_file_path.display()
             );
         }
         // Supplement the CLI arguments with the config file
-        let cfg_file_contents = std::fs::read_to_string(cfg_file_path)
-            .with_context(|| format!("Failed to read pageserver config at '{cfg_file_path}'"))?;
+        let cfg_file_contents = std::fs::read_to_string(cfg_file_path).with_context(|| {
+            format!(
+                "Failed to read pageserver config at '{}'",
+                cfg_file_path.display()
+            )
+        })?;
         (
             cfg_file_contents
                 .parse::<toml_edit::Document>()
                 .with_context(|| {
-                    format!("Failed to parse '{cfg_file_path}' as pageserver config")
+                    format!(
+                        "Failed to parse '{}' as pageserver config",
+                        cfg_file_path.display()
+                    )
                 })?,
             true,
         )
     } else if cfg_file_path.exists() {
-        anyhow::bail!("Config file '{cfg_file_path}' exists but is not a regular file");
+        anyhow::bail!(
+            "Config file '{}' exists but is not a regular file",
+            cfg_file_path.display()
+        );
     } else {
         // We're initializing the tenant, so there's no config file yet
         (
@@ -174,7 +190,7 @@ fn initialize_config(
 
             for (key, item) in doc.iter() {
                 if config_file_exists && update_config && key == "id" && toml.contains_key(key) {
-                    anyhow::bail!("Pageserver config file exists at '{cfg_file_path}' and has node id already, it cannot be overridden");
+                    anyhow::bail!("Pageserver config file exists at '{}' and has node id already, it cannot be overridden", cfg_file_path.display());
                 }
                 toml.insert(key, item.clone());
             }
@@ -186,11 +202,18 @@ fn initialize_config(
         .context("Failed to parse pageserver configuration")?;
 
     if update_config {
-        info!("Writing pageserver config to '{cfg_file_path}'");
+        info!("Writing pageserver config to '{}'", cfg_file_path.display());
 
-        std::fs::write(cfg_file_path, toml.to_string())
-            .with_context(|| format!("Failed to write pageserver config to '{cfg_file_path}'"))?;
-        info!("Config successfully written to '{cfg_file_path}'")
+        std::fs::write(cfg_file_path, toml.to_string()).with_context(|| {
+            format!(
+                "Failed to write pageserver config to '{}'",
+                cfg_file_path.display()
+            )
+        })?;
+        info!(
+            "Config successfully written to '{}'",
+            cfg_file_path.display()
+        )
     }
 
     Ok(if init {
@@ -323,22 +346,9 @@ fn start_pageserver(
         }
     };
 
-    // Top-level cancellation token for the process
-    let shutdown_pageserver = tokio_util::sync::CancellationToken::new();
-
     // Set up remote storage client
     let remote_storage = create_remote_storage_client(conf)?;
 
-    // Set up deletion queue
-    let (deletion_queue, deletion_workers) = DeletionQueue::new(
-        remote_storage.clone(),
-        ControlPlaneClient::new(conf, &shutdown_pageserver),
-        conf,
-    );
-    if let Some(deletion_workers) = deletion_workers {
-        deletion_workers.spawn_with(BACKGROUND_RUNTIME.handle());
-    }
-
     // Up to this point no significant I/O has been done: this should have been fast.  Record
     // duration prior to starting I/O intensive phase of startup.
     startup_checkpoint("initial", "Starting loading tenants");
@@ -369,13 +379,13 @@ fn start_pageserver(
     };
 
     // Scan the local 'tenants/' directory and start loading the tenants
-    let deletion_queue_client = deletion_queue.new_client();
+    let shutdown_pageserver = tokio_util::sync::CancellationToken::new();
+
     BACKGROUND_RUNTIME.block_on(mgr::init_tenant_mgr(
         conf,
         TenantSharedResources {
             broker_client: broker_client.clone(),
             remote_storage: remote_storage.clone(),
-            deletion_queue_client,
         },
         order,
         shutdown_pageserver.clone(),
@@ -467,20 +477,16 @@ fn start_pageserver(
     {
         let _rt_guard = MGMT_REQUEST_RUNTIME.enter();
 
-        let router_state = Arc::new(
-            http::routes::State::new(
-                conf,
-                http_auth.clone(),
-                remote_storage.clone(),
-                broker_client.clone(),
-                disk_usage_eviction_state,
-                deletion_queue.new_client(),
-            )
-            .context("Failed to initialize router state")?,
-        );
-        let router = http::make_router(router_state, launch_ts, http_auth.clone())?
-            .build()
-            .map_err(|err| anyhow!(err))?;
+        let router = http::make_router(
+            conf,
+            launch_ts,
+            http_auth,
+            broker_client.clone(),
+            remote_storage,
+            disk_usage_eviction_state,
+        )?
+        .build()
+        .map_err(|err| anyhow!(err))?;
         let service = utils::http::RouterService::new(router).unwrap();
         let server = hyper::Server::from_tcp(http_listener)?
             .serve(service)
@@ -509,9 +515,6 @@ fn start_pageserver(
             // creates a child context with the right DownloadBehavior.
             DownloadBehavior::Error,
         );
-
-        let local_disk_storage = conf.workdir.join("last_consumption_metrics.json");
-
         task_mgr::spawn(
             crate::BACKGROUND_RUNTIME.handle(),
             TaskKind::MetricsCollection,
@@ -538,7 +541,6 @@ fn start_pageserver(
                     conf.cached_metric_collection_interval,
                     conf.synthetic_size_calculation_interval,
                     conf.id,
-                    local_disk_storage,
                     metrics_ctx,
                 )
                 .instrument(info_span!("metrics_collection"))
@@ -602,12 +604,7 @@ fn start_pageserver(
             // Right now that tree doesn't reach very far, and `task_mgr` is used instead.
             // The plan is to change that over time.
             shutdown_pageserver.take();
-            let bg_remote_storage = remote_storage.clone();
-            let bg_deletion_queue = deletion_queue.clone();
-            BACKGROUND_RUNTIME.block_on(pageserver::shutdown_pageserver(
-                bg_remote_storage.map(|_| bg_deletion_queue),
-                0,
-            ));
+            BACKGROUND_RUNTIME.block_on(pageserver::shutdown_pageserver(0));
             unreachable!()
         }
     })
@@ -619,7 +616,7 @@ fn create_remote_storage_client(
     let config = if let Some(config) = &conf.remote_storage_config {
         config
     } else {
-        tracing::warn!("no remote storage configured, this is a deprecated configuration");
+        // No remote storage configured.
         return Ok(None);
     };
 

pageserver/src/config.rs

@@ -11,18 +11,17 @@ use std::env;
 use storage_broker::Uri;
 use utils::crashsafe::path_with_suffix_extension;
 use utils::id::ConnectionId;
-use utils::logging::SecretString;
 
 use once_cell::sync::OnceCell;
 use reqwest::Url;
 use std::num::NonZeroUsize;
+use std::path::{Path, PathBuf};
 use std::str::FromStr;
 use std::sync::Arc;
 use std::time::Duration;
 use toml_edit;
 use toml_edit::{Document, Item};
 
-use camino::{Utf8Path, Utf8PathBuf};
 use postgres_backend::AuthType;
 use utils::{
     id::{NodeId, TenantId, TimelineId},
@@ -65,7 +64,7 @@ pub mod defaults {
         super::ConfigurableSemaphore::DEFAULT_INITIAL.get();
 
     pub const DEFAULT_METRIC_COLLECTION_INTERVAL: &str = "10 min";
-    pub const DEFAULT_CACHED_METRIC_COLLECTION_INTERVAL: &str = "0s";
+    pub const DEFAULT_CACHED_METRIC_COLLECTION_INTERVAL: &str = "1 hour";
     pub const DEFAULT_METRIC_COLLECTION_ENDPOINT: Option<reqwest::Url> = None;
     pub const DEFAULT_SYNTHETIC_SIZE_CALCULATION_INTERVAL: &str = "10 min";
     pub const DEFAULT_BACKGROUND_TASK_MAXIMUM_DELAY: &str = "10s";
@@ -74,7 +73,7 @@ pub mod defaults {
     /// Default built-in configuration file.
     ///
     pub const DEFAULT_CONFIG_FILE: &str = formatcp!(
-        r#"
+        r###"
 # Initial configuration file created by 'pageserver --init'
 #listen_pg_addr = '{DEFAULT_PG_LISTEN_ADDR}'
 #listen_http_addr = '{DEFAULT_HTTP_LISTEN_ADDR}'
@@ -119,7 +118,7 @@ pub mod defaults {
 
 [remote_storage]
 
-"#
+"###
     );
 }
 
@@ -153,9 +152,9 @@ pub struct PageServerConf {
     // that during unit testing, because the current directory is global
     // to the process but different unit tests work on different
     // repositories.
-    pub workdir: Utf8PathBuf,
+    pub workdir: PathBuf,
 
-    pub pg_distrib_dir: Utf8PathBuf,
+    pub pg_distrib_dir: PathBuf,
 
     // Authentication
     /// authentication method for the HTTP mgmt API
@@ -164,7 +163,7 @@ pub struct PageServerConf {
     pub pg_auth_type: AuthType,
     /// Path to a file containing public key for verifying JWT tokens.
     /// Used for both mgmt and compute auth, if enabled.
-    pub auth_validation_public_key_path: Option<Utf8PathBuf>,
+    pub auth_validation_public_key_path: Option<PathBuf>,
 
     pub remote_storage_config: Option<RemoteStorageConfig>,
 
@@ -208,9 +207,6 @@ pub struct PageServerConf {
     pub background_task_maximum_delay: Duration,
 
     pub control_plane_api: Option<Url>,
-
-    /// JWT token for use with the control plane API.
-    pub control_plane_api_token: Option<SecretString>,
 }
 
 /// We do not want to store this in a PageServerConf because the latter may be logged
@@ -253,15 +249,15 @@ struct PageServerConfigBuilder {
     page_cache_size: BuilderValue<usize>,
     max_file_descriptors: BuilderValue<usize>,
 
-    workdir: BuilderValue<Utf8PathBuf>,
+    workdir: BuilderValue<PathBuf>,
 
-    pg_distrib_dir: BuilderValue<Utf8PathBuf>,
+    pg_distrib_dir: BuilderValue<PathBuf>,
 
     http_auth_type: BuilderValue<AuthType>,
     pg_auth_type: BuilderValue<AuthType>,
 
     //
-    auth_validation_public_key_path: BuilderValue<Option<Utf8PathBuf>>,
+    auth_validation_public_key_path: BuilderValue<Option<PathBuf>>,
     remote_storage_config: BuilderValue<Option<RemoteStorageConfig>>,
 
     id: BuilderValue<NodeId>,
@@ -287,7 +283,6 @@ struct PageServerConfigBuilder {
     background_task_maximum_delay: BuilderValue<Duration>,
 
     control_plane_api: BuilderValue<Option<Url>>,
-    control_plane_api_token: BuilderValue<Option<SecretString>>,
 }
 
 impl Default for PageServerConfigBuilder {
@@ -305,12 +300,10 @@ impl Default for PageServerConfigBuilder {
             superuser: Set(DEFAULT_SUPERUSER.to_string()),
             page_cache_size: Set(DEFAULT_PAGE_CACHE_SIZE),
             max_file_descriptors: Set(DEFAULT_MAX_FILE_DESCRIPTORS),
-            workdir: Set(Utf8PathBuf::new()),
-            pg_distrib_dir: Set(Utf8PathBuf::from_path_buf(
-                env::current_dir().expect("cannot access current directory"),
-            )
-            .expect("non-Unicode path")
-            .join("pg_install")),
+            workdir: Set(PathBuf::new()),
+            pg_distrib_dir: Set(env::current_dir()
+                .expect("cannot access current directory")
+                .join("pg_install")),
             http_auth_type: Set(AuthType::Trust),
             pg_auth_type: Set(AuthType::Trust),
             auth_validation_public_key_path: Set(None),
@@ -354,7 +347,6 @@ impl Default for PageServerConfigBuilder {
             .unwrap()),
 
             control_plane_api: Set(None),
-            control_plane_api_token: Set(None),
         }
     }
 }
@@ -392,11 +384,11 @@ impl PageServerConfigBuilder {
         self.max_file_descriptors = BuilderValue::Set(max_file_descriptors)
     }
 
-    pub fn workdir(&mut self, workdir: Utf8PathBuf) {
+    pub fn workdir(&mut self, workdir: PathBuf) {
         self.workdir = BuilderValue::Set(workdir)
     }
 
-    pub fn pg_distrib_dir(&mut self, pg_distrib_dir: Utf8PathBuf) {
+    pub fn pg_distrib_dir(&mut self, pg_distrib_dir: PathBuf) {
         self.pg_distrib_dir = BuilderValue::Set(pg_distrib_dir)
     }
 
@@ -410,7 +402,7 @@ impl PageServerConfigBuilder {
 
     pub fn auth_validation_public_key_path(
         &mut self,
-        auth_validation_public_key_path: Option<Utf8PathBuf>,
+        auth_validation_public_key_path: Option<PathBuf>,
     ) {
         self.auth_validation_public_key_path = BuilderValue::Set(auth_validation_public_key_path)
     }
@@ -483,12 +475,8 @@ impl PageServerConfigBuilder {
         self.background_task_maximum_delay = BuilderValue::Set(delay);
     }
 
-    pub fn control_plane_api(&mut self, api: Option<Url>) {
-        self.control_plane_api = BuilderValue::Set(api)
-    }
-
-    pub fn control_plane_api_token(&mut self, token: Option<SecretString>) {
-        self.control_plane_api_token = BuilderValue::Set(token)
+    pub fn control_plane_api(&mut self, api: Url) {
+        self.control_plane_api = BuilderValue::Set(Some(api))
     }
 
     pub fn build(self) -> anyhow::Result<PageServerConf> {
@@ -579,9 +567,6 @@ impl PageServerConfigBuilder {
             control_plane_api: self
                 .control_plane_api
                 .ok_or(anyhow!("missing control_plane_api"))?,
-            control_plane_api_token: self
-                .control_plane_api_token
-                .ok_or(anyhow!("missing control_plane_api_token"))?,
         })
     }
 }
@@ -591,55 +576,34 @@ impl PageServerConf {
     // Repository paths, relative to workdir.
     //
 
-    pub fn tenants_path(&self) -> Utf8PathBuf {
+    pub fn tenants_path(&self) -> PathBuf {
         self.workdir.join(TENANTS_SEGMENT_NAME)
     }
 
-    pub fn deletion_prefix(&self) -> Utf8PathBuf {
-        self.workdir.join("deletion")
-    }
-
-    pub fn deletion_list_path(&self, sequence: u64) -> Utf8PathBuf {
-        // Encode a version in the filename, so that if we ever switch away from JSON we can
-        // increment this.
-        const VERSION: u8 = 1;
-
-        self.deletion_prefix()
-            .join(format!("{sequence:016x}-{VERSION:02x}.list"))
-    }
-
-    pub fn deletion_header_path(&self) -> Utf8PathBuf {
-        // Encode a version in the filename, so that if we ever switch away from JSON we can
-        // increment this.
-        const VERSION: u8 = 1;
-
-        self.deletion_prefix().join(format!("header-{VERSION:02x}"))
-    }
-
-    pub fn tenant_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
+    pub fn tenant_path(&self, tenant_id: &TenantId) -> PathBuf {
         self.tenants_path().join(tenant_id.to_string())
     }
 
-    pub fn tenant_attaching_mark_file_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
+    pub fn tenant_attaching_mark_file_path(&self, tenant_id: &TenantId) -> PathBuf {
         self.tenant_path(tenant_id)
             .join(TENANT_ATTACHING_MARKER_FILENAME)
     }
 
-    pub fn tenant_ignore_mark_file_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
+    pub fn tenant_ignore_mark_file_path(&self, tenant_id: &TenantId) -> PathBuf {
         self.tenant_path(tenant_id).join(IGNORED_TENANT_FILE_NAME)
     }
 
     /// Points to a place in pageserver's local directory,
     /// where certain tenant's tenantconf file should be located.
-    pub fn tenant_config_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
+    pub fn tenant_config_path(&self, tenant_id: &TenantId) -> PathBuf {
         self.tenant_path(tenant_id).join(TENANT_CONFIG_NAME)
     }
 
-    pub fn timelines_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
+    pub fn timelines_path(&self, tenant_id: &TenantId) -> PathBuf {
         self.tenant_path(tenant_id).join(TIMELINES_SEGMENT_NAME)
     }
 
-    pub fn timeline_path(&self, tenant_id: &TenantId, timeline_id: &TimelineId) -> Utf8PathBuf {
+    pub fn timeline_path(&self, tenant_id: &TenantId, timeline_id: &TimelineId) -> PathBuf {
         self.timelines_path(tenant_id).join(timeline_id.to_string())
     }
 
@@ -647,7 +611,7 @@ impl PageServerConf {
         &self,
         tenant_id: TenantId,
         timeline_id: TimelineId,
-    ) -> Utf8PathBuf {
+    ) -> PathBuf {
         path_with_suffix_extension(
             self.timeline_path(&tenant_id, &timeline_id),
             TIMELINE_UNINIT_MARK_SUFFIX,
@@ -658,19 +622,19 @@ impl PageServerConf {
         &self,
         tenant_id: TenantId,
         timeline_id: TimelineId,
-    ) -> Utf8PathBuf {
+    ) -> PathBuf {
         path_with_suffix_extension(
             self.timeline_path(&tenant_id, &timeline_id),
             TIMELINE_DELETE_MARK_SUFFIX,
         )
     }
 
-    pub fn tenant_deleted_mark_file_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
+    pub fn tenant_deleted_mark_file_path(&self, tenant_id: &TenantId) -> PathBuf {
         self.tenant_path(tenant_id)
             .join(TENANT_DELETED_MARKER_FILE_NAME)
     }
 
-    pub fn traces_path(&self) -> Utf8PathBuf {
+    pub fn traces_path(&self) -> PathBuf {
         self.workdir.join("traces")
     }
 
@@ -679,7 +643,7 @@ impl PageServerConf {
         tenant_id: &TenantId,
         timeline_id: &TimelineId,
         connection_id: &ConnectionId,
-    ) -> Utf8PathBuf {
+    ) -> PathBuf {
         self.traces_path()
             .join(tenant_id.to_string())
             .join(timeline_id.to_string())
@@ -688,41 +652,49 @@ impl PageServerConf {
 
     /// Points to a place in pageserver's local directory,
     /// where certain timeline's metadata file should be located.
-    pub fn metadata_path(&self, tenant_id: &TenantId, timeline_id: &TimelineId) -> Utf8PathBuf {
+    pub fn metadata_path(&self, tenant_id: &TenantId, timeline_id: &TimelineId) -> PathBuf {
         self.timeline_path(tenant_id, timeline_id)
             .join(METADATA_FILE_NAME)
     }
 
     /// Turns storage remote path of a file into its local path.
-    pub fn local_path(&self, remote_path: &RemotePath) -> Utf8PathBuf {
+    pub fn local_path(&self, remote_path: &RemotePath) -> PathBuf {
         remote_path.with_base(&self.workdir)
     }
 
     //
     // Postgres distribution paths
     //
-    pub fn pg_distrib_dir(&self, pg_version: u32) -> anyhow::Result<Utf8PathBuf> {
+    pub fn pg_distrib_dir(&self, pg_version: u32) -> anyhow::Result<PathBuf> {
         let path = self.pg_distrib_dir.clone();
 
-        #[allow(clippy::manual_range_patterns)]
         match pg_version {
-            14 | 15 | 16 => Ok(path.join(format!("v{pg_version}"))),
+            14 => Ok(path.join(format!("v{pg_version}"))),
+            15 => Ok(path.join(format!("v{pg_version}"))),
             _ => bail!("Unsupported postgres version: {}", pg_version),
         }
     }
 
-    pub fn pg_bin_dir(&self, pg_version: u32) -> anyhow::Result<Utf8PathBuf> {
-        Ok(self.pg_distrib_dir(pg_version)?.join("bin"))
+    pub fn pg_bin_dir(&self, pg_version: u32) -> anyhow::Result<PathBuf> {
+        match pg_version {
+            14 => Ok(self.pg_distrib_dir(pg_version)?.join("bin")),
+            15 => Ok(self.pg_distrib_dir(pg_version)?.join("bin")),
+            _ => bail!("Unsupported postgres version: {}", pg_version),
+        }
     }
-    pub fn pg_lib_dir(&self, pg_version: u32) -> anyhow::Result<Utf8PathBuf> {
-        Ok(self.pg_distrib_dir(pg_version)?.join("lib"))
+    pub fn pg_lib_dir(&self, pg_version: u32) -> anyhow::Result<PathBuf> {
+        match pg_version {
+            14 => Ok(self.pg_distrib_dir(pg_version)?.join("lib")),
+            15 => Ok(self.pg_distrib_dir(pg_version)?.join("lib")),
+            _ => bail!("Unsupported postgres version: {}", pg_version),
+        }
     }
 
     /// Parse a configuration file (pageserver.toml) into a PageServerConf struct,
     /// validating the input and failing on errors.
     ///
     /// This leaves any options not present in the file in the built-in defaults.
-    pub fn parse_and_validate(toml: &Document, workdir: &Utf8Path) -> anyhow::Result<Self> {
+    pub fn parse_and_validate(toml: &Document, workdir: &Path) -> anyhow::Result<Self> {
         let mut builder = PageServerConfigBuilder::default();
         builder.workdir(workdir.to_owned());
 
@@ -741,10 +713,10 @@ impl PageServerConf {
                     builder.max_file_descriptors(parse_toml_u64(key, item)? as usize)
                 }
                 "pg_distrib_dir" => {
-                    builder.pg_distrib_dir(Utf8PathBuf::from(parse_toml_string(key, item)?))
+                    builder.pg_distrib_dir(PathBuf::from(parse_toml_string(key, item)?))
                 }
                 "auth_validation_public_key_path" => builder.auth_validation_public_key_path(Some(
-                    Utf8PathBuf::from(parse_toml_string(key, item)?),
+                    PathBuf::from(parse_toml_string(key, item)?),
                 )),
                 "http_auth_type" => builder.http_auth_type(parse_toml_from_str(key, item)?),
                 "pg_auth_type" => builder.pg_auth_type(parse_toml_from_str(key, item)?),
@@ -783,22 +755,7 @@ impl PageServerConf {
                 },
                 "ondemand_download_behavior_treat_error_as_warn" => builder.ondemand_download_behavior_treat_error_as_warn(parse_toml_bool(key, item)?),
                 "background_task_maximum_delay" => builder.background_task_maximum_delay(parse_toml_duration(key, item)?),
-                "control_plane_api" => {
-                    let parsed = parse_toml_string(key, item)?;
-                    if parsed.is_empty() {
-                        builder.control_plane_api(None)
-                    } else {
-                        builder.control_plane_api(Some(parsed.parse().context("failed to parse control plane URL")?))
-                    }
-                },
-                "control_plane_api_token" => {
-                    let parsed = parse_toml_string(key, item)?;
-                    if parsed.is_empty() {
-                        builder.control_plane_api_token(None)
-                    } else {
-                        builder.control_plane_api_token(Some(parsed.into()))
-                    }
-                },
+                "control_plane_api" => builder.control_plane_api(parse_toml_string(key, item)?.parse().context("failed to parse control plane URL")?),
                 _ => bail!("unrecognized pageserver option '{key}'"),
             }
         }
@@ -812,7 +769,8 @@ impl PageServerConf {
             ensure!(
                 auth_validation_public_key_path.exists(),
                 format!(
-                    "Can't find auth_validation_public_key at '{auth_validation_public_key_path}'",
+                    "Can't find auth_validation_public_key at '{}'",
+                    auth_validation_public_key_path.display()
                 )
             );
         }
@@ -928,12 +886,12 @@ impl PageServerConf {
     }
 
     #[cfg(test)]
-    pub fn test_repo_dir(test_name: &str) -> Utf8PathBuf {
-        Utf8PathBuf::from(format!("../tmp_check/test_{test_name}"))
+    pub fn test_repo_dir(test_name: &str) -> PathBuf {
+        PathBuf::from(format!("../tmp_check/test_{test_name}"))
     }
 
-    pub fn dummy_conf(repo_dir: Utf8PathBuf) -> Self {
-        let pg_distrib_dir = Utf8PathBuf::from(env!("CARGO_MANIFEST_DIR")).join("../pg_install");
+    pub fn dummy_conf(repo_dir: PathBuf) -> Self {
+        let pg_distrib_dir = PathBuf::from(env!("CARGO_MANIFEST_DIR")).join("../pg_install");
 
         PageServerConf {
             id: NodeId(0),
@@ -967,7 +925,6 @@ impl PageServerConf {
             ondemand_download_behavior_treat_error_as_warn: false,
             background_task_maximum_delay: Duration::ZERO,
             control_plane_api: None,
-            control_plane_api_token: None,
         }
     }
 }
@@ -1100,8 +1057,8 @@ mod tests {
         num::{NonZeroU32, NonZeroUsize},
     };
 
-    use camino_tempfile::{tempdir, Utf8TempDir};
     use remote_storage::{RemoteStorageKind, S3Config};
+    use tempfile::{tempdir, TempDir};
     use utils::serde_percent::Percent;
 
     use super::*;
@@ -1140,7 +1097,8 @@ background_task_maximum_delay = '334 s'
         let broker_endpoint = storage_broker::DEFAULT_ENDPOINT;
         // we have to create dummy values to overcome the validation errors
         let config_string = format!(
-            "pg_distrib_dir='{pg_distrib_dir}'\nid=10\nbroker_endpoint = '{broker_endpoint}'",
+            "pg_distrib_dir='{}'\nid=10\nbroker_endpoint = '{broker_endpoint}'",
+            pg_distrib_dir.display()
         );
         let toml = config_string.parse()?;
 
@@ -1190,8 +1148,7 @@ background_task_maximum_delay = '334 s'
                 background_task_maximum_delay: humantime::parse_duration(
                     defaults::DEFAULT_BACKGROUND_TASK_MAXIMUM_DELAY
                 )?,
-                control_plane_api: None,
-                control_plane_api_token: None
+                control_plane_api: None
             },
             "Correct defaults should be used when no config values are provided"
         );
@@ -1206,7 +1163,8 @@ background_task_maximum_delay = '334 s'
         let broker_endpoint = storage_broker::DEFAULT_ENDPOINT;
 
         let config_string = format!(
-            "{ALL_BASE_VALUES_TOML}pg_distrib_dir='{pg_distrib_dir}'\nbroker_endpoint = '{broker_endpoint}'",
+            "{ALL_BASE_VALUES_TOML}pg_distrib_dir='{}'\nbroker_endpoint = '{broker_endpoint}'",
+            pg_distrib_dir.display()
         );
         let toml = config_string.parse()?;
 
