bench: lower attach concurrency

it has been observed that we get transaction serialization issues otherwise. evidence: https://neon-github-public-dev.s3.amazonaws.com/reports/main/10402057346/index.html#suites/c62b105f3a4f00dd6be4ad88810e0e02/7c03011d9dbd5ec4/
2026-05-23 16:10:37 +00:00 · 2024-08-15 12:24:38 +00:00
230 changed files with 2861 additions and 7417 deletions
--- a/.config/hakari.toml
+++ b/.config/hakari.toml
@@ -23,30 +23,10 @@ platforms = [
 ]

 [final-excludes]
-workspace-members = [
-    # vm_monitor benefits from the same Cargo.lock as the rest of our artifacts, but
-    # it is built primarly in separate repo neondatabase/autoscaling and thus is excluded
-    # from depending on workspace-hack because most of the dependencies are not used.
-    "vm_monitor",
-    # All of these exist in libs and are not usually built independently.
-    # Putting workspace hack there adds a bottleneck for cargo builds.
-    "compute_api",
-    "consumption_metrics",
-    "desim",
-    "metrics",
-    "pageserver_api",
-    "postgres_backend",
-    "postgres_connection",
-    "postgres_ffi",
-    "pq_proto",
-    "remote_storage",
-    "safekeeper_api",
-    "tenant_size_model",
-    "tracing-utils",
-    "utils",
-    "wal_craft",
-    "walproposer",
-]
+# vm_monitor benefits from the same Cargo.lock as the rest of our artifacts, but
+# it is built primarly in separate repo neondatabase/autoscaling and thus is excluded
+# from depending on workspace-hack because most of the dependencies are not used.
+workspace-members = ["vm_monitor"]

 # Write out exact versions rather than a semver range. (Defaults to false.)
 # exact-versions = true
--- a/.github/actionlint.yml
+++ b/.github/actionlint.yml
@@ -1,6 +1,7 @@
 self-hosted-runner:
  labels:
    - arm64
+    - gen3
    - large
    - large-arm64
    - small
--- a/.github/actions/run-python-test-set/action.yml
+++ b/.github/actions/run-python-test-set/action.yml
@@ -43,7 +43,7 @@ inputs:
  pg_version:
    description: 'Postgres version to use for tests'
    required: false
-    default: 'v16'
+    default: 'v14'
  benchmark_durations:
    description: 'benchmark durations JSON'
    required: false
@@ -83,6 +83,7 @@ runs:
      uses: actions/checkout@v4
      with:
        submodules: true
+        fetch-depth: 1

    - name: Cache poetry deps
      uses: actions/cache@v4
@@ -169,8 +170,10 @@ runs:
          EXTRA_PARAMS="--durations-path $TEST_OUTPUT/benchmark_durations.json $EXTRA_PARAMS"
        fi

-        if [[ $BUILD_TYPE == "debug" && $RUNNER_ARCH == 'X64' ]]; then
+        if [[ "${{ inputs.build_type }}" == "debug" ]]; then
          cov_prefix=(scripts/coverage "--profraw-prefix=$GITHUB_JOB" --dir=/tmp/coverage run)
+        elif [[ "${{ inputs.build_type }}" == "release" ]]; then
+          cov_prefix=()
        else
          cov_prefix=()
        fi
--- a/.github/workflows/_benchmarking_preparation.yml
+++ b/.github/workflows/_benchmarking_preparation.yml
@@ -48,8 +48,6 @@ jobs:

        echo "connstr=${CONNSTR}" >> $GITHUB_OUTPUT  

-    - uses: actions/checkout@v4
-
    - name: Download Neon artifact
      uses: ./.github/actions/download
      with:
--- a/.github/workflows/_build-and-test-locally.yml
+++ b/.github/workflows/_build-and-test-locally.yml
@@ -70,6 +70,7 @@ jobs:
      - uses: actions/checkout@v4
        with:
          submodules: true
+          fetch-depth: 1

      - name: Set pg 14 revision for caching
        id: pg_v14_rev
@@ -94,16 +95,11 @@ jobs:
      # We run tests with addtional features, that are turned off by default (e.g. in release builds), see
      # corresponding Cargo.toml files for their descriptions.
      - name: Set env variables
-        env:
-          ARCH: ${{ inputs.arch }}
        run: |
          CARGO_FEATURES="--features testing"
-          if [[ $BUILD_TYPE == "debug" && $ARCH == 'x64' ]]; then
+          if [[ $BUILD_TYPE == "debug" ]]; then
            cov_prefix="scripts/coverage --profraw-prefix=$GITHUB_JOB --dir=/tmp/coverage run"
            CARGO_FLAGS="--locked"
-          elif [[ $BUILD_TYPE == "debug" ]]; then
-            cov_prefix=""
-            CARGO_FLAGS="--locked"
          elif [[ $BUILD_TYPE == "release" ]]; then
            cov_prefix=""
            CARGO_FLAGS="--locked --release"
@@ -163,8 +159,6 @@ jobs:
      # Do install *before* running rust tests because they might recompile the
      # binaries with different features/flags.
      - name: Install rust binaries
-        env:
-          ARCH: ${{ inputs.arch }}
        run: |
          # Install target binaries
          mkdir -p /tmp/neon/bin/
@@ -179,7 +173,7 @@ jobs:
          done

          # Install test executables and write list of all binaries (for code coverage)
-          if [[ $BUILD_TYPE == "debug" && $ARCH == 'x64' ]]; then
+          if [[ $BUILD_TYPE == "debug" ]]; then
            # Keep bloated coverage data files away from the rest of the artifact
            mkdir -p /tmp/coverage/

@@ -214,12 +208,10 @@ jobs:
          export LD_LIBRARY_PATH

          #nextest does not yet support running doctests
-          ${cov_prefix} cargo test --doc $CARGO_FLAGS $CARGO_FEATURES
+          cargo test --doc $CARGO_FLAGS $CARGO_FEATURES

          for io_engine in std-fs tokio-epoll-uring ; do
-            for io_buffer_alignment in 0 1 512 ; do
-              NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOENGINE=$io_engine NEON_PAGESERVER_UNIT_TEST_IO_BUFFER_ALIGNMENT=$io_buffer_alignment ${cov_prefix} cargo nextest run $CARGO_FLAGS $CARGO_FEATURES
-            done
+            NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOENGINE=$io_engine ${cov_prefix} cargo nextest run $CARGO_FLAGS $CARGO_FEATURES
          done

          # Run separate tests for real S3
@@ -252,8 +244,8 @@ jobs:
        uses: ./.github/actions/save-coverage-data

  regress-tests:
-    # Don't run regression tests on debug arm64 builds
-    if: inputs.build-type != 'debug' || inputs.arch != 'arm64'
+    # Run test on x64 only
+    if: inputs.arch == 'x64'
    needs: [ build-neon ]
    runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', inputs.arch == 'arm64' && 'large-arm64' || 'large')) }}
    container:
@@ -271,6 +263,7 @@ jobs:
      - uses: actions/checkout@v4
        with:
          submodules: true
+          fetch-depth: 1

      - name: Pytest regression tests
        uses: ./.github/actions/run-python-test-set
--- a/.github/workflows/actionlint.yml
+++ b/.github/workflows/actionlint.yml
@@ -44,7 +44,7 @@ jobs:
            grep -ERl $PAT .github/workflows |\
            while read -r f
            do
-              l=$(grep -nE $PAT $f | awk -F: '{print $1}' | head -1)
+              l=$(grep -nE $PAT .github/workflows/release.yml | awk -F: '{print $1}' | head -1)
              echo "::error file=$f,line=$l::Please use 'ubuntu-22.04' instead of 'ubuntu-latest'"
            done
            exit 1
--- a/.github/workflows/benchmarking.yml
+++ b/.github/workflows/benchmarking.yml
@@ -96,7 +96,7 @@ jobs:
      uses: aws-actions/configure-aws-credentials@v4
      with:
        aws-region: eu-central-1
-        role-to-assume: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
+        role-to-assume: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}  
        role-duration-seconds: 18000 # 5 hours

    - name: Download Neon artifact
@@ -146,7 +146,6 @@ jobs:
        api_key: ${{ secrets.NEON_STAGING_API_KEY }}

    - name: Create Allure report
-      id: create-allure-report
      if: ${{ !cancelled() }}
      uses: ./.github/actions/allure-report-generate

@@ -155,10 +154,7 @@ jobs:
      uses: slackapi/slack-github-action@v1
      with:
        channel-id: "C033QLM5P7D" # dev-staging-stream
-        slack-message: |
-          Periodic perf testing: ${{ job.status }}
-          <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
-          <${{ steps.create-allure-report.outputs.report-url }}|Allure report>
+        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 }}

@@ -180,7 +176,7 @@ jobs:
    steps:
    - uses: actions/checkout@v4

-
+    
    - name: Download Neon artifact
      uses: ./.github/actions/download
      with:
@@ -219,23 +215,15 @@ jobs:
        NEON_API_KEY: ${{ secrets.NEON_STAGING_API_KEY }}

    - name: Create Allure report
-      id: create-allure-report
      if: ${{ !cancelled() }}
      uses: ./.github/actions/allure-report-generate
-      with:
-        store-test-results-into-db: true
-      env:
-        REGRESS_TEST_RESULT_CONNSTR_NEW: ${{ secrets.REGRESS_TEST_RESULT_CONNSTR_NEW }}

    - name: Post to a Slack channel
      if: ${{ github.event.schedule && failure() }}
      uses: slackapi/slack-github-action@v1
      with:
-        channel-id: "C06T9AMNDQQ" # on-call-compute-staging-stream
-        slack-message: |
-          Periodic replication testing: ${{ job.status }}
-          <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
-          <${{ steps.create-allure-report.outputs.report-url }}|Allure report>
+        channel-id: "C033QLM5P7D" # dev-staging-stream
+        slack-message: "Periodic replication testing: ${{ job.status }}\n${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
      env:
        SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}

@@ -337,7 +325,7 @@ jobs:
  prepare_AWS_RDS_databases:
    uses: ./.github/workflows/_benchmarking_preparation.yml
    secrets: inherit
-
+  
  pgbench-compare:
    if: ${{ github.event.inputs.run_only_pgvector_tests == 'false' || github.event.inputs.run_only_pgvector_tests == null }}
    needs: [ generate-matrices, prepare_AWS_RDS_databases ]
@@ -377,7 +365,7 @@ jobs:
        aws-region: eu-central-1
        role-to-assume: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
        role-duration-seconds: 18000 # 5 hours
-
+        
    - name: Download Neon artifact
      uses: ./.github/actions/download
      with:
@@ -472,7 +460,6 @@ jobs:
        api_key: ${{ secrets.NEON_STAGING_API_KEY }}

    - name: Create Allure report
-      id: create-allure-report
      if: ${{ !cancelled() }}
      uses: ./.github/actions/allure-report-generate

@@ -481,10 +468,7 @@ jobs:
      uses: slackapi/slack-github-action@v1
      with:
        channel-id: "C033QLM5P7D" # dev-staging-stream
-        slack-message: |
-          Periodic perf testing on ${{ matrix.platform }}: ${{ job.status }}
-          <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
-          <${{ steps.create-allure-report.outputs.report-url }}|Allure report>
+        slack-message: "Periodic perf testing ${{ matrix.platform }}: ${{ job.status }}\n${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
      env:
        SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}

@@ -558,7 +542,7 @@ jobs:
        esac

        echo "connstr=${CONNSTR}" >> $GITHUB_OUTPUT
-
+        
    - name: Configure AWS credentials # necessary on Azure runners to read/write from/to S3
      uses: aws-actions/configure-aws-credentials@v4
      with:
@@ -593,9 +577,8 @@ jobs:
        BENCHMARK_CONNSTR: ${{ steps.set-up-connstr.outputs.connstr }}
        VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
        PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
-
+    
    - name: Create Allure report
-      id: create-allure-report
      if: ${{ !cancelled() }}
      uses: ./.github/actions/allure-report-generate

@@ -604,10 +587,7 @@ jobs:
      uses: slackapi/slack-github-action@v1
      with:
        channel-id: "C033QLM5P7D" # dev-staging-stream
-        slack-message: |
-          Periodic perf testing on ${{ env.PLATFORM }}: ${{ job.status }}
-          <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
-          <${{ steps.create-allure-report.outputs.report-url }}|Allure report>
+        slack-message: "Periodic perf testing ${PLATFORM}: ${{ job.status }}\n${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
      env:
        SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}

@@ -690,7 +670,6 @@ jobs:
        TEST_OLAP_SCALE: 10

    - name: Create Allure report
-      id: create-allure-report
      if: ${{ !cancelled() }}
      uses: ./.github/actions/allure-report-generate

@@ -699,10 +678,7 @@ jobs:
      uses: slackapi/slack-github-action@v1
      with:
        channel-id: "C033QLM5P7D" # dev-staging-stream
-        slack-message: |
-          Periodic OLAP perf testing on ${{ matrix.platform }}: ${{ job.status }}
-          <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
-          <${{ steps.create-allure-report.outputs.report-url }}|Allure report>
+        slack-message: "Periodic OLAP perf testing ${{ matrix.platform }}: ${{ job.status }}\n${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
      env:
        SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}

@@ -788,7 +764,6 @@ jobs:
        TEST_OLAP_SCALE: ${{ matrix.scale }}

    - name: Create Allure report
-      id: create-allure-report
      if: ${{ !cancelled() }}
      uses: ./.github/actions/allure-report-generate

@@ -797,10 +772,7 @@ jobs:
      uses: slackapi/slack-github-action@v1
      with:
        channel-id: "C033QLM5P7D" # dev-staging-stream
-        slack-message: |
-          Periodic TPC-H perf testing on ${{ matrix.platform }}: ${{ job.status }}
-          <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
-          <${{ steps.create-allure-report.outputs.report-url }}|Allure report>
+        slack-message: "Periodic TPC-H perf testing ${{ matrix.platform }}: ${{ job.status }}\n${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
      env:
        SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}

@@ -871,7 +843,6 @@ jobs:
        BENCHMARK_CONNSTR: ${{ steps.set-up-connstr.outputs.connstr }}

    - name: Create Allure report
-      id: create-allure-report
      if: ${{ !cancelled() }}
      uses: ./.github/actions/allure-report-generate

@@ -880,10 +851,6 @@ jobs:
      uses: slackapi/slack-github-action@v1
      with:
        channel-id: "C033QLM5P7D" # dev-staging-stream
-        slack-message: |
-          Periodic TPC-H perf testing on ${{ matrix.platform }}: ${{ job.status }}
-          <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
-          <${{ steps.create-allure-report.outputs.report-url }}|Allure report>
-
+        slack-message: "Periodic User example perf testing ${{ matrix.platform }}: ${{ job.status }}\n${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
      env:
        SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}
--- a/.github/workflows/build-build-tools-image.yml
+++ b/.github/workflows/build-build-tools-image.yml
@@ -38,7 +38,7 @@ jobs:
      matrix:
        arch: [ x64, arm64 ]

-    runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}
+    runs-on: ${{ fromJson(format('["self-hosted", "gen3", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}

    env:
      IMAGE_TAG: ${{ inputs.image-tag }}
--- a/.github/workflows/build_and_test.yml
+++ b/.github/workflows/build_and_test.yml
@@ -48,7 +48,7 @@ jobs:

  tag:
    needs: [ check-permissions ]
-    runs-on: [ self-hosted, small ]
+    runs-on: [ self-hosted, gen3, small ]
    container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/base:pinned
    outputs:
      build-tag: ${{steps.build-tag.outputs.tag}}
@@ -90,7 +90,7 @@ jobs:

  check-codestyle-python:
    needs: [ check-permissions, build-build-tools-image ]
-    runs-on: [ self-hosted, small ]
+    runs-on: [ self-hosted, gen3, small ]
    container:
      image: ${{ needs.build-build-tools-image.outputs.image }}
      credentials:
@@ -101,6 +101,9 @@ jobs:
    steps:
      - name: Checkout
        uses: actions/checkout@v4
+        with:
+          submodules: false
+          fetch-depth: 1

      - name: Cache poetry deps
        uses: actions/cache@v4
@@ -139,6 +142,7 @@ jobs:
        uses: actions/checkout@v4
        with:
          submodules: true
+          fetch-depth: 1

 #      Disabled for now
 #      - name: Restore cargo deps cache
@@ -198,9 +202,9 @@ jobs:
    strategy:
      fail-fast: false
      matrix:
-        arch: [ x64, arm64 ]
+        arch: [ x64 ]
        # Do not build or run tests in debug for release branches
-        build-type: ${{ fromJson((startsWith(github.ref_name, 'release') && github.event_name == 'push') && '["release"]' || '["debug", "release"]') }}
+        build-type: ${{ fromJson((startsWith(github.ref_name, 'release' && github.event_name == 'push')) && '["release"]' || '["debug", "release"]') }}
        include:
          - build-type: release
            arch: arm64
@@ -220,7 +224,7 @@ jobs:
    outputs:
      json: ${{ steps.get-benchmark-durations.outputs.json }}
    needs: [ check-permissions, build-build-tools-image ]
-    runs-on: [ self-hosted, small ]
+    runs-on: [ self-hosted, gen3, small ]
    container:
      image: ${{ needs.build-build-tools-image.outputs.image }}
      credentials:
@@ -253,7 +257,7 @@ jobs:
  benchmarks:
    if: github.ref_name == 'main' || contains(github.event.pull_request.labels.*.name, 'run-benchmarks')
    needs: [ check-permissions, build-and-test-locally, build-build-tools-image, get-benchmarks-durations ]
-    runs-on: [ self-hosted, small ]
+    runs-on: [ self-hosted, gen3, small ]
    container:
      image: ${{ needs.build-build-tools-image.outputs.image }}
      credentials:
@@ -280,7 +284,6 @@ jobs:
          save_perf_report: ${{ github.ref_name == 'main' }}
          extra_params: --splits 5 --group ${{ matrix.pytest_split_group }}
          benchmark_durations: ${{ needs.get-benchmarks-durations.outputs.json }}
-          pg_version: v16
        env:
          VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
          PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
@@ -299,8 +302,9 @@ jobs:
      with:
        channel-id: C060CNA47S9 # on-call-staging-storage-stream
        slack-message: |
-          Benchmarks failed on main <${{ github.event.head_commit.url }}|${{ github.sha }}>
-          <${{ needs.create-test-report.outputs.report-url }}|Allure report>
+          Benchmarks failed on main: ${{ github.event.head_commit.url }}
+
+          Allure report: ${{ needs.create-test-report.outputs.report-url }}
      env:
        SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}

@@ -310,7 +314,7 @@ jobs:
    outputs:
      report-url: ${{ steps.create-allure-report.outputs.report-url }}

-    runs-on: [ self-hosted, small ]
+    runs-on: [ self-hosted, gen3, small ]
    container:
      image: ${{ needs.build-build-tools-image.outputs.image }}
      credentials:
@@ -357,7 +361,7 @@ jobs:

  coverage-report:
    needs: [ check-permissions, build-build-tools-image, build-and-test-locally ]
-    runs-on: [ self-hosted, small ]
+    runs-on: [ self-hosted, gen3, small ]
    container:
      image: ${{ needs.build-build-tools-image.outputs.image }}
      credentials:
@@ -471,7 +475,7 @@ jobs:
      matrix:
        arch: [ x64, arm64 ]

-    runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}
+    runs-on: ${{ fromJson(format('["self-hosted", "gen3", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}

    steps:
      - name: Checkout
@@ -499,10 +503,7 @@ jobs:
      - uses: docker/build-push-action@v6
        with:
          context: .
-          # ARM-specific flags are recommended for Graviton ≥ 2, these flags are also supported by Ampere Altra (Azure)
-          # https://github.com/aws/aws-graviton-getting-started/blob/57dc813626d0266f1cc12ef83474745bb1f31fb4/rust.md
          build-args: |
-            ADDITIONAL_RUSTFLAGS=${{ matrix.arch == 'arm64' && '-Ctarget-feature=+lse -Ctarget-cpu=neoverse-n1' || '' }}
            GIT_VERSION=${{ github.event.pull_request.head.sha || github.sha }}
            BUILD_TAG=${{ needs.tag.outputs.build-tag }}
            TAG=${{ needs.build-build-tools-image.outputs.image-tag }}
@@ -550,7 +551,7 @@ jobs:
        version: [ v14, v15, v16 ]
        arch: [ x64, arm64 ]

-    runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}
+    runs-on: ${{ fromJson(format('["self-hosted", "gen3", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}

    steps:
      - name: Checkout
@@ -695,7 +696,7 @@ jobs:

  vm-compute-node-image:
    needs: [ check-permissions, tag, compute-node-image ]
-    runs-on: [ self-hosted, large ]
+    runs-on: [ self-hosted, gen3, large ]
    strategy:
      fail-fast: false
      matrix:
@@ -744,7 +745,7 @@ jobs:
      matrix:
        arch: [ x64, arm64 ]

-    runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', matrix.arch == 'arm64' && 'small-arm64' || 'small')) }}
+    runs-on: ${{ fromJson(format('["self-hosted", "gen3", "{0}"]', matrix.arch == 'arm64' && 'small-arm64' || 'small')) }}

    steps:
      - name: Checkout
@@ -959,7 +960,7 @@ jobs:
    needs: [ check-permissions, promote-images, tag, build-and-test-locally, trigger-custom-extensions-build-and-wait ]
    if: github.ref_name == 'main' || github.ref_name == 'release'|| github.ref_name == 'release-proxy'

-    runs-on: [ self-hosted, small ]
+    runs-on: [ self-hosted, gen3, small ]
    container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/ansible:latest
    steps:
      - name: Fix git ownership
@@ -979,6 +980,7 @@ jobs:
      - name: Checkout
        uses: actions/checkout@v4
        with:
+          submodules: false
          fetch-depth: 0

      - name: Trigger deploy workflow
@@ -986,10 +988,10 @@ jobs:
          GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
        run: |
          if [[ "$GITHUB_REF_NAME" == "main" ]]; then
-            gh workflow --repo neondatabase/infra run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=false
+            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=false
            gh workflow --repo neondatabase/azure run deploy.yml -f dockerTag=${{needs.tag.outputs.build-tag}}
          elif [[ "$GITHUB_REF_NAME" == "release" ]]; then
-            gh workflow --repo neondatabase/infra run deploy-dev.yml --ref main \
+            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main \
              -f deployPgSniRouter=false \
              -f deployProxy=false \
              -f deployStorage=true \
@@ -999,14 +1001,14 @@ jobs:
              -f dockerTag=${{needs.tag.outputs.build-tag}} \
              -f deployPreprodRegion=true

-            gh workflow --repo neondatabase/infra run deploy-prod.yml --ref main \
+            gh workflow --repo neondatabase/aws run deploy-prod.yml --ref main \
              -f deployStorage=true \
              -f deployStorageBroker=true \
              -f deployStorageController=true \
              -f branch=main \
              -f dockerTag=${{needs.tag.outputs.build-tag}}
          elif [[ "$GITHUB_REF_NAME" == "release-proxy" ]]; then
-            gh workflow --repo neondatabase/infra run deploy-dev.yml --ref main \
+            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main \
              -f deployPgSniRouter=true \
              -f deployProxy=true \
              -f deployStorage=false \
@@ -1016,7 +1018,7 @@ jobs:
              -f dockerTag=${{needs.tag.outputs.build-tag}} \
              -f deployPreprodRegion=true

-            gh workflow --repo neondatabase/infra run deploy-proxy-prod.yml --ref main \
+            gh workflow --repo neondatabase/aws run deploy-proxy-prod.yml --ref main \
              -f deployPgSniRouter=true \
              -f deployProxy=true \
              -f branch=main \
@@ -1059,7 +1061,7 @@ jobs:
    needs: [ check-permissions, promote-images, tag, build-and-test-locally ]
    if: github.ref_name == 'release'

-    runs-on: [ self-hosted, small ]
+    runs-on: [ self-hosted, gen3, small ]
    container:
      image: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/base:pinned
      options: --init
@@ -1115,12 +1117,10 @@ jobs:
    # Format `needs` differently to make the list more readable.
    # Usually we do `needs: [...]`
    needs:
-      - build-and-test-locally
      - check-codestyle-python
      - check-codestyle-rust
-      - promote-images
+      - build-and-test-locally
      - test-images
-      - trigger-custom-extensions-build-and-wait
    runs-on: ubuntu-22.04
    steps:
      # The list of possible results:
--- a/.github/workflows/label-for-external-users.yml
+++ b/.github/workflows/label-for-external-users.yml
@@ -4,7 +4,7 @@ on:
  issues:
    types:
      - opened
-  pull_request_target:
+  pull_request:
    types:
      - opened

@@ -15,40 +15,21 @@ env:
  LABEL: external

 jobs:
-  check-user:
-    runs-on: ubuntu-22.04
-
-    outputs:
-      is-member: ${{ steps.check-user.outputs.is-member }}
-
-    steps:
-    - name: Check whether `${{ github.actor }}` is a member of `${{ github.repository_owner }}`
-      id: check-user
-      env:
-        GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
-      run: |
-        if gh api -H "Accept: application/vnd.github+json" -H "X-GitHub-Api-Version: 2022-11-28" "/orgs/${GITHUB_REPOSITORY_OWNER}/members/${GITHUB_ACTOR}"; then
-          is_member=true
-        else
-          is_member=false
-        fi
-
-        echo "is-member=${is_member}" | tee -a ${GITHUB_OUTPUT}
-
  add-label:
-    if: needs.check-user.outputs.is-member == 'false'
-    needs: [ check-user ]
+    # This workflow uses `author_association` for PRs and issues to determine if the user is an external user.
+    # Possible values for `author_association`: https://docs.github.com/en/graphql/reference/enums#commentauthorassociation
+    if: ${{ !contains(fromJSON('["OWNER", "MEMBER", "COLLABORATOR"]'), github.event[github.event_name == 'pull_request' && 'pull_request' || 'issue'].author_association) }}

    runs-on: ubuntu-22.04
    permissions:
-      pull-requests: write # for `gh pr edit`
-      issues: write        # for `gh issue edit`
+      pull-requests: write
+      issues: write

    steps:
-    - name: Add `${{ env.LABEL }}` label
+    - name: Label new ${{ github.event_name }}
      env:
        GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        ITEM_NUMBER: ${{ github.event[github.event_name == 'pull_request_target' && 'pull_request' || 'issue'].number }}
-        GH_CLI_COMMAND: ${{ github.event_name == 'pull_request_target' && 'pr' || 'issue' }}
+        ITEM_NUMBER: ${{ github.event[github.event_name == 'pull_request' && 'pull_request' || 'issue'].number }}
+        GH_CLI_COMMAND: ${{ github.event_name == 'pull_request' && 'pr' || 'issue' }}
      run: |
        gh ${GH_CLI_COMMAND} --repo ${GITHUB_REPOSITORY} edit --add-label=${LABEL} ${ITEM_NUMBER}
--- a/.github/workflows/neon_extra_builds.yml
+++ b/.github/workflows/neon_extra_builds.yml
@@ -56,6 +56,7 @@ jobs:
        uses: actions/checkout@v4
        with:
          submodules: true
+          fetch-depth: 1

      - name: Install macOS postgres dependencies
        run: brew install flex bison openssl protobuf icu4c pkg-config
@@ -157,6 +158,7 @@ jobs:
        uses: actions/checkout@v4
        with:
          submodules: true
+          fetch-depth: 1

      # Some of our rust modules use FFI and need those to be checked
      - name: Get postgres headers
--- a/.github/workflows/periodic_pagebench.yml
+++ b/.github/workflows/periodic_pagebench.yml
@@ -27,7 +27,7 @@ concurrency:

 jobs:
  trigger_bench_on_ec2_machine_in_eu_central_1:
-    runs-on: [ self-hosted, small ]
+    runs-on: [ self-hosted, gen3, small ]
    container:
      image: neondatabase/build-tools:pinned
      credentials:
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -484,7 +484,7 @@ dependencies = [
 "http 0.2.9",
 "http 1.1.0",
 "once_cell",
- "p256 0.11.1",
+ "p256",
 "percent-encoding",
 "ring 0.17.6",
 "sha2",
@@ -848,12 +848,6 @@ version = "0.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "349a06037c7bf932dd7e7d1f653678b2038b9ad46a74102f1fc7bd7872678cce"

-[[package]]
-name = "base16ct"
-version = "0.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4c7f02d4ea65f2c1853089ffd8d2787bdbc63de2f0d29dedbcf8ccdfa0ccd4cf"
-
 [[package]]
 name = "base64"
 version = "0.13.1"
@@ -977,9 +971,9 @@ checksum = "a3e2c3daef883ecc1b5d58c15adae93470a91d425f3532ba1695849656af3fc1"

 [[package]]
 name = "bytemuck"
-version = "1.16.3"
+version = "1.16.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "102087e286b4677862ea56cf8fc58bb2cdfa8725c40ffb80fe3a008eb7f2fc83"
+checksum = "78834c15cb5d5efe3452d58b1e8ba890dd62d21907f867f383358198e56ebca5"

 [[package]]
 name = "byteorder"
@@ -1208,6 +1202,7 @@ dependencies = [
 "serde_json",
 "serde_with",
 "utils",
+ "workspace_hack",
 ]

 [[package]]
@@ -1320,6 +1315,7 @@ dependencies = [
 "serde",
 "serde_with",
 "utils",
+ "workspace_hack",
 ]

 [[package]]
@@ -1530,10 +1526,8 @@ version = "0.5.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0dc92fb57ca44df6db8059111ab3af99a63d5d0f8375d9972e319a379c6bab76"
 dependencies = [
- "generic-array",
 "rand_core 0.6.4",
 "subtle",
- "zeroize",
 ]

 [[package]]
@@ -1627,7 +1621,6 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "fffa369a668c8af7dbf8b5e56c9f744fbd399949ed171606040001947de40b1c"
 dependencies = [
 "const-oid",
- "pem-rfc7468",
 "zeroize",
 ]

@@ -1668,13 +1661,14 @@ dependencies = [
 "smallvec",
 "tracing",
 "utils",
+ "workspace_hack",
 ]

 [[package]]
 name = "diesel"
-version = "2.2.3"
+version = "2.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "65e13bab2796f412722112327f3e575601a3e9cdcbe426f0d30dbf43f3f5dc71"
+checksum = "62d6dcd069e7b5fe49a302411f759d4cf1cf2c27fe798ef46fb8baefc053dd2b"
 dependencies = [
 "bitflags 2.4.1",
 "byteorder",
@@ -1726,7 +1720,6 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9ed9a281f7bc9b7576e61468ba615a66a5c8cfdff42420a70aa82701a3b1e292"
 dependencies = [
 "block-buffer",
- "const-oid",
 "crypto-common",
 "subtle",
 ]
@@ -1778,25 +1771,11 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "413301934810f597c1d19ca71c8710e99a3f1ba28a0d2ebc01551a2daeea3c5c"
 dependencies = [
 "der 0.6.1",
- "elliptic-curve 0.12.3",
- "rfc6979 0.3.1",
+ "elliptic-curve",
+ "rfc6979",
 "signature 1.6.4",
 ]

-[[package]]
-name = "ecdsa"
-version = "0.16.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ee27f32b5c5292967d2d4a9d7f1e0b0aed2c15daded5a60300e4abb9d8020bca"
-dependencies = [
- "der 0.7.8",
- "digest",
- "elliptic-curve 0.13.8",
- "rfc6979 0.4.0",
- "signature 2.2.0",
- "spki 0.7.3",
-]
-
 [[package]]
 name = "either"
 version = "1.8.1"
@@ -1809,36 +1788,16 @@ version = "0.12.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e7bb888ab5300a19b8e5bceef25ac745ad065f3c9f7efc6de1b91958110891d3"
 dependencies = [
- "base16ct 0.1.1",
+ "base16ct",
 "crypto-bigint 0.4.9",
 "der 0.6.1",
 "digest",
- "ff 0.12.1",
+ "ff",
 "generic-array",
- "group 0.12.1",
- "pkcs8 0.9.0",
+ "group",
+ "pkcs8",
 "rand_core 0.6.4",
- "sec1 0.3.0",
- "subtle",
- "zeroize",
-]
-
-[[package]]
-name = "elliptic-curve"
-version = "0.13.8"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b5e6043086bf7973472e0c7dff2142ea0b680d30e18d9cc40f267efbf222bd47"
-dependencies = [
- "base16ct 0.2.0",
- "crypto-bigint 0.5.5",
- "digest",
- "ff 0.13.0",
- "generic-array",
- "group 0.13.0",
- "pem-rfc7468",
- "pkcs8 0.10.2",
- "rand_core 0.6.4",
- "sec1 0.7.3",
+ "sec1",
 "subtle",
 "zeroize",
 ]
@@ -1992,16 +1951,6 @@ dependencies = [
 "subtle",
 ]

-[[package]]
-name = "ff"
-version = "0.13.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ded41244b729663b1e574f1b4fb731469f69f79c17667b5d776b16cda0479449"
-dependencies = [
- "rand_core 0.6.4",
- "subtle",
-]
-
 [[package]]
 name = "filetime"
 version = "0.2.22"
@@ -2199,7 +2148,6 @@ checksum = "85649ca51fd72272d7821adaf274ad91c288277713d9c18820d8499a7ff69e9a"
 dependencies = [
 "typenum",
 "version_check",
- "zeroize",
 ]

 [[package]]
@@ -2266,18 +2214,7 @@ version = "0.12.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5dfbfb3a6cfbd390d5c9564ab283a0349b9b9fcd46a706c1eb10e0db70bfbac7"
 dependencies = [
- "ff 0.12.1",
- "rand_core 0.6.4",
- "subtle",
-]
-
-[[package]]
-name = "group"
-version = "0.13.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f0f9ef7462f7c099f518d754361858f86d8a07af53ba9af0fe635bbccb151a63"
-dependencies = [
- "ff 0.13.0",
+ "ff",
 "rand_core 0.6.4",
 "subtle",
 ]
@@ -2839,42 +2776,6 @@ dependencies = [
 "libc",
 ]

-[[package]]
-name = "jose-b64"
-version = "0.1.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bec69375368709666b21c76965ce67549f2d2db7605f1f8707d17c9656801b56"
-dependencies = [
- "base64ct",
- "serde",
- "subtle",
- "zeroize",
-]
-
-[[package]]
-name = "jose-jwa"
-version = "0.1.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9ab78e053fe886a351d67cf0d194c000f9d0dcb92906eb34d853d7e758a4b3a7"
-dependencies = [
- "serde",
-]
-
-[[package]]
-name = "jose-jwk"
-version = "0.1.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "280fa263807fe0782ecb6f2baadc28dffc04e00558a58e33bfdb801d11fd58e7"
-dependencies = [
- "jose-b64",
- "jose-jwa",
- "p256 0.13.2",
- "p384",
- "rsa",
- "serde",
- "zeroize",
-]
-
 [[package]]
 name = "js-sys"
 version = "0.3.69"
@@ -2934,9 +2835,6 @@ name = "lazy_static"
 version = "1.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e2abad23fbc42b3700f2f279844dc832adb2b2eb069b2df918f455c4e18cc646"
-dependencies = [
- "spin 0.5.2",
-]

 [[package]]
 name = "lazycell"
@@ -3144,6 +3042,7 @@ dependencies = [
 "rand 0.8.5",
 "rand_distr",
 "twox-hash",
+ "workspace_hack",
 ]

 [[package]]
@@ -3305,23 +3204,6 @@ dependencies = [
 "num-traits",
 ]

-[[package]]
-name = "num-bigint-dig"
-version = "0.8.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dc84195820f291c7697304f3cbdadd1cb7199c0efc917ff5eafd71225c136151"
-dependencies = [
- "byteorder",
- "lazy_static",
- "libm",
- "num-integer",
- "num-iter",
- "num-traits",
- "rand 0.8.5",
- "smallvec",
- "zeroize",
-]
-
 [[package]]
 name = "num-complex"
 version = "0.4.4"
@@ -3599,33 +3481,11 @@ version = "0.11.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "51f44edd08f51e2ade572f141051021c5af22677e42b7dd28a88155151c33594"
 dependencies = [
- "ecdsa 0.14.8",
- "elliptic-curve 0.12.3",
+ "ecdsa",
+ "elliptic-curve",
 "sha2",
 ]

-[[package]]
-name = "p256"
-version = "0.13.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c9863ad85fa8f4460f9c48cb909d38a0d689dba1f6f6988a5e3e0d31071bcd4b"
-dependencies = [
- "ecdsa 0.16.9",
- "elliptic-curve 0.13.8",
- "primeorder",
- "sha2",
-]
-
-[[package]]
-name = "p384"
-version = "0.13.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "70786f51bcc69f6a4c0360e063a4cac5419ef7c5cd5b3c99ad70f3be5ba79209"
-dependencies = [
- "elliptic-curve 0.13.8",
- "primeorder",
-]
-
 [[package]]
 name = "pagebench"
 version = "0.1.0"
@@ -3787,6 +3647,7 @@ dependencies = [
 "strum_macros",
 "thiserror",
 "utils",
+ "workspace_hack",
 ]

 [[package]]
@@ -3986,15 +3847,6 @@ dependencies = [
 "serde",
 ]

-[[package]]
-name = "pem-rfc7468"
-version = "0.7.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "88b39c9bfcfc231068454382784bb460aae594343fb030d46e9f50a645418412"
-dependencies = [
- "base64ct",
-]
-
 [[package]]
 name = "percent-encoding"
 version = "2.2.0"
@@ -4061,17 +3913,6 @@ version = "0.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"

