Try larger sleep

This test failed consistently on `main` now. It's better to temporarily disable it to avoid blocking others' PRs while investigating the root cause for the test failure.

See: #2255, #2256

.circleci/config.yml

@@ -1,369 +0,0 @@
-version: 2.1
-
-executors:
-  neon-xlarge-executor:
-    resource_class: xlarge
-    docker:
-      # NB: when changed, do not forget to update rust image tag in all Dockerfiles
-      - image: neondatabase/rust:1.58
-  neon-executor:
-    docker:
-      - image: neondatabase/rust:1.58
-
-jobs:
-  # A job to build postgres
-  build-postgres:
-    executor: neon-xlarge-executor
-    parameters:
-      build_type:
-        type: enum
-        enum: ["debug", "release"]
-    environment:
-      BUILD_TYPE: << parameters.build_type >>
-    steps:
-        # Checkout the git repo (circleci doesn't have a flag to enable submodules here)
-      - checkout
-
-        # Grab the postgres git revision to build a cache key.
-        # Append makefile as it could change the way postgres is built.
-        # Note this works even though the submodule hasn't been checkout out yet.
-      - run:
-          name: Get postgres cache key
-          command: |
-              git rev-parse HEAD:vendor/postgres > /tmp/cache-key-postgres
-              cat Makefile >> /tmp/cache-key-postgres
-
-      - restore_cache:
-          name: Restore postgres cache
-          keys:
-            # Restore ONLY if the rev key matches exactly
-            - v05-postgres-cache-<< parameters.build_type >>-{{ checksum "/tmp/cache-key-postgres" }}
-
-        # Build postgres if the restore_cache didn't find a build.
-        # `make` can't figure out whether the cache is valid, since
-        # it only compares file timestamps.
-      - run:
-          name: build postgres
-          command: |
-            if [ ! -e tmp_install/bin/postgres ]; then
-              # "depth 1" saves some time by not cloning the whole repo
-              git submodule update --init --depth 1
-              # bail out on any warnings
-              COPT='-Werror' mold -run make postgres -j$(nproc)
-            fi
-
-      - save_cache:
-          name: Save postgres cache
-          key: v05-postgres-cache-<< parameters.build_type >>-{{ checksum "/tmp/cache-key-postgres" }}
-          paths:
-            - tmp_install
-
-  # A job to build Neon rust code
-  build-neon:
-    executor: neon-xlarge-executor
-    parameters:
-      build_type:
-        type: enum
-        enum: ["debug", "release"]
-    environment:
-      BUILD_TYPE: << parameters.build_type >>
-    steps:
-        # Checkout the git repo (without submodules)
-      - checkout
-
-        # Grab the postgres git revision to build a cache key.
-        # Append makefile as it could change the way postgres is built.
-        # Note this works even though the submodule hasn't been checkout out yet.
-      - run:
-          name: Get postgres cache key
-          command: |
-            git rev-parse HEAD:vendor/postgres > /tmp/cache-key-postgres
-            cat Makefile >> /tmp/cache-key-postgres
-
-
-      - restore_cache:
-          name: Restore postgres cache
-          keys:
-            # Restore ONLY if the rev key matches exactly
-            - v05-postgres-cache-<< parameters.build_type >>-{{ checksum "/tmp/cache-key-postgres" }}
-
-      - restore_cache:
-          name: Restore rust cache
-          keys:
-            # Require an exact match. While an out of date cache might speed up the build,
-            # there's no way to clean out old packages, so the cache grows every time something
-            # changes.
-            - v05-rust-cache-deps-<< parameters.build_type >>-{{ checksum "Cargo.lock" }}
-
-        # Build the rust code, including test binaries
-      - run:
-          name: Rust build << parameters.build_type >>
-          command: |
-            if [[ $BUILD_TYPE == "debug" ]]; then
-              CARGO_FLAGS=
-            elif [[ $BUILD_TYPE == "release" ]]; then
-              CARGO_FLAGS="--release --features profiling"
-            fi
-
-            export CARGO_INCREMENTAL=0
-            export CACHEPOT_BUCKET=zenith-rust-cachepot
-            export RUSTC_WRAPPER=""
-            export AWS_ACCESS_KEY_ID="${CACHEPOT_AWS_ACCESS_KEY_ID}"
-            export AWS_SECRET_ACCESS_KEY="${CACHEPOT_AWS_SECRET_ACCESS_KEY}"
-            mold -run cargo build $CARGO_FLAGS --features failpoints --bins --tests
-            cachepot -s
-
-      - save_cache:
-          name: Save rust cache
-          key: v05-rust-cache-deps-<< parameters.build_type >>-{{ checksum "Cargo.lock" }}
-          paths:
-            - ~/.cargo/registry
-            - ~/.cargo/git
-            - target
-
-        # Run rust unit tests
-      - run:
-          name: cargo test
-          command: |
-            if [[ $BUILD_TYPE == "debug" ]]; then
-              CARGO_FLAGS=
-            elif [[ $BUILD_TYPE == "release" ]]; then
-              CARGO_FLAGS=--release
-            fi
-
-            cargo test $CARGO_FLAGS
-
-        # Install the rust binaries, for use by test jobs
-      - run:
-          name: Install rust binaries
-          command: |
-            binaries=$(
-              cargo metadata --format-version=1 --no-deps |
-              jq -r '.packages[].targets[] | select(.kind | index("bin")) | .name'
-            )
-
-            mkdir -p /tmp/zenith/bin
-            mkdir -p /tmp/zenith/test_bin
-            mkdir -p /tmp/zenith/etc
-
-            # Install target binaries
-            for bin in $binaries; do
-              SRC=target/$BUILD_TYPE/$bin
-              DST=/tmp/zenith/bin/$bin
-              cp $SRC $DST
-            done
-
-        # Install the postgres binaries, for use by test jobs
-      - run:
-          name: Install postgres binaries
-          command: |
-            cp -a tmp_install /tmp/zenith/pg_install
-
-      # Save rust binaries for other jobs in the workflow
-      - persist_to_workspace:
-          root: /tmp/zenith
-          paths:
-            - "*"
-
-  check-codestyle-python:
-    executor: neon-executor
-    steps:
-      - checkout
-      - restore_cache:
-          keys:
-            - v2-python-deps-{{ checksum "poetry.lock" }}
-      - run:
-          name: Install deps
-          command: ./scripts/pysync
-      - save_cache:
-          key: v2-python-deps-{{ checksum "poetry.lock" }}
-          paths:
-            - /home/circleci/.cache/pypoetry/virtualenvs
-      - run:
-          name: Print versions
-          when: always
-          command: |
-              poetry run python --version
-              poetry show
-      - run:
-          name: Run yapf to ensure code format
-          when: always
-          command: poetry run yapf --recursive --diff .
-      - run:
-          name: Run mypy to check types
-          when: always
-          command: poetry run mypy .
-
-  run-pytest:
-    executor: neon-executor
-    parameters:
-      # pytest args to specify the tests to run.
-      #
-      # This can be a test file name, e.g. 'test_pgbench.py, or a subdirectory,
-      # or '-k foobar' to run tests containing string 'foobar'. See pytest man page
-      # section SPECIFYING TESTS / SELECTING TESTS for details.
-      #
-      # Select the type of Rust build. Must be "release" or "debug".
-      build_type:
-        type: string
-        default: "debug"
-      # This parameter is required, to prevent the mistake of running all tests in one job.
-      test_selection:
-        type: string
-        default: ""
-      # Arbitrary parameters to pytest. For example "-s" to prevent capturing stdout/stderr
-      extra_params:
-        type: string
-        default: ""
-      needs_postgres_source:
-        type: boolean
-        default: false
-      run_in_parallel:
-        type: boolean
-        default: true
-      save_perf_report:
-        type: boolean
-        default: false
-    environment:
-      BUILD_TYPE: << parameters.build_type >>
-    steps:
-      - attach_workspace:
-          at: /tmp/zenith
-      - checkout
-      - when:
-          condition: << parameters.needs_postgres_source >>
-          steps:
-            - run: git submodule update --init --depth 1
-      - restore_cache:
-          keys:
-            - v2-python-deps-{{ checksum "poetry.lock" }}
-      - run:
-          name: Install deps
-          command: ./scripts/pysync
-      - save_cache:
-          key: v2-python-deps-{{ checksum "poetry.lock" }}
-          paths:
-            - /home/circleci/.cache/pypoetry/virtualenvs
-      - run:
-          name: Run pytest
-          # pytest doesn't output test logs in real time, so CI job may fail with
-          # `Too long with no output` error, if a test is running for a long time.
-          # In that case, tests should have internal timeouts that are less than
-          # no_output_timeout, specified here.
-          no_output_timeout: 10m
-          environment:
-            - NEON_BIN: /tmp/zenith/bin
-            - POSTGRES_DISTRIB_DIR: /tmp/zenith/pg_install
-            - TEST_OUTPUT: /tmp/test_output
-            # this variable will be embedded in perf test report
-            # and is needed to distinguish different environments
-            - PLATFORM: zenith-local-ci
-          command: |
-            PERF_REPORT_DIR="$(realpath test_runner/perf-report-local)"
-            rm -rf $PERF_REPORT_DIR
-
-            TEST_SELECTION="test_runner/<< parameters.test_selection >>"
-            EXTRA_PARAMS="<< parameters.extra_params >>"
-            if [ -z "$TEST_SELECTION" ]; then
-              echo "test_selection must be set"
-              exit 1
-            fi
-            if << parameters.run_in_parallel >>; then
-              EXTRA_PARAMS="-n4 $EXTRA_PARAMS"
-            fi
-            if << parameters.save_perf_report >>; then
-              if [[ $CIRCLE_BRANCH == "main" ]]; then
-                mkdir -p "$PERF_REPORT_DIR"
-                EXTRA_PARAMS="--out-dir $PERF_REPORT_DIR $EXTRA_PARAMS"
-              fi
-            fi
-
-            export GITHUB_SHA=$CIRCLE_SHA1
-
-            # Run the tests.
-            #
-            # The junit.xml file allows CircleCI to display more fine-grained test information
-            # in its "Tests" tab in the results page.
-            # --verbose prints name of each test (helpful when there are
-            # multiple tests in one file)
-            # -rA prints summary in the end
-            # -n4 uses four processes to run tests via pytest-xdist
-            # -s is not used to prevent pytest from capturing output, because tests are running
-            # in parallel and logs are mixed between different tests
-            ./scripts/pytest \
-              --junitxml=$TEST_OUTPUT/junit.xml \
-              --tb=short \
-              --verbose \
-              -m "not remote_cluster" \
-              -rA $TEST_SELECTION $EXTRA_PARAMS
-
-            if << parameters.save_perf_report >>; then
-              if [[ $CIRCLE_BRANCH == "main" ]]; then
-                export REPORT_FROM="$PERF_REPORT_DIR"
-                export REPORT_TO=local
-                scripts/generate_and_push_perf_report.sh
-              fi
-            fi
-      - run:
-          # CircleCI artifacts are preserved one file at a time, so skipping
-          # this step isn't a good idea. If you want to extract the
-          # pageserver state, perhaps a tarball would be a better idea.
-          name: Delete all data but logs
-          when: always
-          command: |
-            du -sh /tmp/test_output/*
-            find /tmp/test_output -type f ! -name "*.log" ! -name "regression.diffs" ! -name "junit.xml" ! -name "*.filediff" ! -name "*.stdout" ! -name "*.stderr" ! -name "flamegraph.svg" ! -name "*.metrics" -delete
-            du -sh /tmp/test_output/*
-      - store_artifacts:
-          path: /tmp/test_output
-      # The store_test_results step tells CircleCI where to find the junit.xml file.
-      - store_test_results:
-          path: /tmp/test_output
-      # Save data (if any)
-      - persist_to_workspace:
-          root: /tmp/zenith
-          paths:
-            - "*"
-
-workflows:
-  build_and_test:
-    jobs:
-      - check-codestyle-python
-      - build-postgres:
-          name: build-postgres-<< matrix.build_type >>
-          matrix:
-            parameters:
-              build_type: ["debug", "release"]
-      - build-neon:
-          name: build-neon-<< matrix.build_type >>
-          matrix:
-            parameters:
-              build_type: ["debug", "release"]
-          requires:
-            - build-postgres-<< matrix.build_type >>
-      - run-pytest:
-          name: pg_regress-tests-<< matrix.build_type >>
-          matrix:
-            parameters:
-              build_type: ["debug", "release"]
-          test_selection: batch_pg_regress
-          needs_postgres_source: true
-          requires:
-            - build-neon-<< matrix.build_type >>
-      - run-pytest:
-          name: other-tests-<< matrix.build_type >>
-          matrix:
-            parameters:
-              build_type: ["debug", "release"]
-          test_selection: batch_others
-          requires:
-            - build-neon-<< matrix.build_type >>
-      - run-pytest:
-          name: benchmarks
-          context: PERF_TEST_RESULT_CONNSTR
-          build_type: release
-          test_selection: performance
-          run_in_parallel: false
-          save_perf_report: true
-          requires:
-            - build-neon-release

.github/actions/download/action.yml

@@ -0,0 +1,56 @@
+name: "Download an artifact"
+description: "Custom download action"
+inputs:
+  name:
+    description: "Artifact name"
+    required: true
+  path:
+    description: "A directory to put artifact into"
+    default: "."
+    required: false
+  skip-if-does-not-exist:
+    description: "Allow to skip if file doesn't exist, fail otherwise"
+    default: false
+    required: false
+
+runs:
+  using: "composite"
+  steps:
+    - name: Download artifact
+      id: download-artifact
+      shell: bash -euxo pipefail {0}
+      env:
+        TARGET: ${{ inputs.path }}
+        ARCHIVE: /tmp/downloads/${{ inputs.name }}.tar.zst
+        SKIP_IF_DOES_NOT_EXIST: ${{ inputs.skip-if-does-not-exist }}
+      run: |
+        BUCKET=neon-github-public-dev
+        PREFIX=artifacts/${GITHUB_RUN_ID}
+        FILENAME=$(basename $ARCHIVE)
+
+        S3_KEY=$(aws s3api list-objects-v2 --bucket ${BUCKET} --prefix ${PREFIX} | jq -r '.Contents[].Key' | grep ${FILENAME} | sort --version-sort | tail -1 || true)
+        if [ -z "${S3_KEY}" ]; then
+          if [ "${SKIP_IF_DOES_NOT_EXIST}" = "true" ]; then
+            echo '::set-output name=SKIPPED::true'
+            exit 0
+          else
+            echo 2>&1 "Neither s3://${BUCKET}/${PREFIX}/${GITHUB_RUN_ATTEMPT}/${FILENAME} nor its version from previous attempts exist"
+            exit 1
+          fi
+        fi
+
+        echo '::set-output name=SKIPPED::false'
+
+        mkdir -p $(dirname $ARCHIVE)
+        time aws s3 cp --only-show-errors s3://${BUCKET}/${S3_KEY} ${ARCHIVE}
+
+    - name: Extract artifact
+      if: ${{ steps.download-artifact.outputs.SKIPPED == 'false' }}
+      shell: bash -euxo pipefail {0}
+      env:
+        TARGET: ${{ inputs.path }}
+        ARCHIVE: /tmp/downloads/${{ inputs.name }}.tar.zst
+      run: |
+        mkdir -p ${TARGET}
+        time tar -xf ${ARCHIVE} -C ${TARGET}
+        rm -f ${ARCHIVE}

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

@@ -27,22 +27,35 @@ inputs:
     description: 'Whether to upload the performance report'
     required: false
     default: 'false'
+  run_with_real_s3:
+    description: 'Whether to pass real s3 credentials to the test suite'
+    required: false
+    default: 'false'
+  real_s3_bucket:
+    description: 'Bucket name for real s3 tests'
+    required: false
+    default: ''
+  real_s3_region:
+    description: 'Region name for real s3 tests'
+    required: false
+    default: ''
+  real_s3_access_key_id:
+    description: 'Access key id'
+    required: false
+    default: ''
+  real_s3_secret_access_key:
+    description: 'Secret access key'
+    required: false
+    default: ''
 
 runs:
   using: "composite"
   steps:
-    - name: Get Neon artifact for restoration
-      uses: actions/download-artifact@v3
+    - name: Get Neon artifact
+      uses: ./.github/actions/download
       with:
         name: neon-${{ runner.os }}-${{ inputs.build_type }}-${{ inputs.rust_toolchain }}-artifact
-        path: ./neon-artifact/
-
-    - name: Extract Neon artifact
-      shell: bash -ex {0}
-      run: |
-        mkdir -p /tmp/neon/
-        tar -xf ./neon-artifact/neon.tgz -C /tmp/neon/
-        rm -rf ./neon-artifact/
+        path: /tmp/neon
 
     - name: Checkout
       if: inputs.needs_postgres_source == 'true'
@@ -59,7 +72,7 @@ runs:
         key: v1-${{ runner.os }}-python-deps-${{ hashFiles('poetry.lock') }}
 
     - name: Install Python deps
-      shell: bash -ex {0}
+      shell: bash -euxo pipefail {0}
       run: ./scripts/pysync
 
     - name: Run pytest
@@ -70,7 +83,10 @@ runs:
         # this variable will be embedded in perf test report
         # and is needed to distinguish different environments
         PLATFORM: github-actions-selfhosted
-      shell: bash -ex {0}
+        BUILD_TYPE: ${{ inputs.build_type }}
+        AWS_ACCESS_KEY_ID: ${{ inputs.real_s3_access_key_id }}
+        AWS_SECRET_ACCESS_KEY: ${{ inputs.real_s3_secret_access_key }}
+      shell: bash -euxo pipefail {0}
       run: |
         PERF_REPORT_DIR="$(realpath test_runner/perf-report-local)"
         rm -rf $PERF_REPORT_DIR
@@ -84,6 +100,14 @@ runs:
         if [[ "${{ inputs.run_in_parallel }}" == "true" ]]; then
           EXTRA_PARAMS="-n4 $EXTRA_PARAMS"
         fi
+
+        if [[ "${{ inputs.run_with_real_s3 }}" == "true" ]]; then
+          echo "REAL S3 ENABLED"
+          export ENABLE_REAL_S3_REMOTE_STORAGE=nonempty
+          export REMOTE_STORAGE_S3_BUCKET=${{ inputs.real_s3_bucket }}
+          export REMOTE_STORAGE_S3_REGION=${{ inputs.real_s3_region }}
+        fi
+
         if [[ "${{ inputs.save_perf_report }}" == "true" ]]; then
           if [[ "$GITHUB_REF" == "refs/heads/main" ]]; then
             mkdir -p "$PERF_REPORT_DIR"
@@ -99,7 +123,7 @@ runs:
 
         # Run the tests.
         #
-        # The junit.xml file allows CircleCI to display more fine-grained test information
+        # The junit.xml file allows CI tools to display more fine-grained test information
         # in its "Tests" tab in the results page.
         # --verbose prints name of each test (helpful when there are
         # multiple tests in one file)
@@ -123,7 +147,7 @@ runs:
         fi
 
     - name: Delete all data but logs
-      shell: bash -ex {0}
+      shell: bash -euxo pipefail {0}
       if: always()
       run: |
         du -sh /tmp/test_output/*
@@ -132,9 +156,7 @@ runs:
 
     - name: Upload python test logs
       if: always()
-      uses: actions/upload-artifact@v3
+      uses: ./.github/actions/upload
       with:
-        retention-days: 7
-        if-no-files-found: error
         name: python-test-${{ inputs.test_selection }}-${{ runner.os }}-${{ inputs.build_type }}-${{ inputs.rust_toolchain }}-logs
         path: /tmp/test_output/

.github/actions/save-coverage-data/action.yml

@@ -5,13 +5,18 @@ runs:
   using: "composite"
   steps:
     - name: Merge coverage data
-      shell: bash -ex {0}
+      shell: bash -euxo pipefail {0}
       run: scripts/coverage "--profraw-prefix=$GITHUB_JOB" --dir=/tmp/coverage merge
 
-    - name: Upload coverage data
-      uses: actions/upload-artifact@v3
+    - name: Download previous coverage data into the same directory
+      uses: ./.github/actions/download
       with:
-        retention-days: 7
-        if-no-files-found: error
         name: coverage-data-artifact
-        path: /tmp/coverage/
+        path: /tmp/coverage
+        skip-if-does-not-exist: true # skip if there's no previous coverage to download
+
+    - name: Upload coverage data
+      uses: ./.github/actions/upload
+      with:
+        name: coverage-data-artifact
+        path: /tmp/coverage

.github/actions/upload/action.yml

@@ -0,0 +1,51 @@
+name: "Upload an artifact"
+description: "Custom upload action"
+inputs:
+  name:
+    description: "Artifact name"
+    required: true
+  path:
+    description: "A directory or file to upload"
+    required: true
+
+runs:
+  using: "composite"
+  steps:
+    - name: Prepare artifact
+      shell: bash -euxo pipefail {0}
+      env:
+        SOURCE: ${{ inputs.path }}
+        ARCHIVE: /tmp/uploads/${{ inputs.name }}.tar.zst
+      run: |
+        mkdir -p $(dirname $ARCHIVE)
+
+        if [ -f ${ARCHIVE} ]; then
+          echo 2>&1 "File ${ARCHIVE} already exist. Something went wrong before"
+          exit 1
+        fi
+
+        ZSTD_NBTHREADS=0
+        if [ -d  ${SOURCE} ]; then
+          time tar -C ${SOURCE} -cf ${ARCHIVE} --zstd .
+        elif [ -f ${SOURCE} ]; then
+          time tar -cf ${ARCHIVE} --zstd ${SOURCE}
+        else
+          echo 2>&1 "${SOURCE} neither directory nor file, don't know how to handle it"
+        fi
+
+    - name: Upload artifact
+      shell: bash -euxo pipefail {0}
+      env:
+        SOURCE: ${{ inputs.path }}
+        ARCHIVE: /tmp/uploads/${{ inputs.name }}.tar.zst
+      run: |
+        BUCKET=neon-github-public-dev
+        PREFIX=artifacts/${GITHUB_RUN_ID}
+        FILENAME=$(basename $ARCHIVE)
+
+        FILESIZE=$(du -sh ${ARCHIVE} | cut -f1)
+
+        time aws s3 mv --only-show-errors ${ARCHIVE} s3://${BUCKET}/${PREFIX}/${GITHUB_RUN_ATTEMPT}/${FILENAME}
+
+        # Ref https://docs.github.com/en/actions/using-workflows/workflow-commands-for-github-actions#adding-a-job-summary
+        echo "[${FILENAME}](https://${BUCKET}.s3.amazonaws.com/${PREFIX}/${GITHUB_RUN_ATTEMPT}/${FILENAME}) ${FILESIZE}" >> ${GITHUB_STEP_SUMMARY}

.github/ansible/production.hosts

@@ -17,4 +17,4 @@ env_name = prod-1
 console_mgmt_base_url = http://console-release.local
 bucket_name           = zenith-storage-oregon
 bucket_region         = us-west-2
-etcd_endpoints        = etcd-release.local:2379
+etcd_endpoints        = zenith-1-etcd.local:2379

.github/ansible/scripts/init_safekeeper.sh

@@ -12,10 +12,9 @@ cat <<EOF | tee /tmp/payload
   "version": 1,
   "host": "${HOST}",
   "port": 6500,
+  "http_port": 7676,
   "region_id": {{ console_region_id }},
-  "instance_id": "${INSTANCE_ID}",
-  "http_host": "${HOST}",
-  "http_port": 7676
+  "instance_id": "${INSTANCE_ID}"
 }
 EOF
 

.github/ansible/staging.hosts

@@ -17,4 +17,4 @@ env_name = us-stage
 console_mgmt_base_url = http://console-staging.local
 bucket_name           = zenith-staging-storage-us-east-1
 bucket_region         = us-east-1
-etcd_endpoints        = etcd-staging.local:2379
+etcd_endpoints        = zenith-us-stage-etcd.local:2379

.github/workflows/benchmarking.yml

@@ -11,7 +11,7 @@ on:
     #          │ │ ┌───────────── day of the month (1 - 31)
     #          │ │ │ ┌───────────── month (1 - 12 or JAN-DEC)
     #          │ │ │ │ ┌───────────── day of the week (0 - 6 or SUN-SAT)
-    - cron:  '36 7 * * *' # run once a day, timezone is utc
+    - cron:  '36 4 * * *' # run once a day, timezone is utc
 
   workflow_dispatch: # adds ability to run this manually
 
@@ -60,7 +60,7 @@ jobs:
     - name: Setup cluster
       env:
         BENCHMARK_CONNSTR: "${{ secrets.BENCHMARK_STAGING_CONNSTR }}"
-      shell: bash
+      shell: bash -euxo pipefail {0}
       run: |
         set -e
 
@@ -104,3 +104,12 @@ jobs:
         PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
       run: |
         REPORT_FROM=$(realpath perf-report-staging) REPORT_TO=staging scripts/generate_and_push_perf_report.sh
+
+    - name: Post to a Slack channel
+      if: ${{ github.event.schedule && failure() }}
+      uses: slackapi/slack-github-action@v1
+      with:
+        channel-id: "C033QLM5P7D" # dev-staging-stream
+        slack-message: "Periodic perf testing: ${{ job.status }}\n${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
+      env:
+        SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}

.github/workflows/build_and_test.yml

@@ -1,14 +1,15 @@
-name: Test
+name: Test and Deploy
 
 on:
   push:
     branches:
-    - main
+      - main
+      - release
   pull_request:
 
 defaults:
   run:
-    shell: bash -ex {0}
+    shell: bash -euxo pipefail {0}
 
 concurrency:
   # Allow only one workflow per any non-`main` branch.
@@ -20,8 +21,9 @@ env:
   COPT: '-Werror'
 
 jobs:
-  build-postgres:
-    runs-on: [ self-hosted, Linux, k8s-runner ]
+  build-neon:
+    runs-on: dev
+    container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rustlegacy:2746987948
     strategy:
       fail-fast: false
       matrix:
@@ -30,7 +32,19 @@ jobs:
 
     env:
       BUILD_TYPE: ${{ matrix.build_type }}
+      GIT_VERSION: ${{ github.sha }}
+
     steps:
+      - name: Fix git ownerwhip
+        run: |
+          # Workaround for `fatal: detected dubious ownership in repository at ...`
+          #
+          # Use both ${{ github.workspace }} and ${GITHUB_WORKSPACE} because they're different on host and in containers
+          #   Ref https://github.com/actions/checkout/issues/785
+          #
+          git config --global --add safe.directory ${{ github.workspace }}
+          git config --global --add safe.directory ${GITHUB_WORKSPACE}
+
       - name: Checkout
         uses: actions/checkout@v3
         with:
@@ -41,6 +55,47 @@ jobs:
         id: pg_ver
         run: echo ::set-output name=pg_rev::$(git rev-parse HEAD:vendor/postgres)
 
+      # Set some environment variables used by all the steps.
+      #
+      # CARGO_FLAGS is extra options to pass to "cargo build", "cargo test" etc.
+      #   It also includes --features, if any
+      #
+      # CARGO_FEATURES is passed to "cargo metadata". It is separate from CARGO_FLAGS,
+      #   because "cargo metadata" doesn't accept --release or --debug options
+      #
+      - name: Set env variables
+        run: |
+          if [[ $BUILD_TYPE == "debug" ]]; then
+            cov_prefix="scripts/coverage --profraw-prefix=$GITHUB_JOB --dir=/tmp/coverage run"
+            CARGO_FEATURES=""
+            CARGO_FLAGS=""
+          elif [[ $BUILD_TYPE == "release" ]]; then
+            cov_prefix=""
+            CARGO_FEATURES="--features profiling"
+            CARGO_FLAGS="--release $CARGO_FEATURES"
+          fi
+          echo "cov_prefix=${cov_prefix}" >> $GITHUB_ENV
+          echo "CARGO_FEATURES=${CARGO_FEATURES}" >> $GITHUB_ENV
+          echo "CARGO_FLAGS=${CARGO_FLAGS}" >> $GITHUB_ENV
+
+      # Don't include the ~/.cargo/registry/src directory. It contains just
+      # uncompressed versions of the crates in ~/.cargo/registry/cache
+      # directory, and it's faster to let 'cargo' to rebuild it from the
+      # compressed crates.
+      - name: Cache cargo deps
+        id: cache_cargo
+        uses: actions/cache@v3
+        with:
+          path: |
+            ~/.cargo/registry/
+            !~/.cargo/registry/src
+            ~/.cargo/git/
+            target/
+          # Fall back to older versions of the key, if no cache for current Cargo.lock was found
+          key: |
+            v3-${{ runner.os }}-${{ matrix.build_type }}-cargo-${{ matrix.rust_toolchain }}-${{ hashFiles('Cargo.lock') }}
+            v3-${{ runner.os }}-${{ matrix.build_type }}-cargo-${{ matrix.rust_toolchain }}-
+
       - name: Cache postgres build
         id: cache_pg
         uses: actions/cache@v3
@@ -52,111 +107,22 @@ jobs:
         if: steps.cache_pg.outputs.cache-hit != 'true'
         run: mold -run make postgres -j$(nproc)
 
-      # actions/cache@v3 does not allow concurrently using the same cache across job steps, so use a separate cache
-      - name: Prepare postgres artifact
-        run: tar -C tmp_install/ -czf ./pg.tgz .
-      - name: Upload postgres artifact
-        uses: actions/upload-artifact@v3
-        with:
-          retention-days: 7
-          if-no-files-found: error
-          name: postgres-${{ runner.os }}-${{ matrix.build_type }}-artifact
-          path: ./pg.tgz
-
-
-  build-neon:
-    runs-on: [ self-hosted, Linux, k8s-runner ]
-    needs: [ build-postgres ]
-    strategy:
-      fail-fast: false
-      matrix:
-        build_type: [ debug, release ]
-        rust_toolchain: [ 1.58 ]
-
-    env:
-      BUILD_TYPE: ${{ matrix.build_type }}
-
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v3
-        with:
-          submodules: true
-          fetch-depth: 1
-
-      - name: Get postgres artifact for restoration
-        uses: actions/download-artifact@v3
-        with:
-          name: postgres-${{ runner.os }}-${{ matrix.build_type }}-artifact
-          path: ./postgres-artifact/
-      - name: Extract postgres artifact
-        run: |
-          mkdir ./tmp_install/
-          tar -xf ./postgres-artifact/pg.tgz -C ./tmp_install/
-          rm -rf ./postgres-artifact/
-
-      - name: Cache cargo deps
-        id: cache_cargo
-        uses: actions/cache@v3
-        with:
-          path: |
-            ~/.cargo/registry/
-            ~/.cargo/git/
-            target/
-          # Fall back to older versions of the key, if no cache for current Cargo.lock was found
-          key: |
-            v2-${{ runner.os }}-${{ matrix.build_type }}-cargo-${{ matrix.rust_toolchain }}-${{ hashFiles('Cargo.lock') }}
-            v2-${{ runner.os }}-${{ matrix.build_type }}-cargo-${{ matrix.rust_toolchain }}-
-
       - name: Run cargo build
         run: |
-          if [[ $BUILD_TYPE == "debug" ]]; then
-            cov_prefix=(scripts/coverage "--profraw-prefix=$GITHUB_JOB" --dir=/tmp/coverage run)
-            CARGO_FLAGS=
-          elif [[ $BUILD_TYPE == "release" ]]; then
-            cov_prefix=()
-            CARGO_FLAGS="--release --features profiling"
-          fi
-
-          "${cov_prefix[@]}" mold -run cargo build $CARGO_FLAGS --features failpoints --bins --tests
+          ${cov_prefix} mold -run cargo build $CARGO_FLAGS --features failpoints --bins --tests
 
       - name: Run cargo test
         run: |
-          if [[ $BUILD_TYPE == "debug" ]]; then
-            cov_prefix=(scripts/coverage "--profraw-prefix=$GITHUB_JOB" --dir=/tmp/coverage run)
-            CARGO_FLAGS=
-          elif [[ $BUILD_TYPE == "release" ]]; then
-            cov_prefix=()
-            CARGO_FLAGS=--release
-          fi
-
-          "${cov_prefix[@]}" cargo test $CARGO_FLAGS
+          ${cov_prefix} cargo test $CARGO_FLAGS
 
       - name: Install rust binaries
         run: |
-          if [[ $BUILD_TYPE == "debug" ]]; then
-            cov_prefix=(scripts/coverage "--profraw-prefix=$GITHUB_JOB" --dir=/tmp/coverage run)
-          elif [[ $BUILD_TYPE == "release" ]]; then
-            cov_prefix=()
-          fi
-
+          # Install target binaries
+          mkdir -p /tmp/neon/bin/
           binaries=$(
-            "${cov_prefix[@]}" cargo metadata --format-version=1 --no-deps |
+            ${cov_prefix} cargo metadata $CARGO_FEATURES --format-version=1 --no-deps |
             jq -r '.packages[].targets[] | select(.kind | index("bin")) | .name'
           )
-
-          test_exe_paths=$(
-            "${cov_prefix[@]}" cargo test --message-format=json --no-run |
-            jq -r '.executable | select(. != null)'
-          )
-
-          mkdir -p /tmp/neon/bin/
-          mkdir -p /tmp/neon/test_bin/
-          mkdir -p /tmp/neon/etc/
-
-          # Keep bloated coverage data files away from the rest of the artifact
-          mkdir -p /tmp/coverage/
-
-          # Install target binaries
           for bin in $binaries; do
             SRC=target/$BUILD_TYPE/$bin
             DST=/tmp/neon/bin/$bin
@@ -165,9 +131,15 @@ jobs:
 
           # Install test executables and write list of all binaries (for code coverage)
           if [[ $BUILD_TYPE == "debug" ]]; then
-            for bin in $binaries; do
-              echo "/tmp/neon/bin/$bin" >> /tmp/coverage/binaries.list
-            done
+            # Keep bloated coverage data files away from the rest of the artifact
+            mkdir -p /tmp/coverage/
+
+            mkdir -p /tmp/neon/test_bin/
+
+            test_exe_paths=$(
+              ${cov_prefix} cargo test $CARGO_FLAGS --message-format=json --no-run |
+              jq -r '.executable | select(. != null)'
+            )
             for bin in $test_exe_paths; do
               SRC=$bin
               DST=/tmp/neon/test_bin/$(basename $bin)
@@ -177,30 +149,29 @@ jobs:
               strip "$SRC" -o "$DST"
               echo "$DST" >> /tmp/coverage/binaries.list
             done
+
+            for bin in $binaries; do
+              echo "/tmp/neon/bin/$bin" >> /tmp/coverage/binaries.list
+            done
           fi
 
       - name: Install postgres binaries
         run: cp -a tmp_install /tmp/neon/pg_install
 
-      - name: Prepare neon artifact
-        run: tar -C /tmp/neon/ -czf ./neon.tgz .
-
-      - name: Upload neon binaries
-        uses: actions/upload-artifact@v3
+      - name: Upload Neon artifact
+        uses: ./.github/actions/upload
         with:
-          retention-days: 7
-          if-no-files-found: error
           name: neon-${{ runner.os }}-${{ matrix.build_type }}-${{ matrix.rust_toolchain }}-artifact
-          path: ./neon.tgz
+          path: /tmp/neon
 
       # XXX: keep this after the binaries.list is formed, so the coverage can properly work later
       - name: Merge and upload coverage data
         if: matrix.build_type == 'debug'
         uses: ./.github/actions/save-coverage-data
 
-
   pg_regress-tests:
-    runs-on: [ self-hosted, Linux, k8s-runner ]
+    runs-on: dev
+    container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rustlegacy:2746987948
     needs: [ build-neon ]
     strategy:
       fail-fast: false
@@ -227,7 +198,8 @@ jobs:
         uses: ./.github/actions/save-coverage-data
 
   other-tests:
-    runs-on: [ self-hosted, Linux, k8s-runner ]
+    runs-on: dev
+    container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rustlegacy:2746987948
     needs: [ build-neon ]
     strategy:
       fail-fast: false
@@ -247,14 +219,20 @@ jobs:
           build_type: ${{ matrix.build_type }}
           rust_toolchain: ${{ matrix.rust_toolchain }}
           test_selection: batch_others
-
+          run_with_real_s3: true
+          real_s3_bucket: ci-tests-s3
+          real_s3_region: us-west-2
+          real_s3_access_key_id: "${{ secrets.AWS_ACCESS_KEY_ID_CI_TESTS_S3 }}"
+          real_s3_secret_access_key: "${{ secrets.AWS_SECRET_ACCESS_KEY_CI_TESTS_S3 }}"
       - name: Merge and upload coverage data
         if: matrix.build_type == 'debug'
         uses: ./.github/actions/save-coverage-data
 
   benchmarks:
-    runs-on: [ self-hosted, Linux, k8s-runner ]
+    runs-on: dev
+    container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rustlegacy:2746987948
     needs: [ build-neon ]
+    if: github.ref_name == 'main' || contains(github.event.pull_request.labels.*.name, 'run-benchmarks')
     strategy:
       fail-fast: false
       matrix:
@@ -282,7 +260,8 @@ jobs:
       # while coverage is currently collected for the debug ones
 
   coverage-report:
-    runs-on: [ self-hosted, Linux, k8s-runner ]
+    runs-on: dev
+    container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/rustlegacy:2746987948
     needs: [ other-tests, pg_regress-tests ]
     strategy:
       fail-fast: false
@@ -302,27 +281,22 @@ jobs:
         with:
           path: |
             ~/.cargo/registry/
+            !~/.cargo/registry/src
             ~/.cargo/git/
             target/
-          key: v2-${{ runner.os }}-${{ matrix.build_type }}-cargo-${{ matrix.rust_toolchain }}-${{ hashFiles('Cargo.lock') }}
+          key: v3-${{ runner.os }}-${{ matrix.build_type }}-cargo-${{ matrix.rust_toolchain }}-${{ hashFiles('Cargo.lock') }}
 
-      - name: Get Neon artifact for restoration
-        uses: actions/download-artifact@v3
+      - name: Get Neon artifact
+        uses: ./.github/actions/download
         with:
           name: neon-${{ runner.os }}-${{ matrix.build_type }}-${{ matrix.rust_toolchain }}-artifact
-          path: ./neon-artifact/
+          path: /tmp/neon
 
-      - name: Extract Neon artifact
-        run: |
-          mkdir -p /tmp/neon/
-          tar -xf ./neon-artifact/neon.tgz -C /tmp/neon/
-          rm -rf ./neon-artifact/
-
-      - name: Restore coverage data
-        uses: actions/download-artifact@v3
+      - name: Get coverage artifact
+        uses: ./.github/actions/download
         with:
           name: coverage-data-artifact
-          path: /tmp/coverage/
+          path: /tmp/coverage
 
       - name: Merge coverage data
         run: scripts/coverage "--profraw-prefix=$GITHUB_JOB" --dir=/tmp/coverage merge
@@ -360,40 +334,40 @@ jobs:
             }"
 
   trigger-e2e-tests:
-   runs-on: [ self-hosted, Linux, k8s-runner ]
-   needs: [ build-neon ]
-   steps:
-     - name: Set PR's status to pending and request a remote CI test
-       run: |
-         COMMIT_SHA=${{ github.event.pull_request.head.sha }}
-         COMMIT_SHA=${COMMIT_SHA:-${{ github.sha }}}
+    runs-on: [ self-hosted, Linux, k8s-runner ]
+    needs: [ build-neon ]
+    steps:
+      - name: Set PR's status to pending and request a remote CI test
+        run: |
+          COMMIT_SHA=${{ github.event.pull_request.head.sha }}
+          COMMIT_SHA=${COMMIT_SHA:-${{ github.sha }}}
 
-         REMOTE_REPO="${{ github.repository_owner }}/cloud"
+          REMOTE_REPO="${{ github.repository_owner }}/cloud"
 
-         curl -f -X POST \
-         https://api.github.com/repos/${{ github.repository }}/statuses/$COMMIT_SHA \
-         -H "Accept: application/vnd.github.v3+json" \
-         --user "${{ secrets.CI_ACCESS_TOKEN }}" \
-         --data \
-           "{
-             \"state\": \"pending\",
-             \"context\": \"neon-cloud-e2e\",
-             \"description\": \"[$REMOTE_REPO] Remote CI job is about to start\"
-           }"
+          curl -f -X POST \
+          https://api.github.com/repos/${{ github.repository }}/statuses/$COMMIT_SHA \
+          -H "Accept: application/vnd.github.v3+json" \
+          --user "${{ secrets.CI_ACCESS_TOKEN }}" \
+          --data \
+            "{
+              \"state\": \"pending\",
+              \"context\": \"neon-cloud-e2e\",
+              \"description\": \"[$REMOTE_REPO] Remote CI job is about to start\"
+            }"
 
-         curl -f -X POST \
-         https://api.github.com/repos/$REMOTE_REPO/actions/workflows/testing.yml/dispatches \
-         -H "Accept: application/vnd.github.v3+json" \
-         --user "${{ secrets.CI_ACCESS_TOKEN }}" \
-         --data \
-           "{
-             \"ref\": \"main\",
-             \"inputs\": {
-               \"ci_job_name\": \"neon-cloud-e2e\",
-               \"commit_hash\": \"$COMMIT_SHA\",
-               \"remote_repo\": \"${{ github.repository }}\"
-             }
-           }"
+          curl -f -X POST \
+          https://api.github.com/repos/$REMOTE_REPO/actions/workflows/testing.yml/dispatches \
+          -H "Accept: application/vnd.github.v3+json" \
+          --user "${{ secrets.CI_ACCESS_TOKEN }}" \
+          --data \
+            "{
+              \"ref\": \"main\",
+              \"inputs\": {
+                \"ci_job_name\": \"neon-cloud-e2e\",
+                \"commit_hash\": \"$COMMIT_SHA\",
+                \"remote_repo\": \"${{ github.repository }}\"
+              }
+            }"
 
   docker-image:
     runs-on: [ self-hosted, Linux, k8s-runner ]
@@ -436,9 +410,9 @@ jobs:
       - name: Get legacy build tag
         run: |
           if [[ "$GITHUB_REF_NAME" == "main" ]]; then
-            echo "::set-output name=tag::latest
+            echo "::set-output name=tag::latest"
           elif [[ "$GITHUB_REF_NAME" == "release" ]]; then
-            echo "::set-output name=tag::release
+            echo "::set-output name=tag::release"
           else
             echo "GITHUB_REF_NAME (value '$GITHUB_REF_NAME') is not set to either 'main' or 'release'"
             exit 1
@@ -498,9 +472,9 @@ jobs:
       - name: Get legacy build tag
         run: |
           if [[ "$GITHUB_REF_NAME" == "main" ]]; then
-            echo "::set-output name=tag::latest
+            echo "::set-output name=tag::latest"
           elif [[ "$GITHUB_REF_NAME" == "release" ]]; then
-            echo "::set-output name=tag::release
+            echo "::set-output name=tag::release"
           else
             echo "GITHUB_REF_NAME (value '$GITHUB_REF_NAME') is not set to either 'main' or 'release'"
             exit 1

.github/workflows/codestyle.yml

@@ -8,7 +8,7 @@ on:
 
 defaults:
   run:
-    shell: bash -ex {0}
+    shell: bash -euxo pipefail {0}
 
 concurrency:
   # Allow only one workflow per any non-`main` branch.
@@ -27,7 +27,7 @@ jobs:
         # Rust toolchains (e.g. nightly or 1.37.0), add them here.
         rust_toolchain: [1.58]
         os: [ubuntu-latest, macos-latest]
-    timeout-minutes: 50
+    timeout-minutes: 60
     name: run regression test suite
     runs-on: ${{ matrix.os }}
 
@@ -98,9 +98,10 @@ jobs:
         with:
           path: |
             ~/.cargo/registry
+            !~/.cargo/registry/src
             ~/.cargo/git
             target
-          key: ${{ runner.os }}-cargo-${{ hashFiles('./Cargo.lock') }}-rust-${{ matrix.rust_toolchain }}
+          key: v1-${{ runner.os }}-cargo-${{ hashFiles('./Cargo.lock') }}-rust-${{ matrix.rust_toolchain }}
 
       - name: Run cargo clippy
         run: ./run_clippy.sh

.github/workflows/pg_clients.yml

@@ -40,7 +40,7 @@ jobs:
         key: v1-${{ runner.os }}-python-deps-${{ hashFiles('poetry.lock') }}
 
     - name: Install Python deps
-      shell: bash -ex {0}
+      shell: bash -euxo pipefail {0}
       run: ./scripts/pysync
 
     - name: Run pytest
@@ -49,7 +49,7 @@ jobs:
         BENCHMARK_CONNSTR: "${{ secrets.BENCHMARK_STAGING_CONNSTR }}"
         TEST_OUTPUT: /tmp/test_output
         POSTGRES_DISTRIB_DIR: /tmp/neon/pg_install
-      shell: bash -ex {0}
+      shell: bash -euxo pipefail {0}
       run: |
         # Test framework expects we have psql binary;
         # but since we don't really need it in this test, let's mock it

CONTRIBUTING.md

@@ -11,17 +11,15 @@ than it was before.
 