@@ -1246,8 +1204,7 @@ background_task_maximum_delay = '334 s'
                 test_remote_failures: 0,
                 ondemand_download_behavior_treat_error_as_warn: false,
                 background_task_maximum_delay: Duration::from_secs(334),
-                control_plane_api: None,
-                control_plane_api_token: None
+                control_plane_api: None
             },
             "Should be able to parse all basic config values correctly"
         );
@@ -1266,18 +1223,23 @@ background_task_maximum_delay = '334 s'
         let identical_toml_declarations = &[
             format!(
                 r#"[remote_storage]
-local_path = '{local_storage_path}'"#,
+local_path = '{}'"#,
+                local_storage_path.display()
+            ),
+            format!(
+                "remote_storage={{local_path='{}'}}",
+                local_storage_path.display()
             ),
-            format!("remote_storage={{local_path='{local_storage_path}'}}"),
         ];
 
         for remote_storage_config_str in identical_toml_declarations {
             let config_string = format!(
                 r#"{ALL_BASE_VALUES_TOML}
-pg_distrib_dir='{pg_distrib_dir}'
+pg_distrib_dir='{}'
 broker_endpoint = '{broker_endpoint}'
 
 {remote_storage_config_str}"#,
+                pg_distrib_dir.display(),
             );
 
             let toml = config_string.parse()?;
@@ -1340,10 +1302,11 @@ concurrency_limit = {s3_concurrency_limit}"#
         for remote_storage_config_str in identical_toml_declarations {
             let config_string = format!(
                 r#"{ALL_BASE_VALUES_TOML}
-pg_distrib_dir='{pg_distrib_dir}'
+pg_distrib_dir='{}'
 broker_endpoint = '{broker_endpoint}'
 
 {remote_storage_config_str}"#,
+                pg_distrib_dir.display(),
             );
 
             let toml = config_string.parse()?;
@@ -1385,11 +1348,12 @@ broker_endpoint = '{broker_endpoint}'
 
         let config_string = format!(
             r#"{ALL_BASE_VALUES_TOML}
-pg_distrib_dir='{pg_distrib_dir}'
+pg_distrib_dir='{}'
 broker_endpoint = '{broker_endpoint}'
 
 [tenant_config]
 trace_read_requests = {trace_read_requests}"#,
+            pg_distrib_dir.display(),
         );
 
         let toml = config_string.parse()?;
@@ -1409,7 +1373,7 @@ trace_read_requests = {trace_read_requests}"#,
         let (workdir, pg_distrib_dir) = prepare_fs(&tempdir)?;
 
         let pageserver_conf_toml = format!(
-            r#"pg_distrib_dir = "{pg_distrib_dir}"
+            r#"pg_distrib_dir = "{}"
 metric_collection_endpoint = "http://sample.url"
 metric_collection_interval = "10min"
 id = 222
@@ -1427,6 +1391,7 @@ kind = "LayerAccessThreshold"
 period = "20m"
 threshold = "20m"
 "#,
+            pg_distrib_dir.display(),
         );
         let toml: Document = pageserver_conf_toml.parse()?;
         let conf = PageServerConf::parse_and_validate(&toml, &workdir)?;
@@ -1467,7 +1432,7 @@ threshold = "20m"
         Ok(())
     }
 
-    fn prepare_fs(tempdir: &Utf8TempDir) -> anyhow::Result<(Utf8PathBuf, Utf8PathBuf)> {
+    fn prepare_fs(tempdir: &TempDir) -> anyhow::Result<(PathBuf, PathBuf)> {
         let tempdir_path = tempdir.path();
 
         let workdir = tempdir_path.join("workdir");

pageserver/src/consumption_metrics.rs

@@ -1,54 +1,188 @@
+//!
 //! Periodically collect consumption metrics for all active tenants
 //! and push them to a HTTP endpoint.
+//! Cache metrics to send only the updated ones.
+//!
 use crate::context::{DownloadBehavior, RequestContext};
 use crate::task_mgr::{self, TaskKind, BACKGROUND_RUNTIME};
 use crate::tenant::{mgr, LogicalSizeCalculationCause};
-use camino::Utf8PathBuf;
-use consumption_metrics::EventType;
+use anyhow;
+use chrono::{DateTime, Utc};
+use consumption_metrics::{idempotency_key, Event, EventChunk, EventType, CHUNK_SIZE};
 use pageserver_api::models::TenantState;
 use reqwest::Url;
+use serde::Serialize;
+use serde_with::{serde_as, DisplayFromStr};
 use std::collections::HashMap;
 use std::sync::Arc;
 use std::time::{Duration, SystemTime};
 use tracing::*;
-use utils::id::NodeId;
-
-mod metrics;
-use metrics::MetricsKey;
-mod disk_cache;
-mod upload;
+use utils::id::{NodeId, TenantId, TimelineId};
+use utils::lsn::Lsn;
 
 const DEFAULT_HTTP_REPORTING_TIMEOUT: Duration = Duration::from_secs(60);
 
-/// Basically a key-value pair, but usually in a Vec except for [`Cache`].
-///
-/// This is as opposed to `consumption_metrics::Event` which is the externally communicated form.
-/// Difference is basically the missing idempotency key, which lives only for the duration of
-/// upload attempts.
-type RawMetric = (MetricsKey, (EventType, u64));
+#[serde_as]
+#[derive(Serialize, Debug, Clone, Copy)]
+struct Ids {
+    #[serde_as(as = "DisplayFromStr")]
+    tenant_id: TenantId,
+    #[serde_as(as = "Option<DisplayFromStr>")]
+    #[serde(skip_serializing_if = "Option::is_none")]
+    timeline_id: Option<TimelineId>,
+}
 
-/// Caches the [`RawMetric`]s
-///
-/// In practice, during startup, last sent values are stored here to be used in calculating new
-/// ones. After successful uploading, the cached values are updated to cache. This used to be used
-/// for deduplication, but that is no longer needed.
-type Cache = HashMap<MetricsKey, (EventType, u64)>;
+/// Key that uniquely identifies the object, this metric describes.
+#[derive(Debug, Clone, PartialEq, Eq, Hash)]
+struct MetricsKey {
+    tenant_id: TenantId,
+    timeline_id: Option<TimelineId>,
+    metric: &'static str,
+}
+
+impl MetricsKey {
+    const fn absolute_values(self) -> AbsoluteValueFactory {
+        AbsoluteValueFactory(self)
+    }
+    const fn incremental_values(self) -> IncrementalValueFactory {
+        IncrementalValueFactory(self)
+    }
+}
+
+/// Helper type which each individual metric kind can return to produce only absolute values.
+struct AbsoluteValueFactory(MetricsKey);
+
+impl AbsoluteValueFactory {
+    fn at(self, time: DateTime<Utc>, val: u64) -> (MetricsKey, (EventType, u64)) {
+        let key = self.0;
+        (key, (EventType::Absolute { time }, val))
+    }
+}
+
+/// Helper type which each individual metric kind can return to produce only incremental values.
+struct IncrementalValueFactory(MetricsKey);
+
+impl IncrementalValueFactory {
+    #[allow(clippy::wrong_self_convention)]
+    fn from_previous_up_to(
+        self,
+        prev_end: DateTime<Utc>,
+        up_to: DateTime<Utc>,
+        val: u64,
+    ) -> (MetricsKey, (EventType, u64)) {
+        let key = self.0;
+        // cannot assert prev_end < up_to because these are realtime clock based
+        (
+            key,
+            (
+                EventType::Incremental {
+                    start_time: prev_end,
+                    stop_time: up_to,
+                },
+                val,
+            ),
+        )
+    }
+
+    fn key(&self) -> &MetricsKey {
+        &self.0
+    }
+}
+
+// the static part of a MetricsKey
+impl MetricsKey {
+    /// Absolute value of [`Timeline::get_last_record_lsn`].
+    ///
+    /// [`Timeline::get_last_record_lsn`]: crate::tenant::Timeline::get_last_record_lsn
+    const fn written_size(tenant_id: TenantId, timeline_id: TimelineId) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: Some(timeline_id),
+            metric: "written_size",
+        }
+        .absolute_values()
+    }
+
+    /// Values will be the difference of the latest [`MetricsKey::written_size`] to what we
+    /// previously sent, starting from the previously sent incremental time range ending at the
+    /// latest absolute measurement.
+    const fn written_size_delta(
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+    ) -> IncrementalValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: Some(timeline_id),
+            // the name here is correctly about data not size, because that is what is wanted by
+            // downstream pipeline
+            metric: "written_data_bytes_delta",
+        }
+        .incremental_values()
+    }
+
+    /// Exact [`Timeline::get_current_logical_size`].
+    ///
+    /// [`Timeline::get_current_logical_size`]: crate::tenant::Timeline::get_current_logical_size
+    const fn timeline_logical_size(
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+    ) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: Some(timeline_id),
+            metric: "timeline_logical_size",
+        }
+        .absolute_values()
+    }
+
+    /// [`Tenant::remote_size`]
+    ///
+    /// [`Tenant::remote_size`]: crate::tenant::Tenant::remote_size
+    const fn remote_storage_size(tenant_id: TenantId) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: None,
+            metric: "remote_storage_size",
+        }
+        .absolute_values()
+    }
+
+    /// Sum of [`Timeline::resident_physical_size`] for each `Tenant`.
+    ///
+    /// [`Timeline::resident_physical_size`]: crate::tenant::Timeline::resident_physical_size
+    const fn resident_size(tenant_id: TenantId) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: None,
+            metric: "resident_size",
+        }
+        .absolute_values()
+    }
+
+    /// [`Tenant::cached_synthetic_size`] as refreshed by [`calculate_synthetic_size_worker`].
+    ///
+    /// [`Tenant::cached_synthetic_size`]: crate::tenant::Tenant::cached_synthetic_size
+    const fn synthetic_size(tenant_id: TenantId) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: None,
+            metric: "synthetic_storage_size",
+        }
+        .absolute_values()
+    }
+}
 
 /// Main thread that serves metrics collection
 pub async fn collect_metrics(
     metric_collection_endpoint: &Url,
     metric_collection_interval: Duration,
-    _cached_metric_collection_interval: Duration,
+    cached_metric_collection_interval: Duration,
     synthetic_size_calculation_interval: Duration,
     node_id: NodeId,
-    local_disk_storage: Utf8PathBuf,
     ctx: RequestContext,
 ) -> anyhow::Result<()> {
-    if _cached_metric_collection_interval != Duration::ZERO {
-        tracing::warn!(
-            "cached_metric_collection_interval is no longer used, please set it to zero."
-        )
-    }
+    let mut ticker = tokio::time::interval(metric_collection_interval);
+    info!("starting collect_metrics");
 
     // spin up background worker that caclulates tenant sizes
     let worker_ctx =
@@ -68,216 +202,543 @@ pub async fn collect_metrics(
         },
     );
 
-    let path: Arc<Utf8PathBuf> = Arc::new(local_disk_storage);
-
-    let cancel = task_mgr::shutdown_token();
-
-    let restore_and_reschedule = restore_and_reschedule(&path, metric_collection_interval);
-
-    let mut cached_metrics = tokio::select! {
-        _ = cancel.cancelled() => return Ok(()),
-        ret = restore_and_reschedule => ret,
-    };
-
     // define client here to reuse it for all requests
     let client = reqwest::ClientBuilder::new()
         .timeout(DEFAULT_HTTP_REPORTING_TIMEOUT)
         .build()
         .expect("Failed to create http client with timeout");
-
-    let node_id = node_id.to_string();
-
-    // reminder: ticker is ready immediatedly
-    let mut ticker = tokio::time::interval(metric_collection_interval);
+    let mut cached_metrics = HashMap::new();
+    let mut prev_iteration_time: std::time::Instant = std::time::Instant::now();
 
     loop {
-        let tick_at = tokio::select! {
-            _ = cancel.cancelled() => return Ok(()),
-            tick_at = ticker.tick() => tick_at,
-        };
+        tokio::select! {
+            _ = task_mgr::shutdown_watcher() => {
+                info!("collect_metrics received cancellation request");
+                return Ok(());
+            },
+            tick_at = ticker.tick() => {
 
-        // these are point in time, with variable "now"
-        let metrics = metrics::collect_all_metrics(&cached_metrics, &ctx).await;
+                // send cached metrics every cached_metric_collection_interval
+                let send_cached = prev_iteration_time.elapsed() >= cached_metric_collection_interval;
 
-        if metrics.is_empty() {
-            continue;
+                if send_cached {
+                    prev_iteration_time = std::time::Instant::now();
+                }
+
+                collect_metrics_iteration(&client, &mut cached_metrics, metric_collection_endpoint, node_id, &ctx, send_cached).await;
+
+                crate::tenant::tasks::warn_when_period_overrun(
+                    tick_at.elapsed(),
+                    metric_collection_interval,
+                    "consumption_metrics_collect_metrics",
+                );
+            }
         }
-
-        let metrics = Arc::new(metrics);
-
-        // why not race cancellation here? because we are one of the last tasks, and if we are
-        // already here, better to try to flush the new values.
-
-        let flush = async {
-            match disk_cache::flush_metrics_to_disk(&metrics, &path).await {
-                Ok(()) => {
-                    tracing::debug!("flushed metrics to disk");
-                }
-                Err(e) => {
-                    // idea here is that if someone creates a directory as our path, then they
-                    // might notice it from the logs before shutdown and remove it
-                    tracing::error!("failed to persist metrics to {path:?}: {e:#}");
-                }
-            }
-        };
-
-        let upload = async {
-            let res = upload::upload_metrics(
-                &client,
-                metric_collection_endpoint,
-                &cancel,
-                &node_id,
-                &metrics,
-                &mut cached_metrics,
-            )
-            .await;
-            if let Err(e) = res {
-                // serialization error which should never happen
-                tracing::error!("failed to upload due to {e:#}");
-            }
-        };
-
-        // let these run concurrently
-        let (_, _) = tokio::join!(flush, upload);
-
-        crate::tenant::tasks::warn_when_period_overrun(
-            tick_at.elapsed(),
-            metric_collection_interval,
-            "consumption_metrics_collect_metrics",
-        );
     }
 }
 
-/// Called on the first iteration in an attempt to join the metric uploading schedule from previous
-/// pageserver session. Pageserver is supposed to upload at intervals regardless of restarts.
+/// One iteration of metrics collection
 ///
-/// Cancellation safe.
-async fn restore_and_reschedule(
-    path: &Arc<Utf8PathBuf>,
-    metric_collection_interval: Duration,
-) -> Cache {
-    let (cached, earlier_metric_at) = match disk_cache::read_metrics_from_disk(path.clone()).await {
-        Ok(found_some) => {
-            // there is no min needed because we write these sequentially in
-            // collect_all_metrics
-            let earlier_metric_at = found_some
-                .iter()
-                .map(|(_, (et, _))| et.recorded_at())
-                .copied()
-                .next();
+/// Gather per-tenant and per-timeline metrics and send them to the `metric_collection_endpoint`.
+/// Cache metrics to avoid sending the same metrics multiple times.
+///
+/// This function handles all errors internally
+/// and doesn't break iteration if just one tenant fails.
+///
+/// TODO
+/// - refactor this function (chunking+sending part) to reuse it in proxy module;
+async fn collect_metrics_iteration(
+    client: &reqwest::Client,
+    cached_metrics: &mut HashMap<MetricsKey, (EventType, u64)>,
+    metric_collection_endpoint: &reqwest::Url,
+    node_id: NodeId,
+    ctx: &RequestContext,
+    send_cached: bool,
+) {
+    let mut current_metrics: Vec<(MetricsKey, (EventType, u64))> = Vec::new();
+    trace!(
+        "starting collect_metrics_iteration. metric_collection_endpoint: {}",
+        metric_collection_endpoint
+    );
 
-            let cached = found_some.into_iter().collect::<Cache>();
-
-            (cached, earlier_metric_at)
-        }
-        Err(e) => {
-            use std::io::{Error, ErrorKind};
-
-            let root = e.root_cause();
-            let maybe_ioerr = root.downcast_ref::<Error>();
-            let is_not_found = maybe_ioerr.is_some_and(|e| e.kind() == ErrorKind::NotFound);
-
-            if !is_not_found {
-                tracing::info!("failed to read any previous metrics from {path:?}: {e:#}");
-            }
-
-            (HashMap::new(), None)
+    // get list of tenants
+    let tenants = match mgr::list_tenants().await {
+        Ok(tenants) => tenants,
+        Err(err) => {
+            error!("failed to list tenants: {:?}", err);
+            return;
         }
     };
 
-    if let Some(earlier_metric_at) = earlier_metric_at {
-        let earlier_metric_at: SystemTime = earlier_metric_at.into();
+    // iterate through list of Active tenants and collect metrics
+    for (tenant_id, tenant_state) in tenants {
+        if tenant_state != TenantState::Active {
+            continue;
+        }
 
-        let error = reschedule(earlier_metric_at, metric_collection_interval).await;
-
-        if let Some(error) = error {
-            if error.as_secs() >= 60 {
-                tracing::info!(
-                    error_ms = error.as_millis(),
-                    "startup scheduling error due to restart"
-                )
+        let tenant = match mgr::get_tenant(tenant_id, true).await {
+            Ok(tenant) => tenant,
+            Err(err) => {
+                // It is possible that tenant was deleted between
+                // `list_tenants` and `get_tenant`, so just warn about it.
+                warn!("failed to get tenant {tenant_id:?}: {err:?}");
+                continue;
             }
+        };
+
+        let mut tenant_resident_size = 0;
+
+        // iterate through list of timelines in tenant
+        for timeline in tenant.list_timelines() {
+            // collect per-timeline metrics only for active timelines
+
+            let timeline_id = timeline.timeline_id;
+
+            match TimelineSnapshot::collect(&timeline, ctx) {
+                Ok(Some(snap)) => {
+                    snap.to_metrics(
+                        tenant_id,
+                        timeline_id,
+                        Utc::now(),
+                        &mut current_metrics,
+                        cached_metrics,
+                    );
+                }
+                Ok(None) => {}
+                Err(e) => {
+                    error!(
+                        "failed to get metrics values for tenant {tenant_id} timeline {}: {e:#?}",
+                        timeline.timeline_id
+                    );
+                    continue;
+                }
+            }
+
+            tenant_resident_size += timeline.resident_physical_size();
+        }
+
+        current_metrics
+            .push(MetricsKey::remote_storage_size(tenant_id).at(Utc::now(), tenant.remote_size()));
+
+        current_metrics
+            .push(MetricsKey::resident_size(tenant_id).at(Utc::now(), tenant_resident_size));
+
+        // Note that this metric is calculated in a separate bgworker
+        // Here we only use cached value, which may lag behind the real latest one
+        let synthetic_size = tenant.cached_synthetic_size();
+
+        if synthetic_size != 0 {
+            // only send non-zeroes because otherwise these show up as errors in logs
+            current_metrics
+                .push(MetricsKey::synthetic_size(tenant_id).at(Utc::now(), synthetic_size));
         }
     }
 
-    cached
+    // Filter metrics, unless we want to send all metrics, including cached ones.
+    // See: https://github.com/neondatabase/neon/issues/3485
+    if !send_cached {
+        current_metrics.retain(|(curr_key, (kind, curr_val))| {
+            if kind.is_incremental() {
+                // incremental values (currently only written_size_delta) should not get any cache
+                // deduplication because they will be used by upstream for "is still alive."
+                true
+            } else {
+                match cached_metrics.get(curr_key) {
+                    Some((_, val)) => val != curr_val,
+                    None => true,
+                }
+            }
+        });
+    }
+
+    if current_metrics.is_empty() {
+        trace!("no new metrics to send");
+        return;
+    }
+
+    // Send metrics.
+    // Split into chunks of 1000 metrics to avoid exceeding the max request size
+    let chunks = current_metrics.chunks(CHUNK_SIZE);
+
+    let mut chunk_to_send: Vec<Event<Ids>> = Vec::with_capacity(CHUNK_SIZE);
+
+    let node_id = node_id.to_string();
+
+    for chunk in chunks {
+        chunk_to_send.clear();
+
+        // enrich metrics with type,timestamp and idempotency key before sending
+        chunk_to_send.extend(chunk.iter().map(|(curr_key, (when, curr_val))| Event {
+            kind: *when,
+            metric: curr_key.metric,
+            idempotency_key: idempotency_key(&node_id),
+            value: *curr_val,
+            extra: Ids {
+                tenant_id: curr_key.tenant_id,
+                timeline_id: curr_key.timeline_id,
+            },
+        }));
+
+        const MAX_RETRIES: u32 = 3;
+
+        for attempt in 0..MAX_RETRIES {
+            let res = client
+                .post(metric_collection_endpoint.clone())
+                .json(&EventChunk {
+                    events: (&chunk_to_send).into(),
+                })
+                .send()
+                .await;
+
+            match res {
+                Ok(res) => {
+                    if res.status().is_success() {
+                        // update cached metrics after they were sent successfully
+                        for (curr_key, curr_val) in chunk.iter() {
+                            cached_metrics.insert(curr_key.clone(), *curr_val);
+                        }
+                    } else {
+                        error!("metrics endpoint refused the sent metrics: {:?}", res);
+                        for metric in chunk_to_send
+                            .iter()
+                            .filter(|metric| metric.value > (1u64 << 40))
+                        {
+                            // Report if the metric value is suspiciously large
+                            error!("potentially abnormal metric value: {:?}", metric);
+                        }
+                    }
+                    break;
+                }
+                Err(err) if err.is_timeout() => {
+                    error!(attempt, "timeout sending metrics, retrying immediately");
+                    continue;
+                }
+                Err(err) => {
+                    error!(attempt, ?err, "failed to send metrics");
+                    break;
+                }
+            }
+        }
+    }
 }
 
-async fn reschedule(
-    earlier_metric_at: SystemTime,
-    metric_collection_interval: Duration,
-) -> Option<Duration> {
-    let now = SystemTime::now();
-    match now.duration_since(earlier_metric_at) {
-        Ok(from_last_send) if from_last_send < metric_collection_interval => {
-            let sleep_for = metric_collection_interval - from_last_send;
+/// Internal type to make timeline metric production testable.
+///
+/// As this value type contains all of the information needed from a timeline to produce the
+/// metrics, it can easily be created with different values in test.
+struct TimelineSnapshot {
+    loaded_at: (Lsn, SystemTime),
+    last_record_lsn: Lsn,
+    current_exact_logical_size: Option<u64>,
+}
 
-            let deadline = std::time::Instant::now() + sleep_for;
+impl TimelineSnapshot {
+    /// Collect the metrics from an actual timeline.
+    ///
+    /// Fails currently only when [`Timeline::get_current_logical_size`] fails.
+    ///
+    /// [`Timeline::get_current_logical_size`]: crate::tenant::Timeline::get_current_logical_size
+    fn collect(
+        t: &Arc<crate::tenant::Timeline>,
+        ctx: &RequestContext,
+    ) -> anyhow::Result<Option<Self>> {
+        use anyhow::Context;
 
-            tokio::time::sleep_until(deadline.into()).await;
+        if !t.is_active() {
+            // no collection for broken or stopping needed, we will still keep the cached values
+            // though at the caller.
+            Ok(None)
+        } else {
+            let loaded_at = t.loaded_at;
+            let last_record_lsn = t.get_last_record_lsn();
 
-            let now = std::time::Instant::now();
+            let current_exact_logical_size = {
+                let span = info_span!("collect_metrics_iteration", tenant_id = %t.tenant_id, timeline_id = %t.timeline_id);
+                let res = span
+                    .in_scope(|| t.get_current_logical_size(ctx))
+                    .context("get_current_logical_size");
+                match res? {
+                    // Only send timeline logical size when it is fully calculated.
+                    (size, is_exact) if is_exact => Some(size),
+                    (_, _) => None,
+                }
+            };
 
-            // executor threads might be busy, add extra measurements
-            Some(if now < deadline {
-                deadline - now
-            } else {
-                now - deadline
-            })
+            Ok(Some(TimelineSnapshot {
+                loaded_at,
+                last_record_lsn,
+                current_exact_logical_size,
+            }))
         }
-        Ok(from_last_send) => Some(from_last_send.saturating_sub(metric_collection_interval)),
-        Err(_) => {
-            tracing::warn!(
-                ?now,
-                ?earlier_metric_at,
-                "oldest recorded metric is in future; first values will come out with inconsistent timestamps"
-            );
-            earlier_metric_at.duration_since(now).ok()
+    }
+
+    /// Produce the timeline consumption metrics into the `metrics` argument.
+    fn to_metrics(
+        &self,
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+        now: DateTime<Utc>,
+        metrics: &mut Vec<(MetricsKey, (EventType, u64))>,
+        cache: &HashMap<MetricsKey, (EventType, u64)>,
+    ) {
+        let timeline_written_size = u64::from(self.last_record_lsn);
+
+        let (key, written_size_now) =
+            MetricsKey::written_size(tenant_id, timeline_id).at(now, timeline_written_size);
+
+        // last_record_lsn can only go up, right now at least, TODO: #2592 or related
+        // features might change this.
+
+        let written_size_delta_key = MetricsKey::written_size_delta(tenant_id, timeline_id);
+
+        // use this when available, because in a stream of incremental values, it will be
+        // accurate where as when last_record_lsn stops moving, we will only cache the last
+        // one of those.
+        let last_stop_time = cache
+            .get(written_size_delta_key.key())
+            .map(|(until, _val)| {
+                until
+                    .incremental_timerange()
+                    .expect("never create EventType::Absolute for written_size_delta")
+                    .end
+            });
+
+        // by default, use the last sent written_size as the basis for
+        // calculating the delta. if we don't yet have one, use the load time value.
+        let prev = cache
+            .get(&key)
+            .map(|(prev_at, prev)| {
+                // use the prev time from our last incremental update, or default to latest
+                // absolute update on the first round.
+                let prev_at = prev_at
+                    .absolute_time()
+                    .expect("never create EventType::Incremental for written_size");
+                let prev_at = last_stop_time.unwrap_or(prev_at);
+                (*prev_at, *prev)
+            })
+            .unwrap_or_else(|| {
+                // if we don't have a previous point of comparison, compare to the load time
+                // lsn.
+                let (disk_consistent_lsn, loaded_at) = &self.loaded_at;
+                (DateTime::from(*loaded_at), disk_consistent_lsn.0)
+            });
+
+        // written_size_bytes_delta
+        metrics.extend(
+            if let Some(delta) = written_size_now.1.checked_sub(prev.1) {
+                let up_to = written_size_now
+                    .0
+                    .absolute_time()
+                    .expect("never create EventType::Incremental for written_size");
+                let key_value = written_size_delta_key.from_previous_up_to(prev.0, *up_to, delta);
+                Some(key_value)
+            } else {
+                None
+            },
+        );
+
+        // written_size
+        metrics.push((key, written_size_now));
+
+        if let Some(size) = self.current_exact_logical_size {
+            metrics.push(MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, size));
         }
     }
 }
 
 /// Caclculate synthetic size for each active tenant
-async fn calculate_synthetic_size_worker(
+pub async fn calculate_synthetic_size_worker(
     synthetic_size_calculation_interval: Duration,
     ctx: &RequestContext,
 ) -> anyhow::Result<()> {
     info!("starting calculate_synthetic_size_worker");
 