-[[package]]
-name = "pkcs1"
-version = "0.7.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c8ffb9f10fa047879315e6625af03c164b16962a5368d724ed16323b68ace47f"
-dependencies = [
- "der 0.7.8",
- "pkcs8 0.10.2",
- "spki 0.7.3",
-]
-
 [[package]]
 name = "pkcs8"
 version = "0.9.0"
@@ -4082,16 +3923,6 @@ dependencies = [
 "spki 0.6.0",
 ]

-[[package]]
-name = "pkcs8"
-version = "0.10.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f950b2377845cebe5cf8b5165cb3cc1a5e0fa5cfa3e1f7f55707d8fd82e0a7b7"
-dependencies = [
- "der 0.7.8",
- "spki 0.7.3",
-]
-
 [[package]]
 name = "pkg-config"
 version = "0.3.27"
@@ -4188,6 +4019,7 @@ dependencies = [
 "tokio-rustls 0.25.0",
 "tokio-util",
 "tracing",
+ "workspace_hack",
 ]

 [[package]]
@@ -4200,6 +4032,7 @@ dependencies = [
 "postgres",
 "tokio-postgres",
 "url",
+ "workspace_hack",
 ]

 [[package]]
@@ -4222,6 +4055,7 @@ dependencies = [
 "serde",
 "thiserror",
 "utils",
+ "workspace_hack",
 ]

 [[package]]
@@ -4259,6 +4093,7 @@ dependencies = [
 "thiserror",
 "tokio",
 "tracing",
+ "workspace_hack",
 ]

 [[package]]
@@ -4281,15 +4116,6 @@ dependencies = [
 "syn 2.0.52",
 ]

-[[package]]
-name = "primeorder"
-version = "0.13.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "353e1ca18966c16d9deb1c69278edbc5f194139612772bd9537af60ac231e1e6"
-dependencies = [
- "elliptic-curve 0.13.8",
-]
-
 [[package]]
 name = "proc-macro-hack"
 version = "0.5.20+deprecated"
@@ -4407,7 +4233,6 @@ version = "0.1.0"
 dependencies = [
 "ahash",
 "anyhow",
- "arc-swap",
 "async-compression",
 "async-trait",
 "atomic-take",
@@ -4425,7 +4250,6 @@ dependencies = [
 "consumption_metrics",
 "crossbeam-deque",
 "dashmap",
- "ecdsa 0.16.9",
 "env_logger",
 "fallible-iterator",
 "framed-websockets",
@@ -4446,15 +4270,12 @@ dependencies = [
 "indexmap 2.0.1",
 "ipnet",
 "itertools 0.10.5",
- "jose-jwa",
- "jose-jwk",
 "lasso",
 "md5",
 "measured",
 "metrics",
 "once_cell",
 "opentelemetry",
- "p256 0.13.2",
 "parking_lot 0.12.1",
 "parquet",
 "parquet_derive",
@@ -4475,7 +4296,6 @@ dependencies = [
 "reqwest-retry",
 "reqwest-tracing",
 "routerify",
- "rsa",
 "rstest",
 "rustc-hash",
 "rustls 0.22.4",
@@ -4485,7 +4305,6 @@ dependencies = [
 "serde",
 "serde_json",
 "sha2",
- "signature 2.2.0",
 "smallvec",
 "smol_str",
 "socket2 0.5.5",
@@ -4823,6 +4642,7 @@ dependencies = [
 "toml_edit 0.19.10",
 "tracing",
 "utils",
+ "workspace_hack",
 ]

 [[package]]
@@ -4987,16 +4807,6 @@ dependencies = [
 "zeroize",
 ]

-[[package]]
-name = "rfc6979"
-version = "0.4.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f8dd2a808d456c4a54e300a23e9f5a67e122c3024119acbfd73e3bf664491cb2"
-dependencies = [
- "hmac",
- "subtle",
-]
-
 [[package]]
 name = "ring"
 version = "0.16.20"
@@ -5057,26 +4867,6 @@ dependencies = [
 "archery",
 ]

-[[package]]
-name = "rsa"
-version = "0.9.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5d0e5124fcb30e76a7e79bfee683a2746db83784b86289f6251b54b7950a0dfc"
-dependencies = [
- "const-oid",
- "digest",
- "num-bigint-dig",
- "num-integer",
- "num-traits",
- "pkcs1",
- "pkcs8 0.10.2",
- "rand_core 0.6.4",
- "signature 2.2.0",
- "spki 0.7.3",
- "subtle",
- "zeroize",
-]
-
 [[package]]
 name = "rstest"
 version = "0.18.2"
@@ -5347,6 +5137,7 @@ dependencies = [
 "serde",
 "serde_with",
 "utils",
+ "workspace_hack",
 ]

 [[package]]
@@ -5404,24 +5195,10 @@ version = "0.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3be24c1842290c45df0a7bf069e0c268a747ad05a192f2fd7dcfdbc1cba40928"
 dependencies = [
- "base16ct 0.1.1",
+ "base16ct",
 "der 0.6.1",
 "generic-array",
- "pkcs8 0.9.0",
- "subtle",
- "zeroize",
-]
-
-[[package]]
-name = "sec1"
-version = "0.7.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d3e97a565f76233a6003f9f5c54be1d9c5bdfa3eccfb189469f11ec4901c47dc"
-dependencies = [
- "base16ct 0.2.0",
- "der 0.7.8",
- "generic-array",
- "pkcs8 0.10.2",
+ "pkcs8",
 "subtle",
 "zeroize",
 ]
@@ -5590,12 +5367,11 @@ dependencies = [

 [[package]]
 name = "serde_json"
-version = "1.0.125"
+version = "1.0.96"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "83c8e735a073ccf5be70aa8066aa984eaf2fa000db6c8d0100ae605b366d31ed"
+checksum = "057d394a50403bcac12672b2b18fb387ab6d289d957dab67dd201875391e52f1"
 dependencies = [
 "itoa",
- "memchr",
 "ryu",
 "serde",
 ]
@@ -5769,7 +5545,6 @@ version = "2.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "77549399552de45a898a580c1b41d445bf730df867cc44e6c0233bbc4b8329de"
 dependencies = [
- "digest",
 "rand_core 0.6.4",
 ]

@@ -6183,6 +5958,7 @@ dependencies = [
 "anyhow",
 "serde",
 "serde_json",
+ "workspace_hack",
 ]

 [[package]]
@@ -6783,6 +6559,7 @@ dependencies = [
 "tracing",
 "tracing-opentelemetry",
 "tracing-subscriber",
+ "workspace_hack",
 ]

 [[package]]
@@ -7000,6 +6777,7 @@ dependencies = [
 "url",
 "uuid",
 "walkdir",
+ "workspace_hack",
 ]

 [[package]]
@@ -7078,6 +6856,7 @@ dependencies = [
 "postgres_ffi",
 "regex",
 "utils",
+ "workspace_hack",
 ]

 [[package]]
@@ -7098,6 +6877,7 @@ dependencies = [
 "bindgen",
 "postgres_ffi",
 "utils",
+ "workspace_hack",
 ]

 [[package]]
@@ -7599,17 +7379,13 @@ dependencies = [
 "clap",
 "clap_builder",
 "crossbeam-utils",
- "crypto-bigint 0.5.5",
- "der 0.7.8",
 "deranged",
- "digest",
 "either",
 "fail",
 "futures-channel",
 "futures-executor",
 "futures-io",
 "futures-util",
- "generic-array",
 "getrandom 0.2.11",
 "hashbrown 0.14.5",
 "hex",
@@ -7617,7 +7393,6 @@ dependencies = [
 "hyper 0.14.26",
 "indexmap 1.9.3",
 "itertools 0.10.5",
- "lazy_static",
 "libc",
 "log",
 "memchr",
@@ -7641,9 +7416,7 @@ dependencies = [
 "serde",
 "serde_json",
 "sha2",
- "signature 2.2.0",
 "smallvec",
- "spki 0.7.3",
 "subtle",
 "syn 1.0.109",
 "syn 2.0.52",
@@ -7654,6 +7427,8 @@ dependencies = [
 "tokio",
 "tokio-rustls 0.24.0",
 "tokio-util",
+ "toml_datetime",
+ "toml_edit 0.19.10",
 "tonic",
 "tower",
 "tracing",
@@ -7752,7 +7527,6 @@ version = "1.7.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "525b4ec142c6b68a2d10f01f7bbf6755599ca3f81ea53b8431b7dd348f5fdb2d"
 dependencies = [
- "serde",
 "zeroize_derive",
 ]

--- a/3
+++ b/3
@@ -35,9 +35,8 @@ COPY --from=pg-build /home/nonroot/pg_install/v16/include/postgresql/server pg_i
 COPY --from=pg-build /home/nonroot/pg_install/v16/lib                       pg_install/v16/lib
 COPY --chown=nonroot . .

-ARG ADDITIONAL_RUSTFLAGS
 RUN set -e \
-    && PQ_LIB_DIR=$(pwd)/pg_install/v16/lib RUSTFLAGS="-Clinker=clang -Clink-arg=-fuse-ld=mold -Clink-arg=-Wl,--no-rosegment ${ADDITIONAL_RUSTFLAGS}" cargo build \
+    && PQ_LIB_DIR=$(pwd)/pg_install/v16/lib RUSTFLAGS="-Clinker=clang -Clink-arg=-fuse-ld=mold -Clink-arg=-Wl,--no-rosegment" cargo build \
      --bin pg_sni_router  \
      --bin pageserver  \
      --bin pagectl  \
--- a/README.md
+++ b/README.md
@@ -126,7 +126,7 @@ make -j`sysctl -n hw.logicalcpu` -s
 To run the `psql` client, install the `postgresql-client` package or modify `PATH` and `LD_LIBRARY_PATH` to include `pg_install/bin` and `pg_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 the python3 packages using `./scripts/pysync` (requires [poetry>=1.8](https://python-poetry.org/)) in the project directory.
+Python (3.9 or higher), and install the python3 packages using `./scripts/pysync` (requires [poetry>=1.3](https://python-poetry.org/)) in the project directory.


 #### Running neon database
@@ -262,7 +262,7 @@ By default, this runs both debug and release modes, and all supported postgres v
 testing locally, it is convenient to run just one set of permutations, like this:

 ```sh
-DEFAULT_PG_VERSION=16 BUILD_TYPE=release ./scripts/pytest
+DEFAULT_PG_VERSION=15 BUILD_TYPE=release ./scripts/pytest
 ```

 ## Flamegraphs
--- a/control_plane/src/background_process.rs
+++ b/control_plane/src/background_process.rs
@@ -379,7 +379,7 @@ where
    }
 }

-pub(crate) fn process_has_stopped(pid: Pid) -> anyhow::Result<bool> {
+fn process_has_stopped(pid: Pid) -> anyhow::Result<bool> {
    match kill(pid, None) {
        // Process exists, keep waiting
        Ok(_) => Ok(false),
--- a/control_plane/src/bin/neon_local.rs
+++ b/control_plane/src/bin/neon_local.rs
@@ -15,9 +15,7 @@ use control_plane::local_env::{
 };
 use control_plane::pageserver::PageServerNode;
 use control_plane::safekeeper::SafekeeperNode;
-use control_plane::storage_controller::{
-    NeonStorageControllerStartArgs, NeonStorageControllerStopArgs, StorageController,
-};
+use control_plane::storage_controller::StorageController;
 use control_plane::{broker, local_env};
 use pageserver_api::config::{
    DEFAULT_HTTP_LISTEN_PORT as DEFAULT_PAGESERVER_HTTP_PORT,
@@ -54,7 +52,7 @@ const DEFAULT_PAGESERVER_ID: NodeId = NodeId(1);
 const DEFAULT_BRANCH_NAME: &str = "main";
 project_git_version!(GIT_VERSION);

-const DEFAULT_PG_VERSION: &str = "16";
+const DEFAULT_PG_VERSION: &str = "15";

 const DEFAULT_PAGESERVER_CONTROL_PLANE_API: &str = "http://127.0.0.1:1234/upcall/v1/";

@@ -1054,36 +1052,6 @@ fn get_start_timeout(args: &ArgMatches) -> &Duration {
    humantime_duration.as_ref()
 }

-fn storage_controller_start_args(args: &ArgMatches) -> NeonStorageControllerStartArgs {
-    let maybe_instance_id = args.get_one::<u8>("instance-id");
-
-    let base_port = args.get_one::<u16>("base-port");
-
-    if maybe_instance_id.is_some() && base_port.is_none() {
-        panic!("storage-controller start specificied instance-id but did not provide base-port");
-    }
-
-    let start_timeout = args
-        .get_one::<humantime::Duration>("start-timeout")
-        .expect("invalid value for start-timeout");
-
-    NeonStorageControllerStartArgs {
-        instance_id: maybe_instance_id.copied().unwrap_or(1),
-        base_port: base_port.copied(),
-        start_timeout: *start_timeout,
-    }
-}
-
-fn storage_controller_stop_args(args: &ArgMatches) -> NeonStorageControllerStopArgs {
-    let maybe_instance_id = args.get_one::<u8>("instance-id");
-    let immediate = args.get_one::<String>("stop-mode").map(|s| s.as_str()) == Some("immediate");
-
-    NeonStorageControllerStopArgs {
-        instance_id: maybe_instance_id.copied().unwrap_or(1),
-        immediate,
-    }
-}
-
 async fn handle_pageserver(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
    match sub_match.subcommand() {
        Some(("start", subcommand_args)) => {
@@ -1145,14 +1113,19 @@ async fn handle_storage_controller(
    let svc = StorageController::from_env(env);
    match sub_match.subcommand() {
        Some(("start", start_match)) => {
-            if let Err(e) = svc.start(storage_controller_start_args(start_match)).await {
+            if let Err(e) = svc.start(get_start_timeout(start_match)).await {
                eprintln!("start failed: {e}");
                exit(1);
            }
        }

        Some(("stop", stop_match)) => {
-            if let Err(e) = svc.stop(storage_controller_stop_args(stop_match)).await {
+            let immediate = stop_match
+                .get_one::<String>("stop-mode")
+                .map(|s| s.as_str())
+                == Some("immediate");
+
+            if let Err(e) = svc.stop(immediate).await {
                eprintln!("stop failed: {}", e);
                exit(1);
            }
@@ -1255,12 +1228,7 @@ async fn handle_start_all(
    // Only start the storage controller if the pageserver is configured to need it
    if env.control_plane_api.is_some() {
        let storage_controller = StorageController::from_env(env);
-        if let Err(e) = storage_controller
-            .start(NeonStorageControllerStartArgs::with_default_instance_id(
-                (*retry_timeout).into(),
-            ))
-            .await
-        {
+        if let Err(e) = storage_controller.start(retry_timeout).await {
            eprintln!("storage_controller start failed: {:#}", e);
            try_stop_all(env, true).await;
            exit(1);
@@ -1390,21 +1358,10 @@ async fn try_stop_all(env: &local_env::LocalEnv, immediate: bool) {
        eprintln!("neon broker stop failed: {e:#}");
    }

-    // Stop all storage controller instances. In the most common case there's only one,
-    // but iterate though the base data directory in order to discover the instances.
-    let storcon_instances = env
-        .storage_controller_instances()
-        .await
-        .expect("Must inspect data dir");
-    for (instance_id, _instance_dir_path) in storcon_instances {
+    if env.control_plane_api.is_some() {
        let storage_controller = StorageController::from_env(env);
-        let stop_args = NeonStorageControllerStopArgs {
-            instance_id,
-            immediate,
-        };
-
-        if let Err(e) = storage_controller.stop(stop_args).await {
-            eprintln!("Storage controller instance {instance_id} stop failed: {e:#}");
+        if let Err(e) = storage_controller.stop(immediate).await {
+            eprintln!("storage controller stop failed: {e:#}");
        }
    }
 }
@@ -1544,18 +1501,6 @@ fn cli() -> Command {
        .action(ArgAction::SetTrue)
        .required(false);

-    let instance_id = Arg::new("instance-id")
-        .long("instance-id")
-        .help("Identifier used to distinguish storage controller instances (default 1)")
-        .value_parser(value_parser!(u8))
-        .required(false);
-
-    let base_port = Arg::new("base-port")
-        .long("base-port")
-        .help("Base port for the storage controller instance idenfified by instance-id (defaults to pagserver cplane api)")
-        .value_parser(value_parser!(u16))
-        .required(false);
-
    Command::new("Neon CLI")
        .arg_required_else_help(true)
        .version(GIT_VERSION)
@@ -1664,12 +1609,9 @@ fn cli() -> Command {
                .arg_required_else_help(true)
                .about("Manage storage_controller")
                .subcommand(Command::new("start").about("Start storage controller")
-                            .arg(timeout_arg.clone())
-                            .arg(instance_id.clone())
-                            .arg(base_port))
+                            .arg(timeout_arg.clone()))
                .subcommand(Command::new("stop").about("Stop storage controller")
-                            .arg(stop_mode_arg.clone())
-                            .arg(instance_id))
+                            .arg(stop_mode_arg.clone()))
        )
        .subcommand(
            Command::new("safekeeper")
--- a/control_plane/src/local_env.rs
+++ b/control_plane/src/local_env.rs
@@ -27,7 +27,7 @@ use crate::pageserver::PageServerNode;
 use crate::pageserver::PAGESERVER_REMOTE_STORAGE_DIR;
 use crate::safekeeper::SafekeeperNode;

-pub const DEFAULT_PG_VERSION: u32 = 16;
+pub const DEFAULT_PG_VERSION: u32 = 15;

 //
 // This data structures represents neon_local CLI config
@@ -156,11 +156,6 @@ pub struct NeonStorageControllerConf {
    #[serde(with = "humantime_serde")]
    pub max_warming_up: Duration,

-    pub start_as_candidate: bool,
-
-    /// Database url used when running multiple storage controller instances
-    pub database_url: Option<SocketAddr>,
-
    /// Threshold for auto-splitting a tenant into shards
    pub split_threshold: Option<u64>,

@@ -179,8 +174,6 @@ impl Default for NeonStorageControllerConf {
        Self {
            max_offline: Self::DEFAULT_MAX_OFFLINE_INTERVAL,
            max_warming_up: Self::DEFAULT_MAX_WARMING_UP_INTERVAL,
-            start_as_candidate: false,
-            database_url: None,
            split_threshold: None,
            max_secondary_lag_bytes: None,
        }
@@ -399,36 +392,6 @@ impl LocalEnv {
        }
    }

-    /// Inspect the base data directory and extract the instance id and instance directory path
-    /// for all storage controller instances
-    pub async fn storage_controller_instances(&self) -> std::io::Result<Vec<(u8, PathBuf)>> {
-        let mut instances = Vec::default();
-
-        let dir = std::fs::read_dir(self.base_data_dir.clone())?;
-        for dentry in dir {
-            let dentry = dentry?;
-            let is_dir = dentry.metadata()?.is_dir();
-            let filename = dentry.file_name().into_string().unwrap();
-            let parsed_instance_id = match filename.strip_prefix("storage_controller_") {
-                Some(suffix) => suffix.parse::<u8>().ok(),
-                None => None,
-            };
-
-            let is_instance_dir = is_dir && parsed_instance_id.is_some();
-
-            if !is_instance_dir {
-                continue;
-            }
-
-            instances.push((
-                parsed_instance_id.expect("Checked previously"),
-                dentry.path(),
-            ));
-        }
-
-        Ok(instances)
-    }
-
    pub fn register_branch_mapping(
        &mut self,
        branch_name: String,
--- a/control_plane/src/storage_controller.rs
+++ b/control_plane/src/storage_controller.rs
@@ -3,8 +3,6 @@ use crate::{
    local_env::{LocalEnv, NeonStorageControllerConf},
 };
 use camino::{Utf8Path, Utf8PathBuf};
-use hyper::Uri;
-use nix::unistd::Pid;
 use pageserver_api::{
    controller_api::{
        NodeConfigureRequest, NodeDescribeResponse, NodeRegisterRequest, TenantCreateRequest,
@@ -20,7 +18,7 @@ use pageserver_client::mgmt_api::ResponseErrorMessageExt;
 use postgres_backend::AuthType;
 use reqwest::Method;
 use serde::{de::DeserializeOwned, Deserialize, Serialize};
-use std::{fs, net::SocketAddr, path::PathBuf, str::FromStr, sync::OnceLock};
+use std::{fs, str::FromStr, time::Duration};
 use tokio::process::Command;
 use tracing::instrument;
 use url::Url;
@@ -31,14 +29,12 @@ use utils::{

 pub struct StorageController {
    env: LocalEnv,
+    listen: String,
    private_key: Option<Vec<u8>>,
    public_key: Option<String>,
+    postgres_port: u16,
    client: reqwest::Client,
    config: NeonStorageControllerConf,
-
-    // The listen addresses is learned when starting the storage controller,
-    // hence the use of OnceLock to init it at the right time.
-    listen: OnceLock<SocketAddr>,
 }

 const COMMAND: &str = "storage_controller";
@@ -47,36 +43,6 @@ const STORAGE_CONTROLLER_POSTGRES_VERSION: u32 = 16;

 const DB_NAME: &str = "storage_controller";

-pub struct NeonStorageControllerStartArgs {
-    pub instance_id: u8,
-    pub base_port: Option<u16>,
-    pub start_timeout: humantime::Duration,
-}
-
-impl NeonStorageControllerStartArgs {
-    pub fn with_default_instance_id(start_timeout: humantime::Duration) -> Self {
-        Self {
-            instance_id: 1,
-            base_port: None,
-            start_timeout,
-        }
-    }
-}
-
-pub struct NeonStorageControllerStopArgs {
-    pub instance_id: u8,
-    pub immediate: bool,
-}
-
-impl NeonStorageControllerStopArgs {
-    pub fn with_default_instance_id(immediate: bool) -> Self {
-        Self {
-            instance_id: 1,
-            immediate,
-        }
-    }
-}
-
 #[derive(Serialize, Deserialize)]
 pub struct AttachHookRequest {
    pub tenant_shard_id: TenantShardId,
@@ -101,6 +67,23 @@ pub struct InspectResponse {

 impl StorageController {
    pub fn from_env(env: &LocalEnv) -> Self {
+        // Makes no sense to construct this if pageservers aren't going to use it: assume
+        // pageservers have control plane API set
+        let listen_url = env.control_plane_api.clone().unwrap();
+
+        let listen = format!(
+            "{}:{}",
+            listen_url.host_str().unwrap(),
+            listen_url.port().unwrap()
+        );
+
+        // Convention: NeonEnv in python tests reserves the next port after the control_plane_api
+        // port, for use by our captive postgres.
+        let postgres_port = listen_url
+            .port()
+            .expect("Control plane API setting should always have a port")
+            + 1;
+
        // Assume all pageservers have symmetric auth configuration: this service
        // expects to use one JWT token to talk to all of them.
        let ps_conf = env
@@ -143,28 +126,20 @@ impl StorageController {

        Self {
            env: env.clone(),
+            listen,
            private_key,
            public_key,
+            postgres_port,
            client: reqwest::ClientBuilder::new()
                .build()
                .expect("Failed to construct http client"),
            config: env.storage_controller.clone(),
-            listen: OnceLock::default(),
        }
    }

-    fn storage_controller_instance_dir(&self, instance_id: u8) -> PathBuf {
-        self.env
-            .base_data_dir
-            .join(format!("storage_controller_{}", instance_id))
-    }
-
-    fn pid_file(&self, instance_id: u8) -> Utf8PathBuf {
-        Utf8PathBuf::from_path_buf(
-            self.storage_controller_instance_dir(instance_id)
-                .join("storage_controller.pid"),
-        )
-        .expect("non-Unicode path")
+    fn pid_file(&self) -> Utf8PathBuf {
+        Utf8PathBuf::from_path_buf(self.env.base_data_dir.join("storage_controller.pid"))
+            .expect("non-Unicode path")
    }

    /// PIDFile for the postgres instance used to store storage controller state
@@ -209,23 +184,23 @@ impl StorageController {
    }

    /// Readiness check for our postgres process
-    async fn pg_isready(&self, pg_bin_dir: &Utf8Path, postgres_port: u16) -> anyhow::Result<bool> {
+    async fn pg_isready(&self, pg_bin_dir: &Utf8Path) -> anyhow::Result<bool> {
        let bin_path = pg_bin_dir.join("pg_isready");
-        let args = ["-h", "localhost", "-p", &format!("{}", postgres_port)];
+        let args = ["-h", "localhost", "-p", &format!("{}", self.postgres_port)];
        let exitcode = Command::new(bin_path).args(args).spawn()?.wait().await?;

        Ok(exitcode.success())
    }

-    /// Create our database if it doesn't exist
+    /// Create our database if it doesn't exist, and run migrations.
    ///
    /// This function is equivalent to the `diesel setup` command in the diesel CLI.  We implement
    /// the same steps by hand to avoid imposing a dependency on installing diesel-cli for developers
    /// who just want to run `cargo neon_local` without knowing about diesel.
    ///
    /// Returns the database url
-    pub async fn setup_database(&self, postgres_port: u16) -> anyhow::Result<String> {
-        let database_url = format!("postgresql://localhost:{}/{DB_NAME}", postgres_port);
+    pub async fn setup_database(&self) -> anyhow::Result<String> {
+        let database_url = format!("postgresql://localhost:{}/{DB_NAME}", self.postgres_port);

        let pg_bin_dir = self.get_pg_bin_dir().await?;
        let createdb_path = pg_bin_dir.join("createdb");
@@ -234,7 +209,7 @@ impl StorageController {
                "-h",
                "localhost",
                "-p",
-                &format!("{}", postgres_port),
+                &format!("{}", self.postgres_port),
                DB_NAME,
            ])
            .output()
@@ -255,14 +230,13 @@ impl StorageController {

    pub async fn connect_to_database(
        &self,
-        postgres_port: u16,
    ) -> anyhow::Result<(
        tokio_postgres::Client,
        tokio_postgres::Connection<tokio_postgres::Socket, tokio_postgres::tls::NoTlsStream>,
    )> {
        tokio_postgres::Config::new()
            .host("localhost")
-            .port(postgres_port)
+            .port(self.postgres_port)
            // The user is the ambient operating system user name.
            // That is an impurity which we want to fix in => TODO https://github.com/neondatabase/neon/issues/8400
            //
@@ -278,114 +252,72 @@ impl StorageController {
            .map_err(anyhow::Error::new)
    }

-    pub async fn start(&self, start_args: NeonStorageControllerStartArgs) -> anyhow::Result<()> {
-        let instance_dir = self.storage_controller_instance_dir(start_args.instance_id);
-        if let Err(err) = tokio::fs::create_dir(&instance_dir).await {
-            if err.kind() != std::io::ErrorKind::AlreadyExists {
-                panic!("Failed to create instance dir {instance_dir:?}");
-            }
-        }
+    pub async fn start(&self, retry_timeout: &Duration) -> anyhow::Result<()> {
+        // Start a vanilla Postgres process used by the storage controller for persistence.
+        let pg_data_path = Utf8PathBuf::from_path_buf(self.env.base_data_dir.clone())
+            .unwrap()
+            .join("storage_controller_db");
+        let pg_bin_dir = self.get_pg_bin_dir().await?;
+        let pg_lib_dir = self.get_pg_lib_dir().await?;
+        let pg_log_path = pg_data_path.join("postgres.log");

-        let (listen, postgres_port) = {
-            if let Some(base_port) = start_args.base_port {
-                (
-                    format!("127.0.0.1:{base_port}"),
-                    self.config
-                        .database_url
-                        .expect("--base-port requires NeonStorageControllerConf::database_url")
-                        .port(),
-                )
-            } else {
-                let listen_url = self.env.control_plane_api.clone().unwrap();
-
-                let listen = format!(
-                    "{}:{}",
-                    listen_url.host_str().unwrap(),
-                    listen_url.port().unwrap()
-                );
-
-                (listen, listen_url.port().unwrap() + 1)
+        if !tokio::fs::try_exists(&pg_data_path).await? {
+            // Initialize empty database
+            let initdb_path = pg_bin_dir.join("initdb");
+            let mut child = Command::new(&initdb_path)
+                .envs(vec![
+                    ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
+                    ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
+                ])
+                .args(["-D", pg_data_path.as_ref()])
+                .spawn()
+                .expect("Failed to spawn initdb");
+            let status = child.wait().await?;
+            if !status.success() {
+                anyhow::bail!("initdb failed with status {status}");
            }
        };

-        let socket_addr = listen
-            .parse()
-            .expect("listen address is a valid socket address");
-        self.listen
-            .set(socket_addr)
-            .expect("StorageController::listen is only set here");
+        // Write a minimal config file:
+        // - Specify the port, since this is chosen dynamically
+        // - Switch off fsync, since we're running on lightweight test environments and when e.g. scale testing
+        //   the storage controller we don't want a slow local disk to interfere with that.
+        //
+        // NB: it's important that we rewrite this file on each start command so we propagate changes
+        // from `LocalEnv`'s config file (`.neon/config`).
+        tokio::fs::write(
+            &pg_data_path.join("postgresql.conf"),
+            format!("port = {}\nfsync=off\n", self.postgres_port),
+        )
+        .await?;

-        // Do we remove the pid file on stop?
-        let pg_started = self.is_postgres_running().await?;
-        let pg_lib_dir = self.get_pg_lib_dir().await?;
+        println!("Starting storage controller database...");
+        let db_start_args = [
+            "-w",
+            "-D",
+            pg_data_path.as_ref(),
+            "-l",
+            pg_log_path.as_ref(),
+            "start",
+        ];

-        if !pg_started {
-            // Start a vanilla Postgres process used by the storage controller for persistence.
-            let pg_data_path = Utf8PathBuf::from_path_buf(self.env.base_data_dir.clone())
-                .unwrap()
-                .join("storage_controller_db");
-            let pg_bin_dir = self.get_pg_bin_dir().await?;
-            let pg_log_path = pg_data_path.join("postgres.log");
+        background_process::start_process(
+            "storage_controller_db",
+            &self.env.base_data_dir,
+            pg_bin_dir.join("pg_ctl").as_std_path(),
+            db_start_args,
+            vec![
+                ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
+                ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
+            ],
+            background_process::InitialPidFile::Create(self.postgres_pid_file()),
+            retry_timeout,
+            || self.pg_isready(&pg_bin_dir),
+        )
+        .await?;

-            if !tokio::fs::try_exists(&pg_data_path).await? {
-                // Initialize empty database
-                let initdb_path = pg_bin_dir.join("initdb");
-                let mut child = Command::new(&initdb_path)
-                    .envs(vec![
-                        ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
-                        ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
-                    ])
-                    .args(["-D", pg_data_path.as_ref()])
-                    .spawn()
-                    .expect("Failed to spawn initdb");
-                let status = child.wait().await?;
-                if !status.success() {
-                    anyhow::bail!("initdb failed with status {status}");
-                }
-            };
-
-            // Write a minimal config file:
-            // - Specify the port, since this is chosen dynamically
-            // - Switch off fsync, since we're running on lightweight test environments and when e.g. scale testing
-            //   the storage controller we don't want a slow local disk to interfere with that.
-            //
-            // NB: it's important that we rewrite this file on each start command so we propagate changes
-            // from `LocalEnv`'s config file (`.neon/config`).
-            tokio::fs::write(
-                &pg_data_path.join("postgresql.conf"),
-                format!("port = {}\nfsync=off\n", postgres_port),
-            )
-            .await?;
-
-            println!("Starting storage controller database...");
-            let db_start_args = [
-                "-w",
-                "-D",
-                pg_data_path.as_ref(),
-                "-l",
-                pg_log_path.as_ref(),
-                "start",
-            ];
-
-            background_process::start_process(
-                "storage_controller_db",
-                &self.env.base_data_dir,
-                pg_bin_dir.join("pg_ctl").as_std_path(),
-                db_start_args,
-                vec![
-                    ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
-                    ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
-                ],
-                background_process::InitialPidFile::Create(self.postgres_pid_file()),
-                &start_args.start_timeout,
-                || self.pg_isready(&pg_bin_dir, postgres_port),
-            )
-            .await?;
-
-            self.setup_database(postgres_port).await?;
-        }
-
-        let database_url = format!("postgresql://localhost:{}/{DB_NAME}", postgres_port);
+        // Run migrations on every startup, in case something changed.
+        let database_url = self.setup_database().await?;

        // We support running a startup SQL script to fiddle with the database before we launch storcon.
        // This is used by the test suite.
@@ -407,7 +339,7 @@ impl StorageController {
                }
            }
        };
-        let (mut client, conn) = self.connect_to_database(postgres_port).await?;
+        let (mut client, conn) = self.connect_to_database().await?;
        let conn = tokio::spawn(conn);
        let tx = client.build_transaction();
        let tx = tx.start().await?;
@@ -416,20 +348,9 @@ impl StorageController {
        drop(client);
        conn.await??;

-        let listen = self
-            .listen
-            .get()
-            .expect("cell is set earlier in this function");
-        let address_for_peers = Uri::builder()
-            .scheme("http")
-            .authority(format!("{}:{}", listen.ip(), listen.port()))
-            .path_and_query("")
-            .build()
-            .unwrap();
-
        let mut args = vec![
            "-l",
-            &listen.to_string(),
+            &self.listen,
            "--dev",
            "--database-url",
            &database_url,
@@ -437,27 +358,15 @@ impl StorageController {
            &humantime::Duration::from(self.config.max_offline).to_string(),
            "--max-warming-up-interval",
            &humantime::Duration::from(self.config.max_warming_up).to_string(),
-            "--address-for-peers",
-            &address_for_peers.to_string(),
        ]
        .into_iter()
        .map(|s| s.to_string())
        .collect::<Vec<_>>();
-
-        if self.config.start_as_candidate {
-            args.push("--start-as-candidate".to_string());
-        }
-
        if let Some(private_key) = &self.private_key {
            let claims = Claims::new(None, Scope::PageServerApi);
            let jwt_token =
                encode_from_key_file(&claims, private_key).expect("failed to generate jwt token");
            args.push(format!("--jwt-token={jwt_token}"));
-
-            let peer_claims = Claims::new(None, Scope::Admin);
-            let peer_jwt_token = encode_from_key_file(&peer_claims, private_key)
-                .expect("failed to generate jwt token");
-            args.push(format!("--peer-jwt-token={peer_jwt_token}"));
        }

        if let Some(public_key) = &self.public_key {
@@ -485,15 +394,15 @@ impl StorageController {

        background_process::start_process(
            COMMAND,
-            &instance_dir,
+            &self.env.base_data_dir,
            &self.env.storage_controller_bin(),
            args,
            vec![
                ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
                ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
            ],
-            background_process::InitialPidFile::Create(self.pid_file(start_args.instance_id)),
-            &start_args.start_timeout,
+            background_process::InitialPidFile::Create(self.pid_file()),
+            retry_timeout,
            || async {
                match self.ready().await {
                    Ok(_) => Ok(true),
@@ -506,35 +415,8 @@ impl StorageController {
        Ok(())
    }

-    pub async fn stop(&self, stop_args: NeonStorageControllerStopArgs) -> anyhow::Result<()> {
-        background_process::stop_process(
-            stop_args.immediate,
-            COMMAND,
-            &self.pid_file(stop_args.instance_id),
-        )?;
-
-        let storcon_instances = self.env.storage_controller_instances().await?;
-        for (instance_id, instanced_dir_path) in storcon_instances {
-            if instance_id == stop_args.instance_id {
-                continue;
-            }
-
-            let pid_file = instanced_dir_path.join("storage_controller.pid");
-            let pid = tokio::fs::read_to_string(&pid_file)
-                .await
-                .map_err(|err| {
-                    anyhow::anyhow!("Failed to read storcon pid file at {pid_file:?}: {err}")
-                })?
-                .parse::<i32>()
-                .expect("pid is valid i32");
-
-            let other_proc_alive = !background_process::process_has_stopped(Pid::from_raw(pid))?;
-            if other_proc_alive {
-                // There is another storage controller instance running, so we return
-                // and leave the database running.
-                return Ok(());
-            }
-        }
+    pub async fn stop(&self, immediate: bool) -> anyhow::Result<()> {
+        background_process::stop_process(immediate, COMMAND, &self.pid_file())?;

        let pg_data_path = self.env.base_data_dir.join("storage_controller_db");
        let pg_bin_dir = self.get_pg_bin_dir().await?;
@@ -547,51 +429,27 @@ impl StorageController {
            .wait()
            .await?;
        if !stop_status.success() {
-            match self.is_postgres_running().await {
-                Ok(false) => {
-                    println!("Storage controller database is already stopped");
-                    return Ok(());
-                }
-                Ok(true) => {
-                    anyhow::bail!("Failed to stop storage controller database");
-                }
-                Err(err) => {
-                    anyhow::bail!("Failed to stop storage controller database: {err}");
-                }
+            let pg_status_args = ["-D", &pg_data_path.to_string_lossy(), "status"];
+            let status_exitcode = Command::new(pg_bin_dir.join("pg_ctl"))
+                .args(pg_status_args)
+                .spawn()?
+                .wait()
+                .await?;
+
+            // pg_ctl status returns this exit code if postgres is not running: in this case it is
+            // fine that stop failed.  Otherwise it is an error that stop failed.
+            const PG_STATUS_NOT_RUNNING: i32 = 3;
+            if Some(PG_STATUS_NOT_RUNNING) == status_exitcode.code() {
+                println!("Storage controller database is already stopped");
+                return Ok(());
+            } else {
+                anyhow::bail!("Failed to stop storage controller database: {stop_status}")
            }
        }