 ## Submitting changes
 
-1. Make a PR for every change.
-
-   Even seemingly trivial patches can break things in surprising ways.
-Use of common sense is OK. If you're only fixing a typo in a comment,
-it's probably fine to just push it. But if in doubt, open a PR.
-
-2. Get at least one +1 on your PR before you push.
+1. Get at least one +1 on your PR before you push.
 
    For simple patches, it will only take a minute for someone to review
 it.
 
+2. Don't force push small changes after making the PR ready for review.
+Doing so will force readers to re-read your entire PR, which will delay
+the review process.
+
 3. Always keep the CI green.
 
    Do not push, if the CI failed on your PR. Even if you think it's not

Cargo.lock

@@ -154,9 +154,9 @@ checksum = "d468802bab17cbc0cc575e9b053f41e72aa36bfa6b7f55e3529ffa43161b97fa"
 
 [[package]]
 name = "axum"
-version = "0.5.12"
+version = "0.5.13"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d16705af05732b7d3258ec0f7b73c03a658a28925e050d8852d5b568ee8bcf4e"
+checksum = "6b9496f0c1d1afb7a2af4338bbe1d969cddfead41d87a9fb3aaa6d0bbc7af648"
 dependencies = [
  "async-trait",
  "axum-core",
@@ -317,15 +317,6 @@ dependencies = [
  "serde",
 ]
 
-[[package]]
-name = "cast"
-version = "0.2.7"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4c24dab4283a142afa2fdca129b80ad2c6284e073930f964c3a1293c225ee39a"
-dependencies = [
- "rustc_version",
-]
-
 [[package]]
 name = "cast"
 version = "0.3.0"
@@ -467,7 +458,6 @@ dependencies = [
  "clap 3.2.12",
  "env_logger",
  "hyper",
- "libc",
  "log",
  "postgres",
  "regex",
@@ -505,8 +495,8 @@ name = "control_plane"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "lazy_static",
  "nix",
+ "once_cell",
  "pageserver",
  "postgres",
  "regex",
@@ -517,7 +507,6 @@ dependencies = [
  "tar",
  "thiserror",
  "toml",
- "url",
  "utils",
  "workspace_hack",
 ]
@@ -581,7 +570,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b01d6de93b2b6c65e17c634a26653a29d107b3c98c607c765bf38d041531cd8f"
 dependencies = [
  "atty",
- "cast 0.3.0",
+ "cast",
  "clap 2.34.0",
  "criterion-plot",
  "csv",
@@ -602,11 +591,11 @@ dependencies = [
 
 [[package]]
 name = "criterion-plot"
-version = "0.4.4"
+version = "0.4.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d00996de9f2f7559f7f4dc286073197f83e92256a59ed395f9aac01fe717da57"
+checksum = "2673cc8207403546f45f5fd319a974b1e6983ad1a3ee7e6041650013be041876"
 dependencies = [
- "cast 0.2.7",
+ "cast",
  "itertools",
 ]
 
@@ -682,9 +671,9 @@ dependencies = [
 
 [[package]]
 name = "crypto-common"
-version = "0.1.5"
+version = "0.1.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2ccfd8c0ee4cce11e45b3fd6f9d5e69e0cc62912aa6a0cb1bf4617b0eba5a12f"
+checksum = "1bfb12502f3fc46cca1bb51ac28df9d618d813cdc3d2f25b9fe775a34af26bb3"
 dependencies = [
  "generic-array",
  "typenum",
@@ -1118,9 +1107,9 @@ dependencies = [
 
 [[package]]
 name = "gimli"
-version = "0.26.1"
+version = "0.26.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "78cc372d058dcf6d5ecd98510e7fbc9e5aec4d21de70f65fea8fecebcd881bd4"
+checksum = "22030e2c5a68ec659fde1e949a745124b48e6fa8b045b7ed5bd1fe4ccc5c4e5d"
 
 [[package]]
 name = "git-version"
@@ -1186,9 +1175,9 @@ dependencies = [
 
 [[package]]
 name = "hashbrown"
-version = "0.12.2"
+version = "0.12.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "607c8a29735385251a339424dd462993c0fed8fa09d378f259377df08c126022"
+checksum = "8a9ee70c43aaf417c914396645a0fa852624801b24ebb7ae78fe8272889ac888"
 
 [[package]]
 name = "heck"
@@ -1390,7 +1379,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "10a35a97730320ffe8e2d410b5d3b69279b98d2c14bdb8b70ea89ecf7888d41e"
 dependencies = [
  "autocfg",
- "hashbrown 0.12.2",
+ "hashbrown 0.12.3",
 ]
 
 [[package]]
@@ -1602,7 +1591,6 @@ dependencies = [
 name = "metrics"
 version = "0.1.0"
 dependencies = [
- "lazy_static",
  "libc",
  "once_cell",
  "prometheus",
@@ -1677,7 +1665,6 @@ dependencies = [
  "git-version",
  "pageserver",
  "postgres",
- "postgres_ffi",
  "safekeeper",
  "serde_json",
  "utils",
@@ -1855,9 +1842,9 @@ dependencies = [
 
 [[package]]
 name = "os_str_bytes"
-version = "6.1.0"
+version = "6.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "21326818e99cfe6ce1e524c2a805c189a99b5ae555a35d19f9a284b427d86afa"
+checksum = "648001efe5d5c0102d8cea768e348da85d90af8ba91f0bea908f157951493cd4"
 
 [[package]]
 name = "pageserver"
@@ -1883,7 +1870,6 @@ dependencies = [
  "humantime-serde",
  "hyper",
  "itertools",
- "lazy_static",
  "metrics",
  "nix",
  "once_cell",
@@ -1905,7 +1891,6 @@ dependencies = [
  "thiserror",
  "tokio",
  "tokio-postgres",
- "tokio-stream",
  "toml_edit",
  "tracing",
  "url",
@@ -2130,9 +2115,9 @@ dependencies = [
  "crc32c",
  "env_logger",
  "hex",
- "lazy_static",
  "log",
  "memoffset",
+ "once_cell",
  "postgres",
  "rand",
  "regex",
@@ -2292,9 +2277,9 @@ dependencies = [
  "hex",
  "hmac 0.12.1",
  "hyper",
- "lazy_static",
  "md5",
  "metrics",
+ "once_cell",
  "parking_lot 0.12.1",
  "pin-project-lite",
  "rand",
@@ -2740,9 +2725,9 @@ dependencies = [
 
 [[package]]
 name = "rustversion"
-version = "1.0.7"
+version = "1.0.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a0a5f7c728f5d284929a1cccb5bc19884422bfe6ef4d6c409da2c41838983fcf"
+checksum = "24c8ad4f0c00e1eb5bc7614d236a7f1300e3dbd76b68cac8e06fb00b015ad8d8"
 
 [[package]]
 name = "ryu"
@@ -2764,12 +2749,10 @@ dependencies = [
  "daemonize",
  "etcd_broker",
  "fs2",
- "futures",
  "git-version",
  "hex",
  "humantime",
  "hyper",
- "lazy_static",
  "metrics",
  "once_cell",
  "postgres",
@@ -2784,12 +2767,10 @@ dependencies = [
  "tempfile",
  "tokio",
  "tokio-postgres",
- "tokio-util",
  "toml_edit",
  "tracing",
  "url",
  "utils",
- "walkdir",
  "workspace_hack",
 ]
 
@@ -3625,9 +3606,9 @@ checksum = "099b7128301d285f79ddd55b9a83d5e6b9e97c92e0ea0daebee7263e932de992"
 
 [[package]]
 name = "unicode-ident"
-version = "1.0.1"
+version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5bd2fe26506023ed7b5e1e315add59d6f584c621d037f9368fea9cfb988f368c"
+checksum = "15c61ba63f9235225a22310255a29b806b907c9b8c964bcbd0a2c70f3f2deea7"
 
 [[package]]
 name = "unicode-normalization"
@@ -3688,9 +3669,9 @@ dependencies = [
  "hex-literal",
  "hyper",
  "jsonwebtoken",
- "lazy_static",
  "metrics",
  "nix",
+ "once_cell",
  "pin-project-lite",
  "postgres",
  "postgres-protocol",

Dockerfile

@@ -17,6 +17,10 @@ RUN set -e \
 FROM neondatabase/rust:1.58 AS build
 ARG GIT_VERSION=local
 
+# Enable https://github.com/paritytech/cachepot to cache Rust crates' compilation results in Docker builds.
+# Set up cachepot to use an AWS S3 bucket for cache results, to reuse it between `docker build` invocations.
+# cachepot falls back to local filesystem if S3 is misconfigured, not failing the build.
+ARG RUSTC_WRAPPER=cachepot
 ARG CACHEPOT_BUCKET=zenith-rust-cachepot
 ARG AWS_ACCESS_KEY_ID
 ARG AWS_SECRET_ACCESS_KEY

Dockerfile.compute-tools

@@ -1,7 +1,11 @@
 # First transient image to build compute_tools binaries
-# NB: keep in sync with rust image version in .circle/config.yml
+# NB: keep in sync with rust image version in .github/workflows/build_and_test.yml
 FROM neondatabase/rust:1.58 AS rust-build
 
+# Enable https://github.com/paritytech/cachepot to cache Rust crates' compilation results in Docker builds.
+# Set up cachepot to use an AWS S3 bucket for cache results, to reuse it between `docker build` invocations.
+# cachepot falls back to local filesystem if S3 is misconfigured, not failing the build.
+ARG RUSTC_WRAPPER=cachepot
 ARG CACHEPOT_BUCKET=zenith-rust-cachepot
 ARG AWS_ACCESS_KEY_ID
 ARG AWS_SECRET_ACCESS_KEY

Makefile

@@ -29,9 +29,11 @@ else
 endif
 
 # macOS with brew-installed openssl requires explicit paths
+# It can be configured with OPENSSL_PREFIX variable
 UNAME_S := $(shell uname -s)
 ifeq ($(UNAME_S),Darwin)
-    PG_CONFIGURE_OPTS += --with-includes=$(HOMEBREW_PREFIX)/opt/openssl/include --with-libraries=$(HOMEBREW_PREFIX)/opt/openssl/lib
+    OPENSSL_PREFIX ?= $(shell brew --prefix openssl@3)
+    PG_CONFIGURE_OPTS += --with-includes=$(OPENSSL_PREFIX)/include --with-libraries=$(OPENSSL_PREFIX)/lib
 endif
 
 # Choose whether we should be silent or verbose

README.md

@@ -1,6 +1,6 @@
 # Neon
 
-Neon is a serverless open source alternative to AWS Aurora Postgres. It separates storage and compute and substitutes PostgreSQL storage layer by redistributing data across a cluster of nodes.
+Neon is a serverless open-source alternative to AWS Aurora Postgres. It separates storage and compute and substitutes the PostgreSQL storage layer by redistributing data across a cluster of nodes.
 
 The project used to be called "Zenith". Many of the commands and code comments
 still refer to "zenith", but we are in the process of renaming things.
@@ -12,32 +12,31 @@ Alternatively, compile and run the project [locally](#running-local-installation
 
 ## Architecture overview
 
-A Neon installation consists of compute nodes and Neon storage engine.
+A Neon installation consists of compute nodes and a Neon storage engine.
 
-Compute nodes are stateless PostgreSQL nodes, backed by Neon storage engine.
+Compute nodes are stateless PostgreSQL nodes backed by the Neon storage engine.
 
-Neon storage engine consists of two major components:
-- Pageserver. Scalable storage backend for compute nodes.
-- WAL service. The service that receives WAL from compute node and ensures that it is stored durably.
+The Neon storage engine consists of two major components:
+- Pageserver. Scalable storage backend for the compute nodes.
+- WAL service. The service receives WAL from the compute node and ensures that it is stored durably.
 
 Pageserver consists of:
 - Repository - Neon storage implementation.
 - WAL receiver - service that receives WAL from WAL service and stores it in the repository.
 - Page service - service that communicates with compute nodes and responds with pages from the repository.
-- WAL redo - service that builds pages from base images and WAL records on Page service request.
-
+- WAL redo - service that builds pages from base images and WAL records on Page service request
 ## Running local installation
 
 
 #### Installing dependencies on Linux
-1. Install build dependencies and other useful packages
+1. Install build dependencies and other applicable packages
 
-* On Ubuntu or Debian this set of packages should be sufficient to build the code:
+* On Ubuntu or Debian, this set of packages should be sufficient to build the code:
 ```bash
 apt install build-essential libtool libreadline-dev zlib1g-dev flex bison libseccomp-dev \
 libssl-dev clang pkg-config libpq-dev etcd cmake postgresql-client
 ```
-* On Fedora these packages are needed:
+* On Fedora, these packages are needed:
 ```bash
 dnf install flex bison readline-devel zlib-devel openssl-devel \
   libseccomp-devel perl clang cmake etcd postgresql postgresql-contrib
@@ -69,7 +68,7 @@ brew install libpq
 brew link --force libpq
 ```
 
-#### Building on Linux and OSX
+#### Building on Linux
 
 1. Build neon and patched postgres
 ```
@@ -80,19 +79,35 @@ cd neon
 
 # The preferred and default is to make a debug build. This will create a 
 # demonstrably slower build than a release build. If you want to use a release
-# build, utilize "`BUILD_TYPE=release make -j`nproc``" 
+# build, utilize "BUILD_TYPE=release make -j`nproc`" 
 
 make -j`nproc`
 ```
 
-#### dependency installation notes
+#### Building on OSX
+
+1. Build neon and patched postgres
+```
+# Note: The path to the neon sources can not contain a space.
+
+git clone --recursive https://github.com/neondatabase/neon.git
+cd neon
+
+# The preferred and default is to make a debug build. This will create a 
+# demonstrably slower build than a release build. If you want to use a release
+# build, utilize "BUILD_TYPE=release make -j`sysctl -n hw.logicalcpu`" 
+
+make -j`sysctl -n hw.logicalcpu`
+```
+
+#### Dependency installation notes
 To run the `psql` client, install the `postgresql-client` package or modify `PATH` and `LD_LIBRARY_PATH` to include `tmp_install/bin` and `tmp_install/lib`, respectively.
 
 To run the integration tests or Python scripts (not required to use the code), install
-Python (3.9 or higher), and install python3 packages using `./scripts/pysync` (requires poetry) in the project directory.
+Python (3.9 or higher), and install python3 packages using `./scripts/pysync` (requires [poetry](https://python-poetry.org/)) in the project directory.
 
 
-#### running neon database
+#### Running neon database
 1. Start pageserver and postgres on top of it (should be called from repo root):
 ```sh
 # Create repository in .neon with proper paths to binaries and data
@@ -123,7 +138,7 @@ Starting postgres node at 'host=127.0.0.1 port=55432 user=cloud_admin dbname=pos
  main  127.0.0.1:55432  de200bd42b49cc1814412c7e592dd6e9  main         0/16B5BA8  running
 ```
 
-2. Now it is possible to connect to postgres and run some queries:
+2. Now, it is possible to connect to postgres and run some queries:
 ```text
 > psql -p55432 -h 127.0.0.1 -U cloud_admin postgres
 postgres=# CREATE TABLE t(key int primary key, value text);
@@ -181,14 +196,16 @@ postgres=# select * from t;
 (1 row)
 ```
 
-4. If you want to run tests afterwards (see below), you have to stop all the running the pageserver, safekeeper and postgres instances
-   you have just started. You can stop them all with one command:
+4. If you want to run tests afterward (see below), you must stop all the running of the pageserver, safekeeper, and postgres instances
+   you have just started. You can terminate them all with one command:
 ```sh
 > ./target/debug/neon_local stop
 ```
 