-    // reminder: ticker is ready immediatedly
     let mut ticker = tokio::time::interval(synthetic_size_calculation_interval);
-    let cause = LogicalSizeCalculationCause::ConsumptionMetricsSyntheticSize;
 
     loop {
-        let tick_at = tokio::select! {
-            _ = task_mgr::shutdown_watcher() => return Ok(()),
-            tick_at = ticker.tick() => tick_at,
-        };
+        tokio::select! {
+            _ = task_mgr::shutdown_watcher() => {
+                return Ok(());
+            },
+            tick_at = ticker.tick() => {
 
-        let tenants = match mgr::list_tenants().await {
-            Ok(tenants) => tenants,
-            Err(e) => {
-                warn!("cannot get tenant list: {e:#}");
-                continue;
-            }
-        };
+                let tenants = match mgr::list_tenants().await {
+                    Ok(tenants) => tenants,
+                    Err(e) => {
+                        warn!("cannot get tenant list: {e:#}");
+                        continue;
+                    }
+                };
+                // iterate through list of Active tenants and collect metrics
+                for (tenant_id, tenant_state) in tenants {
 
-        for (tenant_id, tenant_state) in tenants {
-            if tenant_state != TenantState::Active {
-                continue;
-            }
+                    if tenant_state != TenantState::Active {
+                        continue;
+                    }
+
+                    if let Ok(tenant) = mgr::get_tenant(tenant_id, true).await
+                    {
+                        if let Err(e) = tenant.calculate_synthetic_size(
+                            LogicalSizeCalculationCause::ConsumptionMetricsSyntheticSize,
+                            ctx).await {
+                            error!("failed to calculate synthetic size for tenant {}: {}", tenant_id, e);
+                        }
+                    }
 
-            if let Ok(tenant) = mgr::get_tenant(tenant_id, true).await {
-                if let Err(e) = tenant.calculate_synthetic_size(cause, ctx).await {
-                    error!("failed to calculate synthetic size for tenant {tenant_id}: {e:#}");
                 }
+
+                crate::tenant::tasks::warn_when_period_overrun(
+                    tick_at.elapsed(),
+                    synthetic_size_calculation_interval,
+                    "consumption_metrics_synthetic_size_worker",
+                );
             }
         }
-
-        crate::tenant::tasks::warn_when_period_overrun(
-            tick_at.elapsed(),
-            synthetic_size_calculation_interval,
-            "consumption_metrics_synthetic_size_worker",
-        );
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::collections::HashMap;
+
+    use std::time::SystemTime;
+    use utils::{
+        id::{TenantId, TimelineId},
+        lsn::Lsn,
+    };
+
+    use crate::consumption_metrics::MetricsKey;
+
+    use super::TimelineSnapshot;
+    use chrono::{DateTime, Utc};
+
+    #[test]
+    fn startup_collected_timeline_metrics_before_advancing() {
+        let tenant_id = TenantId::generate();
+        let timeline_id = TimelineId::generate();
+
+        let mut metrics = Vec::new();
+        let cache = HashMap::new();
+
+        let initdb_lsn = Lsn(0x10000);
+        let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
+
+        let snap = TimelineSnapshot {
+            loaded_at: (disk_consistent_lsn, SystemTime::now()),
+            last_record_lsn: disk_consistent_lsn,
+            current_exact_logical_size: Some(0x42000),
+        };
+
+        let now = DateTime::<Utc>::from(SystemTime::now());
+
+        snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
+
+        assert_eq!(
+            metrics,
+            &[
+                MetricsKey::written_size_delta(tenant_id, timeline_id).from_previous_up_to(
+                    snap.loaded_at.1.into(),
+                    now,
+                    0
+                ),
+                MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
+                MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
+            ]
+        );
+    }
+
+    #[test]
+    fn startup_collected_timeline_metrics_second_round() {
+        let tenant_id = TenantId::generate();
+        let timeline_id = TimelineId::generate();
+
+        let [now, before, init] = time_backwards();
+
+        let now = DateTime::<Utc>::from(now);
+        let before = DateTime::<Utc>::from(before);
+
+        let initdb_lsn = Lsn(0x10000);
+        let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
+
+        let mut metrics = Vec::new();
+        let cache = HashMap::from([
+            MetricsKey::written_size(tenant_id, timeline_id).at(before, disk_consistent_lsn.0)
+        ]);
+
+        let snap = TimelineSnapshot {
+            loaded_at: (disk_consistent_lsn, init),
+            last_record_lsn: disk_consistent_lsn,
+            current_exact_logical_size: Some(0x42000),
+        };
+
+        snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
+
+        assert_eq!(
+            metrics,
+            &[
+                MetricsKey::written_size_delta(tenant_id, timeline_id)
+                    .from_previous_up_to(before, now, 0),
+                MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
+                MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
+            ]
+        );
+    }
+
+    #[test]
+    fn startup_collected_timeline_metrics_nth_round_at_same_lsn() {
+        let tenant_id = TenantId::generate();
+        let timeline_id = TimelineId::generate();
+
+        let [now, just_before, before, init] = time_backwards();
+
+        let now = DateTime::<Utc>::from(now);
+        let just_before = DateTime::<Utc>::from(just_before);
+        let before = DateTime::<Utc>::from(before);
+
+        let initdb_lsn = Lsn(0x10000);
+        let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
+
+        let mut metrics = Vec::new();
+        let cache = HashMap::from([
+            // at t=before was the last time the last_record_lsn changed
+            MetricsKey::written_size(tenant_id, timeline_id).at(before, disk_consistent_lsn.0),
+            // end time of this event is used for the next ones
+            MetricsKey::written_size_delta(tenant_id, timeline_id).from_previous_up_to(
+                before,
+                just_before,
+                0,
+            ),
+        ]);
+
+        let snap = TimelineSnapshot {
+            loaded_at: (disk_consistent_lsn, init),
+            last_record_lsn: disk_consistent_lsn,
+            current_exact_logical_size: Some(0x42000),
+        };
+
+        snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
+
+        assert_eq!(
+            metrics,
+            &[
+                MetricsKey::written_size_delta(tenant_id, timeline_id).from_previous_up_to(
+                    just_before,
+                    now,
+                    0
+                ),
+                MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
+                MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
+            ]
+        );
+    }
+
+    fn time_backwards<const N: usize>() -> [std::time::SystemTime; N] {
+        let mut times = [std::time::SystemTime::UNIX_EPOCH; N];
+        times[0] = std::time::SystemTime::now();
+        for behind in 1..N {
+            times[behind] = times[0] - std::time::Duration::from_secs(behind as u64);
+        }
+
+        times
     }
 }

pageserver/src/consumption_metrics/disk_cache.rs

@@ -1,119 +0,0 @@
-use anyhow::Context;
-use camino::{Utf8Path, Utf8PathBuf};
-use std::sync::Arc;
-
-use super::RawMetric;
-
-pub(super) async fn read_metrics_from_disk(
-    path: Arc<Utf8PathBuf>,
-) -> anyhow::Result<Vec<RawMetric>> {
-    // do not add context to each error, callsite will log with full path
-    let span = tracing::Span::current();
-    tokio::task::spawn_blocking(move || {
-        let _e = span.entered();
-
-        if let Some(parent) = path.parent() {
-            if let Err(e) = scan_and_delete_with_same_prefix(&path) {
-                tracing::info!("failed to cleanup temporary files in {parent:?}: {e:#}");
-            }
-        }
-
-        let mut file = std::fs::File::open(&*path)?;
-        let reader = std::io::BufReader::new(&mut file);
-        anyhow::Ok(serde_json::from_reader::<_, Vec<RawMetric>>(reader)?)
-    })
-    .await
-    .context("read metrics join error")
-    .and_then(|x| x)
-}
-
-fn scan_and_delete_with_same_prefix(path: &Utf8Path) -> std::io::Result<()> {
-    let it = std::fs::read_dir(path.parent().expect("caller checked"))?;
-
-    let prefix = path.file_name().expect("caller checked").to_string();
-
-    for entry in it {
-        let entry = entry?;
-        if !entry.metadata()?.is_file() {
-            continue;
-        }
-        let file_name = entry.file_name();
-
-        if path.file_name().unwrap() == file_name {
-            // do not remove our actual file
-            continue;
-        }
-
-        let file_name = file_name.to_string_lossy();
-
-        if !file_name.starts_with(&*prefix) {
-            continue;
-        }
-
-        let path = entry.path();
-
-        if let Err(e) = std::fs::remove_file(&path) {
-            tracing::warn!("cleaning up old tempfile {file_name:?} failed: {e:#}");
-        } else {
-            tracing::info!("cleaned up old tempfile {file_name:?}");
-        }
-    }
-
-    Ok(())
-}
-
-pub(super) async fn flush_metrics_to_disk(
-    current_metrics: &Arc<Vec<RawMetric>>,
-    path: &Arc<Utf8PathBuf>,
-) -> anyhow::Result<()> {
-    use std::io::Write;
-
-    anyhow::ensure!(path.parent().is_some(), "path must have parent: {path:?}");
-    anyhow::ensure!(
-        path.file_name().is_some(),
-        "path must have filename: {path:?}"
-    );
-
-    let span = tracing::Span::current();
-    tokio::task::spawn_blocking({
-        let current_metrics = current_metrics.clone();
-        let path = path.clone();
-        move || {
-            let _e = span.entered();
-
-            let parent = path.parent().expect("existence checked");
-            let file_name = path.file_name().expect("existence checked");
-            let mut tempfile = camino_tempfile::Builder::new()
-                .prefix(file_name)
-                .suffix(".tmp")
-                .tempfile_in(parent)?;
-
-            tracing::debug!("using tempfile {:?}", tempfile.path());
-
-            // write out all of the raw metrics, to be read out later on restart as cached values
-            {
-                let mut writer = std::io::BufWriter::new(&mut tempfile);
-                serde_json::to_writer(&mut writer, &*current_metrics)
-                    .context("serialize metrics")?;
-                writer
-                    .into_inner()
-                    .map_err(|_| anyhow::anyhow!("flushing metrics failed"))?;
-            }
-
-            tempfile.flush()?;
-            tempfile.as_file().sync_all()?;
-
-            fail::fail_point!("before-persist-last-metrics-collected");
-
-            drop(tempfile.persist(&*path).map_err(|e| e.error)?);
-
-            let f = std::fs::File::open(path.parent().unwrap())?;
-            f.sync_all()?;
-
-            anyhow::Ok(())
-        }
-    })
-    .await
-    .with_context(|| format!("write metrics to {path:?} join error"))
-    .and_then(|x| x.with_context(|| format!("write metrics to {path:?}")))
-}

pageserver/src/consumption_metrics/metrics.rs

@@ -1,455 +0,0 @@
-use crate::context::RequestContext;
-use anyhow::Context;
-use chrono::{DateTime, Utc};
-use consumption_metrics::EventType;
-use futures::stream::StreamExt;
-use serde_with::serde_as;
-use std::{sync::Arc, time::SystemTime};
-use utils::{
-    id::{TenantId, TimelineId},
-    lsn::Lsn,
-};
-
-use super::{Cache, RawMetric};
-
-/// Name of the metric, used by `MetricsKey` factory methods and `deserialize_cached_events`
-/// instead of static str.
-// Do not rename any of these without first consulting with data team and partner
-// management.
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, serde::Serialize, serde::Deserialize)]
-pub(super) enum Name {
-    /// Timeline last_record_lsn, absolute
-    #[serde(rename = "written_size")]
-    WrittenSize,
-    /// Timeline last_record_lsn, incremental
-    #[serde(rename = "written_data_bytes_delta")]
-    WrittenSizeDelta,
-    /// Timeline logical size
-    #[serde(rename = "timeline_logical_size")]
-    LogicalSize,
-    /// Tenant remote size
-    #[serde(rename = "remote_storage_size")]
-    RemoteSize,
-    /// Tenant resident size
-    #[serde(rename = "resident_size")]
-    ResidentSize,
-    /// Tenant synthetic size
-    #[serde(rename = "synthetic_storage_size")]
-    SyntheticSize,
-}
-
-/// Key that uniquely identifies the object this metric describes.
-///
-/// This is a denormalization done at the MetricsKey const methods; these should not be constructed
-/// elsewhere.
-#[serde_with::serde_as]
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, serde::Serialize, serde::Deserialize)]
-pub(crate) struct MetricsKey {
-    #[serde_as(as = "serde_with::DisplayFromStr")]
-    pub(super) tenant_id: TenantId,
-
-    #[serde_as(as = "Option<serde_with::DisplayFromStr>")]
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub(super) timeline_id: Option<TimelineId>,
-
-    pub(super) metric: Name,
-}
-
-impl MetricsKey {
-    const fn absolute_values(self) -> AbsoluteValueFactory {
-        AbsoluteValueFactory(self)
-    }
-    const fn incremental_values(self) -> IncrementalValueFactory {
-        IncrementalValueFactory(self)
-    }
-}
-
-/// Helper type which each individual metric kind can return to produce only absolute values.
-struct AbsoluteValueFactory(MetricsKey);
-
-impl AbsoluteValueFactory {
-    const fn at(self, time: DateTime<Utc>, val: u64) -> RawMetric {
-        let key = self.0;
-        (key, (EventType::Absolute { time }, val))
-    }
-
-    fn key(&self) -> &MetricsKey {
-        &self.0
-    }
-}
-
-/// Helper type which each individual metric kind can return to produce only incremental values.
-struct IncrementalValueFactory(MetricsKey);
-
-impl IncrementalValueFactory {
-    #[allow(clippy::wrong_self_convention)]
-    const fn from_until(
-        self,
-        prev_end: DateTime<Utc>,
-        up_to: DateTime<Utc>,
-        val: u64,
-    ) -> RawMetric {
-        let key = self.0;
-        // cannot assert prev_end < up_to because these are realtime clock based
-        let when = EventType::Incremental {
-            start_time: prev_end,
-            stop_time: up_to,
-        };
-        (key, (when, val))
-    }
-
-    fn key(&self) -> &MetricsKey {
-        &self.0
-    }
-}
-
-// the static part of a MetricsKey
-impl MetricsKey {
-    /// Absolute value of [`Timeline::get_last_record_lsn`].
-    ///
-    /// [`Timeline::get_last_record_lsn`]: crate::tenant::Timeline::get_last_record_lsn
-    const fn written_size(tenant_id: TenantId, timeline_id: TimelineId) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: Some(timeline_id),
-            metric: Name::WrittenSize,
-        }
-        .absolute_values()
-    }
-
-    /// Values will be the difference of the latest [`MetricsKey::written_size`] to what we
-    /// previously sent, starting from the previously sent incremental time range ending at the
-    /// latest absolute measurement.
-    const fn written_size_delta(
-        tenant_id: TenantId,
-        timeline_id: TimelineId,
-    ) -> IncrementalValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: Some(timeline_id),
-            metric: Name::WrittenSizeDelta,
-        }
-        .incremental_values()
-    }
-
-    /// Exact [`Timeline::get_current_logical_size`].
-    ///
-    /// [`Timeline::get_current_logical_size`]: crate::tenant::Timeline::get_current_logical_size
-    const fn timeline_logical_size(
-        tenant_id: TenantId,
-        timeline_id: TimelineId,
-    ) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: Some(timeline_id),
-            metric: Name::LogicalSize,
-        }
-        .absolute_values()
-    }
-
-    /// [`Tenant::remote_size`]
-    ///
-    /// [`Tenant::remote_size`]: crate::tenant::Tenant::remote_size
-    const fn remote_storage_size(tenant_id: TenantId) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: None,
-            metric: Name::RemoteSize,
-        }
-        .absolute_values()
-    }
-
-    /// Sum of [`Timeline::resident_physical_size`] for each `Tenant`.
-    ///
-    /// [`Timeline::resident_physical_size`]: crate::tenant::Timeline::resident_physical_size
-    const fn resident_size(tenant_id: TenantId) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: None,
-            metric: Name::ResidentSize,
-        }
-        .absolute_values()
-    }
-
-    /// [`Tenant::cached_synthetic_size`] as refreshed by [`calculate_synthetic_size_worker`].
-    ///
-    /// [`Tenant::cached_synthetic_size`]: crate::tenant::Tenant::cached_synthetic_size
-    /// [`calculate_synthetic_size_worker`]: super::calculate_synthetic_size_worker
-    const fn synthetic_size(tenant_id: TenantId) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: None,
-            metric: Name::SyntheticSize,
-        }
-        .absolute_values()
-    }
-}
-
-pub(super) async fn collect_all_metrics(
-    cached_metrics: &Cache,
-    ctx: &RequestContext,
-) -> Vec<RawMetric> {
-    use pageserver_api::models::TenantState;
-
-    let started_at = std::time::Instant::now();
-
-    let tenants = match crate::tenant::mgr::list_tenants().await {
-        Ok(tenants) => tenants,
-        Err(err) => {
-            tracing::error!("failed to list tenants: {:?}", err);
-            return vec![];
-        }
-    };
-
-    let tenants = futures::stream::iter(tenants).filter_map(|(id, state)| async move {
-        if state != TenantState::Active {
-            None
-        } else {
-            crate::tenant::mgr::get_tenant(id, true)
-                .await
-                .ok()
-                .map(|tenant| (id, tenant))
-        }
-    });
-
-    let res = collect(tenants, cached_metrics, ctx).await;
-
-    tracing::info!(
-        elapsed_ms = started_at.elapsed().as_millis(),
-        total = res.len(),
-        "collected metrics"
-    );
-
-    res
-}
-
-async fn collect<S>(tenants: S, cache: &Cache, ctx: &RequestContext) -> Vec<RawMetric>
-where
-    S: futures::stream::Stream<Item = (TenantId, Arc<crate::tenant::Tenant>)>,
-{
-    let mut current_metrics: Vec<RawMetric> = Vec::new();
-
-    let mut tenants = std::pin::pin!(tenants);
-
-    while let Some((tenant_id, tenant)) = tenants.next().await {
-        let mut tenant_resident_size = 0;
-
-        for timeline in tenant.list_timelines() {
-            let timeline_id = timeline.timeline_id;
-
-            match TimelineSnapshot::collect(&timeline, ctx) {
-                Ok(Some(snap)) => {
-                    snap.to_metrics(
-                        tenant_id,
-                        timeline_id,
-                        Utc::now(),
-                        &mut current_metrics,
-                        cache,
-                    );
-                }
-                Ok(None) => {}
-                Err(e) => {
-                    tracing::error!(
-                        "failed to get metrics values for tenant {tenant_id} timeline {}: {e:#?}",
-                        timeline.timeline_id
-                    );
-                    continue;
-                }
-            }
-
-            tenant_resident_size += timeline.resident_physical_size();
-        }
-
-        let snap = TenantSnapshot::collect(&tenant, tenant_resident_size);
-        snap.to_metrics(tenant_id, Utc::now(), cache, &mut current_metrics);
-    }
-
-    current_metrics
-}
-
-/// In-between abstraction to allow testing metrics without actual Tenants.
-struct TenantSnapshot {
-    resident_size: u64,
-    remote_size: u64,
-    synthetic_size: u64,
-}
-
-impl TenantSnapshot {
-    /// Collect tenant status to have metrics created out of it.
-    ///
-    /// `resident_size` is calculated of the timelines we had access to for other metrics, so we
-    /// cannot just list timelines here.
-    fn collect(t: &Arc<crate::tenant::Tenant>, resident_size: u64) -> Self {
-        TenantSnapshot {
-            resident_size,
-            remote_size: t.remote_size(),
-            // Note that this metric is calculated in a separate bgworker
-            // Here we only use cached value, which may lag behind the real latest one
-            synthetic_size: t.cached_synthetic_size(),
-        }
-    }
-
-    fn to_metrics(
-        &self,
-        tenant_id: TenantId,
-        now: DateTime<Utc>,
-        cached: &Cache,
-        metrics: &mut Vec<RawMetric>,
-    ) {
-        let remote_size = MetricsKey::remote_storage_size(tenant_id).at(now, self.remote_size);
-
-        let resident_size = MetricsKey::resident_size(tenant_id).at(now, self.resident_size);
-
-        let synthetic_size = {
-            let factory = MetricsKey::synthetic_size(tenant_id);
-            let mut synthetic_size = self.synthetic_size;
-
-            if synthetic_size == 0 {
-                if let Some((_, value)) = cached.get(factory.key()) {
-                    // use the latest value from previous session
-                    synthetic_size = *value;
-                }
-            }
-
-            if synthetic_size != 0 {
-                // only send non-zeroes because otherwise these show up as errors in logs
-                Some(factory.at(now, synthetic_size))
-            } else {
-                None
-            }
-        };
-
-        metrics.extend(
-            [Some(remote_size), Some(resident_size), synthetic_size]
-                .into_iter()
-                .flatten(),
-        );
-    }
-}
-
-/// Internal type to make timeline metric production testable.
-///
-/// As this value type contains all of the information needed from a timeline to produce the
-/// metrics, it can easily be created with different values in test.
-struct TimelineSnapshot {
-    loaded_at: (Lsn, SystemTime),
-    last_record_lsn: Lsn,
-    current_exact_logical_size: Option<u64>,
-}
-
-impl TimelineSnapshot {
-    /// Collect the metrics from an actual timeline.
-    ///
-    /// Fails currently only when [`Timeline::get_current_logical_size`] fails.
-    ///
-    /// [`Timeline::get_current_logical_size`]: crate::tenant::Timeline::get_current_logical_size
-    fn collect(
-        t: &Arc<crate::tenant::Timeline>,
-        ctx: &RequestContext,
-    ) -> anyhow::Result<Option<Self>> {
-        if !t.is_active() {
-            // no collection for broken or stopping needed, we will still keep the cached values
-            // though at the caller.
-            Ok(None)
-        } else {
-            let loaded_at = t.loaded_at;
-            let last_record_lsn = t.get_last_record_lsn();
-
-            let current_exact_logical_size = {
-                let span = tracing::info_span!("collect_metrics_iteration", tenant_id = %t.tenant_id, timeline_id = %t.timeline_id);
-                let res = span
-                    .in_scope(|| t.get_current_logical_size(ctx))
-                    .context("get_current_logical_size");
-                match res? {
-                    // Only send timeline logical size when it is fully calculated.
-                    (size, is_exact) if is_exact => Some(size),
-                    (_, _) => None,
-                }
-            };
-
-            Ok(Some(TimelineSnapshot {
-                loaded_at,
-                last_record_lsn,
-                current_exact_logical_size,
-            }))
-        }
-    }
-
-    /// Produce the timeline consumption metrics into the `metrics` argument.
-    fn to_metrics(
-        &self,
-        tenant_id: TenantId,
-        timeline_id: TimelineId,
-        now: DateTime<Utc>,
-        metrics: &mut Vec<RawMetric>,
-        cache: &Cache,
-    ) {
-        let timeline_written_size = u64::from(self.last_record_lsn);
-
-        let written_size_delta_key = MetricsKey::written_size_delta(tenant_id, timeline_id);
-
-        let last_stop_time = cache
-            .get(written_size_delta_key.key())
-            .map(|(until, _val)| {
-                until
-                    .incremental_timerange()
-                    .expect("never create EventType::Absolute for written_size_delta")
-                    .end
-            });
-
-        let (key, written_size_now) =
-            MetricsKey::written_size(tenant_id, timeline_id).at(now, timeline_written_size);
-
-        // by default, use the last sent written_size as the basis for
-        // calculating the delta. if we don't yet have one, use the load time value.
-        let prev = cache
-            .get(&key)
-            .map(|(prev_at, prev)| {
-                // use the prev time from our last incremental update, or default to latest
-                // absolute update on the first round.
-                let prev_at = prev_at
-                    .absolute_time()
-                    .expect("never create EventType::Incremental for written_size");
-                let prev_at = last_stop_time.unwrap_or(prev_at);
-                (*prev_at, *prev)
-            })
-            .unwrap_or_else(|| {
-                // if we don't have a previous point of comparison, compare to the load time
-                // lsn.
-                let (disk_consistent_lsn, loaded_at) = &self.loaded_at;
-                (DateTime::from(*loaded_at), disk_consistent_lsn.0)
-            });
-
-        let up_to = now;
-
-        if let Some(delta) = written_size_now.1.checked_sub(prev.1) {
-            let key_value = written_size_delta_key.from_until(prev.0, up_to, delta);
-            // written_size_delta
-            metrics.push(key_value);
-            // written_size
-            metrics.push((key, written_size_now));
-        } else {
-            // the cached value was ahead of us, report zero until we've caught up
-            metrics.push(written_size_delta_key.from_until(prev.0, up_to, 0));
-            // the cached value was ahead of us, report the same until we've caught up
-            metrics.push((key, (written_size_now.0, prev.1)));
-        }
-
-        {
-            let factory = MetricsKey::timeline_logical_size(tenant_id, timeline_id);
-            let current_or_previous = self
-                .current_exact_logical_size
-                .or_else(|| cache.get(factory.key()).map(|(_, val)| *val));
-
-            if let Some(size) = current_or_previous {
-                metrics.push(factory.at(now, size));
-            }
-        }
-    }
-}
-
-#[cfg(test)]
-mod tests;
-
-#[cfg(test)]
-pub(crate) use tests::metric_examples;