        Ok(())
    }

-    async fn is_postgres_running(&self) -> anyhow::Result<bool> {
-        let pg_data_path = self.env.base_data_dir.join("storage_controller_db");
-        let pg_bin_dir = self.get_pg_bin_dir().await?;
-
-        let pg_status_args = ["-D", &pg_data_path.to_string_lossy(), "status"];
-        let status_exitcode = Command::new(pg_bin_dir.join("pg_ctl"))
-            .args(pg_status_args)
-            .spawn()?
-            .wait()
-            .await?;
-
-        // pg_ctl status returns this exit code if postgres is not running: in this case it is
-        // fine that stop failed.  Otherwise it is an error that stop failed.
-        const PG_STATUS_NOT_RUNNING: i32 = 3;
-        const PG_NO_DATA_DIR: i32 = 4;
-        const PG_STATUS_RUNNING: i32 = 0;
-        match status_exitcode.code() {
-            Some(PG_STATUS_NOT_RUNNING) => Ok(false),
-            Some(PG_NO_DATA_DIR) => Ok(false),
-            Some(PG_STATUS_RUNNING) => Ok(true),
-            Some(code) => Err(anyhow::anyhow!(
-                "pg_ctl status returned unexpected status code: {:?}",
-                code
-            )),
-            None => Err(anyhow::anyhow!("pg_ctl status returned no status code")),
-        }
-    }
-
    fn get_claims_for_path(path: &str) -> anyhow::Result<Option<Claims>> {
        let category = match path.find('/') {
            Some(idx) => &path[..idx],
@@ -617,31 +475,15 @@ impl StorageController {
        RQ: Serialize + Sized,
        RS: DeserializeOwned + Sized,
    {
-        // In the special case of the `storage_controller start` subcommand, we wish
-        // to use the API endpoint of the newly started storage controller in order
-        // to pass the readiness check. In this scenario [`Self::listen`] will be set
-        // (see [`Self::start`]).
-        //
-        // Otherwise, we infer the storage controller api endpoint from the configured
-        // control plane API.
-        let url = if let Some(socket_addr) = self.listen.get() {
-            Url::from_str(&format!(
-                "http://{}:{}/{path}",
-                socket_addr.ip().to_canonical(),
-                socket_addr.port()
-            ))
-            .unwrap()
-        } else {
-            // The configured URL has the /upcall path prefix for pageservers to use: we will strip that out
-            // for general purpose API access.
-            let listen_url = self.env.control_plane_api.clone().unwrap();
-            Url::from_str(&format!(
-                "http://{}:{}/{path}",
-                listen_url.host_str().unwrap(),
-                listen_url.port().unwrap()
-            ))
-            .unwrap()
-        };
+        // The configured URL has the /upcall path prefix for pageservers to use: we will strip that out
+        // for general purpose API access.
+        let listen_url = self.env.control_plane_api.clone().unwrap();
+        let url = Url::from_str(&format!(
+            "http://{}:{}/{path}",
+            listen_url.host_str().unwrap(),
+            listen_url.port().unwrap()
+        ))
+        .unwrap();

        let mut builder = self.client.request(method, url);
        if let Some(body) = body {
--- a/control_plane/storcon_cli/src/main.rs
+++ b/control_plane/storcon_cli/src/main.rs
@@ -147,9 +147,9 @@ enum Command {
        #[arg(long)]
        threshold: humantime::Duration,
    },
-    // Migrate away from a set of specified pageservers by moving the primary attachments to pageservers
+    // Drain a set of specified pageservers by moving the primary attachments to pageservers
    // outside of the specified set.
-    BulkMigrate {
+    Drain {
        // Set of pageserver node ids to drain.
        #[arg(long)]
        nodes: Vec<NodeId>,
@@ -163,34 +163,6 @@ enum Command {
        #[arg(long)]
        dry_run: Option<bool>,
    },
-    /// Start draining the specified pageserver.
-    /// The drain is complete when the schedulling policy returns to active.
-    StartDrain {
-        #[arg(long)]
-        node_id: NodeId,
-    },
-    /// Cancel draining the specified pageserver and wait for `timeout`
-    /// for the operation to be canceled. May be retried.
-    CancelDrain {
-        #[arg(long)]
-        node_id: NodeId,
-        #[arg(long)]
-        timeout: humantime::Duration,
-    },
-    /// Start filling the specified pageserver.
-    /// The drain is complete when the schedulling policy returns to active.
-    StartFill {
-        #[arg(long)]
-        node_id: NodeId,
-    },
-    /// Cancel filling the specified pageserver and wait for `timeout`
-    /// for the operation to be canceled. May be retried.
-    CancelFill {
-        #[arg(long)]
-        node_id: NodeId,
-        #[arg(long)]
-        timeout: humantime::Duration,
-    },
 }

 #[derive(Parser)]
@@ -277,34 +249,6 @@ impl FromStr for NodeAvailabilityArg {
    }
 }

-async fn wait_for_scheduling_policy<F>(
-    client: Client,
-    node_id: NodeId,
-    timeout: Duration,
-    f: F,
-) -> anyhow::Result<NodeSchedulingPolicy>
-where
-    F: Fn(NodeSchedulingPolicy) -> bool,
-{
-    let waiter = tokio::time::timeout(timeout, async move {
-        loop {
-            let node = client
-                .dispatch::<(), NodeDescribeResponse>(
-                    Method::GET,
-                    format!("control/v1/node/{node_id}"),
-                    None,
-                )
-                .await?;
-
-            if f(node.scheduling) {
-                return Ok::<NodeSchedulingPolicy, mgmt_api::Error>(node.scheduling);
-            }
-        }
-    });
-
-    Ok(waiter.await??)
-}
-
 #[tokio::main]
 async fn main() -> anyhow::Result<()> {
    let cli = Cli::parse();
@@ -678,13 +622,12 @@ async fn main() -> anyhow::Result<()> {
                                threshold: threshold.into(),
                            },
                        )),
-                        heatmap_period: Some("300s".to_string()),
                        ..Default::default()
                    },
                })
                .await?;
        }
-        Command::BulkMigrate {
+        Command::Drain {
            nodes,
            concurrency,
            max_shards,
@@ -713,7 +656,7 @@ async fn main() -> anyhow::Result<()> {
            }

            if nodes.len() != node_to_drain_descs.len() {
-                anyhow::bail!("Bulk migration requested away from node which doesn't exist.")
+                anyhow::bail!("Drain requested for node which doesn't exist.")
            }

            node_to_fill_descs.retain(|desc| {
@@ -725,7 +668,7 @@ async fn main() -> anyhow::Result<()> {
            });

            if node_to_fill_descs.is_empty() {
-                anyhow::bail!("There are no nodes to migrate to")
+                anyhow::bail!("There are no nodes to drain to")
            }

            // Set the node scheduling policy to draining for the nodes which
@@ -746,7 +689,7 @@ async fn main() -> anyhow::Result<()> {
                    .await?;
            }

-            // Perform the migration: move each tenant shard scheduled on a node to
+            // Perform the drain: move each tenant shard scheduled on a node to
            // be drained to a node which is being filled. A simple round robin
            // strategy is used to pick the new node.
            let tenants = storcon_client
@@ -759,13 +702,13 @@ async fn main() -> anyhow::Result<()> {

            let mut selected_node_idx = 0;

-            struct MigrationMove {
+            struct DrainMove {
                tenant_shard_id: TenantShardId,
                from: NodeId,
                to: NodeId,
            }

-            let mut moves: Vec<MigrationMove> = Vec::new();
+            let mut moves: Vec<DrainMove> = Vec::new();

            let shards = tenants
                .into_iter()
@@ -795,7 +738,7 @@ async fn main() -> anyhow::Result<()> {
                    continue;
                }

-                moves.push(MigrationMove {
+                moves.push(DrainMove {
                    tenant_shard_id: shard.tenant_shard_id,
                    from: shard
                        .node_attached
@@ -872,67 +815,6 @@ async fn main() -> anyhow::Result<()> {
                failure
            );
        }
-        Command::StartDrain { node_id } => {
-            storcon_client
-                .dispatch::<(), ()>(
-                    Method::PUT,
-                    format!("control/v1/node/{node_id}/drain"),
-                    None,
-                )
-                .await?;
-            println!("Drain started for {node_id}");
-        }
-        Command::CancelDrain { node_id, timeout } => {
-            storcon_client
-                .dispatch::<(), ()>(
-                    Method::DELETE,
-                    format!("control/v1/node/{node_id}/drain"),
-                    None,
-                )
-                .await?;
-
-            println!("Waiting for node {node_id} to quiesce on scheduling policy ...");
-
-            let final_policy =
-                wait_for_scheduling_policy(storcon_client, node_id, *timeout, |sched| {
-                    use NodeSchedulingPolicy::*;
-                    matches!(sched, Active | PauseForRestart)
-                })
-                .await?;
-
-            println!(
-                "Drain was cancelled for node {node_id}. Schedulling policy is now {final_policy:?}"
-            );
-        }
-        Command::StartFill { node_id } => {
-            storcon_client
-                .dispatch::<(), ()>(Method::PUT, format!("control/v1/node/{node_id}/fill"), None)
-                .await?;
-
-            println!("Fill started for {node_id}");
-        }
-        Command::CancelFill { node_id, timeout } => {
-            storcon_client
-                .dispatch::<(), ()>(
-                    Method::DELETE,
-                    format!("control/v1/node/{node_id}/fill"),
-                    None,
-                )
-                .await?;
-
-            println!("Waiting for node {node_id} to quiesce on scheduling policy ...");
-
-            let final_policy =
-                wait_for_scheduling_policy(storcon_client, node_id, *timeout, |sched| {
-                    use NodeSchedulingPolicy::*;
-                    matches!(sched, Active)
-                })
-                .await?;
-
-            println!(
-                "Fill was cancelled for node {node_id}. Schedulling policy is now {final_policy:?}"
-            );
-        }
    }

    Ok(())
--- a/deny.toml
+++ b/deny.toml
@@ -22,10 +22,7 @@ feature-depth = 1
 [advisories]
 db-urls = ["https://github.com/rustsec/advisory-db"]
 yanked = "warn"
-
-[[advisories.ignore]]
-id = "RUSTSEC-2023-0071"
-reason = "the marvin attack only affects private key decryption, not public key signature verification"
+ignore = []

 # This section is considered when running `cargo deny check licenses`
 # More documentation for the licenses section can be found here:
--- a/docs/rfcs/033-storage-controller-drain-and-fill.md
+++ b/docs/rfcs/033-storage-controller-drain-and-fill.md
@@ -14,7 +14,7 @@ picked tenant (which requested on-demand activation) for around 30 seconds
 during the restart at 2024-04-03 16:37 UTC.

 Note that lots of shutdowns on loaded pageservers do not finish within the
-[10 second systemd enforced timeout](https://github.com/neondatabase/infra/blob/0a5280b383e43c063d43cbf87fa026543f6d6ad4/.github/ansible/systemd/pageserver.service#L16). This means we are shutting down without flushing ephemeral layers
+[10 second systemd enforced timeout](https://github.com/neondatabase/aws/blob/0a5280b383e43c063d43cbf87fa026543f6d6ad4/.github/ansible/systemd/pageserver.service#L16). This means we are shutting down without flushing ephemeral layers
 and have to reingest data in order to serve requests after restarting, potentially making first request latencies worse.

 This problem is not yet very acutely felt in storage controller managed pageservers since
--- a/docs/rfcs/036-physical-replication.md
+++ b/docs/rfcs/036-physical-replication.md
@@ -1,265 +0,0 @@
-# Physical Replication
-
-This RFC is a bit special in that we have already implemented physical
-replication a long time ago. However, we never properly wrote down all
-the decisions and assumptions, and in the last months when more users
-have started to use the feature, numerous issues have surfaced.
-
-This RFC documents the design decisions that have been made.
-
-## Summary
-
-PostgreSQL has a feature called streaming replication, where a replica
-streams WAL from the primary and continuously applies it. It is also
-known as "physical replication", to distinguish it from logical
-replication.  In PostgreSQL, a replica is initialized by taking a
-physical backup of the primary. In Neon, the replica is initialized
-from a slim "base backup" from the pageserver, just like a primary,
-and the primary and the replicas connect to the same pageserver,
-sharing the storage.
-
-There are two kinds of read-only replicas in Neon:
- replicas that follow the primary, and
- "static" replicas that are pinned at a particular LSN.
-
-A static replica is useful e.g. for performing time-travel queries and
-running one-off slow queries without affecting the primary. A replica
-that follows the primary can be used e.g. to scale out read-only
-workloads.
-
-## Motivation
-
-Read-only replicas allow offloading read-only queries. It's useful for
-isolation, if you want to make sure that read-only queries don't
-affect the primary, and it's also an easy way to provide guaranteed
-read-only access to an application, without having to mess with access
-controls.
-
-## Non Goals (if relevant)
-
-This RFC is all about WAL-based *physical* replication. Logical
-replication is a different feature.
-
-Neon also has the capability to launch "static" read-only nodes which
-do not follow the primary, but are pinned to a particular LSN. They
-can be used for long-running one-off queries, or for Point-in-time
-queries. They work similarly to read replicas that follow the primary,
-but some things are simpler: there are no concerns about cache
-invalidation when the data changes on the primary, or worrying about
-transactions that are in-progress on the primary.
-
-## Impacted components (e.g. pageserver, safekeeper, console, etc)
-
- Control plane launches the replica
- Replica Postgres instance connects to the safekeepers, to stream the WAL
- The primary does not know about the standby, except for the hot standby feedback
- The primary and replicas all connect to the same pageservers
-
-
-# Context
-
-Some useful things to know about hot standby and replicas in
-PostgreSQL.
-
-## PostgreSQL startup sequence
-
-"Running" and "start up" terms are little imprecise. PostgreSQL
-replica startup goes through several stages:
-
-1. First, the process is started up, and various initialization steps
-   are performed, like initializing shared memory. If you try to
-   connect to the server in this stage, you get an error: ERROR: the
-   database system is starting up. This stage happens very quickly, no
-
-2. Then the server reads the checpoint record from the WAL and starts
-   the WAL replay starting from the checkpoint. This works differently
-   in Neon: we start the WAL replay at the basebackup LSN, not from a
-   checkpoint! If you connect to the server in this state, you get an
-   error: ERROR: the database system is not yet accepting
-   connections. We proceed to the next stage, when the WAL replay sees
-   a running-xacts record. Or in Neon, the "CLOG scanning" mechanism
-   can allow us to move directly to next stage, with all the caveats
-   listed in this RFC.
-
-3. When the running-xacts information is established, the server
-   starts to accept connections normally.
-
-From PostgreSQL's point of view, the server is already running in
-stage 2, even though it's not accepting connections yet. Our
-`compute_ctl` does not consider it as running until stage 3. If the
-transition from stage 2 to 3 doesn't happen fast enough, the control
-plane will mark the start operation as failed.
-
-
-## Decisions, Issues
-
-### Cache invalidation in replica
-
-When a read replica follows the primary in PostgreSQL, it needs to
-stream all the WAL from the primary and apply all the records, to keep
-the local copy of the data consistent with the primary. In Neon, the
-replica can fetch the updated page versions from the pageserver, so
-it's not necessary to apply all the WAL. However, it needs to ensure
-that any pages that are currently in the Postgres buffer cache, or the
-Local File Cache, are either updated, or thrown away so that the next
-read of the page will fetch the latest version.
-
-We choose to apply the WAL records for pages that are already in the
-buffer cache, and skip records for other pages. Somewhat arbitrarily,
-we also apply records affecting catalog relations, fetching the old
-page version from the pageserver if necessary first. See
-`neon_redo_read_buffer_filter()` function.
-
-The replica wouldn't necessarily need to see all the WAL records, only
-the records that apply to cached pages. For simplicity, we do stream
-all the WAL to the replica, and the replica simply ignores WAL records
-that require no action.
-
-Like in PostgreSQL, the read replica maintains a "replay LSN", which
-is the LSN up to which the replica has received and replayed the
-WAL. The replica can lag behind the primary, if it cannot quite keep
-up with the primary, or if a long-running query conflicts with changes
-that are about to be applied, or even intentionally if the user wishes
-to see delayed data (see recovery_min_apply_delay). It's important
-that the replica sees a consistent view of the whole cluster at the
-replay LSN, when it's lagging behind.
-
-In Neon, the replica connects to a safekeeper to get the WAL
-stream. That means that the safekeepers must be able to regurgitate
-the original WAL as far back as the replay LSN of any running read
-replica. (A static read-only node that does not follow the primary
-does not require a WAL stream however). The primary does not need to
-be running, and when it is, the replicas don't incur any extra
-overhead to the primary (see hot standby feedback though).
-
-### In-progress transactions
-
-In PostgreSQL, when a hot standby server starts up, it cannot
-immediately open up for queries (see [PostgreSQL startup
-sequence]). It first needs to establish a complete list of in-progress
-transactions, including subtransactions, that are running at the
-primary, at the current replay LSN. Normally that happens quickly,
-when the replica sees a "running-xacts" WAL record, because the
-primary writes a running-xacts WAL record at every checkpoint, and in
-PostgreSQL the replica always starts the WAL replay from a checkpoint
-REDO point. (A shutdown checkpoint WAL record also implies that all
-the non-prepared transactions have ended.) If there are a lot of
-subtransactions in progress, however, the standby might need to wait
-for old transactions to complete before it can open up for queries.
-
-In Neon that problem is worse: a replica can start at any LSN, so
-there's no guarantee that it will see a running-xacts record any time
-soon. In particular, if the primary is not running when the replica is
-started, it might never see a running-xacts record.
-
-To make things worse, we initially missed this issue, and always
-started accepting queries at replica startup, even if it didn't have
-the transaction information. That could lead to incorrect query
-results and data corruption later. However, as we fixed that, we
-introduced a new problem compared to what we had before: previously
-the replica would always start up, but after fixing that bug, it might
-not. In a superficial way, the old behavior was better (but could lead
-to serious issues later!). That made fixing that bug was very hard,
-because as we fixed it, we made things (superficially) worse for
-others.
-
-See https://github.com/neondatabase/neon/pull/7288 which fixed the
-bug, and follow-up PRs https://github.com/neondatabase/neon/pull/8323
-and https://github.com/neondatabase/neon/pull/8484 to try to claw back
-the cases that started to cause trouble as fixing it. As of this
-writing, there are still cases where a replica might not immediately
-start up, causing the control plane operation to fail, the remaining
-issues are tracked in https://github.com/neondatabase/neon/issues/6211.
-
-One long-term fix for this is to switch to using so-called CSN
-snapshots in read replica. That would make it unnecessary to have the
-full in-progress transaction list in the replica at startup time. See
-https://commitfest.postgresql.org/48/4912/ for a work-in-progress
-patch to upstream to implement that.
-
-Another thing we could do is to teach the control plane about that
-distinction between "starting up" and "running but haven't received
-running-xacts information yet", so that we could keep the replica
-waiting longer in that stage, and also give any client connections the
-same `ERROR: the database system is not yet accepting connections`
-error that you get in standalone PostgreSQL in that state.
-
-
-### Recovery conflicts and Hot standby feedback
-
-It's possible that a tuple version is vacuumed away in the primary,
-even though it is still needed by a running transactions in the
-replica. This is called a "recovery conflict", and PostgreSQL provides
-various options for dealing with it. By default, the WAL replay will
-wait up to 30 s for the conflicting query to finish. After that, it
-will kill the running query, so that the WAL replay can proceed.
-
-Another way to avoid the situation is to enable the
-[`hot_standby_feedback`](https://www.postgresql.org/docs/current/runtime-config-replication.html#GUC-HOT-STANDBY-FEEDBACK)
-option. When it is enabled, the primary will refrain from vacuuming
-tuples that are still needed in the primary. That means potentially
-bloating the primary, which violates the usual rule that read replicas
-don't affect the operations on the primary, which is why it's off by
-default. We leave it to users to decide if they want to turn it on,
-same as PostgreSQL.
-
-Neon supports `hot_standby_feedback` by passing the feedback messages
-from the replica to the safekeepers, and from safekeepers to the
-primary.
-
-### Relationship of settings between primary and replica
-
-In order to enter hot standby mode, some configuration options need to
-be set to the same or larger values in the standby, compared to the
-primary.  See [explanation in the PostgreSQL
-docs](https://www.postgresql.org/docs/current/hot-standby.html#HOT-STANDBY-ADMIN)
-
-In Neon, we have this problem too. To prevent customers from hitting
-it, the control plane automatically adjusts the settings of a replica,
-so that they match or exceed the primary's settings (see
-https://github.com/neondatabase/cloud/issues/14903). However, you
-can still hit the issue if the primary is restarted with larger
-settings, while the replica is running.
-
-
-### Interaction with Pageserver GC
-
-The read replica can lag behind the primary. If there are recovery
-conflicts or the replica cannot keep up for some reason, the lag can
-in principle grow indefinitely. The replica will issue all GetPage
-requests to the pageservers at the current replay LSN, and needs to
-see the old page versions.
-
-If the retention period in the pageserver is set to be small, it may
-have already garbage collected away the old page versions. That will
-cause read errors in the compute, and can mean that the replica cannot
-make progress with the replication anymore.
-
-There is a mechanism for replica to pass information about its replay
-LSN to the pageserver, so that the pageserver refrains from GC'ing
-data that is still needed by the standby. It's called
-'standby_horizon' in the pageserver code, see
-https://github.com/neondatabase/neon/pull/7368. A separate "lease"
-mechanism also is in the works, where the replica could hold a lease
-on the old LSN, preventing the pageserver from advancing the GC
-horizon past that point. The difference is that the standby_horizon
-mechanism relies on a feedback message from replica to safekeeper,
-while the least API is exposed directly from the pageserver. A static
-read-only node is not connected to safekeepers, so it cannot use the
-standby_horizon mechanism.
-
-
-### Synchronous replication
-
-We haven't put any effort into synchronous replication yet.
-
-PostgreSQL provides multiple levels of synchronicity. In the weaker
-levels, a transaction is not acknowledged as committed to the client
-in the primary until the WAL has been streamed to a replica or flushed
-to disk there. Those modes don't make senses in Neon, because the
-safekeepers handle durability.
-
-`synchronous_commit=remote_apply` mode would make sense. In that mode,
-the commit is not acknowledged to the client until it has been
-replayed in the replica. That ensures that after commit, you can see
-the commit in the replica too (aka. read-your-write consistency).
--- a/docs/updating-postgres.md
+++ b/docs/updating-postgres.md
@@ -21,21 +21,30 @@ _Example: 15.4 is the new minor version to upgrade to from 15.3._
 1. Create a new branch based on the stable branch you are updating.

    ```shell
-    git checkout -b my-branch-15 REL_15_STABLE_neon
+    git checkout -b my-branch REL_15_STABLE_neon
    ```

-1. Find the upstream release tags you're looking for. They are of the form `REL_X_Y`.
+1. Tag the last commit on the stable branch you are updating.

-1. Merge the upstream tag into the branch you created on the tag and resolve any conflicts.
+    ```shell
+    git tag REL_15_3_neon
+    ```
+
+1. Push the new tag to the Neon Postgres repository.
+
+    ```shell
+    git push origin REL_15_3_neon
+    ```
+
+1. Find the release tags you're looking for. They are of the form `REL_X_Y`.
+
+1. Rebase the branch you created on the tag and resolve any conflicts.

    ```shell
    git fetch upstream REL_15_4
-    git merge REL_15_4
+    git rebase REL_15_4
    ```

-    In the commit message of the merge commit, mention if there were
-    any non-trivial conflicts or other issues.
-
 1. Run the Postgres test suite to make sure our commits have not affected
 Postgres in a negative way.

@@ -48,7 +57,7 @@ Postgres in a negative way.
 1. Push your branch to the Neon Postgres repository.

    ```shell
-    git push origin my-branch-15
+    git push origin my-branch
    ```

 1. Clone the Neon repository if you have not done so already.
@@ -65,7 +74,7 @@ branch.
 1. Update the Git submodule.

    ```shell
-    git submodule set-branch --branch my-branch-15 vendor/postgres-v15
+    git submodule set-branch --branch my-branch vendor/postgres-v15
    git submodule update --remote vendor/postgres-v15
    ```

@@ -80,12 +89,14 @@ minor Postgres release.

 1. Create a pull request, and wait for CI to go green.

-1. Push the Postgres branches with the merge commits into the Neon Postgres repository.
+1. Force push the rebased Postgres branches into the Neon Postgres repository.

    ```shell
-    git push origin my-branch-15:REL_15_STABLE_neon
+    git push --force origin my-branch:REL_15_STABLE_neon
    ```

+    It may require disabling various branch protections.
+
 1. Update your Neon PR to point at the branches.

    ```shell
--- a/libs/compute_api/Cargo.toml
+++ b/libs/compute_api/Cargo.toml
@@ -14,3 +14,5 @@ regex.workspace = true

 utils = { path = "../utils" }
 remote_storage = { version = "0.1", path = "../remote_storage/" }
+
+workspace_hack.workspace = true
--- a/libs/consumption_metrics/Cargo.toml
+++ b/libs/consumption_metrics/Cargo.toml
@@ -6,8 +6,10 @@ license = "Apache-2.0"

 [dependencies]
 anyhow.workspace = true
-chrono = { workspace = true, features = ["serde"] }
+chrono.workspace = true
 rand.workspace = true
 serde.workspace = true
 serde_with.workspace = true
 utils.workspace = true
+
+workspace_hack.workspace = true
--- a/libs/desim/Cargo.toml
+++ b/libs/desim/Cargo.toml
@@ -14,3 +14,5 @@ parking_lot.workspace = true
 hex.workspace = true
 scopeguard.workspace = true
 smallvec = { workspace = true, features = ["write"] }
+
+workspace_hack.workspace = true
--- a/libs/metrics/Cargo.toml
+++ b/libs/metrics/Cargo.toml
@@ -12,6 +12,8 @@ chrono.workspace = true
 twox-hash.workspace = true
 measured.workspace = true

+workspace_hack.workspace = true
+
 [target.'cfg(target_os = "linux")'.dependencies]
 procfs.workspace = true
 measured-process.workspace = true
--- a/libs/pageserver_api/Cargo.toml
+++ b/libs/pageserver_api/Cargo.toml
@@ -21,9 +21,11 @@ hex.workspace = true
 humantime.workspace = true
 thiserror.workspace = true
 humantime-serde.workspace = true
-chrono = { workspace = true, features = ["serde"] }
+chrono.workspace = true
 itertools.workspace = true

+workspace_hack.workspace = true
+
 [dev-dependencies]
 bincode.workspace = true
 rand.workspace = true
--- a/libs/pageserver_api/src/controller_api.rs
+++ b/libs/pageserver_api/src/controller_api.rs
@@ -8,7 +8,6 @@ use std::time::{Duration, Instant};
 use serde::{Deserialize, Serialize};
 use utils::id::{NodeId, TenantId};

-use crate::models::PageserverUtilization;
 use crate::{
    models::{ShardParameters, TenantConfig},
    shard::{ShardStripeSize, TenantShardId},
@@ -141,11 +140,23 @@ pub struct TenantShardMigrateRequest {
    pub node_id: NodeId,
 }

-#[derive(Serialize, Clone, Debug)]
+/// Utilisation score indicating how good a candidate a pageserver
+/// is for scheduling the next tenant. See [`crate::models::PageserverUtilization`].
+/// Lower values are better.
+#[derive(Serialize, Deserialize, Clone, Copy, Eq, PartialEq, PartialOrd, Ord, Debug)]
+pub struct UtilizationScore(pub u64);
+
+impl UtilizationScore {
+    pub fn worst() -> Self {
+        UtilizationScore(u64::MAX)
+    }
+}
+
+#[derive(Serialize, Clone, Copy, Debug)]
 #[serde(into = "NodeAvailabilityWrapper")]
 pub enum NodeAvailability {
    // Normal, happy state
-    Active(PageserverUtilization),
+    Active(UtilizationScore),
    // Node is warming up, but we expect it to become available soon. Covers
    // the time span between the re-attach response being composed on the storage controller
    // and the first successful heartbeat after the processing of the re-attach response
@@ -184,9 +195,7 @@ impl From<NodeAvailabilityWrapper> for NodeAvailability {
        match val {
            // Assume the worst utilisation score to begin with. It will later be updated by
            // the heartbeats.
-            NodeAvailabilityWrapper::Active => {
-                NodeAvailability::Active(PageserverUtilization::full())
-            }
+            NodeAvailabilityWrapper::Active => NodeAvailability::Active(UtilizationScore::worst()),
            NodeAvailabilityWrapper::WarmingUp => NodeAvailability::WarmingUp(Instant::now()),
            NodeAvailabilityWrapper::Offline => NodeAvailability::Offline,
        }
@@ -304,17 +313,20 @@ pub struct MetadataHealthUpdateRequest {
 pub struct MetadataHealthUpdateResponse {}

 #[derive(Serialize, Deserialize, Debug)]
+
 pub struct MetadataHealthListUnhealthyResponse {
    pub unhealthy_tenant_shards: Vec<TenantShardId>,
 }

 #[derive(Serialize, Deserialize, Debug)]
+
 pub struct MetadataHealthListOutdatedRequest {
    #[serde(with = "humantime_serde")]
    pub not_scrubbed_for: Duration,
 }

 #[derive(Serialize, Deserialize, Debug)]
+
 pub struct MetadataHealthListOutdatedResponse {
    pub health_records: Vec<MetadataHealthRecord>,
 }
--- a/libs/pageserver_api/src/models.rs
+++ b/libs/pageserver_api/src/models.rs
@@ -348,7 +348,7 @@ impl AuxFilePolicy {

    /// If a tenant writes aux files without setting `switch_aux_policy`, this value will be used.
    pub fn default_tenant_config() -> Self {
-        Self::V2
+        Self::V1
    }
 }

--- a/libs/pageserver_api/src/models/utilization.rs
+++ b/libs/pageserver_api/src/models/utilization.rs
@@ -38,7 +38,7 @@ pub struct PageserverUtilization {
    pub max_shard_count: u32,

    /// Cached result of [`Self::score`]
-    pub utilization_score: Option<u64>,
+    pub utilization_score: u64,

    /// When was this snapshot captured, pageserver local time.
    ///
@@ -50,8 +50,6 @@ fn unity_percent() -> Percent {
    Percent::new(0).unwrap()
 }

-pub type RawScore = u64;
-
 impl PageserverUtilization {
    const UTILIZATION_FULL: u64 = 1000000;

@@ -64,7 +62,7 @@ impl PageserverUtilization {
    /// - Negative values are forbidden
    /// - Values over UTILIZATION_FULL indicate an overloaded node, which may show degraded performance due to
    ///   layer eviction.
-    pub fn score(&self) -> RawScore {
+    pub fn score(&self) -> u64 {
        let disk_usable_capacity = ((self.disk_usage_bytes + self.free_space_bytes)
            * self.disk_usable_pct.get() as u64)
            / 100;
@@ -76,30 +74,8 @@ impl PageserverUtilization {
        std::cmp::max(disk_utilization_score, shard_utilization_score)
    }

-    pub fn cached_score(&mut self) -> RawScore {
-        match self.utilization_score {
-            None => {
-                let s = self.score();
-                self.utilization_score = Some(s);
-                s
-            }
-            Some(s) => s,
-        }
-    }
-
-    /// If a node is currently hosting more work than it can comfortably handle.  This does not indicate that
-    /// it will fail, but it is a strong signal that more work should not be added unless there is no alternative.
-    pub fn is_overloaded(score: RawScore) -> bool {
-        score >= Self::UTILIZATION_FULL
-    }
-
-    pub fn adjust_shard_count_max(&mut self, shard_count: u32) {
-        if self.shard_count < shard_count {
-            self.shard_count = shard_count;
-
-            // Dirty cache: this will be calculated next time someone retrives the score
-            self.utilization_score = None;
-        }
+    pub fn refresh_score(&mut self) {
+        self.utilization_score = self.score();
    }

    /// A utilization structure that has a full utilization score: use this as a placeholder when
@@ -112,38 +88,7 @@ impl PageserverUtilization {
            disk_usable_pct: Percent::new(100).unwrap(),
            shard_count: 1,
            max_shard_count: 1,
-            utilization_score: Some(Self::UTILIZATION_FULL),
-            captured_at: serde_system_time::SystemTime(SystemTime::now()),
-        }
-    }
-}
-
-/// Test helper
-pub mod test_utilization {
-    use super::PageserverUtilization;
-    use std::time::SystemTime;
-    use utils::{
-        serde_percent::Percent,
-        serde_system_time::{self},
-    };
-
-    // Parameters of the imaginary node used for test utilization instances
-    const TEST_DISK_SIZE: u64 = 1024 * 1024 * 1024 * 1024;
-    const TEST_SHARDS_MAX: u32 = 1000;
-
-    /// Unit test helper.  Unconditionally compiled because cfg(test) doesn't carry across crates.  Do
-    /// not abuse this function from non-test code.
-    ///
-    /// Emulates a node with a 1000 shard limit and a 1TB disk.
-    pub fn simple(shard_count: u32, disk_wanted_bytes: u64) -> PageserverUtilization {
-        PageserverUtilization {
-            disk_usage_bytes: disk_wanted_bytes,
-            free_space_bytes: TEST_DISK_SIZE - std::cmp::min(disk_wanted_bytes, TEST_DISK_SIZE),
-            disk_wanted_bytes,
-            disk_usable_pct: Percent::new(100).unwrap(),
-            shard_count,
-            max_shard_count: TEST_SHARDS_MAX,
-            utilization_score: None,
+            utilization_score: Self::UTILIZATION_FULL,
            captured_at: serde_system_time::SystemTime(SystemTime::now()),
        }
    }
@@ -175,7 +120,7 @@ mod tests {
            disk_usage_bytes: u64::MAX,
            free_space_bytes: 0,
            disk_wanted_bytes: u64::MAX,
-            utilization_score: Some(13),
+            utilization_score: 13,
            disk_usable_pct: Percent::new(90).unwrap(),
            shard_count: 100,
            max_shard_count: 200,
--- a/libs/postgres_backend/Cargo.toml
+++ b/libs/postgres_backend/Cargo.toml
@@ -18,6 +18,7 @@ tokio-rustls.workspace = true
 tracing.workspace = true

 pq_proto.workspace = true
+workspace_hack.workspace = true

 [dev-dependencies]
 once_cell.workspace = true
--- a/libs/postgres_connection/Cargo.toml
+++ b/libs/postgres_connection/Cargo.toml
@@ -11,5 +11,7 @@ postgres.workspace = true
 tokio-postgres.workspace = true
 url.workspace = true

+workspace_hack.workspace = true
+
 [dev-dependencies]
 once_cell.workspace = true
--- a/libs/postgres_ffi/Cargo.toml
+++ b/libs/postgres_ffi/Cargo.toml
@@ -19,6 +19,8 @@ thiserror.workspace = true
 serde.workspace = true
 utils.workspace = true

+workspace_hack.workspace = true
+
 [dev-dependencies]
 env_logger.workspace = true
 postgres.workspace = true
--- a/libs/postgres_ffi/src/lib.rs
+++ b/libs/postgres_ffi/src/lib.rs
@@ -143,8 +143,8 @@ pub use v14::xlog_utils::XLogFileName;

 pub use v14::bindings::DBState_DB_SHUTDOWNED;

-pub fn bkpimage_is_compressed(bimg_info: u8, version: u32) -> bool {
-    dispatch_pgversion!(version, pgv::bindings::bkpimg_is_compressed(bimg_info))
+pub fn bkpimage_is_compressed(bimg_info: u8, version: u32) -> anyhow::Result<bool> {
+    dispatch_pgversion!(version, Ok(pgv::bindings::bkpimg_is_compressed(bimg_info)))
 }

 pub fn generate_wal_segment(
--- a/libs/postgres_ffi/wal_craft/Cargo.toml
+++ b/libs/postgres_ffi/wal_craft/Cargo.toml
@@ -14,6 +14,8 @@ postgres.workspace = true
 postgres_ffi.workspace = true
 camino-tempfile.workspace = true

+workspace_hack.workspace = true
+
 [dev-dependencies]
 regex.workspace = true
 utils.workspace = true
--- a/libs/pq_proto/Cargo.toml
+++ b/libs/pq_proto/Cargo.toml
@@ -11,7 +11,9 @@ itertools.workspace = true
 pin-project-lite.workspace = true
 postgres-protocol.workspace = true
 rand.workspace = true
-tokio = { workspace = true, features = ["io-util"] }
+tokio.workspace = true
 tracing.workspace = true
 thiserror.workspace = true
 serde.workspace = true
+
+workspace_hack.workspace = true
--- a/libs/remote_storage/Cargo.toml
+++ b/libs/remote_storage/Cargo.toml
@@ -32,7 +32,7 @@ scopeguard.workspace = true
 metrics.workspace = true
 utils.workspace = true
 pin-project-lite.workspace = true
-
+workspace_hack.workspace = true
 azure_core.workspace = true
 azure_identity.workspace = true
 azure_storage.workspace = true
@@ -46,4 +46,3 @@ sync_wrapper = { workspace = true, features = ["futures"] }
 camino-tempfile.workspace = true
 test-context.workspace = true
 rand.workspace = true
-tokio = { workspace = true, features = ["test-util"] }
--- a/libs/remote_storage/src/azure_blob.rs
+++ b/libs/remote_storage/src/azure_blob.rs
@@ -383,48 +383,6 @@ impl RemoteStorage for AzureBlobStorage {
        }
    }