 ## Running tests
 
+Ensure your dependencies are installed as described [here](https://github.com/neondatabase/neon#dependency-installation-notes).
+
 ```sh
 git clone --recursive https://github.com/neondatabase/neon.git
 make # builds also postgres and installs it to ./tmp_install
@@ -205,8 +222,8 @@ To view your `rustdoc` documentation in a browser, try running `cargo doc --no-d
 
 ### Postgres-specific terms
 
-Due to Neon's very close relation with PostgreSQL internals, there are numerous specific terms used.
-Same applies to certain spelling: i.e. we use MB to denote 1024 * 1024 bytes, while MiB would be technically more correct, it's inconsistent with what PostgreSQL code and its documentation use.
+Due to Neon's very close relation with PostgreSQL internals, numerous specific terms are used.
+The same applies to certain spelling: i.e. we use MB to denote 1024 * 1024 bytes, while MiB would be technically more correct, it's inconsistent with what PostgreSQL code and its documentation use.
 
 To get more familiar with this aspect, refer to:
 

compute_tools/Cargo.toml

@@ -4,7 +4,6 @@ version = "0.1.0"
 edition = "2021"
 
 [dependencies]
-libc = "0.2"
 anyhow = "1.0"
 chrono = "0.4"
 clap = "3.0"

compute_tools/src/bin/compute_ctl.rs

@@ -157,7 +157,7 @@ fn main() -> Result<()> {
             exit(code)
         }
         Err(error) => {
-            error!("could not start the compute node: {}", error);
+            error!("could not start the compute node: {:?}", error);
 
             let mut state = compute.state.write().unwrap();
             state.error = Some(format!("{:?}", error));

compute_tools/src/spec.rs

@@ -1,8 +1,7 @@
 use std::path::Path;
 
-use anyhow::{anyhow, Result};
+use anyhow::Result;
 use log::{info, log_enabled, warn, Level};
-use postgres::error::SqlState;
 use postgres::{Client, NoTls};
 use serde::Deserialize;
 
@@ -395,20 +394,34 @@ pub fn handle_grants(node: &ComputeNode, client: &mut Client) -> Result<()> {
 
         // This will only change ownership on the schema itself, not the objects
         // inside it. Without it owner of the `public` schema will be `cloud_admin`
-        // and database owner cannot do anything with it.
-        let alter_query = format!("ALTER SCHEMA public OWNER TO {}", db.owner.quote());
-        let res = db_client.simple_query(&alter_query);
-
-        if let Err(e) = res {
-            if e.code() == Some(&SqlState::INVALID_SCHEMA_NAME) {
-                // This is OK, db just don't have a `public` schema.
-                // Probably user dropped it manually.
-                info!("no 'public' schema found in the database {}", db.name);
-            } else {
-                // Something different happened, propagate the error
-                return Err(anyhow!(e));
-            }
-        }
+        // and database owner cannot do anything with it. SQL procedure ensures
+        // that it won't error out if schema `public` doesn't exist.
+        let alter_query = format!(
+            "DO $$\n\
+                DECLARE\n\
+                    schema_owner TEXT;\n\
+                BEGIN\n\
+                    IF EXISTS(\n\
+                        SELECT nspname\n\
+                        FROM pg_catalog.pg_namespace\n\
+                        WHERE nspname = 'public'\n\
+                    )\n\
+                    THEN\n\
+                        SELECT nspowner::regrole::text\n\
+                            FROM pg_catalog.pg_namespace\n\
+                            WHERE nspname = 'public'\n\
+                            INTO schema_owner;\n\
+                \n\
+                        IF schema_owner = 'cloud_admin' OR schema_owner = 'zenith_admin'\n\
+                        THEN\n\
+                            ALTER SCHEMA public OWNER TO {};\n\
+                        END IF;\n\
+                    END IF;\n\
+                END\n\
+            $$;",
+            db.owner.quote()
+        );
+        db_client.simple_query(&alter_query)?;
     }
 
     Ok(())

control_plane/Cargo.toml

@@ -9,12 +9,11 @@ postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="d052ee8
 serde = { version = "1.0", features = ["derive"] }
 serde_with = "1.12.0"
 toml = "0.5"
-lazy_static = "1.4"
+once_cell = "1.13.0"
 regex = "1"
 anyhow = "1.0"
 thiserror = "1"
 nix = "0.23"
-url = "2.2.2"
 reqwest = { version = "0.11", default-features = false, features = ["blocking", "json", "rustls-tls"] }
 
 pageserver = { path = "../pageserver" }

control_plane/src/etcd.rs

@@ -30,14 +30,14 @@ pub fn start_etcd_process(env: &local_env::LocalEnv) -> anyhow::Result<()> {
     let etcd_stdout_file =
         fs::File::create(etcd_data_dir.join("etcd.stdout.log")).with_context(|| {
             format!(
-                "Failed to create ectd stout file in directory {}",
+                "Failed to create etcd stout file in directory {}",
                 etcd_data_dir.display()
             )
         })?;
     let etcd_stderr_file =
         fs::File::create(etcd_data_dir.join("etcd.stderr.log")).with_context(|| {
             format!(
-                "Failed to create ectd stderr file in directory {}",
+                "Failed to create etcd stderr file in directory {}",
                 etcd_data_dir.display()
             )
         })?;

control_plane/src/lib.rs

@@ -51,7 +51,11 @@ fn fill_rust_env_vars(cmd: &mut Command) -> &mut Command {
 }
 
 fn fill_aws_secrets_vars(mut cmd: &mut Command) -> &mut Command {
-    for env_key in ["AWS_ACCESS_KEY_ID", "AWS_SECRET_ACCESS_KEY"] {
+    for env_key in [
+        "AWS_ACCESS_KEY_ID",
+        "AWS_SECRET_ACCESS_KEY",
+        "AWS_SESSION_TOKEN",
+    ] {
         if let Ok(value) = std::env::var(env_key) {
             cmd = cmd.env(env_key, value);
         }

control_plane/src/postgresql_conf.rs

@@ -5,7 +5,7 @@
 /// enough to extract a few settings we need in Zenith, assuming you don't do
 /// funny stuff like include-directives or funny escaping.
 use anyhow::{bail, Context, Result};
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
 use regex::Regex;
 use std::collections::HashMap;
 use std::fmt;
@@ -19,9 +19,7 @@ pub struct PostgresConf {
     hash: HashMap<String, String>,
 }
 
-lazy_static! {
-    static ref CONF_LINE_RE: Regex = Regex::new(r"^((?:\w|\.)+)\s*=\s*(\S+)$").unwrap();
-}
+static CONF_LINE_RE: Lazy<Regex> = Lazy::new(|| Regex::new(r"^((?:\w|\.)+)\s*=\s*(\S+)$").unwrap());
 
 impl PostgresConf {
     pub fn new() -> PostgresConf {
@@ -139,10 +137,10 @@ fn escape_str(s: &str) -> String {
     //
     // This regex is a bit more conservative than the rules in guc-file.l, so we quote some
     // strings that PostgreSQL would accept without quoting, but that's OK.
-    lazy_static! {
-        static ref UNQUOTED_RE: Regex =
-            Regex::new(r"(^[-+]?[0-9]+[a-zA-Z]*$)|(^[a-zA-Z][a-zA-Z0-9]*$)").unwrap();
-    }
+
+    static UNQUOTED_RE: Lazy<Regex> =
+        Lazy::new(|| Regex::new(r"(^[-+]?[0-9]+[a-zA-Z]*$)|(^[a-zA-Z][a-zA-Z0-9]*$)").unwrap());
+
     if UNQUOTED_RE.is_match(s) {
         s.to_string()
     } else {

control_plane/src/safekeeper.rs

@@ -247,7 +247,7 @@ impl SafekeeperNode {
         // Shutting down may take a long time,
         // if safekeeper flushes a lot of data
         let mut tcp_stopped = false;
-        for _ in 0..100 {
+        for i in 0..600 {
             if !tcp_stopped {
                 if let Err(err) = TcpStream::connect(&address) {
                     tcp_stopped = true;
@@ -272,9 +272,11 @@ impl SafekeeperNode {
                     }
                 }
             }
-            print!(".");
-            io::stdout().flush().unwrap();
-            thread::sleep(Duration::from_secs(1));
+            if i % 10 == 0 {
+                print!(".");
+                io::stdout().flush().unwrap();
+            }
+            thread::sleep(Duration::from_millis(100));
         }
 
         bail!("Failed to stop safekeeper with pid {}", pid);
@@ -304,10 +306,9 @@ impl SafekeeperNode {
         Ok(self
             .http_request(
                 Method::POST,
-                format!("{}/{}", self.http_base_url, "timeline"),
+                format!("{}/tenant/{}/timeline", self.http_base_url, tenant_id),
             )
             .json(&TimelineCreateRequest {
-                tenant_id,
                 timeline_id,
                 peer_ids,
             })

control_plane/src/storage.rs

@@ -12,9 +12,9 @@ use anyhow::{bail, Context};
 use nix::errno::Errno;
 use nix::sys::signal::{kill, Signal};
 use nix::unistd::Pid;
-use pageserver::http::models::{TenantConfigRequest, TenantCreateRequest, TimelineCreateRequest};
-use pageserver::tenant_mgr::TenantInfo;
-use pageserver::timelines::TimelineInfo;
+use pageserver::http::models::{
+    TenantConfigRequest, TenantCreateRequest, TenantInfo, TimelineCreateRequest, TimelineInfo,
+};
 use postgres::{Config, NoTls};
 use reqwest::blocking::{Client, RequestBuilder, Response};
 use reqwest::{IntoUrl, Method};
@@ -318,7 +318,7 @@ impl PageServerNode {
         // Shutting down may take a long time,
         // if pageserver checkpoints a lot of data
         let mut tcp_stopped = false;
-        for _ in 0..100 {
+        for i in 0..600 {
             if !tcp_stopped {
                 if let Err(err) = TcpStream::connect(&address) {
                     tcp_stopped = true;
@@ -344,9 +344,11 @@ impl PageServerNode {
                     }
                 }
             }
-            print!(".");
-            io::stdout().flush().unwrap();
-            thread::sleep(Duration::from_secs(1));
+            if i % 10 == 0 {
+                print!(".");
+                io::stdout().flush().unwrap();
+            }
+            thread::sleep(Duration::from_millis(100));
         }
 
         bail!("Failed to stop pageserver with pid {}", pid);
@@ -399,6 +401,7 @@ impl PageServerNode {
                     .get("checkpoint_distance")
                     .map(|x| x.parse::<u64>())
                     .transpose()?,
+                checkpoint_timeout: settings.get("checkpoint_timeout").map(|x| x.to_string()),
                 compaction_target_size: settings
                     .get("compaction_target_size")
                     .map(|x| x.parse::<u64>())
@@ -453,6 +456,7 @@ impl PageServerNode {
                     .map(|x| x.parse::<u64>())
                     .transpose()
                     .context("Failed to parse 'checkpoint_distance' as an integer")?,
+                checkpoint_timeout: settings.get("checkpoint_timeout").map(|x| x.to_string()),
                 compaction_target_size: settings
                     .get("compaction_target_size")
                     .map(|x| x.parse::<u64>())

docker-entrypoint.sh

@@ -1,6 +1,8 @@
 #!/bin/sh
 set -eux
 
+pageserver_id_param="${NODE_ID:-10}"
+
 broker_endpoints_param="${BROKER_ENDPOINT:-absent}"
 if [ "$broker_endpoints_param" != "absent" ]; then
     broker_endpoints_param="-c broker_endpoints=['$broker_endpoints_param']"
@@ -8,10 +10,12 @@ else
     broker_endpoints_param=''
 fi
 
+remote_storage_param="${REMOTE_STORAGE:-}"
+
 if [ "$1" = 'pageserver' ]; then
     if [ ! -d "/data/tenants" ]; then
         echo "Initializing pageserver data directory"
-        pageserver --init -D /data -c "pg_distrib_dir='/usr/local'" -c "id=10" $broker_endpoints_param
+        pageserver --init -D /data -c "pg_distrib_dir='/usr/local'" -c "id=${pageserver_id_param}" $broker_endpoints_param $remote_storage_param
     fi
     echo "Staring pageserver at 0.0.0.0:6400"
     pageserver -c "listen_pg_addr='0.0.0.0:6400'" -c "listen_http_addr='0.0.0.0:9898'" $broker_endpoints_param -D /data

docs/.gitignore

@@ -0,0 +1 @@
+book

docs/README.md

@@ -1,14 +0,0 @@
-# Zenith documentation
-
-## Table of contents
-
-- [authentication.md](authentication.md) — pageserver JWT authentication.
-- [docker.md](docker.md) — Docker images and building pipeline.
-- [glossary.md](glossary.md) — Glossary of all the terms used in codebase.
-- [multitenancy.md](multitenancy.md) — how multitenancy is organized in the pageserver and Zenith CLI.
-- [sourcetree.md](sourcetree.md) — Overview of the source tree layout.
-- [pageserver/README.md](/pageserver/README.md) — pageserver overview.
-- [postgres_ffi/README.md](/libs/postgres_ffi/README.md) — Postgres FFI overview.
-- [test_runner/README.md](/test_runner/README.md) — tests infrastructure overview.
-- [safekeeper/README.md](/safekeeper/README.md) — WAL service overview.
-- [core_changes.md](core_changes.md) - Description of Zenith changes in Postgres core

docs/SUMMARY.md

@@ -0,0 +1,82 @@
+# Summary
+
+[Introduction]()
+- [Separation of Compute and Storage](./separation-compute-storage.md)
+
+# Architecture
+
+- [Compute]()
+  - [WAL proposer]()
+  - [WAL Backpressure]()
+  - [Postgres changes](./core_changes.md)
+
+- [Pageserver](./pageserver.md)
+    - [Services](./pageserver-services.md)
+    - [Thread management](./pageserver-thread-mgmt.md)
+    - [WAL Redo](./pageserver-walredo.md)
+    - [Page cache](./pageserver-pagecache.md)
+    - [Storage](./pageserver-storage.md)
+        - [Datadir mapping]()
+        - [Layer files]()
+        - [Branching]()
+        - [Garbage collection]()
+    - [Cloud Storage]()
+    - [Processing a GetPage request](./pageserver-processing-getpage.md)
+    - [Processing WAL](./pageserver-processing-wal.md)
+	- [Management API]()
+	- [Tenant Rebalancing]()
+
+- [WAL Service](walservice.md)
+  - [Consensus protocol](safekeeper-protocol.md)
+  - [Management API]()
+  - [Rebalancing]()
+
+- [Control Plane]()
+
+- [Proxy]()
+
+- [Source view](./sourcetree.md)
+  - [docker.md](./docker.md) — Docker images and building pipeline.
+  - [Error handling and logging]()
+  - [Testing]()
+    - [Unit testing]()
+    - [Integration testing]()
+    - [Benchmarks]()
+
+
+- [Glossary](./glossary.md)
+
+# Uncategorized
+
+- [authentication.md](./authentication.md)
+- [multitenancy.md](./multitenancy.md) — how multitenancy is organized in the pageserver and Zenith CLI.
+- [settings.md](./settings.md)
+#FIXME: move these under sourcetree.md
+#- [postgres_ffi/README.md](/libs/postgres_ffi/README.md)
+#- [test_runner/README.md](/test_runner/README.md)
+
+
+# RFCs
+
+- [RFCs](./rfcs/README.md)
+
+- [002-storage](rfcs/002-storage.md)
+- [003-laptop-cli](rfcs/003-laptop-cli.md)
+- [004-durability](rfcs/004-durability.md)
+- [005-zenith_local](rfcs/005-zenith_local.md)
+- [006-laptop-cli-v2-CLI](rfcs/006-laptop-cli-v2-CLI.md)
+- [006-laptop-cli-v2-repository-structure](rfcs/006-laptop-cli-v2-repository-structure.md)
+- [007-serverless-on-laptop](rfcs/007-serverless-on-laptop.md)
+- [008-push-pull](rfcs/008-push-pull.md)
+- [009-snapshot-first-storage-cli](rfcs/009-snapshot-first-storage-cli.md)
+- [009-snapshot-first-storage](rfcs/009-snapshot-first-storage.md)
+- [009-snapshot-first-storage-pitr](rfcs/009-snapshot-first-storage-pitr.md)
+- [010-storage_details](rfcs/010-storage_details.md)
+- [011-retention-policy](rfcs/011-retention-policy.md)
+- [012-background-tasks](rfcs/012-background-tasks.md)
+- [013-term-history](rfcs/013-term-history.md)
+- [014-safekeepers-gossip](rfcs/014-safekeepers-gossip.md)
+- [014-storage-lsm](rfcs/014-storage-lsm.md)
+- [015-storage-messaging](rfcs/015-storage-messaging.md)
+- [016-connection-routing](rfcs/016-connection-routing.md)
+- [cluster-size-limits](rfcs/cluster-size-limits.md)

docs/book.toml

@@ -0,0 +1,5 @@
+[book]
+language = "en"
+multilingual = false
+src = "."
+title = "Neon architecture"

docs/core_changes.md

@@ -1,202 +1,519 @@
-1. Add t_cid to XLOG record
-- Why?
-  The cmin/cmax on a heap page is a real bummer. I don't see any other way to fix that than bite the bullet and modify the WAL-logging routine to include the cmin/cmax.
+# Postgres core changes
 
-  To recap, the problem is that the XLOG_HEAP_INSERT record does not include the command id of the inserted row. And same with deletion/update. So in the primary, a row is inserted with current xmin + cmin. But in the replica, the cmin is always set to 1. That works, because the command id is only relevant to the inserting transaction itself. After commit/abort, no one cares abut it anymore.
+This lists all the changes that have been made to the PostgreSQL
+source tree, as a somewhat logical set of patches. The long-term goal
+is to eliminate all these changes, by submitting patches to upstream
+and refactoring code into extensions, so that you can run unmodified
+PostgreSQL against Neon storage.
 
-- Alternatives?
-  I don't know
+In Neon, we run PostgreSQL in the compute nodes, but we also run a special WAL redo process in the
+page server. We currently use the same binary for both, with --wal-redo runtime flag to launch it in
+the WAL redo mode. Some PostgreSQL changes are needed in the compute node, while others are just for
+the WAL redo process.
 
-2. Add PD_WAL_LOGGED.
-- Why?
-  Postgres sometimes writes data to the page before it is wal-logged. If such page ais swapped out, we  will loose this change. The problem is currently solved by setting PD_WAL_LOGGED bit in page header. When page without this bit set is written to the SMGR, then it is forced to be written to the WAL as FPI using log_newpage_copy() function.
+In addition to core PostgreSQL changes, there is a Neon extension in contrib/neon, to hook into the
+smgr interface. Once all the core changes have been submitted to upstream or eliminated some other
+way, the extension could live outside the postgres repository and build against vanilla PostgreSQL.
 
-  There was wrong assumption that it can happen only during construction of some exotic indexes (like gist). It is not true. The same situation can happen with COPY,VACUUM and when record hint bits are set.
+Below is a list of all the PostgreSQL source code changes, categorized into changes needed for
+compute, and changes needed for the WAL redo process:
 
-- Discussion:
-  https://discord.com/channels/869525774699462656/882681420986851359
+# Changes for Compute node
 
-- Alternatives:
-  Do not store this flag in page header, but associate this bit with shared buffer. Logically it is more correct but in practice we will get not advantages: neither in space, neither in CPU overhead.
+## Add t_cid to heap WAL records
+
+```
+ src/backend/access/heap/heapam.c                            |   26 +-
+ src/include/access/heapam_xlog.h                            |    6 +-
+```
+
+We have added a new t_cid field to heap WAL records. This changes the WAL record format, making Neon WAL format incompatible with vanilla PostgreSQL!
+
+### Problem we're trying to solve
+
+The problem is that the XLOG_HEAP_INSERT record does not include the command id of the inserted row. And same with deletion/update. So in the primary, a row is inserted with current xmin + cmin. But in the replica, the cmin is always set to 1. That works in PostgreSQL, because the command id is only relevant to the inserting transaction itself. After commit/abort, no one cares about it anymore. But with Neon, we rely on WAL replay to reconstruct the page, even while the original transaction is still running.
+
+### How to get rid of the patch
+
+Bite the bullet and submit the patch to PostgreSQL, to add the t_cid to the WAL records. It makes the WAL records larger, which could make this unpopular in the PostgreSQL community. However, it might simplify some logical decoding code; Andres Freund briefly mentioned in PGCon 2022 discussion on Heikki's Neon presentation that logical decoding currently needs to jump through some hoops to reconstruct the same information.
 
 
-3. XLogReadBufferForRedo not always loads and pins requested buffer. So we need to add extra checks that buffer is really pinned. Also do not use BufferGetBlockNumber for buffer returned by XLogReadBufferForRedo.
-- Why?
-  XLogReadBufferForRedo is not pinning pages which are not requested by wal-redo. It is specific only for wal-redo Postgres.
+### Alternatives
+Perhaps we could write an extra WAL record with the t_cid information, when a page is evicted that contains rows that were touched a transaction that's still running. However, that seems very complicated.
 
-- Alternatives?
-  No
+## ginfast.c
+
+```
+diff --git a/src/backend/access/gin/ginfast.c b/src/backend/access/gin/ginfast.c
+index e0d9940946..2d964c02e9 100644
+--- a/src/backend/access/gin/ginfast.c
++++ b/src/backend/access/gin/ginfast.c
+@@ -285,6 +285,17 @@ ginHeapTupleFastInsert(GinState *ginstate, GinTupleCollector *collector)
+                memset(&sublist, 0, sizeof(GinMetaPageData));
+                makeSublist(index, collector->tuples, collector->ntuples, &sublist);
+ 
++               if (metadata->head != InvalidBlockNumber)
++               {
++                       /*
++                        * ZENITH: Get buffer before XLogBeginInsert() to avoid recursive call
++                        * of XLogBeginInsert(). Reading a new buffer might evict a dirty page from
++                        * the buffer cache, and if that page happens to be an FSM or VM page, zenith_write()
++                        * will try to WAL-log an image of the page.
++                        */
++                       buffer = ReadBuffer(index, metadata->tail);
++               }
++
+                if (needWal)
+                        XLogBeginInsert();
+ 
+@@ -316,7 +327,6 @@ ginHeapTupleFastInsert(GinState *ginstate, GinTupleCollector *collector)
+                        data.prevTail = metadata->tail;
+                        data.newRightlink = sublist.head;
+ 
+-                       buffer = ReadBuffer(index, metadata->tail);
+                        LockBuffer(buffer, GIN_EXCLUSIVE);
+                        page = BufferGetPage(buffer);
+```
+
+The problem is explained in the comment above
+
+### How to get rid of the patch
+
+Can we stop WAL-logging FSM or VM pages? Or delay the WAL logging until we're out of the critical
+section or something.
+
+Maybe some bigger rewrite of FSM and VM would help to avoid WAL-logging FSM and VM page images?
 
 
-4. Eliminate reporting of some warnings related with hint bits, for example
-"page is not marked all-visible but visibility map bit is set in relation".
-- Why?
-  Hint bit may be not WAL logged.
+## Mark index builds that use buffer manager without logging explicitly
 
-- Alternative?
-  Always wal log any page changes.
+```
+ src/backend/access/gin/gininsert.c                          |    7 +
+ src/backend/access/gist/gistbuild.c                         |   15 +-
+ src/backend/access/spgist/spginsert.c                       |    8 +-
+
+also some changes in src/backend/storage/smgr/smgr.c
+```
+
+When a GIN index is built, for example, it is built by inserting the entries into the index more or
+less normally, but without WAL-logging anything. After the index has been built, we iterate through
+all pages and write them to the WAL. That doesn't work for Neon, because if a page is not WAL-logged
+and is evicted from the buffer cache, it is lost. We have an check to catch that in the Neon
+extension. To fix that, we've added a few functions to track explicitly when we're performing such
+an operation: `smgr_start_unlogged_build`, `smgr_finish_unlogged_build_phase_1` and
+`smgr_end_unlogged_build`.
 
 
-5. Maintain last written LSN.
-- Why?
-  When compute node requests page from page server, we need to specify LSN. Ideally it should be LSN
-  of WAL record performing last update of this pages. But we do not know it, because we do not have page.
-  We can use current WAL flush position, but in this case there is high probability that page server
-  will be blocked until this peace of WAL is delivered.
-  As better approximation we can keep max LSN of written page. It will be better to take in account LSNs only of evicted pages,
-  but SMGR API doesn't provide such knowledge.
+### How to get rid of the patch
 
-- Alternatives?
-  Maintain map of LSNs of evicted pages.
+I think it would make sense to be more explicit about that in PostgreSQL too. So extract these
+changes to a patch and post to pgsql-hackers.
 
 
-6. Launching Postgres without WAL.
-- Why?
-  According to Zenith architecture compute node is stateless. So when we are launching
-  compute node, we need to provide some dummy PG_DATADIR. Relation pages
-  can be requested on demand from page server. But Postgres still need some non-relational data:
-  control and configuration files, SLRUs,...
-  It is currently implemented  using basebackup (do not mix with pg_basebackup) which is created
-  by pageserver. It includes in this tarball config/control files, SLRUs and required directories.
-  As far as pageserver do not have original (non-scattered) WAL segments, it includes in
-  this tarball dummy WAL segment which contains only SHUTDOWN_CHECKPOINT record at the beginning of segment,
-  which redo field points to the end of wal. It allows to load checkpoint record in more or less
-  standard way with minimal changes of Postgres, but then some special handling is needed,
-  including restoring previous record position from zenith.signal file.
-  Also we have to correctly initialize header of last WAL page (pointed by checkpoint.redo)
-  to pass checks performed by XLogReader.
+## Track last-written page LSN
 
-- Alternatives?
-  We may not include fake WAL segment in tarball at all and modify xlog.c to load checkpoint record
-  in special way. But it may only increase number of changes in xlog.c
+```
+ src/backend/commands/dbcommands.c                           |   17 +-
 
-7. Add redo_read_buffer_filter callback to XLogReadBufferForRedoExtended
-- Why?
-  We need a way in wal-redo Postgres to ignore pages which are not requested by pageserver.
-  So wal-redo Postgres reconstructs only requested page and for all other returns BLK_DONE
-  which means that recovery for them is not needed.
+Also one call to SetLastWrittenPageLSN() in spginsert.c, maybe elsewhere too
+```
 
-- Alternatives?
-  No
+Whenever a page is evicted from the buffer cache, we remember its LSN, so that we can use the same
+LSN in the GetPage@LSN request when reading the page back from the page server. The value is
+conservative: it would be correct to always use the last-inserted LSN, but it would be slow because
+then the page server would need to wait for the recent WAL to be streamed and processed, before
+responding to any GetPage@LSN request.
 
-8. Enforce WAL logging of sequence updates.
-- Why?
-  Due to performance reasons Postgres don't want to log each fetching of a value from a sequence,
-  so we pre-log a few fetches in advance. In the event of crash we can lose
-  (skip over) as many values as we pre-logged.
-  But it doesn't work with Zenith because page with sequence value can be evicted from buffer cache
-  and we will get a gap in sequence values even without crash.
+The last-written page LSN is mostly tracked in the smgrwrite() function, without core code changes,
+but there are a few exceptions where we've had to add explicit calls to the Neon-specific
+SetLastWrittenPageLSN() function.
 
-- Alternatives:
-  Do not try to preserve sequential order but avoid performance penalty.
+There's an open PR to track the LSN in a more-fine grained fashion:
+https://github.com/neondatabase/postgres/pull/177
+
+PostgreSQL v15 introduces a new method to do CREATE DATABASE that WAL-logs the database instead of
+relying copying files and checkpoint. With that method, we probably won't need any special handling.
+The old method is still available, though.
+
+### How to get rid of the patch
+
+Wait until v15?
 
 
-9. Treat unlogged tables as normal (permanent) tables.
-- Why?
-  Unlogged tables are not transient, so them have to survive node restart (unlike temporary tables).
-  But as far as compute node is stateless, we need to persist their data to storage node.
-  And it can only be done through the WAL.
+## Cache relation sizes
 
-- Alternatives?
-  * Store unlogged tables locally (violates requirement of stateless compute nodes).
-  * Prohibit unlogged tables at all.
+The Neon extension contains a little cache for smgrnblocks() and smgrexists() calls, to avoid going
+to the page server every time. It might be useful to cache those in PostgreSQL, maybe in the
+relcache? (I think we do cache nblocks in relcache already, check why that's not good enough for
+Neon)
 
 
-10. Support start Postgres in wal-redo mode
-- Why?
-  To be able to apply WAL record and reconstruct pages at page server.
+## Misc change in vacuumlazy.c
 
-- Alternatives?
-  * Rewrite redo handlers in Rust
-  * Do not reconstruct pages at page server at all and do it at compute node.
+```
+index 8aab6e324e..c684c4fbee 100644
+--- a/src/backend/access/heap/vacuumlazy.c
++++ b/src/backend/access/heap/vacuumlazy.c
+@@ -1487,7 +1487,10 @@ lazy_scan_heap(LVRelState *vacrel, VacuumParams *params, bool aggressive)
+                else if (all_visible_according_to_vm && !PageIsAllVisible(page)
+                                 && VM_ALL_VISIBLE(vacrel->rel, blkno, &vmbuffer))
+                {
+-                       elog(WARNING, "page is not marked all-visible but visibility map bit is set in relation \"%s\" page %u",
++                       /* ZENITH-XXX: all visible hint is not wal-logged
++                        * FIXME: Replay visibilitymap changes in pageserver
++                        */
++                       elog(DEBUG1, "page is not marked all-visible but visibility map bit is set in relation \"%s\" page %u",
+                                 vacrel->relname, blkno);
+                        visibilitymap_clear(vacrel->rel, blkno, vmbuffer,
+                                                                VISIBILITYMAP_VALID_BITS);
+```
 
 
-11. WAL proposer
-- Why?
-  WAL proposer is communicating with safekeeper and ensures WAL durability by quorum writes.
-  It is currently implemented as patch to standard WAL sender.
-
-- Alternatives?
-  Can be moved to extension if some extra callbacks will be added to wal sender code.
+Is this still needed? If that WARNING happens, it looks like potential corruption that we should
+fix!
 
 
-12. Secure Computing BPF API wrapper.
-- Why?
-  Pageserver delegates complex WAL decoding duties to Postgres,
-  which means that the latter might fall victim to carefully designed
-  malicious WAL records and start doing harmful things to the system.
-  To prevent this, it has been decided to limit possible interactions
-  with the outside world using the Secure Computing BPF mode.
+## Use buffer manager when extending VM or FSM
 
-- Alternatives:
-  * Rewrite redo handlers in Rust.
-  * Add more checks to guarantee correctness of WAL records.
-  * Move seccomp.c to extension
-  * Many other discussed approaches to neutralize incorrect WAL records vulnerabilities.
+```
+ src/backend/storage/freespace/freespace.c                   |   14 +-
+ src/backend/access/heap/visibilitymap.c                     |   15 +-
+
+diff --git a/src/backend/access/heap/visibilitymap.c b/src/backend/access/heap/visibilitymap.c
+index e198df65d8..addfe93eac 100644
+--- a/src/backend/access/heap/visibilitymap.c
++++ b/src/backend/access/heap/visibilitymap.c
+@@ -652,10 +652,19 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
+        /* Now extend the file */
+        while (vm_nblocks_now < vm_nblocks)
+        {
+-               PageSetChecksumInplace((Page) pg.data, vm_nblocks_now);
++               /*
++                * ZENITH: Initialize VM pages through buffer cache to prevent loading
++                * them from pageserver.
++                */
++               Buffer  buffer = ReadBufferExtended(rel, VISIBILITYMAP_FORKNUM, P_NEW,
++                                                                                       RBM_ZERO_AND_LOCK, NULL);
++               Page    page = BufferGetPage(buffer);
++
++               PageInit((Page) page, BLCKSZ, 0);
++               PageSetChecksumInplace(page, vm_nblocks_now);
++               MarkBufferDirty(buffer);
++               UnlockReleaseBuffer(buffer);
+ 
+-               smgrextend(rel->rd_smgr, VISIBILITYMAP_FORKNUM, vm_nblocks_now,
+-                                  pg.data, false);
+                vm_nblocks_now++;
+        }
+```
+
+### Problem we're trying to solve
+
+???
+
+### How to get rid of the patch
+
+Maybe this would be a reasonable change in PostgreSQL too?
 
 
-13. Callbacks for replica feedbacks
-- Why?
-  Allowing waproposer to interact with walsender code.
+## Allow startup without reading checkpoint record
 
-- Alternatives
-  Copy walsender code to walproposer.
+In Neon, the compute node is stateless. So when we are launching compute node, we need to provide
+some dummy PG_DATADIR. Relation pages can be requested on demand from page server. But Postgres
+still need some non-relational data: control and configuration files, SLRUs,...  It is currently
+implemented using basebackup (do not mix with pg_basebackup) which is created by pageserver. It
+includes in this tarball config/control files, SLRUs and required directories.
+
+As pageserver does not have the original WAL segments, the basebackup tarball includes an empty WAL
+segment to bootstrap the WAL writing, but it doesn't contain the checkpoint record.  There are some
+changes in xlog.c, to allow starting the compute node without reading the last checkpoint record
+from WAL.
+
+This includes code to read the `zenith.signal` file, which tells the startup code the LSN to start
+at. When the `zenith.signal` file is present, the startup uses that LSN instead of the last
+checkpoint's LSN. The system is known to be consistent at that LSN, without any WAL redo.
 
 
-14. Support multiple SMGR implementations.
-- Why?
-  Postgres provides abstract API for storage manager but it has only one implementation
-  and provides no way to replace it with custom storage manager.
+### How to get rid of the patch
 
-- Alternatives?
-  None.
+???
 
 
-15. Calculate database size as sum of all database relations.
-- Why?
-  Postgres is calculating database size by traversing data directory
-  but as far as Zenith compute node is stateless we can not do it.
+### Alternatives
 
-- Alternatives?
-  Send this request directly to pageserver and calculate real (physical) size
-  of Zenith representation of database/timeline, rather than sum logical size of all relations.
+Include a fake checkpoint record in the tarball. Creating fake WAL is a bit risky, though; I'm
+afraid it might accidentally get streamed to the safekeepers and overwrite or corrupt the real WAL.
+
+## Disable sequence caching
+
+```
+diff --git a/src/backend/commands/sequence.c b/src/backend/commands/sequence.c
+index 0415df9ccb..9f9db3c8bc 100644
+--- a/src/backend/commands/sequence.c
++++ b/src/backend/commands/sequence.c
+@@ -53,7 +53,9 @@
+  * so we pre-log a few fetches in advance. In the event of
+  * crash we can lose (skip over) as many values as we pre-logged.
+  */
+-#define SEQ_LOG_VALS   32
++/* Zenith XXX: to ensure sequence order of sequence in Zenith we need to WAL log each sequence update. */
++/* #define SEQ_LOG_VALS        32 */
++#define SEQ_LOG_VALS   0
+```
+
+Due to performance reasons Postgres don't want to log each fetching of a value from a sequence, so
+it pre-logs a few fetches in advance. In the event of crash we can lose (skip over) as many values
+as we pre-logged. But with Neon, because page with sequence value can be evicted from buffer cache,
+we can get a gap in sequence values even without crash.
+
+### How to get rid of the patch
+
+Maybe we can just remove it, and accept the gaps. Or add some special handling for sequence
+relations in the Neon extension, to WAL log the sequence page when it's about to be evicted. It
+would be weird if the sequence moved backwards though, think of PITR.
+
+Or add a GUC for the amount to prefix to PostgreSQL, and force it to 1 in Neon.
 
 
------------------------------------------------
-Not currently committed but proposed:
+## Walproposer
 
-1. Disable ring buffer buffer manager strategies
-- Why?
-  Postgres tries to avoid cache flushing by bulk operations (copy, seqscan, vacuum,...).
-  Even if there are free space in buffer cache, pages may be evicted.
-  Negative effect of it can be somehow compensated by file system cache, but in case of Zenith
-  cost of requesting page from page server is much higher.
+```
+ src/Makefile                                                |    1 +
+ src/backend/replication/libpqwalproposer/Makefile           |   37 +
+ src/backend/replication/libpqwalproposer/libpqwalproposer.c |  416 ++++++++++++
+ src/backend/postmaster/bgworker.c                           |    4 +
+ src/backend/postmaster/postmaster.c                         |    6 +
+ src/backend/replication/Makefile                            |    4 +-
+ src/backend/replication/walproposer.c                       | 2350 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+ src/backend/replication/walproposer_utils.c                 |  402 +++++++++++
+ src/backend/replication/walreceiver.c                       |    7 +
+ src/backend/replication/walsender.c                         |  320 ++++++---
+ src/backend/storage/ipc/ipci.c                              |    6 +
+ src/include/replication/walproposer.h                       |  565 ++++++++++++++++
+```
 
-- Alternatives?
-  Instead of just prohibiting ring buffer we may try to implement more flexible eviction policy,
-  for example copy evicted page from ring buffer to some other buffer if there is free space
-  in buffer cache.
+WAL proposer is communicating with safekeeper and ensures WAL durability by quorum writes.  It is
+currently implemented as patch to standard WAL sender.
 
-2. Disable marking page as dirty when hint bits are set.
-- Why?
-  Postgres has to modify page twice: first time when some tuple is updated and second time when
-  hint bits are set. Wal logging hint bits updates requires FPI which significantly increase size of WAL.
+### How to get rid of the patch
 
-- Alternatives?
-  Add special WAL record for setting page hints.
+Refactor into an extension. Submit hooks or APIs into upstream if necessary.
 
-3. Prefetching
-- Why?
-  As far as pages in Zenith are loaded on demand, to reduce node startup time
-  and also speedup some massive queries we need some mechanism for bulk loading to
-  reduce page request round-trip overhead.
+@MMeent did some work on this already: https://github.com/neondatabase/postgres/pull/96
 
-  Currently Postgres is supporting prefetching only for bitmap scan.
-  In Zenith we also use prefetch for sequential and index scan. For sequential scan we prefetch
-  some number of following pages. For index scan we prefetch pages of heap relation addressed by TIDs.
+## Ignore unexpected data beyond EOF in bufmgr.c
 
-4. Prewarming.
-- Why?
-  Short downtime (or, in other words, fast compute node restart time) is one of the key feature of Zenith.
-  But overhead of request-response round-trip for loading pages on demand can make started node warm-up quite slow.
-  We can capture state of compute node buffer cache and send bulk request for this pages at startup.
+```
+@@ -922,11 +928,14 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
+                 */
+                bufBlock = isLocalBuf ? LocalBufHdrGetBlock(bufHdr) : BufHdrGetBlock(bufHdr);
+                if (!PageIsNew((Page) bufBlock))
+-                       ereport(ERROR,
++               {
++                        // XXX-ZENITH
++                        MemSet((char *) bufBlock, 0, BLCKSZ);
++                        ereport(DEBUG1,
+                                        (errmsg("unexpected data beyond EOF in block %u of relation %s",
+                                                        blockNum, relpath(smgr->smgr_rnode, forkNum)),
+                                         errhint("This has been seen to occur with buggy kernels; consider updating your system.")));
+-
++               }
+                /*
+                 * We *must* do smgrextend before succeeding, else the page will not
+                 * be reserved by the kernel, and the next P_NEW call will decide to
+```
+
+PostgreSQL is a bit sloppy with extending relations. Usually, the relation is extended with zeros
+first, then the page is filled, and finally the new page WAL-logged. But if multiple backends extend
+a relation at the same time, the pages can be WAL-logged in different order.
+
+I'm not sure what scenario exactly required this change in Neon, though.
+
+### How to get rid of the patch
+
+Submit patches to pgsql-hackers, to tighten up the WAL-logging around relation extension. It's a bit
+confusing even in PostgreSQL. Maybe WAL log the intention to extend first, then extend the relation,
+and finally WAL-log that the extension succeeded.
+
+## Make smgr interface available to extensions
+
+```
+ src/backend/storage/smgr/smgr.c                             |  203 +++---
+ src/include/storage/smgr.h                                  |   72 +-
+```
+
+### How to get rid of the patch
+
+Submit to upstream. This could be useful for the Disk Encryption patches too, or for compression.
+
+
+## Added relpersistence argument to smgropen()
+
+```
+ src/backend/access/heap/heapam_handler.c                    |    2 +-
+ src/backend/catalog/storage.c                               |   10 +-
+ src/backend/commands/tablecmds.c                            |    2 +-
+ src/backend/storage/smgr/md.c                               |    4 +-
+ src/include/utils/rel.h                                     |    3 +-
+```
+
+Neon needs to treat unlogged relations differently from others, so the smgrread(), smgrwrite() etc.
+implementations need to know the 'relpersistence' of the relation. To get that information where
+it's needed, we added the 'relpersistence' field to smgropen().
+
+### How to get rid of the patch
+
+Maybe 'relpersistence' would be useful in PostgreSQL for debugging purposes? Or simply for the
+benefit of extensions like Neon. Should consider this in the patch to make smgr API usable to
+extensions.
+
+## Alternatives
+
+Currently in Neon, unlogged tables live on local disk in the compute node, and are wiped away on
+compute node restart. One alternative would be to instead WAL-log even unlogged tables, essentially
+ignoring the UNLOGGED option. Or prohibit UNLOGGED tables completely. But would we still need the
+relpersistence argument to handle index builds? See item on "Mark index builds that use buffer
+manager without logging explicitly".
+
+## Use smgr and dbsize_hook for size calculations
+
+```
+ src/backend/utils/adt/dbsize.c                              |   61 +-
+```
+
+In PostgreSQL, the rel and db-size functions scan the data directory directly. That won't work in Neon.
+
+### How to get rid of the patch
+
+Send patch to PostgreSQL, to use smgr API functions for relation size calculation instead. Maybe as
+part of the general smgr API patch.
+
+
+
+# WAL redo process changes
+
+Pageserver delegates complex WAL decoding duties to Postgres, which means that the latter might fall
+victim to carefully designed malicious WAL records and start doing harmful things to the system.  To
+prevent this, the redo functions are executed in a separate process that is sandboxed with Linux
+Secure Computing mode (see seccomp(2) man page).
+
+As an alternative to having a separate WAL redo process, we could rewrite all redo handlers in Rust
+This is infeasible. However, it would take a lot of effort to rewrite them, ensure that you've done
+the rewrite correctly, and once you've done that, it would be a lot of ongoing maintenance effort to
+keep the rewritten code in sync over time, across new PostgreSQL versions. That's why we want to
+leverage PostgreSQL code.
+
+Another alternative would be to harden all the PostgreSQL WAL redo functions so that it would be
+safe to call them directly from Rust code, without needing the security sandbox. That's not feasible
+for similar reasons as rewriting them in Rust.
+
+
+## Don't replay change in XLogReadBufferForRedo that are not for the target page we're replaying
+
+```
+ src/backend/access/gin/ginxlog.c                            |   19 +-
+
+Also some changes in xlog.c and xlogutils.c
+
+Example:
+
+@@ -415,21 +416,27 @@ ginRedoSplit(XLogReaderState *record)
+        if (!isLeaf)
+                ginRedoClearIncompleteSplit(record, 3);
+ 
+-       if (XLogReadBufferForRedo(record, 0, &lbuffer) != BLK_RESTORED)
++       action = XLogReadBufferForRedo(record, 0, &lbuffer);
++       if (action != BLK_RESTORED && action != BLK_DONE)
+                elog(ERROR, "GIN split record did not contain a full-page image of left page");
+```
+
+### Problem we're trying to solve
+
+In PostgreSQL, if a WAL redo function calls XLogReadBufferForRead() for a page that has a full-page
+image, it always succeeds. However, Neon WAL redo process is only concerned about replaying changes
+to a singe page, so replaying any changes for other pages is a waste of cycles. We have modified
+XLogReadBufferForRead() to return BLK_DONE for all other pages, to avoid the overhead. That is
+unexpected by code like the above.
+
+### How to get rid of the patch
+
+Submit the changes to upstream, hope the community accepts them. There's no harm to PostgreSQL from
+these changes, although it doesn't have any benefit either.
+
+To make these changes useful to upstream PostgreSQL, we could implement a feature to look ahead the
+WAL, and detect truncated relations. Even in PostgreSQL, it is a waste of cycles to replay changes
+to pages that are later truncated away, so we could have XLogReadBufferForRedo() return BLK_DONE or
+BLK_NOTFOUND for pages that are known to be truncated away later in the WAL stream.
+
+### Alternatives
+
+Maybe we could revert this optimization, and restore pages other than the target page too.
+
+## Add predefined_sysidentifier flag to initdb
+
+```
+ src/backend/bootstrap/bootstrap.c                           |   13 +-
+ src/bin/initdb/initdb.c                                     |    4 +
+
+And some changes in xlog.c
+```
+
+This is used to help with restoring a database when you have all the WAL, all the way back to
+initdb, but no backup. You can reconstruct the missing backup by running initdb again, with the same
+sysidentifier.
+
+
+### How to get rid of the patch
+
+Ignore it. This is only needed for disaster recovery, so once we've eliminated all other Postgres
+patches, we can just keep it around as a patch or as separate branch in a repo.
+
+
+# Not currently committed but proposed
+
+## Disable ring buffer buffer manager strategies
+
+### Why?
+
+Postgres tries to avoid cache flushing by bulk operations (copy, seqscan, vacuum,...).
+Even if there are free space in buffer cache, pages may be evicted.
+Negative effect of it can be somehow compensated by file system cache, but in Neon,
+cost of requesting page from page server is much higher.
+
+### Alternatives?
+
+Instead of just prohibiting ring buffer we may try to implement more flexible eviction policy,
+for example copy evicted page from ring buffer to some other buffer if there is free space
+in buffer cache.
+
+## Disable marking page as dirty when hint bits are set.
+
+### Why?
+
+Postgres has to modify page twice: first time when some tuple is updated and second time when
+hint bits are set. Wal logging hint bits updates requires FPI which significantly increase size of WAL.
+
+### Alternatives?
+
+Add special WAL record for setting page hints.
+
+## Prefetching
+
+### Why?
+
+As far as pages in Neon are loaded on demand, to reduce node startup time
+and also speedup some massive queries we need some mechanism for bulk loading to
+reduce page request round-trip overhead.
+
+Currently Postgres is supporting prefetching only for bitmap scan.
+In Neon we should also use prefetch for sequential and index scans, because the OS is not doing it for us.
+For sequential scan we could prefetch some number of following pages. For index scan we could prefetch pages
+of heap relation addressed by TIDs.
+
+## Prewarming
+
+### Why?
+
+Short downtime (or, in other words, fast compute node restart time) is one of the key feature of Zenith.
+But overhead of request-response round-trip for loading pages on demand can make started node warm-up quite slow.
+We can capture state of compute node buffer cache and send bulk request for this pages at startup.

docs/glossary.md

@@ -75,7 +75,7 @@ layer's Segment and range of LSNs.
 There are two kinds of layers, in-memory and on-disk layers. In-memory
 layers are used to ingest incoming WAL, and provide fast access
 to the recent page versions. On-disk layers are stored as files on disk, and
-are immutable. See pageserver/src/layered_repository/README.md for more.
+are immutable. See [pageserver-storage.md](./pageserver-storage.md) for more.
 
 ### Layer file (on-disk layer)
 
@@ -111,7 +111,7 @@ PostgreSQL LSNs and functions to monitor them:
 * `pg_last_wal_replay_lsn ()` - Returns the last write-ahead log location that has been replayed during recovery. If recovery is still in progress this will increase monotonically.
 [source PostgreSQL documentation](https://www.postgresql.org/docs/devel/functions-admin.html):
 
-Neon safekeeper LSNs. For more check [safekeeper/README_PROTO.md](/safekeeper/README_PROTO.md)
+Neon safekeeper LSNs. See [safekeeper protocol section](safekeeper-protocol.md) for more information.
 * `CommitLSN`: position in WAL confirmed by quorum safekeepers.
 * `RestartLSN`: position in WAL confirmed by all safekeepers.
 * `FlushLSN`: part of WAL persisted to the disk by safekeeper.

docs/pageserver-page-service.md

@@ -0,0 +1,9 @@
+# Page Service
+
+The Page Service listens for GetPage@LSN requests from the Compute Nodes,
+and responds with pages from the repository. On each GetPage@LSN request,
+it calls into the Repository function
+
+A separate thread is spawned for each incoming connection to the page
+service. The page service uses the libpq protocol to communicate with
+the client. The client is a Compute Postgres instance.

docs/pageserver-pagecache.md

@@ -0,0 +1,8 @@
+# Page cache
+
+TODO:
+
+- shared across tenants
+- store pages from layer files
+- store pages from "in-memory layer"
+- store materialized pages

docs/pageserver-processing-getpage.md

@@ -0,0 +1,4 @@
+# Processing a GetPage request
+
+TODO:
+- sequence diagram that shows how a GetPage@LSN request is processed

docs/pageserver-processing-wal.md

@@ -0,0 +1,5 @@
+# Processing WAL
+
+TODO:
+- diagram that shows how incoming WAL is processed
+- explain durability, what is fsync'd when, disk_consistent_lsn

docs/pageserver-services.md

@@ -1,15 +1,4 @@
-## Page server architecture
-
-The Page Server has a few different duties:
-
-- Respond to GetPage@LSN requests from the Compute Nodes
-- Receive WAL from WAL safekeeper
-- Replay WAL that's applicable to the chunks that the Page Server maintains
-- Backup to S3
-
-S3 is the main fault-tolerant storage of all data, as there are no Page Server
-replicas. We use a separate fault-tolerant WAL service to reduce latency. It
-keeps track of WAL records which are not synced to S3 yet.
+# Services
 
 The Page Server consists of multiple threads that operate on a shared
 repository of page versions:
@@ -21,18 +10,22 @@ repository of page versions:
                                    | WAL receiver |
                                    |              |
                                    +--------------+
-                                                                                 +----+
-                  +---------+                              ..........            |    |
-                  |         |                              .        .            |    |
- GetPage@LSN      |         |                              . backup .  ------->  | S3 |
-------------->    |  Page   |         repository           .        .            |    |
-                  | Service |                              ..........            |    |
-   page           |         |                                                    +----+
+                                                                                 ......
+                  +---------+                              +--------+            .    .
+                  |         |                              |        |            .    .
+ GetPage@LSN      |         |                              | backup |  ------->  . S3 .
+------------->    |  Page   |         repository           |        |            .    .
+                  | Service |                              +--------+            .    .
+   page           |         |                                                    ......
 <-------------    |         |
-                  +---------+      +--------------------+
-		                   |   Checkpointing /  |
-				   | Garbage collection |
-                                   +--------------------+
+                  +---------+     +-----------+     +--------------------+
+                                  | WAL redo  |     | Checkpointing,     |
+                  +----------+    | processes |     | Garbage collection |
+                  |          |    +-----------+     +--------------------+
+                  |   HTTP   |
+                  | mgmt API |
+                  |          |
+                  +----------+
 
 Legend:
 
@@ -40,28 +33,75 @@ Legend:
 |  |   A thread or multi-threaded service
 +--+
 
-....
-.  .   Component at its early development phase.
-....
-
 --->   Data flow
 <---
 ```
 
-Page Service
-------------
+## Page Service
 
 The Page Service listens for GetPage@LSN requests from the Compute Nodes,
-and responds with pages from the repository.
+and responds with pages from the repository. On each GetPage@LSN request,
+it calls into the Repository function
+
+A separate thread is spawned for each incoming connection to the page
+service. The page service uses the libpq protocol to communicate with
+the client. The client is a Compute Postgres instance.
+
+## WAL Receiver
+
+The WAL receiver connects to the external WAL safekeeping service
+using PostgreSQL physical streaming replication, and continuously
+receives WAL. It decodes the WAL records, and stores them to the
+repository.
 
 
-WAL Receiver
-------------
+## Backup service
 
-The WAL receiver connects to the external WAL safekeeping service (or
-directly to the primary) using PostgreSQL physical streaming
-replication, and continuously receives WAL. It decodes the WAL records,
-and stores them to the repository.
+The backup service, responsible for storing pageserver recovery data externally.
+
+Currently, pageserver stores its files in a filesystem directory it's pointed to.
+That working directory could be rather ephemeral for such cases as "a pageserver pod running in k8s with no persistent volumes attached".
+Therefore, the server interacts with external, more reliable storage to back up and restore its state.
+
+The code for storage support is extensible and can support arbitrary ones as long as they implement a certain Rust trait.
+There are the following implementations present:
+* local filesystem — to use in tests mainly
+* AWS S3           - to use in production
+
+The backup service is disabled by default and can be enabled to interact with a single remote storage.
+
+CLI examples:
+* Local FS: `${PAGESERVER_BIN} -c "remote_storage={local_path='/some/local/path/'}"`
+* AWS S3  : `env AWS_ACCESS_KEY_ID='SOMEKEYAAAAASADSAH*#' AWS_SECRET_ACCESS_KEY='SOMEsEcReTsd292v' ${PAGESERVER_BIN} -c "remote_storage={bucket_name='some-sample-bucket',bucket_region='eu-north-1', prefix_in_bucket='/test_prefix/'}"`
+
+For Amazon AWS S3, a key id and secret access key could be located in `~/.aws/credentials` if awscli was ever configured to work with the desired bucket, on the AWS Settings page for a certain user. Also note, that the bucket names does not contain any protocols when used on AWS.
+For local S3 installations, refer to the their documentation for name format and credentials.
+
+Similar to other pageserver settings, toml config file can be used to configure either of the storages as backup targets.
+Required sections are:
+
+```toml
+[remote_storage]
+local_path = '/Users/someonetoignore/Downloads/tmp_dir/'
+```
+
+or
+
+```toml
+[remote_storage]
+bucket_name = 'some-sample-bucket'
+bucket_region = 'eu-north-1'
+prefix_in_bucket = '/test_prefix/'
+```
+
+`AWS_SECRET_ACCESS_KEY` and `AWS_ACCESS_KEY_ID` env variables can be used to specify the S3 credentials if needed.
+
+
+## Repository background tasks
+
+The Repository also has a few different background threads and tokio tasks that perform
+background duties like dumping accumulated WAL data from memory to disk, reorganizing
+files for performance (compaction), and garbage collecting old files.
 
 
 Repository
@@ -76,7 +116,7 @@ implemented by the LayeredRepository object in
 `layered_repository.rs`. There is only that one implementation of the
 Repository trait, but it's still a useful abstraction that keeps the
 interface for the low-level storage functionality clean. The layered
-storage format is described in layered_repository/README.md.
+storage format is described in [pageserver-storage.md](./pageserver-storage.md).
 
 Each repository consists of multiple Timelines. Timeline is a
 workhorse that accepts page changes from the WAL, and serves
@@ -116,48 +156,6 @@ Remove old on-disk layer files that are no longer needed according to the
 PITR retention policy
 
 
-### Backup service
-
-The backup service, responsible for storing pageserver recovery data externally.
-
-Currently, pageserver stores its files in a filesystem directory it's pointed to.
-That working directory could be rather ephemeral for such cases as "a pageserver pod running in k8s with no persistent volumes attached".
-Therefore, the server interacts with external, more reliable storage to back up and restore its state.
-
-The code for storage support is extensible and can support arbitrary ones as long as they implement a certain Rust trait.
-There are the following implementations present:
-* local filesystem — to use in tests mainly
-* AWS S3           - to use in production
-
-Implementation details are covered in the [backup readme](./src/remote_storage/README.md) and corresponding Rust file docs, parameters documentation can be found at [settings docs](../docs/settings.md).
-
-The backup service is disabled by default and can be enabled to interact with a single remote storage.
-
-CLI examples:
-* Local FS: `${PAGESERVER_BIN} -c "remote_storage={local_path='/some/local/path/'}"`
-* AWS S3  : `env AWS_ACCESS_KEY_ID='SOMEKEYAAAAASADSAH*#' AWS_SECRET_ACCESS_KEY='SOMEsEcReTsd292v' ${PAGESERVER_BIN} -c "remote_storage={bucket_name='some-sample-bucket',bucket_region='eu-north-1', prefix_in_bucket='/test_prefix/'}"`
-
-For Amazon AWS S3, a key id and secret access key could be located in `~/.aws/credentials` if awscli was ever configured to work with the desired bucket, on the AWS Settings page for a certain user. Also note, that the bucket names does not contain any protocols when used on AWS.
-For local S3 installations, refer to the their documentation for name format and credentials.
-
-Similar to other pageserver settings, toml config file can be used to configure either of the storages as backup targets.
-Required sections are:
-
-```toml
-[remote_storage]
-local_path = '/Users/someonetoignore/Downloads/tmp_dir/'
-```
-
-or
-
-```toml
-[remote_storage]
-bucket_name = 'some-sample-bucket'
-bucket_region = 'eu-north-1'
-prefix_in_bucket = '/test_prefix/'
-```
-
-`AWS_SECRET_ACCESS_KEY` and `AWS_ACCESS_KEY_ID` env variables can be used to specify the S3 credentials if needed.
 
 TODO: Sharding
 --------------------

docs/pageserver-storage.md

@@ -1,4 +1,4 @@
-# Overview
+# Pageserver storage
 
 The main responsibility of the Page Server is to process the incoming WAL, and
 reprocess it into a format that allows reasonably quick access to any page

docs/pageserver-thread-mgmt.md

@@ -0,0 +1,26 @@
+## Thread management
+
+Each thread in the system is tracked by the `thread_mgr` module. It
+maintains a registry of threads, and which tenant or timeline they are
+operating on. This is used for safe shutdown of a tenant, or the whole
+system.
+
+### Handling shutdown
+
+When a tenant or timeline is deleted, we need to shut down all threads
+operating on it, before deleting the data on disk. A thread registered
+in the thread registry can check if it has been requested to shut down,
+by calling `is_shutdown_requested()`. For async operations, there's also
+a `shudown_watcher()` async task that can be used to wake up on shutdown.
+
+### Sync vs async
+
+The primary programming model in the page server is synchronous,
+blocking code. However, there are some places where async code is
+used. Be very careful when mixing sync and async code.
+
+Async is primarily used to wait for incoming data on network
+connections. For example, all WAL receivers have a shared thread pool,
+with one async Task for each connection. Once a piece of WAL has been
+received from the network, the thread calls the blocking functions in
+the Repository to process the WAL.

docs/pageserver-walredo.md

@@ -0,0 +1,77 @@
+# WAL Redo
+
+To reconstruct a particular page version from an image of the page and
+some WAL records, the pageserver needs to replay the WAL records. This
+happens on-demand, when a GetPage@LSN request comes in, or as part of
+background jobs that reorganize data for faster access.
+
+It's important that data cannot leak from one tenant to another, and
+that a corrupt WAL record on one timeline doesn't affect other tenants
+or timelines.
+
+## Multi-tenant security
+
+If you have direct access to the WAL directory, or if you have
+superuser access to a running PostgreSQL server, it's easy to
+construct a malicious or corrupt WAL record that causes the WAL redo
+functions to crash, or to execute arbitrary code. That is not a
+security problem for PostgreSQL; if you have superuser access, you
+have full access to the system anyway.
+
+The Neon pageserver, however, is multi-tenant. It needs to execute WAL
+belonging to different tenants in the same system, and malicious WAL
+in one tenant must not affect other tenants.
+
+A separate WAL redo process is launched for each tenant, and the
+process uses the seccomp(2) system call to restrict its access to the
+bare minimum needed to replay WAL records. The process does not have
+access to the filesystem or network. It can only communicate with the
+parent pageserver process through a pipe.
+
+If an attacker creates a malicious WAL record and injects it into the
+WAL stream of a timeline, he can take control of the WAL redo process
+in the pageserver. However, the WAL redo process cannot access the
+rest of the system. And because there is a separate WAL redo process
+for each tenant, the hijacked WAL redo process can only see WAL and
+data belonging to the same tenant, which the attacker would have
+access to anyway.
+
+## WAL-redo process communication
+
+The WAL redo process runs the 'postgres' executable, launched with a
+Neon-specific command-line option to put it into WAL-redo process
+mode.  The pageserver controls the lifetime of the WAL redo processes,
+launching them as needed. If a tenant is detached from the pageserver,
+any WAL redo processes for that tenant are killed.
+
+The pageserver communicates with each WAL redo process over its
+stdin/stdout/stderr. It works in request-response model with a simple
+custom protocol, described in walredo.rs. To replay a set of WAL
+records for a page, the pageserver sends the "before" image of the
+page and the WAL records over 'stdin', followed by a command to
+perform the replay. The WAL redo process responds with an "after"
+image of the page.
+
+## Special handling of some records
+
+Some WAL record types are handled directly in the pageserver, by
+bespoken Rust code, and are not sent over to the WAL redo process.
+This includes SLRU-related WAL records, like commit records. SLRUs
+don't use the standard Postgres buffer manager, so dealing with them
+in the Neon WAL redo mode would require quite a few changes to
+Postgres code and special handling in the protocol anyway.
+
+Some record types that include a full-page-image (e.g. XLOG_FPI) are
+also handled specially when incoming WAL is processed already, and are
+stored as page images rather than WAL records.
+
+
+## Records that modify multiple pages
+
+Some Postgres WAL records modify multiple pages. Such WAL records are
+duplicated, so that a copy is stored for each affected page. This is
+somewhat wasteful, but because most WAL records only affect one page,
+the overhead is acceptable.
+
+The WAL redo always happens for one particular page. If the WAL record
+coantains changes to other pages, they are ignored.

docs/pageserver.md

@@ -0,0 +1,11 @@
+# Page server architecture
+
+The Page Server has a few different duties:
+
+- Respond to GetPage@LSN requests from the Compute Nodes
+- Receive WAL from WAL safekeeper, and store it
+- Upload data to S3 to make it durable, download files from S3 as needed
+
+S3 is the main fault-tolerant storage of all data, as there are no Page Server
+replicas. We use a separate fault-tolerant WAL service to reduce latency. It
+keeps track of WAL records which are not synced to S3 yet.

docs/rfcs/017-user-management.md

@@ -1,80 +0,0 @@
-# Postgres user and database management
-
-We've accumulated a bunch of problems with our approach to role and database management, namely:
-
-1. we don't allow role and database creation from Postgres, and users are complaining about that
-2. fine-grained role management is not possible both from Postgres and console
-3. web_access and @user are different roles, which creates object access problems in some cases
-
-Right now, we do store users and databases both in console and Postgres, and there are two main reasons for
-that:
-
-* we want to be able to authenticate users in proxy against the console without Postgres involvement. Otherwise,
-malicious brute force attempts will wake up Postgres (expensive) and may exhaust the Postgres connection pool (deny of service).
-* it is handy when we can render console UI without waking up compute (e.g., show database list)
-
-Storing the same information in two systems is a form of replication. And in the current scheme
-the console is primary, and Postgres catalog is a replica.
-
-This RFC proposes to address problems 1. and 2. by making Postgres a source of truth for roles/databases and
-only caching this info in the console. So using the replication analogy, now the Postgres catalog will be primary, and
-the console will be a replica. Problem 3 is a bit different and could be addressed by ditching the web_access
-user and using, e.g., JWT auth for the @username user so that we do not introduce a new user (JWT is needed
-since we don't know users password).
-
-This RFC doesn't talk about giving root access to the database, which is blocked by a secure runtime setup.
-
-## Overview
-
-* Add `/tenant/$tenant/branch/$branch/refresh_catalog` endpoint to console management API which asks `/get_catalog` and updates cached roles/databases info.
-* Whenever user edits list of databases or users postgres signals `compute_ctl` to call `/<...>/refresh_catalog` in the console
-* Add password strenght check in our extension
-
-## Postgres behavior
-
-Default user role (@username) should have `CREATE ROLE`, `CREATE DB`, and `BYPASSRLS` privileges. We expose Postgres port
-to the open internet, so we need to check passwords strength. We can use the `passwordcheck` extension or do the same
-from our extension.
-
-Whenever a user edits a list of databases or users, Postgres sends SIGHUP to `compute_ctl`. `compute_ctl` should write PID to `compute_ctl.pid` file.
-
-
-## Compute_ctl behavior
-
-Upon `SIGHUP` signal `compute_ctl` should call `/tenant/$tenant/branch/$branch/refresh_catalog` to inform console about changes in the database. The console will circle back and load the data from `/get_catalog` on compute (see next section on why this approach instead of direct PUT/PATH to the console). In the case of `/refresh_catalog` failure, we should retry it N times.
-
-Also `compute_ctl` listens for http `/get_catalog` and returns list of databases and users upon request:
-```
-/get_catalog: -> {
-    databases: [{
-        name: "db1",
-        owner: "jack"
-    }],
-    roles: [{
-        name: "jack",
-        rolepassword: "SCRAM-SHA-256..."
-    }]
-}
-```
-
-## Console behavior
-
-Whenever the console receives `/refresh_catalog` on the management API it goes to compute and asks for `/get_catalog`. I suggest using this way instead of accepting a list of databases/roles directly to the console endpoint for the following reasons:
-
-* we, anyway, will need console originated call to compute's `/get_catalog` after historical branch creation
-* If an intruder gains access to some other `/tenant/$tenant/.../refresh_catalog` he won't be able to change the roles list and will just force an unnecessary reload.
-
-`/refresh_catalog` returns HTTP 200 OK on success.
-
-We should have a button in the admin UI to manually force `/refresh_catalog` in case of data desync.
-
-# Scalability
-
-On my laptop, I can create 4200 roles per second. That corresponds to 363 million roles per day. So both `/get_catalog` can become expensive, and our roles database can snowball. While we can address `/get_catalog` size by catching only the latest changes (e.g., maintain the audit table and drain it by the console), it is still not nice that a single tenant can blow up a multi-tenant console database. I would instead propose to limit the number of databases and roles by some big number like 1000 and bump this limit if somebody asks for it with a legit use case. 
-
-
-# QA:
-
-- Why implement `/get_catalog` instead of sending an SQL query from the console to the compute?
-
-- So far, we do not allow remote superuser access to Postgres, and exposing only endpoints with fixed queries beneath them reduces the attack surface.

docs/separation-compute-storage.md

@@ -0,0 +1,8 @@
+# Separation of Compute and Storage
+
+TODO:
+
+- Read path
+- Write path
+- Durability model
+- API auth

docs/settings.md

@@ -15,7 +15,7 @@ listen_pg_addr = '127.0.0.1:64000'
 listen_http_addr = '127.0.0.1:9898'
 
 checkpoint_distance = '268435456' # in bytes
-checkpoint_period = '1 s'
+checkpoint_timeout = '10m'
 
 gc_period = '100 s'
 gc_horizon = '67108864'
@@ -46,7 +46,7 @@ Note the `[remote_storage]` section: it's a [table](https://toml.io/en/v1.0.0#ta
 
 All values can be passed as an argument to the pageserver binary, using the `-c` parameter and specified as a valid TOML string. All tables should be passed in the inline form.
 