pageserver/src/consumption_metrics/metrics/tests.rs

@@ -1,297 +0,0 @@
-use super::*;
-use std::collections::HashMap;
-use std::time::SystemTime;
-use utils::lsn::Lsn;
-
-#[test]
-fn startup_collected_timeline_metrics_before_advancing() {
-    let tenant_id = TenantId::generate();
-    let timeline_id = TimelineId::generate();
-
-    let mut metrics = Vec::new();
-    let cache = HashMap::new();
-
-    let initdb_lsn = Lsn(0x10000);
-    let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
-
-    let snap = TimelineSnapshot {
-        loaded_at: (disk_consistent_lsn, SystemTime::now()),
-        last_record_lsn: disk_consistent_lsn,
-        current_exact_logical_size: Some(0x42000),
-    };
-
-    let now = DateTime::<Utc>::from(SystemTime::now());
-
-    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
-
-    assert_eq!(
-        metrics,
-        &[
-            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(
-                snap.loaded_at.1.into(),
-                now,
-                0
-            ),
-            MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
-            MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
-        ]
-    );
-}
-
-#[test]
-fn startup_collected_timeline_metrics_second_round() {
-    let tenant_id = TenantId::generate();
-    let timeline_id = TimelineId::generate();
-
-    let [now, before, init] = time_backwards();
-
-    let now = DateTime::<Utc>::from(now);
-    let before = DateTime::<Utc>::from(before);
-
-    let initdb_lsn = Lsn(0x10000);
-    let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
-
-    let mut metrics = Vec::new();
-    let cache = HashMap::from([
-        MetricsKey::written_size(tenant_id, timeline_id).at(before, disk_consistent_lsn.0)
-    ]);
-
-    let snap = TimelineSnapshot {
-        loaded_at: (disk_consistent_lsn, init),
-        last_record_lsn: disk_consistent_lsn,
-        current_exact_logical_size: Some(0x42000),
-    };
-
-    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
-
-    assert_eq!(
-        metrics,
-        &[
-            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(before, now, 0),
-            MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
-            MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
-        ]
-    );
-}
-
-#[test]
-fn startup_collected_timeline_metrics_nth_round_at_same_lsn() {
-    let tenant_id = TenantId::generate();
-    let timeline_id = TimelineId::generate();
-
-    let [now, just_before, before, init] = time_backwards();
-
-    let now = DateTime::<Utc>::from(now);
-    let just_before = DateTime::<Utc>::from(just_before);
-    let before = DateTime::<Utc>::from(before);
-
-    let initdb_lsn = Lsn(0x10000);
-    let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
-
-    let mut metrics = Vec::new();
-    let cache = HashMap::from([
-        // at t=before was the last time the last_record_lsn changed
-        MetricsKey::written_size(tenant_id, timeline_id).at(before, disk_consistent_lsn.0),
-        // end time of this event is used for the next ones
-        MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(before, just_before, 0),
-    ]);
-
-    let snap = TimelineSnapshot {
-        loaded_at: (disk_consistent_lsn, init),
-        last_record_lsn: disk_consistent_lsn,
-        current_exact_logical_size: Some(0x42000),
-    };
-
-    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
-
-    assert_eq!(
-        metrics,
-        &[
-            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(just_before, now, 0),
-            MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
-            MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
-        ]
-    );
-}
-
-#[test]
-fn post_restart_written_sizes_with_rolled_back_last_record_lsn() {
-    // it can happen that we lose the inmemorylayer but have previously sent metrics and we
-    // should never go backwards
-
-    let tenant_id = TenantId::generate();
-    let timeline_id = TimelineId::generate();
-
-    let [later, now, at_restart] = time_backwards();
-
-    // FIXME: tests would be so much easier if we did not need to juggle back and forth
-    // SystemTime and DateTime::<Utc> ... Could do the conversion only at upload time?
-    let now = DateTime::<Utc>::from(now);
-    let later = DateTime::<Utc>::from(later);
-    let before_restart = at_restart - std::time::Duration::from_secs(5 * 60);
-    let way_before = before_restart - std::time::Duration::from_secs(10 * 60);
-    let before_restart = DateTime::<Utc>::from(before_restart);
-    let way_before = DateTime::<Utc>::from(way_before);
-
-    let snap = TimelineSnapshot {
-        loaded_at: (Lsn(50), at_restart),
-        last_record_lsn: Lsn(50),
-        current_exact_logical_size: None,
-    };
-
-    let mut cache = HashMap::from([
-        MetricsKey::written_size(tenant_id, timeline_id).at(before_restart, 100),
-        MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(
-            way_before,
-            before_restart,
-            // not taken into account, but the timestamps are important
-            999_999_999,
-        ),
-    ]);
-
-    let mut metrics = Vec::new();
-    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
-
-    assert_eq!(
-        metrics,
-        &[
-            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(
-                before_restart,
-                now,
-                0
-            ),
-            MetricsKey::written_size(tenant_id, timeline_id).at(now, 100),
-        ]
-    );
-
-    // now if we cache these metrics, and re-run while "still in recovery"
-    cache.extend(metrics.drain(..));
-
-    // "still in recovery", because our snapshot did not change
-    snap.to_metrics(tenant_id, timeline_id, later, &mut metrics, &cache);
-
-    assert_eq!(
-        metrics,
-        &[
-            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(now, later, 0),
-            MetricsKey::written_size(tenant_id, timeline_id).at(later, 100),
-        ]
-    );
-}
-
-#[test]
-fn post_restart_current_exact_logical_size_uses_cached() {
-    let tenant_id = TenantId::generate();
-    let timeline_id = TimelineId::generate();
-
-    let [now, at_restart] = time_backwards();
-
-    let now = DateTime::<Utc>::from(now);
-    let before_restart = at_restart - std::time::Duration::from_secs(5 * 60);
-    let before_restart = DateTime::<Utc>::from(before_restart);
-
-    let snap = TimelineSnapshot {
-        loaded_at: (Lsn(50), at_restart),
-        last_record_lsn: Lsn(50),
-        current_exact_logical_size: None,
-    };
-
-    let cache = HashMap::from([
-        MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(before_restart, 100)
-    ]);
-
-    let mut metrics = Vec::new();
-    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
-
-    metrics.retain(|(key, _)| key.metric == Name::LogicalSize);
-
-    assert_eq!(
-        metrics,
-        &[MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 100)]
-    );
-}
-
-#[test]
-fn post_restart_synthetic_size_uses_cached_if_available() {
-    let tenant_id = TenantId::generate();
-
-    let ts = TenantSnapshot {
-        resident_size: 1000,
-        remote_size: 1000,
-        // not yet calculated
-        synthetic_size: 0,
-    };
-
-    let now = SystemTime::now();
-    let before_restart = DateTime::<Utc>::from(now - std::time::Duration::from_secs(5 * 60));
-    let now = DateTime::<Utc>::from(now);
-
-    let cached = HashMap::from([MetricsKey::synthetic_size(tenant_id).at(before_restart, 1000)]);
-
-    let mut metrics = Vec::new();
-    ts.to_metrics(tenant_id, now, &cached, &mut metrics);
-
-    assert_eq!(
-        metrics,
-        &[
-            MetricsKey::remote_storage_size(tenant_id).at(now, 1000),
-            MetricsKey::resident_size(tenant_id).at(now, 1000),
-            MetricsKey::synthetic_size(tenant_id).at(now, 1000),
-        ]
-    );
-}
-
-#[test]
-fn post_restart_synthetic_size_is_not_sent_when_not_cached() {
-    let tenant_id = TenantId::generate();
-
-    let ts = TenantSnapshot {
-        resident_size: 1000,
-        remote_size: 1000,
-        // not yet calculated
-        synthetic_size: 0,
-    };
-
-    let now = SystemTime::now();
-    let now = DateTime::<Utc>::from(now);
-
-    let cached = HashMap::new();
-
-    let mut metrics = Vec::new();
-    ts.to_metrics(tenant_id, now, &cached, &mut metrics);
-
-    assert_eq!(
-        metrics,
-        &[
-            MetricsKey::remote_storage_size(tenant_id).at(now, 1000),
-            MetricsKey::resident_size(tenant_id).at(now, 1000),
-            // no synthetic size here
-        ]
-    );
-}
-
-fn time_backwards<const N: usize>() -> [std::time::SystemTime; N] {
-    let mut times = [std::time::SystemTime::UNIX_EPOCH; N];
-    times[0] = std::time::SystemTime::now();
-    for behind in 1..N {
-        times[behind] = times[0] - std::time::Duration::from_secs(behind as u64);
-    }
-
-    times
-}
-
-pub(crate) const fn metric_examples(
-    tenant_id: TenantId,
-    timeline_id: TimelineId,
-    now: DateTime<Utc>,
-    before: DateTime<Utc>,
-) -> [RawMetric; 6] {
-    [
-        MetricsKey::written_size(tenant_id, timeline_id).at(now, 0),
-        MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(before, now, 0),
-        MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0),
-        MetricsKey::remote_storage_size(tenant_id).at(now, 0),
-        MetricsKey::resident_size(tenant_id).at(now, 0),
-        MetricsKey::synthetic_size(tenant_id).at(now, 1),
-    ]
-}

pageserver/src/consumption_metrics/upload.rs

@@ -1,443 +0,0 @@
-use consumption_metrics::{Event, EventChunk, IdempotencyKey, CHUNK_SIZE};
-use serde_with::serde_as;
-use tokio_util::sync::CancellationToken;
-use tracing::Instrument;
-
-use super::{metrics::Name, Cache, MetricsKey, RawMetric};
-use utils::id::{TenantId, TimelineId};
-
-/// How the metrics from pageserver are identified.
-#[serde_with::serde_as]
-#[derive(serde::Serialize, serde::Deserialize, Debug, Clone, Copy, PartialEq)]
-struct Ids {
-    #[serde_as(as = "serde_with::DisplayFromStr")]
-    pub(super) tenant_id: TenantId,
-    #[serde_as(as = "Option<serde_with::DisplayFromStr>")]
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub(super) timeline_id: Option<TimelineId>,
-}
-
-#[tracing::instrument(skip_all, fields(metrics_total = %metrics.len()))]
-pub(super) async fn upload_metrics(
-    client: &reqwest::Client,
-    metric_collection_endpoint: &reqwest::Url,
-    cancel: &CancellationToken,
-    node_id: &str,
-    metrics: &[RawMetric],
-    cached_metrics: &mut Cache,
-) -> anyhow::Result<()> {
-    let mut uploaded = 0;
-    let mut failed = 0;
-
-    let started_at = std::time::Instant::now();
-
-    let mut iter = serialize_in_chunks(CHUNK_SIZE, metrics, node_id);
-
-    while let Some(res) = iter.next() {
-        let (chunk, body) = res?;
-
-        let event_bytes = body.len();
-
-        let is_last = iter.len() == 0;
-
-        let res = upload(client, metric_collection_endpoint, body, cancel, is_last)
-            .instrument(tracing::info_span!(
-                "upload",
-                %event_bytes,
-                uploaded,
-                total = metrics.len(),
-            ))
-            .await;
-
-        match res {
-            Ok(()) => {
-                for (curr_key, curr_val) in chunk {
-                    cached_metrics.insert(*curr_key, *curr_val);
-                }
-                uploaded += chunk.len();
-            }
-            Err(_) => {
-                // failure(s) have already been logged
-                //
-                // however this is an inconsistency: if we crash here, we will start with the
-                // values as uploaded. in practice, the rejections no longer happen.
-                failed += chunk.len();
-            }
-        }
-    }
-
-    let elapsed = started_at.elapsed();
-
-    tracing::info!(
-        uploaded,
-        failed,
-        elapsed_ms = elapsed.as_millis(),
-        "done sending metrics"
-    );
-
-    Ok(())
-}
-
-// The return type is quite ugly, but we gain testability in isolation
-fn serialize_in_chunks<'a, F>(
-    chunk_size: usize,
-    input: &'a [RawMetric],
-    factory: F,
-) -> impl ExactSizeIterator<Item = Result<(&'a [RawMetric], bytes::Bytes), serde_json::Error>> + 'a
-where
-    F: KeyGen<'a> + 'a,
-{
-    use bytes::BufMut;
-
-    struct Iter<'a, F> {
-        inner: std::slice::Chunks<'a, RawMetric>,
-        chunk_size: usize,
-
-        // write to a BytesMut so that we can cheaply clone the frozen Bytes for retries
-        buffer: bytes::BytesMut,
-        // chunk amount of events are reused to produce the serialized document
-        scratch: Vec<Event<Ids, Name>>,
-        factory: F,
-    }
-
-    impl<'a, F: KeyGen<'a>> Iterator for Iter<'a, F> {
-        type Item = Result<(&'a [RawMetric], bytes::Bytes), serde_json::Error>;
-
-        fn next(&mut self) -> Option<Self::Item> {
-            let chunk = self.inner.next()?;
-
-            if self.scratch.is_empty() {
-                // first round: create events with N strings
-                self.scratch.extend(
-                    chunk
-                        .iter()
-                        .map(|raw_metric| raw_metric.as_event(&self.factory.generate())),
-                );
-            } else {
-                // next rounds: update_in_place to reuse allocations
-                assert_eq!(self.scratch.len(), self.chunk_size);
-                self.scratch
-                    .iter_mut()
-                    .zip(chunk.iter())
-                    .for_each(|(slot, raw_metric)| {
-                        raw_metric.update_in_place(slot, &self.factory.generate())
-                    });
-            }
-
-            let res = serde_json::to_writer(
-                (&mut self.buffer).writer(),
-                &EventChunk {
-                    events: (&self.scratch[..chunk.len()]).into(),
-                },
-            );
-
-            match res {
-                Ok(()) => Some(Ok((chunk, self.buffer.split().freeze()))),
-                Err(e) => Some(Err(e)),
-            }
-        }
-
-        fn size_hint(&self) -> (usize, Option<usize>) {
-            self.inner.size_hint()
-        }
-    }
-
-    impl<'a, F: KeyGen<'a>> ExactSizeIterator for Iter<'a, F> {}
-
-    let buffer = bytes::BytesMut::new();
-    let inner = input.chunks(chunk_size);
-    let scratch = Vec::new();
-
-    Iter {
-        inner,
-        chunk_size,
-        buffer,
-        scratch,
-        factory,
-    }
-}
-
-trait RawMetricExt {
-    fn as_event(&self, key: &IdempotencyKey<'_>) -> Event<Ids, Name>;
-    fn update_in_place(&self, event: &mut Event<Ids, Name>, key: &IdempotencyKey<'_>);
-}
-
-impl RawMetricExt for RawMetric {
-    fn as_event(&self, key: &IdempotencyKey<'_>) -> Event<Ids, Name> {
-        let MetricsKey {
-            metric,
-            tenant_id,
-            timeline_id,
-        } = self.0;
-
-        let (kind, value) = self.1;
-
-        Event {
-            kind,
-            metric,
-            idempotency_key: key.to_string(),
-            value,
-            extra: Ids {
-                tenant_id,
-                timeline_id,
-            },
-        }
-    }
-
-    fn update_in_place(&self, event: &mut Event<Ids, Name>, key: &IdempotencyKey<'_>) {
-        use std::fmt::Write;
-
-        let MetricsKey {
-            metric,
-            tenant_id,
-            timeline_id,
-        } = self.0;
-
-        let (kind, value) = self.1;
-
-        *event = Event {
-            kind,
-            metric,
-            idempotency_key: {
-                event.idempotency_key.clear();
-                write!(event.idempotency_key, "{key}").unwrap();
-                std::mem::take(&mut event.idempotency_key)
-            },
-            value,
-            extra: Ids {
-                tenant_id,
-                timeline_id,
-            },
-        };
-    }
-}
-
-trait KeyGen<'a>: Copy {
-    fn generate(&self) -> IdempotencyKey<'a>;
-}
-
-impl<'a> KeyGen<'a> for &'a str {
-    fn generate(&self) -> IdempotencyKey<'a> {
-        IdempotencyKey::generate(self)
-    }
-}
-
-enum UploadError {
-    Rejected(reqwest::StatusCode),
-    Reqwest(reqwest::Error),
-    Cancelled,
-}
-
-impl std::fmt::Debug for UploadError {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        // use same impl because backoff::retry will log this using both
-        std::fmt::Display::fmt(self, f)
-    }
-}
-
-impl std::fmt::Display for UploadError {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        use UploadError::*;
-
-        match self {
-            Rejected(code) => write!(f, "server rejected the metrics with {code}"),
-            Reqwest(e) => write!(f, "request failed: {e}"),
-            Cancelled => write!(f, "cancelled"),
-        }
-    }
-}
-
-impl UploadError {
-    fn is_reject(&self) -> bool {
-        matches!(self, UploadError::Rejected(_))
-    }
-}
-
-// this is consumed by the test verifiers
-static LAST_IN_BATCH: reqwest::header::HeaderName =
-    reqwest::header::HeaderName::from_static("pageserver-metrics-last-upload-in-batch");
-
-async fn upload(
-    client: &reqwest::Client,
-    metric_collection_endpoint: &reqwest::Url,
-    body: bytes::Bytes,
-    cancel: &CancellationToken,
-    is_last: bool,
-) -> Result<(), UploadError> {
-    let warn_after = 3;
-    let max_attempts = 10;
-    let res = utils::backoff::retry(
-        move || {
-            let body = body.clone();
-            async move {
-                let res = client
-                    .post(metric_collection_endpoint.clone())
-                    .header(reqwest::header::CONTENT_TYPE, "application/json")
-                    .header(
-                        LAST_IN_BATCH.clone(),
-                        if is_last { "true" } else { "false" },
-                    )
-                    .body(body)
-                    .send()
-                    .await;
-
-                let res = res.and_then(|res| res.error_for_status());
-
-                // 10 redirects are normally allowed, so we don't need worry about 3xx
-                match res {
-                    Ok(_response) => Ok(()),
-                    Err(e) => {
-                        let status = e.status().filter(|s| s.is_client_error());
-                        if let Some(status) = status {
-                            // rejection used to be a thing when the server could reject a
-                            // whole batch of metrics if one metric was bad.
-                            Err(UploadError::Rejected(status))
-                        } else {
-                            Err(UploadError::Reqwest(e))
-                        }
-                    }
-                }
-            }
-        },
-        UploadError::is_reject,
-        warn_after,
-        max_attempts,
-        "upload consumption_metrics",
-        utils::backoff::Cancel::new(cancel.clone(), || UploadError::Cancelled),
-    )
-    .await;
-
-    match &res {
-        Ok(_) => {}
-        Err(e) if e.is_reject() => {
-            // permanent errors currently do not get logged by backoff::retry
-            // display alternate has no effect, but keeping it here for easier pattern matching.
-            tracing::error!("failed to upload metrics: {e:#}");
-        }
-        Err(_) => {
-            // these have been logged already
-        }
-    }
-
-    res
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use chrono::{DateTime, Utc};
-    use once_cell::sync::Lazy;
-
-    #[test]
-    fn chunked_serialization() {
-        let examples = metric_samples();
-        assert!(examples.len() > 1);
-
-        let factory = FixedGen::new(Utc::now(), "1", 42);
-
-        // need to use Event here because serde_json::Value uses default hashmap, not linked
-        // hashmap
-        #[derive(serde::Deserialize)]
-        struct EventChunk {
-            events: Vec<Event<Ids, Name>>,
-        }
-
-        let correct = serialize_in_chunks(examples.len(), &examples, factory)
-            .map(|res| res.unwrap().1)
-            .flat_map(|body| serde_json::from_slice::<EventChunk>(&body).unwrap().events)
-            .collect::<Vec<_>>();
-
-        for chunk_size in 1..examples.len() {
-            let actual = serialize_in_chunks(chunk_size, &examples, factory)
-                .map(|res| res.unwrap().1)
-                .flat_map(|body| serde_json::from_slice::<EventChunk>(&body).unwrap().events)
-                .collect::<Vec<_>>();
-
-            // if these are equal, it means that multi-chunking version works as well
-            assert_eq!(correct, actual);
-        }
-    }
-
-    #[derive(Clone, Copy)]
-    struct FixedGen<'a>(chrono::DateTime<chrono::Utc>, &'a str, u16);
-
-    impl<'a> FixedGen<'a> {
-        fn new(now: chrono::DateTime<chrono::Utc>, node_id: &'a str, nonce: u16) -> Self {
-            FixedGen(now, node_id, nonce)
-        }
-    }
-
-    impl<'a> KeyGen<'a> for FixedGen<'a> {
-        fn generate(&self) -> IdempotencyKey<'a> {
-            IdempotencyKey::for_tests(self.0, self.1, self.2)
-        }
-    }
-
-    static SAMPLES_NOW: Lazy<DateTime<Utc>> = Lazy::new(|| {
-        DateTime::parse_from_rfc3339("2023-09-15T00:00:00.123456789Z")
-            .unwrap()
-            .into()
-    });
-
-    #[test]
-    fn metric_image_stability() {
-        // it is important that these strings stay as they are
-
-        let examples = [
-            (
-                line!(),
-                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"written_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000","timeline_id":"ffffffffffffffffffffffffffffffff"}"#,
-            ),
-            (
-                line!(),
-                r#"{"type":"incremental","start_time":"2023-09-14T00:00:00.123456789Z","stop_time":"2023-09-15T00:00:00.123456789Z","metric":"written_data_bytes_delta","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000","timeline_id":"ffffffffffffffffffffffffffffffff"}"#,
-            ),
-            (
-                line!(),
-                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"timeline_logical_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000","timeline_id":"ffffffffffffffffffffffffffffffff"}"#,
-            ),
-            (
-                line!(),
-                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"remote_storage_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000"}"#,
-            ),
-            (
-                line!(),
-                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"resident_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000"}"#,
-            ),
-            (
-                line!(),
-                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"synthetic_storage_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":1,"tenant_id":"00000000000000000000000000000000"}"#,
-            ),
-        ];
-
-        let idempotency_key = consumption_metrics::IdempotencyKey::for_tests(*SAMPLES_NOW, "1", 0);
-        let examples = examples.into_iter().zip(metric_samples());
-
-        for ((line, expected), (key, (kind, value))) in examples {
-            let e = consumption_metrics::Event {
-                kind,
-                metric: key.metric,
-                idempotency_key: idempotency_key.to_string(),
-                value,
-                extra: Ids {
-                    tenant_id: key.tenant_id,
-                    timeline_id: key.timeline_id,
-                },
-            };
-            let actual = serde_json::to_string(&e).unwrap();
-            assert_eq!(expected, actual, "example for {kind:?} from line {line}");
-        }
-    }
-
-    fn metric_samples() -> [RawMetric; 6] {
-        let tenant_id = TenantId::from_array([0; 16]);
-        let timeline_id = TimelineId::from_array([0xff; 16]);
-
-        let before = DateTime::parse_from_rfc3339("2023-09-14T00:00:00.123456789Z")
-            .unwrap()
-            .into();
-        let [now, before] = [*SAMPLES_NOW, before];
-
-        super::super::metrics::metric_examples(tenant_id, timeline_id, now, before)
-    }
-}

pageserver/src/context.rs

@@ -94,18 +94,6 @@ pub struct RequestContext {
     task_kind: TaskKind,
     download_behavior: DownloadBehavior,
     access_stats_behavior: AccessStatsBehavior,
-    page_content_kind: PageContentKind,
-}
-
-/// The kind of access to the page cache.
-#[derive(Clone, Copy, PartialEq, Eq, Debug, enum_map::Enum, strum_macros::IntoStaticStr)]
-pub enum PageContentKind {
-    Unknown,
-    DeltaLayerBtreeNode,
-    DeltaLayerValue,
-    ImageLayerBtreeNode,
-    ImageLayerValue,
-    InMemoryLayer,
 }
 
 /// Desired behavior if the operation requires an on-demand download
@@ -149,7 +137,6 @@ impl RequestContextBuilder {
                 task_kind,
                 download_behavior: DownloadBehavior::Download,
                 access_stats_behavior: AccessStatsBehavior::Update,
-                page_content_kind: PageContentKind::Unknown,
             },
         }
     }
@@ -162,7 +149,6 @@ impl RequestContextBuilder {
                 task_kind: original.task_kind,
                 download_behavior: original.download_behavior,
                 access_stats_behavior: original.access_stats_behavior,
-                page_content_kind: original.page_content_kind,
             },
         }
     }
@@ -181,11 +167,6 @@ impl RequestContextBuilder {
         self
     }
 
-    pub(crate) fn page_content_kind(mut self, k: PageContentKind) -> Self {
-        self.inner.page_content_kind = k;
-        self
-    }
-
     pub fn build(self) -> RequestContext {
         self.inner
     }
@@ -282,8 +263,4 @@ impl RequestContext {
     pub(crate) fn access_stats_behavior(&self) -> AccessStatsBehavior {
         self.access_stats_behavior
     }
-
-    pub(crate) fn page_content_kind(&self) -> PageContentKind {
-        self.page_content_kind
-    }
 }

pageserver/src/control_plane_client.rs

@@ -1,9 +1,7 @@
 use std::collections::HashMap;
 
-use pageserver_api::control_api::{
-    ReAttachRequest, ReAttachResponse, ValidateRequest, ValidateRequestTenant, ValidateResponse,
-};
-use serde::{de::DeserializeOwned, Serialize};
+use hyper::StatusCode;
+use pageserver_api::control_api::{ReAttachRequest, ReAttachResponse};
 use tokio_util::sync::CancellationToken;
 use url::Url;
 use utils::{
@@ -14,34 +12,25 @@ use utils::{
 
 use crate::config::PageServerConf;
 
+// Backoffs when control plane requests do not succeed: compromise between reducing load
+// on control plane, and retrying frequently when we are blocked on a control plane
+// response to make progress.
+const BACKOFF_INCREMENT: f64 = 0.1;
+const BACKOFF_MAX: f64 = 10.0;
+
 /// The Pageserver's client for using the control plane API: this is a small subset
 /// of the overall control plane API, for dealing with generations (see docs/rfcs/025-generation-numbers.md)
-pub struct ControlPlaneClient {
+pub(crate) struct ControlPlaneClient {
     http_client: reqwest::Client,
     base_url: Url,
     node_id: NodeId,
     cancel: CancellationToken,
 }
 
-/// Represent operations which internally retry on all errors other than
-/// cancellation token firing: the only way they can fail is ShuttingDown.
-pub enum RetryForeverError {
-    ShuttingDown,
-}
-
-#[async_trait::async_trait]
-pub trait ControlPlaneGenerationsApi {
-    async fn re_attach(&self) -> Result<HashMap<TenantId, Generation>, RetryForeverError>;
-    async fn validate(
-        &self,
-        tenants: Vec<(TenantId, Generation)>,
-    ) -> Result<HashMap<TenantId, bool>, RetryForeverError>;
-}
-
 impl ControlPlaneClient {
     /// A None return value indicates that the input `conf` object does not have control
     /// plane API enabled.
-    pub fn new(conf: &'static PageServerConf, cancel: &CancellationToken) -> Option<Self> {
+    pub(crate) fn new(conf: &'static PageServerConf, cancel: &CancellationToken) -> Option<Self> {
         let mut url = match conf.control_plane_api.as_ref() {
             Some(u) => u.clone(),
             None => return None,
@@ -53,78 +42,39 @@ impl ControlPlaneClient {
             segs.pop_if_empty().push("");
         }
 
-        let mut client = reqwest::ClientBuilder::new();
-
-        if let Some(jwt) = &conf.control_plane_api_token {
-            let mut headers = hyper::HeaderMap::new();
-            headers.insert("Authorization", jwt.get_contents().parse().unwrap());
-            client = client.default_headers(headers);
-        }
+        let client = reqwest::ClientBuilder::new()
+            .build()
+            .expect("Failed to construct http client");
 
         Some(Self {
-            http_client: client.build().expect("Failed to construct HTTP client"),
+            http_client: client,
             base_url: url,
             node_id: conf.id,
             cancel: cancel.clone(),
         })
     }
 