-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError> {
-        let kind = RequestKind::Head;
-        let _permit = self.permit(kind, cancel).await?;
-
-        let started_at = start_measuring_requests(kind);
-
-        let blob_client = self.client.blob_client(self.relative_path_to_name(key));
-        let properties_future = blob_client.get_properties().into_future();
-
-        let properties_future = tokio::time::timeout(self.timeout, properties_future);
-
-        let res = tokio::select! {
-            res = properties_future => res,
-            _ = cancel.cancelled() => return Err(TimeoutOrCancel::Cancel.into()),
-        };
-
-        if let Ok(inner) = &res {
-            // do not incl. timeouts as errors in metrics but cancellations
-            let started_at = ScopeGuard::into_inner(started_at);
-            crate::metrics::BUCKET_METRICS
-                .req_seconds
-                .observe_elapsed(kind, inner, started_at);
-        }
-
-        let data = match res {
-            Ok(Ok(data)) => Ok(data),
-            Ok(Err(sdk)) => Err(to_download_error(sdk)),
-            Err(_timeout) => Err(DownloadError::Timeout),
-        }?;
-
-        let properties = data.blob.properties;
-        Ok(ListingObject {
-            key: key.to_owned(),
-            last_modified: SystemTime::from(properties.last_modified),
-            size: properties.content_length,
-        })
-    }
-
    async fn upload(
        &self,
        from: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync + 'static,
--- a/libs/remote_storage/src/lib.rs
+++ b/libs/remote_storage/src/lib.rs
@@ -150,7 +150,7 @@ pub enum ListingMode {
    NoDelimiter,
 }

-#[derive(PartialEq, Eq, Debug, Clone)]
+#[derive(PartialEq, Eq, Debug)]
 pub struct ListingObject {
    pub key: RemotePath,
    pub last_modified: SystemTime,
@@ -215,13 +215,6 @@ pub trait RemoteStorage: Send + Sync + 'static {
        Ok(combined)
    }

-    /// Obtain metadata information about an object.
-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError>;
-
    /// Streams the local file contents into remote into the remote storage entry.
    ///
    /// If the operation fails because of timeout or cancellation, the root cause of the error will be
@@ -370,20 +363,6 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
        }
    }

-    // See [`RemoteStorage::head_object`].
-    pub async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError> {
-        match self {
-            Self::LocalFs(s) => s.head_object(key, cancel).await,
-            Self::AwsS3(s) => s.head_object(key, cancel).await,
-            Self::AzureBlob(s) => s.head_object(key, cancel).await,
-            Self::Unreliable(s) => s.head_object(key, cancel).await,
-        }
-    }
-
    /// See [`RemoteStorage::upload`]
    pub async fn upload(
        &self,
@@ -619,7 +598,6 @@ impl ConcurrencyLimiter {
            RequestKind::Delete => &self.write,
            RequestKind::Copy => &self.write,
            RequestKind::TimeTravel => &self.write,
-            RequestKind::Head => &self.read,
        }
    }

--- a/libs/remote_storage/src/local_fs.rs
+++ b/libs/remote_storage/src/local_fs.rs
@@ -445,20 +445,6 @@ impl RemoteStorage for LocalFs {
        }
    }

-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        _cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError> {
-        let target_file_path = key.with_base(&self.storage_root);
-        let metadata = file_metadata(&target_file_path).await?;
-        Ok(ListingObject {
-            key: key.clone(),
-            last_modified: metadata.modified()?,
-            size: metadata.len(),
-        })
-    }
-
    async fn upload(
        &self,
        data: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync,
--- a/libs/remote_storage/src/metrics.rs
+++ b/libs/remote_storage/src/metrics.rs
@@ -13,7 +13,6 @@ pub(crate) enum RequestKind {
    List = 3,
    Copy = 4,
    TimeTravel = 5,
-    Head = 6,
 }

 use scopeguard::ScopeGuard;
@@ -28,7 +27,6 @@ impl RequestKind {
            List => "list_objects",
            Copy => "copy_object",
            TimeTravel => "time_travel_recover",
-            Head => "head_object",
        }
    }
    const fn as_index(&self) -> usize {
@@ -36,8 +34,7 @@ impl RequestKind {
    }
 }

-const REQUEST_KIND_COUNT: usize = 7;
-pub(crate) struct RequestTyped<C>([C; REQUEST_KIND_COUNT]);
+pub(crate) struct RequestTyped<C>([C; 6]);

 impl<C> RequestTyped<C> {
    pub(crate) fn get(&self, kind: RequestKind) -> &C {
@@ -46,8 +43,8 @@ impl<C> RequestTyped<C> {

    fn build_with(mut f: impl FnMut(RequestKind) -> C) -> Self {
        use RequestKind::*;
-        let mut it = [Get, Put, Delete, List, Copy, TimeTravel, Head].into_iter();
-        let arr = std::array::from_fn::<C, REQUEST_KIND_COUNT, _>(|index| {
+        let mut it = [Get, Put, Delete, List, Copy, TimeTravel].into_iter();
+        let arr = std::array::from_fn::<C, 6, _>(|index| {
            let next = it.next().unwrap();
            assert_eq!(index, next.as_index());
            f(next)
--- a/libs/remote_storage/src/s3_bucket.rs
+++ b/libs/remote_storage/src/s3_bucket.rs
@@ -23,7 +23,7 @@ use aws_config::{
 use aws_sdk_s3::{
    config::{AsyncSleep, IdentityCache, Region, SharedAsyncSleep},
    error::SdkError,
-    operation::{get_object::GetObjectError, head_object::HeadObjectError},
+    operation::get_object::GetObjectError,
    types::{Delete, DeleteMarkerEntry, ObjectIdentifier, ObjectVersion, StorageClass},
    Client,
 };
@@ -604,78 +604,6 @@ impl RemoteStorage for S3Bucket {
        }
    }

-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError> {
-        let kind = RequestKind::Head;
-        let _permit = self.permit(kind, cancel).await?;
-
-        let started_at = start_measuring_requests(kind);
-
-        let head_future = self
-            .client
-            .head_object()
-            .bucket(self.bucket_name())
-            .key(self.relative_path_to_s3_object(key))
-            .send();
-
-        let head_future = tokio::time::timeout(self.timeout, head_future);
-
-        let res = tokio::select! {
-            res = head_future => res,
-            _ = cancel.cancelled() => return Err(TimeoutOrCancel::Cancel.into()),
-        };
-
-        let res = res.map_err(|_e| DownloadError::Timeout)?;
-
-        // do not incl. timeouts as errors in metrics but cancellations
-        let started_at = ScopeGuard::into_inner(started_at);
-        crate::metrics::BUCKET_METRICS
-            .req_seconds
-            .observe_elapsed(kind, &res, started_at);
-
-        let data = match res {
-            Ok(object_output) => object_output,
-            Err(SdkError::ServiceError(e)) if matches!(e.err(), HeadObjectError::NotFound(_)) => {
-                // Count this in the AttemptOutcome::Ok bucket, because 404 is not
-                // an error: we expect to sometimes fetch an object and find it missing,
-                // e.g. when probing for timeline indices.
-                crate::metrics::BUCKET_METRICS.req_seconds.observe_elapsed(
-                    kind,
-                    AttemptOutcome::Ok,
-                    started_at,
-                );
-                return Err(DownloadError::NotFound);
-            }
-            Err(e) => {
-                crate::metrics::BUCKET_METRICS.req_seconds.observe_elapsed(
-                    kind,
-                    AttemptOutcome::Err,
-                    started_at,
-                );
-
-                return Err(DownloadError::Other(
-                    anyhow::Error::new(e).context("s3 head object"),
-                ));
-            }
-        };
-
-        let (Some(last_modified), Some(size)) = (data.last_modified, data.content_length) else {
-            return Err(DownloadError::Other(anyhow!(
-                "head_object doesn't contain last_modified or content_length"
-            )))?;
-        };
-        Ok(ListingObject {
-            key: key.to_owned(),
-            last_modified: SystemTime::try_from(last_modified).map_err(|e| {
-                DownloadError::Other(anyhow!("can't convert time '{last_modified}': {e}"))
-            })?,
-            size: size as u64,
-        })
-    }
-
    async fn upload(
        &self,
        from: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync + 'static,
--- a/libs/remote_storage/src/simulate_failures.rs
+++ b/libs/remote_storage/src/simulate_failures.rs
@@ -30,7 +30,6 @@ pub struct UnreliableWrapper {
 #[derive(Debug, Hash, Eq, PartialEq)]
 enum RemoteOp {
    ListPrefixes(Option<RemotePath>),
-    HeadObject(RemotePath),
    Upload(RemotePath),
    Download(RemotePath),
    Delete(RemotePath),
@@ -138,16 +137,6 @@ impl RemoteStorage for UnreliableWrapper {
        self.inner.list(prefix, mode, max_keys, cancel).await
    }

-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<crate::ListingObject, DownloadError> {
-        self.attempt(RemoteOp::HeadObject(key.clone()))
-            .map_err(DownloadError::Other)?;
-        self.inner.head_object(key, cancel).await
-    }
-
    async fn upload(
        &self,
        data: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync + 'static,
--- a/libs/safekeeper_api/Cargo.toml
+++ b/libs/safekeeper_api/Cargo.toml
@@ -9,3 +9,5 @@ serde.workspace = true
 serde_with.workspace = true
 const_format.workspace = true
 utils.workspace = true
+
+workspace_hack.workspace = true
--- a/libs/tenant_size_model/Cargo.toml
+++ b/libs/tenant_size_model/Cargo.toml
@@ -9,3 +9,5 @@ license.workspace = true
 anyhow.workspace = true
 serde.workspace = true
 serde_json.workspace = true
+
+workspace_hack.workspace = true
--- a/libs/tracing-utils/Cargo.toml
+++ b/libs/tracing-utils/Cargo.toml
@@ -14,3 +14,5 @@ tokio = { workspace = true, features = ["rt", "rt-multi-thread"] }
 tracing.workspace = true
 tracing-opentelemetry.workspace = true
 tracing-subscriber.workspace = true
+
+workspace_hack.workspace = true
--- a/libs/utils/Cargo.toml
+++ b/libs/utils/Cargo.toml
@@ -39,7 +39,7 @@ thiserror.workspace = true
 tokio.workspace = true
 tokio-tar.workspace = true
 tokio-util.workspace = true
-toml_edit = { workspace = true, features = ["serde"] }
+toml_edit.workspace = true
 tracing.workspace = true
 tracing-error.workspace = true
 tracing-subscriber = { workspace = true, features = ["json", "registry"] }
@@ -54,6 +54,7 @@ walkdir.workspace = true
 pq_proto.workspace = true
 postgres_connection.workspace = true
 metrics.workspace = true
+workspace_hack.workspace = true

 const_format.workspace = true

@@ -70,7 +71,6 @@ criterion.workspace = true
 hex-literal.workspace = true
 camino-tempfile.workspace = true
 serde_assert.workspace = true
-tokio = { workspace = true, features = ["test-util"] }

 [[bench]]
 name = "benchmarks"
--- a/libs/walproposer/Cargo.toml
+++ b/libs/walproposer/Cargo.toml
@@ -9,6 +9,8 @@ anyhow.workspace = true
 utils.workspace = true
 postgres_ffi.workspace = true

+workspace_hack.workspace = true
+
 [build-dependencies]
 anyhow.workspace = true
 bindgen.workspace = true
--- a/libs/walproposer/build.rs
+++ b/libs/walproposer/build.rs
@@ -95,7 +95,6 @@ fn main() -> anyhow::Result<()> {
        .allowlist_var("ERROR")
        .allowlist_var("FATAL")
        .allowlist_var("PANIC")
-        .allowlist_var("PG_VERSION_NUM")
        .allowlist_var("WPEVENT")
        .allowlist_var("WL_LATCH_SET")
        .allowlist_var("WL_SOCKET_READABLE")
--- a/libs/walproposer/src/walproposer.rs
+++ b/libs/walproposer/src/walproposer.rs
@@ -282,11 +282,7 @@ mod tests {
    use std::cell::UnsafeCell;
    use utils::id::TenantTimelineId;

-    use crate::{
-        api_bindings::Level,
-        bindings::{NeonWALReadResult, PG_VERSION_NUM},
-        walproposer::Wrapper,
-    };
+    use crate::{api_bindings::Level, bindings::NeonWALReadResult, walproposer::Wrapper};

    use super::ApiImpl;

@@ -493,79 +489,41 @@ mod tests {

        let (sender, receiver) = sync_channel(1);

-        // Messages definitions are at walproposer.h
-        // xxx: it would be better to extract them from safekeeper crate and
-        // use serialization/deserialization here.
-        let greeting_tag = (b'g' as u64).to_ne_bytes();
-        let proto_version = 2_u32.to_ne_bytes();
-        let pg_version: [u8; 4] = PG_VERSION_NUM.to_ne_bytes();
-        let proposer_id = [0; 16];
-        let system_id = 0_u64.to_ne_bytes();
-        let tenant_id = ttid.tenant_id.as_arr();
-        let timeline_id = ttid.timeline_id.as_arr();
-        let pg_tli = 1_u32.to_ne_bytes();
-        let wal_seg_size = 16777216_u32.to_ne_bytes();
-        let proposer_greeting = [
-            greeting_tag.as_slice(),
-            proto_version.as_slice(),
-            pg_version.as_slice(),
-            proposer_id.as_slice(),
-            system_id.as_slice(),
-            tenant_id.as_slice(),
-            timeline_id.as_slice(),
-            pg_tli.as_slice(),
-            wal_seg_size.as_slice(),
-        ]
-        .concat();
-
-        let voting_tag = (b'v' as u64).to_ne_bytes();
-        let vote_request_term = 3_u64.to_ne_bytes();
-        let proposer_id = [0; 16];
-        let vote_request = [
-            voting_tag.as_slice(),
-            vote_request_term.as_slice(),
-            proposer_id.as_slice(),
-        ]
-        .concat();
-
-        let acceptor_greeting_term = 2_u64.to_ne_bytes();
-        let acceptor_greeting_node_id = 1_u64.to_ne_bytes();
-        let acceptor_greeting = [
-            greeting_tag.as_slice(),
-            acceptor_greeting_term.as_slice(),
-            acceptor_greeting_node_id.as_slice(),
-        ]
-        .concat();
-
-        let vote_response_term = 3_u64.to_ne_bytes();
-        let vote_given = 1_u64.to_ne_bytes();
-        let flush_lsn = 0x539_u64.to_ne_bytes();
-        let truncate_lsn = 0x539_u64.to_ne_bytes();
-        let th_len = 1_u32.to_ne_bytes();
-        let th_term = 2_u64.to_ne_bytes();
-        let th_lsn = 0x539_u64.to_ne_bytes();
-        let timeline_start_lsn = 0x539_u64.to_ne_bytes();
-        let vote_response = [
-            voting_tag.as_slice(),
-            vote_response_term.as_slice(),
-            vote_given.as_slice(),
-            flush_lsn.as_slice(),
-            truncate_lsn.as_slice(),
-            th_len.as_slice(),
-            th_term.as_slice(),
-            th_lsn.as_slice(),
-            timeline_start_lsn.as_slice(),
-        ]
-        .concat();
-
        let my_impl: Box<dyn ApiImpl> = Box::new(MockImpl {
            wait_events: Cell::new(WaitEventsData {
                sk: std::ptr::null_mut(),
                event_mask: 0,
            }),
-            expected_messages: vec![proposer_greeting, vote_request],
+            expected_messages: vec![
+                // TODO: When updating Postgres versions, this test will cause
+                // problems. Postgres version in message needs updating.
+                //
+                // Greeting(ProposerGreeting { protocol_version: 2, pg_version: 160003, proposer_id: [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], system_id: 0, timeline_id: 9e4c8f36063c6c6e93bc20d65a820f3d, tenant_id: 9e4c8f36063c6c6e93bc20d65a820f3d, tli: 1, wal_seg_size: 16777216 })
+                vec![
+                    103, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 3, 113, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+                    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 158, 76, 143, 54, 6, 60, 108, 110,
+                    147, 188, 32, 214, 90, 130, 15, 61, 158, 76, 143, 54, 6, 60, 108, 110, 147,
+                    188, 32, 214, 90, 130, 15, 61, 1, 0, 0, 0, 0, 0, 0, 1,
+                ],
+                // VoteRequest(VoteRequest { term: 3 })
+                vec![
+                    118, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+                    0, 0, 0, 0, 0, 0,
+                ],
+            ],
            expected_ptr: AtomicUsize::new(0),
-            safekeeper_replies: vec![acceptor_greeting, vote_response],
+            safekeeper_replies: vec![
+                // Greeting(AcceptorGreeting { term: 2, node_id: NodeId(1) })
+                vec![
+                    103, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,
+                ],
+                // VoteResponse(VoteResponse { term: 3, vote_given: 1, flush_lsn: 0/539, truncate_lsn: 0/539, term_history: [(2, 0/539)], timeline_start_lsn: 0/539 })
+                vec![
+                    118, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 57,
+                    5, 0, 0, 0, 0, 0, 0, 57, 5, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0,
+                    0, 57, 5, 0, 0, 0, 0, 0, 0, 57, 5, 0, 0, 0, 0, 0, 0,
+                ],
+            ],
            replies_ptr: AtomicUsize::new(0),
            sync_channel: sender,
            shmem: UnsafeCell::new(crate::api_bindings::empty_shmem()),
--- a/pageserver/benches/bench_ingest.rs
+++ b/pageserver/benches/bench_ingest.rs
@@ -4,13 +4,12 @@ use bytes::Bytes;
 use camino::Utf8PathBuf;
 use criterion::{criterion_group, criterion_main, Criterion};
 use pageserver::{
-    config::{defaults::DEFAULT_IO_BUFFER_ALIGNMENT, PageServerConf},
+    config::PageServerConf,
    context::{DownloadBehavior, RequestContext},
    l0_flush::{L0FlushConfig, L0FlushGlobalState},
    page_cache,
    repository::Value,
    task_mgr::TaskKind,
-    tenant::storage_layer::inmemory_layer::SerializedBatch,
    tenant::storage_layer::InMemoryLayer,
    virtual_file,
 };
@@ -68,16 +67,12 @@ async fn ingest(
    let layer =
        InMemoryLayer::create(conf, timeline_id, tenant_shard_id, lsn, entered, &ctx).await?;

-    let data = Value::Image(Bytes::from(vec![0u8; put_size]));
-    let data_ser_size = data.serialized_size().unwrap() as usize;
+    let data = Value::Image(Bytes::from(vec![0u8; put_size])).ser()?;
    let ctx = RequestContext::new(
        pageserver::task_mgr::TaskKind::WalReceiverConnectionHandler,
        pageserver::context::DownloadBehavior::Download,
    );

-    const BATCH_SIZE: usize = 16;
-    let mut batch = Vec::new();
-
    for i in 0..put_count {
        lsn += put_size as u64;

@@ -100,17 +95,7 @@ async fn ingest(
            }
        }

-        batch.push((key.to_compact(), lsn, data_ser_size, data.clone()));
-        if batch.len() >= BATCH_SIZE {
-            let this_batch = std::mem::take(&mut batch);
-            let serialized = SerializedBatch::from_values(this_batch);
-            layer.put_batch(serialized, &ctx).await?;
-        }
-    }
-    if !batch.is_empty() {
-        let this_batch = std::mem::take(&mut batch);
-        let serialized = SerializedBatch::from_values(this_batch);
-        layer.put_batch(serialized, &ctx).await?;
+        layer.put_value(key.to_compact(), lsn, &data, &ctx).await?;
    }
    layer.freeze(lsn + 1).await;

@@ -164,11 +149,7 @@ fn criterion_benchmark(c: &mut Criterion) {
    let conf: &'static PageServerConf = Box::leak(Box::new(
        pageserver::config::PageServerConf::dummy_conf(temp_dir.path().to_path_buf()),
    ));
-    virtual_file::init(
-        16384,
-        virtual_file::io_engine_for_bench(),
-        DEFAULT_IO_BUFFER_ALIGNMENT,
-    );
+    virtual_file::init(16384, virtual_file::io_engine_for_bench());
    page_cache::init(conf.page_cache_size);

    {
--- a/pageserver/client/src/mgmt_api.rs
+++ b/pageserver/client/src/mgmt_api.rs
@@ -506,16 +506,6 @@ impl Client {
            .map_err(Error::ReceiveBody)
    }

-    /// Configs io buffer alignment at runtime.
-    pub async fn put_io_alignment(&self, align: usize) -> Result<()> {
-        let uri = format!("{}/v1/io_alignment", self.mgmt_api_endpoint);
-        self.request(Method::PUT, uri, align)
-            .await?
-            .json()
-            .await
-            .map_err(Error::ReceiveBody)
-    }
-
    pub async fn get_utilization(&self) -> Result<PageserverUtilization> {
        let uri = format!("{}/v1/utilization", self.mgmt_api_endpoint);
        self.get(uri)
--- a/pageserver/ctl/src/layer_map_analyzer.rs
+++ b/pageserver/ctl/src/layer_map_analyzer.rs
@@ -4,7 +4,6 @@

 use anyhow::Result;
 use camino::{Utf8Path, Utf8PathBuf};
-use pageserver::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
 use pageserver::context::{DownloadBehavior, RequestContext};
 use pageserver::task_mgr::TaskKind;
 use pageserver::tenant::{TENANTS_SEGMENT_NAME, TIMELINES_SEGMENT_NAME};
@@ -145,11 +144,7 @@ pub(crate) async fn main(cmd: &AnalyzeLayerMapCmd) -> Result<()> {
    let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);

    // Initialize virtual_file (file desriptor cache) and page cache which are needed to access layer persistent B-Tree.
-    pageserver::virtual_file::init(
-        10,
-        virtual_file::api::IoEngineKind::StdFs,
-        DEFAULT_IO_BUFFER_ALIGNMENT,
-    );
+    pageserver::virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs);
    pageserver::page_cache::init(100);

    let mut total_delta_layers = 0usize;
--- a/pageserver/ctl/src/layers.rs
+++ b/pageserver/ctl/src/layers.rs
@@ -3,7 +3,6 @@ use std::path::{Path, PathBuf};
 use anyhow::Result;
 use camino::{Utf8Path, Utf8PathBuf};
 use clap::Subcommand;
-use pageserver::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
 use pageserver::context::{DownloadBehavior, RequestContext};
 use pageserver::task_mgr::TaskKind;
 use pageserver::tenant::block_io::BlockCursor;
@@ -60,7 +59,7 @@ pub(crate) enum LayerCmd {

 async fn read_delta_file(path: impl AsRef<Path>, ctx: &RequestContext) -> Result<()> {
    let path = Utf8Path::from_path(path.as_ref()).expect("non-Unicode path");
-    virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs, 1);
+    virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs);
    page_cache::init(100);
    let file = VirtualFile::open(path, ctx).await?;
    let file_id = page_cache::next_file_id();
@@ -190,11 +189,7 @@ pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
            new_tenant_id,
            new_timeline_id,
        } => {
-            pageserver::virtual_file::init(
-                10,
-                virtual_file::api::IoEngineKind::StdFs,
-                DEFAULT_IO_BUFFER_ALIGNMENT,
-            );
+            pageserver::virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs);
            pageserver::page_cache::init(100);

            let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);
--- a/pageserver/ctl/src/main.rs
+++ b/pageserver/ctl/src/main.rs
@@ -20,7 +20,6 @@ use clap::{Parser, Subcommand};
 use index_part::IndexPartCmd;
 use layers::LayerCmd;
 use pageserver::{
-    config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
    context::{DownloadBehavior, RequestContext},
    page_cache,
    task_mgr::TaskKind,
@@ -206,11 +205,7 @@ fn read_pg_control_file(control_file_path: &Utf8Path) -> anyhow::Result<()> {

 async fn print_layerfile(path: &Utf8Path) -> anyhow::Result<()> {
    // Basic initialization of things that don't change after startup
-    virtual_file::init(
-        10,
-        virtual_file::api::IoEngineKind::StdFs,
-        DEFAULT_IO_BUFFER_ALIGNMENT,
-    );
+    virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs);
    page_cache::init(100);
    let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);
    dump_layerfile_from_path(path, true, &ctx).await
--- a/pageserver/pagebench/src/cmd/getpage_latest_lsn.rs
+++ b/pageserver/pagebench/src/cmd/getpage_latest_lsn.rs
@@ -58,11 +58,6 @@ pub(crate) struct Args {
    /// [`pageserver_api::models::virtual_file::IoEngineKind`].
    #[clap(long)]
    set_io_engine: Option<pageserver_api::models::virtual_file::IoEngineKind>,
-
-    /// Before starting the benchmark, live-reconfigure the pageserver to use specified alignment for io buffers.
-    #[clap(long)]
-    set_io_alignment: Option<usize>,
-
    targets: Option<Vec<TenantTimelineId>>,
 }

@@ -129,10 +124,6 @@ async fn main_impl(
        mgmt_api_client.put_io_engine(engine_str).await?;
    }

-    if let Some(align) = args.set_io_alignment {
-        mgmt_api_client.put_io_alignment(align).await?;
-    }
-
    // discover targets
    let timelines: Vec<TenantTimelineId> = crate::util::cli::targets::discover(
        &mgmt_api_client,
--- a/pageserver/src/bin/pageserver.rs
+++ b/pageserver/src/bin/pageserver.rs
@@ -125,7 +125,6 @@ fn main() -> anyhow::Result<()> {
    info!(?conf.virtual_file_io_engine, "starting with virtual_file IO engine");
    info!(?conf.virtual_file_direct_io, "starting with virtual_file Direct IO settings");
    info!(?conf.compact_level0_phase1_value_access, "starting with setting for compact_level0_phase1_value_access");
-    info!(?conf.io_buffer_alignment, "starting with setting for IO buffer alignment");

    let tenants_path = conf.tenants_path();
    if !tenants_path.exists() {
@@ -137,11 +136,7 @@ fn main() -> anyhow::Result<()> {
    let scenario = failpoint_support::init();

    // Basic initialization of things that don't change after startup
-    virtual_file::init(
-        conf.max_file_descriptors,
-        conf.virtual_file_io_engine,
-        conf.io_buffer_alignment,
-    );
+    virtual_file::init(conf.max_file_descriptors, conf.virtual_file_io_engine);
    page_cache::init(conf.page_cache_size);

    start_pageserver(launch_ts, conf).context("Failed to start pageserver")?;
--- a/pageserver/src/config.rs
+++ b/pageserver/src/config.rs
@@ -50,6 +50,7 @@ pub mod defaults {
        DEFAULT_HTTP_LISTEN_ADDR, DEFAULT_HTTP_LISTEN_PORT, DEFAULT_PG_LISTEN_ADDR,
        DEFAULT_PG_LISTEN_PORT,
    };
+    use pageserver_api::models::ImageCompressionAlgorithm;
    pub use storage_broker::DEFAULT_ENDPOINT as BROKER_DEFAULT_ENDPOINT;

    pub const DEFAULT_WAIT_LSN_TIMEOUT: &str = "300 s";
@@ -89,14 +90,13 @@ pub mod defaults {

    pub const DEFAULT_MAX_VECTORED_READ_BYTES: usize = 128 * 1024; // 128 KiB

-    pub const DEFAULT_IMAGE_COMPRESSION: &str = "zstd(1)";
+    pub const DEFAULT_IMAGE_COMPRESSION: ImageCompressionAlgorithm =
+        ImageCompressionAlgorithm::Disabled;

    pub const DEFAULT_VALIDATE_VECTORED_GET: bool = false;

    pub const DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB: usize = 0;

-    pub const DEFAULT_IO_BUFFER_ALIGNMENT: usize = 0;
-
    ///
    /// Default built-in configuration file.
    ///
@@ -291,8 +291,6 @@ pub struct PageServerConf {

    /// Direct IO settings
    pub virtual_file_direct_io: virtual_file::DirectIoMode,
-
-    pub io_buffer_alignment: usize,
 }

 /// We do not want to store this in a PageServerConf because the latter may be logged
@@ -397,8 +395,6 @@ struct PageServerConfigBuilder {
    compact_level0_phase1_value_access: BuilderValue<CompactL0Phase1ValueAccess>,

    virtual_file_direct_io: BuilderValue<virtual_file::DirectIoMode>,
-
-    io_buffer_alignment: BuilderValue<usize>,
 }

 impl PageServerConfigBuilder {
@@ -482,12 +478,11 @@ impl PageServerConfigBuilder {
            max_vectored_read_bytes: Set(MaxVectoredReadBytes(
                NonZeroUsize::new(DEFAULT_MAX_VECTORED_READ_BYTES).unwrap(),
            )),
-            image_compression: Set(DEFAULT_IMAGE_COMPRESSION.parse().unwrap()),
+            image_compression: Set(DEFAULT_IMAGE_COMPRESSION),
            ephemeral_bytes_per_memory_kb: Set(DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB),
            l0_flush: Set(L0FlushConfig::default()),
            compact_level0_phase1_value_access: Set(CompactL0Phase1ValueAccess::default()),
            virtual_file_direct_io: Set(virtual_file::DirectIoMode::default()),
-            io_buffer_alignment: Set(DEFAULT_IO_BUFFER_ALIGNMENT),
        }
    }
 }
@@ -667,10 +662,6 @@ impl PageServerConfigBuilder {
        self.virtual_file_direct_io = BuilderValue::Set(value);
    }

-    pub fn io_buffer_alignment(&mut self, value: usize) {
-        self.io_buffer_alignment = BuilderValue::Set(value);
-    }
-
    pub fn build(self, id: NodeId) -> anyhow::Result<PageServerConf> {
        let default = Self::default_values();

@@ -727,7 +718,6 @@ impl PageServerConfigBuilder {
                l0_flush,
                compact_level0_phase1_value_access,
                virtual_file_direct_io,
-                io_buffer_alignment,
            }
            CUSTOM LOGIC
            {
@@ -997,9 +987,6 @@ impl PageServerConf {
                "virtual_file_direct_io" => {
                    builder.virtual_file_direct_io(utils::toml_edit_ext::deserialize_item(item).context("virtual_file_direct_io")?)
                }
-                "io_buffer_alignment" => {
-                    builder.io_buffer_alignment(parse_toml_u64("io_buffer_alignment", item)? as usize)
-                }
                _ => bail!("unrecognized pageserver option '{key}'"),
            }
        }
@@ -1078,12 +1065,11 @@ impl PageServerConf {
                NonZeroUsize::new(defaults::DEFAULT_MAX_VECTORED_READ_BYTES)
                    .expect("Invalid default constant"),
            ),
-            image_compression: defaults::DEFAULT_IMAGE_COMPRESSION.parse().unwrap(),
+            image_compression: defaults::DEFAULT_IMAGE_COMPRESSION,
            ephemeral_bytes_per_memory_kb: defaults::DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB,
            l0_flush: L0FlushConfig::default(),
            compact_level0_phase1_value_access: CompactL0Phase1ValueAccess::default(),
            virtual_file_direct_io: virtual_file::DirectIoMode::default(),
-            io_buffer_alignment: defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
        }
    }
 }
@@ -1319,12 +1305,11 @@ background_task_maximum_delay = '334 s'
                    NonZeroUsize::new(defaults::DEFAULT_MAX_VECTORED_READ_BYTES)
                        .expect("Invalid default constant")
                ),
-                image_compression: defaults::DEFAULT_IMAGE_COMPRESSION.parse().unwrap(),
+                image_compression: defaults::DEFAULT_IMAGE_COMPRESSION,
                ephemeral_bytes_per_memory_kb: defaults::DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB,
                l0_flush: L0FlushConfig::default(),
                compact_level0_phase1_value_access: CompactL0Phase1ValueAccess::default(),
                virtual_file_direct_io: virtual_file::DirectIoMode::default(),
-                io_buffer_alignment: defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
            },
            "Correct defaults should be used when no config values are provided"
        );
@@ -1393,12 +1378,11 @@ background_task_maximum_delay = '334 s'
                    NonZeroUsize::new(defaults::DEFAULT_MAX_VECTORED_READ_BYTES)
                        .expect("Invalid default constant")
                ),
-                image_compression: defaults::DEFAULT_IMAGE_COMPRESSION.parse().unwrap(),
+                image_compression: defaults::DEFAULT_IMAGE_COMPRESSION,
                ephemeral_bytes_per_memory_kb: defaults::DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB,
                l0_flush: L0FlushConfig::default(),
                compact_level0_phase1_value_access: CompactL0Phase1ValueAccess::default(),
                virtual_file_direct_io: virtual_file::DirectIoMode::default(),
-                io_buffer_alignment: defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
            },
            "Should be able to parse all basic config values correctly"
        );
--- a/pageserver/src/disk_usage_eviction_task.rs
+++ b/pageserver/src/disk_usage_eviction_task.rs
@@ -64,7 +64,7 @@ use crate::{
        mgr::TenantManager,
        remote_timeline_client::LayerFileMetadata,
        secondary::SecondaryTenant,
-        storage_layer::{AsLayerDesc, EvictionError, Layer, LayerName, LayerVisibilityHint},
+        storage_layer::{AsLayerDesc, EvictionError, Layer, LayerName},
    },
    CancellableTask, DiskUsageEvictionTask,
 };
@@ -114,7 +114,7 @@ fn default_highest_layer_count_loses_first() -> bool {
 }

 impl EvictionOrder {
-    fn sort(&self, candidates: &mut [(EvictionPartition, EvictionCandidate)]) {
+    fn sort(&self, candidates: &mut [(MinResidentSizePartition, EvictionCandidate)]) {
        use EvictionOrder::*;

        match self {
@@ -644,7 +644,6 @@ pub(crate) struct EvictionCandidate {
    pub(crate) layer: EvictionLayer,
    pub(crate) last_activity_ts: SystemTime,
    pub(crate) relative_last_activity: finite_f32::FiniteF32,
-    pub(crate) visibility: LayerVisibilityHint,
 }

 impl std::fmt::Display for EvictionLayer {
@@ -686,22 +685,14 @@ impl std::fmt::Debug for EvictionCandidate {
 }

 #[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
-enum EvictionPartition {
-    // A layer that is un-wanted by the tenant: evict all these first, before considering
-    // any other layers
-    EvictNow,
-
-    // Above the minimum size threshold: this layer is a candidate for eviction.
+enum MinResidentSizePartition {
    Above,
-
-    // Below the minimum size threshold: this layer should only be evicted if all the
-    // tenants' layers above the minimum size threshold have already been considered.
    Below,
 }

 enum EvictionCandidates {
    Cancelled,
-    Finished(Vec<(EvictionPartition, EvictionCandidate)>),
+    Finished(Vec<(MinResidentSizePartition, EvictionCandidate)>),
 }

 /// Gather the eviction candidates.
@@ -899,10 +890,8 @@ async fn collect_eviction_candidates(
            max_layer_size
        };

-        // Sort layers most-recently-used first, then calculate [`EvictionPartition`] for each layer,
-        // where the inputs are:
-        //  - whether the layer is visible
-        //  - whether the layer is above/below the min_resident_size cutline
+        // Sort layers most-recently-used first, then partition by
+        // cumsum above/below min_resident_size.
        tenant_candidates
            .sort_unstable_by_key(|layer_info| std::cmp::Reverse(layer_info.last_activity_ts));
        let mut cumsum: i128 = 0;
@@ -919,23 +908,12 @@ async fn collect_eviction_candidates(
                    candidate.relative_last_activity =
                        eviction_order.relative_last_activity(total, i);

-                    let partition = match candidate.visibility {
-                        LayerVisibilityHint::Covered => {
-                            // Covered layers are evicted first
-                            EvictionPartition::EvictNow
-                        }
-                        LayerVisibilityHint::Visible => {
-                            cumsum += i128::from(candidate.layer.get_file_size());
-
-                            if cumsum > min_resident_size as i128 {
-                                EvictionPartition::Above
-                            } else {
-                                // The most recent layers below the min_resident_size threshold
-                                // are the last to be evicted.
-                                EvictionPartition::Below
-                            }
-                        }
+                    let partition = if cumsum > min_resident_size as i128 {
+                        MinResidentSizePartition::Above
+                    } else {
+                        MinResidentSizePartition::Below
                    };
+                    cumsum += i128::from(candidate.layer.get_file_size());

                    (partition, candidate)
                });
@@ -1003,7 +981,7 @@ async fn collect_eviction_candidates(
                        // Secondary locations' layers are always considered above the min resident size,
                        // i.e. secondary locations are permitted to be trimmed to zero layers if all
                        // the layers have sufficiently old access times.
-                        EvictionPartition::Above,
+                        MinResidentSizePartition::Above,
                        candidate,
                    )
                });
@@ -1031,9 +1009,7 @@ async fn collect_eviction_candidates(
        }
    }

-    debug_assert!(EvictionPartition::Above < EvictionPartition::Below,
-        "as explained in the function's doc comment, layers that aren't in the tenant's min_resident_size are evicted first");
-    debug_assert!(EvictionPartition::EvictNow < EvictionPartition::Above,
+    debug_assert!(MinResidentSizePartition::Above < MinResidentSizePartition::Below,
        "as explained in the function's doc comment, layers that aren't in the tenant's min_resident_size are evicted first");

    eviction_order.sort(&mut candidates);
@@ -1046,7 +1022,7 @@ async fn collect_eviction_candidates(
 ///
 /// Returns the amount of candidates selected, with the planned usage.
 fn select_victims<U: Usage>(
-    candidates: &[(EvictionPartition, EvictionCandidate)],
+    candidates: &[(MinResidentSizePartition, EvictionCandidate)],
    usage_pre: U,
 ) -> VictimSelection<U> {
    let mut usage_when_switched = None;
@@ -1058,7 +1034,7 @@ fn select_victims<U: Usage>(
            break;
        }

-        if partition == &EvictionPartition::Below && usage_when_switched.is_none() {
+        if partition == &MinResidentSizePartition::Below && usage_when_switched.is_none() {
            usage_when_switched = Some((usage_planned, i));
        }