-Example: `${PAGESERVER_BIN} -c "checkpoint_period = '100 s'" -c "remote_storage={local_path='/some/local/path/'}"`
+Example: `${PAGESERVER_BIN} -c "checkpoint_timeout = '10 m'" -c "remote_storage={local_path='/some/local/path/'}"`
 
 Note that TOML distinguishes between strings and integers, the former require single or double quotes around them.
 
@@ -82,6 +82,14 @@ S3.
 
 The unit is # of bytes.
 
+#### checkpoint_timeout
+
+Apart from `checkpoint_distance`, open layer flushing is also triggered
+`checkpoint_timeout` after the last flush. This makes WAL eventually uploaded to
+s3 when activity is stopped.
+
+The default is 10m.
+
 #### compaction_period
 
 Every `compaction_period` seconds, the page server checks if

docs/sourcetree.md

@@ -28,7 +28,7 @@ The pageserver has a few different duties:
 - Receive WAL from the WAL service and decode it.
 - Replay WAL that's applicable to the chunks that the Page Server maintains
 
-For more detailed info, see [/pageserver/README](/pageserver/README.md)
+For more detailed info, see [pageserver-services.md](./pageserver-services.md)
 
 `/proxy`:
 
@@ -57,7 +57,7 @@ PostgreSQL extension that contains functions needed for testing and debugging.
 The zenith WAL service that receives WAL from a primary compute nodes and streams it to the pageserver.
 It acts as a holding area and redistribution center for recently generated WAL.
 
-For more detailed info, see [/safekeeper/README](/safekeeper/README.md)
+For more detailed info, see [walservice.md](./walservice.md)
 
 `/workspace_hack`:
 The workspace_hack crate exists only to pin down some dependencies.

docs/walservice.md

@@ -75,8 +75,8 @@ safekeepers. The Paxos and crash recovery algorithm ensures that only
 one primary node can be actively streaming WAL to the quorum of
 safekeepers.
 
-See README_PROTO.md for a more detailed description of the consensus
-protocol. spec/ contains TLA+ specification of it.
+See [this section](safekeeper-protocol.md) for a more detailed description of
+the consensus protocol. spec/ contains TLA+ specification of it.
 
 # Q&A
 

libs/etcd_broker/Cargo.toml

@@ -9,7 +9,7 @@
  serde = { version = "1.0", features = ["derive"] }
  serde_json = "1"
  serde_with = "1.12.0"
- once_cell = "1.8.0"
+ once_cell = "1.13.0"
 
  utils = { path = "../utils" }
  workspace_hack = { version = "0.1", path = "../../workspace_hack" }

libs/metrics/Cargo.toml

@@ -6,6 +6,5 @@ edition = "2021"
 [dependencies]
 prometheus = {version = "0.13", default_features=false, features = ["process"]} # removes protobuf dependency
 libc = "0.2"
-lazy_static = "1.4"
-once_cell = "1.8.0"
+once_cell = "1.13.0"
 workspace_hack = { version = "0.1", path = "../../workspace_hack" }

libs/metrics/src/lib.rs

@@ -2,7 +2,10 @@
 //! make sure that we use the same dep version everywhere.
 //! Otherwise, we might not see all metrics registered via
 //! a default registry.
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
+use prometheus::core::{AtomicU64, GenericGauge, GenericGaugeVec};
+pub use prometheus::opts;
+pub use prometheus::register;
 pub use prometheus::{core, default_registry, proto};
 pub use prometheus::{exponential_buckets, linear_buckets};
 pub use prometheus::{register_gauge, Gauge};
@@ -18,6 +21,17 @@ pub use prometheus::{Encoder, TextEncoder};
 mod wrappers;
 pub use wrappers::{CountedReader, CountedWriter};
 
+pub type UIntGauge = GenericGauge<AtomicU64>;
+pub type UIntGaugeVec = GenericGaugeVec<AtomicU64>;
+
+#[macro_export]
+macro_rules! register_uint_gauge_vec {
+    ($NAME:expr, $HELP:expr, $LABELS_NAMES:expr $(,)?) => {{
+        let gauge_vec = UIntGaugeVec::new($crate::opts!($NAME, $HELP), $LABELS_NAMES).unwrap();
+        $crate::register(Box::new(gauge_vec.clone())).map(|_| gauge_vec)
+    }};
+}
+
 /// Gathers all Prometheus metrics and records the I/O stats just before that.
 ///
 /// Metrics gathering is a relatively simple and standalone operation, so
@@ -27,19 +41,22 @@ pub fn gather() -> Vec<prometheus::proto::MetricFamily> {
     prometheus::gather()
 }
 
-lazy_static! {
-    static ref DISK_IO_BYTES: IntGaugeVec = register_int_gauge_vec!(
+static DISK_IO_BYTES: Lazy<IntGaugeVec> = Lazy::new(|| {
+    register_int_gauge_vec!(
         "libmetrics_disk_io_bytes_total",
         "Bytes written and read from disk, grouped by the operation (read|write)",
         &["io_operation"]
     )
-    .expect("Failed to register disk i/o bytes int gauge vec");
-    static ref MAXRSS_KB: IntGauge = register_int_gauge!(
+    .expect("Failed to register disk i/o bytes int gauge vec")
+});
+
+static MAXRSS_KB: Lazy<IntGauge> = Lazy::new(|| {
+    register_int_gauge!(
         "libmetrics_maxrss_kb",
         "Memory usage (Maximum Resident Set Size)"
     )
-    .expect("Failed to register maxrss_kb int gauge");
-}
+    .expect("Failed to register maxrss_kb int gauge")
+});
 
 pub const DISK_WRITE_SECONDS_BUCKETS: &[f64] = &[
     0.000_050, 0.000_100, 0.000_500, 0.001, 0.003, 0.005, 0.01, 0.05, 0.1, 0.3, 0.5,

libs/metrics/src/wrappers.rs

@@ -10,13 +10,13 @@ use std::io::{Read, Result, Write};
 /// # use std::io::{Result, Read};
 /// # use metrics::{register_int_counter, IntCounter};
 /// # use metrics::CountedReader;
+/// # use once_cell::sync::Lazy;
 /// #
-/// # lazy_static::lazy_static! {
-/// #     static ref INT_COUNTER: IntCounter = register_int_counter!(
+/// # static INT_COUNTER: Lazy<IntCounter> = Lazy::new( || { register_int_counter!(
 /// #         "int_counter",
 /// #         "let's count something!"
-/// #     ).unwrap();
-/// # }
+/// #     ).unwrap()
+/// # });
 /// #
 /// fn do_some_reads(stream: impl Read, count: usize) -> Result<Vec<u8>> {
 ///     let mut reader = CountedReader::new(stream, |cnt| {
@@ -85,13 +85,13 @@ impl<T: Read> Read for CountedReader<'_, T> {
 /// # use std::io::{Result, Write};
 /// # use metrics::{register_int_counter, IntCounter};
 /// # use metrics::CountedWriter;
+/// # use once_cell::sync::Lazy;
 /// #
-/// # lazy_static::lazy_static! {
-/// #     static ref INT_COUNTER: IntCounter = register_int_counter!(
+/// # static INT_COUNTER: Lazy<IntCounter> = Lazy::new( || { register_int_counter!(
 /// #         "int_counter",
 /// #         "let's count something!"
-/// #     ).unwrap();
-/// # }
+/// #     ).unwrap()
+/// # });
 /// #
 /// fn do_some_writes(stream: impl Write, payload: &[u8]) -> Result<()> {
 ///     let mut writer = CountedWriter::new(stream, |cnt| {

libs/postgres_ffi/Cargo.toml

@@ -12,7 +12,7 @@ byteorder = "1.4.3"
 anyhow = "1.0"
 crc32c = "0.6.0"
 hex = "0.4.3"
-lazy_static = "1.4"
+once_cell = "1.13.0"
 log = "0.4.14"
 memoffset = "0.6.2"
 thiserror = "1.0"

libs/postgres_ffi/build.rs

@@ -49,12 +49,12 @@ fn main() {
     // Finding the location of C headers for the Postgres server:
     // - if POSTGRES_INSTALL_DIR is set look into it, otherwise look into `<project_root>/tmp_install`
     // - if there's a `bin/pg_config` file use it for getting include server, otherwise use `<project_root>/tmp_install/include/postgresql/server`
-    let mut pg_install_dir: PathBuf;
-    if let Some(postgres_install_dir) = env::var_os("POSTGRES_INSTALL_DIR") {
-        pg_install_dir = postgres_install_dir.into();
+    let mut pg_install_dir = if let Some(postgres_install_dir) = env::var_os("POSTGRES_INSTALL_DIR")
+    {
+        postgres_install_dir.into()
     } else {
-        pg_install_dir = PathBuf::from("tmp_install")
-    }
+        PathBuf::from("tmp_install")
+    };
 
     if pg_install_dir.is_relative() {
         let cwd = env::current_dir().unwrap();

libs/postgres_ffi/src/relfile_utils.rs

@@ -2,7 +2,7 @@
 //! Common utilities for dealing with PostgreSQL relation files.
 //!
 use crate::pg_constants;
-use lazy_static::lazy_static;
+use once_cell::sync::OnceCell;
 use regex::Regex;
 
 #[derive(Debug, Clone, thiserror::Error, PartialEq)]
@@ -54,11 +54,14 @@ pub fn forknumber_to_name(forknum: u8) -> Option<&'static str> {
 /// See functions relpath() and _mdfd_segpath() in PostgreSQL sources.
 ///
 pub fn parse_relfilename(fname: &str) -> Result<(u32, u8, u32), FilePathError> {
-    lazy_static! {
-        static ref RELFILE_RE: Regex =
-            Regex::new(r"^(?P<relnode>\d+)(_(?P<forkname>[a-z]+))?(\.(?P<segno>\d+))?$").unwrap();
-    }
+    static RELFILE_RE: OnceCell<Regex> = OnceCell::new();
+    RELFILE_RE.get_or_init(|| {
+        Regex::new(r"^(?P<relnode>\d+)(_(?P<forkname>[a-z]+))?(\.(?P<segno>\d+))?$").unwrap()
+    });
+
     let caps = RELFILE_RE
+        .get()
+        .unwrap()
         .captures(fname)
         .ok_or(FilePathError::InvalidFileName)?;
 

libs/postgres_ffi/src/waldecoder.rs

@@ -13,24 +13,30 @@ use super::xlog_utils::*;
 use super::XLogLongPageHeaderData;
 use super::XLogPageHeaderData;
 use super::XLogRecord;
+use super::XLOG_PAGE_MAGIC;
 use bytes::{Buf, BufMut, Bytes, BytesMut};
 use crc32c::*;
 use log::*;
 use std::cmp::min;
+use std::num::NonZeroU32;
 use thiserror::Error;
 use utils::lsn::Lsn;
 
+enum State {
+    WaitingForRecord,
+    ReassemblingRecord {
+        recordbuf: BytesMut,
+        contlen: NonZeroU32,
+    },
+    SkippingEverything {
+        skip_until_lsn: Lsn,
+    },
+}
+
 pub struct WalStreamDecoder {
     lsn: Lsn,
-
-    startlsn: Lsn, // LSN where this record starts
-    contlen: u32,
-    padlen: u32,
-
     inputbuf: BytesMut,
-
-    /// buffer used to reassemble records that cross page boundaries.
-    recordbuf: BytesMut,
+    state: State,
 }
 
 #[derive(Error, Debug, Clone)]
@@ -48,13 +54,8 @@ impl WalStreamDecoder {
     pub fn new(lsn: Lsn) -> WalStreamDecoder {
         WalStreamDecoder {
             lsn,
-
-            startlsn: Lsn(0),
-            contlen: 0,
-            padlen: 0,
-
             inputbuf: BytesMut::new(),
-            recordbuf: BytesMut::new(),
+            state: State::WaitingForRecord,
         }
     }
 
@@ -67,6 +68,58 @@ impl WalStreamDecoder {
         self.inputbuf.extend_from_slice(buf);
     }
 
+    fn validate_page_header(&self, hdr: &XLogPageHeaderData) -> Result<(), WalDecodeError> {
+        let validate_impl = || {
+            if hdr.xlp_magic != XLOG_PAGE_MAGIC as u16 {
+                return Err(format!(
+                    "invalid xlog page header: xlp_magic={}, expected {}",
+                    hdr.xlp_magic, XLOG_PAGE_MAGIC
+                ));
+            }
+            if hdr.xlp_pageaddr != self.lsn.0 {
+                return Err(format!(
+                    "invalid xlog page header: xlp_pageaddr={}, expected {}",
+                    hdr.xlp_pageaddr, self.lsn
+                ));
+            }
+            match self.state {
+                State::WaitingForRecord => {
+                    if hdr.xlp_info & XLP_FIRST_IS_CONTRECORD != 0 {
+                        return Err(
+                            "invalid xlog page header: unexpected XLP_FIRST_IS_CONTRECORD".into(),
+                        );
+                    }
+                    if hdr.xlp_rem_len != 0 {
+                        return Err(format!(
+                            "invalid xlog page header: xlp_rem_len={}, but it's not a contrecord",
+                            hdr.xlp_rem_len
+                        ));
+                    }
+                }
+                State::ReassemblingRecord { contlen, .. } => {
+                    if hdr.xlp_info & XLP_FIRST_IS_CONTRECORD == 0 {
+                        return Err(
+                            "invalid xlog page header: XLP_FIRST_IS_CONTRECORD expected, not found"
+                                .into(),
+                        );
+                    }
+                    if hdr.xlp_rem_len != contlen.get() {
+                        return Err(format!(
+                            "invalid xlog page header: xlp_rem_len={}, expected {}",
+                            hdr.xlp_rem_len,
+                            contlen.get()
+                        ));
+                    }
+                }
+                State::SkippingEverything { .. } => {
+                    panic!("Should not be validating page header in the SkippingEverything state");
+                }
+            };
+            Ok(())
+        };
+        validate_impl().map_err(|msg| WalDecodeError { msg, lsn: self.lsn })
+    }
+
     /// Attempt to decode another WAL record from the input that has been fed to the
     /// decoder so far.
     ///
@@ -76,128 +129,121 @@ impl WalStreamDecoder {
     ///     Err(WalDecodeError): an error occurred while decoding, meaning the input was invalid.
     ///
     pub fn poll_decode(&mut self) -> Result<Option<(Lsn, Bytes)>, WalDecodeError> {
-        let recordbuf;
-
         // Run state machine that validates page headers, and reassembles records
         // that cross page boundaries.
         loop {
             // parse and verify page boundaries as we go
-            if self.padlen > 0 {
-                // We should first skip padding, as we may have to skip some page headers if we're processing the XLOG_SWITCH record.
-                if self.inputbuf.remaining() < self.padlen as usize {
-                    return Ok(None);
-                }
+            // However, we may have to skip some page headers if we're processing the XLOG_SWITCH record or skipping padding for whatever reason.
+            match self.state {
+                State::WaitingForRecord | State::ReassemblingRecord { .. } => {
+                    if self.lsn.segment_offset(pg_constants::WAL_SEGMENT_SIZE) == 0 {
+                        // parse long header
 
-                // skip padding
-                self.inputbuf.advance(self.padlen as usize);
-                self.lsn += self.padlen as u64;
-                self.padlen = 0;
-            } else if self.lsn.segment_offset(pg_constants::WAL_SEGMENT_SIZE) == 0 {
-                // parse long header
+                        if self.inputbuf.remaining() < XLOG_SIZE_OF_XLOG_LONG_PHD {
+                            return Ok(None);
+                        }
 
-                if self.inputbuf.remaining() < XLOG_SIZE_OF_XLOG_LONG_PHD {
-                    return Ok(None);
-                }
+                        let hdr = XLogLongPageHeaderData::from_bytes(&mut self.inputbuf).map_err(
+                            |e| WalDecodeError {
+                                msg: format!("long header deserialization failed {}", e),
+                                lsn: self.lsn,
+                            },
+                        )?;
 
-                let hdr = XLogLongPageHeaderData::from_bytes(&mut self.inputbuf).map_err(|e| {
-                    WalDecodeError {
-                        msg: format!("long header deserialization failed {}", e),
-                        lsn: self.lsn,
+                        self.validate_page_header(&hdr.std)?;
+
+                        self.lsn += XLOG_SIZE_OF_XLOG_LONG_PHD as u64;
+                    } else if self.lsn.block_offset() == 0 {
+                        if self.inputbuf.remaining() < XLOG_SIZE_OF_XLOG_SHORT_PHD {
+                            return Ok(None);
+                        }
+
+                        let hdr =
+                            XLogPageHeaderData::from_bytes(&mut self.inputbuf).map_err(|e| {
+                                WalDecodeError {
+                                    msg: format!("header deserialization failed {}", e),
+                                    lsn: self.lsn,
+                                }
+                            })?;
+
+                        self.validate_page_header(&hdr)?;
+
+                        self.lsn += XLOG_SIZE_OF_XLOG_SHORT_PHD as u64;
                     }
-                })?;
-
-                if hdr.std.xlp_pageaddr != self.lsn.0 {
-                    return Err(WalDecodeError {
-                        msg: "invalid xlog segment header".into(),
-                        lsn: self.lsn,
-                    });
                 }
-                // TODO: verify the remaining fields in the header
-
-                self.lsn += XLOG_SIZE_OF_XLOG_LONG_PHD as u64;
-                continue;
-            } else if self.lsn.block_offset() == 0 {
-                if self.inputbuf.remaining() < XLOG_SIZE_OF_XLOG_SHORT_PHD {
-                    return Ok(None);
-                }
-
-                let hdr = XLogPageHeaderData::from_bytes(&mut self.inputbuf).map_err(|e| {
-                    WalDecodeError {
-                        msg: format!("header deserialization failed {}", e),
-                        lsn: self.lsn,
+                State::SkippingEverything { .. } => {}
+            }
+            match &mut self.state {
+                State::WaitingForRecord => {
+                    // need to have at least the xl_tot_len field
+                    if self.inputbuf.remaining() < 4 {
+                        return Ok(None);
                     }
-                })?;
 
-                if hdr.xlp_pageaddr != self.lsn.0 {
-                    return Err(WalDecodeError {
-                        msg: "invalid xlog page header".into(),
-                        lsn: self.lsn,
-                    });
+                    // peek xl_tot_len at the beginning of the record.
+                    // FIXME: assumes little-endian
+                    let xl_tot_len = (&self.inputbuf[0..4]).get_u32_le();
+                    if (xl_tot_len as usize) < XLOG_SIZE_OF_XLOG_RECORD {
+                        return Err(WalDecodeError {
+                            msg: format!("invalid xl_tot_len {}", xl_tot_len),
+                            lsn: self.lsn,
+                        });
+                    }
+                    // Fast path for the common case that the whole record fits on the page.
+                    let pageleft = self.lsn.remaining_in_block() as u32;
+                    if self.inputbuf.remaining() >= xl_tot_len as usize && xl_tot_len <= pageleft {
+                        self.lsn += xl_tot_len as u64;
+                        let recordbuf = self.inputbuf.copy_to_bytes(xl_tot_len as usize);
+                        return Ok(Some(self.complete_record(recordbuf)?));
+                    } else {
+                        // Need to assemble the record from pieces. Remember the size of the
+                        // record, and loop back. On next iteration, we will reach the 'else'
+                        // branch below, and copy the part of the record that was on this page
+                        // to 'recordbuf'.  Subsequent iterations will skip page headers, and
+                        // append the continuations from the next pages to 'recordbuf'.
+                        self.state = State::ReassemblingRecord {
+                            recordbuf: BytesMut::with_capacity(xl_tot_len as usize),
+                            contlen: NonZeroU32::new(xl_tot_len).unwrap(),
+                        }
+                    }
                 }
-                // TODO: verify the remaining fields in the header
+                State::ReassemblingRecord { recordbuf, contlen } => {
+                    // we're continuing a record, possibly from previous page.
+                    let pageleft = self.lsn.remaining_in_block() as u32;
 
-                self.lsn += XLOG_SIZE_OF_XLOG_SHORT_PHD as u64;
-                continue;
-            } else if self.contlen == 0 {
-                assert!(self.recordbuf.is_empty());
+                    // read the rest of the record, or as much as fits on this page.
+                    let n = min(contlen.get(), pageleft) as usize;
 
-                // need to have at least the xl_tot_len field
-                if self.inputbuf.remaining() < 4 {
-                    return Ok(None);
+                    if self.inputbuf.remaining() < n {
+                        return Ok(None);
+                    }
+
+                    recordbuf.put(self.inputbuf.split_to(n));
+                    self.lsn += n as u64;
+                    *contlen = match NonZeroU32::new(contlen.get() - n as u32) {
+                        Some(x) => x,
+                        None => {
+                            // The record is now complete.
+                            let recordbuf = std::mem::replace(recordbuf, BytesMut::new()).freeze();
+                            return Ok(Some(self.complete_record(recordbuf)?));
+                        }
+                    }
                 }
-
-                // peek xl_tot_len at the beginning of the record.
-                // FIXME: assumes little-endian
-                self.startlsn = self.lsn;
-                let xl_tot_len = (&self.inputbuf[0..4]).get_u32_le();
-                if (xl_tot_len as usize) < XLOG_SIZE_OF_XLOG_RECORD {
-                    return Err(WalDecodeError {
-                        msg: format!("invalid xl_tot_len {}", xl_tot_len),
-                        lsn: self.lsn,
-                    });
+                State::SkippingEverything { skip_until_lsn } => {
+                    assert!(*skip_until_lsn >= self.lsn);
+                    let n = skip_until_lsn.0 - self.lsn.0;
+                    if self.inputbuf.remaining() < n as usize {
+                        return Ok(None);
+                    }
+                    self.inputbuf.advance(n as usize);
+                    self.lsn += n;
+                    self.state = State::WaitingForRecord;
                 }
-
-                // Fast path for the common case that the whole record fits on the page.
-                let pageleft = self.lsn.remaining_in_block() as u32;
-                if self.inputbuf.remaining() >= xl_tot_len as usize && xl_tot_len <= pageleft {
-                    // Take the record from the 'inputbuf', and validate it.
-                    recordbuf = self.inputbuf.copy_to_bytes(xl_tot_len as usize);
-                    self.lsn += xl_tot_len as u64;
-                    break;
-                } else {
-                    // Need to assemble the record from pieces. Remember the size of the
-                    // record, and loop back. On next iteration, we will reach the 'else'
-                    // branch below, and copy the part of the record that was on this page
-                    // to 'recordbuf'.  Subsequent iterations will skip page headers, and
-                    // append the continuations from the next pages to 'recordbuf'.
-                    self.recordbuf.reserve(xl_tot_len as usize);
-                    self.contlen = xl_tot_len;
-                    continue;
-                }
-            } else {
-                // we're continuing a record, possibly from previous page.
-                let pageleft = self.lsn.remaining_in_block() as u32;
-
-                // read the rest of the record, or as much as fits on this page.
-                let n = min(self.contlen, pageleft) as usize;
-
-                if self.inputbuf.remaining() < n {
-                    return Ok(None);
-                }
-
-                self.recordbuf.put(self.inputbuf.split_to(n));
-                self.lsn += n as u64;
-                self.contlen -= n as u32;
-
-                if self.contlen == 0 {
-                    // The record is now complete.
-                    recordbuf = std::mem::replace(&mut self.recordbuf, BytesMut::new()).freeze();
-                    break;
-                }
-                continue;
             }
         }
+    }
 
+    fn complete_record(&mut self, recordbuf: Bytes) -> Result<(Lsn, Bytes), WalDecodeError> {
         // We now have a record in the 'recordbuf' local variable.
         let xlogrec =
             XLogRecord::from_slice(&recordbuf[0..XLOG_SIZE_OF_XLOG_RECORD]).map_err(|e| {
@@ -219,18 +265,20 @@ impl WalStreamDecoder {
 
         // XLOG_SWITCH records are special. If we see one, we need to skip
         // to the next WAL segment.
-        if xlogrec.is_xlog_switch_record() {
+        let next_lsn = if xlogrec.is_xlog_switch_record() {
             trace!("saw xlog switch record at {}", self.lsn);
-            self.padlen = self.lsn.calc_padding(pg_constants::WAL_SEGMENT_SIZE as u64) as u32;
+            self.lsn + self.lsn.calc_padding(pg_constants::WAL_SEGMENT_SIZE as u64)
         } else {
             // Pad to an 8-byte boundary
-            self.padlen = self.lsn.calc_padding(8u32) as u32;
-        }
+            self.lsn.align()
+        };
+        self.state = State::SkippingEverything {
+            skip_until_lsn: next_lsn,
+        };
 
         // We should return LSN of the next record, not the last byte of this record or
         // the byte immediately after. Note that this handles both XLOG_SWITCH and usual
         // records, the former "spans" until the next WAL segment (see test_xlog_switch).
-        let result = (self.lsn + self.padlen as u64, recordbuf);
-        Ok(Some(result))
+        Ok((next_lsn, recordbuf))
     }
 }

libs/postgres_ffi/src/xlog_utils.rs

@@ -15,7 +15,8 @@ use crate::XLogPageHeaderData;
 use crate::XLogRecord;
 use crate::XLOG_PAGE_MAGIC;
 
-use anyhow::{bail, ensure};
+use crate::pg_constants::WAL_SEGMENT_SIZE;
+use anyhow::{anyhow, bail, ensure};
 use byteorder::{ByteOrder, LittleEndian};
 use bytes::BytesMut;
 use bytes::{Buf, Bytes};
@@ -158,7 +159,7 @@ fn find_end_of_wal_segment(
     let mut buf = [0u8; XLOG_BLCKSZ];
     let file_name = XLogFileName(tli, segno, wal_seg_size);
     let mut last_valid_rec_pos: usize = start_offset; // assume at given start_offset begins new record
-    let mut file = File::open(data_dir.join(file_name.clone() + ".partial")).unwrap();
+    let mut file = File::open(data_dir.join(file_name.clone() + ".partial"))?;
     file.seek(SeekFrom::Start(offs as u64))?;
     // xl_crc is the last field in XLogRecord, will not be read into rec_hdr
     const_assert!(XLOG_RECORD_CRC_OFFS + 4 == XLOG_SIZE_OF_XLOG_RECORD);
@@ -395,10 +396,13 @@ pub fn find_end_of_wal(
     let mut high_tli: TimeLineID = 0;
     let mut high_ispartial = false;
 
-    for entry in fs::read_dir(data_dir).unwrap().flatten() {
+    for entry in fs::read_dir(data_dir)?.flatten() {
         let ispartial: bool;
         let entry_name = entry.file_name();
-        let fname = entry_name.to_str().unwrap();
+        let fname = entry_name
+            .to_str()
+            .ok_or_else(|| anyhow!("Invalid file name"))?;
+
         /*
          * Check if the filename looks like an xlog file, or a .partial file.
          */
@@ -410,7 +414,7 @@ pub fn find_end_of_wal(
             continue;
         }
         let (segno, tli) = XLogFromFileName(fname, wal_seg_size);
-        if !ispartial && entry.metadata().unwrap().len() != wal_seg_size as u64 {
+        if !ispartial && entry.metadata()?.len() != wal_seg_size as u64 {
             continue;
         }
         if segno > high_segno
@@ -461,8 +465,7 @@ pub fn find_end_of_wal(
 pub fn main() {
     let mut data_dir = PathBuf::new();
     data_dir.push(".");
-    let wal_seg_size = 16 * 1024 * 1024;
-    let (wal_end, tli) = find_end_of_wal(&data_dir, wal_seg_size, true, Lsn(0)).unwrap();
+    let (wal_end, tli) = find_end_of_wal(&data_dir, WAL_SEGMENT_SIZE, true, Lsn(0)).unwrap();
     println!(
         "wal_end={:>08X}{:>08X}, tli={}",
         (wal_end >> 32) as u32,
@@ -606,10 +609,9 @@ mod tests {
     fn test_end_of_wal<C: wal_craft::Crafter>(
         test_name: &str,
         expected_end_of_wal_non_partial: Lsn,
-        last_segment: &str,
     ) {
         use wal_craft::*;
-        // 1. Generate some WAL
+        // Craft some WAL
         let top_path = PathBuf::from(env!("CARGO_MANIFEST_DIR"))
             .join("..")
             .join("..");
@@ -622,24 +624,71 @@ mod tests {
         }
         cfg.initdb().unwrap();
         let srv = cfg.start_server().unwrap();
-        let expected_wal_end: Lsn =
-            u64::from(C::craft(&mut srv.connect_with_timeout().unwrap()).unwrap()).into();
+        let (intermediate_lsns, expected_end_of_wal_partial) =
+            C::craft(&mut srv.connect_with_timeout().unwrap()).unwrap();
+        let intermediate_lsns: Vec<Lsn> = intermediate_lsns
+            .iter()
+            .map(|&lsn| u64::from(lsn).into())
+            .collect();
+        let expected_end_of_wal_partial: Lsn = u64::from(expected_end_of_wal_partial).into();
         srv.kill();
 
-        // 2. Pick WAL generated by initdb
-        let wal_dir = cfg.datadir.join("pg_wal");
-        let wal_seg_size = 16 * 1024 * 1024;
+        // Check find_end_of_wal on the initial WAL
+        let last_segment = cfg
+            .wal_dir()
+            .read_dir()
+            .unwrap()
+            .map(|f| f.unwrap().file_name().into_string().unwrap())
+            .filter(|fname| IsXLogFileName(fname))
+            .max()
+            .unwrap();
+        check_pg_waldump_end_of_wal(&cfg, &last_segment, expected_end_of_wal_partial);
+        for start_lsn in std::iter::once(Lsn(0))
+            .chain(intermediate_lsns)
+            .chain(std::iter::once(expected_end_of_wal_partial))
+        {
+            // Erase all WAL before `start_lsn` to ensure it's not used by `find_end_of_wal`.
+            // We assume that `start_lsn` is non-decreasing.
+            info!(
+                "Checking with start_lsn={}, erasing WAL before it",
+                start_lsn
+            );
+            for file in fs::read_dir(cfg.wal_dir()).unwrap().flatten() {
+                let fname = file.file_name().into_string().unwrap();
+                if !IsXLogFileName(&fname) {
+                    continue;
+                }
+                let (segno, _) = XLogFromFileName(&fname, WAL_SEGMENT_SIZE);
+                let seg_start_lsn = XLogSegNoOffsetToRecPtr(segno, 0, WAL_SEGMENT_SIZE);
+                if seg_start_lsn > u64::from(start_lsn) {
+                    continue;
+                }
+                let mut f = File::options().write(true).open(file.path()).unwrap();
+                const ZEROS: [u8; WAL_SEGMENT_SIZE] = [0u8; WAL_SEGMENT_SIZE];
+                f.write_all(
+                    &ZEROS[0..min(
+                        WAL_SEGMENT_SIZE,
+                        (u64::from(start_lsn) - seg_start_lsn) as usize,
+                    )],
+                )
+                .unwrap();
+            }
+            check_end_of_wal(
+                &cfg,
+                &last_segment,
+                start_lsn,
+                expected_end_of_wal_non_partial,
+                expected_end_of_wal_partial,
+            );
+        }
+    }
 
-        // 3. Check end_of_wal on non-partial WAL segment (we treat it as fully populated)
-        let (wal_end, tli) = find_end_of_wal(&wal_dir, wal_seg_size, true, Lsn(0)).unwrap();
-        let wal_end = Lsn(wal_end);
-        info!(
-            "find_end_of_wal returned (wal_end={}, tli={})",
-            wal_end, tli
-        );
-        assert_eq!(wal_end, expected_end_of_wal_non_partial);
-
-        // 4. Get the actual end of WAL by pg_waldump
+    fn check_pg_waldump_end_of_wal(
+        cfg: &wal_craft::Conf,
+        last_segment: &str,
+        expected_end_of_wal: Lsn,
+    ) {
+        // Get the actual end of WAL by pg_waldump
         let waldump_output = cfg
             .pg_waldump("000000010000000000000001", last_segment)
             .unwrap()
@@ -658,32 +707,57 @@ mod tests {
         let waldump_wal_end = Lsn::from_str(caps.get(1).unwrap().as_str()).unwrap();
         info!(
             "waldump erred on {}, expected wal end at {}",
-            waldump_wal_end, expected_wal_end
+            waldump_wal_end, expected_end_of_wal
         );
-        assert_eq!(waldump_wal_end, expected_wal_end);
+        assert_eq!(waldump_wal_end, expected_end_of_wal);
+    }
 
-        // 5. Rename file to partial to actually find last valid lsn
-        fs::rename(
-            wal_dir.join(last_segment),
-            wal_dir.join(format!("{}.partial", last_segment)),
-        )
-        .unwrap();
-        let (wal_end, tli) = find_end_of_wal(&wal_dir, wal_seg_size, true, Lsn(0)).unwrap();
+    fn check_end_of_wal(
+        cfg: &wal_craft::Conf,
+        last_segment: &str,
+        start_lsn: Lsn,
+        expected_end_of_wal_non_partial: Lsn,
+        expected_end_of_wal_partial: Lsn,
+    ) {
+        // Check end_of_wal on non-partial WAL segment (we treat it as fully populated)
+        let (wal_end, tli) =
+            find_end_of_wal(&cfg.wal_dir(), WAL_SEGMENT_SIZE, true, start_lsn).unwrap();
         let wal_end = Lsn(wal_end);
         info!(
-            "find_end_of_wal returned (wal_end={}, tli={})",
+            "find_end_of_wal returned (wal_end={}, tli={}) with non-partial WAL segment",
             wal_end, tli
         );
-        assert_eq!(wal_end, waldump_wal_end);
+        assert_eq!(wal_end, expected_end_of_wal_non_partial);
+
+        // Rename file to partial to actually find last valid lsn, then rename it back.
+        fs::rename(
+            cfg.wal_dir().join(&last_segment),
+            cfg.wal_dir().join(format!("{}.partial", last_segment)),
+        )
+        .unwrap();
+        let (wal_end, tli) =
+            find_end_of_wal(&cfg.wal_dir(), WAL_SEGMENT_SIZE, true, start_lsn).unwrap();
+        let wal_end = Lsn(wal_end);
+        info!(
+            "find_end_of_wal returned (wal_end={}, tli={}) with partial WAL segment",
+            wal_end, tli
+        );
+        assert_eq!(wal_end, expected_end_of_wal_partial);
+        fs::rename(
+            cfg.wal_dir().join(format!("{}.partial", last_segment)),
+            cfg.wal_dir().join(last_segment),
+        )
+        .unwrap();
     }
 
+    const_assert!(WAL_SEGMENT_SIZE == 16 * 1024 * 1024);
+
     #[test]
     pub fn test_find_end_of_wal_simple() {
         init_logging();
         test_end_of_wal::<wal_craft::Simple>(
             "test_find_end_of_wal_simple",
             "0/2000000".parse::<Lsn>().unwrap(),
-            "000000010000000000000001",
         );
     }
 
@@ -693,7 +767,6 @@ mod tests {
         test_end_of_wal::<wal_craft::WalRecordCrossingSegmentFollowedBySmallOne>(
             "test_find_end_of_wal_crossing_segment_followed_by_small_one",
             "0/3000000".parse::<Lsn>().unwrap(),
-            "000000010000000000000002",
         );
     }
 
@@ -704,7 +777,6 @@ mod tests {
         test_end_of_wal::<wal_craft::LastWalRecordCrossingSegment>(
             "test_find_end_of_wal_last_crossing_segment",
             "0/3000000".parse::<Lsn>().unwrap(),
-            "000000010000000000000002",
         );
     }
 

libs/postgres_ffi/wal_craft/Cargo.toml

@@ -10,7 +10,7 @@ anyhow = "1.0"
 clap = "3.0"
 env_logger = "0.9"
 log = "0.4"
-once_cell = "1.8.0"
+once_cell = "1.13.0"
 postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="d052ee8b86fff9897c77b0fe89ea9daba0e1fa38" }
 postgres_ffi = { path = "../" }
 tempfile = "3.2"

libs/postgres_ffi/wal_craft/src/bin/wal_craft.rs

@@ -55,7 +55,7 @@ fn main() -> Result<()> {
         .get_matches();
 
     let wal_craft = |arg_matches: &ArgMatches, client| {
-        let lsn = match arg_matches.value_of("type").unwrap() {
+        let (intermediate_lsns, end_of_wal_lsn) = match arg_matches.value_of("type").unwrap() {
             Simple::NAME => Simple::craft(client)?,
             LastWalRecordXlogSwitch::NAME => LastWalRecordXlogSwitch::craft(client)?,
             LastWalRecordXlogSwitchEndsOnPageBoundary::NAME => {
@@ -67,7 +67,10 @@ fn main() -> Result<()> {
             LastWalRecordCrossingSegment::NAME => LastWalRecordCrossingSegment::craft(client)?,
             a => panic!("Unknown --type argument: {}", a),
         };
-        println!("end_of_wal = {}", lsn);
+        for lsn in intermediate_lsns {
+            println!("intermediate_lsn = {}", lsn);
+        }
+        println!("end_of_wal = {}", end_of_wal_lsn);
         Ok(())
     };
 

libs/postgres_ffi/wal_craft/src/lib.rs

@@ -4,6 +4,7 @@ use log::*;
 use once_cell::sync::Lazy;
 use postgres::types::PgLsn;
 use postgres::Client;
+use postgres_ffi::pg_constants::WAL_SEGMENT_SIZE;
 use postgres_ffi::xlog_utils::{
     XLOG_BLCKSZ, XLOG_SIZE_OF_XLOG_RECORD, XLOG_SIZE_OF_XLOG_SHORT_PHD,
 };
@@ -45,6 +46,10 @@ impl Conf {
         self.pg_distrib_dir.join("lib")
     }
 