-    async fn retry_http_forever<R, T>(
+    async fn try_re_attach(
         &self,
-        url: &url::Url,
-        request: R,
-    ) -> Result<T, RetryForeverError>
-    where
-        R: Serialize,
-        T: DeserializeOwned,
-    {
-        #[derive(thiserror::Error, Debug)]
-        enum RemoteAttemptError {
-            #[error("shutdown")]
-            Shutdown,
-            #[error("remote: {0}")]
-            Remote(reqwest::Error),
-        }
-
-        match backoff::retry(
-            || async {
-                let response = self
-                    .http_client
-                    .post(url.clone())
-                    .json(&request)
-                    .send()
-                    .await
-                    .map_err(RemoteAttemptError::Remote)?;
-
-                response
-                    .error_for_status_ref()
-                    .map_err(RemoteAttemptError::Remote)?;
-                response
-                    .json::<T>()
-                    .await
-                    .map_err(RemoteAttemptError::Remote)
-            },
-            |_| false,
-            3,
-            u32::MAX,
-            "calling control plane generation validation API",
-            backoff::Cancel::new(self.cancel.clone(), || RemoteAttemptError::Shutdown),
-        )
-        .await
-        {
-            Err(RemoteAttemptError::Shutdown) => Err(RetryForeverError::ShuttingDown),
-            Err(RemoteAttemptError::Remote(_)) => {
-                panic!("We retry forever, this should never be reached");
+        url: Url,
+        request: &ReAttachRequest,
+    ) -> anyhow::Result<ReAttachResponse> {
+        match self.http_client.post(url).json(request).send().await {
+            Err(e) => Err(anyhow::Error::from(e)),
+            Ok(r) => {
+                if r.status() == StatusCode::OK {
+                    r.json::<ReAttachResponse>()
+                        .await
+                        .map_err(anyhow::Error::from)
+                } else {
+                    Err(anyhow::anyhow!("Unexpected status {}", r.status()))
+                }
             }
-            Ok(r) => Ok(r),
         }
     }
-}
 
-#[async_trait::async_trait]
-impl ControlPlaneGenerationsApi for ControlPlaneClient {
-    /// Block until we get a successful response, or error out if we are shut down
-    async fn re_attach(&self) -> Result<HashMap<TenantId, Generation>, RetryForeverError> {
+    /// Block until we get a successful response
+    pub(crate) async fn re_attach(&self) -> anyhow::Result<HashMap<TenantId, Generation>> {
         let re_attach_path = self
             .base_url
             .join("re-attach")
@@ -133,47 +83,37 @@ impl ControlPlaneGenerationsApi for ControlPlaneClient {
             node_id: self.node_id,
         };
 
-        let response: ReAttachResponse = self.retry_http_forever(&re_attach_path, request).await?;
-        tracing::info!(
-            "Received re-attach response with {} tenants",
-            response.tenants.len()
-        );
+        let mut attempt = 0;
+        loop {
+            let result = self.try_re_attach(re_attach_path.clone(), &request).await;
+            match result {
+                Ok(res) => {
+                    tracing::info!(
+                        "Received re-attach response with {} tenants",
+                        res.tenants.len()
+                    );
 
-        Ok(response
-            .tenants
-            .into_iter()
-            .map(|t| (t.id, Generation::new(t.generation)))
-            .collect::<HashMap<_, _>>())
-    }
-
-    /// Block until we get a successful response, or error out if we are shut down
-    async fn validate(
-        &self,
-        tenants: Vec<(TenantId, Generation)>,
-    ) -> Result<HashMap<TenantId, bool>, RetryForeverError> {
-        let re_attach_path = self
-            .base_url
-            .join("validate")
-            .expect("Failed to build validate path");
-
-        let request = ValidateRequest {
-            tenants: tenants
-                .into_iter()
-                .map(|(id, gen)| ValidateRequestTenant {
-                    id,
-                    gen: gen
-                        .into()
-                        .expect("Generation should always be valid for a Tenant doing deletions"),
-                })
-                .collect(),
-        };
-
-        let response: ValidateResponse = self.retry_http_forever(&re_attach_path, request).await?;
-
-        Ok(response
-            .tenants
-            .into_iter()
-            .map(|rt| (rt.id, rt.valid))
-            .collect())
+                    return Ok(res
+                        .tenants
+                        .into_iter()
+                        .map(|t| (t.id, Generation::new(t.generation)))
+                        .collect::<HashMap<_, _>>());
+                }
+                Err(e) => {
+                    tracing::error!("Error re-attaching tenants, retrying: {e:#}");
+                    backoff::exponential_backoff(
+                        attempt,
+                        BACKOFF_INCREMENT,
+                        BACKOFF_MAX,
+                        &self.cancel,
+                    )
+                    .await;
+                    if self.cancel.is_cancelled() {
+                        return Err(anyhow::anyhow!("Shutting down"));
+                    }
+                    attempt += 1;
+                }
+            }
+        }
     }
 }

pageserver/src/deletion_queue/deleter.rs

@@ -1,156 +0,0 @@
-//! The deleter is the final stage in the deletion queue.  It accumulates remote
-//! paths to delete, and periodically executes them in batches of up to 1000
-//! using the DeleteObjects request.
-//!
-//! Its purpose is to increase efficiency of remote storage I/O by issuing a smaller
-//! number of full-sized DeleteObjects requests, rather than a larger number of
-//! smaller requests.
-
-use remote_storage::GenericRemoteStorage;
-use remote_storage::RemotePath;
-use remote_storage::MAX_KEYS_PER_DELETE;
-use std::time::Duration;
-use tokio_util::sync::CancellationToken;
-use tracing::info;
-use tracing::warn;
-
-use crate::metrics;
-
-use super::DeletionQueueError;
-use super::FlushOp;
-
-const AUTOFLUSH_INTERVAL: Duration = Duration::from_secs(10);
-
-pub(super) enum DeleterMessage {
-    Delete(Vec<RemotePath>),
-    Flush(FlushOp),
-}
-
-/// Non-persistent deletion queue, for coalescing multiple object deletes into
-/// larger DeleteObjects requests.
-pub(super) struct Deleter {
-    // Accumulate up to 1000 keys for the next deletion operation
-    accumulator: Vec<RemotePath>,
-
-    rx: tokio::sync::mpsc::Receiver<DeleterMessage>,
-
-    cancel: CancellationToken,
-    remote_storage: GenericRemoteStorage,
-}
-
-impl Deleter {
-    pub(super) fn new(
-        remote_storage: GenericRemoteStorage,
-        rx: tokio::sync::mpsc::Receiver<DeleterMessage>,
-        cancel: CancellationToken,
-    ) -> Self {
-        Self {
-            remote_storage,
-            rx,
-            cancel,
-            accumulator: Vec::new(),
-        }
-    }
-
-    /// Wrap the remote `delete_objects` with a failpoint
-    async fn remote_delete(&self) -> Result<(), anyhow::Error> {
-        fail::fail_point!("deletion-queue-before-execute", |_| {
-            info!("Skipping execution, failpoint set");
-            metrics::DELETION_QUEUE
-                .remote_errors
-                .with_label_values(&["failpoint"])
-                .inc();
-            Err(anyhow::anyhow!("failpoint hit"))
-        });
-
-        self.remote_storage.delete_objects(&self.accumulator).await
-    }
-
-    /// Block until everything in accumulator has been executed
-    async fn flush(&mut self) -> Result<(), DeletionQueueError> {
-        while !self.accumulator.is_empty() && !self.cancel.is_cancelled() {
-            match self.remote_delete().await {
-                Ok(()) => {
-                    // Note: we assume that the remote storage layer returns Ok(()) if some
-                    // or all of the deleted objects were already gone.
-                    metrics::DELETION_QUEUE
-                        .keys_executed
-                        .inc_by(self.accumulator.len() as u64);
-                    info!(
-                        "Executed deletion batch {}..{}",
-                        self.accumulator
-                            .first()
-                            .expect("accumulator should be non-empty"),
-                        self.accumulator
-                            .last()
-                            .expect("accumulator should be non-empty"),
-                    );
-                    self.accumulator.clear();
-                }
-                Err(e) => {
-                    warn!("DeleteObjects request failed: {e:#}, will retry");
-                    metrics::DELETION_QUEUE
-                        .remote_errors
-                        .with_label_values(&["execute"])
-                        .inc();
-                }
-            };
-        }
-        if self.cancel.is_cancelled() {
-            // Expose an error because we may not have actually flushed everything
-            Err(DeletionQueueError::ShuttingDown)
-        } else {
-            Ok(())
-        }
-    }
-
-    pub(super) async fn background(&mut self) -> Result<(), DeletionQueueError> {
-        self.accumulator.reserve(MAX_KEYS_PER_DELETE);
-
-        loop {
-            if self.cancel.is_cancelled() {
-                return Err(DeletionQueueError::ShuttingDown);
-            }
-
-            let msg = match tokio::time::timeout(AUTOFLUSH_INTERVAL, self.rx.recv()).await {
-                Ok(Some(m)) => m,
-                Ok(None) => {
-                    // All queue senders closed
-                    info!("Shutting down");
-                    return Err(DeletionQueueError::ShuttingDown);
-                }
-                Err(_) => {
-                    // Timeout, we hit deadline to execute whatever we have in hand.  These functions will
-                    // return immediately if no work is pending
-                    self.flush().await?;
-
-                    continue;
-                }
-            };
-
-            match msg {
-                DeleterMessage::Delete(mut list) => {
-                    while !list.is_empty() || self.accumulator.len() == MAX_KEYS_PER_DELETE {
-                        if self.accumulator.len() == MAX_KEYS_PER_DELETE {
-                            self.flush().await?;
-                            // If we have received this number of keys, proceed with attempting to execute
-                            assert_eq!(self.accumulator.len(), 0);
-                        }
-
-                        let available_slots = MAX_KEYS_PER_DELETE - self.accumulator.len();
-                        let take_count = std::cmp::min(available_slots, list.len());
-                        for path in list.drain(list.len() - take_count..) {
-                            self.accumulator.push(path);
-                        }
-                    }
-                }
-                DeleterMessage::Flush(flush_op) => {
-                    // If flush() errors, we drop the flush_op and the caller will get
-                    // an error recv()'ing their oneshot channel.
-                    self.flush().await?;
-                    flush_op.notify();
-                }
-            }
-        }
-    }
-}

pageserver/src/deletion_queue/list_writer.rs

@@ -1,487 +0,0 @@
-//! The list writer is the first stage in the deletion queue.  It accumulates
-//! layers to delete, and periodically writes out these layers into a persistent
-//! DeletionList.
-//!
-//! The purpose of writing DeletionLists is to decouple the decision to
-//! delete an object from the validation required to execute it: even if
-//! validation is not possible, e.g. due to a control plane outage, we can
-//! still persist our intent to delete an object, in a way that would
-//! survive a restart.
-//!
-//! DeletionLists are passed onwards to the Validator.
-
-use super::DeletionHeader;
-use super::DeletionList;
-use super::FlushOp;
-use super::ValidatorQueueMessage;
-
-use std::collections::HashMap;
-use std::fs::create_dir_all;
-use std::time::Duration;
-
-use regex::Regex;
-use remote_storage::RemotePath;
-use tokio_util::sync::CancellationToken;
-use tracing::debug;
-use tracing::info;
-use tracing::warn;
-use utils::generation::Generation;
-use utils::id::TenantId;
-use utils::id::TimelineId;
-
-use crate::config::PageServerConf;
-use crate::deletion_queue::TEMP_SUFFIX;
-use crate::metrics;
-use crate::tenant::remote_timeline_client::remote_layer_path;
-use crate::tenant::storage_layer::LayerFileName;
-use crate::virtual_file;
-use crate::virtual_file::on_fatal_io_error;
-
-// The number of keys in a DeletionList before we will proactively persist it
-// (without reaching a flush deadline).  This aims to deliver objects of the order
-// of magnitude 1MB when we are under heavy delete load.
-const DELETION_LIST_TARGET_SIZE: usize = 16384;
-
-// Ordinarily, we only flush to DeletionList periodically, to bound the window during
-// which we might leak objects from not flushing a DeletionList after
-// the objects are already unlinked from timeline metadata.
-const FRONTEND_DEFAULT_TIMEOUT: Duration = Duration::from_millis(10000);
-
-// If someone is waiting for a flush to DeletionList, only delay a little to accumulate
-// more objects before doing the flush.
-const FRONTEND_FLUSHING_TIMEOUT: Duration = Duration::from_millis(100);
-
-#[derive(Debug)]
-pub(super) struct DeletionOp {
-    pub(super) tenant_id: TenantId,
-    pub(super) timeline_id: TimelineId,
-    // `layers` and `objects` are both just lists of objects.  `layers` is used if you do not
-    // have a config object handy to project it to a remote key, and need the consuming worker
-    // to do it for you.
-    pub(super) layers: Vec<(LayerFileName, Generation)>,
-    pub(super) objects: Vec<RemotePath>,
-
-    /// The _current_ generation of the Tenant attachment in which we are enqueuing
-    /// this deletion.
-    pub(super) generation: Generation,
-}
-
-#[derive(Debug)]
-pub(super) struct RecoverOp {
-    pub(super) attached_tenants: HashMap<TenantId, Generation>,
-}
-
-#[derive(Debug)]
-pub(super) enum ListWriterQueueMessage {
-    Delete(DeletionOp),
-    // Wait until all prior deletions make it into a persistent DeletionList
-    Flush(FlushOp),
-    // Wait until all prior deletions have been executed (i.e. objects are actually deleted)
-    FlushExecute(FlushOp),
-    // Call once after re-attaching to control plane, to notify the deletion queue about
-    // latest attached generations & load any saved deletion lists from disk.
-    Recover(RecoverOp),
-}
-
-pub(super) struct ListWriter {
-    conf: &'static PageServerConf,
-
-    // Incoming frontend requests to delete some keys
-    rx: tokio::sync::mpsc::Receiver<ListWriterQueueMessage>,
-
-    // Outbound requests to the backend to execute deletion lists we have composed.
-    tx: tokio::sync::mpsc::Sender<ValidatorQueueMessage>,
-
-    // The list we are currently building, contains a buffer of keys to delete
-    // and our next sequence number
-    pending: DeletionList,
-
-    // These FlushOps should notify the next time we flush
-    pending_flushes: Vec<FlushOp>,
-
-    // Worker loop is torn down when this fires.
-    cancel: CancellationToken,
-
-    // Safety guard to do recovery exactly once
-    recovered: bool,
-}
-
-impl ListWriter {
-    // Initially DeletionHeader.validated_sequence is zero.  The place we start our
-    // sequence numbers must be higher than that.
-    const BASE_SEQUENCE: u64 = 1;
-
-    pub(super) fn new(
-        conf: &'static PageServerConf,
-        rx: tokio::sync::mpsc::Receiver<ListWriterQueueMessage>,
-        tx: tokio::sync::mpsc::Sender<ValidatorQueueMessage>,
-        cancel: CancellationToken,
-    ) -> Self {
-        Self {
-            pending: DeletionList::new(Self::BASE_SEQUENCE),
-            conf,
-            rx,
-            tx,
-            pending_flushes: Vec::new(),
-            cancel,
-            recovered: false,
-        }
-    }
-
-    /// Try to flush `list` to persistent storage
-    ///
-    /// This does not return errors, because on failure to flush we do not lose
-    /// any state: flushing will be retried implicitly on the next deadline
-    async fn flush(&mut self) {
-        if self.pending.is_empty() {
-            for f in self.pending_flushes.drain(..) {
-                f.notify();
-            }
-            return;
-        }
-
-        match self.pending.save(self.conf).await {
-            Ok(_) => {
-                info!(sequence = self.pending.sequence, "Stored deletion list");
-
-                for f in self.pending_flushes.drain(..) {
-                    f.notify();
-                }
-
-                // Take the list we've accumulated, replace it with a fresh list for the next sequence
-                let next_list = DeletionList::new(self.pending.sequence + 1);
-                let list = std::mem::replace(&mut self.pending, next_list);
-
-                if let Err(e) = self.tx.send(ValidatorQueueMessage::Delete(list)).await {
-                    // This is allowed to fail: it will only happen if the backend worker is shut down,
-                    // so we can just drop this on the floor.
-                    info!("Deletion list dropped, this is normal during shutdown ({e:#})");
-                }
-            }
-            Err(e) => {
-                metrics::DELETION_QUEUE.unexpected_errors.inc();
-                warn!(
-                    sequence = self.pending.sequence,
-                    "Failed to write deletion list, will retry later ({e:#})"
-                );
-            }
-        }
-    }
-
-    /// Load the header, to learn the sequence number up to which deletions
-    /// have been validated.  We will apply validated=true to DeletionLists
-    /// <= this sequence when loading them.
-    ///
-    /// It is not an error for the header to not exist: we return None, and
-    /// the caller should act as if validated_sequence is 0
-    async fn load_validated_sequence(&self) -> Result<Option<u64>, anyhow::Error> {
-        let header_path = self.conf.deletion_header_path();
-        match tokio::fs::read(&header_path).await {
-            Ok(header_bytes) => {
-                match serde_json::from_slice::<DeletionHeader>(&header_bytes) {
-                    Ok(h) => Ok(Some(h.validated_sequence)),
-                    Err(e) => {
-                        warn!(
-                            "Failed to deserialize deletion header, ignoring {header_path}: {e:#}",
-                        );
-                        // This should never happen unless we make a mistake with our serialization.
-                        // Ignoring a deletion header is not consequential for correctnes because all deletions
-                        // are ultimately allowed to fail: worst case we leak some objects for the scrubber to clean up.
-                        metrics::DELETION_QUEUE.unexpected_errors.inc();
-                        Ok(None)
-                    }
-                }
-            }
-            Err(e) => {
-                if e.kind() == std::io::ErrorKind::NotFound {
-                    debug!("Deletion header {header_path} not found, first start?");
-                    Ok(None)
-                } else {
-                    on_fatal_io_error(&e);
-                }
-            }
-        }
-    }
-
-    async fn recover(
-        &mut self,
-        attached_tenants: HashMap<TenantId, Generation>,
-    ) -> Result<(), anyhow::Error> {
-        debug!(
-            "recovering with {} attached tenants",
-            attached_tenants.len()
-        );
-
-        // Load the header
-        let validated_sequence = self.load_validated_sequence().await?.unwrap_or(0);
-
-        self.pending.sequence = validated_sequence + 1;
-
-        let deletion_directory = self.conf.deletion_prefix();
-        let mut dir = match tokio::fs::read_dir(&deletion_directory).await {
-            Ok(d) => d,
-            Err(e) => {
-                warn!("Failed to open deletion list directory {deletion_directory}: {e:#}");
-
-                // This is fatal: any failure to read this local directory indicates a
-                // storage problem or configuration problem of the node.
-                virtual_file::on_fatal_io_error(&e);
-            }
-        };
-
-        let list_name_pattern =
-            Regex::new("(?<sequence>[a-zA-Z0-9]{16})-(?<version>[a-zA-Z0-9]{2}).list").unwrap();
-
-        let header_path = self.conf.deletion_header_path();
-        let mut seqs: Vec<u64> = Vec::new();
-        while let Some(dentry) = dir.next_entry().await? {
-            let file_name = dentry.file_name();
-            let dentry_str = file_name.to_string_lossy();
-
-            if file_name == header_path.file_name().unwrap_or("") {
-                // Don't try and parse the header's name like a list
-                continue;
-            }
-
-            if dentry_str.ends_with(TEMP_SUFFIX) {
-                info!("Cleaning up temporary file {dentry_str}");
-                let absolute_path =
-                    deletion_directory.join(dentry.file_name().to_str().expect("non-Unicode path"));
-                if let Err(e) = tokio::fs::remove_file(&absolute_path).await {
-                    // Non-fatal error: we will just leave the file behind but not
-                    // try and load it.
-                    warn!("Failed to clean up temporary file {absolute_path}: {e:#}");
-
-                    virtual_file::on_fatal_io_error(&e);
-                }
-
-                continue;
-            }
-
-            let file_name = dentry.file_name().to_owned();
-            let basename = file_name.to_string_lossy();
-            let seq_part = if let Some(m) = list_name_pattern.captures(&basename) {
-                m.name("sequence")
-                    .expect("Non optional group should be present")
-                    .as_str()
-            } else {
-                warn!("Unexpected filename in deletion queue: {basename}");
-                metrics::DELETION_QUEUE.unexpected_errors.inc();
-                continue;
-            };
-
-            let seq: u64 = match u64::from_str_radix(seq_part, 16) {
-                Ok(s) => s,
-                Err(e) => {
-                    warn!("Malformed key '{basename}': {e}");
-                    metrics::DELETION_QUEUE.unexpected_errors.inc();
-                    continue;
-                }
-            };
-            seqs.push(seq);
-        }
-        seqs.sort();
-
-        // Start our next deletion list from after the last location validated by
-        // previous process lifetime, or after the last location found (it is updated
-        // below after enumerating the deletion lists)
-        self.pending.sequence = validated_sequence + 1;
-        if let Some(max_list_seq) = seqs.last() {
-            self.pending.sequence = std::cmp::max(self.pending.sequence, max_list_seq + 1);
-        }
-
-        for s in seqs {
-            let list_path = self.conf.deletion_list_path(s);
-
-            let list_bytes = match tokio::fs::read(&list_path).await {
-                Ok(b) => b,
-                Err(e) => {
-                    virtual_file::on_fatal_io_error(&e);
-                }
-            };
-
-            let mut deletion_list = match serde_json::from_slice::<DeletionList>(&list_bytes) {
-                Ok(l) => l,
-                Err(e) => {
-                    // Drop the list on the floor: any objects it referenced will be left behind
-                    // for scrubbing to clean up.  This should never happen unless we have a serialization bug.
-                    warn!(sequence = s, "Failed to deserialize deletion list: {e}");
-                    metrics::DELETION_QUEUE.unexpected_errors.inc();
-                    continue;
-                }
-            };
-
-            if deletion_list.sequence <= validated_sequence {
-                // If the deletion list falls below valid_seq, we may assume that it was
-                // already validated the last time this pageserver ran.  Otherwise, we still
-                // load it, as it may still contain content valid in this generation.
-                deletion_list.validated = true;
-            } else {
-                // Special case optimization: if a tenant is still attached, and no other
-                // generation was issued to another node in the interval while we restarted,
-                // then we may treat deletion lists from the previous generation as if they
-                // belong to our currently attached generation, and proceed to validate & execute.
-                for (tenant_id, tenant_list) in &mut deletion_list.tenants {
-                    if let Some(attached_gen) = attached_tenants.get(tenant_id) {
-                        if attached_gen.previous() == tenant_list.generation {
-                            tenant_list.generation = *attached_gen;
-                        }
-                    }
-                }
-            }
-
-            info!(
-                validated = deletion_list.validated,
-                sequence = deletion_list.sequence,
-                "Recovered deletion list"
-            );
-
-            // We will drop out of recovery if this fails: it indicates that we are shutting down
-            // or the backend has panicked
-            metrics::DELETION_QUEUE
-                .keys_submitted
-                .inc_by(deletion_list.len() as u64);
-            self.tx
-                .send(ValidatorQueueMessage::Delete(deletion_list))
-                .await?;
-        }
-
-        info!(next_sequence = self.pending.sequence, "Replay complete");
-
-        Ok(())
-    }
-
-    /// This is the front-end ingest, where we bundle up deletion requests into DeletionList
-    /// and write them out, for later validation by the backend and execution by the executor.
-    pub(super) async fn background(&mut self) {
-        info!("Started deletion frontend worker");
-
-        // Synchronous, but we only do it once per process lifetime so it's tolerable
-        if let Err(e) = create_dir_all(&self.conf.deletion_prefix()) {
-            tracing::error!(
-                "Failed to create deletion list directory {}, deletions will not be executed ({e})",
-                self.conf.deletion_prefix(),
-            );
-            metrics::DELETION_QUEUE.unexpected_errors.inc();
-            return;
-        }
-
-        while !self.cancel.is_cancelled() {
-            let timeout = if self.pending_flushes.is_empty() {
-                FRONTEND_DEFAULT_TIMEOUT
-            } else {
-                FRONTEND_FLUSHING_TIMEOUT
-            };
-
-            let msg = match tokio::time::timeout(timeout, self.rx.recv()).await {
-                Ok(Some(msg)) => msg,
-                Ok(None) => {
-                    // Queue sender destroyed, shutting down
-                    break;
-                }
-                Err(_) => {
-                    // Hit deadline, flush.
-                    self.flush().await;
-                    continue;
-                }
-            };
-
-            match msg {
-                ListWriterQueueMessage::Delete(op) => {
-                    assert!(
-                        self.recovered,
-                        "Cannot process deletions before recovery.  This is a bug."
-                    );
-
-                    debug!(
-                        "Delete: ingesting {} layers, {} other objects",
-                        op.layers.len(),
-                        op.objects.len()
-                    );
-
-                    let mut layer_paths = Vec::new();
-                    for (layer, generation) in op.layers {
-                        layer_paths.push(remote_layer_path(
-                            &op.tenant_id,
-                            &op.timeline_id,
-                            &layer,
-                            generation,
-                        ));
-                    }
-                    layer_paths.extend(op.objects);
-
-                    if !self.pending.push(
-                        &op.tenant_id,
-                        &op.timeline_id,
-                        op.generation,
-                        &mut layer_paths,
-                    ) {
-                        self.flush().await;
-                        let retry_succeeded = self.pending.push(
-                            &op.tenant_id,
-                            &op.timeline_id,
-                            op.generation,
-                            &mut layer_paths,
-                        );
-                        if !retry_succeeded {
-                            // Unexpected: after we flush, we should have
-                            // drained self.pending, so a conflict on
-                            // generation numbers should be impossible.
-                            tracing::error!(
-                                "Failed to enqueue deletions, leaking objects.  This is a bug."
-                            );
-                            metrics::DELETION_QUEUE.unexpected_errors.inc();
-                        }
-                    }
-                }
-                ListWriterQueueMessage::Flush(op) => {
-                    if self.pending.is_empty() {
-                        // Execute immediately
-                        debug!("Flush: No pending objects, flushing immediately");
-                        op.notify()
-                    } else {
-                        // Execute next time we flush
-                        debug!("Flush: adding to pending flush list for next deadline flush");
-                        self.pending_flushes.push(op);
-                    }
-                }
-                ListWriterQueueMessage::FlushExecute(op) => {
-                    debug!("FlushExecute: passing through to backend");
-                    // We do not flush to a deletion list here: the client sends a Flush before the FlushExecute
-                    if let Err(e) = self.tx.send(ValidatorQueueMessage::Flush(op)).await {
-                        info!("Can't flush, shutting down ({e})");
-                        // Caller will get error when their oneshot sender was dropped.
-                    }
-                }
-                ListWriterQueueMessage::Recover(op) => {
-                    if self.recovered {
-                        tracing::error!(
-                            "Deletion queue recovery called more than once.  This is a bug."
-                        );
-                        metrics::DELETION_QUEUE.unexpected_errors.inc();
-                        // Non-fatal: although this is a bug, since we did recovery at least once we may proceed.
-                        continue;
-                    }
-
-                    if let Err(e) = self.recover(op.attached_tenants).await {
-                        // This should only happen in truly unrecoverable cases, like the recovery finding that the backend
-                        // queue receiver has been dropped, or something is critically broken with
-                        // the local filesystem holding deletion lists.
-                        info!(
-                            "Deletion queue recover aborted, deletion queue will not proceed ({e})"
-                        );
-                        metrics::DELETION_QUEUE.unexpected_errors.inc();
-                        return;
-                    } else {
-                        self.recovered = true;
-                    }
-                }
-            }
-
-            if self.pending.len() > DELETION_LIST_TARGET_SIZE || !self.pending_flushes.is_empty() {
-                self.flush().await;
-            }
-        }
-        info!("Deletion queue shut down.");
-    }
-}