--- a/pageserver/src/http/routes.rs
+++ b/pageserver/src/http/routes.rs
@@ -178,8 +178,10 @@ fn check_permission(request: &Request<Body>, tenant_id: Option<TenantId>) -> Res
 impl From<PageReconstructError> for ApiError {
    fn from(pre: PageReconstructError) -> ApiError {
        match pre {
-            PageReconstructError::Other(other) => ApiError::InternalServerError(other),
-            PageReconstructError::MissingKey(e) => ApiError::InternalServerError(e.into()),
+            PageReconstructError::Other(pre) => ApiError::InternalServerError(pre),
+            PageReconstructError::MissingKey(e) => {
+                ApiError::InternalServerError(anyhow::anyhow!("{e}"))
+            }
            PageReconstructError::Cancelled => ApiError::Cancelled,
            PageReconstructError::AncestorLsnTimeout(e) => ApiError::Timeout(format!("{e}").into()),
            PageReconstructError::WalRedo(pre) => ApiError::InternalServerError(pre),
@@ -1706,6 +1708,11 @@ async fn timeline_compact_handler(
        flags |= CompactFlags::ForceImageLayerCreation;
    }
    if Some(true) == parse_query_param::<_, bool>(&request, "enhanced_gc_bottom_most_compaction")? {
+        if !cfg!(feature = "testing") {
+            return Err(ApiError::InternalServerError(anyhow!(
+                "enhanced_gc_bottom_most_compaction is only available in testing mode"
+            )));
+        }
        flags |= CompactFlags::EnhancedGcBottomMostCompaction;
    }
    let wait_until_uploaded =
@@ -2325,20 +2332,6 @@ async fn put_io_engine_handler(
    json_response(StatusCode::OK, ())
 }

-async fn put_io_alignment_handler(
-    mut r: Request<Body>,
-    _cancel: CancellationToken,
-) -> Result<Response<Body>, ApiError> {
-    check_permission(&r, None)?;
-    let align: usize = json_request(&mut r).await?;
-    crate::virtual_file::set_io_buffer_alignment(align).map_err(|align| {
-        ApiError::PreconditionFailed(
-            format!("Requested io alignment ({align}) is not a power of two").into(),
-        )
-    })?;
-    json_response(StatusCode::OK, ())
-}
-
 /// Polled by control plane.
 ///
 /// See [`crate::utilization`].
@@ -2951,7 +2944,7 @@ pub fn make_router(
        )
        .put(
            "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/compact",
-            |r| api_handler(r, timeline_compact_handler),
+            |r| testing_api_handler("run timeline compaction", r, timeline_compact_handler),
        )
        .put(
            "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/checkpoint",
@@ -3026,9 +3019,6 @@ pub fn make_router(
            |r| api_handler(r, timeline_collect_keyspace),
        )
        .put("/v1/io_engine", |r| api_handler(r, put_io_engine_handler))
-        .put("/v1/io_alignment", |r| {
-            api_handler(r, put_io_alignment_handler)
-        })
        .put(
            "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/force_aux_policy_switch",
            |r| api_handler(r, force_aux_policy_switch_handler),
--- a/pageserver/src/l0_flush.rs
+++ b/pageserver/src/l0_flush.rs
@@ -1,10 +1,15 @@
 use std::{num::NonZeroUsize, sync::Arc};

+use crate::tenant::ephemeral_file;
+
 #[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize)]
 #[serde(tag = "mode", rename_all = "kebab-case", deny_unknown_fields)]
 pub enum L0FlushConfig {
+    PageCached,
    #[serde(rename_all = "snake_case")]
-    Direct { max_concurrency: NonZeroUsize },
+    Direct {
+        max_concurrency: NonZeroUsize,
+    },
 }

 impl Default for L0FlushConfig {
@@ -20,12 +25,14 @@ impl Default for L0FlushConfig {
 pub struct L0FlushGlobalState(Arc<Inner>);

 pub enum Inner {
+    PageCached,
    Direct { semaphore: tokio::sync::Semaphore },
 }

 impl L0FlushGlobalState {
    pub fn new(config: L0FlushConfig) -> Self {
        match config {
+            L0FlushConfig::PageCached => Self(Arc::new(Inner::PageCached)),
            L0FlushConfig::Direct { max_concurrency } => {
                let semaphore = tokio::sync::Semaphore::new(max_concurrency.get());
                Self(Arc::new(Inner::Direct { semaphore }))
@@ -37,3 +44,13 @@ impl L0FlushGlobalState {
        &self.0
    }
 }
+
+impl L0FlushConfig {
+    pub(crate) fn prewarm_on_write(&self) -> ephemeral_file::PrewarmPageCacheOnWrite {
+        use L0FlushConfig::*;
+        match self {
+            PageCached => ephemeral_file::PrewarmPageCacheOnWrite::Yes,
+            Direct { .. } => ephemeral_file::PrewarmPageCacheOnWrite::No,
+        }
+    }
+}
--- a/pageserver/src/lib.rs
+++ b/pageserver/src/lib.rs
@@ -49,7 +49,7 @@ use tracing::{info, info_span};
 /// backwards-compatible changes to the metadata format.
 pub const STORAGE_FORMAT_VERSION: u16 = 3;

-pub const DEFAULT_PG_VERSION: u32 = 16;
+pub const DEFAULT_PG_VERSION: u32 = 15;

 // Magic constants used to identify different kinds of files
 pub const IMAGE_FILE_MAGIC: u16 = 0x5A60;
@@ -88,8 +88,6 @@ pub async fn shutdown_pageserver(
 ) {
    use std::time::Duration;

-    let started_at = std::time::Instant::now();
-
    // If the orderly shutdown below takes too long, we still want to make
    // sure that all walredo processes are killed and wait()ed on by us, not systemd.
    //
@@ -243,10 +241,7 @@ pub async fn shutdown_pageserver(
    walredo_extraordinary_shutdown_thread.join().unwrap();
    info!("walredo_extraordinary_shutdown_thread done");

-    info!(
-        elapsed_ms = started_at.elapsed().as_millis(),
-        "Shut down successfully completed"
-    );
+    info!("Shut down successfully completed");
    std::process::exit(exit_code);
 }

--- a/pageserver/src/metrics.rs
+++ b/pageserver/src/metrics.rs
@@ -1803,23 +1803,6 @@ pub(crate) static SECONDARY_RESIDENT_PHYSICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::n
    .expect("failed to define a metric")
 });

-pub(crate) static NODE_UTILIZATION_SCORE: Lazy<UIntGauge> = Lazy::new(|| {
-    register_uint_gauge!(
-        "pageserver_utilization_score",
-        "The utilization score we report to the storage controller for scheduling, where 0 is empty, 1000000 is full, and anything above is considered overloaded",
-    )
-    .expect("failed to define a metric")
-});
-
-pub(crate) static SECONDARY_HEATMAP_TOTAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
-    register_uint_gauge_vec!(
-        "pageserver_secondary_heatmap_total_size",
-        "The total size in bytes of all layers in the most recently downloaded heatmap.",
-        &["tenant_id", "shard_id"]
-    )
-    .expect("failed to define a metric")
-});
-
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
 pub enum RemoteOpKind {
    Upload,
@@ -1870,64 +1853,16 @@ pub(crate) static TENANT_TASK_EVENTS: Lazy<IntCounterVec> = Lazy::new(|| {
    .expect("Failed to register tenant_task_events metric")
 });

-pub struct BackgroundLoopSemaphoreMetrics {
-    counters: EnumMap<BackgroundLoopKind, IntCounterPair>,
-    durations: EnumMap<BackgroundLoopKind, Counter>,
-}
-
-pub(crate) static BACKGROUND_LOOP_SEMAPHORE: Lazy<BackgroundLoopSemaphoreMetrics> = Lazy::new(
-    || {
-        let counters = register_int_counter_pair_vec!(
-            "pageserver_background_loop_semaphore_wait_start_count",
-            "Counter for background loop concurrency-limiting semaphore acquire calls started",
-            "pageserver_background_loop_semaphore_wait_finish_count",
-            "Counter for background loop concurrency-limiting semaphore acquire calls finished",
-            &["task"],
-        )
-        .unwrap();
-
-        let durations = register_counter_vec!(
-            "pageserver_background_loop_semaphore_wait_duration_seconds",
-            "Sum of wall clock time spent waiting on the background loop concurrency-limiting semaphore acquire calls",
-            &["task"],
-        )
-        .unwrap();
-
-        BackgroundLoopSemaphoreMetrics {
-            counters: enum_map::EnumMap::from_array(std::array::from_fn(|i| {
-                let kind = <BackgroundLoopKind as enum_map::Enum>::from_usize(i);
-                counters.with_label_values(&[kind.into()])
-            })),
-            durations: enum_map::EnumMap::from_array(std::array::from_fn(|i| {
-                let kind = <BackgroundLoopKind as enum_map::Enum>::from_usize(i);
-                durations.with_label_values(&[kind.into()])
-            })),
-        }
-    },
-);
-
-impl BackgroundLoopSemaphoreMetrics {
-    pub(crate) fn measure_acquisition(&self, task: BackgroundLoopKind) -> impl Drop + '_ {
-        struct Record<'a> {
-            metrics: &'a BackgroundLoopSemaphoreMetrics,
-            task: BackgroundLoopKind,
-            _counter_guard: metrics::IntCounterPairGuard,
-            start: Instant,
-        }
-        impl Drop for Record<'_> {
-            fn drop(&mut self) {
-                let elapsed = self.start.elapsed().as_secs_f64();
-                self.metrics.durations[self.task].inc_by(elapsed);
-            }
-        }
-        Record {
-            metrics: self,
-            task,
-            _counter_guard: self.counters[task].guard(),
-            start: Instant::now(),
-        }
-    }
-}
+pub(crate) static BACKGROUND_LOOP_SEMAPHORE_WAIT_GAUGE: Lazy<IntCounterPairVec> = Lazy::new(|| {
+    register_int_counter_pair_vec!(
+        "pageserver_background_loop_semaphore_wait_start_count",
+        "Counter for background loop concurrency-limiting semaphore acquire calls started",
+        "pageserver_background_loop_semaphore_wait_finish_count",
+        "Counter for background loop concurrency-limiting semaphore acquire calls finished",
+        &["task"],
+    )
+    .unwrap()
+});

 pub(crate) static BACKGROUND_LOOP_PERIOD_OVERRUN_COUNT: Lazy<IntCounterVec> = Lazy::new(|| {
    register_int_counter_vec!(
@@ -2609,7 +2544,6 @@ use std::time::{Duration, Instant};
 use crate::context::{PageContentKind, RequestContext};
 use crate::task_mgr::TaskKind;
 use crate::tenant::mgr::TenantSlot;
-use crate::tenant::tasks::BackgroundLoopKind;

 /// Maintain a per timeline gauge in addition to the global gauge.
 pub(crate) struct PerTimelineRemotePhysicalSizeGauge {
--- a/pageserver/src/pgdatadir_mapping.rs
+++ b/pageserver/src/pgdatadir_mapping.rs
@@ -15,11 +15,12 @@ use crate::{aux_file, repository::*};
 use anyhow::{ensure, Context};
 use bytes::{Buf, Bytes, BytesMut};
 use enum_map::Enum;
+use itertools::Itertools;
 use pageserver_api::key::{
    dbdir_key_range, rel_block_to_key, rel_dir_to_key, rel_key_range, rel_size_to_key,
    relmap_file_key, repl_origin_key, repl_origin_key_range, slru_block_to_key, slru_dir_to_key,
    slru_segment_key_range, slru_segment_size_to_key, twophase_file_key, twophase_key_range,
-    CompactKey, AUX_FILES_KEY, CHECKPOINT_KEY, CONTROLFILE_KEY, DBDIR_KEY, TWOPHASEDIR_KEY,
+    AUX_FILES_KEY, CHECKPOINT_KEY, CONTROLFILE_KEY, DBDIR_KEY, TWOPHASEDIR_KEY,
 };
 use pageserver_api::keyspace::SparseKeySpace;
 use pageserver_api::models::AuxFilePolicy;
@@ -36,6 +37,7 @@ use tokio_util::sync::CancellationToken;
 use tracing::{debug, info, trace, warn};
 use utils::bin_ser::DeserializeError;
 use utils::pausable_failpoint;
+use utils::vec_map::{VecMap, VecMapOrdering};
 use utils::{bin_ser::BeSer, lsn::Lsn};

 /// Max delta records appended to the AUX_FILES_KEY (for aux v1). The write path will write a full image once this threshold is reached.
@@ -172,7 +174,6 @@ impl Timeline {
            pending_deletions: Vec::new(),
            pending_nblocks: 0,
            pending_directory_entries: Vec::new(),
-            pending_bytes: 0,
            lsn,
        }
    }
@@ -286,7 +287,10 @@ impl Timeline {
        // then check if the database was already initialized.
        // get_rel_exists can be called before dbdir is created.
        let buf = version.get(self, DBDIR_KEY, ctx).await?;
-        let dbdirs = DbDirectory::des(&buf)?.dbdirs;
+        let dbdirs = match DbDirectory::des(&buf).context("deserialization failure") {
+            Ok(dir) => Ok(dir.dbdirs),
+            Err(e) => Err(PageReconstructError::from(e)),
+        }?;
        if !dbdirs.contains_key(&(tag.spcnode, tag.dbnode)) {
            return Ok(false);
        }
@@ -294,8 +298,13 @@ impl Timeline {
        let key = rel_dir_to_key(tag.spcnode, tag.dbnode);
        let buf = version.get(self, key, ctx).await?;

-        let dir = RelDirectory::des(&buf)?;
-        Ok(dir.rels.contains(&(tag.relnode, tag.forknum)))
+        match RelDirectory::des(&buf).context("deserialization failure") {
+            Ok(dir) => {
+                let exists = dir.rels.contains(&(tag.relnode, tag.forknum));
+                Ok(exists)
+            }
+            Err(e) => Err(PageReconstructError::from(e)),
+        }
    }

    /// Get a list of all existing relations in given tablespace and database.
@@ -314,16 +323,20 @@ impl Timeline {
        let key = rel_dir_to_key(spcnode, dbnode);
        let buf = version.get(self, key, ctx).await?;

-        let dir = RelDirectory::des(&buf)?;
-        let rels: HashSet<RelTag> =
-            HashSet::from_iter(dir.rels.iter().map(|(relnode, forknum)| RelTag {
-                spcnode,
-                dbnode,
-                relnode: *relnode,
-                forknum: *forknum,
-            }));
+        match RelDirectory::des(&buf).context("deserialization failure") {
+            Ok(dir) => {
+                let rels: HashSet<RelTag> =
+                    HashSet::from_iter(dir.rels.iter().map(|(relnode, forknum)| RelTag {
+                        spcnode,
+                        dbnode,
+                        relnode: *relnode,
+                        forknum: *forknum,
+                    }));

-        Ok(rels)
+                Ok(rels)
+            }
+            Err(e) => Err(PageReconstructError::from(e)),
+        }
    }

    /// Get the whole SLRU segment
@@ -385,8 +398,13 @@ impl Timeline {
        let key = slru_dir_to_key(kind);
        let buf = version.get(self, key, ctx).await?;

-        let dir = SlruSegmentDirectory::des(&buf)?;
-        Ok(dir.segments.contains(&segno))
+        match SlruSegmentDirectory::des(&buf).context("deserialization failure") {
+            Ok(dir) => {
+                let exists = dir.segments.contains(&segno);
+                Ok(exists)
+            }
+            Err(e) => Err(PageReconstructError::from(e)),
+        }
    }

    /// Locate LSN, such that all transactions that committed before
@@ -602,7 +620,10 @@ impl Timeline {
        let key = slru_dir_to_key(kind);

        let buf = version.get(self, key, ctx).await?;
-        Ok(SlruSegmentDirectory::des(&buf)?.segments)
+        match SlruSegmentDirectory::des(&buf).context("deserialization failure") {
+            Ok(dir) => Ok(dir.segments),
+            Err(e) => Err(PageReconstructError::from(e)),
+        }
    }

    pub(crate) async fn get_relmap_file(
@@ -626,7 +647,10 @@ impl Timeline {
        // fetch directory entry
        let buf = self.get(DBDIR_KEY, lsn, ctx).await?;

-        Ok(DbDirectory::des(&buf)?.dbdirs)
+        match DbDirectory::des(&buf).context("deserialization failure") {
+            Ok(dir) => Ok(dir.dbdirs),
+            Err(e) => Err(PageReconstructError::from(e)),
+        }
    }

    pub(crate) async fn get_twophase_file(
@@ -648,7 +672,10 @@ impl Timeline {
        // fetch directory entry
        let buf = self.get(TWOPHASEDIR_KEY, lsn, ctx).await?;

-        Ok(TwoPhaseDirectory::des(&buf)?.xids)
+        match TwoPhaseDirectory::des(&buf).context("deserialization failure") {
+            Ok(dir) => Ok(dir.xids),
+            Err(e) => Err(PageReconstructError::from(e)),
+        }
    }

    pub(crate) async fn get_control_file(
@@ -673,7 +700,10 @@ impl Timeline {
        ctx: &RequestContext,
    ) -> Result<HashMap<String, Bytes>, PageReconstructError> {
        match self.get(AUX_FILES_KEY, lsn, ctx).await {
-            Ok(buf) => Ok(AuxFilesDirectory::des(&buf)?.files),
+            Ok(buf) => match AuxFilesDirectory::des(&buf).context("deserialization failure") {
+                Ok(dir) => Ok(dir.files),
+                Err(e) => Err(PageReconstructError::from(e)),
+            },
            Err(e) => {
                // This is expected: historical databases do not have the key.
                debug!("Failed to get info about AUX files: {}", e);
@@ -689,14 +719,13 @@ impl Timeline {
    ) -> Result<HashMap<String, Bytes>, PageReconstructError> {
        let kv = self
            .scan(KeySpace::single(Key::metadata_aux_key_range()), lsn, ctx)
-            .await?;
+            .await
+            .context("scan")?;
        let mut result = HashMap::new();
        let mut sz = 0;
        for (_, v) in kv {
-            let v = v?;
-            let v = aux_file::decode_file_value_bytes(&v)
-                .context("value decode")
-                .map_err(PageReconstructError::Other)?;
+            let v = v.context("get value")?;
+            let v = aux_file::decode_file_value_bytes(&v).context("value decode")?;
            for (fname, content) in v {
                sz += fname.len();
                sz += content.len();
@@ -726,17 +755,7 @@ impl Timeline {
    ) -> Result<HashMap<String, Bytes>, PageReconstructError> {
        let current_policy = self.last_aux_file_policy.load();
        match current_policy {
-            Some(AuxFilePolicy::V1) => {
-                warn!("this timeline is using deprecated aux file policy V1 (policy=V1)");
-                self.list_aux_files_v1(lsn, ctx).await
-            }
-            None => {
-                let res = self.list_aux_files_v1(lsn, ctx).await?;
-                if !res.is_empty() {
-                    warn!("this timeline is using deprecated aux file policy V1 (policy=None)");
-                }
-                Ok(res)
-            }
+            Some(AuxFilePolicy::V1) | None => self.list_aux_files_v1(lsn, ctx).await,
            Some(AuxFilePolicy::V2) => self.list_aux_files_v2(lsn, ctx).await,
            Some(AuxFilePolicy::CrossValidation) => {
                let v1_result = self.list_aux_files_v1(lsn, ctx).await;
@@ -774,10 +793,11 @@ impl Timeline {
    ) -> Result<HashMap<RepOriginId, Lsn>, PageReconstructError> {
        let kv = self
            .scan(KeySpace::single(repl_origin_key_range()), lsn, ctx)
-            .await?;
+            .await
+            .context("scan")?;
        let mut result = HashMap::new();
        for (k, v) in kv {
-            let v = v?;
+            let v = v.context("get value")?;
            let origin_id = k.field6 as RepOriginId;
            let origin_lsn = Lsn::des(&v).unwrap();
            if origin_lsn != Lsn::INVALID {
@@ -1031,33 +1051,21 @@ pub struct DatadirModification<'a> {
    // The put-functions add the modifications here, and they are flushed to the
    // underlying key-value store by the 'finish' function.
    pending_lsns: Vec<Lsn>,
-    pending_updates: HashMap<Key, Vec<(Lsn, usize, Value)>>,
+    pending_updates: HashMap<Key, Vec<(Lsn, Value)>>,
    pending_deletions: Vec<(Range<Key>, Lsn)>,
    pending_nblocks: i64,

    /// For special "directory" keys that store key-value maps, track the size of the map
    /// if it was updated in this modification.
    pending_directory_entries: Vec<(DirectoryKind, usize)>,
-
-    /// An **approximation** of how large our EphemeralFile write will be when committed.
-    pending_bytes: usize,
 }

 impl<'a> DatadirModification<'a> {
-    // When a DatadirModification is committed, we do a monolithic serialization of all its contents.  WAL records can
-    // contain multiple pages, so the pageserver's record-based batch size isn't sufficient to bound this allocation: we
-    // additionally specify a limit on how much payload a DatadirModification may contain before it should be committed.
-    pub(crate) const MAX_PENDING_BYTES: usize = 8 * 1024 * 1024;
-
    /// Get the current lsn
    pub(crate) fn get_lsn(&self) -> Lsn {
        self.lsn
    }

-    pub(crate) fn approx_pending_bytes(&self) -> usize {
-        self.pending_bytes
-    }
-
    /// Set the current lsn
    pub(crate) fn set_lsn(&mut self, lsn: Lsn) -> anyhow::Result<()> {
        ensure!(
@@ -1597,7 +1605,6 @@ impl<'a> DatadirModification<'a> {
                if aux_files_key_v1.is_empty() {
                    None
                } else {
-                    warn!("this timeline is using deprecated aux file policy V1");
                    self.tline.do_switch_aux_policy(AuxFilePolicy::V1)?;
                    Some(AuxFilePolicy::V1)
                }
@@ -1726,17 +1733,12 @@ impl<'a> DatadirModification<'a> {
                    // the original code assumes all other errors are missing keys. Therefore, we keep the code path
                    // the same for now, though in theory, we should only match the `MissingKey` variant.
                    Err(
-                        e @ (PageReconstructError::Other(_)
+                        PageReconstructError::Other(_)
                        | PageReconstructError::WalRedo(_)
-                        | PageReconstructError::MissingKey(_)),
+                        | PageReconstructError::MissingKey { .. },
                    ) => {
                        // Key is missing, we must insert an image as the basis for subsequent deltas.

-                        if !matches!(e, PageReconstructError::MissingKey(_)) {
-                            let e = utils::error::report_compact_sources(&e);
-                            tracing::warn!("treating error as if it was a missing key: {}", e);
-                        }
-
                        let mut dir = AuxFilesDirectory {
                            files: HashMap::new(),
                        };
@@ -1791,25 +1793,21 @@ impl<'a> DatadirModification<'a> {
        // Flush relation and  SLRU data blocks, keep metadata.
        let mut retained_pending_updates = HashMap::<_, Vec<_>>::new();
        for (key, values) in self.pending_updates.drain() {
-            let mut write_batch = Vec::new();
-            for (lsn, value_ser_size, value) in values {
+            for (lsn, value) in values {
                if key.is_rel_block_key() || key.is_slru_block_key() {
                    // This bails out on first error without modifying pending_updates.
                    // That's Ok, cf this function's doc comment.
-                    write_batch.push((key.to_compact(), lsn, value_ser_size, value));
+                    writer.put(key, lsn, &value, ctx).await?;
                } else {
-                    retained_pending_updates.entry(key).or_default().push((
-                        lsn,
-                        value_ser_size,
-                        value,
-                    ));
+                    retained_pending_updates
+                        .entry(key)
+                        .or_default()
+                        .push((lsn, value));
                }
            }
-            writer.put_batch(write_batch, ctx).await?;
        }

        self.pending_updates = retained_pending_updates;
-        self.pending_bytes = 0;

        if pending_nblocks != 0 {
            writer.update_current_logical_size(pending_nblocks * i64::from(BLCKSZ));
@@ -1835,20 +1833,17 @@ impl<'a> DatadirModification<'a> {
        self.pending_nblocks = 0;

        if !self.pending_updates.is_empty() {
-            // Ordering: the items in this batch do not need to be in any global order, but values for
-            // a particular Key must be in Lsn order relative to one another.  InMemoryLayer relies on
-            // this to do efficient updates to its index.
-            let batch: Vec<(CompactKey, Lsn, usize, Value)> = self
-                .pending_updates
-                .drain()
-                .flat_map(|(key, values)| {
-                    values.into_iter().map(move |(lsn, val_ser_size, value)| {
-                        (key.to_compact(), lsn, val_ser_size, value)
-                    })
-                })
-                .collect::<Vec<_>>();
+            // The put_batch call below expects expects the inputs to be sorted by Lsn,
+            // so we do that first.
+            let lsn_ordered_batch: VecMap<Lsn, (Key, Value)> = VecMap::from_iter(
+                self.pending_updates
+                    .drain()
+                    .map(|(key, vals)| vals.into_iter().map(move |(lsn, val)| (lsn, (key, val))))
+                    .kmerge_by(|lhs, rhs| lhs.0 < rhs.0),
+                VecMapOrdering::GreaterOrEqual,
+            );

-            writer.put_batch(batch, ctx).await?;
+            writer.put_batch(lsn_ordered_batch, ctx).await?;
        }

        if !self.pending_deletions.is_empty() {
@@ -1873,8 +1868,6 @@ impl<'a> DatadirModification<'a> {
            writer.update_directory_entries_count(kind, count as u64);
        }

-        self.pending_bytes = 0;
-
        Ok(())
    }

@@ -1891,7 +1884,7 @@ impl<'a> DatadirModification<'a> {
        // Note: we don't check pending_deletions. It is an error to request a
        // value that has been removed, deletion only avoids leaking storage.
        if let Some(values) = self.pending_updates.get(&key) {
-            if let Some((_, _, value)) = values.last() {
+            if let Some((_, value)) = values.last() {
                return if let Value::Image(img) = value {
                    Ok(img.clone())
                } else {
@@ -1900,7 +1893,7 @@ impl<'a> DatadirModification<'a> {
                    // work directly with Images, and we never need to read actual
                    // data pages. We could handle this if we had to, by calling
                    // the walredo manager, but let's keep it simple for now.
-                    Err(PageReconstructError::Other(anyhow::anyhow!(
+                    Err(PageReconstructError::from(anyhow::anyhow!(
                        "unexpected pending WAL record"
                    )))
                };
@@ -1919,17 +1912,13 @@ impl<'a> DatadirModification<'a> {
    fn put(&mut self, key: Key, val: Value) {
        let values = self.pending_updates.entry(key).or_default();
        // Replace the previous value if it exists at the same lsn
-        if let Some((last_lsn, last_value_ser_size, last_value)) = values.last_mut() {
+        if let Some((last_lsn, last_value)) = values.last_mut() {
            if *last_lsn == self.lsn {
-                *last_value_ser_size = val.serialized_size().unwrap() as usize;
                *last_value = val;
                return;
            }
        }
-
-        let val_serialized_size = val.serialized_size().unwrap() as usize;
-        self.pending_bytes += val_serialized_size;
-        values.push((self.lsn, val_serialized_size, val));
+        values.push((self.lsn, val));
    }

    fn delete(&mut self, key_range: Range<Key>) {
@@ -2059,7 +2048,7 @@ mod tests {

        let (tenant, ctx) = harness.load().await;
        let tline = tenant
-            .create_empty_timeline(TIMELINE_ID, Lsn(0x10), DEFAULT_PG_VERSION, &ctx)
+            .create_empty_timeline(TIMELINE_ID, Lsn(0), DEFAULT_PG_VERSION, &ctx)
            .await?;
        let tline = tline.raw_timeline().unwrap();

--- a/pageserver/src/task_mgr.rs
+++ b/pageserver/src/task_mgr.rs
@@ -393,7 +393,7 @@ struct PageServerTask {

    /// Tasks may optionally be launched for a particular tenant/timeline, enabling
    /// later cancelling tasks for that tenant/timeline in [`shutdown_tasks`]
-    tenant_shard_id: TenantShardId,
+    tenant_shard_id: Option<TenantShardId>,
    timeline_id: Option<TimelineId>,

    mutable: Mutex<MutableTaskState>,
@@ -405,7 +405,7 @@ struct PageServerTask {
 pub fn spawn<F>(
    runtime: &tokio::runtime::Handle,
    kind: TaskKind,
-    tenant_shard_id: TenantShardId,
+    tenant_shard_id: Option<TenantShardId>,
    timeline_id: Option<TimelineId>,
    name: &str,
    future: F,
@@ -550,7 +550,7 @@ pub async fn shutdown_tasks(
        let tasks = TASKS.lock().unwrap();
        for task in tasks.values() {
            if (kind.is_none() || Some(task.kind) == kind)
-                && (tenant_shard_id.is_none() || Some(task.tenant_shard_id) == tenant_shard_id)
+                && (tenant_shard_id.is_none() || task.tenant_shard_id == tenant_shard_id)
                && (timeline_id.is_none() || task.timeline_id == timeline_id)
            {
                task.cancel.cancel();
@@ -573,8 +573,13 @@ pub async fn shutdown_tasks(
        };
        if let Some(mut join_handle) = join_handle {
            if log_all {
-                // warn to catch these in tests; there shouldn't be any
-                warn!(name = task.name, tenant_shard_id = ?tenant_shard_id, timeline_id = ?timeline_id, kind = ?task_kind, "stopping left-over");
+                if tenant_shard_id.is_none() {
+                    // there are quite few of these
+                    info!(name = task.name, kind = ?task_kind, "stopping global task");
+                } else {
+                    // warn to catch these in tests; there shouldn't be any
+                    warn!(name = task.name, tenant_shard_id = ?tenant_shard_id, timeline_id = ?timeline_id, kind = ?task_kind, "stopping left-over");
+                }
            }
            if tokio::time::timeout(std::time::Duration::from_secs(1), &mut join_handle)
                .await
--- a/pageserver/src/tenant.rs
+++ b/pageserver/src/tenant.rs
@@ -798,7 +798,7 @@ impl Tenant {
        task_mgr::spawn(
            &tokio::runtime::Handle::current(),
            TaskKind::Attach,
-            tenant_shard_id,
+            Some(tenant_shard_id),
            None,
            "attach tenant",
            async move {
@@ -3741,21 +3741,13 @@ impl Tenant {
    /// less than this (via eviction and on-demand downloads), but this function enables
    /// the Tenant to advertise how much storage it would prefer to have to provide fast I/O
    /// by keeping important things on local disk.
-    ///
-    /// This is a heuristic, not a guarantee: tenants that are long-idle will actually use less
-    /// than they report here, due to layer eviction.  Tenants with many active branches may
-    /// actually use more than they report here.
    pub(crate) fn local_storage_wanted(&self) -> u64 {
+        let mut wanted = 0;
        let timelines = self.timelines.lock().unwrap();
-
-        // Heuristic: we use the max() of the timelines' visible sizes, rather than the sum.  This
-        // reflects the observation that on tenants with multiple large branches, typically only one
-        // of them is used actively enough to occupy space on disk.
-        timelines
-            .values()
-            .map(|t| t.metrics.visible_physical_size_gauge.get())
-            .max()
-            .unwrap_or(0)
+        for timeline in timelines.values() {
+            wanted += timeline.metrics.visible_physical_size_gauge.get();
+        }
+        wanted
    }
 }

@@ -4499,13 +4491,10 @@ mod tests {

        // This needs to traverse to the parent, and fails.
        let err = newtline.get(*TEST_KEY, Lsn(0x50), &ctx).await.unwrap_err();
-        assert!(
-            err.to_string().starts_with(&format!(
-                "bad state on timeline {}: Broken",
-                tline.timeline_id
-            )),
-            "{err}"
-        );
+        assert!(err.to_string().starts_with(&format!(
+            "Bad state on timeline {}: Broken",
+            tline.timeline_id
+        )));

        Ok(())
    }
@@ -5940,10 +5929,10 @@ mod tests {
            .await
            .unwrap();

-        // the default aux file policy to switch is v2 if not set by the admins
+        // the default aux file policy to switch is v1 if not set by the admins
        assert_eq!(
            harness.tenant_conf.switch_aux_file_policy,
-            AuxFilePolicy::default_tenant_config()
+            AuxFilePolicy::V1
        );
        let (tenant, ctx) = harness.load().await;

@@ -5987,8 +5976,8 @@ mod tests {
        );
        assert_eq!(
            tline.last_aux_file_policy.load(),
-            Some(AuxFilePolicy::V2),
-            "aux file is written with switch_aux_file_policy unset (which is v2), so we should use v2 there"
+            Some(AuxFilePolicy::V1),
+            "aux file is written with switch_aux_file_policy unset (which is v1), so we should keep v1"
        );

        // we can read everything from the storage
@@ -6010,8 +5999,8 @@ mod tests {

        assert_eq!(
            tline.last_aux_file_policy.load(),
-            Some(AuxFilePolicy::V2),
-            "keep v2 storage format when new files are written"
+            Some(AuxFilePolicy::V1),
+            "keep v1 storage format when new files are written"
        );

        let files = tline.list_aux_files(lsn, &ctx).await.unwrap();
@@ -6027,7 +6016,7 @@ mod tests {

        // child copies the last flag even if that is not on remote storage yet
        assert_eq!(child.get_switch_aux_file_policy(), AuxFilePolicy::V2);
-        assert_eq!(child.last_aux_file_policy.load(), Some(AuxFilePolicy::V2));
+        assert_eq!(child.last_aux_file_policy.load(), Some(AuxFilePolicy::V1));

        let files = child.list_aux_files(lsn, &ctx).await.unwrap();
        assert_eq!(files.get("pg_logical/mappings/test1"), None);
--- a/pageserver/src/tenant/blob_io.rs
+++ b/pageserver/src/tenant/blob_io.rs
@@ -24,7 +24,6 @@ use tracing::warn;
 use crate::context::RequestContext;
 use crate::page_cache::PAGE_SZ;
 use crate::tenant::block_io::BlockCursor;
-use crate::virtual_file::owned_buffers_io::io_buf_ext::{FullSlice, IoBufExt};
 use crate::virtual_file::VirtualFile;
 use std::cmp::min;
 use std::io::{Error, ErrorKind};
@@ -187,11 +186,11 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
    /// You need to make sure that the internal buffer is empty, otherwise
    /// data will be written in wrong order.
    #[inline(always)]
-    async fn write_all_unbuffered<Buf: IoBuf + Send>(
+    async fn write_all_unbuffered<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
        &mut self,
-        src_buf: FullSlice<Buf>,
+        src_buf: B,
        ctx: &RequestContext,
-    ) -> (FullSlice<Buf>, Result<(), Error>) {
+    ) -> (B::Buf, Result<(), Error>) {
        let (src_buf, res) = self.inner.write_all(src_buf, ctx).await;
        let nbytes = match res {
            Ok(nbytes) => nbytes,
@@ -205,9 +204,8 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
    /// Flushes the internal buffer to the underlying `VirtualFile`.
    pub async fn flush_buffer(&mut self, ctx: &RequestContext) -> Result<(), Error> {
        let buf = std::mem::take(&mut self.buf);
-        let (slice, res) = self.inner.write_all(buf.slice_len(), ctx).await;
+        let (mut buf, res) = self.inner.write_all(buf, ctx).await;
        res?;
-        let mut buf = slice.into_raw_slice().into_inner();
        buf.clear();
        self.buf = buf;
        Ok(())
@@ -224,30 +222,19 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
    }

    /// Internal, possibly buffered, write function
-    async fn write_all<Buf: IoBuf + Send>(
+    async fn write_all<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
        &mut self,
-        src_buf: FullSlice<Buf>,
+        src_buf: B,
        ctx: &RequestContext,
-    ) -> (FullSlice<Buf>, Result<(), Error>) {
-        let src_buf = src_buf.into_raw_slice();
-        let src_buf_bounds = src_buf.bounds();
-        let restore = move |src_buf_slice: Slice<_>| {
-            FullSlice::must_new(Slice::from_buf_bounds(
-                src_buf_slice.into_inner(),
-                src_buf_bounds,
-            ))
-        };
-
+    ) -> (B::Buf, Result<(), Error>) {
        if !BUFFERED {
            assert!(self.buf.is_empty());
-            return self
-                .write_all_unbuffered(FullSlice::must_new(src_buf), ctx)
-                .await;
+            return self.write_all_unbuffered(src_buf, ctx).await;
        }
        let remaining = Self::CAPACITY - self.buf.len();
        let src_buf_len = src_buf.bytes_init();
        if src_buf_len == 0 {
-            return (restore(src_buf), Ok(()));
+            return (Slice::into_inner(src_buf.slice_full()), Ok(()));
        }
        let mut src_buf = src_buf.slice(0..src_buf_len);
        // First try to copy as much as we can into the buffer
@@ -258,7 +245,7 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
        // Then, if the buffer is full, flush it out
        if self.buf.len() == Self::CAPACITY {
            if let Err(e) = self.flush_buffer(ctx).await {
-                return (restore(src_buf), Err(e));
+                return (Slice::into_inner(src_buf), Err(e));
            }
        }
        // Finally, write the tail of src_buf:
@@ -271,29 +258,27 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
                let copied = self.write_into_buffer(&src_buf);
                // We just verified above that src_buf fits into our internal buffer.
                assert_eq!(copied, src_buf.len());
-                restore(src_buf)
+                Slice::into_inner(src_buf)
            } else {
-                let (src_buf, res) = self
-                    .write_all_unbuffered(FullSlice::must_new(src_buf), ctx)
-                    .await;
+                let (src_buf, res) = self.write_all_unbuffered(src_buf, ctx).await;
                if let Err(e) = res {
                    return (src_buf, Err(e));
                }
                src_buf
            }
        } else {
-            restore(src_buf)
+            Slice::into_inner(src_buf)
        };
        (src_buf, Ok(()))
    }