+    pub fn wal_dir(&self) -> PathBuf {
+        self.datadir.join("pg_wal")
+    }
+
     fn new_pg_command(&self, command: impl AsRef<Path>) -> Result<Command> {
         let path = self.pg_bin_dir().join(command);
         ensure!(path.exists(), "Command {:?} does not exist", path);
@@ -211,7 +216,7 @@ pub fn ensure_server_config(client: &mut impl postgres::GenericClient) -> Result
         "Unexpected wal_segment_size unit"
     );
     ensure!(
-        wal_segment_size.get::<_, i64>("setting") == 16 * 1024 * 1024,
+        wal_segment_size.get::<_, i64>("setting") == WAL_SEGMENT_SIZE as i64,
         "Unexpected wal_segment_size in bytes"
     );
 
@@ -221,20 +226,24 @@ pub fn ensure_server_config(client: &mut impl postgres::GenericClient) -> Result
 pub trait Crafter {
     const NAME: &'static str;
 
-    /// Generates WAL using the client `client`. Returns the expected end-of-wal LSN.
-    fn craft(client: &mut impl postgres::GenericClient) -> Result<PgLsn>;
+    /// Generates WAL using the client `client`. Returns a pair of:
+    /// * A vector of some valid "interesting" intermediate LSNs which one may start reading from.
+    ///   May include or exclude Lsn(0) and the end-of-wal.
+    /// * The expected end-of-wal LSN.
+    fn craft(client: &mut impl postgres::GenericClient) -> Result<(Vec<PgLsn>, PgLsn)>;
 }
 
 fn craft_internal<C: postgres::GenericClient>(
     client: &mut C,
-    f: impl Fn(&mut C, PgLsn) -> Result<Option<PgLsn>>,
-) -> Result<PgLsn> {
+    f: impl Fn(&mut C, PgLsn) -> Result<(Vec<PgLsn>, Option<PgLsn>)>,
+) -> Result<(Vec<PgLsn>, PgLsn)> {
     ensure_server_config(client)?;
 
     let initial_lsn = client.pg_current_wal_insert_lsn()?;
     info!("LSN initial = {}", initial_lsn);
 
-    let last_lsn = match f(client, initial_lsn)? {
+    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) => match last_lsn.cmp(&client.pg_current_wal_insert_lsn()?) {
             Ordering::Less => bail!("Some records were inserted after the crafted WAL"),
@@ -242,6 +251,9 @@ fn craft_internal<C: postgres::GenericClient>(
             Ordering::Greater => bail!("Reported LSN is greater than insert_lsn"),
         },
     };
+    if !intermediate_lsns.starts_with(&[initial_lsn]) {
+        intermediate_lsns.insert(0, initial_lsn);
+    }
 
     // Some records may be not flushed, e.g. non-transactional logical messages.
     client.execute("select neon_xlogflush(pg_current_wal_insert_lsn())", &[])?;
@@ -250,16 +262,16 @@ fn craft_internal<C: postgres::GenericClient>(
         Ordering::Equal => {}
         Ordering::Greater => bail!("Reported LSN is greater than flush_lsn"),
     }
-    Ok(last_lsn)
+    Ok((intermediate_lsns, last_lsn))
 }
 
 pub struct Simple;
 impl Crafter for Simple {
     const NAME: &'static str = "simple";
-    fn craft(client: &mut impl postgres::GenericClient) -> Result<PgLsn> {
+    fn craft(client: &mut impl postgres::GenericClient) -> Result<(Vec<PgLsn>, PgLsn)> {
         craft_internal(client, |client, _| {
             client.execute("CREATE table t(x int)", &[])?;
-            Ok(None)
+            Ok((Vec::new(), None))
         })
     }
 }
@@ -267,12 +279,13 @@ impl Crafter for Simple {
 pub struct LastWalRecordXlogSwitch;
 impl Crafter for LastWalRecordXlogSwitch {
     const NAME: &'static str = "last_wal_record_xlog_switch";
-    fn craft(client: &mut impl postgres::GenericClient) -> Result<PgLsn> {
+    fn craft(client: &mut impl postgres::GenericClient) -> Result<(Vec<PgLsn>, PgLsn)> {
         // Do not use generate_internal because here we end up with flush_lsn exactly on
         // the segment boundary and insert_lsn after the initial page header, which is unusual.
         ensure_server_config(client)?;
 
         client.execute("CREATE table t(x int)", &[])?;
+        let before_xlog_switch = client.pg_current_wal_insert_lsn()?;
         let after_xlog_switch: PgLsn = client.query_one("SELECT pg_switch_wal()", &[])?.get(0);
         let next_segment = PgLsn::from(0x0200_0000);
         ensure!(
@@ -281,14 +294,14 @@ impl Crafter for LastWalRecordXlogSwitch {
             after_xlog_switch,
             next_segment
         );
-        Ok(next_segment)
+        Ok((vec![before_xlog_switch, after_xlog_switch], next_segment))
     }
 }
 
 pub struct LastWalRecordXlogSwitchEndsOnPageBoundary;
 impl Crafter for LastWalRecordXlogSwitchEndsOnPageBoundary {
     const NAME: &'static str = "last_wal_record_xlog_switch_ends_on_page_boundary";
-    fn craft(client: &mut impl postgres::GenericClient) -> Result<PgLsn> {
+    fn craft(client: &mut impl postgres::GenericClient) -> Result<(Vec<PgLsn>, PgLsn)> {
         // Do not use generate_internal because here we end up with flush_lsn exactly on
         // the segment boundary and insert_lsn after the initial page header, which is unusual.
         ensure_server_config(client)?;
@@ -334,6 +347,7 @@ impl Crafter for LastWalRecordXlogSwitchEndsOnPageBoundary {
         );
 
         // Emit the XLOG_SWITCH
+        let before_xlog_switch = client.pg_current_wal_insert_lsn()?;
         let after_xlog_switch: PgLsn = client.query_one("SELECT pg_switch_wal()", &[])?.get(0);
         let next_segment = PgLsn::from(0x0200_0000);
         ensure!(
@@ -347,14 +361,14 @@ impl Crafter for LastWalRecordXlogSwitchEndsOnPageBoundary {
             "XLOG_SWITCH message ended not on page boundary: {}",
             after_xlog_switch
         );
-        Ok(next_segment)
+        Ok((vec![before_xlog_switch, after_xlog_switch], next_segment))
     }
 }
 
 fn craft_single_logical_message(
     client: &mut impl postgres::GenericClient,
     transactional: bool,
-) -> Result<PgLsn> {
+) -> Result<(Vec<PgLsn>, PgLsn)> {
     craft_internal(client, |client, initial_lsn| {
         ensure!(
             initial_lsn < PgLsn::from(0x0200_0000 - 1024 * 1024),
@@ -386,9 +400,9 @@ fn craft_single_logical_message(
                 message_lsn < after_message_lsn,
                 "No record found after the emitted message"
             );
-            Ok(Some(after_message_lsn))
+            Ok((vec![message_lsn], Some(after_message_lsn)))
         } else {
-            Ok(Some(message_lsn))
+            Ok((Vec::new(), Some(message_lsn)))
         }
     })
 }
@@ -396,7 +410,7 @@ fn craft_single_logical_message(
 pub struct WalRecordCrossingSegmentFollowedBySmallOne;
 impl Crafter for WalRecordCrossingSegmentFollowedBySmallOne {
     const NAME: &'static str = "wal_record_crossing_segment_followed_by_small_one";
-    fn craft(client: &mut impl postgres::GenericClient) -> Result<PgLsn> {
+    fn craft(client: &mut impl postgres::GenericClient) -> Result<(Vec<PgLsn>, PgLsn)> {
         craft_single_logical_message(client, true)
     }
 }
@@ -404,7 +418,7 @@ impl Crafter for WalRecordCrossingSegmentFollowedBySmallOne {
 pub struct LastWalRecordCrossingSegment;
 impl Crafter for LastWalRecordCrossingSegment {
     const NAME: &'static str = "last_wal_record_crossing_segment";
-    fn craft(client: &mut impl postgres::GenericClient) -> Result<PgLsn> {
+    fn craft(client: &mut impl postgres::GenericClient) -> Result<(Vec<PgLsn>, PgLsn)> {
         craft_single_logical_message(client, false)
     }
 }

libs/remote_storage/Cargo.toml

@@ -7,7 +7,7 @@ edition = "2021"
 anyhow = { version = "1.0", features = ["backtrace"] }
 async-trait = "0.1"
 metrics = { version = "0.1", path = "../metrics" }
-once_cell = "1.8.0"
+once_cell = "1.13.0"
 rusoto_core = "0.48"
 rusoto_s3 = "0.48"
 serde = { version = "1.0", features = ["derive"] }

libs/remote_storage/src/lib.rs

@@ -66,6 +66,9 @@ pub trait RemoteStorage: Send + Sync {
     async fn list(&self) -> anyhow::Result<Vec<Self::RemoteObjectId>>;
 
     /// Lists all top level subdirectories for a given prefix
+    /// Note: here we assume that if the prefix is passed it was obtained via remote_object_id
+    /// which already takes into account any kind of global prefix (prefix_in_bucket for S3 or storage_root for LocalFS)
+    /// so this method doesnt need to.
     async fn list_prefixes(
         &self,
         prefix: Option<Self::RemoteObjectId>,

libs/remote_storage/src/local_fs.rs

@@ -116,7 +116,7 @@ impl RemoteStorage for LocalFs {
         prefix: Option<Self::RemoteObjectId>,
     ) -> anyhow::Result<Vec<Self::RemoteObjectId>> {
         let path = match prefix {
-            Some(prefix) => Cow::Owned(self.storage_root.join(prefix)),
+            Some(prefix) => Cow::Owned(prefix),
             None => Cow::Borrowed(&self.storage_root),
         };
         get_all_files(path.as_ref(), false).await

libs/remote_storage/src/s3_bucket.rs

@@ -171,17 +171,25 @@ impl S3Bucket {
 
         let access_key_id = std::env::var("AWS_ACCESS_KEY_ID").ok();
         let secret_access_key = std::env::var("AWS_SECRET_ACCESS_KEY").ok();
+        // session token is used when authorizing through sso
+        // which is typically the case when testing locally on developer machine
+        let session_token = std::env::var("AWS_SESSION_TOKEN").ok();
 
         let client = if access_key_id.is_none() && secret_access_key.is_none() {
             debug!("Using IAM-based AWS access");
             S3Client::new_with(request_dispatcher, InstanceMetadataProvider::new(), region)
         } else {
-            debug!("Using credentials-based AWS access");
+            debug!(
+                "Using credentials-based AWS access. Session token is set: {}",
+                session_token.is_some()
+            );
             S3Client::new_with(
                 request_dispatcher,
-                StaticProvider::new_minimal(
+                StaticProvider::new(
                     access_key_id.unwrap_or_default(),
                     secret_access_key.unwrap_or_default(),
+                    session_token,
+                    None,
                 ),
                 region,
             )
@@ -304,32 +312,24 @@ impl RemoteStorage for S3Bucket {
         Ok(document_keys)
     }
 
+    /// See the doc for `RemoteStorage::list_prefixes`
     /// Note: it wont include empty "directories"
     async fn list_prefixes(
         &self,
         prefix: Option<Self::RemoteObjectId>,
     ) -> anyhow::Result<Vec<Self::RemoteObjectId>> {
-        let list_prefix = match prefix {
-            Some(prefix) => {
-                let mut prefix_in_bucket = self.prefix_in_bucket.clone().unwrap_or_default();
-                // if there is no trailing / in default prefix and
-                // supplied prefix does not start with "/" insert it
-                if !(prefix_in_bucket.ends_with(S3_PREFIX_SEPARATOR)
-                    || prefix.0.starts_with(S3_PREFIX_SEPARATOR))
-                {
-                    prefix_in_bucket.push(S3_PREFIX_SEPARATOR);
-                }
-
-                prefix_in_bucket.push_str(&prefix.0);
+        // get the passed prefix or if it is not set use prefix_in_bucket value
+        let list_prefix = prefix
+            .map(|p| p.0)
+            .or_else(|| self.prefix_in_bucket.clone())
+            .map(|mut p| {
                 // required to end with a separator
                 // otherwise request will return only the entry of a prefix
-                if !prefix_in_bucket.ends_with(S3_PREFIX_SEPARATOR) {
-                    prefix_in_bucket.push(S3_PREFIX_SEPARATOR);
+                if !p.ends_with(S3_PREFIX_SEPARATOR) {
+                    p.push(S3_PREFIX_SEPARATOR);
                 }
-                Some(prefix_in_bucket)
-            }
-            None => self.prefix_in_bucket.clone(),
-        };
+                p
+            });
 
         let mut document_keys = Vec::new();
 

libs/utils/Cargo.toml

@@ -8,7 +8,6 @@ anyhow = "1.0"
 bincode = "1.3"
 bytes = "1.0.1"
 hyper = { version = "0.14.7", features = ["full"] }
-lazy_static = "1.4.0"
 pin-project-lite = "0.2.7"
 postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="d052ee8b86fff9897c77b0fe89ea9daba0e1fa38" }
 postgres-protocol = { git = "https://github.com/zenithdb/rust-postgres.git", rev="d052ee8b86fff9897c77b0fe89ea9daba0e1fa38" }
@@ -28,6 +27,8 @@ rustls = "0.20.2"
 rustls-split = "0.3.0"
 git-version = "0.3.5"
 serde_with = "1.12.0"
+once_cell = "1.13.0"
+
 
 metrics = { path = "../metrics" }
 workspace_hack = { version = "0.1", path = "../../workspace_hack" }

libs/utils/src/http/endpoint.rs

@@ -4,8 +4,8 @@ use crate::zid::ZTenantId;
 use anyhow::anyhow;
 use hyper::header::AUTHORIZATION;
 use hyper::{header::CONTENT_TYPE, Body, Request, Response, Server};
-use lazy_static::lazy_static;
 use metrics::{register_int_counter, Encoder, IntCounter, TextEncoder};
+use once_cell::sync::Lazy;
 use routerify::ext::RequestExt;
 use routerify::RequestInfo;
 use routerify::{Middleware, Router, RouterBuilder, RouterService};
@@ -16,13 +16,13 @@ use std::net::TcpListener;
 
 use super::error::ApiError;
 
-lazy_static! {
-    static ref SERVE_METRICS_COUNT: IntCounter = register_int_counter!(
+static SERVE_METRICS_COUNT: Lazy<IntCounter> = Lazy::new(|| {
+    register_int_counter!(
         "libmetrics_metric_handler_requests_total",
         "Number of metric requests made"
     )
-    .expect("failed to define a metric");
-}
+    .expect("failed to define a metric")
+});
 
 async fn logger(res: Response<Body>, info: RequestInfo) -> Result<Response<Body>, ApiError> {
     info!("{} {} {}", info.method(), info.uri().path(), res.status(),);

libs/utils/src/pq_proto.rs

@@ -47,10 +47,12 @@ pub enum FeStartupPacket {
     StartupMessage {
         major_version: u32,
         minor_version: u32,
-        params: HashMap<String, String>,
+        params: StartupMessageParams,
     },
 }
 
+pub type StartupMessageParams = HashMap<String, String>;
+
 #[derive(Debug, Hash, PartialEq, Eq, Clone, Copy)]
 pub struct CancelKeyData {
     pub backend_pid: i32,

libs/utils/tests/ssl_test.rs

@@ -7,7 +7,7 @@ use std::{
 
 use byteorder::{BigEndian, ReadBytesExt, WriteBytesExt};
 use bytes::{Buf, BufMut, Bytes, BytesMut};
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
 
 use utils::postgres_backend::{AuthType, Handler, PostgresBackend};
 
@@ -19,16 +19,15 @@ fn make_tcp_pair() -> (TcpStream, TcpStream) {
     (server_stream, client_stream)
 }
 
-lazy_static! {
-    static ref KEY: rustls::PrivateKey = {
-        let mut cursor = Cursor::new(include_bytes!("key.pem"));
-        rustls::PrivateKey(rustls_pemfile::rsa_private_keys(&mut cursor).unwrap()[0].clone())
-    };
-    static ref CERT: rustls::Certificate = {
-        let mut cursor = Cursor::new(include_bytes!("cert.pem"));
-        rustls::Certificate(rustls_pemfile::certs(&mut cursor).unwrap()[0].clone())
-    };
-}
+static KEY: Lazy<rustls::PrivateKey> = Lazy::new(|| {
+    let mut cursor = Cursor::new(include_bytes!("key.pem"));
+    rustls::PrivateKey(rustls_pemfile::rsa_private_keys(&mut cursor).unwrap()[0].clone())
+});
+
+static CERT: Lazy<rustls::Certificate> = Lazy::new(|| {
+    let mut cursor = Cursor::new(include_bytes!("cert.pem"));
+    rustls::Certificate(rustls_pemfile::certs(&mut cursor).unwrap()[0].clone())
+});
 
 #[test]
 fn ssl() {

neon_local/Cargo.toml

@@ -15,6 +15,5 @@ git-version = "0.3.5"
 pageserver = { path = "../pageserver" }
 control_plane = { path = "../control_plane" }
 safekeeper = { path = "../safekeeper" }
-postgres_ffi = { path = "../libs/postgres_ffi" }
 utils = { path = "../libs/utils" }
 workspace_hack = { version = "0.1", path = "../workspace_hack" }

neon_local/src/main.rs

@@ -9,6 +9,7 @@ use pageserver::config::defaults::{
     DEFAULT_HTTP_LISTEN_ADDR as DEFAULT_PAGESERVER_HTTP_ADDR,
     DEFAULT_PG_LISTEN_ADDR as DEFAULT_PAGESERVER_PG_ADDR,
 };
+use pageserver::http::models::TimelineInfo;
 use safekeeper::defaults::{
     DEFAULT_HTTP_LISTEN_PORT as DEFAULT_SAFEKEEPER_HTTP_PORT,
     DEFAULT_PG_LISTEN_PORT as DEFAULT_SAFEKEEPER_PG_PORT,
@@ -25,8 +26,6 @@ use utils::{
     zid::{NodeId, ZTenantId, ZTenantTimelineId, ZTimelineId},
 };
 
-use pageserver::timelines::TimelineInfo;
-
 // Default id of a safekeeper node, if not specified on the command line.
 const DEFAULT_SAFEKEEPER_ID: NodeId = NodeId(1);
 const DEFAULT_PAGESERVER_ID: NodeId = NodeId(1);
@@ -885,7 +884,7 @@ fn handle_pageserver(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> Resul
     match sub_match.subcommand() {
         Some(("start", start_match)) => {
             if let Err(e) = pageserver.start(&pageserver_config_overrides(start_match)) {
-                eprintln!("pageserver start failed: {}", e);
+                eprintln!("pageserver start failed: {e}");
                 exit(1);
             }
         }
@@ -907,10 +906,19 @@ fn handle_pageserver(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> Resul
             }
 
             if let Err(e) = pageserver.start(&pageserver_config_overrides(restart_match)) {
-                eprintln!("pageserver start failed: {}", e);
+                eprintln!("pageserver start failed: {e}");
                 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"),
     }

pageserver/Cargo.toml

@@ -21,7 +21,6 @@ futures = "0.3.13"
 hex = "0.4.3"
 hyper = "0.14"
 itertools = "0.10.3"
-lazy_static = "1.4.0"
 clap = "3.0"
 daemonize = "0.4.1"
 tokio = { version = "1.17", features = ["process", "sync", "macros", "fs", "rt", "io-util", "time"] }
@@ -29,7 +28,6 @@ postgres-types = { git = "https://github.com/zenithdb/rust-postgres.git", rev="d
 postgres-protocol = { git = "https://github.com/zenithdb/rust-postgres.git", rev="d052ee8b86fff9897c77b0fe89ea9daba0e1fa38" }
 postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="d052ee8b86fff9897c77b0fe89ea9daba0e1fa38" }
 tokio-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="d052ee8b86fff9897c77b0fe89ea9daba0e1fa38" }
-tokio-stream = "0.1.8"
 anyhow = { version = "1.0", features = ["backtrace"] }
 crc32c = "0.6.0"
 thiserror = "1.0"
@@ -49,7 +47,7 @@ tracing = "0.1.27"
 signal-hook = "0.3.10"
 url = "2"
 nix = "0.23"
-once_cell = "1.8.0"
+once_cell = "1.13.0"
 crossbeam-utils = "0.8.5"
 fail = "0.5.0"
 git-version = "0.3.5"

pageserver/src/basebackup.rs

@@ -23,8 +23,7 @@ use tar::{Builder, EntryType, Header};
 use tracing::*;
 
 use crate::reltag::{RelTag, SlruKind};
-use crate::repository::Timeline;
-use crate::DatadirTimelineImpl;
+use crate::DatadirTimeline;
 use postgres_ffi::xlog_utils::*;
 use postgres_ffi::*;
 use utils::lsn::Lsn;
@@ -32,12 +31,13 @@ use utils::lsn::Lsn;
 /// This is short-living object only for the time of tarball creation,
 /// created mostly to avoid passing a lot of parameters between various functions
 /// used for constructing tarball.
-pub struct Basebackup<'a, W>
+pub struct Basebackup<'a, W, T>
 where
     W: Write,
+    T: DatadirTimeline,
 {
     ar: Builder<AbortableWrite<W>>,
-    timeline: &'a Arc<DatadirTimelineImpl>,
+    timeline: &'a Arc<T>,
     pub lsn: Lsn,
     prev_record_lsn: Lsn,
     full_backup: bool,
@@ -52,17 +52,18 @@ where
 //  * When working without safekeepers. In this situation it is important to match the lsn
 //    we are taking basebackup on with the lsn that is used in pageserver's walreceiver
 //    to start the replication.
-impl<'a, W> Basebackup<'a, W>
+impl<'a, W, T> Basebackup<'a, W, T>
 where
     W: Write,
+    T: DatadirTimeline,
 {
     pub fn new(
         write: W,
-        timeline: &'a Arc<DatadirTimelineImpl>,
+        timeline: &'a Arc<T>,
         req_lsn: Option<Lsn>,
         prev_lsn: Option<Lsn>,
         full_backup: bool,
-    ) -> Result<Basebackup<'a, W>> {
+    ) -> Result<Basebackup<'a, W, T>> {
         // Compute postgres doesn't have any previous WAL files, but the first
         // record that it's going to write needs to include the LSN of the
         // previous record (xl_prev). We include prev_record_lsn in the
@@ -79,13 +80,13 @@ where
         let (backup_prev, backup_lsn) = if let Some(req_lsn) = req_lsn {
             // Backup was requested at a particular LSN. Wait for it to arrive.
             info!("waiting for {}", req_lsn);
-            timeline.tline.wait_lsn(req_lsn)?;
+            timeline.wait_lsn(req_lsn)?;
 
             // If the requested point is the end of the timeline, we can
             // provide prev_lsn. (get_last_record_rlsn() might return it as
             // zero, though, if no WAL has been generated on this timeline
             // yet.)
-            let end_of_timeline = timeline.tline.get_last_record_rlsn();
+            let end_of_timeline = timeline.get_last_record_rlsn();
             if req_lsn == end_of_timeline.last {
                 (end_of_timeline.prev, req_lsn)
             } else {
@@ -93,7 +94,7 @@ where
             }
         } else {
             // Backup was requested at end of the timeline.
-            let end_of_timeline = timeline.tline.get_last_record_rlsn();
+            let end_of_timeline = timeline.get_last_record_rlsn();
             (end_of_timeline.prev, end_of_timeline.last)
         };
 
@@ -371,7 +372,7 @@ where
         // add zenith.signal file
         let mut zenith_signal = String::new();
         if self.prev_record_lsn == Lsn(0) {
-            if self.lsn == self.timeline.tline.get_ancestor_lsn() {
+            if self.lsn == self.timeline.get_ancestor_lsn() {
                 write!(zenith_signal, "PREV LSN: none")?;
             } else {
                 write!(zenith_signal, "PREV LSN: invalid")?;
@@ -402,9 +403,10 @@ where
     }
 }
 
-impl<'a, W> Drop for Basebackup<'a, W>
+impl<'a, W, T> Drop for Basebackup<'a, W, T>
 where
     W: Write,
+    T: DatadirTimeline,
 {
     /// If the basebackup was not finished, prevent the Archive::drop() from
     /// writing the end-of-archive marker.

pageserver/src/config.rs

@@ -59,6 +59,7 @@ pub mod defaults {
 
 # [tenant_config]
 #checkpoint_distance = {DEFAULT_CHECKPOINT_DISTANCE} # in bytes
+#checkpoint_timeout = {DEFAULT_CHECKPOINT_TIMEOUT}
 #compaction_target_size = {DEFAULT_COMPACTION_TARGET_SIZE} # in bytes
 #compaction_period = '{DEFAULT_COMPACTION_PERIOD}'
 #compaction_threshold = '{DEFAULT_COMPACTION_THRESHOLD}'
@@ -452,6 +453,13 @@ impl PageServerConf {
                 Some(parse_toml_u64("checkpoint_distance", checkpoint_distance)?);
         }
 
+        if let Some(checkpoint_timeout) = item.get("checkpoint_timeout") {
+            t_conf.checkpoint_timeout = Some(parse_toml_duration(
+                "checkpoint_timeout",
+                checkpoint_timeout,
+            )?);
+        }
+
         if let Some(compaction_target_size) = item.get("compaction_target_size") {
             t_conf.compaction_target_size = Some(parse_toml_u64(
                 "compaction_target_size",

pageserver/src/http/models.rs

@@ -7,6 +7,10 @@ use utils::{
     zid::{NodeId, ZTenantId, ZTimelineId},
 };
 
+// These enums are used in the API response fields.
+use crate::repository::LocalTimelineState;
+use crate::tenant_mgr::TenantState;
+
 #[serde_as]
 #[derive(Serialize, Deserialize)]
 pub struct TimelineCreateRequest {
@@ -28,6 +32,7 @@ pub struct TenantCreateRequest {
     #[serde_as(as = "Option<DisplayFromStr>")]
     pub new_tenant_id: Option<ZTenantId>,
     pub checkpoint_distance: Option<u64>,
+    pub checkpoint_timeout: Option<String>,
     pub compaction_target_size: Option<u64>,
     pub compaction_period: Option<String>,
     pub compaction_threshold: Option<usize>,
@@ -66,6 +71,7 @@ pub struct TenantConfigRequest {
     #[serde(default)]
     #[serde_as(as = "Option<DisplayFromStr>")]
     pub checkpoint_distance: Option<u64>,
+    pub checkpoint_timeout: Option<String>,
     pub compaction_target_size: Option<u64>,
     pub compaction_period: Option<String>,
     pub compaction_threshold: Option<usize>,
@@ -83,6 +89,7 @@ impl TenantConfigRequest {
         TenantConfigRequest {
             tenant_id,
             checkpoint_distance: None,
+            checkpoint_timeout: None,
             compaction_target_size: None,
             compaction_period: None,
             compaction_threshold: None,
@@ -97,14 +104,59 @@ impl TenantConfigRequest {
     }
 }
 
-/// A WAL receiver's data stored inside the global `WAL_RECEIVERS`.
-/// We keep one WAL receiver active per timeline.
+#[serde_as]
+#[derive(Serialize, Deserialize, Clone)]
+pub struct TenantInfo {
+    #[serde_as(as = "DisplayFromStr")]
+    pub id: ZTenantId,
+    pub state: Option<TenantState>,
+    pub current_physical_size: Option<u64>, // physical size is only included in `tenant_status` endpoint
+    pub has_in_progress_downloads: Option<bool>,
+}
+
 #[serde_as]
 #[derive(Debug, Serialize, Deserialize, Clone)]
-pub struct WalReceiverEntry {
-    pub wal_producer_connstr: Option<String>,
+pub struct LocalTimelineInfo {
+    #[serde_as(as = "Option<DisplayFromStr>")]
+    pub ancestor_timeline_id: Option<ZTimelineId>,
+    #[serde_as(as = "Option<DisplayFromStr>")]
+    pub ancestor_lsn: Option<Lsn>,
+    #[serde_as(as = "DisplayFromStr")]
+    pub last_record_lsn: Lsn,
+    #[serde_as(as = "Option<DisplayFromStr>")]
+    pub prev_record_lsn: Option<Lsn>,
+    #[serde_as(as = "DisplayFromStr")]
+    pub latest_gc_cutoff_lsn: Lsn,
+    #[serde_as(as = "DisplayFromStr")]
+    pub disk_consistent_lsn: Lsn,
+    pub current_logical_size: Option<usize>, // is None when timeline is Unloaded
+    pub current_physical_size: Option<u64>,  // is None when timeline is Unloaded
+    pub current_logical_size_non_incremental: Option<usize>,
+    pub current_physical_size_non_incremental: Option<u64>,
+    pub timeline_state: LocalTimelineState,
+
+    pub wal_source_connstr: Option<String>,
     #[serde_as(as = "Option<DisplayFromStr>")]
     pub last_received_msg_lsn: Option<Lsn>,
     /// the timestamp (in microseconds) of the last received message
     pub last_received_msg_ts: Option<u128>,
 }
+
+#[serde_as]
+#[derive(Debug, Serialize, Deserialize, Clone)]
+pub struct RemoteTimelineInfo {
+    #[serde_as(as = "DisplayFromStr")]
+    pub remote_consistent_lsn: Lsn,
+    pub awaits_download: bool,
+}
+
+#[serde_as]
+#[derive(Debug, Serialize, Deserialize, Clone)]
+pub struct TimelineInfo {
+    #[serde_as(as = "DisplayFromStr")]
+    pub tenant_id: ZTenantId,
+    #[serde_as(as = "DisplayFromStr")]
+    pub timeline_id: ZTimelineId,
+    pub local: Option<LocalTimelineInfo>,
+    pub remote: Option<RemoteTimelineInfo>,
+}

pageserver/src/http/openapi_spec.yml

@@ -78,6 +78,11 @@ paths:
         schema:
           type: string
           description: Controls calculation of current_logical_size_non_incremental
+      - name: include-non-incremental-physical-size
+        in: query
+        schema:
+          type: string
+          description: Controls calculation of current_physical_size_non_incremental
     get:
       description: Get timelines for tenant
       responses:
@@ -136,6 +141,11 @@ paths:
           schema:
             type: string
           description: Controls calculation of current_logical_size_non_incremental
+        - name: include-non-incremental-physical-size
+          in: query
+          schema:
+            type: string
+            description: Controls calculation of current_physical_size_non_incremental
       responses:
         "200":
           description: TimelineInfo
@@ -197,54 +207,6 @@ paths:
               schema:
                 $ref: "#/components/schemas/Error"
 
-  /v1/tenant/{tenant_id}/timeline/{timeline_id}/wal_receiver:
-    parameters:
-      - name: tenant_id
-        in: path
-        required: true
-        schema:
-          type: string
-          format: hex
-      - name: timeline_id
-        in: path
-        required: true
-        schema:
-          type: string
-          format: hex
-    get:
-      description: Get wal receiver's data attached to the timeline
-      responses:
-        "200":
-          description: WalReceiverEntry
-          content:
-            application/json:
-              schema:
-                $ref: "#/components/schemas/WalReceiverEntry"
-        "401":
-          description: Unauthorized Error
-          content:
-            application/json:
-              schema:
-                $ref: "#/components/schemas/UnauthorizedError"
-        "403":
-          description: Forbidden Error
-          content:
-            application/json:
-              schema:
-                $ref: "#/components/schemas/ForbiddenError"
-        "404":
-          description: Error when no wal receiver is running or found
-          content:
-            application/json:
-              schema:
-                $ref: "#/components/schemas/NotFoundError"
-        "500":
-          description: Generic operation error
-          content:
-            application/json:
-              schema:
-                $ref: "#/components/schemas/Error"
-
   /v1/tenant/{tenant_id}/attach:
     parameters:
       - name: tenant_id
@@ -577,6 +539,8 @@ components:
           type: string
         state:
           type: string
+        current_physical_size:
+          type: integer
         has_in_progress_downloads:
           type: boolean
     TenantCreateInfo:
@@ -596,6 +560,8 @@ components:
           type: string
         checkpoint_distance:
           type: integer
+        checkpoint_timeout:
+          type: string
         compaction_period:
           type: string
         compaction_threshold:
@@ -614,6 +580,8 @@ components:
           type: string
         checkpoint_distance:
           type: integer
+        checkpoint_timeout:
+          type: string
         compaction_period:
           type: string
         compaction_threshold:
@@ -671,18 +639,13 @@ components:
           format: hex
         current_logical_size:
           type: integer
+        current_physical_size:
+          type: integer
         current_logical_size_non_incremental:
           type: integer
-
-    WalReceiverEntry:
-      type: object
-      required:
-        - thread_id
-        - wal_producer_connstr
-      properties:
-        thread_id:
+        current_physical_size_non_incremental:
           type: integer
-        wal_producer_connstr:
+        wal_source_connstr:
           type: string
         last_received_msg_lsn:
           type: string

pageserver/src/http/routes.rs

@@ -6,16 +6,19 @@ use hyper::{Body, Request, Response, Uri};
 use remote_storage::GenericRemoteStorage;
 use tracing::*;
 
+use super::models::{LocalTimelineInfo, RemoteTimelineInfo, TimelineInfo};
 use super::models::{
-    StatusResponse, TenantConfigRequest, TenantCreateRequest, TenantCreateResponse,
+    StatusResponse, TenantConfigRequest, TenantCreateRequest, TenantCreateResponse, TenantInfo,
     TimelineCreateRequest,
 };
-use crate::repository::Repository;
+use crate::layered_repository::metadata::TimelineMetadata;
+use crate::pgdatadir_mapping::DatadirTimeline;
+use crate::repository::{LocalTimelineState, RepositoryTimeline};
+use crate::repository::{Repository, Timeline};
 use crate::storage_sync;
 use crate::storage_sync::index::{RemoteIndex, RemoteTimeline};
 use crate::tenant_config::TenantConfOpt;
-use crate::tenant_mgr::TenantInfo;
-use crate::timelines::{LocalTimelineInfo, RemoteTimelineInfo, TimelineInfo};
+use crate::TimelineImpl;
 use crate::{config::PageServerConf, tenant_mgr, timelines};
 use utils::{
     auth::JwtAuth,
@@ -26,6 +29,7 @@ use utils::{
         request::parse_request_param,
         RequestExt, RouterBuilder,
     },
+    lsn::Lsn,
     zid::{ZTenantId, ZTenantTimelineId, ZTimelineId},
 };
 
@@ -79,6 +83,123 @@ fn get_config(request: &Request<Body>) -> &'static PageServerConf {
     get_state(request).conf
 }
 
+// Helper functions to construct a LocalTimelineInfo struct for a timeline
+
+fn local_timeline_info_from_loaded_timeline(
+    timeline: &TimelineImpl,
+    include_non_incremental_logical_size: bool,
+    include_non_incremental_physical_size: bool,
+) -> anyhow::Result<LocalTimelineInfo> {
+    let last_record_lsn = timeline.get_last_record_lsn();
+    let (wal_source_connstr, last_received_msg_lsn, last_received_msg_ts) = {
+        let guard = timeline.last_received_wal.lock().unwrap();
+        if let Some(info) = guard.as_ref() {
+            (
+                Some(info.wal_source_connstr.clone()),
+                Some(info.last_received_msg_lsn),
+                Some(info.last_received_msg_ts),
+            )
+        } else {
+            (None, None, None)
+        }
+    };
+
+    let info = LocalTimelineInfo {
+        ancestor_timeline_id: timeline.get_ancestor_timeline_id(),
+        ancestor_lsn: {
+            match timeline.get_ancestor_lsn() {
+                Lsn(0) => None,
+                lsn @ Lsn(_) => Some(lsn),
+            }
+        },
+        disk_consistent_lsn: timeline.get_disk_consistent_lsn(),
+        last_record_lsn,
+        prev_record_lsn: Some(timeline.get_prev_record_lsn()),
+        latest_gc_cutoff_lsn: *timeline.get_latest_gc_cutoff_lsn(),
+        timeline_state: LocalTimelineState::Loaded,
+        current_logical_size: Some(timeline.get_current_logical_size()),
+        current_physical_size: Some(timeline.get_physical_size()),
+        current_logical_size_non_incremental: if include_non_incremental_logical_size {
+            Some(timeline.get_current_logical_size_non_incremental(last_record_lsn)?)
+        } else {
+            None
+        },
+        current_physical_size_non_incremental: if include_non_incremental_physical_size {
+            Some(timeline.get_physical_size_non_incremental()?)
+        } else {
+            None
+        },
+        wal_source_connstr,
+        last_received_msg_lsn,
+        last_received_msg_ts,
+    };
+    Ok(info)
+}
+
+fn local_timeline_info_from_unloaded_timeline(metadata: &TimelineMetadata) -> LocalTimelineInfo {
+    LocalTimelineInfo {
+        ancestor_timeline_id: metadata.ancestor_timeline(),
+        ancestor_lsn: {
+            match metadata.ancestor_lsn() {
+                Lsn(0) => None,
+                lsn @ Lsn(_) => Some(lsn),
+            }
+        },
+        disk_consistent_lsn: metadata.disk_consistent_lsn(),
+        last_record_lsn: metadata.disk_consistent_lsn(),
+        prev_record_lsn: metadata.prev_record_lsn(),
+        latest_gc_cutoff_lsn: metadata.latest_gc_cutoff_lsn(),
+        timeline_state: LocalTimelineState::Unloaded,
+        current_logical_size: None,
+        current_physical_size: None,
+        current_logical_size_non_incremental: None,
+        current_physical_size_non_incremental: None,
+        wal_source_connstr: None,
+        last_received_msg_lsn: None,
+        last_received_msg_ts: None,
+    }
+}
+
+fn local_timeline_info_from_repo_timeline(
+    repo_timeline: &RepositoryTimeline<TimelineImpl>,
+    include_non_incremental_logical_size: bool,
+    include_non_incremental_physical_size: bool,
+) -> anyhow::Result<LocalTimelineInfo> {
+    match repo_timeline {
+        RepositoryTimeline::Loaded(timeline) => local_timeline_info_from_loaded_timeline(
+            &*timeline,
+            include_non_incremental_logical_size,
+            include_non_incremental_physical_size,
+        ),
+        RepositoryTimeline::Unloaded { metadata } => {
+            Ok(local_timeline_info_from_unloaded_timeline(metadata))
+        }
+    }
+}
+
+fn list_local_timelines(
+    tenant_id: ZTenantId,
+    include_non_incremental_logical_size: bool,
+    include_non_incremental_physical_size: bool,
+) -> Result<Vec<(ZTimelineId, LocalTimelineInfo)>> {
+    let repo = tenant_mgr::get_repository_for_tenant(tenant_id)
+        .with_context(|| format!("Failed to get repo for tenant {}", tenant_id))?;
+    let repo_timelines = repo.list_timelines();
+
+    let mut local_timeline_info = Vec::with_capacity(repo_timelines.len());
+    for (timeline_id, repository_timeline) in repo_timelines {
+        local_timeline_info.push((
+            timeline_id,
+            local_timeline_info_from_repo_timeline(
+                &repository_timeline,
+                include_non_incremental_logical_size,
+                include_non_incremental_physical_size,
+            )?,
+        ))
+    }
+    Ok(local_timeline_info)
+}
+
 // healthcheck handler
 async fn status_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
     let config = get_config(&request);
@@ -93,16 +214,30 @@ async fn timeline_create_handler(mut request: Request<Body>) -> Result<Response<
 
     let new_timeline_info = tokio::task::spawn_blocking(move || {
         let _enter = info_span!("/timeline_create", tenant = %tenant_id, new_timeline = ?request_data.new_timeline_id, lsn=?request_data.ancestor_start_lsn).entered();
-        timelines::create_timeline(
+
+        match timelines::create_timeline(
             get_config(&request),
             tenant_id,
             request_data.new_timeline_id.map(ZTimelineId::from),
             request_data.ancestor_timeline_id.map(ZTimelineId::from),
             request_data.ancestor_start_lsn,
-        )
+        ) {
+            Ok(Some((new_timeline_id, new_timeline))) => {
+                // Created. Construct a TimelineInfo for it.
+                let local_info = local_timeline_info_from_loaded_timeline(new_timeline.as_ref(), false, false)?;
+                Ok(Some(TimelineInfo {
+                    tenant_id,
+                    timeline_id: new_timeline_id,
+                    local: Some(local_info),
+                    remote: None,
+                }))
+            }
+            Ok(None) => Ok(None), // timeline already exists
+            Err(err) => Err(err),
+        }
     })
     .await
-    .map_err(ApiError::from_err)??;
+        .map_err(ApiError::from_err)??;
 