pageserver/src/deletion_queue/validator.rs

@@ -1,426 +0,0 @@
-//! The validator is responsible for validating DeletionLists for execution,
-//! based on whethe the generation in the DeletionList is still the latest
-//! generation for a tenant.
-//!
-//! The purpose of validation is to ensure split-brain safety in the cluster
-//! of pageservers: a deletion may only be executed if the tenant generation
-//! that originated it is still current.  See docs/rfcs/025-generation-numbers.md
-//! The purpose of accumulating lists before validating them is to reduce load
-//! on the control plane API by issuing fewer, larger requests.
-//!
-//! In addition to validating DeletionLists, the validator validates updates to remote_consistent_lsn
-//! for timelines: these are logically deletions because the safekeepers use remote_consistent_lsn
-//! to decide when old
-//!
-//! Deletions are passed onward to the Deleter.
-
-use std::collections::HashMap;
-use std::sync::Arc;
-use std::time::Duration;
-
-use camino::Utf8PathBuf;
-use tokio_util::sync::CancellationToken;
-use tracing::debug;
-use tracing::info;
-use tracing::warn;
-
-use crate::config::PageServerConf;
-use crate::control_plane_client::ControlPlaneGenerationsApi;
-use crate::control_plane_client::RetryForeverError;
-use crate::metrics;
-use crate::virtual_file::on_fatal_io_error;
-
-use super::deleter::DeleterMessage;
-use super::DeletionHeader;
-use super::DeletionList;
-use super::DeletionQueueError;
-use super::FlushOp;
-use super::VisibleLsnUpdates;
-
-// After this length of time, do any validation work that is pending,
-// even if we haven't accumulated many keys to delete.
-//
-// This also causes updates to remote_consistent_lsn to be validated, even
-// if there were no deletions enqueued.
-const AUTOFLUSH_INTERVAL: Duration = Duration::from_secs(10);
-
-// If we have received this number of keys, proceed with attempting to execute
-const AUTOFLUSH_KEY_COUNT: usize = 16384;
-
-#[derive(Debug)]
-pub(super) enum ValidatorQueueMessage {
-    Delete(DeletionList),
-    Flush(FlushOp),
-}
-pub(super) struct Validator<C>
-where
-    C: ControlPlaneGenerationsApi,
-{
-    conf: &'static PageServerConf,
-    rx: tokio::sync::mpsc::Receiver<ValidatorQueueMessage>,
-    tx: tokio::sync::mpsc::Sender<DeleterMessage>,
-
-    // Client for calling into control plane API for validation of deletes
-    control_plane_client: Option<C>,
-
-    // DeletionLists which are waiting generation validation.  Not safe to
-    // execute until [`validate`] has processed them.
-    pending_lists: Vec<DeletionList>,
-
-    // DeletionLists which have passed validation and are ready to execute.
-    validated_lists: Vec<DeletionList>,
-
-    // Sum of all the lengths of lists in pending_lists
-    pending_key_count: usize,
-
-    // Lsn validation state: we read projected LSNs and write back visible LSNs
-    // after validation.  This is the LSN equivalent of `pending_validation_lists`:
-    // it is drained in [`validate`]
-    lsn_table: Arc<std::sync::RwLock<VisibleLsnUpdates>>,
-
-    // If we failed to rewrite a deletion list due to local filesystem I/O failure,
-    // we must remember that and refuse to advance our persistent validated sequence
-    // number past the failure.
-    list_write_failed: Option<u64>,
-
-    cancel: CancellationToken,
-}
-
-impl<C> Validator<C>
-where
-    C: ControlPlaneGenerationsApi,
-{
-    pub(super) fn new(
-        conf: &'static PageServerConf,
-        rx: tokio::sync::mpsc::Receiver<ValidatorQueueMessage>,
-        tx: tokio::sync::mpsc::Sender<DeleterMessage>,
-        control_plane_client: Option<C>,
-        lsn_table: Arc<std::sync::RwLock<VisibleLsnUpdates>>,
-        cancel: CancellationToken,
-    ) -> Self {
-        Self {
-            conf,
-            rx,
-            tx,
-            control_plane_client,
-            lsn_table,
-            pending_lists: Vec::new(),
-            validated_lists: Vec::new(),
-            pending_key_count: 0,
-            list_write_failed: None,
-            cancel,
-        }
-    }
-    /// Process any outstanding validations of generations of pending LSN updates or pending
-    /// DeletionLists.
-    ///
-    /// Valid LSN updates propagate back to Timelines immediately, valid DeletionLists
-    /// go into the queue of ready-to-execute lists.
-    async fn validate(&mut self) -> Result<(), DeletionQueueError> {
-        // Figure out for each tenant which generation number to validate.
-        //
-        // It is sufficient to validate the max generation number of each tenant because only the
-        // highest generation number can possibly be valid. Hence this map will collect the
-        // highest generation pending validation for each tenant.
-        let mut tenant_generations = HashMap::new();
-        for list in &self.pending_lists {
-            for (tenant_id, tenant_list) in &list.tenants {
-                // Note: DeletionLists are in logical time order, so generation always
-                // goes up.  By doing a simple insert() we will always end up with
-                // the latest generation seen for a tenant.
-                tenant_generations.insert(*tenant_id, tenant_list.generation);
-            }
-        }
-
-        let pending_lsn_updates = {
-            let mut lsn_table = self.lsn_table.write().expect("Lock should not be poisoned");
-            std::mem::take(&mut *lsn_table)
-        };
-        for (tenant_id, update) in &pending_lsn_updates.tenants {
-            let entry = tenant_generations
-                .entry(*tenant_id)
-                .or_insert(update.generation);
-            if update.generation > *entry {
-                *entry = update.generation;
-            }
-        }
-
-        if tenant_generations.is_empty() {
-            // No work to do
-            return Ok(());
-        }
-
-        let tenants_valid = if let Some(control_plane_client) = &self.control_plane_client {
-            match control_plane_client
-                .validate(tenant_generations.iter().map(|(k, v)| (*k, *v)).collect())
-                .await
-            {
-                Ok(tenants) => tenants,
-                Err(RetryForeverError::ShuttingDown) => {
-                    // The only way a validation call returns an error is when the cancellation token fires
-                    return Err(DeletionQueueError::ShuttingDown);
-                }
-            }
-        } else {
-            // Control plane API disabled.  In legacy mode we consider everything valid.
-            tenant_generations.keys().map(|k| (*k, true)).collect()
-        };
-
-        let mut validated_sequence: Option<u64> = None;
-
-        // Apply the validation results to the pending LSN updates
-        for (tenant_id, tenant_lsn_state) in pending_lsn_updates.tenants {
-            let validated_generation = tenant_generations
-                .get(&tenant_id)
-                .expect("Map was built from the same keys we're reading");
-
-            let valid = tenants_valid
-                .get(&tenant_id)
-                .copied()
-                // If the tenant was missing from the validation response, it has been deleted.
-                // The Timeline that requested the LSN update is probably already torn down,
-                // or will be torn down soon.  In this case, drop the update by setting valid=false.
-                .unwrap_or(false);
-
-            if valid && *validated_generation == tenant_lsn_state.generation {
-                for (_timeline_id, pending_lsn) in tenant_lsn_state.timelines {
-                    pending_lsn.result_slot.store(pending_lsn.projected);
-                }
-            } else {
-                // If we failed validation, then do not apply any of the projected updates
-                warn!("Dropped remote consistent LSN updates for tenant {tenant_id} in stale generation {:?}", tenant_lsn_state.generation);
-                metrics::DELETION_QUEUE.dropped_lsn_updates.inc();
-            }
-        }
-
-        // Apply the validation results to the pending deletion lists
-        for list in &mut self.pending_lists {
-            // Filter the list based on whether the server responded valid: true.
-            // If a tenant is omitted in the response, it has been deleted, and we should
-            // proceed with deletion.
-            let mut mutated = false;
-            list.tenants.retain(|tenant_id, tenant| {
-                let validated_generation = tenant_generations
-                    .get(tenant_id)
-                    .expect("Map was built from the same keys we're reading");
-
-                // If the tenant was missing from the validation response, it has been deleted.
-                // This means that a deletion is valid, but also redundant since the tenant's
-                // objects should have already been deleted.  Treat it as invalid to drop the
-                // redundant deletion.
-                let valid = tenants_valid.get(tenant_id).copied().unwrap_or(false);
-
-                // A list is valid if it comes from the current _or previous_ generation.
-                // - The previous generation case is permitted due to how we store deletion lists locally:
-                // if we see the immediately previous generation in a locally stored deletion list,
-                // it proves that this node's disk was used for both current & previous generations,
-                // and therefore no other node was involved in between: the two generations may be
-                // logically treated as the same.
-                // - In that previous generation case, we rewrote it to the current generation
-                // in recover(), so the comparison here is simply an equality.
-
-                let this_list_valid = valid
-                    && (tenant.generation == *validated_generation);
-
-                if !this_list_valid {
-                    warn!("Dropping stale deletions for tenant {tenant_id} in generation {:?}, objects may be leaked", tenant.generation);
-                    metrics::DELETION_QUEUE.keys_dropped.inc_by(tenant.len() as u64);
-                    mutated = true;
-                }
-                this_list_valid
-            });
-            list.validated = true;
-
-            if mutated {
-                // Save the deletion list if we had to make changes due to stale generations.  The
-                // saved list is valid for execution.
-                if let Err(e) = list.save(self.conf).await {
-                    // Highly unexpected.  Could happen if e.g. disk full.
-                    // If we didn't save the trimmed list, it is _not_ valid to execute.
-                    warn!("Failed to save modified deletion list {list}: {e:#}");
-                    metrics::DELETION_QUEUE.unexpected_errors.inc();
-
-                    // Rather than have a complex retry process, just drop it and leak the objects,
-                    // scrubber will clean up eventually.
-                    list.tenants.clear(); // Result is a valid-but-empty list, which is a no-op for execution.
-
-                    // We must remember this failure, to prevent later writing out a header that
-                    // would imply the unwritable list was valid on disk.
-                    if self.list_write_failed.is_none() {
-                        self.list_write_failed = Some(list.sequence);
-                    }
-                }
-            }
-
-            // Assert monotonicity of the list sequence numbers we are processing
-            if let Some(validated) = validated_sequence {
-                assert!(list.sequence >= validated)
-            }
-
-            validated_sequence = Some(list.sequence);
-        }
-
-        if let Some(validated_sequence) = validated_sequence {
-            if let Some(list_write_failed) = self.list_write_failed {
-                // Rare error case: we failed to write out a deletion list to excise invalid
-                // entries, so we cannot advance the header's valid sequence number past that point.
-                //
-                // In this state we will continue to validate, execute and delete deletion lists,
-                // we just cannot update the header.  It should be noticed and fixed by a human due to
-                // the nonzero value of our unexpected_errors metric.
-                warn!(
-                    sequence_number = list_write_failed,
-                    "Cannot write header because writing a deletion list failed earlier",
-                );
-            } else {
-                // Write the queue header to record how far validation progressed.  This avoids having
-                // to rewrite each DeletionList to set validated=true in it.
-                let header = DeletionHeader::new(validated_sequence);
-
-                // Drop result because the validated_sequence is an optimization.  If we fail to save it,
-                // then restart, we will drop some deletion lists, creating work for scrubber.
-                // The save() function logs a warning on error.
-                if let Err(e) = header.save(self.conf).await {
-                    warn!("Failed to write deletion queue header: {e:#}");
-                    metrics::DELETION_QUEUE.unexpected_errors.inc();
-                }
-            }
-        }
-
-        // Transfer the validated lists to the validated queue, for eventual execution
-        self.validated_lists.append(&mut self.pending_lists);
-
-        Ok(())
-    }
-
-    async fn cleanup_lists(&mut self, list_paths: Vec<Utf8PathBuf>) {
-        for list_path in list_paths {
-            debug!("Removing deletion list {list_path}");
-
-            if let Err(e) = tokio::fs::remove_file(&list_path).await {
-                // Unexpected: we should have permissions and nothing else should
-                // be touching these files.  We will leave the file behind.  Subsequent
-                // pageservers will try and load it again: hopefully whatever storage
-                // issue (probably permissions) has been fixed by then.
-                tracing::error!("Failed to delete {list_path}: {e:#}");
-                metrics::DELETION_QUEUE.unexpected_errors.inc();
-
-                on_fatal_io_error(&e);
-            }
-        }
-    }
-
-    async fn flush(&mut self) -> Result<(), DeletionQueueError> {
-        tracing::debug!("Flushing with {} pending lists", self.pending_lists.len());
-
-        // Issue any required generation validation calls to the control plane
-        self.validate().await?;
-
-        // After successful validation, nothing is pending: any lists that
-        // made it through validation will be in validated_lists.
-        assert!(self.pending_lists.is_empty());
-        self.pending_key_count = 0;
-
-        tracing::debug!(
-            "Validation complete, have {} validated lists",
-            self.validated_lists.len()
-        );
-
-        // Return quickly if we have no validated lists to execute.  This avoids flushing the
-        // executor when an idle backend hits its autoflush interval
-        if self.validated_lists.is_empty() {
-            return Ok(());
-        }
-
-        // Drain `validated_lists` into the executor
-        let mut executing_lists = Vec::new();
-        for list in self.validated_lists.drain(..) {
-            let list_path = self.conf.deletion_list_path(list.sequence);
-            let objects = list.into_remote_paths();
-            self.tx
-                .send(DeleterMessage::Delete(objects))
-                .await
-                .map_err(|_| DeletionQueueError::ShuttingDown)?;
-            executing_lists.push(list_path);
-        }
-
-        self.flush_executor().await?;
-
-        // Erase the deletion lists whose keys have all be deleted from remote storage
-        self.cleanup_lists(executing_lists).await;
-
-        Ok(())
-    }
-
-    async fn flush_executor(&mut self) -> Result<(), DeletionQueueError> {
-        // Flush the executor, so that all the keys referenced by these deletion lists
-        // are actually removed from remote storage.  This is a precondition to deleting
-        // the deletion lists themselves.
-        let (flush_op, rx) = FlushOp::new();
-        self.tx
-            .send(DeleterMessage::Flush(flush_op))
-            .await
-            .map_err(|_| DeletionQueueError::ShuttingDown)?;
-
-        rx.await.map_err(|_| DeletionQueueError::ShuttingDown)
-    }
-
-    pub(super) async fn background(&mut self) {
-        tracing::info!("Started deletion backend worker");
-
-        while !self.cancel.is_cancelled() {
-            let msg = match tokio::time::timeout(AUTOFLUSH_INTERVAL, self.rx.recv()).await {
-                Ok(Some(m)) => m,
-                Ok(None) => {
-                    // All queue senders closed
-                    info!("Shutting down");
-                    break;
-                }
-                Err(_) => {
-                    // Timeout, we hit deadline to execute whatever we have in hand.  These functions will
-                    // return immediately if no work is pending.
-                    match self.flush().await {
-                        Ok(()) => {}
-                        Err(DeletionQueueError::ShuttingDown) => {
-                            // If we are shutting down, then auto-flush can safely be skipped
-                        }
-                    }
-
-                    continue;
-                }
-            };
-
-            match msg {
-                ValidatorQueueMessage::Delete(list) => {
-                    if list.validated {
-                        // A pre-validated list may only be seen during recovery, if we are recovering
-                        // a DeletionList whose on-disk state has validated=true
-                        self.validated_lists.push(list)
-                    } else {
-                        self.pending_key_count += list.len();
-                        self.pending_lists.push(list);
-                    }
-
-                    if self.pending_key_count > AUTOFLUSH_KEY_COUNT {
-                        match self.flush().await {
-                            Ok(()) => {}
-                            Err(DeletionQueueError::ShuttingDown) => {
-                                // If we are shutting down, then auto-flush can safely be skipped
-                            }
-                        }
-                    }
-                }
-                ValidatorQueueMessage::Flush(op) => {
-                    match self.flush().await {
-                        Ok(()) => {
-                            op.notify();
-                        }
-                        Err(DeletionQueueError::ShuttingDown) => {
-                            // If we fail due to shutting down, we will just drop `op` to propagate that status.
-                        }
-                    }
-                }
-            }
-        }
-    }
-}

pageserver/src/disk_usage_eviction_task.rs

@@ -43,12 +43,12 @@
 
 use std::{
     collections::HashMap,
+    path::Path,
     sync::Arc,
     time::{Duration, SystemTime},
 };
 
 use anyhow::Context;
-use camino::Utf8Path;
 use remote_storage::GenericRemoteStorage;
 use serde::{Deserialize, Serialize};
 use tokio::time::Instant;
@@ -122,7 +122,7 @@ async fn disk_usage_eviction_task(
     state: &State,
     task_config: &DiskUsageEvictionTaskConfig,
     storage: GenericRemoteStorage,
-    tenants_dir: &Utf8Path,
+    tenants_dir: &Path,
     cancel: CancellationToken,
 ) {
     scopeguard::defer! {
@@ -184,7 +184,7 @@ async fn disk_usage_eviction_task_iteration(
     state: &State,
     task_config: &DiskUsageEvictionTaskConfig,
     storage: &GenericRemoteStorage,
-    tenants_dir: &Utf8Path,
+    tenants_dir: &Path,
     cancel: &CancellationToken,
 ) -> anyhow::Result<()> {
     let usage_pre = filesystem_level_usage::get(tenants_dir, task_config)
@@ -620,8 +620,9 @@ impl std::ops::Deref for TimelineKey {
 }
 
 mod filesystem_level_usage {
+    use std::path::Path;
+
     use anyhow::Context;
-    use camino::Utf8Path;
 
     use crate::statvfs::Statvfs;
 
@@ -663,7 +664,7 @@ mod filesystem_level_usage {
     }
 
     pub fn get<'a>(
-        tenants_dir: &Utf8Path,
+        tenants_dir: &Path,
         config: &'a DiskUsageEvictionTaskConfig,
     ) -> anyhow::Result<Usage<'a>> {
         let mock_config = {

pageserver/src/http/openapi_spec.yml

@@ -1093,9 +1093,6 @@ components:
         remote_consistent_lsn:
           type: string
           format: hex
-        remote_consistent_lsn_visible:
-          type: string
-          format: hex
         ancestor_timeline_id:
           type: string
           format: hex

pageserver/src/http/routes.rs

@@ -5,7 +5,6 @@ use std::collections::HashMap;
 use std::sync::Arc;
 
 use anyhow::{anyhow, Context, Result};
-use futures::TryFutureExt;
 use hyper::StatusCode;
 use hyper::{Body, Request, Response, Uri};
 use metrics::launch_timestamp::LaunchTimestamp;
@@ -13,6 +12,7 @@ use pageserver_api::models::{
     DownloadRemoteLayersTaskSpawnRequest, TenantAttachRequest, TenantLoadRequest,
 };
 use remote_storage::GenericRemoteStorage;
+use storage_broker::BrokerClientChannel;
 use tenant_size_model::{SizeResult, StorageModel};
 use tokio_util::sync::CancellationToken;
 use tracing::*;
@@ -25,7 +25,6 @@ use super::models::{
     TimelineCreateRequest, TimelineGcRequest, TimelineInfo,
 };
 use crate::context::{DownloadBehavior, RequestContext};
-use crate::deletion_queue::DeletionQueueClient;
 use crate::metrics::{StorageTimeOperation, STORAGE_TIME_GLOBAL};
 use crate::pgdatadir_mapping::LsnForTimestamp;
 use crate::task_mgr::TaskKind;
@@ -36,7 +35,7 @@ use crate::tenant::mgr::{
 use crate::tenant::size::ModelInputs;
 use crate::tenant::storage_layer::LayerAccessStatsReset;
 use crate::tenant::timeline::Timeline;
-use crate::tenant::{LogicalSizeCalculationCause, PageReconstructError, TenantSharedResources};
+use crate::tenant::{LogicalSizeCalculationCause, PageReconstructError};
 use crate::{config::PageServerConf, tenant::mgr};
 use crate::{disk_usage_eviction_task, tenant};
 use utils::{
@@ -56,24 +55,22 @@ use utils::{
 // Imports only used for testing APIs
 use super::models::ConfigureFailpointsRequest;
 
-pub struct State {
+struct State {
     conf: &'static PageServerConf,
     auth: Option<Arc<JwtAuth>>,
     allowlist_routes: Vec<Uri>,
     remote_storage: Option<GenericRemoteStorage>,
     broker_client: storage_broker::BrokerClientChannel,
     disk_usage_eviction_state: Arc<disk_usage_eviction_task::State>,
-    deletion_queue_client: DeletionQueueClient,
 }
 
 impl State {
-    pub fn new(
+    fn new(
         conf: &'static PageServerConf,
         auth: Option<Arc<JwtAuth>>,
         remote_storage: Option<GenericRemoteStorage>,
         broker_client: storage_broker::BrokerClientChannel,
         disk_usage_eviction_state: Arc<disk_usage_eviction_task::State>,
-        deletion_queue_client: DeletionQueueClient,
     ) -> anyhow::Result<Self> {
         let allowlist_routes = ["/v1/status", "/v1/doc", "/swagger.yml"]
             .iter()
@@ -86,17 +83,8 @@ impl State {
             remote_storage,
             broker_client,
             disk_usage_eviction_state,
-            deletion_queue_client,
         })
     }
-
-    fn tenant_resources(&self) -> TenantSharedResources {
-        TenantSharedResources {
-            broker_client: self.broker_client.clone(),
-            remote_storage: self.remote_storage.clone(),
-            deletion_queue_client: self.deletion_queue_client.clone(),
-        }
-    }
 }
 
 #[inline(always)]
@@ -132,7 +120,7 @@ impl From<PageReconstructError> for ApiError {
                 ApiError::InternalServerError(anyhow::anyhow!("request was cancelled"))
             }
             PageReconstructError::AncestorStopping(_) => {
-                ApiError::ResourceUnavailable(format!("{pre}"))
+                ApiError::InternalServerError(anyhow::Error::new(pre))
             }
             PageReconstructError::WalRedo(pre) => {
                 ApiError::InternalServerError(anyhow::Error::new(pre))
@@ -145,7 +133,7 @@ impl From<TenantMapInsertError> for ApiError {
     fn from(tmie: TenantMapInsertError) -> ApiError {
         match tmie {
             TenantMapInsertError::StillInitializing | TenantMapInsertError::ShuttingDown => {
-                ApiError::ResourceUnavailable(format!("{tmie}"))
+                ApiError::InternalServerError(anyhow::Error::new(tmie))
             }
             TenantMapInsertError::TenantAlreadyExists(id, state) => {
                 ApiError::Conflict(format!("tenant {id} already exists, state: {state:?}"))
@@ -159,12 +147,6 @@ impl From<TenantStateError> for ApiError {
     fn from(tse: TenantStateError) -> ApiError {
         match tse {
             TenantStateError::NotFound(tid) => ApiError::NotFound(anyhow!("tenant {}", tid).into()),
-            TenantStateError::NotActive(_) => {
-                ApiError::ResourceUnavailable("Tenant not yet active".into())
-            }
-            TenantStateError::IsStopping(_) => {
-                ApiError::ResourceUnavailable("Tenant is stopping".into())
-            }
             _ => ApiError::InternalServerError(anyhow::Error::new(tse)),
         }
     }
@@ -174,17 +156,14 @@ impl From<GetTenantError> for ApiError {
     fn from(tse: GetTenantError) -> ApiError {
         match tse {
             GetTenantError::NotFound(tid) => ApiError::NotFound(anyhow!("tenant {}", tid).into()),
-            GetTenantError::Broken(reason) => {
-                ApiError::InternalServerError(anyhow!("tenant is broken: {}", reason))
-            }
-            GetTenantError::NotActive(_) => {
+            e @ GetTenantError::NotActive(_) => {
                 // Why is this not `ApiError::NotFound`?
                 // Because we must be careful to never return 404 for a tenant if it does
                 // in fact exist locally. If we did, the caller could draw the conclusion
                 // that it can attach the tenant to another PS and we'd be in split-brain.
                 //
                 // (We can produce this variant only in `mgr::get_tenant(..., active=true)` calls).
-                ApiError::ResourceUnavailable("Tenant not yet active".into())
+                ApiError::InternalServerError(anyhow::Error::new(e))
             }
         }
     }
@@ -305,14 +284,7 @@ async fn build_timeline_info_common(
     };
     let current_physical_size = Some(timeline.layer_size_sum().await);
     let state = timeline.current_state();
-    let remote_consistent_lsn_projected = timeline
-        .get_remote_consistent_lsn_projected()
-        .unwrap_or(Lsn(0));
-    let remote_consistent_lsn_visible = timeline
-        .get_remote_consistent_lsn_visible()
-        .unwrap_or(Lsn(0));
-
-    let walreceiver_status = timeline.walreceiver_status();
+    let remote_consistent_lsn = timeline.get_remote_consistent_lsn().unwrap_or(Lsn(0));
 
     let info = TimelineInfo {
         tenant_id: timeline.tenant_id,
@@ -320,8 +292,7 @@ async fn build_timeline_info_common(
         ancestor_timeline_id,
         ancestor_lsn,
         disk_consistent_lsn: timeline.get_disk_consistent_lsn(),
-        remote_consistent_lsn: remote_consistent_lsn_projected,
-        remote_consistent_lsn_visible,
+        remote_consistent_lsn,
         last_record_lsn,
         prev_record_lsn: Some(timeline.get_prev_record_lsn()),
         latest_gc_cutoff_lsn: *timeline.get_latest_gc_cutoff_lsn(),
@@ -335,8 +306,6 @@ async fn build_timeline_info_common(
         pg_version: timeline.pg_version,
 
         state,
-
-        walreceiver_status,
     };
     Ok(info)
 }
@@ -520,23 +489,24 @@ async fn tenant_attach_handler(
 
     let generation = get_request_generation(state, maybe_body.as_ref().and_then(|r| r.generation))?;
 
-    if state.remote_storage.is_none() {
+    if let Some(remote_storage) = &state.remote_storage {
+        mgr::attach_tenant(
+            state.conf,
+            tenant_id,
+            generation,
+            tenant_conf,
+            state.broker_client.clone(),
+            remote_storage.clone(),
+            &ctx,
+        )
+        .instrument(info_span!("tenant_attach", %tenant_id))
+        .await?;
+    } else {
         return Err(ApiError::BadRequest(anyhow!(
             "attach_tenant is not possible because pageserver was configured without remote storage"
         )));
     }
 
-    mgr::attach_tenant(
-        state.conf,
-        tenant_id,
-        generation,
-        tenant_conf,
-        state.tenant_resources(),
-        &ctx,
-    )
-    .instrument(info_span!("tenant_attach", %tenant_id))
-    .await?;
-
     json_response(StatusCode::ACCEPTED, ())
 }
 