    /// Write a blob of data. Returns the offset that it was written to,
    /// which can be used to retrieve the data later.
-    pub async fn write_blob<Buf: IoBuf + Send>(
+    pub async fn write_blob<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
        &mut self,
-        srcbuf: FullSlice<Buf>,
+        srcbuf: B,
        ctx: &RequestContext,
-    ) -> (FullSlice<Buf>, Result<u64, Error>) {
+    ) -> (B::Buf, Result<u64, Error>) {
        let (buf, res) = self
            .write_blob_maybe_compressed(srcbuf, ctx, ImageCompressionAlgorithm::Disabled)
            .await;
@@ -302,40 +287,43 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {

    /// Write a blob of data. Returns the offset that it was written to,
    /// which can be used to retrieve the data later.
-    pub(crate) async fn write_blob_maybe_compressed<Buf: IoBuf + Send>(
+    pub async fn write_blob_maybe_compressed<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
        &mut self,
-        srcbuf: FullSlice<Buf>,
+        srcbuf: B,
        ctx: &RequestContext,
        algorithm: ImageCompressionAlgorithm,
-    ) -> (FullSlice<Buf>, Result<(u64, CompressionInfo), Error>) {
+    ) -> (B::Buf, Result<(u64, CompressionInfo), Error>) {
        let offset = self.offset;
        let mut compression_info = CompressionInfo {
            written_compressed: false,
            compressed_size: None,
        };

-        let len = srcbuf.len();
+        let len = srcbuf.bytes_init();

        let mut io_buf = self.io_buf.take().expect("we always put it back below");
        io_buf.clear();
        let mut compressed_buf = None;
-        let ((io_buf_slice, hdr_res), srcbuf) = async {
+        let ((io_buf, hdr_res), srcbuf) = async {
            if len < 128 {
                // Short blob. Write a 1-byte length header
                io_buf.put_u8(len as u8);
-                (self.write_all(io_buf.slice_len(), ctx).await, srcbuf)
+                (
+                    self.write_all(io_buf, ctx).await,
+                    srcbuf.slice_full().into_inner(),
+                )
            } else {
                // Write a 4-byte length header
                if len > MAX_SUPPORTED_LEN {
                    return (
                        (
-                            io_buf.slice_len(),
+                            io_buf,
                            Err(Error::new(
                                ErrorKind::Other,
                                format!("blob too large ({len} bytes)"),
                            )),
                        ),
-                        srcbuf,
+                        srcbuf.slice_full().into_inner(),
                    );
                }
                let (high_bit_mask, len_written, srcbuf) = match algorithm {
@@ -348,7 +336,8 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
                        } else {
                            async_compression::tokio::write::ZstdEncoder::new(Vec::new())
                        };
-                        encoder.write_all(&srcbuf[..]).await.unwrap();
+                        let slice = srcbuf.slice_full();
+                        encoder.write_all(&slice[..]).await.unwrap();
                        encoder.shutdown().await.unwrap();
                        let compressed = encoder.into_inner();
                        compression_info.compressed_size = Some(compressed.len());
@@ -356,29 +345,31 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
                            compression_info.written_compressed = true;
                            let compressed_len = compressed.len();
                            compressed_buf = Some(compressed);
-                            (BYTE_ZSTD, compressed_len, srcbuf)
+                            (BYTE_ZSTD, compressed_len, slice.into_inner())
                        } else {
-                            (BYTE_UNCOMPRESSED, len, srcbuf)
+                            (BYTE_UNCOMPRESSED, len, slice.into_inner())
                        }
                    }
-                    ImageCompressionAlgorithm::Disabled => (BYTE_UNCOMPRESSED, len, srcbuf),
+                    ImageCompressionAlgorithm::Disabled => {
+                        (BYTE_UNCOMPRESSED, len, srcbuf.slice_full().into_inner())
+                    }
                };
                let mut len_buf = (len_written as u32).to_be_bytes();
                assert_eq!(len_buf[0] & 0xf0, 0);
                len_buf[0] |= high_bit_mask;
                io_buf.extend_from_slice(&len_buf[..]);
-                (self.write_all(io_buf.slice_len(), ctx).await, srcbuf)
+                (self.write_all(io_buf, ctx).await, srcbuf)
            }
        }
        .await;
-        self.io_buf = Some(io_buf_slice.into_raw_slice().into_inner());
+        self.io_buf = Some(io_buf);
        match hdr_res {
            Ok(_) => (),
-            Err(e) => return (srcbuf, Err(e)),
+            Err(e) => return (Slice::into_inner(srcbuf.slice(..)), Err(e)),
        }
        let (srcbuf, res) = if let Some(compressed_buf) = compressed_buf {
-            let (_buf, res) = self.write_all(compressed_buf.slice_len(), ctx).await;
-            (srcbuf, res)
+            let (_buf, res) = self.write_all(compressed_buf, ctx).await;
+            (Slice::into_inner(srcbuf.slice(..)), res)
        } else {
            self.write_all(srcbuf, ctx).await
        };
@@ -441,21 +432,21 @@ pub(crate) mod tests {
                let (_, res) = if compression {
                    let res = wtr
                        .write_blob_maybe_compressed(
-                            blob.clone().slice_len(),
+                            blob.clone(),
                            ctx,
                            ImageCompressionAlgorithm::Zstd { level: Some(1) },
                        )
                        .await;
                    (res.0, res.1.map(|(off, _)| off))
                } else {
-                    wtr.write_blob(blob.clone().slice_len(), ctx).await
+                    wtr.write_blob(blob.clone(), ctx).await
                };
                let offs = res?;
                offsets.push(offs);
            }
            // Write out one page worth of zeros so that we can
            // read again with read_blk
-            let (_, res) = wtr.write_blob(vec![0; PAGE_SZ].slice_len(), ctx).await;
+            let (_, res) = wtr.write_blob(vec![0; PAGE_SZ], ctx).await;
            let offs = res?;
            println!("Writing final blob at offs={offs}");
            wtr.flush_buffer(ctx).await?;
--- a/pageserver/src/tenant/ephemeral_file.rs
+++ b/pageserver/src/tenant/ephemeral_file.rs
@@ -21,6 +21,7 @@ pub struct EphemeralFile {
 }

 mod page_caching;
+pub(crate) use page_caching::PrewarmOnWrite as PrewarmPageCacheOnWrite;
 mod zero_padded_read_write;

 impl EphemeralFile {
@@ -51,10 +52,12 @@ impl EphemeralFile {
        )
        .await?;

+        let prewarm = conf.l0_flush.prewarm_on_write();
+
        Ok(EphemeralFile {
            _tenant_shard_id: tenant_shard_id,
            _timeline_id: timeline_id,
-            rw: page_caching::RW::new(file, gate_guard),
+            rw: page_caching::RW::new(file, prewarm, gate_guard),
        })
    }

@@ -79,8 +82,6 @@ impl EphemeralFile {
        self.rw.read_blk(blknum, ctx).await
    }

-    #[cfg(test)]
-    // This is a test helper: outside of tests, we are always written to via a pre-serialized batch.
    pub(crate) async fn write_blob(
        &mut self,
        srcbuf: &[u8],
@@ -88,30 +89,17 @@ impl EphemeralFile {
    ) -> Result<u64, io::Error> {
        let pos = self.rw.bytes_written();

-        let mut len_bytes = std::io::Cursor::new(Vec::new());
-        crate::tenant::storage_layer::inmemory_layer::SerializedBatch::write_blob_length(
-            srcbuf.len(),
-            &mut len_bytes,
-        );
-        let len_bytes = len_bytes.into_inner();
-
        // Write the length field
-        self.rw.write_all_borrowed(&len_bytes, ctx).await?;
+        if srcbuf.len() < 0x80 {
+            // short one-byte length header
+            let len_buf = [srcbuf.len() as u8];

-        // Write the payload
-        self.rw.write_all_borrowed(srcbuf, ctx).await?;
-
-        Ok(pos)
-    }
-
-    /// Returns the offset at which the first byte of the input was written, for use
-    /// in constructing indices over the written value.
-    pub(crate) async fn write_raw(
-        &mut self,
-        srcbuf: &[u8],
-        ctx: &RequestContext,
-    ) -> Result<u64, io::Error> {
-        let pos = self.rw.bytes_written();
+            self.rw.write_all_borrowed(&len_buf, ctx).await?;
+        } else {
+            let mut len_buf = u32::to_be_bytes(srcbuf.len() as u32);
+            len_buf[0] |= 0x80;
+            self.rw.write_all_borrowed(&len_buf, ctx).await?;
+        }

        // Write the payload
        self.rw.write_all_borrowed(srcbuf, ctx).await?;
--- a/pageserver/src/tenant/ephemeral_file/page_caching.rs
+++ b/pageserver/src/tenant/ephemeral_file/page_caching.rs
@@ -1,15 +1,14 @@
 //! Wrapper around [`super::zero_padded_read_write::RW`] that uses the
 //! [`crate::page_cache`] to serve reads that need to go to the underlying [`VirtualFile`].
-//!
-//! Subject to removal in <https://github.com/neondatabase/neon/pull/8537>

 use crate::context::RequestContext;
 use crate::page_cache::{self, PAGE_SZ};
 use crate::tenant::block_io::BlockLease;
-use crate::virtual_file::owned_buffers_io::util::size_tracking_writer;
 use crate::virtual_file::VirtualFile;

-use std::io::{self};
+use once_cell::sync::Lazy;
+use std::io::{self, ErrorKind};
+use std::ops::{Deref, Range};
 use tokio_epoll_uring::BoundedBuf;
 use tracing::*;

@@ -18,17 +17,33 @@ use super::zero_padded_read_write;
 /// See module-level comment.
 pub struct RW {
    page_cache_file_id: page_cache::FileId,
-    rw: super::zero_padded_read_write::RW<size_tracking_writer::Writer<VirtualFile>>,
+    rw: super::zero_padded_read_write::RW<PreWarmingWriter>,
    /// Gate guard is held on as long as we need to do operations in the path (delete on drop).
    _gate_guard: utils::sync::gate::GateGuard,
 }

+/// When we flush a block to the underlying [`crate::virtual_file::VirtualFile`],
+/// should we pre-warm the [`crate::page_cache`] with the contents?
+#[derive(Clone, Copy)]
+pub enum PrewarmOnWrite {
+    Yes,
+    No,
+}
+
 impl RW {
-    pub fn new(file: VirtualFile, _gate_guard: utils::sync::gate::GateGuard) -> Self {
+    pub fn new(
+        file: VirtualFile,
+        prewarm_on_write: PrewarmOnWrite,
+        _gate_guard: utils::sync::gate::GateGuard,
+    ) -> Self {
        let page_cache_file_id = page_cache::next_file_id();
        Self {
            page_cache_file_id,
-            rw: super::zero_padded_read_write::RW::new(size_tracking_writer::Writer::new(file)),
+            rw: super::zero_padded_read_write::RW::new(PreWarmingWriter::new(
+                page_cache_file_id,
+                file,
+                prewarm_on_write,
+            )),
            _gate_guard,
        }
    }
@@ -68,10 +83,10 @@ impl RW {
        let vec = Vec::with_capacity(size);

        // read from disk what we've already flushed
-        let file_size_tracking_writer = self.rw.as_writer();
-        let flushed_range = 0..usize::try_from(file_size_tracking_writer.bytes_written()).unwrap();
-        let mut vec = file_size_tracking_writer
-            .as_inner()
+        let writer = self.rw.as_writer();
+        let flushed_range = writer.written_range();
+        let mut vec = writer
+            .file
            .read_exact_at(
                vec.slice(0..(flushed_range.end - flushed_range.start)),
                u64::try_from(flushed_range.start).unwrap(),
@@ -106,7 +121,7 @@ impl RW {
                            format!(
                                "ephemeral file: read immutable page #{}: {}: {:#}",
                                blknum,
-                                self.rw.as_writer().as_inner().path,
+                                self.rw.as_writer().file.path,
                                e,
                            ),
                        )
@@ -116,7 +131,7 @@ impl RW {
                    }
                    page_cache::ReadBufResult::NotFound(write_guard) => {
                        let write_guard = writer
-                            .as_inner()
+                            .file
                            .read_exact_at_page(write_guard, blknum as u64 * PAGE_SZ as u64, ctx)
                            .await?;
                        let read_guard = write_guard.mark_valid();
@@ -138,16 +153,153 @@ impl Drop for RW {

        // unlink the file
        // we are clear to do this, because we have entered a gate
-        let path = &self.rw.as_writer().as_inner().path;
-        let res = std::fs::remove_file(path);
+        let res = std::fs::remove_file(&self.rw.as_writer().file.path);
        if let Err(e) = res {
            if e.kind() != std::io::ErrorKind::NotFound {
                // just never log the not found errors, we cannot do anything for them; on detach
                // the tenant directory is already gone.
                //
                // not found files might also be related to https://github.com/neondatabase/neon/issues/2442
-                error!("could not remove ephemeral file '{path}': {e}");
+                error!(
+                    "could not remove ephemeral file '{}': {}",
+                    self.rw.as_writer().file.path,
+                    e
+                );
            }
        }
    }
 }
+
+struct PreWarmingWriter {
+    prewarm_on_write: PrewarmOnWrite,
+    nwritten_blocks: u32,
+    page_cache_file_id: page_cache::FileId,
+    file: VirtualFile,
+}
+
+impl PreWarmingWriter {
+    fn new(
+        page_cache_file_id: page_cache::FileId,
+        file: VirtualFile,
+        prewarm_on_write: PrewarmOnWrite,
+    ) -> Self {
+        Self {
+            prewarm_on_write,
+            nwritten_blocks: 0,
+            page_cache_file_id,
+            file,
+        }
+    }
+
+    /// Return the byte range within `file` that has been written though `write_all`.
+    ///
+    /// The returned range would be invalidated by another `write_all`. To prevent that, we capture `&_`.
+    fn written_range(&self) -> (impl Deref<Target = Range<usize>> + '_) {
+        let nwritten_blocks = usize::try_from(self.nwritten_blocks).unwrap();
+        struct Wrapper(Range<usize>);
+        impl Deref for Wrapper {
+            type Target = Range<usize>;
+            fn deref(&self) -> &Range<usize> {
+                &self.0
+            }
+        }
+        Wrapper(0..nwritten_blocks * PAGE_SZ)
+    }
+}
+
+impl crate::virtual_file::owned_buffers_io::write::OwnedAsyncWriter for PreWarmingWriter {
+    async fn write_all<
+        B: tokio_epoll_uring::BoundedBuf<Buf = Buf>,
+        Buf: tokio_epoll_uring::IoBuf + Send,
+    >(
+        &mut self,
+        buf: B,
+        ctx: &RequestContext,
+    ) -> std::io::Result<(usize, B::Buf)> {
+        let buf = buf.slice(..);
+        let saved_bounds = buf.bounds(); // save for reconstructing the Slice from iobuf after the IO is done
+        let check_bounds_stuff_works = if cfg!(test) && cfg!(debug_assertions) {
+            Some(buf.to_vec())
+        } else {
+            None
+        };
+        let buflen = buf.len();
+        assert_eq!(
+            buflen % PAGE_SZ,
+            0,
+            "{buflen} ; we know TAIL_SZ is a PAGE_SZ multiple, and write_buffered_borrowed is used"
+        );
+
+        // Do the IO.
+        let iobuf = match self.file.write_all(buf, ctx).await {
+            (iobuf, Ok(nwritten)) => {
+                assert_eq!(nwritten, buflen);
+                iobuf
+            }
+            (_, Err(e)) => {
+                return Err(std::io::Error::new(
+                    ErrorKind::Other,
+                    // order error before path because path is long and error is short
+                    format!(
+                        "ephemeral_file: write_blob: write-back tail self.nwritten_blocks={}, buflen={}, {:#}: {}",
+                        self.nwritten_blocks, buflen, e, self.file.path,
+                    ),
+                ));
+            }
+        };
+
+        // Reconstruct the Slice (the write path consumed the Slice and returned us the underlying IoBuf)
+        let buf = tokio_epoll_uring::Slice::from_buf_bounds(iobuf, saved_bounds);
+        if let Some(check_bounds_stuff_works) = check_bounds_stuff_works {
+            assert_eq!(&check_bounds_stuff_works, &*buf);
+        }
+
+        let nblocks = buflen / PAGE_SZ;
+        let nblocks32 = u32::try_from(nblocks).unwrap();
+
+        if matches!(self.prewarm_on_write, PrewarmOnWrite::Yes) {
+            // Pre-warm page cache with the contents.
+            // At least in isolated bulk ingest benchmarks (test_bulk_insert.py), the pre-warming
+            // benefits the code that writes InMemoryLayer=>L0 layers.
+
+            let cache = page_cache::get();
+            static CTX: Lazy<RequestContext> = Lazy::new(|| {
+                RequestContext::new(
+                    crate::task_mgr::TaskKind::EphemeralFilePreWarmPageCache,
+                    crate::context::DownloadBehavior::Error,
+                )
+            });
+            for blknum_in_buffer in 0..nblocks {
+                let blk_in_buffer =
+                    &buf[blknum_in_buffer * PAGE_SZ..(blknum_in_buffer + 1) * PAGE_SZ];
+                let blknum = self
+                    .nwritten_blocks
+                    .checked_add(blknum_in_buffer as u32)
+                    .unwrap();
+                match cache
+                    .read_immutable_buf(self.page_cache_file_id, blknum, &CTX)
+                    .await
+                {
+                    Err(e) => {
+                        error!("ephemeral_file write_blob failed to get immutable buf to pre-warm page cache: {e:?}");
+                        // fail gracefully, it's not the end of the world if we can't pre-warm the cache here
+                    }
+                    Ok(v) => match v {
+                        page_cache::ReadBufResult::Found(_guard) => {
+                            // This function takes &mut self, so, it shouldn't be possible to reach this point.
+                            unreachable!("we just wrote block {blknum} to the VirtualFile, which is owned by Self, \
+                                      and this function takes &mut self, so, no concurrent read_blk is possible");
+                        }
+                        page_cache::ReadBufResult::NotFound(mut write_guard) => {
+                            write_guard.copy_from_slice(blk_in_buffer);
+                            let _ = write_guard.mark_valid();
+                        }
+                    },
+                }
+            }
+        }
+
+        self.nwritten_blocks = self.nwritten_blocks.checked_add(nblocks32).unwrap();
+        Ok((buflen, buf.into_inner()))
+    }
+}
--- a/pageserver/src/tenant/ephemeral_file/zero_padded_read_write/zero_padded.rs
+++ b/pageserver/src/tenant/ephemeral_file/zero_padded_read_write/zero_padded.rs
@@ -5,8 +5,6 @@

 use std::mem::MaybeUninit;

-use crate::virtual_file::owned_buffers_io::io_buf_ext::FullSlice;
-
 /// See module-level comment.
 pub struct Buffer<const N: usize> {
    allocation: Box<[u8; N]>,
@@ -62,10 +60,10 @@ impl<const N: usize> crate::virtual_file::owned_buffers_io::write::Buffer for Bu
        self.written
    }

-    fn flush(self) -> FullSlice<Self> {
+    fn flush(self) -> tokio_epoll_uring::Slice<Self> {
        self.invariants();
        let written = self.written;
-        FullSlice::must_new(tokio_epoll_uring::BoundedBuf::slice(self, 0..written))
+        tokio_epoll_uring::BoundedBuf::slice(self, 0..written)
    }

    fn reuse_after_flush(iobuf: Self::IoBuf) -> Self {
--- a/pageserver/src/tenant/layer_map.rs
+++ b/pageserver/src/tenant/layer_map.rs
@@ -464,7 +464,7 @@ impl LayerMap {
    pub(self) fn insert_historic_noflush(&mut self, layer_desc: PersistentLayerDesc) {
        // TODO: See #3869, resulting #4088, attempted fix and repro #4094

-        if Self::is_l0(&layer_desc.key_range, layer_desc.is_delta) {
+        if Self::is_l0(&layer_desc.key_range) {
            self.l0_delta_layers.push(layer_desc.clone().into());
        }

@@ -483,7 +483,7 @@ impl LayerMap {
        self.historic
            .remove(historic_layer_coverage::LayerKey::from(layer_desc));
        let layer_key = layer_desc.key();
-        if Self::is_l0(&layer_desc.key_range, layer_desc.is_delta) {
+        if Self::is_l0(&layer_desc.key_range) {
            let len_before = self.l0_delta_layers.len();
            let mut l0_delta_layers = std::mem::take(&mut self.l0_delta_layers);
            l0_delta_layers.retain(|other| other.key() != layer_key);
@@ -600,8 +600,8 @@ impl LayerMap {
    }

    /// Check if the key range resembles that of an L0 layer.
-    pub fn is_l0(key_range: &Range<Key>, is_delta_layer: bool) -> bool {
-        is_delta_layer && key_range == &(Key::MIN..Key::MAX)
+    pub fn is_l0(key_range: &Range<Key>) -> bool {
+        key_range == &(Key::MIN..Key::MAX)
    }

    /// This function determines which layers are counted in `count_deltas`:
@@ -628,7 +628,7 @@ impl LayerMap {
    ///      than just the current partition_range.
    pub fn is_reimage_worthy(layer: &PersistentLayerDesc, partition_range: &Range<Key>) -> bool {
        // Case 1
-        if !Self::is_l0(&layer.key_range, layer.is_delta) {
+        if !Self::is_l0(&layer.key_range) {
            return true;
        }

--- a/pageserver/src/tenant/metadata.rs
+++ b/pageserver/src/tenant/metadata.rs
@@ -565,7 +565,7 @@ mod tests {
        );
        let expected_bytes = vec![
            /* TimelineMetadataHeader */
-            74, 104, 158, 105, 0, 70, 0, 4, // checksum, size, format_version (4 + 2 + 2)
+            4, 37, 101, 34, 0, 70, 0, 4, // checksum, size, format_version (4 + 2 + 2)
            /* TimelineMetadataBodyV2 */
            0, 0, 0, 0, 0, 0, 2, 0, // disk_consistent_lsn (8 bytes)
            1, 0, 0, 0, 0, 0, 0, 1, 0, // prev_record_lsn (9 bytes)
@@ -574,7 +574,7 @@ mod tests {
            0, 0, 0, 0, 0, 0, 0, 0, // ancestor_lsn (8 bytes)
            0, 0, 0, 0, 0, 0, 0, 0, // latest_gc_cutoff_lsn (8 bytes)
            0, 0, 0, 0, 0, 0, 0, 0, // initdb_lsn (8 bytes)
-            0, 0, 0, 16, // pg_version (4 bytes)
+            0, 0, 0, 15, // pg_version (4 bytes)
            /* padding bytes */
            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
--- a/pageserver/src/tenant/remote_timeline_client.rs
+++ b/pageserver/src/tenant/remote_timeline_client.rs
@@ -1728,7 +1728,7 @@ impl RemoteTimelineClient {
            task_mgr::spawn(
                &self.runtime,
                TaskKind::RemoteUploadTask,
-                self.tenant_shard_id,
+                Some(self.tenant_shard_id),
                Some(self.timeline_id),
                "remote upload",
                async move {
--- a/pageserver/src/tenant/remote_timeline_client/download.rs
+++ b/pageserver/src/tenant/remote_timeline_client/download.rs
@@ -23,8 +23,6 @@ use crate::span::debug_assert_current_span_has_tenant_and_timeline_id;
 use crate::tenant::remote_timeline_client::{remote_layer_path, remote_timelines_path};
 use crate::tenant::storage_layer::LayerName;
 use crate::tenant::Generation;
-#[cfg_attr(target_os = "macos", allow(unused_imports))]
-use crate::virtual_file::owned_buffers_io::io_buf_ext::IoBufExt;
 use crate::virtual_file::{on_fatal_io_error, MaybeFatalIo, VirtualFile};
 use crate::TEMP_FILE_SUFFIX;
 use remote_storage::{DownloadError, GenericRemoteStorage, ListingMode, RemotePath};
@@ -221,7 +219,9 @@ async fn download_object<'a>(
                            Ok(chunk) => chunk,
                            Err(e) => return Err(e),
                        };
-                        buffered.write_buffered(chunk.slice_len(), ctx).await?;
+                        buffered
+                            .write_buffered(tokio_epoll_uring::BoundedBuf::slice_full(chunk), ctx)
+                            .await?;
                    }
                    let size_tracking = buffered.flush_and_into_inner(ctx).await?;
                    Ok(size_tracking.into_inner())
--- a/pageserver/src/tenant/secondary.rs
+++ b/pageserver/src/tenant/secondary.rs
@@ -8,7 +8,6 @@ use std::{sync::Arc, time::SystemTime};
 use crate::{
    context::RequestContext,
    disk_usage_eviction_task::DiskUsageEvictionInfo,
-    metrics::SECONDARY_HEATMAP_TOTAL_SIZE,
    task_mgr::{self, TaskKind, BACKGROUND_RUNTIME},
 };

@@ -106,9 +105,6 @@ pub(crate) struct SecondaryTenant {

    // Sum of layer sizes on local disk
    pub(super) resident_size_metric: UIntGauge,
-
-    // Sum of layer sizes in the most recently downloaded heatmap
-    pub(super) heatmap_total_size_metric: UIntGauge,
 }

 impl Drop for SecondaryTenant {
@@ -116,7 +112,6 @@ impl Drop for SecondaryTenant {
        let tenant_id = self.tenant_shard_id.tenant_id.to_string();
        let shard_id = format!("{}", self.tenant_shard_id.shard_slug());
        let _ = SECONDARY_RESIDENT_PHYSICAL_SIZE.remove_label_values(&[&tenant_id, &shard_id]);
-        let _ = SECONDARY_HEATMAP_TOTAL_SIZE.remove_label_values(&[&tenant_id, &shard_id]);
    }
 }

@@ -133,10 +128,6 @@ impl SecondaryTenant {
            .get_metric_with_label_values(&[&tenant_id, &shard_id])
            .unwrap();

-        let heatmap_total_size_metric = SECONDARY_HEATMAP_TOTAL_SIZE
-            .get_metric_with_label_values(&[&tenant_id, &shard_id])
-            .unwrap();
-
        Arc::new(Self {
            tenant_shard_id,
            // todo: shall we make this a descendent of the
@@ -154,7 +145,6 @@ impl SecondaryTenant {
            progress: std::sync::Mutex::default(),

            resident_size_metric,
-            heatmap_total_size_metric,
        })
    }

--- a/pageserver/src/tenant/secondary/downloader.rs
+++ b/pageserver/src/tenant/secondary/downloader.rs
@@ -22,7 +22,7 @@ use crate::{
            FAILED_REMOTE_OP_RETRIES,
        },
        span::debug_assert_current_span_has_tenant_id,
-        storage_layer::{layer::local_layer_path, LayerName, LayerVisibilityHint},
+        storage_layer::{layer::local_layer_path, LayerName},
        tasks::{warn_when_period_overrun, BackgroundLoopKind},
    },
    virtual_file::{on_fatal_io_error, MaybeFatalIo, VirtualFile},
@@ -296,9 +296,6 @@ impl SecondaryDetail {
                        }),
                        last_activity_ts: ods.access_time,
                        relative_last_activity: finite_f32::FiniteF32::ZERO,
-                        // Secondary location layers are presumed visible, because Covered layers
-                        // are excluded from the heatmap
-                        visibility: LayerVisibilityHint::Visible,
                    }
                }));

@@ -829,12 +826,6 @@ impl<'a> TenantDownloader<'a> {
            layers_downloaded: 0,
            bytes_downloaded: 0,
        };
-
-        // Also expose heatmap bytes_total as a metric
-        self.secondary_state
-            .heatmap_total_size_metric
-            .set(heatmap_stats.bytes);
-
        // Accumulate list of things to delete while holding the detail lock, for execution after dropping the lock
        let mut delete_layers = Vec::new();
        let mut delete_timelines = Vec::new();
--- a/pageserver/src/tenant/secondary/heatmap.rs
+++ b/pageserver/src/tenant/secondary/heatmap.rs
@@ -29,16 +29,16 @@ pub(super) struct HeatMapTenant {
 #[derive(Serialize, Deserialize)]
 pub(crate) struct HeatMapTimeline {
    #[serde_as(as = "DisplayFromStr")]
-    pub(crate) timeline_id: TimelineId,
+    pub(super) timeline_id: TimelineId,

-    pub(crate) layers: Vec<HeatMapLayer>,
+    pub(super) layers: Vec<HeatMapLayer>,
 }

 #[serde_as]
 #[derive(Serialize, Deserialize)]
 pub(crate) struct HeatMapLayer {
-    pub(crate) name: LayerName,
-    pub(crate) metadata: LayerFileMetadata,
+    pub(super) name: LayerName,
+    pub(super) metadata: LayerFileMetadata,

    #[serde_as(as = "TimestampSeconds<i64>")]
    pub(super) access_time: SystemTime,
--- a/pageserver/src/tenant/storage_layer.rs
+++ b/pageserver/src/tenant/storage_layer.rs
@@ -2,7 +2,7 @@

 pub mod delta_layer;
 pub mod image_layer;
-pub mod inmemory_layer;
+pub(crate) mod inmemory_layer;
 pub(crate) mod layer;
 mod layer_desc;
 mod layer_name;
--- a/pageserver/src/tenant/storage_layer/delta_layer.rs
+++ b/pageserver/src/tenant/storage_layer/delta_layer.rs
@@ -39,10 +39,9 @@ use crate::tenant::disk_btree::{
 use crate::tenant::timeline::GetVectoredError;
 use crate::tenant::vectored_blob_io::{
    BlobFlag, MaxVectoredReadBytes, StreamingVectoredReadPlanner, VectoredBlobReader, VectoredRead,
-    VectoredReadCoalesceMode, VectoredReadPlanner,
+    VectoredReadPlanner,
 };
 use crate::tenant::PageReconstructError;
-use crate::virtual_file::owned_buffers_io::io_buf_ext::{FullSlice, IoBufExt};
 use crate::virtual_file::{self, VirtualFile};
 use crate::{walrecord, TEMP_FILE_SUFFIX};
 use crate::{DELTA_FILE_MAGIC, STORAGE_FORMAT_VERSION};
@@ -64,7 +63,6 @@ use std::os::unix::fs::FileExt;
 use std::str::FromStr;
 use std::sync::Arc;
 use tokio::sync::OnceCell;
-use tokio_epoll_uring::IoBufMut;
 use tracing::*;

 use utils::{
@@ -232,18 +230,6 @@ pub struct DeltaLayerInner {
    max_vectored_read_bytes: Option<MaxVectoredReadBytes>,
 }

-impl DeltaLayerInner {
-    pub(crate) fn layer_dbg_info(&self) -> String {
-        format!(
-            "delta {}..{} {}..{}",
-            self.key_range().start,
-            self.key_range().end,
-            self.lsn_range().start,
-            self.lsn_range().end
-        )
-    }
-}
-
 impl std::fmt::Debug for DeltaLayerInner {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.debug_struct("DeltaLayerInner")
@@ -450,28 +436,19 @@ impl DeltaLayerWriterInner {
        ctx: &RequestContext,
    ) -> anyhow::Result<()> {
        let (_, res) = self
-            .put_value_bytes(
-                key,
-                lsn,
-                Value::ser(&val)?.slice_len(),
-                val.will_init(),
-                ctx,
-            )
+            .put_value_bytes(key, lsn, Value::ser(&val)?, val.will_init(), ctx)
            .await;
        res
    }

-    async fn put_value_bytes<Buf>(
+    async fn put_value_bytes(
        &mut self,
        key: Key,
        lsn: Lsn,
-        val: FullSlice<Buf>,
+        val: Vec<u8>,
        will_init: bool,
        ctx: &RequestContext,
-    ) -> (FullSlice<Buf>, anyhow::Result<()>)
-    where
-        Buf: IoBufMut + Send,
-    {
+    ) -> (Vec<u8>, anyhow::Result<()>) {
        assert!(
            self.lsn_range.start <= lsn,
            "lsn_start={}, lsn={}",
@@ -537,7 +514,7 @@ impl DeltaLayerWriterInner {
        file.seek(SeekFrom::Start(index_start_blk as u64 * PAGE_SZ as u64))
            .await?;
        for buf in block_buf.blocks {
-            let (_buf, res) = file.write_all(buf.slice_len(), ctx).await;
+            let (_buf, res) = file.write_all(buf, ctx).await;
            res?;
        }
        assert!(self.lsn_range.start < self.lsn_range.end);
@@ -557,7 +534,7 @@ impl DeltaLayerWriterInner {
        // TODO: could use smallvec here but it's a pain with Slice<T>
        Summary::ser_into(&summary, &mut buf)?;
        file.seek(SeekFrom::Start(0)).await?;
-        let (_buf, res) = file.write_all(buf.slice_len(), ctx).await;
+        let (_buf, res) = file.write_all(buf, ctx).await;
        res?;

        let metadata = file
@@ -669,17 +646,14 @@ impl DeltaLayerWriter {
            .await
    }

-    pub async fn put_value_bytes<Buf>(
+    pub async fn put_value_bytes(
        &mut self,
        key: Key,
        lsn: Lsn,
-        val: FullSlice<Buf>,
+        val: Vec<u8>,
        will_init: bool,
        ctx: &RequestContext,
-    ) -> (FullSlice<Buf>, anyhow::Result<()>)
-    where
-        Buf: IoBufMut + Send,
-    {
+    ) -> (Vec<u8>, anyhow::Result<()>) {
        self.inner
            .as_mut()
            .unwrap()
@@ -769,7 +743,7 @@ impl DeltaLayer {
        // TODO: could use smallvec here, but it's a pain with Slice<T>
        Summary::ser_into(&new_summary, &mut buf).context("serialize")?;
        file.seek(SeekFrom::Start(0)).await?;
-        let (_buf, res) = file.write_all(buf.slice_len(), ctx).await;
+        let (_buf, res) = file.write_all(buf, ctx).await;
        res?;
        Ok(())
    }
@@ -1046,7 +1020,7 @@ impl DeltaLayerInner {
                    for (_, blob_meta) in read.blobs_at.as_slice() {
                        reconstruct_state.on_key_error(
                            blob_meta.key,
-                            PageReconstructError::Other(anyhow!(
+                            PageReconstructError::from(anyhow!(
                                "Failed to read blobs from virtual file {}: {}",
                                self.file.path,
                                kind
@@ -1073,7 +1047,7 @@ impl DeltaLayerInner {
                    Err(e) => {
                        reconstruct_state.on_key_error(
                            meta.meta.key,
-                            PageReconstructError::Other(anyhow!(e).context(format!(
+                            PageReconstructError::from(anyhow!(e).context(format!(
                                "Failed to deserialize blob from virtual file {}",
                                self.file.path,
                            ))),
@@ -1207,7 +1181,6 @@ impl DeltaLayerInner {
        let mut prev: Option<(Key, Lsn, BlobRef)> = None;

        let mut read_builder: Option<VectoredReadBuilder> = None;
-        let read_mode = VectoredReadCoalesceMode::get();

        let max_read_size = self
            .max_vectored_read_bytes
@@ -1256,7 +1229,6 @@ impl DeltaLayerInner {
                        offsets.end.pos(),
                        meta,
                        max_read_size,
-                        read_mode,
                    ))
                }
            } else {
@@ -1319,12 +1291,12 @@ impl DeltaLayerInner {
                        .put_value_bytes(
                            key,
                            lsn,
-                            std::mem::take(&mut per_blob_copy).slice_len(),
+                            std::mem::take(&mut per_blob_copy),
                            will_init,
                            ctx,
                        )
                        .await;
-                    per_blob_copy = tmp.into_raw_slice().into_inner();
+                    per_blob_copy = tmp;

                    res?;

@@ -1541,10 +1513,6 @@ pub struct DeltaLayerIterator<'a> {
 }

 impl<'a> DeltaLayerIterator<'a> {
-    pub(crate) fn layer_dbg_info(&self) -> String {
-        self.delta_layer.layer_dbg_info()
-    }
-
    /// Retrieve a batch of key-value pairs into the iterator buffer.
    async fn next_batch(&mut self) -> anyhow::Result<()> {
        assert!(self.key_values_batch.is_empty());
@@ -1903,7 +1871,7 @@ pub(crate) mod test {

        for entry in entries {
            let (_, res) = writer
-                .put_value_bytes(entry.key, entry.lsn, entry.value.slice_len(), false, &ctx)
+                .put_value_bytes(entry.key, entry.lsn, entry.value, false, &ctx)
                .await;
            res?;
        }
@@ -2285,7 +2253,7 @@ pub(crate) mod test {
            .await
            .unwrap();
        let delta_layer = resident_layer.get_as_delta(&ctx).await.unwrap();
-        for max_read_size in [1, 2048] {
+        for max_read_size in [1, 1024] {
            for batch_size in [1, 2, 4, 8, 3, 7, 13] {
                println!("running with batch_size={batch_size} max_read_size={max_read_size}");
                // Test if the batch size is correctly determined
--- a/pageserver/src/tenant/storage_layer/image_layer.rs
+++ b/pageserver/src/tenant/storage_layer/image_layer.rs
@@ -38,7 +38,6 @@ use crate::tenant::vectored_blob_io::{
    VectoredReadPlanner,
 };
 use crate::tenant::{PageReconstructError, Timeline};
-use crate::virtual_file::owned_buffers_io::io_buf_ext::IoBufExt;
 use crate::virtual_file::{self, VirtualFile};
 use crate::{IMAGE_FILE_MAGIC, STORAGE_FORMAT_VERSION, TEMP_FILE_SUFFIX};
 use anyhow::{anyhow, bail, ensure, Context, Result};
@@ -167,17 +166,6 @@ pub struct ImageLayerInner {
    max_vectored_read_bytes: Option<MaxVectoredReadBytes>,
 }