     Ok(match new_timeline_info {
         Some(info) => json_response(StatusCode::CREATED, info)?,
@@ -113,10 +248,17 @@ async fn timeline_create_handler(mut request: Request<Body>) -> Result<Response<
 async fn timeline_list_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
     let tenant_id: ZTenantId = parse_request_param(&request, "tenant_id")?;
     check_permission(&request, Some(tenant_id))?;
-    let include_non_incremental_logical_size = get_include_non_incremental_logical_size(&request);
+    let include_non_incremental_logical_size =
+        query_param_present(&request, "include-non-incremental-logical-size");
+    let include_non_incremental_physical_size =
+        query_param_present(&request, "include-non-incremental-physical-size");
     let local_timeline_infos = tokio::task::spawn_blocking(move || {
         let _enter = info_span!("timeline_list", tenant = %tenant_id).entered();
-        crate::timelines::get_local_timelines(tenant_id, include_non_incremental_logical_size)
+        list_local_timelines(
+            tenant_id,
+            include_non_incremental_logical_size,
+            include_non_incremental_physical_size,
+        )
     })
     .await
     .map_err(ApiError::from_err)??;
@@ -145,17 +287,15 @@ async fn timeline_list_handler(request: Request<Body>) -> Result<Response<Body>,
     json_response(StatusCode::OK, response_data)
 }
 
-// Gate non incremental logical size calculation behind a flag
-// after pgbench -i -s100 calculation took 28ms so if multiplied by the number of timelines
-// and tenants it can take noticeable amount of time. Also the value currently used only in tests
-fn get_include_non_incremental_logical_size(request: &Request<Body>) -> bool {
+/// Checks if a query param is present in the request's URL
+fn query_param_present(request: &Request<Body>, param: &str) -> bool {
     request
         .uri()
         .query()
         .map(|v| {
             url::form_urlencoded::parse(v.as_bytes())
                 .into_owned()
-                .any(|(param, _)| param == "include-non-incremental-logical-size")
+                .any(|(p, _)| p == param)
         })
         .unwrap_or(false)
 }
@@ -165,7 +305,10 @@ async fn timeline_detail_handler(request: Request<Body>) -> Result<Response<Body
     check_permission(&request, Some(tenant_id))?;
 
     let timeline_id: ZTimelineId = parse_request_param(&request, "timeline_id")?;
-    let include_non_incremental_logical_size = get_include_non_incremental_logical_size(&request);
+    let include_non_incremental_logical_size =
+        query_param_present(&request, "include-non-incremental-logical-size");
+    let include_non_incremental_physical_size =
+        query_param_present(&request, "include-non-incremental-physical-size");
 
     let (local_timeline_info, remote_timeline_info) = async {
         // any error here will render local timeline as None
@@ -176,11 +319,10 @@ async fn timeline_detail_handler(request: Request<Body>) -> Result<Response<Body
                 repo.get_timeline(timeline_id)
                     .as_ref()
                     .map(|timeline| {
-                        LocalTimelineInfo::from_repo_timeline(
-                            tenant_id,
-                            timeline_id,
+                        local_timeline_info_from_repo_timeline(
                             timeline,
                             include_non_incremental_logical_size,
+                            include_non_incremental_physical_size,
                         )
                     })
                     .transpose()?
@@ -225,23 +367,6 @@ async fn timeline_detail_handler(request: Request<Body>) -> Result<Response<Body
     json_response(StatusCode::OK, timeline_info)
 }
 
-async fn wal_receiver_get_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
-    let tenant_id: ZTenantId = parse_request_param(&request, "tenant_id")?;
-    check_permission(&request, Some(tenant_id))?;
-
-    let timeline_id: ZTimelineId = parse_request_param(&request, "timeline_id")?;
-    let wal_receiver_entry = crate::walreceiver::get_wal_receiver_entry(tenant_id, timeline_id)
-        .instrument(info_span!("wal_receiver_get", tenant = %tenant_id, timeline = %timeline_id))
-        .await
-        .ok_or_else(|| {
-            ApiError::NotFound(format!(
-                "WAL receiver data not found for tenant {tenant_id} and timeline {timeline_id}"
-            ))
-        })?;
-
-    json_response(StatusCode::OK, &wal_receiver_entry)
-}
-
 // TODO makes sense to provide tenant config right away the same way as it handled in tenant_create
 async fn tenant_attach_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
     let tenant_id: ZTenantId = parse_request_param(&request, "tenant_id")?;
@@ -429,14 +554,36 @@ async fn tenant_status(request: Request<Body>) -> Result<Response<Body>, ApiErro
     let index_accessor = remote_index.read().await;
     let has_in_progress_downloads = index_accessor
         .tenant_entry(&tenant_id)
-        .ok_or_else(|| ApiError::NotFound("Tenant not found in remote index".to_string()))?
-        .has_in_progress_downloads();
+        .map(|t| t.has_in_progress_downloads())
+        .unwrap_or_else(|| {
+            info!("Tenant {tenant_id} not found in remote index");
+            false
+        });
+
+    let current_physical_size =
+        match tokio::task::spawn_blocking(move || list_local_timelines(tenant_id, false, false))
+            .await
+            .map_err(ApiError::from_err)?
+        {
+            Err(err) => {
+                // Getting local timelines can fail when no local repo is on disk (e.g, when tenant data is being downloaded).
+                // In that case, put a warning message into log and operate normally.
+                warn!("Failed to get local timelines for tenant {tenant_id}: {err}");
+                None
+            }
+            Ok(local_timeline_infos) => Some(
+                local_timeline_infos
+                    .into_iter()
+                    .fold(0, |acc, x| acc + x.1.current_physical_size.unwrap()),
+            ),
+        };
 
     json_response(
         StatusCode::OK,
         TenantInfo {
             id: tenant_id,
             state: tenant_state,
+            current_physical_size,
             has_in_progress_downloads: Some(has_in_progress_downloads),
         },
     )
@@ -476,6 +623,11 @@ async fn tenant_create_handler(mut request: Request<Body>) -> Result<Response<Bo
     }
 
     tenant_conf.checkpoint_distance = request_data.checkpoint_distance;
+    if let Some(checkpoint_timeout) = request_data.checkpoint_timeout {
+        tenant_conf.checkpoint_timeout =
+            Some(humantime::parse_duration(&checkpoint_timeout).map_err(ApiError::from_err)?);
+    }
+
     tenant_conf.compaction_target_size = request_data.compaction_target_size;
     tenant_conf.compaction_threshold = request_data.compaction_threshold;
 
@@ -536,6 +688,10 @@ async fn tenant_config_handler(mut request: Request<Body>) -> Result<Response<Bo
     }
 
     tenant_conf.checkpoint_distance = request_data.checkpoint_distance;
+    if let Some(checkpoint_timeout) = request_data.checkpoint_timeout {
+        tenant_conf.checkpoint_timeout =
+            Some(humantime::parse_duration(&checkpoint_timeout).map_err(ApiError::from_err)?);
+    }
     tenant_conf.compaction_target_size = request_data.compaction_target_size;
     tenant_conf.compaction_threshold = request_data.compaction_threshold;
 
@@ -606,9 +762,5 @@ pub fn make_router(
             "/v1/tenant/:tenant_id/timeline/:timeline_id/detach",
             timeline_delete_handler,
         )
-        .get(
-            "/v1/tenant/:tenant_id/timeline/:timeline_id/wal_receiver",
-            wal_receiver_get_handler,
-        )
         .any(handler_404))
 }

pageserver/src/import_datadir.rs

@@ -13,9 +13,8 @@ use walkdir::WalkDir;
 
 use crate::pgdatadir_mapping::*;
 use crate::reltag::{RelTag, SlruKind};
-use crate::repository::Repository;
-use crate::repository::Timeline;
 use crate::walingest::WalIngest;
+use crate::walrecord::DecodedWALRecord;
 use postgres_ffi::relfile_utils::*;
 use postgres_ffi::waldecoder::*;
 use postgres_ffi::xlog_utils::*;
@@ -29,9 +28,9 @@ use utils::lsn::Lsn;
 /// This is currently only used to import a cluster freshly created by initdb.
 /// The code that deals with the checkpoint would not work right if the
 /// cluster was not shut down cleanly.
-pub fn import_timeline_from_postgres_datadir<R: Repository>(
+pub fn import_timeline_from_postgres_datadir<T: DatadirTimeline>(
     path: &Path,
-    tline: &mut DatadirTimeline<R>,
+    tline: &T,
     lsn: Lsn,
 ) -> Result<()> {
     let mut pg_control: Option<ControlFileData> = None;
@@ -89,8 +88,8 @@ pub fn import_timeline_from_postgres_datadir<R: Repository>(
 }
 
 // subroutine of import_timeline_from_postgres_datadir(), to load one relation file.
-fn import_rel<R: Repository, Reader: Read>(
-    modification: &mut DatadirModification<R>,
+fn import_rel<T: DatadirTimeline, Reader: Read>(
+    modification: &mut DatadirModification<T>,
     path: &Path,
     spcoid: Oid,
     dboid: Oid,
@@ -169,8 +168,8 @@ fn import_rel<R: Repository, Reader: Read>(
 
 /// Import an SLRU segment file
 ///
-fn import_slru<R: Repository, Reader: Read>(
-    modification: &mut DatadirModification<R>,
+fn import_slru<T: DatadirTimeline, Reader: Read>(
+    modification: &mut DatadirModification<T>,
     slru: SlruKind,
     path: &Path,
     mut reader: Reader,
@@ -225,9 +224,9 @@ fn import_slru<R: Repository, Reader: Read>(
 
 /// Scan PostgreSQL WAL files in given directory and load all records between
 /// 'startpoint' and 'endpoint' into the repository.
-fn import_wal<R: Repository>(
+fn import_wal<T: DatadirTimeline>(
     walpath: &Path,
-    tline: &mut DatadirTimeline<R>,
+    tline: &T,
     startpoint: Lsn,
     endpoint: Lsn,
 ) -> Result<()> {
@@ -268,9 +267,11 @@ fn import_wal<R: Repository>(
         waldecoder.feed_bytes(&buf);
 
         let mut nrecords = 0;
+        let mut modification = tline.begin_modification(endpoint);
+        let mut decoded = DecodedWALRecord::default();
         while last_lsn <= endpoint {
             if let Some((lsn, recdata)) = waldecoder.poll_decode()? {
-                walingest.ingest_record(tline, recdata, lsn)?;
+                walingest.ingest_record(recdata, lsn, &mut modification, &mut decoded)?;
                 last_lsn = lsn;
 
                 nrecords += 1;
@@ -294,8 +295,8 @@ fn import_wal<R: Repository>(
     Ok(())
 }
 
-pub fn import_basebackup_from_tar<R: Repository, Reader: Read>(
-    tline: &mut DatadirTimeline<R>,
+pub fn import_basebackup_from_tar<T: DatadirTimeline, Reader: Read>(
+    tline: &T,
     reader: Reader,
     base_lsn: Lsn,
 ) -> Result<()> {
@@ -336,8 +337,8 @@ pub fn import_basebackup_from_tar<R: Repository, Reader: Read>(
     Ok(())
 }
 
-pub fn import_wal_from_tar<R: Repository, Reader: Read>(
-    tline: &mut DatadirTimeline<R>,
+pub fn import_wal_from_tar<T: DatadirTimeline, Reader: Read>(
+    tline: &T,
     reader: Reader,
     start_lsn: Lsn,
     end_lsn: Lsn,
@@ -384,9 +385,11 @@ pub fn import_wal_from_tar<R: Repository, Reader: Read>(
 
         waldecoder.feed_bytes(&bytes[offset..]);
 
+        let mut modification = tline.begin_modification(end_lsn);
+        let mut decoded = DecodedWALRecord::default();
         while last_lsn <= end_lsn {
             if let Some((lsn, recdata)) = waldecoder.poll_decode()? {
-                walingest.ingest_record(tline, recdata, lsn)?;
+                walingest.ingest_record(recdata, lsn, &mut modification, &mut decoded)?;
                 last_lsn = lsn;
 
                 debug!("imported record at {} (end {})", lsn, end_lsn);
@@ -415,8 +418,8 @@ pub fn import_wal_from_tar<R: Repository, Reader: Read>(
     Ok(())
 }
 
-pub fn import_file<R: Repository, Reader: Read>(
-    modification: &mut DatadirModification<R>,
+pub fn import_file<T: DatadirTimeline, Reader: Read>(
+    modification: &mut DatadirModification<T>,
     file_path: &Path,
     reader: Reader,
     len: usize,
@@ -535,7 +538,7 @@ pub fn import_file<R: Repository, Reader: Read>(
         // zenith.signal is not necessarily the last file, that we handle
         // but it is ok to call `finish_write()`, because final `modification.commit()`
         // will update lsn once more to the final one.
-        let writer = modification.tline.tline.writer();
+        let writer = modification.tline.writer();
         writer.finish_write(prev_lsn);
 
         debug!("imported zenith signal {}", prev_lsn);

pageserver/src/layered_repository/block_io.rs

@@ -5,7 +5,7 @@
 use crate::page_cache;
 use crate::page_cache::{ReadBufResult, PAGE_SZ};
 use bytes::Bytes;
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
 use std::ops::{Deref, DerefMut};
 use std::os::unix::fs::FileExt;
 use std::sync::atomic::AtomicU64;
@@ -117,9 +117,7 @@ where
     }
 }
 
-lazy_static! {
-    static ref NEXT_ID: AtomicU64 = AtomicU64::new(1);
-}
+static NEXT_ID: Lazy<AtomicU64> = Lazy::new(|| AtomicU64::new(1));
 
 /// An adapter for reading a (virtual) file using the page cache.
 ///

pageserver/src/layered_repository/delta_layer.rs

@@ -316,6 +316,18 @@ impl Layer for DeltaLayer {
         }
     }
 
+    fn key_iter<'a>(&'a self) -> Box<dyn Iterator<Item = (Key, Lsn, u64)> + 'a> {
+        let inner = match self.load() {
+            Ok(inner) => inner,
+            Err(e) => panic!("Failed to load a delta layer: {e:?}"),
+        };
+
+        match DeltaKeyIter::new(inner) {
+            Ok(iter) => Box::new(iter),
+            Err(e) => panic!("Layer index is corrupted: {e:?}"),
+        }
+    }
+
     fn delete(&self) -> Result<()> {
         // delete underlying file
         fs::remove_file(self.path())?;
@@ -660,11 +672,21 @@ impl DeltaLayerWriter {
     /// The values must be appended in key, lsn order.
     ///
     pub fn put_value(&mut self, key: Key, lsn: Lsn, val: Value) -> Result<()> {
+        self.put_value_bytes(key, lsn, &Value::ser(&val)?, val.will_init())
+    }
+
+    pub fn put_value_bytes(
+        &mut self,
+        key: Key,
+        lsn: Lsn,
+        val: &[u8],
+        will_init: bool,
+    ) -> Result<()> {
         assert!(self.lsn_range.start <= lsn);
 
-        let off = self.blob_writer.write_blob(&Value::ser(&val)?)?;
+        let off = self.blob_writer.write_blob(val)?;
 
-        let blob_ref = BlobRef::new(off, val.will_init());
+        let blob_ref = BlobRef::new(off, will_init);
 
         let delta_key = DeltaKey::from_key_lsn(&key, lsn);
         self.tree.append(&delta_key.0, blob_ref.0)?;
@@ -822,3 +844,75 @@ impl<'a> DeltaValueIter<'a> {
         }
     }
 }
+///
+/// Iterator over all keys stored in a delta layer
+///
+/// FIXME: This creates a Vector to hold all keys.
+/// That takes up quite a lot of memory. Should do this in a more streaming
+/// fashion.
+///
+struct DeltaKeyIter {
+    all_keys: Vec<(DeltaKey, u64)>,
+    next_idx: usize,
+}
+
+impl Iterator for DeltaKeyIter {
+    type Item = (Key, Lsn, u64);
+
+    fn next(&mut self) -> Option<Self::Item> {
+        if self.next_idx < self.all_keys.len() {
+            let (delta_key, size) = &self.all_keys[self.next_idx];
+
+            let key = delta_key.key();
+            let lsn = delta_key.lsn();
+
+            self.next_idx += 1;
+            Some((key, lsn, *size))
+        } else {
+            None
+        }
+    }
+}
+
+impl<'a> DeltaKeyIter {
+    fn new(inner: RwLockReadGuard<'a, DeltaLayerInner>) -> Result<Self> {
+        let file = inner.file.as_ref().unwrap();
+        let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
+            inner.index_start_blk,
+            inner.index_root_blk,
+            file,
+        );
+
+        let mut all_keys: Vec<(DeltaKey, u64)> = Vec::new();
+        tree_reader.visit(
+            &[0u8; DELTA_KEY_SIZE],
+            VisitDirection::Forwards,
+            |key, value| {
+                let delta_key = DeltaKey::from_slice(key);
+                let pos = BlobRef(value).pos();
+                if let Some(last) = all_keys.last_mut() {
+                    if last.0.key() == delta_key.key() {
+                        return true;
+                    } else {
+                        // subtract offset of new key BLOB and first blob of this key
+                        // to get total size if values associated with this key
+                        let first_pos = last.1;
+                        last.1 = pos - first_pos;
+                    }
+                }
+                all_keys.push((delta_key, pos));
+                true
+            },
+        )?;
+        if let Some(last) = all_keys.last_mut() {
+            // Last key occupies all space till end of layer
+            last.1 = std::fs::metadata(&file.file.path)?.len() - last.1;
+        }
+        let iter = DeltaKeyIter {
+            all_keys,
+            next_idx: 0,
+        };
+
+        Ok(iter)
+    }
+}

pageserver/src/layered_repository/ephemeral_file.rs

@@ -8,7 +8,7 @@ use crate::page_cache;
 use crate::page_cache::PAGE_SZ;
 use crate::page_cache::{ReadBufResult, WriteBufResult};
 use crate::virtual_file::VirtualFile;
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
 use std::cmp::min;
 use std::collections::HashMap;
 use std::fs::OpenOptions;
@@ -21,15 +21,15 @@ use utils::zid::{ZTenantId, ZTimelineId};
 
 use std::os::unix::fs::FileExt;
 
-lazy_static! {
-    ///
-    /// This is the global cache of file descriptors (File objects).
-    ///
-    static ref EPHEMERAL_FILES: RwLock<EphemeralFiles> = RwLock::new(EphemeralFiles {
+///
+/// This is the global cache of file descriptors (File objects).
+///
+static EPHEMERAL_FILES: Lazy<RwLock<EphemeralFiles>> = Lazy::new(|| {
+    RwLock::new(EphemeralFiles {
         next_file_id: 1,
         files: HashMap::new(),
-    });
-}
+    })
+});
 
 pub struct EphemeralFiles {
     next_file_id: u64,
@@ -43,7 +43,7 @@ pub struct EphemeralFile {
     _timelineid: ZTimelineId,
     file: Arc<VirtualFile>,
 
-    size: u64,
+    pub size: u64,
 }
 
 impl EphemeralFile {

pageserver/src/layered_repository/inmemory_layer.rs

@@ -15,6 +15,7 @@ use crate::layered_repository::storage_layer::{
 use crate::repository::{Key, Value};
 use crate::walrecord;
 use anyhow::{bail, ensure, Result};
+use std::cell::RefCell;
 use std::collections::HashMap;
 use tracing::*;
 use utils::{
@@ -30,6 +31,12 @@ use std::ops::Range;
 use std::path::PathBuf;
 use std::sync::RwLock;
 
+thread_local! {
+    /// A buffer for serializing object during [`InMemoryLayer::put_value`].
+    /// This buffer is reused for each serialization to avoid additional malloc calls.
+    static SER_BUFFER: RefCell<Vec<u8>> = RefCell::new(Vec::new());
+}
+
 pub struct InMemoryLayer {
     conf: &'static PageServerConf,
     tenantid: ZTenantId,
@@ -233,6 +240,14 @@ impl Layer for InMemoryLayer {
 }
 
 impl InMemoryLayer {
+    ///
+    /// Get layer size on the disk
+    ///
+    pub fn size(&self) -> Result<u64> {
+        let inner = self.inner.read().unwrap();
+        Ok(inner.file.size)
+    }
+
     ///
     /// Create a new, empty, in-memory layer
     ///
@@ -270,10 +285,17 @@ impl InMemoryLayer {
     pub fn put_value(&self, key: Key, lsn: Lsn, val: &Value) -> Result<()> {
         trace!("put_value key {} at {}/{}", key, self.timelineid, lsn);
         let mut inner = self.inner.write().unwrap();
-
         inner.assert_writeable();
 
-        let off = inner.file.write_blob(&Value::ser(val)?)?;
+        let off = {
+            SER_BUFFER.with(|x| -> Result<_> {
+                let mut buf = x.borrow_mut();
+                buf.clear();
+                val.ser_into(&mut (*buf))?;
+                let off = inner.file.write_blob(&buf)?;
+                Ok(off)
+            })?
+        };
 
         let vec_map = inner.index.entry(key).or_default();
         let old = vec_map.append_or_update_last(lsn, off).unwrap().0;
@@ -342,8 +364,8 @@ impl InMemoryLayer {
             // Write all page versions
             for (lsn, pos) in vec_map.as_slice() {
                 cursor.read_blob_into_buf(*pos, &mut buf)?;
-                let val = Value::des(&buf)?;
-                delta_layer_writer.put_value(key, *lsn, val)?;
+                let will_init = Value::des(&buf)?.will_init();
+                delta_layer_writer.put_value_bytes(key, *lsn, &buf, will_init)?;
             }
         }
 

pageserver/src/layered_repository/layer_map.rs

@@ -10,24 +10,23 @@
 //! corresponding files are written to disk.
 //!
 
+use crate::layered_repository::inmemory_layer::InMemoryLayer;
 use crate::layered_repository::storage_layer::Layer;
 use crate::layered_repository::storage_layer::{range_eq, range_overlaps};
-use crate::layered_repository::InMemoryLayer;
 use crate::repository::Key;
 use anyhow::Result;
-use lazy_static::lazy_static;
 use metrics::{register_int_gauge, IntGauge};
+use once_cell::sync::Lazy;
 use std::collections::VecDeque;
 use std::ops::Range;
 use std::sync::Arc;
 use tracing::*;
 use utils::lsn::Lsn;
 
-lazy_static! {
-    static ref NUM_ONDISK_LAYERS: IntGauge =
-        register_int_gauge!("pageserver_ondisk_layers", "Number of layers on-disk")
-            .expect("failed to define a metric");
-}
+static NUM_ONDISK_LAYERS: Lazy<IntGauge> = Lazy::new(|| {
+    register_int_gauge!("pageserver_ondisk_layers", "Number of layers on-disk")
+        .expect("failed to define a metric")
+});
 
 ///
 /// LayerMap tracks what layers exist on a timeline.

pageserver/src/layered_repository/storage_layer.rs

@@ -139,6 +139,12 @@ pub trait Layer: Send + Sync {
     /// Iterate through all keys and values stored in the layer
     fn iter(&self) -> Box<dyn Iterator<Item = Result<(Key, Lsn, Value)>> + '_>;
 
+    /// Iterate through all keys stored in the layer. Returns key, lsn and value size
+    /// It is used only for compaction and so is currently implemented only for DeltaLayer
+    fn key_iter(&self) -> Box<dyn Iterator<Item = (Key, Lsn, u64)> + '_> {
+        panic!("Not implemented")
+    }
+
     /// Permanently remove this layer from disk.
     fn delete(&self) -> Result<()>;
 

pageserver/src/lib.rs

@@ -22,7 +22,7 @@ pub mod walreceiver;
 pub mod walrecord;
 pub mod walredo;
 
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
 use tracing::info;
 
 use crate::thread_mgr::ThreadKind;
@@ -42,14 +42,14 @@ pub const STORAGE_FORMAT_VERSION: u16 = 3;
 pub const IMAGE_FILE_MAGIC: u16 = 0x5A60;
 pub const DELTA_FILE_MAGIC: u16 = 0x5A61;
 
-lazy_static! {
-    static ref LIVE_CONNECTIONS_COUNT: IntGaugeVec = register_int_gauge_vec!(
+static LIVE_CONNECTIONS_COUNT: Lazy<IntGaugeVec> = Lazy::new(|| {
+    register_int_gauge_vec!(
         "pageserver_live_connections",
         "Number of live network connections",
         &["pageserver_connection_kind"]
     )
-    .expect("failed to define a metric");
-}
+    .expect("failed to define a metric")
+});
 
 pub const LOG_FILE_NAME: &str = "pageserver.log";
 
@@ -63,8 +63,7 @@ pub enum CheckpointConfig {
 }
 
 pub type RepositoryImpl = LayeredRepository;
-
-pub type DatadirTimelineImpl = DatadirTimeline<RepositoryImpl>;
+pub type TimelineImpl = <LayeredRepository as repository::Repository>::Timeline;
 
 pub fn shutdown_pageserver(exit_code: i32) {
     // Shut down the libpq endpoint thread. This prevents new connections from
@@ -94,3 +93,56 @@ pub fn shutdown_pageserver(exit_code: i32) {
     info!("Shut down successfully completed");
     std::process::exit(exit_code);
 }
+
+const DEFAULT_BASE_BACKOFF_SECONDS: f64 = 0.1;
+const DEFAULT_MAX_BACKOFF_SECONDS: f64 = 3.0;
+
+async fn exponential_backoff(n: u32, base_increment: f64, max_seconds: f64) {
+    let backoff_duration_seconds =
+        exponential_backoff_duration_seconds(n, base_increment, max_seconds);
+    if backoff_duration_seconds > 0.0 {
+        info!(
+            "Backoff: waiting {backoff_duration_seconds} seconds before processing with the task",
+        );
+        tokio::time::sleep(std::time::Duration::from_secs_f64(backoff_duration_seconds)).await;
+    }
+}
+
+fn exponential_backoff_duration_seconds(n: u32, base_increment: f64, max_seconds: f64) -> f64 {
+    if n == 0 {
+        0.0
+    } else {
+        (1.0 + base_increment).powf(f64::from(n)).min(max_seconds)
+    }
+}
+
+#[cfg(test)]
+mod backoff_defaults_tests {
+    use super::*;
+
+    #[test]
+    fn backoff_defaults_produce_growing_backoff_sequence() {
+        let mut current_backoff_value = None;
+
+        for i in 0..10_000 {
+            let new_backoff_value = exponential_backoff_duration_seconds(
+                i,
+                DEFAULT_BASE_BACKOFF_SECONDS,
+                DEFAULT_MAX_BACKOFF_SECONDS,
+            );
+
+            if let Some(old_backoff_value) = current_backoff_value.replace(new_backoff_value) {
+                assert!(
+                    old_backoff_value <= new_backoff_value,
+                    "{i}th backoff value {new_backoff_value} is smaller than the previous one {old_backoff_value}"
+                )
+            }
+        }
+
+        assert_eq!(
+            current_backoff_value.expect("Should have produced backoff values to compare"),
+            DEFAULT_MAX_BACKOFF_SECONDS,
+            "Given big enough of retries, backoff should reach its allowed max value"
+        );
+    }
+}

pageserver/src/page_service.rs

@@ -11,7 +11,7 @@
 
 use anyhow::{bail, ensure, Context, Result};
 use bytes::{Buf, BufMut, Bytes, BytesMut};
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
 use regex::Regex;
 use std::io::{self, Read};
 use std::net::TcpListener;
@@ -30,7 +30,6 @@ use utils::{
 use crate::basebackup;
 use crate::config::{PageServerConf, ProfilingConfig};
 use crate::import_datadir::{import_basebackup_from_tar, import_wal_from_tar};
-use crate::layered_repository::LayeredRepository;
 use crate::pgdatadir_mapping::{DatadirTimeline, LsnForTimestamp};
 use crate::profiling::profpoint_start;
 use crate::reltag::RelTag;
@@ -435,15 +434,15 @@ const TIME_BUCKETS: &[f64] = &[
     0.1,  // 1/10 s
 ];
 
-lazy_static! {
-    static ref SMGR_QUERY_TIME: HistogramVec = register_histogram_vec!(
+static SMGR_QUERY_TIME: Lazy<HistogramVec> = Lazy::new(|| {
+    register_histogram_vec!(
         "pageserver_smgr_query_seconds",
         "Time spent on smgr query handling",
         &["smgr_query_type", "tenant_id", "timeline_id"],
         TIME_BUCKETS.into()
     )
-    .expect("failed to define a metric");
-}
+    .expect("failed to define a metric")
+});
 
 impl PageServerHandler {
     pub fn new(conf: &'static PageServerConf, auth: Option<Arc<JwtAuth>>) -> Self {
@@ -555,9 +554,6 @@ impl PageServerHandler {
         info!("creating new timeline");
         let repo = tenant_mgr::get_repository_for_tenant(tenant_id)?;
         let timeline = repo.create_empty_timeline(timeline_id, base_lsn)?;
-        let repartition_distance = repo.get_checkpoint_distance();
-        let mut datadir_timeline =
-            DatadirTimeline::<LayeredRepository>::new(timeline, repartition_distance);
 
         // TODO mark timeline as not ready until it reaches end_lsn.
         // We might have some wal to import as well, and we should prevent compute
@@ -573,7 +569,7 @@ impl PageServerHandler {
         info!("importing basebackup");
         pgb.write_message(&BeMessage::CopyInResponse)?;
         let reader = CopyInReader::new(pgb);
-        import_basebackup_from_tar(&mut datadir_timeline, reader, base_lsn)?;
+        import_basebackup_from_tar(&*timeline, reader, base_lsn)?;
 
         // TODO check checksum
         // Meanwhile you can verify client-side by taking fullbackup
@@ -583,7 +579,7 @@ impl PageServerHandler {
 
         // Flush data to disk, then upload to s3
         info!("flushing layers");
-        datadir_timeline.tline.checkpoint(CheckpointConfig::Flush)?;
+        timeline.checkpoint(CheckpointConfig::Flush)?;
 
         info!("done");
         Ok(())
@@ -605,10 +601,6 @@ impl PageServerHandler {
         let timeline = repo.get_timeline_load(timeline_id)?;
         ensure!(timeline.get_last_record_lsn() == start_lsn);
 
-        let repartition_distance = repo.get_checkpoint_distance();
-        let mut datadir_timeline =
-            DatadirTimeline::<LayeredRepository>::new(timeline, repartition_distance);
-
         // TODO leave clean state on error. For now you can use detach to clean
         // up broken state from a failed import.
 
@@ -616,16 +608,16 @@ impl PageServerHandler {
         info!("importing wal");
         pgb.write_message(&BeMessage::CopyInResponse)?;
         let reader = CopyInReader::new(pgb);
-        import_wal_from_tar(&mut datadir_timeline, reader, start_lsn, end_lsn)?;
+        import_wal_from_tar(&*timeline, reader, start_lsn, end_lsn)?;
 
         // TODO Does it make sense to overshoot?
-        ensure!(datadir_timeline.tline.get_last_record_lsn() >= end_lsn);
+        ensure!(timeline.get_last_record_lsn() >= end_lsn);
 
         // Flush data to disk, then upload to s3. No need for a forced checkpoint.
         // We only want to persist the data, and it doesn't matter if it's in the
         // shape of deltas or images.
         info!("flushing layers");
-        datadir_timeline.tline.checkpoint(CheckpointConfig::Flush)?;
+        timeline.checkpoint(CheckpointConfig::Flush)?;
 
         info!("done");
         Ok(())
@@ -643,8 +635,8 @@ impl PageServerHandler {
     /// In either case, if the page server hasn't received the WAL up to the
     /// requested LSN yet, we will wait for it to arrive. The return value is
     /// the LSN that should be used to look up the page versions.
-    fn wait_or_get_last_lsn<R: Repository>(
-        timeline: &DatadirTimeline<R>,
+    fn wait_or_get_last_lsn<T: DatadirTimeline>(
+        timeline: &T,
         mut lsn: Lsn,
         latest: bool,
         latest_gc_cutoff_lsn: &RwLockReadGuard<Lsn>,
@@ -671,7 +663,7 @@ impl PageServerHandler {
             if lsn <= last_record_lsn {
                 lsn = last_record_lsn;
             } else {
-                timeline.tline.wait_lsn(lsn)?;
+                timeline.wait_lsn(lsn)?;
                 // Since we waited for 'lsn' to arrive, that is now the last
                 // record LSN. (Or close enough for our purposes; the
                 // last-record LSN can advance immediately after we return
@@ -681,7 +673,7 @@ impl PageServerHandler {
             if lsn == Lsn(0) {
                 bail!("invalid LSN(0) in request");
             }
-            timeline.tline.wait_lsn(lsn)?;
+            timeline.wait_lsn(lsn)?;
         }
         ensure!(
             lsn >= **latest_gc_cutoff_lsn,
@@ -691,14 +683,14 @@ impl PageServerHandler {
         Ok(lsn)
     }
 
-    fn handle_get_rel_exists_request<R: Repository>(
+    fn handle_get_rel_exists_request<T: DatadirTimeline>(
         &self,
-        timeline: &DatadirTimeline<R>,
+        timeline: &T,
         req: &PagestreamExistsRequest,
     ) -> Result<PagestreamBeMessage> {
         let _enter = info_span!("get_rel_exists", rel = %req.rel, req_lsn = %req.lsn).entered();
 
-        let latest_gc_cutoff_lsn = timeline.tline.get_latest_gc_cutoff_lsn();
+        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest, &latest_gc_cutoff_lsn)?;
 
         let exists = timeline.get_rel_exists(req.rel, lsn)?;
@@ -708,13 +700,13 @@ impl PageServerHandler {
         }))
     }
 
-    fn handle_get_nblocks_request<R: Repository>(
+    fn handle_get_nblocks_request<T: DatadirTimeline>(
         &self,
-        timeline: &DatadirTimeline<R>,
+        timeline: &T,
         req: &PagestreamNblocksRequest,
     ) -> Result<PagestreamBeMessage> {
         let _enter = info_span!("get_nblocks", rel = %req.rel, req_lsn = %req.lsn).entered();
-        let latest_gc_cutoff_lsn = timeline.tline.get_latest_gc_cutoff_lsn();
+        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest, &latest_gc_cutoff_lsn)?;
 
         let n_blocks = timeline.get_rel_size(req.rel, lsn)?;
@@ -724,13 +716,13 @@ impl PageServerHandler {
         }))
     }
 
-    fn handle_db_size_request<R: Repository>(
+    fn handle_db_size_request<T: DatadirTimeline>(
         &self,
-        timeline: &DatadirTimeline<R>,
+        timeline: &T,
         req: &PagestreamDbSizeRequest,
     ) -> Result<PagestreamBeMessage> {
         let _enter = info_span!("get_db_size", dbnode = %req.dbnode, req_lsn = %req.lsn).entered();
-        let latest_gc_cutoff_lsn = timeline.tline.get_latest_gc_cutoff_lsn();
+        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest, &latest_gc_cutoff_lsn)?;
 
         let total_blocks =
@@ -743,14 +735,14 @@ impl PageServerHandler {
         }))
     }
 
-    fn handle_get_page_at_lsn_request<R: Repository>(
+    fn handle_get_page_at_lsn_request<T: DatadirTimeline>(
         &self,
-        timeline: &DatadirTimeline<R>,
+        timeline: &T,
         req: &PagestreamGetPageRequest,
     ) -> Result<PagestreamBeMessage> {
         let _enter = info_span!("get_page", rel = %req.rel, blkno = &req.blkno, req_lsn = %req.lsn)
             .entered();
-        let latest_gc_cutoff_lsn = timeline.tline.get_latest_gc_cutoff_lsn();
+        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest, &latest_gc_cutoff_lsn)?;
         /*
         // Add a 1s delay to some requests. The delayed causes the requests to
@@ -783,7 +775,7 @@ impl PageServerHandler {
         // check that the timeline exists
         let timeline = tenant_mgr::get_local_timeline_with_load(tenantid, timelineid)
             .context("Cannot load local timeline")?;
-        let latest_gc_cutoff_lsn = timeline.tline.get_latest_gc_cutoff_lsn();
+        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         if let Some(lsn) = lsn {
             timeline
                 .check_lsn_is_in_scope(lsn, &latest_gc_cutoff_lsn)
@@ -921,7 +913,7 @@ impl postgres_backend::Handler for PageServerHandler {
             let timeline = tenant_mgr::get_local_timeline_with_load(tenantid, timelineid)
                 .context("Cannot load local timeline")?;
 
-            let end_of_timeline = timeline.tline.get_last_record_rlsn();
+            let end_of_timeline = timeline.get_last_record_rlsn();
 
             pgb.write_message_noflush(&BeMessage::RowDescription(&[
                 RowDescriptor::text_col(b"prev_lsn"),
@@ -1052,6 +1044,7 @@ impl postgres_backend::Handler for PageServerHandler {
             let repo = tenant_mgr::get_repository_for_tenant(tenantid)?;
             pgb.write_message_noflush(&BeMessage::RowDescription(&[
                 RowDescriptor::int8_col(b"checkpoint_distance"),
+                RowDescriptor::int8_col(b"checkpoint_timeout"),
                 RowDescriptor::int8_col(b"compaction_target_size"),
                 RowDescriptor::int8_col(b"compaction_period"),
                 RowDescriptor::int8_col(b"compaction_threshold"),
@@ -1062,6 +1055,12 @@ impl postgres_backend::Handler for PageServerHandler {
             ]))?
             .write_message_noflush(&BeMessage::DataRow(&[
                 Some(repo.get_checkpoint_distance().to_string().as_bytes()),
+                Some(
+                    repo.get_checkpoint_timeout()
+                        .as_secs()
+                        .to_string()
+                        .as_bytes(),
+                ),
                 Some(repo.get_compaction_target_size().to_string().as_bytes()),
                 Some(
                     repo.get_compaction_period()
@@ -1139,7 +1138,7 @@ impl postgres_backend::Handler for PageServerHandler {
             let timelineid = ZTimelineId::from_str(caps.get(2).unwrap().as_str())?;
             let timeline = tenant_mgr::get_local_timeline_with_load(tenantid, timelineid)
                 .context("Couldn't load timeline")?;
-            timeline.tline.compact()?;
+            timeline.compact()?;
 
             pgb.write_message_noflush(&SINGLE_COL_ROWDESC)?
                 .write_message_noflush(&BeMessage::CommandComplete(b"SELECT 1"))?;
@@ -1159,13 +1158,8 @@ impl postgres_backend::Handler for PageServerHandler {
             let timeline = tenant_mgr::get_local_timeline_with_load(tenantid, timelineid)
                 .context("Cannot load local timeline")?;
 
-            timeline.tline.checkpoint(CheckpointConfig::Forced)?;
-
-            // Also compact it.
-            //
-            // FIXME: This probably shouldn't be part of a "checkpoint" command, but a
-            // separate operation. Update the tests if you change this.
-            timeline.tline.compact()?;
+            // Checkpoint the timeline and also compact it (due to `CheckpointConfig::Forced`).
+            timeline.checkpoint(CheckpointConfig::Forced)?;
 
             pgb.write_message_noflush(&SINGLE_COL_ROWDESC)?
                 .write_message_noflush(&BeMessage::CommandComplete(b"SELECT 1"))?;

pageserver/src/pgdatadir_mapping.rs

@@ -6,10 +6,10 @@
 //! walingest.rs handles a few things like implicit relation creation and extension.
 //! Clarify that)
 //!
-use crate::keyspace::{KeyPartitioning, KeySpace, KeySpaceAccum};
+use crate::keyspace::{KeySpace, KeySpaceAccum};
 use crate::reltag::{RelTag, SlruKind};
+use crate::repository::Timeline;
 use crate::repository::*;
-use crate::repository::{Repository, Timeline};
 use crate::walrecord::ZenithWalRecord;
 use anyhow::{bail, ensure, Result};
 use bytes::{Buf, Bytes};
@@ -18,34 +18,12 @@ use postgres_ffi::{pg_constants, Oid, TransactionId};
 use serde::{Deserialize, Serialize};
 use std::collections::{HashMap, HashSet};
 use std::ops::Range;
-use std::sync::atomic::{AtomicIsize, Ordering};
-use std::sync::{Arc, Mutex, RwLockReadGuard};
-use tracing::{debug, error, trace, warn};
+use tracing::{debug, trace, warn};
 use utils::{bin_ser::BeSer, lsn::Lsn};
 