@@ -597,7 +567,6 @@ async fn tenant_load_handler(
         generation,
         state.broker_client.clone(),
         state.remote_storage.clone(),
-        state.deletion_queue_client.clone(),
         &ctx,
     )
     .instrument(info_span!("load", %tenant_id))
@@ -631,9 +600,8 @@ async fn tenant_list_handler(
     let response_data = mgr::list_tenants()
         .instrument(info_span!("tenant_list"))
         .await
-        .map_err(|_| {
-            ApiError::ResourceUnavailable("Tenant map is initializing or shutting down".to_string())
-        })?
+        .map_err(anyhow::Error::new)
+        .map_err(ApiError::InternalServerError)?
         .iter()
         .map(|(id, state)| TenantInfo {
             id: *id,
@@ -940,7 +908,8 @@ async fn tenant_create_handler(
         tenant_conf,
         target_tenant_id,
         generation,
-        state.tenant_resources(),
+        state.broker_client.clone(),
+        state.remote_storage.clone(),
         &ctx,
     )
     .instrument(info_span!("tenant_create", tenant_id = %target_tenant_id))
@@ -1157,39 +1126,6 @@ async fn timeline_download_remote_layers_handler_get(
     json_response(StatusCode::OK, info)
 }
 
-async fn deletion_queue_flush(
-    r: Request<Body>,
-    cancel: CancellationToken,
-) -> Result<Response<Body>, ApiError> {
-    let state = get_state(&r);
-
-    if state.remote_storage.is_none() {
-        // Nothing to do if remote storage is disabled.
-        return json_response(StatusCode::OK, ());
-    }
-
-    let execute = parse_query_param(&r, "execute")?.unwrap_or(false);
-
-    let flush = async {
-        if execute {
-            state.deletion_queue_client.flush_execute().await
-        } else {
-            state.deletion_queue_client.flush().await
-        }
-    }
-    // DeletionQueueError's only case is shutting down.
-    .map_err(|_| ApiError::ShuttingDown);
-
-    tokio::select! {
-        res = flush => {
-            res.map(|()| json_response(StatusCode::OK, ()))?
-        }
-        _ = cancel.cancelled() => {
-            Err(ApiError::ShuttingDown)
-        }
-    }
-}
-
 async fn active_timeline_of_active_tenant(
     tenant_id: TenantId,
     timeline_id: TimelineId,
@@ -1418,9 +1354,12 @@ where
 }
 
 pub fn make_router(
-    state: Arc<State>,
+    conf: &'static PageServerConf,
     launch_ts: &'static LaunchTimestamp,
     auth: Option<Arc<JwtAuth>>,
+    broker_client: BrokerClientChannel,
+    remote_storage: Option<GenericRemoteStorage>,
+    disk_usage_eviction_state: Arc<disk_usage_eviction_task::State>,
 ) -> anyhow::Result<RouterBuilder<hyper::Body, ApiError>> {
     let spec = include_bytes!("openapi_spec.yml");
     let mut router = attach_openapi_ui(endpoint::make_router(), spec, "/swagger.yml", "/v1/doc");
@@ -1444,7 +1383,16 @@ pub fn make_router(
     );
 
     Ok(router
-        .data(state)
+        .data(Arc::new(
+            State::new(
+                conf,
+                auth,
+                remote_storage,
+                broker_client,
+                disk_usage_eviction_state,
+            )
+            .context("Failed to initialize router state")?,
+        ))
         .get("/v1/status", |r| api_handler(r, status_handler))
         .put("/v1/failpoints", |r| {
             testing_api_handler("manage failpoints", r, failpoints_handler)
@@ -1524,9 +1472,6 @@ pub fn make_router(
         .put("/v1/disk_usage_eviction/run", |r| {
             api_handler(r, disk_usage_eviction_run)
         })
-        .put("/v1/deletion_queue/flush", |r| {
-            api_handler(r, deletion_queue_flush)
-        })
         .put("/v1/tenant/:tenant_id/break", |r| {
             testing_api_handler("set tenant state to broken", r, handle_tenant_break)
         })

pageserver/src/import_datadir.rs

@@ -6,7 +6,6 @@ use std::path::{Path, PathBuf};
 
 use anyhow::{bail, ensure, Context, Result};
 use bytes::Bytes;
-use camino::Utf8Path;
 use futures::StreamExt;
 use tokio::io::{AsyncRead, AsyncReadExt};
 use tokio_tar::Archive;
@@ -30,7 +29,7 @@ use postgres_ffi::{BLCKSZ, WAL_SEGMENT_SIZE};
 use utils::lsn::Lsn;
 
 // Returns checkpoint LSN from controlfile
-pub fn get_lsn_from_controlfile(path: &Utf8Path) -> Result<Lsn> {
+pub fn get_lsn_from_controlfile(path: &Path) -> Result<Lsn> {
     // Read control file to extract the LSN
     let controlfile_path = path.join("global").join("pg_control");
     let controlfile = ControlFileData::decode(&std::fs::read(controlfile_path)?)?;
@@ -47,7 +46,7 @@ pub fn get_lsn_from_controlfile(path: &Utf8Path) -> Result<Lsn> {
 /// cluster was not shut down cleanly.
 pub async fn import_timeline_from_postgres_datadir(
     tline: &Timeline,
-    pgdata_path: &Utf8Path,
+    pgdata_path: &Path,
     pgdata_lsn: Lsn,
     ctx: &RequestContext,
 ) -> Result<()> {
@@ -76,12 +75,12 @@ pub async fn import_timeline_from_postgres_datadir(
             {
                 pg_control = Some(control_file);
             }
-            modification.flush(ctx).await?;
+            modification.flush().await?;
         }
     }
 
     // We're done importing all the data files.
-    modification.commit(ctx).await?;
+    modification.commit().await?;
 
     // We expect the Postgres server to be shut down cleanly.
     let pg_control = pg_control.context("pg_control file not found")?;
@@ -257,7 +256,7 @@ async fn import_slru(
 /// Scan PostgreSQL WAL files in given directory and load all records between
 /// 'startpoint' and 'endpoint' into the repository.
 async fn import_wal(
-    walpath: &Utf8Path,
+    walpath: &Path,
     tline: &Timeline,
     startpoint: Lsn,
     endpoint: Lsn,
@@ -360,7 +359,7 @@ pub async fn import_basebackup_from_tar(
                     // We found the pg_control file.
                     pg_control = Some(res);
                 }
-                modification.flush(ctx).await?;
+                modification.flush().await?;
             }
             tokio_tar::EntryType::Directory => {
                 debug!("directory {:?}", file_path);
@@ -378,7 +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(ctx).await?;
+    modification.commit().await?;
     Ok(())
 }
 

pageserver/src/lib.rs

@@ -3,8 +3,7 @@ pub mod basebackup;
 pub mod config;
 pub mod consumption_metrics;
 pub mod context;
-pub mod control_plane_client;
-pub mod deletion_queue;
+mod control_plane_client;
 pub mod disk_usage_eviction_task;
 pub mod http;
 pub mod import_datadir;
@@ -25,9 +24,9 @@ pub mod walredo;
 
 pub mod failpoint_support;
 
+use std::path::Path;
+
 use crate::task_mgr::TaskKind;
-use camino::Utf8Path;
-use deletion_queue::DeletionQueue;
 use tracing::info;
 
 /// Current storage format version
@@ -49,8 +48,8 @@ static ZERO_PAGE: bytes::Bytes = bytes::Bytes::from_static(&[0u8; 8192]);
 
 pub use crate::metrics::preinitialize_metrics;
 
-#[tracing::instrument(skip_all, fields(%exit_code))]
-pub async fn shutdown_pageserver(deletion_queue: Option<DeletionQueue>, exit_code: i32) {
+#[tracing::instrument]
+pub async fn shutdown_pageserver(exit_code: i32) {
     use std::time::Duration;
     // Shut down the libpq endpoint task. This prevents new connections from
     // being accepted.
@@ -78,11 +77,6 @@ pub async fn shutdown_pageserver(deletion_queue: Option<DeletionQueue>, exit_cod
     )
     .await;
 
-    // Best effort to persist any outstanding deletions, to avoid leaking objects
-    if let Some(mut deletion_queue) = deletion_queue {
-        deletion_queue.shutdown(Duration::from_secs(5)).await;
-    }
-
     // Shut down the HTTP endpoint last, so that you can still check the server's
     // status while it's shutting down.
     // FIXME: We should probably stop accepting commands like attach/detach earlier.
@@ -131,25 +125,25 @@ pub const TIMELINE_DELETE_MARK_SUFFIX: &str = "___delete";
 /// Full path: `tenants/<tenant_id>/___ignored_tenant`.
 pub const IGNORED_TENANT_FILE_NAME: &str = "___ignored_tenant";
 
-pub fn is_temporary(path: &Utf8Path) -> bool {
+pub fn is_temporary(path: &Path) -> bool {
     match path.file_name() {
-        Some(name) => name.ends_with(TEMP_FILE_SUFFIX),
+        Some(name) => name.to_string_lossy().ends_with(TEMP_FILE_SUFFIX),
         None => false,
     }
 }
 
-fn ends_with_suffix(path: &Utf8Path, suffix: &str) -> bool {
+fn ends_with_suffix(path: &Path, suffix: &str) -> bool {
     match path.file_name() {
-        Some(name) => name.ends_with(suffix),
+        Some(name) => name.to_string_lossy().ends_with(suffix),
         None => false,
     }
 }
 
-pub fn is_uninit_mark(path: &Utf8Path) -> bool {
+pub fn is_uninit_mark(path: &Path) -> bool {
     ends_with_suffix(path, TIMELINE_UNINIT_MARK_SUFFIX)
 }
 
-pub fn is_delete_mark(path: &Utf8Path) -> bool {
+pub fn is_delete_mark(path: &Path) -> bool {
     ends_with_suffix(path, TIMELINE_DELETE_MARK_SUFFIX)
 }
 

pageserver/src/metrics.rs

@@ -1,4 +1,3 @@
-use enum_map::EnumMap;
 use metrics::metric_vec_duration::DurationResultObserver;
 use metrics::{
     register_counter_vec, register_gauge_vec, register_histogram, register_histogram_vec,
@@ -94,35 +93,15 @@ pub(crate) static READ_NUM_FS_LAYERS: Lazy<Histogram> = Lazy::new(|| {
 });
 
 // Metrics collected on operations on the storage repository.
-
-pub(crate) struct ReconstructTimeMetrics {
-    ok: Histogram,
-    err: Histogram,
-}
-
-pub(crate) static RECONSTRUCT_TIME: Lazy<ReconstructTimeMetrics> = Lazy::new(|| {
-    let inner = register_histogram_vec!(
+pub(crate) static RECONSTRUCT_TIME: Lazy<Histogram> = Lazy::new(|| {
+    register_histogram!(
         "pageserver_getpage_reconstruct_seconds",
         "Time spent in reconstruct_value (reconstruct a page from deltas)",
-        &["result"],
         CRITICAL_OP_BUCKETS.into(),
     )
-    .expect("failed to define a metric");
-    ReconstructTimeMetrics {
-        ok: inner.get_metric_with_label_values(&["ok"]).unwrap(),
-        err: inner.get_metric_with_label_values(&["err"]).unwrap(),
-    }
+    .expect("failed to define a metric")
 });
 
-impl ReconstructTimeMetrics {
-    pub(crate) fn for_result<T, E>(&self, result: &Result<T, E>) -> &Histogram {
-        match result {
-            Ok(_) => &self.ok,
-            Err(_) => &self.err,
-        }
-    }
-}
-
 pub(crate) static MATERIALIZED_PAGE_CACHE_HIT_DIRECT: Lazy<IntCounter> = Lazy::new(|| {
     register_int_counter!(
         "pageserver_materialized_cache_hits_direct_total",
@@ -148,24 +127,22 @@ pub(crate) static MATERIALIZED_PAGE_CACHE_HIT: Lazy<IntCounter> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
-pub struct PageCacheMetricsForTaskKind {
+pub struct PageCacheMetrics {
     pub read_accesses_materialized_page: IntCounter,
+    pub read_accesses_ephemeral: IntCounter,
     pub read_accesses_immutable: IntCounter,
 
+    pub read_hits_ephemeral: IntCounter,
     pub read_hits_immutable: IntCounter,
     pub read_hits_materialized_page_exact: IntCounter,
     pub read_hits_materialized_page_older_lsn: IntCounter,
 }
 
-pub struct PageCacheMetrics {
-    map: EnumMap<TaskKind, EnumMap<PageContentKind, PageCacheMetricsForTaskKind>>,
-}
-
 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",
-        &["task_kind", "key_kind", "content_kind", "hit_kind"]
+        &["key_kind", "hit_kind"]
     )
     .expect("failed to define a metric")
 });
@@ -174,73 +151,55 @@ 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",
-        &["task_kind", "key_kind", "content_kind"]
+        &["key_kind"]
     )
     .expect("failed to define a metric")
 });
 
 pub static PAGE_CACHE: Lazy<PageCacheMetrics> = Lazy::new(|| PageCacheMetrics {
-    map: EnumMap::from_array(std::array::from_fn(|task_kind| {
-        let task_kind = <TaskKind as enum_map::Enum>::from_usize(task_kind);
-        let task_kind: &'static str = task_kind.into();
-        EnumMap::from_array(std::array::from_fn(|content_kind| {
-            let content_kind = <PageContentKind as enum_map::Enum>::from_usize(content_kind);
-            let content_kind: &'static str = content_kind.into();
-            PageCacheMetricsForTaskKind {
-                read_accesses_materialized_page: {
-                    PAGE_CACHE_READ_ACCESSES
-                        .get_metric_with_label_values(&[
-                            task_kind,
-                            "materialized_page",
-                            content_kind,
-                        ])
-                        .unwrap()
-                },
+    read_accesses_materialized_page: {
+        PAGE_CACHE_READ_ACCESSES
+            .get_metric_with_label_values(&["materialized_page"])
+            .unwrap()
+    },
 
-                read_accesses_immutable: {
-                    PAGE_CACHE_READ_ACCESSES
-                        .get_metric_with_label_values(&[task_kind, "immutable", content_kind])
-                        .unwrap()
-                },
+    read_accesses_ephemeral: {
+        PAGE_CACHE_READ_ACCESSES
+            .get_metric_with_label_values(&["ephemeral"])
+            .unwrap()
+    },
 
-                read_hits_immutable: {
-                    PAGE_CACHE_READ_HITS
-                        .get_metric_with_label_values(&[task_kind, "immutable", content_kind, "-"])
-                        .unwrap()
-                },
+    read_accesses_immutable: {
+        PAGE_CACHE_READ_ACCESSES
+            .get_metric_with_label_values(&["immutable"])
+            .unwrap()
+    },
 
-                read_hits_materialized_page_exact: {
-                    PAGE_CACHE_READ_HITS
-                        .get_metric_with_label_values(&[
-                            task_kind,
-                            "materialized_page",
-                            content_kind,
-                            "exact",
-                        ])
-                        .unwrap()
-                },
+    read_hits_ephemeral: {
+        PAGE_CACHE_READ_HITS
+            .get_metric_with_label_values(&["ephemeral", "-"])
+            .unwrap()
+    },
 
-                read_hits_materialized_page_older_lsn: {
-                    PAGE_CACHE_READ_HITS
-                        .get_metric_with_label_values(&[
-                            task_kind,
-                            "materialized_page",
-                            content_kind,
-                            "older_lsn",
-                        ])
-                        .unwrap()
-                },
-            }
-        }))
-    })),
+    read_hits_immutable: {
+        PAGE_CACHE_READ_HITS
+            .get_metric_with_label_values(&["immutable", "-"])
+            .unwrap()
+    },
+
+    read_hits_materialized_page_exact: {
+        PAGE_CACHE_READ_HITS
+            .get_metric_with_label_values(&["materialized_page", "exact"])
+            .unwrap()
+    },
+
+    read_hits_materialized_page_older_lsn: {
+        PAGE_CACHE_READ_HITS
+            .get_metric_with_label_values(&["materialized_page", "older_lsn"])
+            .unwrap()
+    },
 });
 
-impl PageCacheMetrics {
-    pub(crate) fn for_ctx(&self, ctx: &RequestContext) -> &PageCacheMetricsForTaskKind {
-        &self.map[ctx.task_kind()][ctx.page_content_kind()]
-    }
-}
-
 pub struct PageCacheSizeMetrics {
     pub max_bytes: UIntGauge,
 
@@ -284,46 +243,6 @@ pub static PAGE_CACHE_SIZE: Lazy<PageCacheSizeMetrics> = Lazy::new(|| PageCacheS
     },
 });
 
-pub(crate) static PAGE_CACHE_ACQUIRE_PINNED_SLOT_TIME: Lazy<Histogram> = Lazy::new(|| {
-    register_histogram!(
-        "pageserver_page_cache_acquire_pinned_slot_seconds",
-        "Time spent acquiring a pinned slot in the page cache",
-        CRITICAL_OP_BUCKETS.into(),
-    )
-    .expect("failed to define a metric")
-});
-
-pub(crate) static PAGE_CACHE_FIND_VICTIMS_ITERS_TOTAL: Lazy<IntCounter> = Lazy::new(|| {
-    register_int_counter!(
-        "pageserver_page_cache_find_victim_iters_total",
-        "Counter for the number of iterations in the find_victim loop",
-    )
-    .expect("failed to define a metric")
-});
-
-static PAGE_CACHE_ERRORS: Lazy<IntCounterVec> = Lazy::new(|| {
-    register_int_counter_vec!(
-        "page_cache_errors_total",
-        "Number of timeouts while acquiring a pinned slot in the page cache",
-        &["error_kind"]
-    )
-    .expect("failed to define a metric")
-});
-
-#[derive(IntoStaticStr)]
-#[strum(serialize_all = "kebab_case")]
-pub(crate) enum PageCacheErrorKind {
-    AcquirePinnedSlotTimeout,
-    EvictIterLimit,
-}
-
-pub(crate) fn page_cache_errors_inc(error_kind: PageCacheErrorKind) {
-    PAGE_CACHE_ERRORS
-        .get_metric_with_label_values(&[error_kind.into()])
-        .unwrap()
-        .inc();
-}
-
 pub(crate) static WAIT_LSN_TIME: Lazy<Histogram> = Lazy::new(|| {
     register_histogram!(
         "pageserver_wait_lsn_seconds",
@@ -351,14 +270,6 @@ static RESIDENT_PHYSICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
-pub(crate) static RESIDENT_PHYSICAL_SIZE_GLOBAL: Lazy<UIntGauge> = Lazy::new(|| {
-    register_uint_gauge!(
-        "pageserver_resident_physical_size_global",
-        "Like `pageserver_resident_physical_size`, but without tenant/timeline dimensions."
-    )
-    .expect("failed to define a metric")
-});
-
 static REMOTE_PHYSICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
     register_uint_gauge_vec!(
         "pageserver_remote_physical_size",
@@ -369,14 +280,6 @@ static REMOTE_PHYSICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
-static REMOTE_PHYSICAL_SIZE_GLOBAL: Lazy<UIntGauge> = Lazy::new(|| {
-    register_uint_gauge!(
-        "pageserver_remote_physical_size_global",
-        "Like `pageserver_remote_physical_size`, but without tenant/timeline dimensions."
-    )
-    .expect("failed to define a metric")
-});
-
 pub(crate) static REMOTE_ONDEMAND_DOWNLOADED_LAYERS: Lazy<IntCounter> = Lazy::new(|| {
     register_int_counter!(
         "pageserver_remote_ondemand_downloaded_layers_total",
@@ -634,7 +537,7 @@ const STORAGE_IO_TIME_BUCKETS: &[f64] = &[
     30.000,   // 30000 ms
 ];
 
-/// VirtualFile fs operation variants.
+/// Tracks time taken by fs operations near VirtualFile.
 ///
 /// Operations:
 /// - open ([`std::fs::OpenOptions::open`])
@@ -645,66 +548,15 @@ const STORAGE_IO_TIME_BUCKETS: &[f64] = &[
 /// - seek (modify internal position or file length query)
 /// - fsync ([`std::fs::File::sync_all`])
 /// - metadata ([`std::fs::File::metadata`])
-#[derive(
-    Debug, Clone, Copy, strum_macros::EnumCount, strum_macros::EnumIter, strum_macros::FromRepr,
-)]
-pub(crate) enum StorageIoOperation {
-    Open,
-    Close,
-    CloseByReplace,
-    Read,
-    Write,
-    Seek,
-    Fsync,
-    Metadata,
-}
-
-impl StorageIoOperation {
-    pub fn as_str(&self) -> &'static str {
-        match self {
-            StorageIoOperation::Open => "open",
-            StorageIoOperation::Close => "close",
-            StorageIoOperation::CloseByReplace => "close-by-replace",
-            StorageIoOperation::Read => "read",
-            StorageIoOperation::Write => "write",
-            StorageIoOperation::Seek => "seek",
-            StorageIoOperation::Fsync => "fsync",
-            StorageIoOperation::Metadata => "metadata",
-        }
-    }
-}
-
-/// Tracks time taken by fs operations near VirtualFile.
-#[derive(Debug)]
-pub(crate) struct StorageIoTime {
-    metrics: [Histogram; StorageIoOperation::COUNT],
-}
-
-impl StorageIoTime {
-    fn new() -> Self {
-        let storage_io_histogram_vec = register_histogram_vec!(
-            "pageserver_io_operations_seconds",
-            "Time spent in IO operations",
-            &["operation"],
-            STORAGE_IO_TIME_BUCKETS.into()
-        )
-        .expect("failed to define a metric");
-        let metrics = std::array::from_fn(|i| {
-            let op = StorageIoOperation::from_repr(i).unwrap();
-            let metric = storage_io_histogram_vec
-                .get_metric_with_label_values(&[op.as_str()])
-                .unwrap();
-            metric
-        });
-        Self { metrics }
-    }
-
-    pub(crate) fn get(&self, op: StorageIoOperation) -> &Histogram {
-        &self.metrics[op as usize]
-    }
-}
-
-pub(crate) static STORAGE_IO_TIME_METRIC: Lazy<StorageIoTime> = Lazy::new(StorageIoTime::new);
+pub(crate) static STORAGE_IO_TIME: Lazy<HistogramVec> = Lazy::new(|| {
+    register_histogram_vec!(
+        "pageserver_io_operations_seconds",
+        "Time spent in IO operations",
+        &["operation"],
+        STORAGE_IO_TIME_BUCKETS.into()
+    )
+    .expect("failed to define a metric")
+});
 
 const STORAGE_IO_SIZE_OPERATIONS: &[&str] = &["read", "write"];
 
@@ -963,54 +815,6 @@ static REMOTE_TIMELINE_CLIENT_BYTES_FINISHED_COUNTER: Lazy<IntCounterVec> = Lazy
     .expect("failed to define a metric")
 });
 
-pub(crate) struct DeletionQueueMetrics {
-    pub(crate) keys_submitted: IntCounter,
-    pub(crate) keys_dropped: IntCounter,
-    pub(crate) keys_executed: IntCounter,
-    pub(crate) dropped_lsn_updates: IntCounter,
-    pub(crate) unexpected_errors: IntCounter,
-    pub(crate) remote_errors: IntCounterVec,
-}
-pub(crate) static DELETION_QUEUE: Lazy<DeletionQueueMetrics> = Lazy::new(|| {
-    DeletionQueueMetrics{
-
-    keys_submitted: register_int_counter!(
-        "pageserver_deletion_queue_submitted_total",
-        "Number of objects submitted for deletion"
-    )
-    .expect("failed to define a metric"),
-
-    keys_dropped: register_int_counter!(
-        "pageserver_deletion_queue_dropped_total",
-        "Number of object deletions dropped due to stale generation."
-    )
-    .expect("failed to define a metric"),
-
-    keys_executed: register_int_counter!(
-        "pageserver_deletion_queue_executed_total",
-        "Number of objects deleted. Only includes objects that we actually deleted, sum with pageserver_deletion_queue_dropped_total for the total number of keys processed."
-    )
-    .expect("failed to define a metric"),
-
-    dropped_lsn_updates: register_int_counter!(
-        "pageserver_deletion_queue_dropped_lsn_updates_total",
-        "Updates to remote_consistent_lsn dropped due to stale generation number."
-    )
-    .expect("failed to define a metric"),
-    unexpected_errors: register_int_counter!(
-        "pageserver_deletion_queue_unexpected_errors_total",
-        "Number of unexpected condiions that may stall the queue: any value above zero is unexpected."
-    )
-    .expect("failed to define a metric"),
-    remote_errors: register_int_counter_vec!(
-        "pageserver_deletion_queue_remote_errors_total",
-        "Retryable remote I/O errors while executing deletions, for example 503 responses to DeleteObjects",
-        &["op_kind"],
-    )
-    .expect("failed to define a metric")
-}
-});
-
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
 pub enum RemoteOpKind {
     Upload,
@@ -1285,7 +1089,7 @@ pub struct TimelineMetrics {
     pub load_layer_map_histo: StorageTimeMetrics,
     pub garbage_collect_histo: StorageTimeMetrics,
     pub last_record_gauge: IntGauge,
-    resident_physical_size_gauge: UIntGauge,
+    pub resident_physical_size_gauge: UIntGauge,
     /// copy of LayeredTimeline.current_logical_size
     pub current_logical_size_gauge: UIntGauge,
     pub num_persistent_files_created: IntCounter,
@@ -1361,31 +1165,6 @@ impl TimelineMetrics {
             ),
         }
     }
-
-    pub fn record_new_file_metrics(&self, sz: u64) {
-        self.resident_physical_size_add(sz);
-        self.num_persistent_files_created.inc_by(1);
-        self.persistent_bytes_written.inc_by(sz);
-    }
-
-    pub fn resident_physical_size_sub(&self, sz: u64) {
-        self.resident_physical_size_gauge.sub(sz);
-        crate::metrics::RESIDENT_PHYSICAL_SIZE_GLOBAL.sub(sz);
-    }
-
-    pub fn resident_physical_size_add(&self, sz: u64) {
-        self.resident_physical_size_gauge.add(sz);
-        crate::metrics::RESIDENT_PHYSICAL_SIZE_GLOBAL.add(sz);
-    }
-
-    pub fn resident_physical_size_set(&self, sz: u64) {
-        self.resident_physical_size_gauge.set(sz);
-        crate::metrics::RESIDENT_PHYSICAL_SIZE_GLOBAL.set(sz);
-    }
-
-    pub fn resident_physical_size_get(&self) -> u64 {
-        self.resident_physical_size_gauge.get()
-    }
 }
 
 impl Drop for TimelineMetrics {
@@ -1393,10 +1172,7 @@ impl Drop for TimelineMetrics {
         let tenant_id = &self.tenant_id;
         let timeline_id = &self.timeline_id;
         let _ = LAST_RECORD_LSN.remove_label_values(&[tenant_id, timeline_id]);
-        {
-            RESIDENT_PHYSICAL_SIZE_GLOBAL.sub(self.resident_physical_size_get());
-            let _ = RESIDENT_PHYSICAL_SIZE.remove_label_values(&[tenant_id, timeline_id]);
-        }
+        let _ = RESIDENT_PHYSICAL_SIZE.remove_label_values(&[tenant_id, timeline_id]);
         let _ = CURRENT_LOGICAL_SIZE.remove_label_values(&[tenant_id, timeline_id]);
         let _ = NUM_PERSISTENT_FILES_CREATED.remove_label_values(&[tenant_id, timeline_id]);
         let _ = PERSISTENT_BYTES_WRITTEN.remove_label_values(&[tenant_id, timeline_id]);
@@ -1447,46 +1223,10 @@ use std::sync::{Arc, Mutex};
 use std::task::{Context, Poll};
 use std::time::{Duration, Instant};
 