-impl ImageLayerInner {
-    pub(crate) fn layer_dbg_info(&self) -> String {
-        format!(
-            "image {}..{} {}",
-            self.key_range().start,
-            self.key_range().end,
-            self.lsn()
-        )
-    }
-}
-
 impl std::fmt::Debug for ImageLayerInner {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.debug_struct("ImageLayerInner")
@@ -366,7 +354,7 @@ impl ImageLayer {
        // TODO: could use smallvec here but it's a pain with Slice<T>
        Summary::ser_into(&new_summary, &mut buf).context("serialize")?;
        file.seek(SeekFrom::Start(0)).await?;
-        let (_buf, res) = file.write_all(buf.slice_len(), ctx).await;
+        let (_buf, res) = file.write_all(buf, ctx).await;
        res?;
        Ok(())
    }
@@ -798,7 +786,7 @@ impl ImageLayerWriterInner {
        self.num_keys += 1;
        let (_img, res) = self
            .blob_writer
-            .write_blob_maybe_compressed(img.slice_len(), ctx, compression)
+            .write_blob_maybe_compressed(img, ctx, compression)
            .await;
        // TODO: re-use the buffer for `img` further upstack
        let (off, compression_info) = res?;
@@ -850,7 +838,7 @@ impl ImageLayerWriterInner {
            .await?;
        let (index_root_blk, block_buf) = self.tree.finish()?;
        for buf in block_buf.blocks {
-            let (_buf, res) = file.write_all(buf.slice_len(), ctx).await;
+            let (_buf, res) = file.write_all(buf, ctx).await;
            res?;
        }

@@ -870,7 +858,7 @@ impl ImageLayerWriterInner {
        // TODO: could use smallvec here but it's a pain with Slice<T>
        Summary::ser_into(&summary, &mut buf)?;
        file.seek(SeekFrom::Start(0)).await?;
-        let (_buf, res) = file.write_all(buf.slice_len(), ctx).await;
+        let (_buf, res) = file.write_all(buf, ctx).await;
        res?;

        let metadata = file
@@ -1035,10 +1023,6 @@ pub struct ImageLayerIterator<'a> {
 }

 impl<'a> ImageLayerIterator<'a> {
-    pub(crate) fn layer_dbg_info(&self) -> String {
-        self.image_layer.layer_dbg_info()
-    }
-
    /// Retrieve a batch of key-value pairs into the iterator buffer.
    async fn next_batch(&mut self) -> anyhow::Result<()> {
        assert!(self.key_values_batch.is_empty());
@@ -1376,7 +1360,7 @@ mod test {
                .await
                .unwrap();
        let img_layer = resident_layer.get_as_image(&ctx).await.unwrap();
-        for max_read_size in [1, 2048] {
+        for max_read_size in [1, 1024] {
            for batch_size in [1, 2, 4, 8, 3, 7, 13] {
                println!("running with batch_size={batch_size} max_read_size={max_read_size}");
                // Test if the batch size is correctly determined
--- a/pageserver/src/tenant/storage_layer/inmemory_layer.rs
+++ b/pageserver/src/tenant/storage_layer/inmemory_layer.rs
@@ -12,8 +12,7 @@ use crate::tenant::block_io::{BlockCursor, BlockReader, BlockReaderRef};
 use crate::tenant::ephemeral_file::EphemeralFile;
 use crate::tenant::timeline::GetVectoredError;
 use crate::tenant::PageReconstructError;
-use crate::virtual_file::owned_buffers_io::io_buf_ext::IoBufExt;
-use crate::{l0_flush, page_cache};
+use crate::{l0_flush, page_cache, walrecord};
 use anyhow::{anyhow, Result};
 use camino::Utf8PathBuf;
 use pageserver_api::key::CompactKey;
@@ -33,7 +32,7 @@ use std::fmt::Write;
 use std::ops::Range;
 use std::sync::atomic::Ordering as AtomicOrdering;
 use std::sync::atomic::{AtomicU64, AtomicUsize};
-use tokio::sync::RwLock;
+use tokio::sync::{RwLock, RwLockWriteGuard};

 use super::{
    DeltaLayerWriter, PersistentLayerDesc, ValueReconstructSituation, ValuesReconstructState,
@@ -249,7 +248,9 @@ impl InMemoryLayer {
    /// debugging function to print out the contents of the layer
    ///
    /// this is likely completly unused
-    pub async fn dump(&self, _verbose: bool, _ctx: &RequestContext) -> Result<()> {
+    pub async fn dump(&self, verbose: bool, ctx: &RequestContext) -> Result<()> {
+        let inner = self.inner.read().await;
+
        let end_str = self.end_lsn_or_max();

        println!(
@@ -257,6 +258,39 @@ impl InMemoryLayer {
            self.timeline_id, self.start_lsn, end_str,
        );

+        if !verbose {
+            return Ok(());
+        }
+
+        let cursor = inner.file.block_cursor();
+        let mut buf = Vec::new();
+        for (key, vec_map) in inner.index.iter() {
+            for (lsn, pos) in vec_map.as_slice() {
+                let mut desc = String::new();
+                cursor.read_blob_into_buf(*pos, &mut buf, ctx).await?;
+                let val = Value::des(&buf);
+                match val {
+                    Ok(Value::Image(img)) => {
+                        write!(&mut desc, " img {} bytes", img.len())?;
+                    }
+                    Ok(Value::WalRecord(rec)) => {
+                        let wal_desc = walrecord::describe_wal_record(&rec).unwrap();
+                        write!(
+                            &mut desc,
+                            " rec {} bytes will_init: {} {}",
+                            buf.len(),
+                            rec.will_init(),
+                            wal_desc
+                        )?;
+                    }
+                    Err(err) => {
+                        write!(&mut desc, " DESERIALIZATION ERROR: {}", err)?;
+                    }
+                }
+                println!("  key {} at {}: {}", key, lsn, desc);
+            }
+        }
+
        Ok(())
    }

@@ -320,82 +354,6 @@ impl InMemoryLayer {
    }
 }

-/// Offset of a particular Value within a serialized batch.
-struct SerializedBatchOffset {
-    key: CompactKey,
-    lsn: Lsn,
-    /// offset in bytes from the start of the batch's buffer to the Value's serialized size header.
-    offset: u64,
-}
-
-pub struct SerializedBatch {
-    /// Blobs serialized in EphemeralFile's native format, ready for passing to [`EphemeralFile::write_raw`].
-    pub(crate) raw: Vec<u8>,
-
-    /// Index of values in [`Self::raw`], using offsets relative to the start of the buffer.
-    offsets: Vec<SerializedBatchOffset>,
-
-    /// The highest LSN of any value in the batch
-    pub(crate) max_lsn: Lsn,
-}
-
-impl SerializedBatch {
-    /// Write a blob length in the internal format of the EphemeralFile
-    pub(crate) fn write_blob_length(len: usize, cursor: &mut std::io::Cursor<Vec<u8>>) {
-        use std::io::Write;
-
-        if len < 0x80 {
-            // short one-byte length header
-            let len_buf = [len as u8];
-
-            cursor
-                .write_all(&len_buf)
-                .expect("Writing to Vec is infallible");
-        } else {
-            let mut len_buf = u32::to_be_bytes(len as u32);
-            len_buf[0] |= 0x80;
-            cursor
-                .write_all(&len_buf)
-                .expect("Writing to Vec is infallible");
-        }
-    }
-
-    pub fn from_values(batch: Vec<(CompactKey, Lsn, usize, Value)>) -> Self {
-        // Pre-allocate a big flat buffer to write into. This should be large but not huge: it is soft-limited in practice by
-        // [`crate::pgdatadir_mapping::DatadirModification::MAX_PENDING_BYTES`]
-        let buffer_size = batch.iter().map(|i| i.2).sum::<usize>() + 4 * batch.len();
-        let mut cursor = std::io::Cursor::new(Vec::<u8>::with_capacity(buffer_size));
-
-        let mut offsets: Vec<SerializedBatchOffset> = Vec::with_capacity(batch.len());
-        let mut max_lsn: Lsn = Lsn(0);
-        for (key, lsn, val_ser_size, val) in batch {
-            let relative_off = cursor.position();
-
-            Self::write_blob_length(val_ser_size, &mut cursor);
-            val.ser_into(&mut cursor)
-                .expect("Writing into in-memory buffer is infallible");
-
-            offsets.push(SerializedBatchOffset {
-                key,
-                lsn,
-                offset: relative_off,
-            });
-            max_lsn = std::cmp::max(max_lsn, lsn);
-        }
-
-        let buffer = cursor.into_inner();
-
-        // Assert that we didn't do any extra allocations while building buffer.
-        debug_assert!(buffer.len() <= buffer_size);
-
-        Self {
-            raw: buffer,
-            offsets,
-            max_lsn,
-        }
-    }
-}
-
 fn inmem_layer_display(mut f: impl Write, start_lsn: Lsn, end_lsn: Lsn) -> std::fmt::Result {
    write!(f, "inmem-{:016X}-{:016X}", start_lsn.0, end_lsn.0)
 }
@@ -456,20 +414,37 @@ impl InMemoryLayer {
        })
    }

-    // Write path.
-    pub async fn put_batch(
+    // Write operations
+
+    /// Common subroutine of the public put_wal_record() and put_page_image() functions.
+    /// Adds the page version to the in-memory tree
+    pub async fn put_value(
        &self,
-        serialized_batch: SerializedBatch,
+        key: CompactKey,
+        lsn: Lsn,
+        buf: &[u8],
        ctx: &RequestContext,
    ) -> Result<()> {
        let mut inner = self.inner.write().await;
        self.assert_writable();
+        self.put_value_locked(&mut inner, key, lsn, buf, ctx).await
+    }

-        let base_off = {
-            inner
+    async fn put_value_locked(
+        &self,
+        locked_inner: &mut RwLockWriteGuard<'_, InMemoryLayerInner>,
+        key: CompactKey,
+        lsn: Lsn,
+        buf: &[u8],
+        ctx: &RequestContext,
+    ) -> Result<()> {
+        trace!("put_value key {} at {}/{}", key, self.timeline_id, lsn);
+
+        let off = {
+            locked_inner
                .file
-                .write_raw(
-                    &serialized_batch.raw,
+                .write_blob(
+                    buf,
                    &RequestContextBuilder::extend(ctx)
                        .page_content_kind(PageContentKind::InMemoryLayer)
                        .build(),
@@ -477,23 +452,15 @@ impl InMemoryLayer {
                .await?
        };

-        for SerializedBatchOffset {
-            key,
-            lsn,
-            offset: relative_off,
-        } in serialized_batch.offsets
-        {
-            let off = base_off + relative_off;
-            let vec_map = inner.index.entry(key).or_default();
-            let old = vec_map.append_or_update_last(lsn, off).unwrap().0;
-            if old.is_some() {
-                // We already had an entry for this LSN. That's odd..
-                warn!("Key {} at {} already exists", key, lsn);
-            }
+        let vec_map = locked_inner.index.entry(key).or_default();
+        let old = vec_map.append_or_update_last(lsn, off).unwrap().0;
+        if old.is_some() {
+            // We already had an entry for this LSN. That's odd..
+            warn!("Key {} at {} already exists", key, lsn);
        }

-        let size = inner.file.len();
-        inner.resource_units.maybe_publish_size(size);
+        let size = locked_inner.file.len();
+        locked_inner.resource_units.maybe_publish_size(size);

        Ok(())
    }
@@ -568,6 +535,7 @@ impl InMemoryLayer {

        use l0_flush::Inner;
        let _concurrency_permit = match l0_flush_global_state {
+            Inner::PageCached => None,
            Inner::Direct { semaphore, .. } => Some(semaphore.acquire().await),
        };

@@ -599,6 +567,28 @@ impl InMemoryLayer {
        .await?;

        match l0_flush_global_state {
+            l0_flush::Inner::PageCached => {
+                let ctx = RequestContextBuilder::extend(ctx)
+                    .page_content_kind(PageContentKind::InMemoryLayer)
+                    .build();
+
+                let mut buf = Vec::new();
+
+                let cursor = inner.file.block_cursor();
+
+                for (key, vec_map) in inner.index.iter() {
+                    // Write all page versions
+                    for (lsn, pos) in vec_map.as_slice() {
+                        cursor.read_blob_into_buf(*pos, &mut buf, &ctx).await?;
+                        let will_init = Value::des(&buf)?.will_init();
+                        let res;
+                        (buf, res) = delta_layer_writer
+                            .put_value_bytes(Key::from_compact(*key), *lsn, buf, will_init, &ctx)
+                            .await;
+                        res?;
+                    }
+                }
+            }
            l0_flush::Inner::Direct { .. } => {
                let file_contents: Vec<u8> = inner.file.load_to_vec(ctx).await?;
                assert_eq!(
@@ -630,17 +620,11 @@ impl InMemoryLayer {
                        // => https://github.com/neondatabase/neon/issues/8183
                        cursor.read_blob_into_buf(*pos, &mut buf, ctx).await?;
                        let will_init = Value::des(&buf)?.will_init();
-                        let (tmp, res) = delta_layer_writer
-                            .put_value_bytes(
-                                Key::from_compact(*key),
-                                *lsn,
-                                buf.slice_len(),
-                                will_init,
-                                ctx,
-                            )
+                        let res;
+                        (buf, res) = delta_layer_writer
+                            .put_value_bytes(Key::from_compact(*key), *lsn, buf, will_init, ctx)
                            .await;
                        res?;
-                        buf = tmp.into_raw_slice().into_inner();
                    }
                }
            }
--- a/pageserver/src/tenant/storage_layer/layer.rs
+++ b/pageserver/src/tenant/storage_layer/layer.rs
@@ -312,9 +312,7 @@ impl Layer {
            .get_or_maybe_download(true, Some(ctx))
            .await
            .map_err(|err| match err {
-                DownloadError::TimelineShutdown | DownloadError::DownloadCancelled => {
-                    GetVectoredError::Cancelled
-                }
+                DownloadError::DownloadCancelled => GetVectoredError::Cancelled,
                other => GetVectoredError::Other(anyhow::anyhow!(other)),
            })?;

@@ -1296,10 +1294,7 @@ impl LayerInner {
                lsn_end: lsn_range.end,
                remote: !resident,
                access_stats,
-                l0: crate::tenant::layer_map::LayerMap::is_l0(
-                    &self.layer_desc().key_range,
-                    self.layer_desc().is_delta,
-                ),
+                l0: crate::tenant::layer_map::LayerMap::is_l0(&self.layer_desc().key_range),
            }
        } else {
            let lsn = self.desc.image_layer_lsn();
--- a/pageserver/src/tenant/storage_layer/layer_name.rs
+++ b/pageserver/src/tenant/storage_layer/layer_name.rs
@@ -256,10 +256,6 @@ impl LayerName {
            LayerName::Delta(layer) => &layer.key_range,
        }
    }
-
-    pub fn is_delta(&self) -> bool {
-        matches!(self, LayerName::Delta(_))
-    }
 }

 impl fmt::Display for LayerName {
--- a/pageserver/src/tenant/storage_layer/merge_iterator.rs
+++ b/pageserver/src/tenant/storage_layer/merge_iterator.rs
@@ -3,7 +3,6 @@ use std::{
    collections::{binary_heap, BinaryHeap},
 };

-use anyhow::bail;
 use pageserver_api::key::Key;
 use utils::lsn::Lsn;

@@ -27,13 +26,6 @@ impl<'a> LayerRef<'a> {
            Self::Delta(x) => LayerIterRef::Delta(x.iter(ctx)),
        }
    }
-
-    fn layer_dbg_info(&self) -> String {
-        match self {
-            Self::Image(x) => x.layer_dbg_info(),
-            Self::Delta(x) => x.layer_dbg_info(),
-        }
-    }
 }

 enum LayerIterRef<'a> {
@@ -48,13 +40,6 @@ impl LayerIterRef<'_> {
            Self::Image(x) => x.next().await,
        }
    }
-
-    fn layer_dbg_info(&self) -> String {
-        match self {
-            Self::Image(x) => x.layer_dbg_info(),
-            Self::Delta(x) => x.layer_dbg_info(),
-        }
-    }
 }

 /// This type plays several roles at once
@@ -90,11 +75,6 @@ impl<'a> PeekableLayerIterRef<'a> {
    async fn next(&mut self) -> anyhow::Result<Option<(Key, Lsn, Value)>> {
        let result = self.peeked.take();
        self.peeked = self.iter.next().await?;
-        if let (Some((k1, l1, _)), Some((k2, l2, _))) = (&self.peeked, &result) {
-            if (k1, l1) < (k2, l2) {
-                bail!("iterator is not ordered: {}", self.iter.layer_dbg_info());
-            }
-        }
        Ok(result)
    }
 }
@@ -198,12 +178,7 @@ impl<'a> IteratorWrapper<'a> {
        let iter = PeekableLayerIterRef::create(iter).await?;
        if let Some((k1, l1, _)) = iter.peek() {
            let (k2, l2) = first_key_lower_bound;
-            if (k1, l1) < (k2, l2) {
-                bail!(
-                    "layer key range did not include the first key in the layer: {}",
-                    layer.layer_dbg_info()
-                );
-            }
+            debug_assert!((k1, l1) >= (k2, l2));
        }
        *self = Self::Loaded { iter };
        Ok(())
--- a/pageserver/src/tenant/storage_layer/split_writer.rs
+++ b/pageserver/src/tenant/storage_layer/split_writer.rs
@@ -208,8 +208,6 @@ impl SplitDeltaLayerWriter {

 #[cfg(test)]
 mod tests {
-    use rand::{RngCore, SeedableRng};
-
    use crate::{
        tenant::{
            harness::{TenantHarness, TIMELINE_ID},
@@ -231,10 +229,7 @@ mod tests {
    }

    fn get_large_img() -> Bytes {
-        let mut rng = rand::rngs::SmallRng::seed_from_u64(42);
-        let mut data = vec![0; 8192];
-        rng.fill_bytes(&mut data);
-        data.into()
+        vec![0; 8192].into()
    }

    #[tokio::test]
--- a/pageserver/src/tenant/tasks.rs
+++ b/pageserver/src/tenant/tasks.rs
@@ -61,12 +61,21 @@ impl BackgroundLoopKind {
    }
 }

+static PERMIT_GAUGES: once_cell::sync::Lazy<
+    enum_map::EnumMap<BackgroundLoopKind, metrics::IntCounterPair>,
+> = once_cell::sync::Lazy::new(|| {
+    enum_map::EnumMap::from_array(std::array::from_fn(|i| {
+        let kind = <BackgroundLoopKind as enum_map::Enum>::from_usize(i);
+        crate::metrics::BACKGROUND_LOOP_SEMAPHORE_WAIT_GAUGE.with_label_values(&[kind.into()])
+    }))
+});
+
 /// Cancellation safe.
 pub(crate) async fn concurrent_background_tasks_rate_limit_permit(
    loop_kind: BackgroundLoopKind,
    _ctx: &RequestContext,
 ) -> tokio::sync::SemaphorePermit<'static> {
-    let _guard = crate::metrics::BACKGROUND_LOOP_SEMAPHORE.measure_acquisition(loop_kind);
+    let _guard = PERMIT_GAUGES[loop_kind].guard();

    pausable_failpoint!(
        "initial-size-calculation-permit-pause",
@@ -89,7 +98,7 @@ pub fn start_background_loops(
    task_mgr::spawn(
        BACKGROUND_RUNTIME.handle(),
        TaskKind::Compaction,
-        tenant_shard_id,
+        Some(tenant_shard_id),
        None,
        &format!("compactor for tenant {tenant_shard_id}"),
        {
@@ -112,7 +121,7 @@ pub fn start_background_loops(
    task_mgr::spawn(
        BACKGROUND_RUNTIME.handle(),
        TaskKind::GarbageCollector,
-        tenant_shard_id,
+        Some(tenant_shard_id),
        None,
        &format!("garbage collector for tenant {tenant_shard_id}"),
        {
@@ -135,7 +144,7 @@ pub fn start_background_loops(
    task_mgr::spawn(
        BACKGROUND_RUNTIME.handle(),
        TaskKind::IngestHousekeeping,
-        tenant_shard_id,
+        Some(tenant_shard_id),
        None,
        &format!("ingest housekeeping for tenant {tenant_shard_id}"),
        {
--- a/pageserver/src/tenant/timeline.rs
+++ b/pageserver/src/tenant/timeline.rs
@@ -22,8 +22,8 @@ use handle::ShardTimelineId;
 use once_cell::sync::Lazy;
 use pageserver_api::{
    key::{
-        CompactKey, KEY_SIZE, METADATA_KEY_BEGIN_PREFIX, METADATA_KEY_END_PREFIX,
-        NON_INHERITED_RANGE, NON_INHERITED_SPARSE_RANGE,
+        KEY_SIZE, METADATA_KEY_BEGIN_PREFIX, METADATA_KEY_END_PREFIX, NON_INHERITED_RANGE,
+        NON_INHERITED_SPARSE_RANGE,
    },
    keyspace::{KeySpaceAccum, KeySpaceRandomAccum, SparseKeyPartitioning},
    models::{
@@ -44,8 +44,10 @@ use tokio::{
 use tokio_util::sync::CancellationToken;
 use tracing::*;
 use utils::{
+    bin_ser::BeSer,
    fs_ext, pausable_failpoint,
    sync::gate::{Gate, GateGuard},
+    vec_map::VecMap,
 };

 use std::pin::pin;
@@ -135,10 +137,7 @@ use self::layer_manager::LayerManager;
 use self::logical_size::LogicalSize;
 use self::walreceiver::{WalReceiver, WalReceiverConf};

-use super::{
-    config::TenantConf, storage_layer::inmemory_layer, storage_layer::LayerVisibilityHint,
-    upload_queue::NotInitialized,
-};
+use super::{config::TenantConf, storage_layer::LayerVisibilityHint, upload_queue::NotInitialized};
 use super::{debug_assert_current_span_has_tenant_and_timeline_id, AttachedTenantConf};
 use super::{remote_timeline_client::index::IndexPart, storage_layer::LayerFringe};
 use super::{
@@ -512,7 +511,7 @@ pub(crate) struct TimelineVisitOutcome {
 #[derive(thiserror::Error, Debug)]
 pub(crate) enum PageReconstructError {
    #[error(transparent)]
-    Other(anyhow::Error),
+    Other(#[from] anyhow::Error),

    #[error("Ancestor LSN wait error: {0}")]
    AncestorLsnTimeout(WaitLsnError),
@@ -528,22 +527,6 @@ pub(crate) enum PageReconstructError {
    MissingKey(MissingKeyError),
 }

-impl From<anyhow::Error> for PageReconstructError {
-    fn from(value: anyhow::Error) -> Self {
-        // with walingest.rs many PageReconstructError are wrapped in as anyhow::Error
-        match value.downcast::<PageReconstructError>() {
-            Ok(pre) => pre,
-            Err(other) => PageReconstructError::Other(other),
-        }
-    }
-}
-
-impl From<utils::bin_ser::DeserializeError> for PageReconstructError {
-    fn from(value: utils::bin_ser::DeserializeError) -> Self {
-        PageReconstructError::Other(anyhow::Error::new(value).context("deserialization failure"))
-    }
-}
-
 impl From<layer_manager::Shutdown> for PageReconstructError {
    fn from(_: layer_manager::Shutdown) -> Self {
        PageReconstructError::Cancelled
@@ -563,7 +546,6 @@ impl From<layer_manager::Shutdown> for GetVectoredError {
    }
 }

-#[derive(thiserror::Error)]
 pub struct MissingKeyError {
    key: Key,
    shard: ShardNumber,
@@ -603,8 +585,11 @@ impl PageReconstructError {
    pub(crate) fn is_stopping(&self) -> bool {
        use PageReconstructError::*;
        match self {
+            Other(_) => false,
+            AncestorLsnTimeout(_) => false,
            Cancelled => true,
-            Other(_) | AncestorLsnTimeout(_) | WalRedo(_) | MissingKey(_) => false,
+            WalRedo(_) => false,
+            MissingKey { .. } => false,
        }
    }
 }
@@ -614,11 +599,11 @@ pub(crate) enum CreateImageLayersError {
    #[error("timeline shutting down")]
    Cancelled,

-    #[error("read failed")]
-    GetVectoredError(#[source] GetVectoredError),
+    #[error(transparent)]
+    GetVectoredError(GetVectoredError),

-    #[error("reconstruction failed")]
-    PageReconstructError(#[source] PageReconstructError),
+    #[error(transparent)]
+    PageReconstructError(PageReconstructError),

    #[error(transparent)]
    Other(#[from] anyhow::Error),
@@ -642,10 +627,10 @@ pub(crate) enum FlushLayerError {

    // Arc<> the following non-clonable error types: we must be Clone-able because the flush error is propagated from the flush
    // loop via a watch channel, where we can only borrow it.
-    #[error("create image layers (shared)")]
+    #[error(transparent)]
    CreateImageLayersError(Arc<CreateImageLayersError>),

-    #[error("other (shared)")]
+    #[error(transparent)]
    Other(#[from] Arc<anyhow::Error>),
 }

@@ -678,46 +663,34 @@ pub(crate) enum GetVectoredError {
    #[error("timeline shutting down")]
    Cancelled,

-    #[error("requested too many keys: {0} > {}", Timeline::MAX_GET_VECTORED_KEYS)]
+    #[error("Requested too many keys: {0} > {}", Timeline::MAX_GET_VECTORED_KEYS)]
    Oversized(u64),

-    #[error("requested at invalid LSN: {0}")]
+    #[error("Requested at invalid LSN: {0}")]
    InvalidLsn(Lsn),

-    #[error("requested key not found: {0}")]
+    #[error("Requested key not found: {0}")]
    MissingKey(MissingKeyError),

-    #[error("ancestry walk")]
-    GetReadyAncestorError(#[source] GetReadyAncestorError),
+    #[error(transparent)]
+    GetReadyAncestorError(GetReadyAncestorError),

    #[error(transparent)]
    Other(#[from] anyhow::Error),
 }

-impl From<GetReadyAncestorError> for GetVectoredError {
-    fn from(value: GetReadyAncestorError) -> Self {
-        use GetReadyAncestorError::*;
-        match value {
-            Cancelled => GetVectoredError::Cancelled,
-            AncestorLsnTimeout(_) | BadState { .. } => {
-                GetVectoredError::GetReadyAncestorError(value)
-            }
-        }
-    }
-}
-
 #[derive(thiserror::Error, Debug)]
 pub(crate) enum GetReadyAncestorError {
-    #[error("ancestor LSN wait error")]
+    #[error("Ancestor LSN wait error: {0}")]
    AncestorLsnTimeout(#[from] WaitLsnError),

-    #[error("bad state on timeline {timeline_id}: {state:?}")]
+    #[error("Bad state on timeline {timeline_id}: {state:?}")]
    BadState {
        timeline_id: TimelineId,
        state: TimelineState,
    },

-    #[error("cancelled")]
+    #[error("Cancelled")]
    Cancelled,
 }

@@ -1646,20 +1619,6 @@ impl Timeline {
        self.last_record_lsn.shutdown();

        if try_freeze_and_flush {
-            if let Some((open, frozen)) = self
-                .layers
-                .read()
-                .await
-                .layer_map()
-                .map(|lm| (lm.open_layer.is_some(), lm.frozen_layers.len()))
-                .ok()
-                .filter(|(open, frozen)| *open || *frozen > 0)
-            {
-                tracing::info!(?open, frozen, "flushing and freezing on shutdown");
-            } else {
-                // this is double-shutdown, ignore it
-            }
-
            // we shut down walreceiver above, so, we won't add anything more
            // to the InMemoryLayer; freeze it and wait for all frozen layers
            // to reach the disk & upload queue, then shut the upload queue and
@@ -2234,11 +2193,6 @@ impl Timeline {

                handles: Default::default(),
            };
-
-            if aux_file_policy == Some(AuxFilePolicy::V1) {
-                warn!("this timeline is using deprecated aux file policy V1");
-            }
-
            result.repartition_threshold =
                result.get_checkpoint_distance() / REPARTITION_FREQ_IN_CHECKPOINT_DISTANCE;

@@ -2287,7 +2241,7 @@ impl Timeline {
        task_mgr::spawn(
            task_mgr::BACKGROUND_RUNTIME.handle(),
            task_mgr::TaskKind::LayerFlushTask,
-            self.tenant_shard_id,
+            Some(self.tenant_shard_id),
            Some(self.timeline_id),
            "layer flush task",
            async move {
@@ -2641,7 +2595,7 @@ impl Timeline {
        task_mgr::spawn(
            task_mgr::BACKGROUND_RUNTIME.handle(),
            task_mgr::TaskKind::InitialLogicalSizeCalculation,
-            self.tenant_shard_id,
+            Some(self.tenant_shard_id),
            Some(self.timeline_id),
            "initial size calculation",
            // NB: don't log errors here, task_mgr will do that.
@@ -2809,7 +2763,7 @@ impl Timeline {
        task_mgr::spawn(
            task_mgr::BACKGROUND_RUNTIME.handle(),
            task_mgr::TaskKind::OndemandLogicalSizeCalculation,
-            self.tenant_shard_id,
+            Some(self.tenant_shard_id),
            Some(self.timeline_id),
            "ondemand logical size calculation",
            async move {
@@ -2983,7 +2937,11 @@ impl Timeline {
                LayerVisibilityHint::Visible => {
                    // Layer is visible to one or more read LSNs: elegible for inclusion in layer map
                    let last_activity_ts = layer.latest_activity();
-                    Some((layer.layer_desc(), layer.metadata(), last_activity_ts))
+                    Some(HeatMapLayer::new(
+                        layer.layer_desc().layer_name(),
+                        layer.metadata(),
+                        last_activity_ts,
+                    ))
                }
                LayerVisibilityHint::Covered => {
                    // Layer is resident but unlikely to be read: not elegible for inclusion in heatmap.
@@ -2992,26 +2950,7 @@ impl Timeline {
            }
        });

-        let mut layers = resident.collect::<Vec<_>>();
-
-        // Sort layers in order of which to download first.  For a large set of layers to download, we
-        // want to prioritize those layers which are most likely to still be in the resident many minutes
-        // or hours later:
-        // - Download L0s last, because they churn the fastest: L0s on a fast-writing tenant might
-        //   only exist for a few minutes before being compacted into L1s.
-        // - For L1 & image layers, download most recent LSNs first: the older the LSN, the sooner
-        //   the layer is likely to be covered by an image layer during compaction.
-        layers.sort_by_key(|(desc, _meta, _atime)| {
-            std::cmp::Reverse((
-                !LayerMap::is_l0(&desc.key_range, desc.is_delta),
-                desc.lsn_range.end,
-            ))
-        });
-
-        let layers = layers
-            .into_iter()
-            .map(|(desc, meta, atime)| HeatMapLayer::new(desc.layer_name(), meta, atime))
-            .collect();
+        let layers = resident.collect();

        Some(HeatMapTimeline::new(self.timeline_id, layers))
    }
@@ -3107,7 +3046,8 @@ impl Timeline {
            cont_lsn = std::cmp::min(Lsn(request_lsn.0 + 1), Lsn(timeline.ancestor_lsn.0 + 1));
            timeline_owned = timeline
                .get_ready_ancestor_timeline(ancestor_timeline, ctx)
-                .await?;
+                .await
+                .map_err(GetVectoredError::GetReadyAncestorError)?;
            timeline = &*timeline_owned;
        };

@@ -3598,6 +3538,34 @@ impl Timeline {
                return Err(FlushLayerError::Cancelled);
            }

+            // FIXME(auxfilesv2): support multiple metadata key partitions might need initdb support as well?
+            // This code path will not be hit during regression tests. After #7099 we have a single partition
+            // with two key ranges. If someone wants to fix initdb optimization in the future, this might need
+            // to be fixed.
+
+            // For metadata, always create delta layers.
+            let delta_layer = if !metadata_partition.parts.is_empty() {
+                assert_eq!(
+                    metadata_partition.parts.len(),
+                    1,
+                    "currently sparse keyspace should only contain a single metadata keyspace"
+                );
+                let metadata_keyspace = &metadata_partition.parts[0];
+                self.create_delta_layer(
+                    &frozen_layer,
+                    Some(
+                        metadata_keyspace.0.ranges.first().unwrap().start
+                            ..metadata_keyspace.0.ranges.last().unwrap().end,
+                    ),
+                    ctx,
+                )
+                .await
+                .map_err(|e| FlushLayerError::from_anyhow(self, e))?
+            } else {
+                None
+            };
+
+            // For image layers, we add them immediately into the layer map.
            let mut layers_to_upload = Vec::new();
            layers_to_upload.extend(
                self.create_image_layers(
@@ -3608,27 +3576,13 @@ impl Timeline {
                )
                .await?,
            );
-            if !metadata_partition.parts.is_empty() {
-                assert_eq!(
-                    metadata_partition.parts.len(),
-                    1,
-                    "currently sparse keyspace should only contain a single metadata keyspace"
-                );
-                layers_to_upload.extend(
-                    self.create_image_layers(
-                        // Safety: create_image_layers treat sparse keyspaces differently that it does not scan
-                        // every single key within the keyspace, and therefore, it's safe to force converting it
-                        // into a dense keyspace before calling this function.
-                        &metadata_partition.into_dense(),
-                        self.initdb_lsn,
-                        ImageLayerCreationMode::Initial,
-                        ctx,
-                    )
-                    .await?,
-                );
-            }

-            (layers_to_upload, None)
+            if let Some(delta_layer) = delta_layer {
+                layers_to_upload.push(delta_layer.clone());
+                (layers_to_upload, Some(delta_layer))
+            } else {
+                (layers_to_upload, None)
+            }
        } else {
            // Normal case, write out a L0 delta layer file.
            // `create_delta_layer` will not modify the layer map.
@@ -3990,7 +3944,7 @@ impl Timeline {
                                    warn!("could not reconstruct FSM or VM key {img_key}, filling with zeros: {err:?}");
                                    ZERO_PAGE.clone()
                                } else {
-                                    return Err(CreateImageLayersError::from(err));
+                                    return Err(CreateImageLayersError::PageReconstructError(err));
                                }
                            }
                        };
@@ -4038,6 +3992,8 @@ impl Timeline {
        mode: ImageLayerCreationMode,
        start: Key,
    ) -> Result<ImageLayerCreationOutcome, CreateImageLayersError> {
+        assert!(!matches!(mode, ImageLayerCreationMode::Initial));
+
        // Metadata keys image layer creation.
        let mut reconstruct_state = ValuesReconstructState::default();
        let data = self
@@ -4048,7 +4004,7 @@ impl Timeline {
            let mut total_kb_retrieved = 0;
            let mut total_keys_retrieved = 0;
            for (k, v) in data {
-                let v = v?;
+                let v = v.map_err(CreateImageLayersError::PageReconstructError)?;
                total_kb_retrieved += KEY_SIZE + v.len();
                total_keys_retrieved += 1;
                new_data.insert(k, v);
@@ -4203,13 +4159,15 @@ impl Timeline {
                        "metadata keys must be partitioned separately"
                    );
                }
+                if mode == ImageLayerCreationMode::Initial {
+                    return Err(CreateImageLayersError::Other(anyhow::anyhow!("no image layer should be created for metadata keys when flushing frozen layers")));
+                }
                if mode == ImageLayerCreationMode::Try && !check_for_image_layers {
                    // Skip compaction if there are not enough updates. Metadata compaction will do a scan and
                    // might mess up with evictions.
                    start = img_range.end;
                    continue;
                }
-                // For initial and force modes, we always generate image layers for metadata keys.
            } else if let ImageLayerCreationMode::Try = mode {
                // check_for_image_layers = false -> skip
                // check_for_image_layers = true -> check time_for_new_image_layer -> skip/generate
@@ -4217,8 +4175,7 @@ impl Timeline {
                    start = img_range.end;
                    continue;
                }
-            }
-            if let ImageLayerCreationMode::Force = mode {
+            } else if let ImageLayerCreationMode::Force = mode {
                // When forced to create image layers, we might try and create them where they already
                // exist.  This mode is only used in tests/debug.
                let layers = self.layers.read().await;
@@ -4232,7 +4189,6 @@ impl Timeline {
                        img_range.start,
                        img_range.end
                    );
-                    start = img_range.end;
                    continue;
                }
            }
@@ -4521,7 +4477,6 @@ impl DurationRecorder {
 /// the layer descriptor requires the user to provide the ranges, which should cover all
 /// keys specified in the `data` field.
 #[cfg(test)]
-#[derive(Clone)]
 pub struct DeltaLayerTestDesc {
    pub lsn_range: Range<Lsn>,
    pub key_range: Range<Key>,
@@ -4551,13 +4506,6 @@ impl DeltaLayerTestDesc {
            data,
        }
    }
-
-    pub(crate) fn layer_name(&self) -> LayerName {
-        LayerName::Delta(super::storage_layer::DeltaLayerName {
-            key_range: self.key_range.clone(),
-            lsn_range: self.lsn_range.clone(),
-        })
-    }
 }

 impl Timeline {
@@ -4567,12 +4515,7 @@ impl Timeline {
        new_images: &[ResidentLayer],
        layers_to_remove: &[Layer],
    ) -> Result<(), CompactionError> {
-        let mut guard = tokio::select! {
-            guard = self.layers.write() => guard,
-            _ = self.cancel.cancelled() => {
-                return Err(CompactionError::ShuttingDown);
-            }
-        };
+        let mut guard = self.layers.write().await;

        let mut duplicated_layers = HashSet::new();