 /// Block number within a relation or SLRU. This matches PostgreSQL's BlockNumber type.
 pub type BlockNumber = u32;
 
-pub struct DatadirTimeline<R>
-where
-    R: Repository,
-{
-    /// The underlying key-value store. Callers should not read or modify the
-    /// data in the underlying store directly. However, it is exposed to have
-    /// access to information like last-LSN, ancestor, and operations like
-    /// compaction.
-    pub tline: Arc<R::Timeline>,
-
-    /// When did we last calculate the partitioning?
-    partitioning: Mutex<(KeyPartitioning, Lsn)>,
-
-    /// Configuration: how often should the partitioning be recalculated.
-    repartition_threshold: u64,
-
-    /// Current logical size of the "datadir", at the last LSN.
-    current_logical_size: AtomicIsize,
-}
-
 #[derive(Debug)]
 pub enum LsnForTimestamp {
     Present(Lsn),
@@ -54,33 +32,29 @@ pub enum LsnForTimestamp {
     NoData(Lsn),
 }
 
-impl<R: Repository> DatadirTimeline<R> {
-    pub fn new(tline: Arc<R::Timeline>, repartition_threshold: u64) -> Self {
-        DatadirTimeline {
-            tline,
-            partitioning: Mutex::new((KeyPartitioning::new(), Lsn(0))),
-            current_logical_size: AtomicIsize::new(0),
-            repartition_threshold,
-        }
-    }
-
-    /// (Re-)calculate the logical size of the database at the latest LSN.
-    ///
-    /// This can be a slow operation.
-    pub fn init_logical_size(&self) -> Result<()> {
-        let last_lsn = self.tline.get_last_record_lsn();
-        self.current_logical_size.store(
-            self.get_current_logical_size_non_incremental(last_lsn)? as isize,
-            Ordering::SeqCst,
-        );
-        Ok(())
-    }
-
+///
+/// This trait provides all the functionality to store PostgreSQL relations, SLRUs,
+/// and other special kinds of files, in a versioned key-value store. The
+/// Timeline trait provides the key-value store.
+///
+/// This is a trait, so that we can easily include all these functions in a Timeline
+/// implementation. You're not expected to have different implementations of this trait,
+/// rather, this provides an interface and implementation, over Timeline.
+///
+/// If you wanted to store other kinds of data in the Neon repository, e.g.
+/// flat files or MySQL, you would create a new trait like this, with all the
+/// functions that make sense for the kind of data you're storing. For flat files,
+/// for example, you might have a function like "fn read(path, offset, size)".
+/// We might also have that situation in the future, to support multiple PostgreSQL
+/// versions, if there are big changes in how the data is organized in the data
+/// directory, or if new special files are introduced.
+///
+pub trait DatadirTimeline: Timeline {
     /// Start ingesting a WAL record, or other atomic modification of
     /// the timeline.
     ///
     /// This provides a transaction-like interface to perform a bunch
-    /// of modifications atomically, all stamped with one LSN.
+    /// of modifications atomically.
     ///
     /// To ingest a WAL record, call begin_modification(lsn) to get a
     /// DatadirModification object. Use the functions in the object to
@@ -88,18 +62,27 @@ impl<R: Repository> DatadirTimeline<R> {
     /// that the WAL record affects. When you're done, call commit() to
     /// commit the changes.
     ///
+    /// Lsn stored in modification is advanced by `ingest_record` and
+    /// is used by `commit()` to update `last_record_lsn`.
+    ///
+    /// Calling commit() will flush all the changes and reset the state,
+    /// so the `DatadirModification` struct can be reused to perform the next modification.
+    ///
     /// Note that any pending modifications you make through the
     /// modification object won't be visible to calls to the 'get' and list
     /// functions of the timeline until you finish! And if you update the
     /// same page twice, the last update wins.
     ///
-    pub fn begin_modification(&self, lsn: Lsn) -> DatadirModification<R> {
+    fn begin_modification(&self, lsn: Lsn) -> DatadirModification<Self>
+    where
+        Self: Sized,
+    {
         DatadirModification {
             tline: self,
-            lsn,
             pending_updates: HashMap::new(),
             pending_deletions: Vec::new(),
             pending_nblocks: 0,
+            lsn,
         }
     }
 
@@ -108,7 +91,7 @@ impl<R: Repository> DatadirTimeline<R> {
     //------------------------------------------------------------------------------
 
     /// Look up given page version.
-    pub fn get_rel_page_at_lsn(&self, tag: RelTag, blknum: BlockNumber, lsn: Lsn) -> Result<Bytes> {
+    fn get_rel_page_at_lsn(&self, tag: RelTag, blknum: BlockNumber, lsn: Lsn) -> Result<Bytes> {
         ensure!(tag.relnode != 0, "invalid relnode");
 
         let nblocks = self.get_rel_size(tag, lsn)?;
@@ -121,11 +104,11 @@ impl<R: Repository> DatadirTimeline<R> {
         }
 
         let key = rel_block_to_key(tag, blknum);
-        self.tline.get(key, lsn)
+        self.get(key, lsn)
     }
 
     // Get size of a database in blocks
-    pub fn get_db_size(&self, spcnode: Oid, dbnode: Oid, lsn: Lsn) -> Result<usize> {
+    fn get_db_size(&self, spcnode: Oid, dbnode: Oid, lsn: Lsn) -> Result<usize> {
         let mut total_blocks = 0;
 
         let rels = self.list_rels(spcnode, dbnode, lsn)?;
@@ -138,9 +121,13 @@ impl<R: Repository> DatadirTimeline<R> {
     }
 
     /// Get size of a relation file
-    pub fn get_rel_size(&self, tag: RelTag, lsn: Lsn) -> Result<BlockNumber> {
+    fn get_rel_size(&self, tag: RelTag, lsn: Lsn) -> Result<BlockNumber> {
         ensure!(tag.relnode != 0, "invalid relnode");
 
+        if let Some(nblocks) = self.get_cached_rel_size(&tag, lsn) {
+            return Ok(nblocks);
+        }
+
         if (tag.forknum == pg_constants::FSM_FORKNUM
             || tag.forknum == pg_constants::VISIBILITYMAP_FORKNUM)
             && !self.get_rel_exists(tag, lsn)?
@@ -153,17 +140,25 @@ impl<R: Repository> DatadirTimeline<R> {
         }
 
         let key = rel_size_to_key(tag);
-        let mut buf = self.tline.get(key, lsn)?;
-        Ok(buf.get_u32_le())
+        let mut buf = self.get(key, lsn)?;
+        let nblocks = buf.get_u32_le();
+
+        // Update relation size cache
+        self.update_cached_rel_size(tag, lsn, nblocks);
+        Ok(nblocks)
     }
 
     /// Does relation exist?
-    pub fn get_rel_exists(&self, tag: RelTag, lsn: Lsn) -> Result<bool> {
+    fn get_rel_exists(&self, tag: RelTag, lsn: Lsn) -> Result<bool> {
         ensure!(tag.relnode != 0, "invalid relnode");
 
+        // first try to lookup relation in cache
+        if let Some(_nblocks) = self.get_cached_rel_size(&tag, lsn) {
+            return Ok(true);
+        }
         // fetch directory listing
         let key = rel_dir_to_key(tag.spcnode, tag.dbnode);
-        let buf = self.tline.get(key, lsn)?;
+        let buf = self.get(key, lsn)?;
         let dir = RelDirectory::des(&buf)?;
 
         let exists = dir.rels.get(&(tag.relnode, tag.forknum)).is_some();
@@ -172,10 +167,10 @@ impl<R: Repository> DatadirTimeline<R> {
     }
 
     /// Get a list of all existing relations in given tablespace and database.
-    pub fn list_rels(&self, spcnode: Oid, dbnode: Oid, lsn: Lsn) -> Result<HashSet<RelTag>> {
+    fn list_rels(&self, spcnode: Oid, dbnode: Oid, lsn: Lsn) -> Result<HashSet<RelTag>> {
         // fetch directory listing
         let key = rel_dir_to_key(spcnode, dbnode);
-        let buf = self.tline.get(key, lsn)?;
+        let buf = self.get(key, lsn)?;
         let dir = RelDirectory::des(&buf)?;
 
         let rels: HashSet<RelTag> =
@@ -190,7 +185,7 @@ impl<R: Repository> DatadirTimeline<R> {
     }
 
     /// Look up given SLRU page version.
-    pub fn get_slru_page_at_lsn(
+    fn get_slru_page_at_lsn(
         &self,
         kind: SlruKind,
         segno: u32,
@@ -198,26 +193,21 @@ impl<R: Repository> DatadirTimeline<R> {
         lsn: Lsn,
     ) -> Result<Bytes> {
         let key = slru_block_to_key(kind, segno, blknum);
-        self.tline.get(key, lsn)
+        self.get(key, lsn)
     }
 
     /// Get size of an SLRU segment
-    pub fn get_slru_segment_size(
-        &self,
-        kind: SlruKind,
-        segno: u32,
-        lsn: Lsn,
-    ) -> Result<BlockNumber> {
+    fn get_slru_segment_size(&self, kind: SlruKind, segno: u32, lsn: Lsn) -> Result<BlockNumber> {
         let key = slru_segment_size_to_key(kind, segno);
-        let mut buf = self.tline.get(key, lsn)?;
+        let mut buf = self.get(key, lsn)?;
         Ok(buf.get_u32_le())
     }
 
     /// Get size of an SLRU segment
-    pub fn get_slru_segment_exists(&self, kind: SlruKind, segno: u32, lsn: Lsn) -> Result<bool> {
+    fn get_slru_segment_exists(&self, kind: SlruKind, segno: u32, lsn: Lsn) -> Result<bool> {
         // fetch directory listing
         let key = slru_dir_to_key(kind);
-        let buf = self.tline.get(key, lsn)?;
+        let buf = self.get(key, lsn)?;
         let dir = SlruSegmentDirectory::des(&buf)?;
 
         let exists = dir.segments.get(&segno).is_some();
@@ -231,10 +221,10 @@ impl<R: Repository> DatadirTimeline<R> {
     /// so it's not well defined which LSN you get if there were multiple commits
     /// "in flight" at that point in time.
     ///
-    pub fn find_lsn_for_timestamp(&self, search_timestamp: TimestampTz) -> Result<LsnForTimestamp> {
-        let gc_cutoff_lsn_guard = self.tline.get_latest_gc_cutoff_lsn();
+    fn find_lsn_for_timestamp(&self, search_timestamp: TimestampTz) -> Result<LsnForTimestamp> {
+        let gc_cutoff_lsn_guard = self.get_latest_gc_cutoff_lsn();
         let min_lsn = *gc_cutoff_lsn_guard;
-        let max_lsn = self.tline.get_last_record_lsn();
+        let max_lsn = self.get_last_record_lsn();
 
         // LSNs are always 8-byte aligned. low/mid/high represent the
         // LSN divided by 8.
@@ -325,88 +315,51 @@ impl<R: Repository> DatadirTimeline<R> {
     }
 
     /// Get a list of SLRU segments
-    pub fn list_slru_segments(&self, kind: SlruKind, lsn: Lsn) -> Result<HashSet<u32>> {
+    fn list_slru_segments(&self, kind: SlruKind, lsn: Lsn) -> Result<HashSet<u32>> {
         // fetch directory entry
         let key = slru_dir_to_key(kind);
 
-        let buf = self.tline.get(key, lsn)?;
+        let buf = self.get(key, lsn)?;
         let dir = SlruSegmentDirectory::des(&buf)?;
 
         Ok(dir.segments)
     }
 
-    pub fn get_relmap_file(&self, spcnode: Oid, dbnode: Oid, lsn: Lsn) -> Result<Bytes> {
+    fn get_relmap_file(&self, spcnode: Oid, dbnode: Oid, lsn: Lsn) -> Result<Bytes> {
         let key = relmap_file_key(spcnode, dbnode);
 
-        let buf = self.tline.get(key, lsn)?;
+        let buf = self.get(key, lsn)?;
         Ok(buf)
     }
 
-    pub fn list_dbdirs(&self, lsn: Lsn) -> Result<HashMap<(Oid, Oid), bool>> {
+    fn list_dbdirs(&self, lsn: Lsn) -> Result<HashMap<(Oid, Oid), bool>> {
         // fetch directory entry
-        let buf = self.tline.get(DBDIR_KEY, lsn)?;
+        let buf = self.get(DBDIR_KEY, lsn)?;
         let dir = DbDirectory::des(&buf)?;
 
         Ok(dir.dbdirs)
     }
 
-    pub fn get_twophase_file(&self, xid: TransactionId, lsn: Lsn) -> Result<Bytes> {
+    fn get_twophase_file(&self, xid: TransactionId, lsn: Lsn) -> Result<Bytes> {
         let key = twophase_file_key(xid);
-        let buf = self.tline.get(key, lsn)?;
+        let buf = self.get(key, lsn)?;
         Ok(buf)
     }
 
-    pub fn list_twophase_files(&self, lsn: Lsn) -> Result<HashSet<TransactionId>> {
+    fn list_twophase_files(&self, lsn: Lsn) -> Result<HashSet<TransactionId>> {
         // fetch directory entry
-        let buf = self.tline.get(TWOPHASEDIR_KEY, lsn)?;
+        let buf = self.get(TWOPHASEDIR_KEY, lsn)?;
         let dir = TwoPhaseDirectory::des(&buf)?;
 
         Ok(dir.xids)
     }
 
-    pub fn get_control_file(&self, lsn: Lsn) -> Result<Bytes> {
-        self.tline.get(CONTROLFILE_KEY, lsn)
+    fn get_control_file(&self, lsn: Lsn) -> Result<Bytes> {
+        self.get(CONTROLFILE_KEY, lsn)
     }
 
-    pub fn get_checkpoint(&self, lsn: Lsn) -> Result<Bytes> {
-        self.tline.get(CHECKPOINT_KEY, lsn)
-    }
-
-    /// Get the LSN of the last ingested WAL record.
-    ///
-    /// This is just a convenience wrapper that calls through to the underlying
-    /// repository.
-    pub fn get_last_record_lsn(&self) -> Lsn {
-        self.tline.get_last_record_lsn()
-    }
-
-    /// Check that it is valid to request operations with that lsn.
-    ///
-    /// This is just a convenience wrapper that calls through to the underlying
-    /// repository.
-    pub fn check_lsn_is_in_scope(
-        &self,
-        lsn: Lsn,
-        latest_gc_cutoff_lsn: &RwLockReadGuard<Lsn>,
-    ) -> Result<()> {
-        self.tline.check_lsn_is_in_scope(lsn, latest_gc_cutoff_lsn)
-    }
-
-    /// Retrieve current logical size of the timeline
-    ///
-    /// NOTE: counted incrementally, includes ancestors,
-    pub fn get_current_logical_size(&self) -> usize {
-        let current_logical_size = self.current_logical_size.load(Ordering::Acquire);
-        match usize::try_from(current_logical_size) {
-            Ok(sz) => sz,
-            Err(_) => {
-                error!(
-                    "current_logical_size is out of range: {}",
-                    current_logical_size
-                );
-                0
-            }
-        }
+    fn get_checkpoint(&self, lsn: Lsn) -> Result<Bytes> {
+        self.get(CHECKPOINT_KEY, lsn)
     }
 
     /// Does the same as get_current_logical_size but counted on demand.
@@ -414,16 +367,16 @@ impl<R: Repository> DatadirTimeline<R> {
     ///
     /// Only relation blocks are counted currently. That excludes metadata,
     /// SLRUs, twophase files etc.
-    pub fn get_current_logical_size_non_incremental(&self, lsn: Lsn) -> Result<usize> {
+    fn get_current_logical_size_non_incremental(&self, lsn: Lsn) -> Result<usize> {
         // Fetch list of database dirs and iterate them
-        let buf = self.tline.get(DBDIR_KEY, lsn)?;
+        let buf = self.get(DBDIR_KEY, lsn)?;
         let dbdir = DbDirectory::des(&buf)?;
 
         let mut total_size: usize = 0;
         for (spcnode, dbnode) in dbdir.dbdirs.keys() {
             for rel in self.list_rels(*spcnode, *dbnode, lsn)? {
                 let relsize_key = rel_size_to_key(rel);
-                let mut buf = self.tline.get(relsize_key, lsn)?;
+                let mut buf = self.get(relsize_key, lsn)?;
                 let relsize = buf.get_u32_le();
 
                 total_size += relsize as usize;
@@ -444,7 +397,7 @@ impl<R: Repository> DatadirTimeline<R> {
         result.add_key(DBDIR_KEY);
 
         // Fetch list of database dirs and iterate them
-        let buf = self.tline.get(DBDIR_KEY, lsn)?;
+        let buf = self.get(DBDIR_KEY, lsn)?;
         let dbdir = DbDirectory::des(&buf)?;
 
         let mut dbs: Vec<(Oid, Oid)> = dbdir.dbdirs.keys().cloned().collect();
@@ -461,7 +414,7 @@ impl<R: Repository> DatadirTimeline<R> {
             rels.sort_unstable();
             for rel in rels {
                 let relsize_key = rel_size_to_key(rel);
-                let mut buf = self.tline.get(relsize_key, lsn)?;
+                let mut buf = self.get(relsize_key, lsn)?;
                 let relsize = buf.get_u32_le();
 
                 result.add_range(rel_block_to_key(rel, 0)..rel_block_to_key(rel, relsize));
@@ -477,13 +430,13 @@ impl<R: Repository> DatadirTimeline<R> {
         ] {
             let slrudir_key = slru_dir_to_key(kind);
             result.add_key(slrudir_key);
-            let buf = self.tline.get(slrudir_key, lsn)?;
+            let buf = self.get(slrudir_key, lsn)?;
             let dir = SlruSegmentDirectory::des(&buf)?;
             let mut segments: Vec<u32> = dir.segments.iter().cloned().collect();
             segments.sort_unstable();
             for segno in segments {
                 let segsize_key = slru_segment_size_to_key(kind, segno);
-                let mut buf = self.tline.get(segsize_key, lsn)?;
+                let mut buf = self.get(segsize_key, lsn)?;
                 let segsize = buf.get_u32_le();
 
                 result.add_range(
@@ -495,7 +448,7 @@ impl<R: Repository> DatadirTimeline<R> {
 
         // Then pg_twophase
         result.add_key(TWOPHASEDIR_KEY);
-        let buf = self.tline.get(TWOPHASEDIR_KEY, lsn)?;
+        let buf = self.get(TWOPHASEDIR_KEY, lsn)?;
         let twophase_dir = TwoPhaseDirectory::des(&buf)?;
         let mut xids: Vec<TransactionId> = twophase_dir.xids.iter().cloned().collect();
         xids.sort_unstable();
@@ -509,31 +462,31 @@ impl<R: Repository> DatadirTimeline<R> {
         Ok(result.to_keyspace())
     }
 
-    pub fn repartition(&self, lsn: Lsn, partition_size: u64) -> Result<(KeyPartitioning, Lsn)> {
-        let mut partitioning_guard = self.partitioning.lock().unwrap();
-        if partitioning_guard.1 == Lsn(0)
-            || lsn.0 - partitioning_guard.1 .0 > self.repartition_threshold
-        {
-            let keyspace = self.collect_keyspace(lsn)?;
-            let partitioning = keyspace.partition(partition_size);
-            *partitioning_guard = (partitioning, lsn);
-            return Ok((partitioning_guard.0.clone(), lsn));
-        }
-        Ok((partitioning_guard.0.clone(), partitioning_guard.1))
-    }
+    /// Get cached size of relation if it not updated after specified LSN
+    fn get_cached_rel_size(&self, tag: &RelTag, lsn: Lsn) -> Option<BlockNumber>;
+
+    /// Update cached relation size if there is no more recent update
+    fn update_cached_rel_size(&self, tag: RelTag, lsn: Lsn, nblocks: BlockNumber);
+
+    /// Store cached relation size
+    fn set_cached_rel_size(&self, tag: RelTag, lsn: Lsn, nblocks: BlockNumber);
+
+    /// Remove cached relation size
+    fn remove_cached_rel_size(&self, tag: &RelTag);
 }
 
 /// DatadirModification represents an operation to ingest an atomic set of
 /// updates to the repository. It is created by the 'begin_record'
 /// function. It is called for each WAL record, so that all the modifications
 /// by a one WAL record appear atomic.
-pub struct DatadirModification<'a, R: Repository> {
+pub struct DatadirModification<'a, T: DatadirTimeline> {
     /// The timeline this modification applies to. You can access this to
     /// read the state, but note that any pending updates are *not* reflected
     /// in the state in 'tline' yet.
-    pub tline: &'a DatadirTimeline<R>,
+    pub tline: &'a T,
 
-    lsn: Lsn,
+    /// Lsn assigned by begin_modification
+    pub lsn: Lsn,
 
     // The modifications are not applied directly to the underlying key-value store.
     // The put-functions add the modifications here, and they are flushed to the
@@ -543,7 +496,7 @@ pub struct DatadirModification<'a, R: Repository> {
     pending_nblocks: isize,
 }
 
-impl<'a, R: Repository> DatadirModification<'a, R> {
+impl<'a, T: DatadirTimeline> DatadirModification<'a, T> {
     /// Initialize a completely new repository.
     ///
     /// This inserts the directory metadata entries that are assumed to
@@ -744,26 +697,36 @@ impl<'a, R: Repository> DatadirModification<'a, R> {
 
         self.pending_nblocks += nblocks as isize;
 
+        // Update relation size cache
+        self.tline.set_cached_rel_size(rel, self.lsn, nblocks);
+
         // Even if nblocks > 0, we don't insert any actual blocks here. That's up to the
         // caller.
-
         Ok(())
     }
 
     /// Truncate relation
     pub fn put_rel_truncation(&mut self, rel: RelTag, nblocks: BlockNumber) -> Result<()> {
         ensure!(rel.relnode != 0, "invalid relnode");
-        let size_key = rel_size_to_key(rel);
+        let last_lsn = self.tline.get_last_record_lsn();
+        if self.tline.get_rel_exists(rel, last_lsn)? {
+            let size_key = rel_size_to_key(rel);
+            // Fetch the old size first
+            let old_size = self.get(size_key)?.get_u32_le();
 
-        // Fetch the old size first
-        let old_size = self.get(size_key)?.get_u32_le();
+            // Update the entry with the new size.
+            let buf = nblocks.to_le_bytes();
+            self.put(size_key, Value::Image(Bytes::from(buf.to_vec())));
 
-        // Update the entry with the new size.
-        let buf = nblocks.to_le_bytes();
-        self.put(size_key, Value::Image(Bytes::from(buf.to_vec())));
+            // Update relation size cache
+            self.tline.set_cached_rel_size(rel, self.lsn, nblocks);
 
-        // Update logical database size.
-        self.pending_nblocks -= old_size as isize - nblocks as isize;
+            // Update relation size cache
+            self.tline.set_cached_rel_size(rel, self.lsn, nblocks);
+
+            // Update logical database size.
+            self.pending_nblocks -= old_size as isize - nblocks as isize;
+        }
         Ok(())
     }
 
@@ -781,6 +744,9 @@ impl<'a, R: Repository> DatadirModification<'a, R> {
             let buf = nblocks.to_le_bytes();
             self.put(size_key, Value::Image(Bytes::from(buf.to_vec())));
 
+            // Update relation size cache
+            self.tline.set_cached_rel_size(rel, self.lsn, nblocks);
+
             self.pending_nblocks += nblocks as isize - old_size as isize;
         }
         Ok(())
@@ -806,6 +772,9 @@ impl<'a, R: Repository> DatadirModification<'a, R> {
         let old_size = self.get(size_key)?.get_u32_le();
         self.pending_nblocks -= old_size as isize;
 
+        // Remove enty from relation size cache
+        self.tline.remove_cached_rel_size(&rel);
+
         // Delete size entry, as well as all blocks
         self.delete(rel_key_range(rel));
 
@@ -928,7 +897,7 @@ impl<'a, R: Repository> DatadirModification<'a, R> {
             return Ok(());
         }
 
-        let writer = self.tline.tline.writer();
+        let writer = self.tline.writer();
 
         // Flush relation and  SLRU data blocks, keep metadata.
         let mut result: Result<()> = Ok(());
@@ -943,10 +912,7 @@ impl<'a, R: Repository> DatadirModification<'a, R> {
         result?;
 
         if pending_nblocks != 0 {
-            self.tline.current_logical_size.fetch_add(
-                pending_nblocks * pg_constants::BLCKSZ as isize,
-                Ordering::SeqCst,
-            );
+            writer.update_current_logical_size(pending_nblocks * pg_constants::BLCKSZ as isize);
             self.pending_nblocks = 0;
         }
 
@@ -956,26 +922,25 @@ impl<'a, R: Repository> DatadirModification<'a, R> {
     ///
     /// Finish this atomic update, writing all the updated keys to the
     /// underlying timeline.
+    /// All the modifications in this atomic update are stamped by the specified LSN.
     ///
-    pub fn commit(self) -> Result<()> {
-        let writer = self.tline.tline.writer();
-
+    pub fn commit(&mut self) -> Result<()> {
+        let writer = self.tline.writer();
+        let lsn = self.lsn;
         let pending_nblocks = self.pending_nblocks;
+        self.pending_nblocks = 0;
 
-        for (key, value) in self.pending_updates {
-            writer.put(key, self.lsn, &value)?;
+        for (key, value) in self.pending_updates.drain() {
+            writer.put(key, lsn, &value)?;
         }
-        for key_range in self.pending_deletions {
-            writer.delete(key_range.clone(), self.lsn)?;
+        for key_range in self.pending_deletions.drain(..) {
+            writer.delete(key_range, lsn)?;
         }
 
-        writer.finish_write(self.lsn);
+        writer.finish_write(lsn);
 
         if pending_nblocks != 0 {
-            self.tline.current_logical_size.fetch_add(
-                pending_nblocks * pg_constants::BLCKSZ as isize,
-                Ordering::SeqCst,
-            );
+            writer.update_current_logical_size(pending_nblocks * pg_constants::BLCKSZ as isize);
         }
 
         Ok(())
@@ -1001,8 +966,8 @@ impl<'a, R: Repository> DatadirModification<'a, R> {
                 bail!("unexpected pending WAL record");
             }
         } else {
-            let last_lsn = self.tline.get_last_record_lsn();
-            self.tline.tline.get(key, last_lsn)
+            let lsn = Lsn::max(self.tline.get_last_record_lsn(), self.lsn);
+            self.tline.get(key, lsn)
         }
     }
 
@@ -1404,13 +1369,12 @@ fn is_slru_block_key(key: Key) -> bool {
 pub fn create_test_timeline<R: Repository>(
     repo: R,
     timeline_id: utils::zid::ZTimelineId,
-) -> Result<Arc<crate::DatadirTimeline<R>>> {
+) -> Result<std::sync::Arc<R::Timeline>> {
     let tline = repo.create_empty_timeline(timeline_id, Lsn(8))?;
-    let tline = DatadirTimeline::new(tline, 256 * 1024);
     let mut m = tline.begin_modification(Lsn(8));
     m.init_empty()?;
     m.commit()?;
-    Ok(Arc::new(tline))
+    Ok(tline)
 }
 
 #[allow(clippy::bool_assert_comparison)]
@@ -1483,7 +1447,7 @@ mod tests {
             .contains(&TESTREL_A));
 
         // Run checkpoint and garbage collection and check that it's still not visible
-        newtline.tline.checkpoint(CheckpointConfig::Forced)?;
+        newtline.checkpoint(CheckpointConfig::Forced)?;
         repo.gc_iteration(Some(NEW_TIMELINE_ID), 0, true)?;
 
         assert!(!newtline

pageserver/src/repository.rs

@@ -185,7 +185,7 @@ impl Value {
 /// A repository corresponds to one .neon directory. One repository holds multiple
 /// timelines, forked off from the same initial call to 'initdb'.
 pub trait Repository: Send + Sync {
-    type Timeline: Timeline;
+    type Timeline: crate::DatadirTimeline;
 
     /// Updates timeline based on the `TimelineSyncStatusUpdate`, received from the remote storage synchronization.
     /// See [`crate::remote_storage`] for more details about the synchronization.
@@ -277,15 +277,6 @@ pub enum LocalTimelineState {
     Unloaded,
 }
 
-impl<'a, T> From<&'a RepositoryTimeline<T>> for LocalTimelineState {
-    fn from(local_timeline_entry: &'a RepositoryTimeline<T>) -> Self {
-        match local_timeline_entry {
-            RepositoryTimeline::Loaded(_) => LocalTimelineState::Loaded,
-            RepositoryTimeline::Unloaded { .. } => LocalTimelineState::Unloaded,
-        }
-    }
-}
-
 ///
 /// Result of performing GC
 ///
@@ -382,6 +373,11 @@ pub trait Timeline: Send + Sync {
         lsn: Lsn,
         latest_gc_cutoff_lsn: &RwLockReadGuard<Lsn>,
     ) -> Result<()>;
+
+    /// Get the physical size of the timeline at the latest LSN
+    fn get_physical_size(&self) -> u64;
+    /// Get the physical size of the timeline at the latest LSN non incrementally
+    fn get_physical_size_non_incremental(&self) -> Result<u64>;
 }
 
 /// Various functions to mutate the timeline.
@@ -405,12 +401,14 @@ pub trait TimelineWriter<'a> {
     /// the 'lsn' or anything older. The previous last record LSN is stored alongside
     /// the latest and can be read.
     fn finish_write(&self, lsn: Lsn);
+
+    fn update_current_logical_size(&self, delta: isize);
 }
 
 #[cfg(test)]
 pub mod repo_harness {
     use bytes::BytesMut;
-    use lazy_static::lazy_static;
+    use once_cell::sync::Lazy;
     use std::sync::{Arc, RwLock, RwLockReadGuard, RwLockWriteGuard};
     use std::{fs, path::PathBuf};
 
@@ -441,14 +439,13 @@ pub mod repo_harness {
         buf.freeze()
     }
 
-    lazy_static! {
-        static ref LOCK: RwLock<()> = RwLock::new(());
-    }
+    static LOCK: Lazy<RwLock<()>> = Lazy::new(|| RwLock::new(()));
 
     impl From<TenantConf> for TenantConfOpt {
         fn from(tenant_conf: TenantConf) -> Self {
             Self {
                 checkpoint_distance: Some(tenant_conf.checkpoint_distance),
+                checkpoint_timeout: Some(tenant_conf.checkpoint_timeout),
                 compaction_target_size: Some(tenant_conf.compaction_target_size),
                 compaction_period: Some(tenant_conf.compaction_period),
                 compaction_threshold: Some(tenant_conf.compaction_threshold),
@@ -591,11 +588,10 @@ mod tests {
     //use std::sync::Arc;
     use bytes::BytesMut;
     use hex_literal::hex;
-    use lazy_static::lazy_static;
+    use once_cell::sync::Lazy;
 
-    lazy_static! {
-        static ref TEST_KEY: Key = Key::from_slice(&hex!("112222222233333333444444445500000001"));
-    }
+    static TEST_KEY: Lazy<Key> =
+        Lazy::new(|| Key::from_slice(&hex!("112222222233333333444444445500000001")));
 
     #[test]
     fn test_basic() -> Result<()> {

pageserver/src/storage_sync.rs

@@ -155,8 +155,7 @@ use std::{
 
 use anyhow::{anyhow, bail, Context};
 use futures::stream::{FuturesUnordered, StreamExt};
-use lazy_static::lazy_static;
-use once_cell::sync::OnceCell;
+use once_cell::sync::{Lazy, OnceCell};
 use remote_storage::{GenericRemoteStorage, RemoteStorage};
 use tokio::{
     fs,
@@ -173,10 +172,10 @@ use self::{
 };
 use crate::{
     config::PageServerConf,
+    exponential_backoff,
     layered_repository::{
         ephemeral_file::is_ephemeral_file,
         metadata::{metadata_path, TimelineMetadata, METADATA_FILE_NAME},
-        LayeredRepository,
     },
     storage_sync::{self, index::RemoteIndex},
     tenant_mgr::attach_downloaded_tenants,
@@ -185,8 +184,8 @@ use crate::{
 };
 
 use metrics::{
-    register_histogram_vec, register_int_counter, register_int_counter_vec, register_int_gauge,
-    HistogramVec, IntCounter, IntCounterVec, IntGauge,
+    register_histogram_vec, register_int_counter_vec, register_int_gauge, HistogramVec,
+    IntCounterVec, IntGauge,
 };
 use utils::zid::{ZTenantId, ZTenantTimelineId, ZTimelineId};
 
@@ -194,32 +193,33 @@ use self::download::download_index_parts;
 pub use self::download::gather_tenant_timelines_index_parts;
 pub use self::download::TEMP_DOWNLOAD_EXTENSION;
 
-lazy_static! {
-    static ref REMAINING_SYNC_ITEMS: IntGauge = register_int_gauge!(
+static REMAINING_SYNC_ITEMS: Lazy<IntGauge> = Lazy::new(|| {
+    register_int_gauge!(
         "pageserver_remote_storage_remaining_sync_items",
         "Number of storage sync items left in the queue"
     )
-    .expect("failed to register pageserver remote storage remaining sync items int gauge");
-    static ref FATAL_TASK_FAILURES: IntCounter = register_int_counter!(
-        "pageserver_remote_storage_fatal_task_failures_total",
-        "Number of critically failed tasks"
-    )
-    .expect("failed to register pageserver remote storage remaining sync items int gauge");
-    static ref IMAGE_SYNC_TIME: HistogramVec = register_histogram_vec!(
+    .expect("failed to register pageserver remote storage remaining sync items int gauge")
+});
+
+static IMAGE_SYNC_TIME: Lazy<HistogramVec> = Lazy::new(|| {
+    register_histogram_vec!(
         "pageserver_remote_storage_image_sync_seconds",
         "Time took to synchronize (download or upload) a whole pageserver image. \
         Grouped by tenant and timeline ids, `operation_kind` (upload|download) and `status` (success|failure)",
         &["tenant_id", "timeline_id", "operation_kind", "status"],
         vec![0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1.0, 3.0, 10.0, 20.0]
     )
-    .expect("failed to register pageserver image sync time histogram vec");
-    static ref REMOTE_INDEX_UPLOAD: IntCounterVec = register_int_counter_vec!(
+    .expect("failed to register pageserver image sync time histogram vec")
+});
+
+static REMOTE_INDEX_UPLOAD: Lazy<IntCounterVec> = Lazy::new(|| {
+    register_int_counter_vec!(
         "pageserver_remote_storage_remote_index_uploads_total",
         "Number of remote index uploads",
         &["tenant_id", "timeline_id"],
     )
-    .expect("failed to register pageserver remote index upload vec");
-}
+    .expect("failed to register pageserver remote index upload vec")
+});
 
 static SYNC_QUEUE: OnceCell<SyncQueue> = OnceCell::new();
 
@@ -928,7 +928,7 @@ fn storage_sync_loop<P, S>(
                     );
                     let mut sync_status_updates: HashMap<ZTenantId, HashSet<ZTimelineId>> =
                         HashMap::new();
-                    let index_accessor = runtime.block_on(index.write());
+                    let index_accessor = runtime.block_on(index.read());
                     for tenant_id in updated_tenants {
                         let tenant_entry = match index_accessor.tenant_entry(&tenant_id) {
                             Some(tenant_entry) => tenant_entry,
@@ -970,14 +970,19 @@ fn storage_sync_loop<P, S>(
     }
 }
 
-// needed to check whether the download happened
-// more informative than just a bool
 #[derive(Debug)]
-enum DownloadMarker {
+enum DownloadStatus {
     Downloaded,
     Nothing,
 }
 
+#[derive(Debug)]
+enum UploadStatus {
+    Uploaded,
+    Failed,
+    Nothing,
+}
+
 async fn process_batches<P, S>(
     conf: &'static PageServerConf,
     max_sync_errors: NonZeroU32,
@@ -1017,7 +1022,7 @@ where
             "Finished storage sync task for sync id {sync_id} download marker {:?}",
             download_marker
         );
-        if matches!(download_marker, DownloadMarker::Downloaded) {
+        if matches!(download_marker, DownloadStatus::Downloaded) {
             downloaded_timelines.insert(sync_id.tenant_id);
         }
     }
@@ -1031,7 +1036,7 @@ async fn process_sync_task_batch<P, S>(
     max_sync_errors: NonZeroU32,
     sync_id: ZTenantTimelineId,
     batch: SyncTaskBatch,
-) -> DownloadMarker
+) -> DownloadStatus
 where
     P: Debug + Send + Sync + 'static,
     S: RemoteStorage<RemoteObjectId = P> + Send + Sync + 'static,
@@ -1048,7 +1053,7 @@ where
     // When operating in a system without tasks failing over the error threshold,
     // current batching and task processing systems aim to update the layer set and metadata files (remote and local),
     // without "losing" such layer files.
-    let (upload_result, status_update) = tokio::join!(
+    let (upload_status, download_status) = tokio::join!(
         async {
             if let Some(upload_data) = upload_data {
                 match validate_task_retries(upload_data, max_sync_errors)
@@ -1066,7 +1071,7 @@ where
                             "upload",
                         )
                         .await;
-                        return Some(());
+                        UploadStatus::Uploaded
                     }
                     ControlFlow::Break(failed_upload_data) => {
                         if let Err(e) = update_remote_data(
@@ -1083,10 +1088,13 @@ where
                         {
                             error!("Failed to update remote timeline {sync_id}: {e:?}");
                         }
+
+                        UploadStatus::Failed
                     }
                 }
+            } else {
+                UploadStatus::Nothing
             }
-            None
         }
         .instrument(info_span!("upload_timeline_data")),
         async {
@@ -1116,50 +1124,53 @@ where
                     }
                 }
             }
-            DownloadMarker::Nothing
+            DownloadStatus::Nothing
         }
         .instrument(info_span!("download_timeline_data")),
     );
 
     if let Some(delete_data) = batch.delete {
-        if upload_result.is_some() {
-            match validate_task_retries(delete_data, max_sync_errors)
-                .instrument(info_span!("retries_validation"))
-                .await
-            {
-                ControlFlow::Continue(new_delete_data) => {
-                    delete_timeline_data(
-                        conf,
-                        (storage.as_ref(), &index, sync_queue),
-                        sync_id,
-                        new_delete_data,
-                        sync_start,
-                        "delete",
-                    )
-                    .instrument(info_span!("delete_timeline_data"))
-                    .await;
-                }
-                ControlFlow::Break(failed_delete_data) => {
-                    if let Err(e) = update_remote_data(
-                        conf,
-                        storage.as_ref(),
-                        &index,
-                        sync_id,
-                        RemoteDataUpdate::Delete(&failed_delete_data.data.deleted_layers),
-                    )
+        match upload_status {
+            UploadStatus::Uploaded | UploadStatus::Nothing => {
+                match validate_task_retries(delete_data, max_sync_errors)
+                    .instrument(info_span!("retries_validation"))
                     .await
-                    {
-                        error!("Failed to update remote timeline {sync_id}: {e:?}");
+                {
+                    ControlFlow::Continue(new_delete_data) => {
+                        delete_timeline_data(
+                            conf,
+                            (storage.as_ref(), &index, sync_queue),
+                            sync_id,
+                            new_delete_data,
+                            sync_start,
+                            "delete",
+                        )
+                        .instrument(info_span!("delete_timeline_data"))
+                        .await;
+                    }
+                    ControlFlow::Break(failed_delete_data) => {
+                        if let Err(e) = update_remote_data(
+                            conf,
+                            storage.as_ref(),
+                            &index,
+                            sync_id,
+                            RemoteDataUpdate::Delete(&failed_delete_data.data.deleted_layers),
+                        )
+                        .await
+                        {
+                            error!("Failed to update remote timeline {sync_id}: {e:?}");
+                        }
                     }
                 }
             }
-        } else {
-            sync_queue.push(sync_id, SyncTask::Delete(delete_data));
-            warn!("Skipping delete task due to failed upload tasks, reenqueuing");
+            UploadStatus::Failed => {
+                warn!("Skipping delete task due to failed upload tasks, reenqueuing");
+                sync_queue.push(sync_id, SyncTask::Delete(delete_data));
+            }
         }
     }
 
-    status_update
+    download_status
 }
 
 async fn download_timeline_data<P, S>(
@@ -1170,7 +1181,7 @@ async fn download_timeline_data<P, S>(
     new_download_data: SyncData<LayersDownload>,
     sync_start: Instant,
     task_name: &str,
-) -> DownloadMarker
+) -> DownloadStatus
 where
     P: Debug + Send + Sync + 'static,
     S: RemoteStorage<RemoteObjectId = P> + Send + Sync + 'static,
@@ -1199,7 +1210,7 @@ where
                 Ok(()) => match index.write().await.set_awaits_download(&sync_id, false) {
                     Ok(()) => {
                         register_sync_status(sync_id, sync_start, task_name, Some(true));
-                        return DownloadMarker::Downloaded;
+                        return DownloadStatus::Downloaded;
                     }
                     Err(e) => {
                         error!("Timeline {sync_id} was expected to be in the remote index after a successful download, but it's absent: {e:?}");
@@ -1215,7 +1226,7 @@ where
         }
     }
 