-use crate::context::{PageContentKind, RequestContext};
-use crate::task_mgr::TaskKind;
-
-/// Maintain a per timeline gauge in addition to the global gauge.
-struct PerTimelineRemotePhysicalSizeGauge {
-    last_set: u64,
-    gauge: UIntGauge,
-}
-
-impl PerTimelineRemotePhysicalSizeGauge {
-    fn new(per_timeline_gauge: UIntGauge) -> Self {
-        Self {
-            last_set: per_timeline_gauge.get(),
-            gauge: per_timeline_gauge,
-        }
-    }
-    fn set(&mut self, sz: u64) {
-        self.gauge.set(sz);
-        if sz < self.last_set {
-            REMOTE_PHYSICAL_SIZE_GLOBAL.sub(self.last_set - sz);
-        } else {
-            REMOTE_PHYSICAL_SIZE_GLOBAL.add(sz - self.last_set);
-        };
-        self.last_set = sz;
-    }
-    fn get(&self) -> u64 {
-        self.gauge.get()
-    }
-}
-
-impl Drop for PerTimelineRemotePhysicalSizeGauge {
-    fn drop(&mut self) {
-        REMOTE_PHYSICAL_SIZE_GLOBAL.sub(self.last_set);
-    }
-}
-
 pub struct RemoteTimelineClientMetrics {
     tenant_id: String,
     timeline_id: String,
-    remote_physical_size_gauge: Mutex<Option<PerTimelineRemotePhysicalSizeGauge>>,
+    remote_physical_size_gauge: Mutex<Option<UIntGauge>>,
     calls_unfinished_gauge: Mutex<HashMap<(&'static str, &'static str), IntGauge>>,
     bytes_started_counter: Mutex<HashMap<(&'static str, &'static str), IntCounter>>,
     bytes_finished_counter: Mutex<HashMap<(&'static str, &'static str), IntCounter>>,
@@ -1504,24 +1244,18 @@ impl RemoteTimelineClientMetrics {
         }
     }
 
-    pub(crate) fn remote_physical_size_set(&self, sz: u64) {
+    pub fn remote_physical_size_gauge(&self) -> UIntGauge {
         let mut guard = self.remote_physical_size_gauge.lock().unwrap();
-        let gauge = guard.get_or_insert_with(|| {
-            PerTimelineRemotePhysicalSizeGauge::new(
+        guard
+            .get_or_insert_with(|| {
                 REMOTE_PHYSICAL_SIZE
                     .get_metric_with_label_values(&[
                         &self.tenant_id.to_string(),
                         &self.timeline_id.to_string(),
                     ])
-                    .unwrap(),
-            )
-        });
-        gauge.set(sz);
-    }
-
-    pub(crate) fn remote_physical_size_get(&self) -> u64 {
-        let guard = self.remote_physical_size_gauge.lock().unwrap();
-        guard.as_ref().map(|gauge| gauge.get()).unwrap_or(0)
+                    .unwrap()
+            })
+            .clone()
     }
 
     pub fn remote_operation_time(
@@ -1860,9 +1594,6 @@ pub fn preinitialize_metrics() {
         Lazy::force(c);
     });
 
-    // Deletion queue stats
-    Lazy::force(&DELETION_QUEUE);
-
     // countervecs
     [&BACKGROUND_LOOP_PERIOD_OVERRUN_COUNT]
         .into_iter()
@@ -1876,6 +1607,7 @@ pub fn preinitialize_metrics() {
     // histograms
     [
         &READ_NUM_FS_LAYERS,
+        &RECONSTRUCT_TIME,
         &WAIT_LSN_TIME,
         &WAL_REDO_TIME,
         &WAL_REDO_WAIT_TIME,
@@ -1886,7 +1618,4 @@ pub fn preinitialize_metrics() {
     .for_each(|h| {
         Lazy::force(h);
     });
-
-    // Custom
-    Lazy::force(&RECONSTRUCT_TIME);
 }

pageserver/src/page_cache.rs

@@ -75,11 +75,7 @@
 use std::{
     collections::{hash_map::Entry, HashMap},
     convert::TryInto,
-    sync::{
-        atomic::{AtomicU64, AtomicU8, AtomicUsize, Ordering},
-        Arc, Weak,
-    },
-    time::Duration,
+    sync::atomic::{AtomicU64, AtomicU8, AtomicUsize, Ordering},
 };
 
 use anyhow::Context;
@@ -89,7 +85,7 @@ use utils::{
     lsn::Lsn,
 };
 
-use crate::{context::RequestContext, metrics::PageCacheSizeMetrics, repository::Key};
+use crate::{metrics::PageCacheSizeMetrics, repository::Key};
 
 static PAGE_CACHE: OnceCell<PageCache> = OnceCell::new();
 const TEST_PAGE_CACHE_SIZE: usize = 50;
@@ -169,8 +165,6 @@ struct Slot {
 
 struct SlotInner {
     key: Option<CacheKey>,
-    // for `coalesce_readers_permit`
-    permit: std::sync::Mutex<Weak<PinnedSlotsPermit>>,
     buf: &'static mut [u8; PAGE_SZ],
 }
 
@@ -213,22 +207,6 @@ impl Slot {
     }
 }
 
-impl SlotInner {
-    /// If there is aready a reader, drop our permit and share its permit, just like we share read access.
-    fn coalesce_readers_permit(&self, permit: PinnedSlotsPermit) -> Arc<PinnedSlotsPermit> {
-        let mut guard = self.permit.lock().unwrap();
-        if let Some(existing_permit) = guard.upgrade() {
-            drop(guard);
-            drop(permit);
-            existing_permit
-        } else {
-            let permit = Arc::new(permit);
-            *guard = Arc::downgrade(&permit);
-            permit
-        }
-    }
-}
-
 pub struct PageCache {
     /// This contains the mapping from the cache key to buffer slot that currently
     /// contains the page, if any.
@@ -246,8 +224,6 @@ pub struct PageCache {
     /// The actual buffers with their metadata.
     slots: Box<[Slot]>,
 
-    pinned_slots: Arc<tokio::sync::Semaphore>,
-
     /// Index of the next candidate to evict, for the Clock replacement algorithm.
     /// This is interpreted modulo the page cache size.
     next_evict_slot: AtomicUsize,
@@ -255,28 +231,23 @@ pub struct PageCache {
     size_metrics: &'static PageCacheSizeMetrics,
 }
 
-struct PinnedSlotsPermit(tokio::sync::OwnedSemaphorePermit);
-
 ///
 /// PageReadGuard is a "lease" on a buffer, for reading. The page is kept locked
 /// until the guard is dropped.
 ///
-pub struct PageReadGuard<'i> {
-    _permit: Arc<PinnedSlotsPermit>,
-    slot_guard: tokio::sync::RwLockReadGuard<'i, SlotInner>,
-}
+pub struct PageReadGuard<'i>(tokio::sync::RwLockReadGuard<'i, SlotInner>);
 
 impl std::ops::Deref for PageReadGuard<'_> {
     type Target = [u8; PAGE_SZ];
 
     fn deref(&self) -> &Self::Target {
-        self.slot_guard.buf
+        self.0.buf
     }
 }
 
 impl AsRef<[u8; PAGE_SZ]> for PageReadGuard<'_> {
     fn as_ref(&self) -> &[u8; PAGE_SZ] {
-        self.slot_guard.buf
+        self.0.buf
     }
 }
 
@@ -293,8 +264,6 @@ impl AsRef<[u8; PAGE_SZ]> for PageReadGuard<'_> {
 pub struct PageWriteGuard<'i> {
     inner: tokio::sync::RwLockWriteGuard<'i, SlotInner>,
 
-    _permit: PinnedSlotsPermit,
-
     // Are the page contents currently valid?
     // Used to mark pages as invalid that are assigned but not yet filled with data.
     valid: bool,
@@ -377,14 +346,8 @@ impl PageCache {
         timeline_id: TimelineId,
         key: &Key,
         lsn: Lsn,
-        ctx: &RequestContext,
     ) -> Option<(Lsn, PageReadGuard)> {
-        let Ok(permit) = self.try_get_pinned_slot_permit().await else {
-            return None;
-        };
-
         crate::metrics::PAGE_CACHE
-            .for_ctx(ctx)
             .read_accesses_materialized_page
             .inc();
 
@@ -397,10 +360,7 @@ impl PageCache {
             lsn,
         };
 
-        if let Some(guard) = self
-            .try_lock_for_read(&mut cache_key, &mut Some(permit))
-            .await
-        {
+        if let Some(guard) = self.try_lock_for_read(&mut cache_key).await {
             if let CacheKey::MaterializedPage {
                 hash_key: _,
                 lsn: available_lsn,
@@ -408,12 +368,10 @@ impl PageCache {
             {
                 if available_lsn == lsn {
                     crate::metrics::PAGE_CACHE
-                        .for_ctx(ctx)
                         .read_hits_materialized_page_exact
                         .inc();
                 } else {
                     crate::metrics::PAGE_CACHE
-                        .for_ctx(ctx)
                         .read_hits_materialized_page_older_lsn
                         .inc();
                 }
@@ -468,11 +426,10 @@ impl PageCache {
         &self,
         file_id: FileId,
         blkno: u32,
-        ctx: &RequestContext,
     ) -> anyhow::Result<ReadBufResult> {
         let mut cache_key = CacheKey::ImmutableFilePage { file_id, blkno };
 
-        self.lock_for_read(&mut cache_key, ctx).await
+        self.lock_for_read(&mut cache_key).await
     }
 
     //
@@ -483,29 +440,6 @@ impl PageCache {
     // "mappings" after this section. But the routines in this section should
     // not require changes.
 
-    async fn try_get_pinned_slot_permit(&self) -> anyhow::Result<PinnedSlotsPermit> {
-        let timer = crate::metrics::PAGE_CACHE_ACQUIRE_PINNED_SLOT_TIME.start_timer();
-        match tokio::time::timeout(
-            // Choose small timeout, neon_smgr does its own retries.
-            // https://neondb.slack.com/archives/C04DGM6SMTM/p1694786876476869
-            Duration::from_secs(10),
-            Arc::clone(&self.pinned_slots).acquire_owned(),
-        )
-        .await
-        {
-            Ok(res) => Ok(PinnedSlotsPermit(
-                res.expect("this semaphore is never closed"),
-            )),
-            Err(_timeout) => {
-                timer.stop_and_discard();
-                crate::metrics::page_cache_errors_inc(
-                    crate::metrics::PageCacheErrorKind::AcquirePinnedSlotTimeout,
-                );
-                anyhow::bail!("timeout: there were page guards alive for all page cache slots")
-            }
-        }
-    }
-
     /// Look up a page in the cache.
     ///
     /// If the search criteria is not exact, *cache_key is updated with the key
@@ -515,11 +449,7 @@ impl PageCache {
     ///
     /// If no page is found, returns None and *cache_key is left unmodified.
     ///
-    async fn try_lock_for_read(
-        &self,
-        cache_key: &mut CacheKey,
-        permit: &mut Option<PinnedSlotsPermit>,
-    ) -> Option<PageReadGuard> {
+    async fn try_lock_for_read(&self, cache_key: &mut CacheKey) -> Option<PageReadGuard> {
         let cache_key_orig = cache_key.clone();
         if let Some(slot_idx) = self.search_mapping(cache_key) {
             // The page was found in the mapping. Lock the slot, and re-check
@@ -529,10 +459,7 @@ impl PageCache {
             let inner = slot.inner.read().await;
             if inner.key.as_ref() == Some(cache_key) {
                 slot.inc_usage_count();
-                return Some(PageReadGuard {
-                    _permit: inner.coalesce_readers_permit(permit.take().unwrap()),
-                    slot_guard: inner,
-                });
+                return Some(PageReadGuard(inner));
             } else {
                 // search_mapping might have modified the search key; restore it.
                 *cache_key = cache_key_orig;
@@ -570,22 +497,14 @@ impl PageCache {
     /// }
     /// ```
     ///
-    async fn lock_for_read(
-        &self,
-        cache_key: &mut CacheKey,
-        ctx: &RequestContext,
-    ) -> anyhow::Result<ReadBufResult> {
-        let mut permit = Some(self.try_get_pinned_slot_permit().await?);
-
+    async 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::ImmutableFilePage { .. } => (
-                &crate::metrics::PAGE_CACHE
-                    .for_ctx(ctx)
-                    .read_accesses_immutable,
-                &crate::metrics::PAGE_CACHE.for_ctx(ctx).read_hits_immutable,
+                &crate::metrics::PAGE_CACHE.read_accesses_immutable,
+                &crate::metrics::PAGE_CACHE.read_hits_immutable,
             ),
         };
         read_access.inc();
@@ -593,21 +512,17 @@ impl PageCache {
         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, &mut permit).await {
-                debug_assert!(permit.is_none());
+            if let Some(read_guard) = self.try_lock_for_read(cache_key).await {
                 if is_first_iteration {
                     hit.inc();
                 }
                 return Ok(ReadBufResult::Found(read_guard));
             }
-            debug_assert!(permit.is_some());
             is_first_iteration = false;
 
             // Not found. Find a victim buffer
-            let (slot_idx, mut inner) = self
-                .find_victim(permit.as_ref().unwrap())
-                .await
-                .context("Failed to find evict victim")?;
+            let (slot_idx, mut inner) =
+                self.find_victim().context("Failed to find evict victim")?;
 
             // Insert mapping for this. At this point, we may find that another
             // thread did the same thing concurrently. In that case, we evicted
@@ -629,16 +544,7 @@ impl PageCache {
             inner.key = Some(cache_key.clone());
             slot.set_usage_count(1);
 
-            debug_assert!(
-                {
-                    let guard = inner.permit.lock().unwrap();
-                    guard.upgrade().is_none()
-                },
-                "we hold a write lock, so, no one else should have a permit"
-            );
-
             return Ok(ReadBufResult::NotFound(PageWriteGuard {
-                _permit: permit.take().unwrap(),
                 inner,
                 valid: false,
             }));
@@ -649,11 +555,7 @@ impl PageCache {
     /// found, returns None.
     ///
     /// When locking a page for writing, the search criteria is always "exact".
-    async fn try_lock_for_write(
-        &self,
-        cache_key: &CacheKey,
-        permit: &mut Option<PinnedSlotsPermit>,
-    ) -> Option<PageWriteGuard> {
+    async fn try_lock_for_write(&self, cache_key: &CacheKey) -> Option<PageWriteGuard> {
         if let Some(slot_idx) = self.search_mapping_for_write(cache_key) {
             // The page was found in the mapping. Lock the slot, and re-check
             // that it's still what we expected (because we don't released the mapping
@@ -662,18 +564,7 @@ impl PageCache {
             let inner = slot.inner.write().await;
             if inner.key.as_ref() == Some(cache_key) {
                 slot.inc_usage_count();
-                debug_assert!(
-                    {
-                        let guard = inner.permit.lock().unwrap();
-                        guard.upgrade().is_none()
-                    },
-                    "we hold a write lock, so, no one else should have a permit"
-                );
-                return Some(PageWriteGuard {
-                    _permit: permit.take().unwrap(),
-                    inner,
-                    valid: true,
-                });
+                return Some(PageWriteGuard { inner, valid: true });
             }
         }
         None
@@ -684,20 +575,15 @@ impl PageCache {
     /// Similar to lock_for_read(), but the returned buffer is write-locked and
     /// may be modified by the caller even if it's already found in the cache.
     async fn lock_for_write(&self, cache_key: &CacheKey) -> anyhow::Result<WriteBufResult> {
-        let mut permit = Some(self.try_get_pinned_slot_permit().await?);
         loop {
             // First check if the key already exists in the cache.
-            if let Some(write_guard) = self.try_lock_for_write(cache_key, &mut permit).await {
-                debug_assert!(permit.is_none());
+            if let Some(write_guard) = self.try_lock_for_write(cache_key).await {
                 return Ok(WriteBufResult::Found(write_guard));
             }
-            debug_assert!(permit.is_some());
 
             // Not found. Find a victim buffer
-            let (slot_idx, mut inner) = self
-                .find_victim(permit.as_ref().unwrap())
-                .await
-                .context("Failed to find evict victim")?;
+            let (slot_idx, mut inner) =
+                self.find_victim().context("Failed to find evict victim")?;
 
             // Insert mapping for this. At this point, we may find that another
             // thread did the same thing concurrently. In that case, we evicted
@@ -719,16 +605,7 @@ impl PageCache {
             inner.key = Some(cache_key.clone());
             slot.set_usage_count(1);
 
-            debug_assert!(
-                {
-                    let guard = inner.permit.lock().unwrap();
-                    guard.upgrade().is_none()
-                },
-                "we hold a write lock, so, no one else should have a permit"
-            );
-
             return Ok(WriteBufResult::NotFound(PageWriteGuard {
-                _permit: permit.take().unwrap(),
                 inner,
                 valid: false,
             }));
@@ -881,10 +758,7 @@ impl PageCache {
     /// Find a slot to evict.
     ///
     /// On return, the slot is empty and write-locked.
-    async fn find_victim(
-        &self,
-        _permit_witness: &PinnedSlotsPermit,
-    ) -> anyhow::Result<(usize, tokio::sync::RwLockWriteGuard<SlotInner>)> {
+    fn find_victim(&self) -> anyhow::Result<(usize, tokio::sync::RwLockWriteGuard<SlotInner>)> {
         let iter_limit = self.slots.len() * 10;
         let mut iters = 0;
         loop {
@@ -897,40 +771,13 @@ impl PageCache {
                 let mut inner = match slot.inner.try_write() {
                     Ok(inner) => inner,
                     Err(_err) => {
+                        // If we have looped through the whole buffer pool 10 times
+                        // and still haven't found a victim buffer, something's wrong.
+                        // Maybe all the buffers were in locked. That could happen in
+                        // theory, if you have more threads holding buffers locked than
+                        // there are buffers in the pool. In practice, with a reasonably
+                        // large buffer pool it really shouldn't happen.
                         if iters > iter_limit {
-                            // NB: Even with the permits, there's no hard guarantee that we will find a slot with
-                            // any particular number of iterations: other threads might race ahead and acquire and
-                            // release pins just as we're scanning the array.
-                            //
-                            // Imagine that nslots is 2, and as starting point, usage_count==1 on all
-                            // slots. There are two threads running concurrently, A and B. A has just
-                            // acquired the permit from the semaphore.
-                            //
-                            //   A: Look at slot 1. Its usage_count == 1, so decrement it to zero, and continue the search
-                            //   B: Acquire permit.
-                            //   B: Look at slot 2, decrement its usage_count to zero and continue the search
-                            //   B: Look at slot 1. Its usage_count is zero, so pin it and bump up its usage_count to 1.
-                            //   B: Release pin and permit again
-                            //   B: Acquire permit.
-                            //   B: Look at slot 2. Its usage_count is zero, so pin it and bump up its usage_count to 1.
-                            //   B: Release pin and permit again
-                            //
-                            // Now we're back in the starting situation that both slots have
-                            // usage_count 1, but A has now been through one iteration of the
-                            // find_victim() loop. This can repeat indefinitely and on each
-                            // iteration, A's iteration count increases by one.
-                            //
-                            // So, even though the semaphore for the permits is fair, the victim search
-                            // itself happens in parallel and is not fair.
-                            // Hence even with a permit, a task can theoretically be starved.
-                            // To avoid this, we'd need tokio to give priority to tasks that are holding
-                            // permits for longer.
-                            // Note that just yielding to tokio during iteration without such
-                            // priority boosting is likely counter-productive. We'd just give more opportunities
-                            // for B to bump usage count, further starving A.
-                            crate::metrics::page_cache_errors_inc(
-                                crate::metrics::PageCacheErrorKind::EvictIterLimit,
-                            );
                             anyhow::bail!("exceeded evict iter limit");
                         }
                         continue;
@@ -941,7 +788,6 @@ impl PageCache {
                     self.remove_mapping(old_key);
                     inner.key = None;
                 }
-                crate::metrics::PAGE_CACHE_FIND_VICTIMS_ITERS_TOTAL.inc_by(iters as u64);
                 return Ok((slot_idx, inner));
             }
         }
@@ -953,9 +799,8 @@ impl PageCache {
     fn new(num_pages: usize) -> Self {
         assert!(num_pages > 0, "page cache size must be > 0");
 
-        // We could use Vec::leak here, but that potentially also leaks
-        // uninitialized reserved capacity. With into_boxed_slice and Box::leak
-        // this is avoided.
+        // We use Box::leak here and into_boxed_slice to avoid leaking uninitialized
+        // memory that Vec's might contain.
         let page_buffer = Box::leak(vec![0u8; num_pages * PAGE_SZ].into_boxed_slice());
 
         let size_metrics = &crate::metrics::PAGE_CACHE_SIZE;
@@ -969,11 +814,7 @@ impl PageCache {
                 let buf: &mut [u8; PAGE_SZ] = chunk.try_into().unwrap();
 
                 Slot {
-                    inner: tokio::sync::RwLock::new(SlotInner {
-                        key: None,
-                        buf,
-                        permit: std::sync::Mutex::new(Weak::new()),
-                    }),
+                    inner: tokio::sync::RwLock::new(SlotInner { key: None, buf }),
                     usage_count: AtomicU8::new(0),
                 }
             })
@@ -985,7 +826,6 @@ impl PageCache {
             slots,
             next_evict_slot: AtomicUsize::new(0),
             size_metrics,
-            pinned_slots: Arc::new(tokio::sync::Semaphore::new(num_pages)),
         }
     }
 }

pageserver/src/page_service.rs

@@ -412,53 +412,31 @@ impl PageServerHandler {
             // TODO: We could create a new per-request context here, with unique ID.
             // Currently we use the same per-timeline context for all requests
 
-            let (response, span) = match neon_fe_msg {
+            let response = match neon_fe_msg {
                 PagestreamFeMessage::Exists(req) => {
                     let _timer = metrics.start_timer(metrics::SmgrQueryType::GetRelExists);
-                    let span = tracing::info_span!("handle_get_rel_exists_request", rel = %req.rel, req_lsn = %req.lsn);
-                    (
-                        self.handle_get_rel_exists_request(&timeline, &req, &ctx)
-                            .instrument(span.clone())
-                            .await,
-                        span,
-                    )
+                    self.handle_get_rel_exists_request(&timeline, &req, &ctx)
+                        .await
                 }
                 PagestreamFeMessage::Nblocks(req) => {
                     let _timer = metrics.start_timer(metrics::SmgrQueryType::GetRelSize);
-                    let span = tracing::info_span!("handle_get_nblocks_request", rel = %req.rel, req_lsn = %req.lsn);
-                    (
-                        self.handle_get_nblocks_request(&timeline, &req, &ctx)
-                            .instrument(span.clone())
-                            .await,
-                        span,
-                    )
+                    self.handle_get_nblocks_request(&timeline, &req, &ctx).await
                 }
                 PagestreamFeMessage::GetPage(req) => {
                     let _timer = metrics.start_timer(metrics::SmgrQueryType::GetPageAtLsn);
-                    let span = tracing::info_span!("handle_get_page_at_lsn_request", rel = %req.rel, blkno = %req.blkno, req_lsn = %req.lsn);
-                    (
-                        self.handle_get_page_at_lsn_request(&timeline, &req, &ctx)
-                            .instrument(span.clone())
-                            .await,
-                        span,
-                    )
+                    self.handle_get_page_at_lsn_request(&timeline, &req, &ctx)
+                        .await
                 }
                 PagestreamFeMessage::DbSize(req) => {
                     let _timer = metrics.start_timer(metrics::SmgrQueryType::GetDbSize);
-                    let span = tracing::info_span!("handle_db_size_request", dbnode = %req.dbnode, req_lsn = %req.lsn);
-                    (
-                        self.handle_db_size_request(&timeline, &req, &ctx)
-                            .instrument(span.clone())
-                            .await,
-                        span,
-                    )
+                    self.handle_db_size_request(&timeline, &req, &ctx).await
                 }
             };
 
             let response = response.unwrap_or_else(|e| {
                 // print the all details to the log with {:#}, but for the client the
                 // error message is enough
-                span.in_scope(|| error!("error reading relation or page version: {:#}", e));
+                error!("error reading relation or page version: {:?}", e);
                 PagestreamBeMessage::Error(PagestreamErrorResponse {
                     message: e.to_string(),
                 })
@@ -649,6 +627,7 @@ impl PageServerHandler {
         Ok(lsn)
     }
 
+    #[instrument(skip(self, timeline, req, ctx), fields(rel = %req.rel, req_lsn = %req.lsn))]
     async fn handle_get_rel_exists_request(
         &self,
         timeline: &Timeline,
@@ -669,6 +648,7 @@ impl PageServerHandler {
         }))
     }
 
+    #[instrument(skip(self, timeline, req, ctx), fields(rel = %req.rel, req_lsn = %req.lsn))]
     async fn handle_get_nblocks_request(
         &self,
         timeline: &Timeline,
@@ -687,6 +667,7 @@ impl PageServerHandler {
         }))
     }
 
+    #[instrument(skip(self, timeline, req, ctx), fields(dbnode = %req.dbnode, req_lsn = %req.lsn))]
     async fn handle_db_size_request(
         &self,
         timeline: &Timeline,
@@ -708,6 +689,7 @@ impl PageServerHandler {
         }))
     }
 
+    #[instrument(skip(self, timeline, req, ctx), fields(rel = %req.rel, blkno = %req.blkno, req_lsn = %req.lsn))]
     async fn handle_get_page_at_lsn_request(
         &self,
         timeline: &Timeline,
@@ -1283,10 +1265,7 @@ async fn get_active_tenant_with_timeout(
         Ok(tenant) => tenant,
         Err(e @ GetTenantError::NotFound(_)) => return Err(GetActiveTenantError::NotFound(e)),
         Err(GetTenantError::NotActive(_)) => {
-            unreachable!("we're calling get_tenant with active_only=false")
-        }
-        Err(GetTenantError::Broken(_)) => {
-            unreachable!("we're calling get_tenant with active_only=false")
+            unreachable!("we're calling get_tenant with active=false")
         }
     };
     let wait_time = Duration::from_secs(30);