-    DownloadMarker::Nothing
+    DownloadStatus::Nothing
 }
 
 async fn update_local_metadata(
@@ -1257,7 +1268,13 @@ async fn update_local_metadata(
             timeline_id,
         } = sync_id;
         tokio::task::spawn_blocking(move || {
-            LayeredRepository::save_metadata(conf, timeline_id, tenant_id, &cloned_metadata, true)
+            crate::layered_repository::save_metadata(
+                conf,
+                timeline_id,
+                tenant_id,
+                &cloned_metadata,
+                true,
+            )
         })
         .await
         .with_context(|| {
@@ -1487,11 +1504,7 @@ async fn validate_task_retries<T>(
         return ControlFlow::Break(sync_data);
     }
 
-    if current_attempt > 0 {
-        let seconds_to_wait = 2.0_f64.powf(current_attempt as f64 - 1.0).min(30.0);
-        info!("Waiting {seconds_to_wait} seconds before starting the task");
-        tokio::time::sleep(Duration::from_secs_f64(seconds_to_wait)).await;
-    }
+    exponential_backoff(current_attempt, 1.0, 30.0).await;
     ControlFlow::Continue(sync_data)
 }
 
@@ -1557,6 +1570,7 @@ fn schedule_first_sync_tasks(
     local_timeline_init_statuses
 }
 
+/// bool in return value stands for awaits_download
 fn compare_local_and_remote_timeline(
     new_sync_tasks: &mut VecDeque<(ZTenantTimelineId, SyncTask)>,
     sync_id: ZTenantTimelineId,
@@ -1566,14 +1580,6 @@ fn compare_local_and_remote_timeline(
 ) -> (LocalTimelineInitStatus, bool) {
     let remote_files = remote_entry.stored_files();
 
-    // TODO probably here we need more sophisticated logic,
-    //   if more data is available remotely can we just download what's there?
-    //   without trying to upload something. It may be tricky, needs further investigation.
-    //   For now looks strange that we can request upload
-    //   and download for the same timeline simultaneously.
-    //   (upload needs to be only for previously unsynced files, not whole timeline dir).
-    //   If one of the tasks fails they will be reordered in the queue which can lead
-    //   to timeline being stuck in evicted state
     let number_of_layers_to_download = remote_files.difference(&local_files).count();
     let (initial_timeline_status, awaits_download) = if number_of_layers_to_download > 0 {
         new_sync_tasks.push_back((

pageserver/src/storage_sync/download.rs

@@ -3,12 +3,13 @@
 use std::{
     collections::{HashMap, HashSet},
     fmt::Debug,
+    mem,
     path::Path,
 };
 
 use anyhow::Context;
 use futures::stream::{FuturesUnordered, StreamExt};
-use remote_storage::{path_with_suffix_extension, RemoteObjectName, RemoteStorage};
+use remote_storage::{path_with_suffix_extension, DownloadError, RemoteObjectName, RemoteStorage};
 use tokio::{
     fs,
     io::{self, AsyncWriteExt},
@@ -27,28 +28,50 @@ use super::{
 
 pub const TEMP_DOWNLOAD_EXTENSION: &str = "temp_download";
 
-/// FIXME: Needs cleanup. Currently it swallows errors. Here we need to ensure that
-/// we successfully downloaded all metadata parts for one tenant.
-/// And successful includes absence of index_part in the remote. Because it is valid situation
-/// when timeline was just created and pageserver restarted before upload of index part was completed.
-/// But currently RemoteStorage interface does not provide this knowledge because it uses
-/// anyhow::Error as an error type. So this needs a refactoring.
-///
-/// In other words we need to yield only complete sets of tenant timelines.
-/// Failure for one timeline of a tenant should exclude whole tenant from returned hashmap.
-/// So there are two requirements: keep everything in one futures unordered
-/// to allow higher concurrency. Mark tenants as failed independently.
-/// That requires some bookeeping.
+// We collect timelines remotely available for each tenant
+// in case we failed to gather all index parts (due to an error)
+// Poisoned variant is returned.
+// When data is received succesfully without errors Present variant is used.
+pub enum TenantIndexParts {
+    Poisoned {
+        present: HashMap<ZTimelineId, IndexPart>,
+        missing: HashSet<ZTimelineId>,
+    },
+    Present(HashMap<ZTimelineId, IndexPart>),
+}
+
+impl TenantIndexParts {
+    fn add_poisoned(&mut self, timeline_id: ZTimelineId) {
+        match self {
+            TenantIndexParts::Poisoned { missing, .. } => {
+                missing.insert(timeline_id);
+            }
+            TenantIndexParts::Present(present) => {
+                *self = TenantIndexParts::Poisoned {
+                    present: mem::take(present),
+                    missing: HashSet::from([timeline_id]),
+                }
+            }
+        }
+    }
+}
+
+impl Default for TenantIndexParts {
+    fn default() -> Self {
+        TenantIndexParts::Present(HashMap::default())
+    }
+}
+
 pub async fn download_index_parts<P, S>(
     conf: &'static PageServerConf,
     storage: &S,
     keys: HashSet<ZTenantTimelineId>,
-) -> HashMap<ZTenantId, HashMap<ZTimelineId, IndexPart>>
+) -> HashMap<ZTenantId, TenantIndexParts>
 where
     P: Debug + Send + Sync + 'static,
     S: RemoteStorage<RemoteObjectId = P> + Send + Sync + 'static,
 {
-    let mut index_parts: HashMap<ZTenantId, HashMap<ZTimelineId, IndexPart>> = HashMap::new();
+    let mut index_parts: HashMap<ZTenantId, TenantIndexParts> = HashMap::new();
 
     let mut part_downloads = keys
         .into_iter()
@@ -59,12 +82,29 @@ where
         match part_upload_result {
             Ok(index_part) => {
                 debug!("Successfully fetched index part for {id}");
-                index_parts
-                    .entry(id.tenant_id)
-                    .or_default()
-                    .insert(id.timeline_id, index_part);
+                match index_parts.entry(id.tenant_id).or_default() {
+                    TenantIndexParts::Poisoned { present, .. } => {
+                        present.insert(id.timeline_id, index_part);
+                    }
+                    TenantIndexParts::Present(parts) => {
+                        parts.insert(id.timeline_id, index_part);
+                    }
+                }
+            }
+            Err(download_error) => {
+                match download_error {
+                    DownloadError::NotFound => {
+                        // thats ok because it means that we didnt upload something we have locally for example
+                    }
+                    e => {
+                        let tenant_parts = index_parts.entry(id.tenant_id).or_default();
+                        tenant_parts.add_poisoned(id.timeline_id);
+                        error!(
+                            "Failed to fetch index part for {id}: {e} poisoning tenant index parts"
+                        );
+                    }
+                }
             }
-            Err(e) => error!("Failed to fetch index part for {id}: {e}"),
         }
     }
 
@@ -90,6 +130,7 @@ where
             tenant_path.display()
         )
     })?;
+
     let timelines = storage
         .list_prefixes(Some(tenant_storage_path))
         .await
@@ -100,6 +141,13 @@ where
             )
         })?;
 
+    if timelines.is_empty() {
+        anyhow::bail!(
+            "no timelines found on the remote storage for tenant {}",
+            tenant_id
+        )
+    }
+
     let mut sync_ids = HashSet::new();
 
     for timeline_remote_storage_key in timelines {
@@ -119,10 +167,16 @@ where
         });
     }
 
-    download_index_parts(conf, storage, sync_ids)
+    match download_index_parts(conf, storage, sync_ids)
         .await
         .remove(&tenant_id)
-        .ok_or_else(|| anyhow::anyhow!("Missing tenant index parts. This is a bug."))
+        .ok_or_else(|| anyhow::anyhow!("Missing tenant index parts. This is a bug."))?
+    {
+        TenantIndexParts::Poisoned { missing, .. } => {
+            anyhow::bail!("Failed to download index parts for all timelines. Missing {missing:?}")
+        }
+        TenantIndexParts::Present(parts) => Ok(parts),
+    }
 }
 
 /// Retrieves index data from the remote storage for a given timeline.
@@ -130,7 +184,7 @@ async fn download_index_part<P, S>(
     conf: &'static PageServerConf,
     storage: &S,
     sync_id: ZTenantTimelineId,
-) -> anyhow::Result<IndexPart>
+) -> Result<IndexPart, DownloadError>
 where
     P: Debug + Send + Sync + 'static,
     S: RemoteStorage<RemoteObjectId = P> + Send + Sync + 'static,
@@ -145,15 +199,11 @@ where
                 "Failed to get the index part storage path for local path '{}'",
                 index_part_path.display()
             )
-        })?;
+        })
+        .map_err(DownloadError::BadInput)?;
+
+    let mut index_part_download = storage.download(&part_storage_path).await?;
 
-    let mut index_part_download =
-        storage
-            .download(&part_storage_path)
-            .await
-            .with_context(|| {
-                format!("Failed to open download stream for for storage path {part_storage_path:?}")
-            })?;
     let mut index_part_bytes = Vec::new();
     io::copy(
         &mut index_part_download.download_stream,
@@ -162,11 +212,16 @@ where
     .await
     .with_context(|| {
         format!("Failed to download an index part from storage path {part_storage_path:?}")
-    })?;
+    })
+    .map_err(DownloadError::Other)?;
 
-    let index_part: IndexPart = serde_json::from_slice(&index_part_bytes).with_context(|| {
-        format!("Failed to deserialize index part file from storage path '{part_storage_path:?}'")
-    })?;
+    let index_part: IndexPart = serde_json::from_slice(&index_part_bytes)
+        .with_context(|| {
+            format!(
+                "Failed to deserialize index part file from storage path '{part_storage_path:?}'"
+            )
+        })
+        .map_err(DownloadError::Other)?;
 
     let missing_files = index_part.missing_files();
     if !missing_files.is_empty() {

pageserver/src/storage_sync/index.rs

@@ -13,6 +13,7 @@ use anyhow::{anyhow, Context, Ok};
 use serde::{Deserialize, Serialize};
 use serde_with::{serde_as, DisplayFromStr};
 use tokio::sync::RwLock;
+use tracing::log::warn;
 
 use crate::{config::PageServerConf, layered_repository::metadata::TimelineMetadata};
 use utils::{
@@ -20,6 +21,8 @@ use utils::{
     zid::{ZTenantId, ZTenantTimelineId, ZTimelineId},
 };
 
+use super::download::TenantIndexParts;
+
 /// A part of the filesystem path, that needs a root to become a path again.
 #[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize, Deserialize)]
 #[serde(transparent)]
@@ -88,21 +91,27 @@ pub struct RemoteIndex(Arc<RwLock<RemoteTimelineIndex>>);
 impl RemoteIndex {
     pub fn from_parts(
         conf: &'static PageServerConf,
-        index_parts: HashMap<ZTenantId, HashMap<ZTimelineId, IndexPart>>,
+        index_parts: HashMap<ZTenantId, TenantIndexParts>,
     ) -> anyhow::Result<Self> {
         let mut entries: HashMap<ZTenantId, TenantEntry> = HashMap::new();
 
-        for (tenant_id, timelines) in index_parts {
-            for (timeline_id, index_part) in timelines {
-                let timeline_path = conf.timeline_path(&timeline_id, &tenant_id);
-                let remote_timeline =
-                    RemoteTimeline::from_index_part(&timeline_path, index_part)
-                        .context("Failed to restore remote timeline data from index part")?;
+        for (tenant_id, index_parts) in index_parts {
+            match index_parts {
+                // TODO: should we schedule a retry so it can be recovered? otherwise we can revive it only through detach/attach or pageserver restart
+                TenantIndexParts::Poisoned { missing, ..} => warn!("skipping tenant_id set up for remote index because the index download has failed for timeline(s): {missing:?}"),
+                TenantIndexParts::Present(timelines) => {
+                    for (timeline_id, index_part) in timelines {
+                        let timeline_path = conf.timeline_path(&timeline_id, &tenant_id);
+                        let remote_timeline =
+                            RemoteTimeline::from_index_part(&timeline_path, index_part)
+                                .context("Failed to restore remote timeline data from index part")?;
 
-                entries
-                    .entry(tenant_id)
-                    .or_default()
-                    .insert(timeline_id, remote_timeline);
+                        entries
+                            .entry(tenant_id)
+                            .or_default()
+                            .insert(timeline_id, remote_timeline);
+                    }
+                },
             }
         }
 

pageserver/src/storage_sync/upload.rs

@@ -4,7 +4,7 @@ use std::{fmt::Debug, path::PathBuf};
 
 use anyhow::Context;
 use futures::stream::{FuturesUnordered, StreamExt};
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
 use remote_storage::RemoteStorage;
 use tokio::fs;
 use tracing::{debug, error, info, warn};
@@ -20,14 +20,14 @@ use crate::{
 };
 use metrics::{register_int_counter_vec, IntCounterVec};
 
-lazy_static! {
-    static ref NO_LAYERS_UPLOAD: IntCounterVec = register_int_counter_vec!(
+static NO_LAYERS_UPLOAD: Lazy<IntCounterVec> = Lazy::new(|| {
+    register_int_counter_vec!(
         "pageserver_remote_storage_no_layers_uploads_total",
         "Number of skipped uploads due to no layers",
         &["tenant_id", "timeline_id"],
     )
-    .expect("failed to register pageserver no layers upload vec");
-}
+    .expect("failed to register pageserver no layers upload vec")
+});
 
 /// Serializes and uploads the given index part data to the remote storage.
 pub(super) async fn upload_index_part<P, S>(

pageserver/src/tenant_config.rs

@@ -23,6 +23,7 @@ pub mod defaults {
     // which is good for now to trigger bugs.
     // This parameter actually determines L0 layer file size.
     pub const DEFAULT_CHECKPOINT_DISTANCE: u64 = 256 * 1024 * 1024;
+    pub const DEFAULT_CHECKPOINT_TIMEOUT: &str = "10 m";
 
     // Target file size, when creating image and delta layers.
     // This parameter determines L1 layer file size.
@@ -48,6 +49,9 @@ pub struct TenantConf {
     // page server crashes.
     // This parameter actually determines L0 layer file size.
     pub checkpoint_distance: u64,
+    // Inmemory layer is also flushed at least once in checkpoint_timeout to
+    // eventually upload WAL after activity is stopped.
+    pub checkpoint_timeout: Duration,
     // Target file size, when creating image and delta layers.
     // This parameter determines L1 layer file size.
     pub compaction_target_size: u64,
@@ -90,6 +94,7 @@ pub struct TenantConf {
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, Default)]
 pub struct TenantConfOpt {
     pub checkpoint_distance: Option<u64>,
+    pub checkpoint_timeout: Option<Duration>,
     pub compaction_target_size: Option<u64>,
     #[serde(with = "humantime_serde")]
     pub compaction_period: Option<Duration>,
@@ -113,6 +118,9 @@ impl TenantConfOpt {
             checkpoint_distance: self
                 .checkpoint_distance
                 .unwrap_or(global_conf.checkpoint_distance),
+            checkpoint_timeout: self
+                .checkpoint_timeout
+                .unwrap_or(global_conf.checkpoint_timeout),
             compaction_target_size: self
                 .compaction_target_size
                 .unwrap_or(global_conf.compaction_target_size),
@@ -142,6 +150,9 @@ impl TenantConfOpt {
         if let Some(checkpoint_distance) = other.checkpoint_distance {
             self.checkpoint_distance = Some(checkpoint_distance);
         }
+        if let Some(checkpoint_timeout) = other.checkpoint_timeout {
+            self.checkpoint_timeout = Some(checkpoint_timeout);
+        }
         if let Some(compaction_target_size) = other.compaction_target_size {
             self.compaction_target_size = Some(compaction_target_size);
         }
@@ -181,6 +192,8 @@ impl TenantConf {
 
         TenantConf {
             checkpoint_distance: DEFAULT_CHECKPOINT_DISTANCE,
+            checkpoint_timeout: humantime::parse_duration(DEFAULT_CHECKPOINT_TIMEOUT)
+                .expect("cannot parse default checkpoint timeout"),
             compaction_target_size: DEFAULT_COMPACTION_TARGET_SIZE,
             compaction_period: humantime::parse_duration(DEFAULT_COMPACTION_PERIOD)
                 .expect("cannot parse default compaction period"),
@@ -212,6 +225,7 @@ impl TenantConf {
     pub fn dummy_conf() -> Self {
         TenantConf {
             checkpoint_distance: defaults::DEFAULT_CHECKPOINT_DISTANCE,
+            checkpoint_timeout: Duration::from_secs(600),
             compaction_target_size: 4 * 1024 * 1024,
             compaction_period: Duration::from_secs(10),
             compaction_threshold: defaults::DEFAULT_COMPACTION_THRESHOLD,

pageserver/src/tenant_mgr.rs

@@ -2,8 +2,8 @@
 //! page server.
 
 use crate::config::PageServerConf;
+use crate::http::models::TenantInfo;
 use crate::layered_repository::{load_metadata, LayeredRepository};
-use crate::pgdatadir_mapping::DatadirTimeline;
 use crate::repository::Repository;
 use crate::storage_sync::index::{RemoteIndex, RemoteTimelineIndex};
 use crate::storage_sync::{self, LocalTimelineInitStatus, SyncStartupData};
@@ -12,10 +12,9 @@ use crate::thread_mgr::ThreadKind;
 use crate::timelines::CreateRepo;
 use crate::walredo::PostgresRedoManager;
 use crate::{thread_mgr, timelines, walreceiver};
-use crate::{DatadirTimelineImpl, RepositoryImpl};
+use crate::{RepositoryImpl, TimelineImpl};
 use anyhow::Context;
 use serde::{Deserialize, Serialize};
-use serde_with::{serde_as, DisplayFromStr};
 use std::collections::hash_map::Entry;
 use std::collections::{HashMap, HashSet};
 use std::fmt;
@@ -28,23 +27,25 @@ use utils::zid::{ZTenantId, ZTenantTimelineId, ZTimelineId};
 
 mod tenants_state {
     use anyhow::ensure;
+    use once_cell::sync::Lazy;
     use std::{
         collections::HashMap,
         sync::{RwLock, RwLockReadGuard, RwLockWriteGuard},
     };
     use tokio::sync::mpsc;
     use tracing::{debug, error};
-
     use utils::zid::ZTenantId;
 
     use crate::tenant_mgr::{LocalTimelineUpdate, Tenant};
 
-    lazy_static::lazy_static! {
-        static ref TENANTS: RwLock<HashMap<ZTenantId, Tenant>> = RwLock::new(HashMap::new());
-        /// Sends updates to the local timelines (creation and deletion) to the WAL receiver,
-        /// so that it can enable/disable corresponding processes.
-        static ref TIMELINE_UPDATE_SENDER: RwLock<Option<mpsc::UnboundedSender<LocalTimelineUpdate>>> = RwLock::new(None);
-    }
+    static TENANTS: Lazy<RwLock<HashMap<ZTenantId, Tenant>>> =
+        Lazy::new(|| RwLock::new(HashMap::new()));
+
+    /// Sends updates to the local timelines (creation and deletion) to the WAL receiver,
+    /// so that it can enable/disable corresponding processes.
+    static TIMELINE_UPDATE_SENDER: Lazy<
+        RwLock<Option<mpsc::UnboundedSender<LocalTimelineUpdate>>>,
+    > = Lazy::new(|| RwLock::new(None));
 
     pub(super) fn read_tenants() -> RwLockReadGuard<'static, HashMap<ZTenantId, Tenant>> {
         TENANTS
@@ -101,7 +102,7 @@ struct Tenant {
     ///
     /// Local timelines have more metadata that's loaded into memory,
     /// that is located in the `repo.timelines` field, [`crate::layered_repository::LayeredTimelineEntry`].
-    local_timelines: HashMap<ZTimelineId, Arc<DatadirTimelineImpl>>,
+    local_timelines: HashMap<ZTimelineId, Arc<<RepositoryImpl as Repository>::Timeline>>,
 }
 
 #[derive(Debug, Serialize, Deserialize, Clone, Copy, PartialEq, Eq)]
@@ -178,7 +179,7 @@ pub enum LocalTimelineUpdate {
     },
     Attach {
         id: ZTenantTimelineId,
-        datadir: Arc<DatadirTimelineImpl>,
+        datadir: Arc<<RepositoryImpl as Repository>::Timeline>,
     },
 }
 
@@ -382,7 +383,7 @@ pub fn get_repository_for_tenant(tenant_id: ZTenantId) -> anyhow::Result<Arc<Rep
 pub fn get_local_timeline_with_load(
     tenant_id: ZTenantId,
     timeline_id: ZTimelineId,
-) -> anyhow::Result<Arc<DatadirTimelineImpl>> {
+) -> anyhow::Result<Arc<TimelineImpl>> {
     let mut m = tenants_state::write_tenants();
     let tenant = m
         .get_mut(&tenant_id)
@@ -489,34 +490,23 @@ pub fn detach_tenant(conf: &'static PageServerConf, tenant_id: ZTenantId) -> any
 fn load_local_timeline(
     repo: &RepositoryImpl,
     timeline_id: ZTimelineId,
-) -> anyhow::Result<Arc<DatadirTimeline<LayeredRepository>>> {
+) -> anyhow::Result<Arc<TimelineImpl>> {
     let inmem_timeline = repo.get_timeline_load(timeline_id).with_context(|| {
         format!("Inmem timeline {timeline_id} not found in tenant's repository")
     })?;
-    let repartition_distance = repo.get_checkpoint_distance() / 10;
-    let page_tline = Arc::new(DatadirTimelineImpl::new(
-        inmem_timeline,
-        repartition_distance,
-    ));
-    page_tline.init_logical_size()?;
+    inmem_timeline.init_logical_size()?;
 
     tenants_state::try_send_timeline_update(LocalTimelineUpdate::Attach {
         id: ZTenantTimelineId::new(repo.tenant_id(), timeline_id),
-        datadir: Arc::clone(&page_tline),
+        datadir: Arc::clone(&inmem_timeline),
     });
 
-    Ok(page_tline)
-}
-
-#[serde_as]
-#[derive(Serialize, Deserialize, Clone)]
-pub struct TenantInfo {
-    #[serde_as(as = "DisplayFromStr")]
-    pub id: ZTenantId,
-    pub state: Option<TenantState>,
-    pub has_in_progress_downloads: Option<bool>,
+    Ok(inmem_timeline)
 }
 
+///
+/// Get list of tenants, for the mgmt API
+///
 pub fn list_tenants(remote_index: &RemoteTimelineIndex) -> Vec<TenantInfo> {
     tenants_state::read_tenants()
         .iter()
@@ -532,6 +522,7 @@ pub fn list_tenants(remote_index: &RemoteTimelineIndex) -> Vec<TenantInfo> {
             TenantInfo {
                 id: *id,
                 state: Some(tenant.state),
+                current_physical_size: None,
                 has_in_progress_downloads,
             }
         })

pageserver/src/tenant_tasks.rs

@@ -120,6 +120,10 @@ pub fn init_tenant_task_pool() -> anyhow::Result<()> {
     let runtime = tokio::runtime::Builder::new_multi_thread()
         .thread_name("tenant-task-worker")
         .enable_all()
+        .on_thread_start(|| {
+            thread_mgr::register(ThreadKind::TenantTaskWorker, "tenant-task-worker")
+        })
+        .on_thread_stop(thread_mgr::deregister)
         .build()?;
 
     let (gc_send, mut gc_recv) = mpsc::channel::<ZTenantId>(100);

pageserver/src/thread_mgr.rs

@@ -45,21 +45,20 @@ use tokio::sync::watch;
 
 use tracing::{debug, error, info, warn};
 
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
 
 use utils::zid::{ZTenantId, ZTimelineId};
 
 use crate::shutdown_pageserver;
 
-lazy_static! {
-    /// Each thread that we track is associated with a "thread ID". It's just
-    /// an increasing number that we assign, not related to any system thread
-    /// id.
-    static ref NEXT_THREAD_ID: AtomicU64 = AtomicU64::new(1);
+/// Each thread that we track is associated with a "thread ID". It's just
+/// an increasing number that we assign, not related to any system thread
+/// id.
+static NEXT_THREAD_ID: Lazy<AtomicU64> = Lazy::new(|| AtomicU64::new(1));
 
-    /// Global registry of threads
-    static ref THREADS: Mutex<HashMap<u64, Arc<PageServerThread>>> = Mutex::new(HashMap::new());
-}
+/// Global registry of threads
+static THREADS: Lazy<Mutex<HashMap<u64, Arc<PageServerThread>>>> =
+    Lazy::new(|| Mutex::new(HashMap::new()));
 
 // There is a Tokio watch channel for each thread, which can be used to signal the
 // thread that it needs to shut down. This thread local variable holds the receiving
@@ -97,6 +96,9 @@ pub enum ThreadKind {
     // Thread that schedules new compaction and gc jobs
     TenantTaskManager,
 
+    // Worker thread for tenant tasks thread pool
+    TenantTaskWorker,
+
     // Thread that flushes frozen in-memory layers to disk
     LayerFlushThread,
 
@@ -105,18 +107,20 @@ pub enum ThreadKind {
     StorageSync,
 }
 
+#[derive(Default)]
 struct MutableThreadState {
     /// Tenant and timeline that this thread is associated with.
     tenant_id: Option<ZTenantId>,
     timeline_id: Option<ZTimelineId>,
 
     /// Handle for waiting for the thread to exit. It can be None, if the
-    /// the thread has already exited.
+    /// the thread has already exited. OR if this thread is managed externally
+    /// and was not spawned through thread_mgr.rs::spawn function.
     join_handle: Option<JoinHandle<()>>,
 }
 
 struct PageServerThread {
-    _thread_id: u64,
+    thread_id: u64,
 
     kind: ThreadKind,
 
@@ -147,7 +151,7 @@ where
     let (shutdown_tx, shutdown_rx) = watch::channel(());
     let thread_id = NEXT_THREAD_ID.fetch_add(1, Ordering::Relaxed);
     let thread = Arc::new(PageServerThread {
-        _thread_id: thread_id,
+        thread_id,
         kind,
         name: name.to_string(),
         shutdown_requested: AtomicBool::new(false),
@@ -315,8 +319,10 @@ pub fn shutdown_threads(
             drop(thread_mut);
             let _ = join_handle.join();
         } else {
-            // The thread had not even fully started yet. Or it was shut down
-            // concurrently and already exited
+            // Possibly one of:
+            //  * The thread had not even fully started yet.
+            //  * It was shut down concurrently and already exited
+            //  * Is managed through `register`/`deregister` fns without providing a join handle
         }
     }
 }
@@ -348,3 +354,56 @@ pub fn is_shutdown_requested() -> bool {
         }
     })
 }
+
+/// Needed to register threads that were not spawned through spawn function.
+/// For example tokio blocking threads. This function is expected to be used
+/// in tandem with `deregister`.
+/// NOTE: threads registered through this function cannot be joined
+pub fn register(kind: ThreadKind, name: &str) {
+    CURRENT_THREAD.with(|ct| {
+        let mut borrowed = ct.borrow_mut();
+        if borrowed.is_some() {
+            panic!("thread already registered")
+        };
+        let (shutdown_tx, shutdown_rx) = watch::channel(());
+        let thread_id = NEXT_THREAD_ID.fetch_add(1, Ordering::Relaxed);
+
+        let thread = Arc::new(PageServerThread {
+            thread_id,
+            kind,
+            name: name.to_owned(),
+            shutdown_requested: AtomicBool::new(false),
+            shutdown_tx,
+            mutable: Mutex::new(MutableThreadState {
+                tenant_id: None,
+                timeline_id: None,
+                join_handle: None,
+            }),
+        });
+
+        *borrowed = Some(Arc::clone(&thread));
+
+        SHUTDOWN_RX.with(|rx| {
+            *rx.borrow_mut() = Some(shutdown_rx);
+        });
+
+        THREADS.lock().unwrap().insert(thread_id, thread);
+    });
+}
+
+// Expected to be used in tandem with `register`. See the doc for `register` for more details
+pub fn deregister() {
+    CURRENT_THREAD.with(|ct| {
+        let mut borrowed = ct.borrow_mut();
+        let thread = match borrowed.take() {
+            Some(thread) => thread,
+            None => panic!("calling deregister on unregistered thread"),
+        };
+
+        SHUTDOWN_RX.with(|rx| {
+            *rx.borrow_mut() = None;
+        });
+
+        THREADS.lock().unwrap().remove(&thread.thread_id)
+    });
+}

pageserver/src/timelines.rs

@@ -4,8 +4,6 @@
 
 use anyhow::{bail, ensure, Context, Result};
 use postgres_ffi::ControlFileData;
-use serde::{Deserialize, Serialize};
-use serde_with::{serde_as, DisplayFromStr};
 use std::{
     fs,
     path::Path,
@@ -20,123 +18,15 @@ use utils::{
     zid::{ZTenantId, ZTimelineId},
 };
 
+use crate::tenant_mgr;
 use crate::{
-    config::PageServerConf,
-    layered_repository::metadata::TimelineMetadata,
-    repository::{LocalTimelineState, Repository},
-    storage_sync::index::RemoteIndex,
-    tenant_config::TenantConfOpt,
-    DatadirTimeline, RepositoryImpl,
+    config::PageServerConf, repository::Repository, storage_sync::index::RemoteIndex,
+    tenant_config::TenantConfOpt, RepositoryImpl, TimelineImpl,
 };
 use crate::{import_datadir, LOG_FILE_NAME};
 use crate::{layered_repository::LayeredRepository, walredo::WalRedoManager};
-use crate::{repository::RepositoryTimeline, tenant_mgr};
 use crate::{repository::Timeline, CheckpointConfig};
 
-#[serde_as]
-#[derive(Debug, Serialize, Deserialize, Clone)]
-pub struct LocalTimelineInfo {
-    #[serde_as(as = "Option<DisplayFromStr>")]
-    pub ancestor_timeline_id: Option<ZTimelineId>,
-    #[serde_as(as = "Option<DisplayFromStr>")]
-    pub ancestor_lsn: Option<Lsn>,
-    #[serde_as(as = "DisplayFromStr")]
-    pub last_record_lsn: Lsn,
-    #[serde_as(as = "Option<DisplayFromStr>")]
-    pub prev_record_lsn: Option<Lsn>,
-    #[serde_as(as = "DisplayFromStr")]
-    pub latest_gc_cutoff_lsn: Lsn,
-    #[serde_as(as = "DisplayFromStr")]
-    pub disk_consistent_lsn: Lsn,
-    pub current_logical_size: Option<usize>, // is None when timeline is Unloaded
-    pub current_logical_size_non_incremental: Option<usize>,
-    pub timeline_state: LocalTimelineState,
-}
-
-impl LocalTimelineInfo {
-    pub fn from_loaded_timeline<R: Repository>(
-        datadir_tline: &DatadirTimeline<R>,
-        include_non_incremental_logical_size: bool,
-    ) -> anyhow::Result<Self> {
-        let last_record_lsn = datadir_tline.tline.get_last_record_lsn();
-        let info = LocalTimelineInfo {
-            ancestor_timeline_id: datadir_tline.tline.get_ancestor_timeline_id(),
-            ancestor_lsn: {
-                match datadir_tline.tline.get_ancestor_lsn() {
-                    Lsn(0) => None,
-                    lsn @ Lsn(_) => Some(lsn),
-                }
-            },
-            disk_consistent_lsn: datadir_tline.tline.get_disk_consistent_lsn(),
-            last_record_lsn,
-            prev_record_lsn: Some(datadir_tline.tline.get_prev_record_lsn()),
-            latest_gc_cutoff_lsn: *datadir_tline.tline.get_latest_gc_cutoff_lsn(),
-            timeline_state: LocalTimelineState::Loaded,
-            current_logical_size: Some(datadir_tline.get_current_logical_size()),
-            current_logical_size_non_incremental: if include_non_incremental_logical_size {
-                Some(datadir_tline.get_current_logical_size_non_incremental(last_record_lsn)?)
-            } else {
-                None
-            },
-        };
-        Ok(info)
-    }
-
-    pub fn from_unloaded_timeline(metadata: &TimelineMetadata) -> Self {
-        LocalTimelineInfo {
-            ancestor_timeline_id: metadata.ancestor_timeline(),
-            ancestor_lsn: {
-                match metadata.ancestor_lsn() {
-                    Lsn(0) => None,
-                    lsn @ Lsn(_) => Some(lsn),
-                }
-            },
-            disk_consistent_lsn: metadata.disk_consistent_lsn(),
-            last_record_lsn: metadata.disk_consistent_lsn(),
-            prev_record_lsn: metadata.prev_record_lsn(),
-            latest_gc_cutoff_lsn: metadata.latest_gc_cutoff_lsn(),
-            timeline_state: LocalTimelineState::Unloaded,
-            current_logical_size: None,
-            current_logical_size_non_incremental: None,
-        }
-    }
-
-    pub fn from_repo_timeline<T>(
-        tenant_id: ZTenantId,
-        timeline_id: ZTimelineId,
-        repo_timeline: &RepositoryTimeline<T>,
-        include_non_incremental_logical_size: bool,
-    ) -> anyhow::Result<Self> {
-        match repo_timeline {
-            RepositoryTimeline::Loaded(_) => {
-                let datadir_tline =
-                    tenant_mgr::get_local_timeline_with_load(tenant_id, timeline_id)?;
-                Self::from_loaded_timeline(&datadir_tline, include_non_incremental_logical_size)
-            }
-            RepositoryTimeline::Unloaded { metadata } => Ok(Self::from_unloaded_timeline(metadata)),
-        }
-    }
-}
-
-#[serde_as]
-#[derive(Debug, Serialize, Deserialize, Clone)]
-pub struct RemoteTimelineInfo {
-    #[serde_as(as = "DisplayFromStr")]
-    pub remote_consistent_lsn: Lsn,
-    pub awaits_download: bool,
-}
-
-#[serde_as]
-#[derive(Debug, Serialize, Deserialize, Clone)]
-pub struct TimelineInfo {
-    #[serde_as(as = "DisplayFromStr")]
-    pub tenant_id: ZTenantId,
-    #[serde_as(as = "DisplayFromStr")]
-    pub timeline_id: ZTimelineId,
-    pub local: Option<LocalTimelineInfo>,
-    pub remote: Option<RemoteTimelineInfo>,
-}
-
 #[derive(Debug, Clone, Copy)]
 pub struct PointInTime {
     pub timeline_id: ZTimelineId,
@@ -298,19 +188,18 @@ fn bootstrap_timeline<R: Repository>(
     // Initdb lsn will be equal to last_record_lsn which will be set after import.
     // Because we know it upfront avoid having an option or dummy zero value by passing it to create_empty_timeline.
     let timeline = repo.create_empty_timeline(tli, lsn)?;
-    let mut page_tline: DatadirTimeline<R> = DatadirTimeline::new(timeline, u64::MAX);
-    import_datadir::import_timeline_from_postgres_datadir(&pgdata_path, &mut page_tline, lsn)?;
+    import_datadir::import_timeline_from_postgres_datadir(&pgdata_path, &*timeline, lsn)?;
 
     fail::fail_point!("before-checkpoint-new-timeline", |_| {
         bail!("failpoint before-checkpoint-new-timeline");
     });
 
-    page_tline.tline.checkpoint(CheckpointConfig::Forced)?;
+    timeline.checkpoint(CheckpointConfig::Forced)?;
 
     info!(
         "created root timeline {} timeline.lsn {}",
         tli,
-        page_tline.tline.get_last_record_lsn()
+        timeline.get_last_record_lsn()
     );
 
     // Remove temp dir. We don't need it anymore
@@ -319,36 +208,22 @@ fn bootstrap_timeline<R: Repository>(
     Ok(())
 }
 
-pub(crate) fn get_local_timelines(
-    tenant_id: ZTenantId,
-    include_non_incremental_logical_size: bool,
-) -> Result<Vec<(ZTimelineId, LocalTimelineInfo)>> {
-    let repo = tenant_mgr::get_repository_for_tenant(tenant_id)
-        .with_context(|| format!("Failed to get repo for tenant {}", tenant_id))?;
-    let repo_timelines = repo.list_timelines();
-
-    let mut local_timeline_info = Vec::with_capacity(repo_timelines.len());
-    for (timeline_id, repository_timeline) in repo_timelines {
-        local_timeline_info.push((
-            timeline_id,
-            LocalTimelineInfo::from_repo_timeline(
-                tenant_id,
-                timeline_id,
-                &repository_timeline,
-                include_non_incremental_logical_size,
-            )?,
-        ))
-    }
-    Ok(local_timeline_info)
-}
-
+///
+/// Create a new timeline.
+///
+/// Returns the new timeline ID and reference to its Timeline object.
+///
+/// If the caller specified the timeline ID to use (`new_timeline_id`), and timeline with
+/// the same timeline ID already exists, returns None. If `new_timeline_id` is not given,
+/// a new unique ID is generated.
+///
 pub(crate) fn create_timeline(
     conf: &'static PageServerConf,
     tenant_id: ZTenantId,
     new_timeline_id: Option<ZTimelineId>,
     ancestor_timeline_id: Option<ZTimelineId>,
     mut ancestor_start_lsn: Option<Lsn>,
-) -> Result<Option<TimelineInfo>> {
+) -> Result<Option<(ZTimelineId, Arc<TimelineImpl>)>> {
     let new_timeline_id = new_timeline_id.unwrap_or_else(ZTimelineId::generate);
     let repo = tenant_mgr::get_repository_for_tenant(tenant_id)?;
 
@@ -357,7 +232,7 @@ pub(crate) fn create_timeline(
         return Ok(None);
     }
 
-    let new_timeline_info = match ancestor_timeline_id {
+    match ancestor_timeline_id {
         Some(ancestor_timeline_id) => {
             let ancestor_timeline = repo
                 .get_timeline_load(ancestor_timeline_id)
@@ -385,26 +260,13 @@ pub(crate) fn create_timeline(
                 }
             }
 
-            repo.branch_timeline(ancestor_timeline_id, new_timeline_id, ancestor_start_lsn)?;
-            // load the timeline into memory
-            let loaded_timeline =
-                tenant_mgr::get_local_timeline_with_load(tenant_id, new_timeline_id)?;
-            LocalTimelineInfo::from_loaded_timeline(&loaded_timeline, false)
-                .context("cannot fill timeline info")?
-        }
-        None => {
-            bootstrap_timeline(conf, tenant_id, new_timeline_id, repo.as_ref())?;
-            // load the timeline into memory
-            let new_timeline =
-                tenant_mgr::get_local_timeline_with_load(tenant_id, new_timeline_id)?;
-            LocalTimelineInfo::from_loaded_timeline(&new_timeline, false)
-                .context("cannot fill timeline info")?
+            repo.branch_timeline(ancestor_timeline_id, new_timeline_id, ancestor_start_lsn)?
         }
+        None => bootstrap_timeline(conf, tenant_id, new_timeline_id, repo.as_ref())?,
     };
-    Ok(Some(TimelineInfo {
-        tenant_id,
-        timeline_id: new_timeline_id,
-        local: Some(new_timeline_info),
-        remote: None,
-    }))
+
+    // load the timeline into memory
+    let loaded_timeline = tenant_mgr::get_local_timeline_with_load(tenant_id, new_timeline_id)?;
+
+    Ok(Some((new_timeline_id, loaded_timeline)))
 }

pageserver/src/virtual_file.rs

@@ -10,7 +10,7 @@
 //! This is similar to PostgreSQL's virtual file descriptor facility in
 //! src/backend/storage/file/fd.c
 //!
-use lazy_static::lazy_static;
+use once_cell::sync::Lazy;
 use once_cell::sync::OnceCell;
 use std::fs::{File, OpenOptions};
 use std::io::{Error, ErrorKind, Read, Seek, SeekFrom, Write};
@@ -32,23 +32,24 @@ const STORAGE_IO_TIME_BUCKETS: &[f64] = &[
     1.0,      // 1 sec
 ];
 
-lazy_static! {
-    static ref STORAGE_IO_TIME: HistogramVec = register_histogram_vec!(
+static STORAGE_IO_TIME: Lazy<HistogramVec> = Lazy::new(|| {
+    register_histogram_vec!(
         "pageserver_io_operations_seconds",
         "Time spent in IO operations",
         &["operation", "tenant_id", "timeline_id"],
         STORAGE_IO_TIME_BUCKETS.into()
     )
-    .expect("failed to define a metric");
-}
-lazy_static! {
-    static ref STORAGE_IO_SIZE: IntGaugeVec = register_int_gauge_vec!(
+    .expect("failed to define a metric")
+});
+
+static STORAGE_IO_SIZE: Lazy<IntGaugeVec> = Lazy::new(|| {
+    register_int_gauge_vec!(
         "pageserver_io_operations_bytes_total",
         "Total amount of bytes read/written in IO operations",
         &["operation", "tenant_id", "timeline_id"]
     )
-    .expect("failed to define a metric");
-}
+    .expect("failed to define a metric")
+});
 
 ///
 /// A virtual file descriptor. You can use this just like std::fs::File, but internally