@@ -4588,7 +4531,7 @@ impl Timeline {
                // for compact_level0_phase1 creating an L0, which does not happen in practice
                // because we have not implemented L0 => L0 compaction.
                duplicated_layers.insert(l.layer_desc().key());
-            } else if LayerMap::is_l0(&l.layer_desc().key_range, l.layer_desc().is_delta) {
+            } else if LayerMap::is_l0(&l.layer_desc().key_range) {
                return Err(CompactionError::Other(anyhow::anyhow!("compaction generates a L0 layer file as output, which will cause infinite compaction.")));
            } else {
                insert_layers.push(l.clone());
@@ -5155,7 +5098,7 @@ impl Timeline {
        let task_id = task_mgr::spawn(
            task_mgr::BACKGROUND_RUNTIME.handle(),
            task_mgr::TaskKind::DownloadAllRemoteLayers,
-            self.tenant_shard_id,
+            Some(self.tenant_shard_id),
            Some(self.timeline_id),
            "download all remote layers task",
            async move {
@@ -5318,7 +5261,6 @@ impl Timeline {
                    layer: layer.to_owned().into(),
                    last_activity_ts,
                    relative_last_activity: finite_f32::FiniteF32::ZERO,
-                    visibility: layer.visibility(),
                }
            })
            .collect();
@@ -5583,6 +5525,44 @@ enum OpenLayerAction {
 }

 impl<'a> TimelineWriter<'a> {
+    /// Put a new page version that can be constructed from a WAL record
+    ///
+    /// This will implicitly extend the relation, if the page is beyond the
+    /// current end-of-file.
+    pub(crate) async fn put(
+        &mut self,
+        key: Key,
+        lsn: Lsn,
+        value: &Value,
+        ctx: &RequestContext,
+    ) -> anyhow::Result<()> {
+        // Avoid doing allocations for "small" values.
+        // In the regression test suite, the limit of 256 avoided allocations in 95% of cases:
+        // https://github.com/neondatabase/neon/pull/5056#discussion_r1301975061
+        let mut buf = smallvec::SmallVec::<[u8; 256]>::new();
+        value.ser_into(&mut buf)?;
+        let buf_size: u64 = buf.len().try_into().expect("oversized value buf");
+
+        let action = self.get_open_layer_action(lsn, buf_size);
+        let layer = self.handle_open_layer_action(lsn, action, ctx).await?;
+        let res = layer.put_value(key.to_compact(), lsn, &buf, ctx).await;
+
+        if res.is_ok() {
+            // Update the current size only when the entire write was ok.
+            // In case of failures, we may have had partial writes which
+            // render the size tracking out of sync. That's ok because
+            // the checkpoint distance should be significantly smaller
+            // than the S3 single shot upload limit of 5GiB.
+            let state = self.write_guard.as_mut().unwrap();
+
+            state.current_size += buf_size;
+            state.prev_lsn = Some(lsn);
+            state.max_lsn = std::cmp::max(state.max_lsn, Some(lsn));
+        }
+
+        res
+    }
+
    async fn handle_open_layer_action(
        &mut self,
        at: Lsn,
@@ -5688,58 +5668,18 @@ impl<'a> TimelineWriter<'a> {
    }

    /// Put a batch of keys at the specified Lsns.
+    ///
+    /// The batch is sorted by Lsn (enforced by usage of [`utils::vec_map::VecMap`].
    pub(crate) async fn put_batch(
        &mut self,
-        batch: Vec<(CompactKey, Lsn, usize, Value)>,
+        batch: VecMap<Lsn, (Key, Value)>,
        ctx: &RequestContext,
    ) -> anyhow::Result<()> {
-        if batch.is_empty() {
-            return Ok(());
+        for (lsn, (key, val)) in batch {
+            self.put(key, lsn, &val, ctx).await?
        }

-        let serialized_batch = inmemory_layer::SerializedBatch::from_values(batch);
-        let batch_max_lsn = serialized_batch.max_lsn;
-        let buf_size: u64 = serialized_batch.raw.len() as u64;
-
-        let action = self.get_open_layer_action(batch_max_lsn, buf_size);
-        let layer = self
-            .handle_open_layer_action(batch_max_lsn, action, ctx)
-            .await?;
-
-        let res = layer.put_batch(serialized_batch, ctx).await;
-
-        if res.is_ok() {
-            // Update the current size only when the entire write was ok.
-            // In case of failures, we may have had partial writes which
-            // render the size tracking out of sync. That's ok because
-            // the checkpoint distance should be significantly smaller
-            // than the S3 single shot upload limit of 5GiB.
-            let state = self.write_guard.as_mut().unwrap();
-
-            state.current_size += buf_size;
-            state.prev_lsn = Some(batch_max_lsn);
-            state.max_lsn = std::cmp::max(state.max_lsn, Some(batch_max_lsn));
-        }
-
-        res
-    }
-
-    #[cfg(test)]
-    /// Test helper, for tests that would like to poke individual values without composing a batch
-    pub(crate) async fn put(
-        &mut self,
-        key: Key,
-        lsn: Lsn,
-        value: &Value,
-        ctx: &RequestContext,
-    ) -> anyhow::Result<()> {
-        use utils::bin_ser::BeSer;
-        let val_ser_size = value.serialized_size().unwrap() as usize;
-        self.put_batch(
-            vec![(key.to_compact(), lsn, val_ser_size, value.clone())],
-            ctx,
-        )
-        .await
+        Ok(())
    }

    pub(crate) async fn delete_batch(
@@ -5783,110 +5723,12 @@ fn is_send() {

 #[cfg(test)]
 mod tests {
-    use pageserver_api::key::Key;
    use utils::{id::TimelineId, lsn::Lsn};

-    use crate::{
-        repository::Value,
-        tenant::{
-            harness::{test_img, TenantHarness},
-            layer_map::LayerMap,
-            storage_layer::{Layer, LayerName},
-            timeline::{DeltaLayerTestDesc, EvictionError},
-            Timeline,
-        },
+    use crate::tenant::{
+        harness::TenantHarness, storage_layer::Layer, timeline::EvictionError, Timeline,
    };

-    #[tokio::test]
-    async fn test_heatmap_generation() {
-        let harness = TenantHarness::create("heatmap_generation").await.unwrap();
-
-        let covered_delta = DeltaLayerTestDesc::new_with_inferred_key_range(
-            Lsn(0x10)..Lsn(0x20),
-            vec![(
-                Key::from_hex("620000000033333333444444445500000000").unwrap(),
-                Lsn(0x11),
-                Value::Image(test_img("foo")),
-            )],
-        );
-        let visible_delta = DeltaLayerTestDesc::new_with_inferred_key_range(
-            Lsn(0x10)..Lsn(0x20),
-            vec![(
-                Key::from_hex("720000000033333333444444445500000000").unwrap(),
-                Lsn(0x11),
-                Value::Image(test_img("foo")),
-            )],
-        );
-        let l0_delta = DeltaLayerTestDesc::new(
-            Lsn(0x20)..Lsn(0x30),
-            Key::from_hex("000000000000000000000000000000000000").unwrap()
-                ..Key::from_hex("FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF").unwrap(),
-            vec![(
-                Key::from_hex("720000000033333333444444445500000000").unwrap(),
-                Lsn(0x25),
-                Value::Image(test_img("foo")),
-            )],
-        );
-        let delta_layers = vec![
-            covered_delta.clone(),
-            visible_delta.clone(),
-            l0_delta.clone(),
-        ];
-
-        let image_layer = (
-            Lsn(0x40),
-            vec![(
-                Key::from_hex("620000000033333333444444445500000000").unwrap(),
-                test_img("bar"),
-            )],
-        );
-        let image_layers = vec![image_layer];
-
-        let (tenant, ctx) = harness.load().await;
-        let timeline = tenant
-            .create_test_timeline_with_layers(
-                TimelineId::generate(),
-                Lsn(0x10),
-                14,
-                &ctx,
-                delta_layers,
-                image_layers,
-                Lsn(0x100),
-            )
-            .await
-            .unwrap();
-
-        // Layer visibility is an input to heatmap generation, so refresh it first
-        timeline.update_layer_visibility().await.unwrap();
-
-        let heatmap = timeline
-            .generate_heatmap()
-            .await
-            .expect("Infallible while timeline is not shut down");
-
-        assert_eq!(heatmap.timeline_id, timeline.timeline_id);
-
-        // L0 should come last
-        assert_eq!(heatmap.layers.last().unwrap().name, l0_delta.layer_name());
-
-        let mut last_lsn = Lsn::MAX;
-        for layer in heatmap.layers {
-            // Covered layer should be omitted
-            assert!(layer.name != covered_delta.layer_name());
-
-            let layer_lsn = match &layer.name {
-                LayerName::Delta(d) => d.lsn_range.end,
-                LayerName::Image(i) => i.lsn,
-            };
-
-            // Apart from L0s, newest Layers should come first
-            if !LayerMap::is_l0(layer.name.key_range(), layer.name.is_delta()) {
-                assert!(layer_lsn <= last_lsn);
-                last_lsn = layer_lsn;
-            }
-        }
-    }
-
    #[tokio::test]
    async fn two_layer_eviction_attempts_at_the_same_time() {
        let harness = TenantHarness::create("two_layer_eviction_attempts_at_the_same_time")
--- a/pageserver/src/tenant/timeline/compaction.rs
+++ b/pageserver/src/tenant/timeline/compaction.rs
@@ -1048,22 +1048,11 @@ impl Timeline {
        let mut dup_end_lsn: Lsn = Lsn::INVALID; // end LSN of layer containing values of the single key
        let mut next_hole = 0; // index of next hole in holes vector

-        let mut keys = 0;
-
        while let Some((key, lsn, value)) = all_values_iter
            .next(ctx)
            .await
            .map_err(CompactionError::Other)?
        {
-            keys += 1;
-
-            if keys % 32_768 == 0 && self.cancel.is_cancelled() {
-                // avoid hitting the cancellation token on every key. in benches, we end up
-                // shuffling an order of million keys per layer, this means we'll check it
-                // around tens of times per layer.
-                return Err(CompactionError::ShuttingDown);
-            }
-
            let same_key = prev_key.map_or(false, |prev_key| prev_key == key);
            // We need to check key boundaries once we reach next key or end of layer with the same key
            if !same_key || lsn == dup_end_lsn {
@@ -1168,8 +1157,6 @@ impl Timeline {
                        .await
                        .map_err(CompactionError::Other)?,
                    );
-
-                    keys = 0;
                }

                writer
@@ -2338,7 +2325,7 @@ impl CompactionJobExecutor for TimelineAdaptor {
                key_range,
            ))
        } else {
-            // The current compaction implementation only ever requests the key space
+            // The current compaction implementatin only ever requests the key space
            // at the compaction end LSN.
            anyhow::bail!("keyspace not available for requested lsn");
        }
--- a/pageserver/src/tenant/timeline/delete.rs
+++ b/pageserver/src/tenant/timeline/delete.rs
@@ -395,7 +395,7 @@ impl DeleteTimelineFlow {
        task_mgr::spawn(
            task_mgr::BACKGROUND_RUNTIME.handle(),
            TaskKind::TimelineDeletionWorker,
-            tenant_shard_id,
+            Some(tenant_shard_id),
            Some(timeline_id),
            "timeline_delete",
            async move {
--- a/pageserver/src/tenant/timeline/eviction_task.rs
+++ b/pageserver/src/tenant/timeline/eviction_task.rs
@@ -30,8 +30,7 @@ use crate::{
    pgdatadir_mapping::CollectKeySpaceError,
    task_mgr::{self, TaskKind, BACKGROUND_RUNTIME},
    tenant::{
-        storage_layer::LayerVisibilityHint, tasks::BackgroundLoopKind, timeline::EvictionError,
-        LogicalSizeCalculationCause, Tenant,
+        tasks::BackgroundLoopKind, timeline::EvictionError, LogicalSizeCalculationCause, Tenant,
    },
 };

@@ -60,7 +59,7 @@ impl Timeline {
        task_mgr::spawn(
            BACKGROUND_RUNTIME.handle(),
            TaskKind::Eviction,
-            self.tenant_shard_id,
+            Some(self.tenant_shard_id),
            Some(self.timeline_id),
            &format!(
                "layer eviction for {}/{}",
@@ -242,22 +241,7 @@ impl Timeline {
                        }
                    };

-                    match layer.visibility() {
-                        LayerVisibilityHint::Visible => {
-                            // Usual case: a visible layer might be read any time, and we will keep it
-                            // resident until it hits our configured TTL threshold.
-                            no_activity_for > p.threshold
-                        }
-                        LayerVisibilityHint::Covered => {
-                            // Covered layers: this is probably a layer that was recently covered by
-                            // an image layer during compaction.  We don't evict it immediately, but
-                            // it doesn't stay resident for the full `threshold`: we just keep it
-                            // for a shorter time in case
-                            // - it is used for Timestamp->LSN lookups
-                            // - a new branch is created in recent history which will read this layer
-                            no_activity_for > p.period
-                        }
-                    }
+                    no_activity_for > p.threshold
                })
                .cloned()
                .for_each(|layer| {
--- a/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs
+++ b/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs
@@ -27,8 +27,8 @@ use super::TaskStateUpdate;
 use crate::{
    context::RequestContext,
    metrics::{LIVE_CONNECTIONS, WALRECEIVER_STARTED_CONNECTIONS, WAL_INGEST},
-    pgdatadir_mapping::DatadirModification,
-    task_mgr::{TaskKind, WALRECEIVER_RUNTIME},
+    task_mgr::TaskKind,
+    task_mgr::WALRECEIVER_RUNTIME,
    tenant::{debug_assert_current_span_has_tenant_and_timeline_id, Timeline, WalReceiverInfo},
    walingest::WalIngest,
    walrecord::DecodedWALRecord,
@@ -345,10 +345,7 @@ pub(super) async fn handle_walreceiver_connection(
                        // Commit every ingest_batch_size records. Even if we filtered out
                        // all records, we still need to call commit to advance the LSN.
                        uncommitted_records += 1;
-                        if uncommitted_records >= ingest_batch_size
-                            || modification.approx_pending_bytes()
-                                > DatadirModification::MAX_PENDING_BYTES
-                        {
+                        if uncommitted_records >= ingest_batch_size {
                            WAL_INGEST
                                .records_committed
                                .inc_by(uncommitted_records - filtered_records);
--- a/pageserver/src/tenant/vectored_blob_io.rs
+++ b/pageserver/src/tenant/vectored_blob_io.rs
@@ -25,10 +25,9 @@ use tokio_epoll_uring::BoundedBuf;
 use utils::lsn::Lsn;
 use utils::vec_map::VecMap;

-use crate::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
 use crate::context::RequestContext;
 use crate::tenant::blob_io::{BYTE_UNCOMPRESSED, BYTE_ZSTD, LEN_COMPRESSION_BIT_MASK};
-use crate::virtual_file::{self, VirtualFile};
+use crate::virtual_file::VirtualFile;

 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
 pub struct MaxVectoredReadBytes(pub NonZeroUsize);
@@ -61,7 +60,7 @@ pub struct VectoredBlobsBuf {
 pub struct VectoredRead {
    pub start: u64,
    pub end: u64,
-    /// Start offset and metadata for each blob in this read
+    /// Starting offsets and metadata for each blob in this read
    pub blobs_at: VecMap<u64, BlobMeta>,
 }

@@ -77,109 +76,14 @@ pub(crate) enum VectoredReadExtended {
    No,
 }

-#[derive(Copy, Clone, Debug, PartialEq, Eq)]
-pub enum VectoredReadCoalesceMode {
-    /// Only coalesce exactly adjacent reads.
-    AdjacentOnly,
-    /// In addition to adjacent reads, also consider reads whose corresponding
-    /// `end` and `start` offsets reside at the same chunk.
-    Chunked(usize),
-}
-
-impl VectoredReadCoalesceMode {
-    /// [`AdjacentVectoredReadBuilder`] is used if alignment requirement is 0,
-    /// whereas [`ChunkedVectoredReadBuilder`] is used for alignment requirement 1 and higher.
-    pub(crate) fn get() -> Self {
-        let align = virtual_file::get_io_buffer_alignment_raw();
-        if align == DEFAULT_IO_BUFFER_ALIGNMENT {
-            VectoredReadCoalesceMode::AdjacentOnly
-        } else {
-            VectoredReadCoalesceMode::Chunked(align)
-        }
-    }
-}
-
-pub(crate) enum VectoredReadBuilder {
-    Adjacent(AdjacentVectoredReadBuilder),
-    Chunked(ChunkedVectoredReadBuilder),
-}
-
-impl VectoredReadBuilder {
-    fn new_impl(
-        start_offset: u64,
-        end_offset: u64,
-        meta: BlobMeta,
-        max_read_size: Option<usize>,
-        mode: VectoredReadCoalesceMode,
-    ) -> Self {
-        match mode {
-            VectoredReadCoalesceMode::AdjacentOnly => Self::Adjacent(
-                AdjacentVectoredReadBuilder::new(start_offset, end_offset, meta, max_read_size),
-            ),
-            VectoredReadCoalesceMode::Chunked(chunk_size) => {
-                Self::Chunked(ChunkedVectoredReadBuilder::new(
-                    start_offset,
-                    end_offset,
-                    meta,
-                    max_read_size,
-                    chunk_size,
-                ))
-            }
-        }
-    }
-
-    pub(crate) fn new(
-        start_offset: u64,
-        end_offset: u64,
-        meta: BlobMeta,
-        max_read_size: usize,
-        mode: VectoredReadCoalesceMode,
-    ) -> Self {
-        Self::new_impl(start_offset, end_offset, meta, Some(max_read_size), mode)
-    }
-
-    pub(crate) fn new_streaming(
-        start_offset: u64,
-        end_offset: u64,
-        meta: BlobMeta,
-        mode: VectoredReadCoalesceMode,
-    ) -> Self {
-        Self::new_impl(start_offset, end_offset, meta, None, mode)
-    }
-
-    pub(crate) fn extend(&mut self, start: u64, end: u64, meta: BlobMeta) -> VectoredReadExtended {
-        match self {
-            VectoredReadBuilder::Adjacent(builder) => builder.extend(start, end, meta),
-            VectoredReadBuilder::Chunked(builder) => builder.extend(start, end, meta),
-        }
-    }
-
-    pub(crate) fn build(self) -> VectoredRead {
-        match self {
-            VectoredReadBuilder::Adjacent(builder) => builder.build(),
-            VectoredReadBuilder::Chunked(builder) => builder.build(),
-        }
-    }
-
-    pub(crate) fn size(&self) -> usize {
-        match self {
-            VectoredReadBuilder::Adjacent(builder) => builder.size(),
-            VectoredReadBuilder::Chunked(builder) => builder.size(),
-        }
-    }
-}
-
-pub(crate) struct AdjacentVectoredReadBuilder {
-    /// Start offset of the read.
+pub(crate) struct VectoredReadBuilder {
    start: u64,
-    // End offset of the read.
    end: u64,
-    /// Start offset and metadata for each blob in this read
    blobs_at: VecMap<u64, BlobMeta>,
    max_read_size: Option<usize>,
 }

-impl AdjacentVectoredReadBuilder {
+impl VectoredReadBuilder {
    /// Start building a new vectored read.
    ///
    /// Note that by design, this does not check against reading more than `max_read_size` to
@@ -189,7 +93,7 @@ impl AdjacentVectoredReadBuilder {
        start_offset: u64,
        end_offset: u64,
        meta: BlobMeta,
-        max_read_size: Option<usize>,
+        max_read_size: usize,
    ) -> Self {
        let mut blobs_at = VecMap::default();
        blobs_at
@@ -200,7 +104,7 @@ impl AdjacentVectoredReadBuilder {
            start: start_offset,
            end: end_offset,
            blobs_at,
-            max_read_size,
+            max_read_size: Some(max_read_size),
        }
    }
    /// Attempt to extend the current read with a new blob if the start
@@ -209,15 +113,13 @@ impl AdjacentVectoredReadBuilder {
    pub(crate) fn extend(&mut self, start: u64, end: u64, meta: BlobMeta) -> VectoredReadExtended {
        tracing::trace!(start, end, "trying to extend");
        let size = (end - start) as usize;
-        let not_limited_by_max_read_size = {
+        if self.end == start && {
            if let Some(max_read_size) = self.max_read_size {
                self.size() + size <= max_read_size
            } else {
                true
            }
-        };
-
-        if self.end == start && not_limited_by_max_read_size {
+        } {
            self.end = end;
            self.blobs_at
                .append(start, meta)
@@ -242,107 +144,6 @@ impl AdjacentVectoredReadBuilder {
    }
 }

-pub(crate) struct ChunkedVectoredReadBuilder {
-    /// Start block number
-    start_blk_no: usize,
-    /// End block number (exclusive).
-    end_blk_no: usize,
-    /// Start offset and metadata for each blob in this read
-    blobs_at: VecMap<u64, BlobMeta>,
-    max_read_size: Option<usize>,
-    /// Chunk size reads are coalesced into.
-    chunk_size: usize,
-}
-
-/// Computes x / d rounded up.
-fn div_round_up(x: usize, d: usize) -> usize {
-    (x + (d - 1)) / d
-}
-
-impl ChunkedVectoredReadBuilder {
-    /// Start building a new vectored read.
-    ///
-    /// Note that by design, this does not check against reading more than `max_read_size` to
-    /// support reading larger blobs than the configuration value. The builder will be single use
-    /// however after that.
-    pub(crate) fn new(
-        start_offset: u64,
-        end_offset: u64,
-        meta: BlobMeta,
-        max_read_size: Option<usize>,
-        chunk_size: usize,
-    ) -> Self {
-        let mut blobs_at = VecMap::default();
-        blobs_at
-            .append(start_offset, meta)
-            .expect("First insertion always succeeds");
-
-        let start_blk_no = start_offset as usize / chunk_size;
-        let end_blk_no = div_round_up(end_offset as usize, chunk_size);
-        Self {
-            start_blk_no,
-            end_blk_no,
-            blobs_at,
-            max_read_size,
-            chunk_size,
-        }
-    }
-
-    /// Attempts to extend the current read with a new blob if the new blob resides in the same or the immediate next chunk.
-    ///
-    /// The resulting size also must be below the max read size.
-    pub(crate) fn extend(&mut self, start: u64, end: u64, meta: BlobMeta) -> VectoredReadExtended {
-        tracing::trace!(start, end, "trying to extend");
-        let start_blk_no = start as usize / self.chunk_size;
-        let end_blk_no = div_round_up(end as usize, self.chunk_size);
-
-        let not_limited_by_max_read_size = {
-            if let Some(max_read_size) = self.max_read_size {
-                let coalesced_size = (end_blk_no - self.start_blk_no) * self.chunk_size;
-                coalesced_size <= max_read_size
-            } else {
-                true
-            }
-        };
-
-        // True if the second block starts in the same block or the immediate next block where the first block ended.
-        //
-        // Note: This automatically handles the case where two blocks are adjacent to each other,
-        // whether they starts on chunk size boundary or not.
-        let is_adjacent_chunk_read = {
-            // 1. first.end & second.start are in the same block
-            self.end_blk_no == start_blk_no + 1 ||
-            // 2. first.end ends one block before second.start
-            self.end_blk_no == start_blk_no
-        };
-
-        if is_adjacent_chunk_read && not_limited_by_max_read_size {
-            self.end_blk_no = end_blk_no;
-            self.blobs_at
-                .append(start, meta)
-                .expect("LSNs are ordered within vectored reads");
-
-            return VectoredReadExtended::Yes;
-        }
-
-        VectoredReadExtended::No
-    }
-
-    pub(crate) fn size(&self) -> usize {
-        (self.end_blk_no - self.start_blk_no) * self.chunk_size
-    }
-
-    pub(crate) fn build(self) -> VectoredRead {
-        let start = (self.start_blk_no * self.chunk_size) as u64;
-        let end = (self.end_blk_no * self.chunk_size) as u64;
-        VectoredRead {
-            start,
-            end,
-            blobs_at: self.blobs_at,
-        }
-    }
-}
-
 #[derive(Copy, Clone, Debug)]
 pub enum BlobFlag {
    None,
@@ -365,18 +166,14 @@ pub struct VectoredReadPlanner {
    prev: Option<(Key, Lsn, u64, BlobFlag)>,

    max_read_size: usize,
-
-    mode: VectoredReadCoalesceMode,
 }

 impl VectoredReadPlanner {
    pub fn new(max_read_size: usize) -> Self {
-        let mode = VectoredReadCoalesceMode::get();
        Self {
            blobs: BTreeMap::new(),
            prev: None,
            max_read_size,
-            mode,
        }
    }

@@ -455,7 +252,6 @@ impl VectoredReadPlanner {
                        end_offset,
                        BlobMeta { key, lsn },
                        self.max_read_size,
-                        self.mode,
                    );

                    let prev_read_builder = current_read_builder.replace(next_read_builder);
@@ -507,18 +303,6 @@ impl<'a> VectoredBlobReader<'a> {
            read.size(),
            buf.capacity()
        );
-
-        if cfg!(debug_assertions) {
-            let align = virtual_file::get_io_buffer_alignment() as u64;
-            debug_assert_eq!(
-                read.start % align,
-                0,
-                "Read start at {} does not satisfy the required io buffer alignment ({} bytes)",
-                read.start,
-                align
-            );
-        }
-
        let mut buf = self
            .file
            .read_exact_at(buf.slice(0..read.size()), read.start, ctx)
@@ -526,20 +310,27 @@ impl<'a> VectoredBlobReader<'a> {
            .into_inner();

        let blobs_at = read.blobs_at.as_slice();
-
-        let start_offset = read.start;
+        let start_offset = blobs_at.first().expect("VectoredRead is never empty").0;

        let mut metas = Vec::with_capacity(blobs_at.len());
+
        // Blobs in `read` only provide their starting offset. The end offset
        // of a blob is implicit: the start of the next blob if one exists
        // or the end of the read.
+        let pairs = blobs_at.iter().zip(
+            blobs_at
+                .iter()
+                .map(Some)
+                .skip(1)
+                .chain(std::iter::once(None)),
+        );

        // Some scratch space, put here for reusing the allocation
        let mut decompressed_vec = Vec::new();

-        for (blob_start, meta) in blobs_at {
-            let blob_start_in_buf = blob_start - start_offset;
-            let first_len_byte = buf[blob_start_in_buf as usize];
+        for ((offset, meta), next) in pairs {
+            let offset_in_buf = offset - start_offset;
+            let first_len_byte = buf[offset_in_buf as usize];

            // Each blob is prefixed by a header containing its size and compression information.
            // Extract the size and skip that header to find the start of the data.
@@ -549,7 +340,7 @@ impl<'a> VectoredBlobReader<'a> {
                (1, first_len_byte as u64, BYTE_UNCOMPRESSED)
            } else {
                let mut blob_size_buf = [0u8; 4];
-                let offset_in_buf = blob_start_in_buf as usize;
+                let offset_in_buf = offset_in_buf as usize;

                blob_size_buf.copy_from_slice(&buf[offset_in_buf..offset_in_buf + 4]);
                blob_size_buf[0] &= !LEN_COMPRESSION_BIT_MASK;
@@ -562,8 +353,12 @@ impl<'a> VectoredBlobReader<'a> {
                )
            };

-            let start_raw = blob_start_in_buf + size_length;
-            let end_raw = start_raw + blob_size;
+            let start_raw = offset_in_buf + size_length;
+            let end_raw = match next {
+                Some((next_blob_start_offset, _)) => next_blob_start_offset - start_offset,
+                None => start_raw + blob_size,
+            };
+            assert_eq!(end_raw - start_raw, blob_size);
            let (start, end);
            if compression_bits == BYTE_UNCOMPRESSED {
                start = start_raw as usize;
@@ -612,22 +407,18 @@ pub struct StreamingVectoredReadPlanner {
    max_cnt: usize,
    /// Size of the current batch
    cnt: usize,
-
-    mode: VectoredReadCoalesceMode,
 }

 impl StreamingVectoredReadPlanner {
    pub fn new(max_read_size: u64, max_cnt: usize) -> Self {
        assert!(max_cnt > 0);
        assert!(max_read_size > 0);
-        let mode = VectoredReadCoalesceMode::get();
        Self {
            read_builder: None,
            prev: None,
            max_cnt,
            max_read_size,
            cnt: 0,
-            mode,
        }
    }

@@ -676,12 +467,17 @@ impl StreamingVectoredReadPlanner {
            }
            None => {
                self.read_builder = {
-                    Some(VectoredReadBuilder::new_streaming(
-                        start_offset,
-                        end_offset,
-                        BlobMeta { key, lsn },
-                        self.mode,
-                    ))
+                    let mut blobs_at = VecMap::default();
+                    blobs_at
+                        .append(start_offset, BlobMeta { key, lsn })
+                        .expect("First insertion always succeeds");
+
+                    Some(VectoredReadBuilder {
+                        start: start_offset,
+                        end: end_offset,
+                        blobs_at,
+                        max_read_size: None,
+                    })
                };
            }
        }
@@ -715,9 +511,7 @@ mod tests {
    use super::*;

    fn validate_read(read: &VectoredRead, offset_range: &[(Key, Lsn, u64, BlobFlag)]) {
-        let align = virtual_file::get_io_buffer_alignment() as u64;
-        assert_eq!(read.start % align, 0);
-        assert_eq!(read.start / align, offset_range.first().unwrap().2 / align);
+        assert_eq!(read.start, offset_range.first().unwrap().2);

        let expected_offsets_in_read: Vec<_> = offset_range.iter().map(|o| o.2).collect();

@@ -731,63 +525,6 @@ mod tests {
        assert_eq!(expected_offsets_in_read, offsets_in_read);
    }

-    #[test]
-    fn planner_chunked_coalesce_all_test() {
-        use crate::virtual_file;
-
-        const CHUNK_SIZE: u64 = 512;
-        virtual_file::set_io_buffer_alignment(CHUNK_SIZE as usize).unwrap();
-        let max_read_size = CHUNK_SIZE as usize * 8;
-        let key = Key::MIN;
-        let lsn = Lsn(0);
-
-        let blob_descriptions = [
-            (key, lsn, CHUNK_SIZE / 8, BlobFlag::None), // Read 1 BEGIN
-            (key, lsn, CHUNK_SIZE / 4, BlobFlag::Ignore), // Gap
-            (key, lsn, CHUNK_SIZE / 2, BlobFlag::None),
-            (key, lsn, CHUNK_SIZE - 2, BlobFlag::Ignore), // Gap
-            (key, lsn, CHUNK_SIZE, BlobFlag::None),
-            (key, lsn, CHUNK_SIZE * 2 - 1, BlobFlag::None),
-            (key, lsn, CHUNK_SIZE * 2 + 1, BlobFlag::Ignore), // Gap
-            (key, lsn, CHUNK_SIZE * 3 + 1, BlobFlag::None),
-            (key, lsn, CHUNK_SIZE * 5 + 1, BlobFlag::None),
-            (key, lsn, CHUNK_SIZE * 6 + 1, BlobFlag::Ignore), // skipped chunk size, but not a chunk: should coalesce.
-            (key, lsn, CHUNK_SIZE * 7 + 1, BlobFlag::None),
-            (key, lsn, CHUNK_SIZE * 8, BlobFlag::None), // Read 2 BEGIN (b/c max_read_size)
-            (key, lsn, CHUNK_SIZE * 9, BlobFlag::Ignore), // ==== skipped a chunk
-            (key, lsn, CHUNK_SIZE * 10, BlobFlag::None), // Read 3 BEGIN (cannot coalesce)
-        ];
-
-        let ranges = [
-            &[
-                blob_descriptions[0],
-                blob_descriptions[2],
-                blob_descriptions[4],
-                blob_descriptions[5],
-                blob_descriptions[7],
-                blob_descriptions[8],
-                blob_descriptions[10],
-            ],
-            &blob_descriptions[11..12],
-            &blob_descriptions[13..],
-        ];
-
-        let mut planner = VectoredReadPlanner::new(max_read_size);
-        for (key, lsn, offset, flag) in blob_descriptions {
-            planner.handle(key, lsn, offset, flag);
-        }
-
-        planner.handle_range_end(652 * 1024);
-
-        let reads = planner.finish();
-
-        assert_eq!(reads.len(), ranges.len());
-
-        for (idx, read) in reads.iter().enumerate() {
-            validate_read(read, ranges[idx]);
-        }
-    }
-
    #[test]
    fn planner_max_read_size_test() {
        let max_read_size = 128 * 1024;
@@ -1000,7 +737,6 @@ mod tests {
        let reserved_bytes = blobs.iter().map(|bl| bl.len()).max().unwrap() * 2 + 16;
        let mut buf = BytesMut::with_capacity(reserved_bytes);

-        let mode = VectoredReadCoalesceMode::get();
        let vectored_blob_reader = VectoredBlobReader::new(&file);
        let meta = BlobMeta {
            key: Key::MIN,
@@ -1012,7 +748,7 @@ mod tests {
            if idx + 1 == offsets.len() {
                continue;
            }
-            let read_builder = VectoredReadBuilder::new(*offset, *end, meta, 16 * 4096, mode);
+            let read_builder = VectoredReadBuilder::new(*offset, *end, meta, 16 * 4096);
            let read = read_builder.build();
            let result = vectored_blob_reader.read_blobs(&read, buf, &ctx).await?;
            assert_eq!(result.blobs.len(), 1);
@@ -1048,12 +784,4 @@ mod tests {
        round_trip_test_compressed(&blobs, true).await?;
        Ok(())
    }
-
-    #[test]
-    fn test_div_round_up() {
-        const CHUNK_SIZE: usize = 512;
-        assert_eq!(1, div_round_up(200, CHUNK_SIZE));
-        assert_eq!(1, div_round_up(CHUNK_SIZE, CHUNK_SIZE));
-        assert_eq!(2, div_round_up(CHUNK_SIZE + 1, CHUNK_SIZE));
-    }
 }
--- a/pageserver/src/utilization.rs
+++ b/pageserver/src/utilization.rs
@@ -9,7 +9,7 @@ use utils::serde_percent::Percent;

 use pageserver_api::models::PageserverUtilization;

-use crate::{config::PageServerConf, metrics::NODE_UTILIZATION_SCORE, tenant::mgr::TenantManager};
+use crate::{config::PageServerConf, tenant::mgr::TenantManager};

 pub(crate) fn regenerate(
    conf: &PageServerConf,
@@ -58,13 +58,13 @@ pub(crate) fn regenerate(
        disk_usable_pct,
        shard_count,
        max_shard_count: MAX_SHARDS,
-        utilization_score: None,
+        utilization_score: 0,
        captured_at: utils::serde_system_time::SystemTime(captured_at),
    };

-    // Initialize `PageserverUtilization::utilization_score`
-    let score = doc.cached_score();
-    NODE_UTILIZATION_SCORE.set(score);
+    doc.refresh_score();
+
+    // TODO: make utilization_score into a metric

    Ok(doc)
 }
--- a/pageserver/src/virtual_file.rs
+++ b/pageserver/src/virtual_file.rs
@@ -10,7 +10,6 @@
 //! This is similar to PostgreSQL's virtual file descriptor facility in
 //! src/backend/storage/file/fd.c
 //!
-use crate::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
 use crate::context::RequestContext;
 use crate::metrics::{StorageIoOperation, STORAGE_IO_SIZE, STORAGE_IO_TIME_METRIC};

@@ -18,7 +17,6 @@ use crate::page_cache::{PageWriteGuard, PAGE_SZ};
 use crate::tenant::TENANTS_SEGMENT_NAME;
 use camino::{Utf8Path, Utf8PathBuf};
 use once_cell::sync::OnceCell;
-use owned_buffers_io::io_buf_ext::FullSlice;
 use pageserver_api::shard::TenantShardId;
 use std::fs::File;
 use std::io::{Error, ErrorKind, Seek, SeekFrom};
@@ -52,7 +50,6 @@ pub(crate) mod owned_buffers_io {
    //! but for the time being we're proving out the primitives in the neon.git repo
    //! for faster iteration.

-    pub(crate) mod io_buf_ext;
    pub(crate) mod slice;
    pub(crate) mod write;
    pub(crate) mod util {
@@ -640,24 +637,24 @@ impl VirtualFile {
    }

    // Copied from https://doc.rust-lang.org/1.72.0/src/std/os/unix/fs.rs.html#219-235
-    pub async fn write_all_at<Buf: IoBuf + Send>(
+    pub async fn write_all_at<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
        &self,
-        buf: FullSlice<Buf>,
+        buf: B,
        mut offset: u64,
        ctx: &RequestContext,
-    ) -> (FullSlice<Buf>, Result<(), Error>) {
-        let buf = buf.into_raw_slice();
-        let bounds = buf.bounds();
-        let restore =
-            |buf: Slice<_>| FullSlice::must_new(Slice::from_buf_bounds(buf.into_inner(), bounds));
-        let mut buf = buf;
+    ) -> (B::Buf, Result<(), Error>) {
+        let buf_len = buf.bytes_init();
+        if buf_len == 0 {
+            return (Slice::into_inner(buf.slice_full()), Ok(()));
+        }
+        let mut buf = buf.slice(0..buf_len);
        while !buf.is_empty() {
-            let (tmp, res) = self.write_at(FullSlice::must_new(buf), offset, ctx).await;
-            buf = tmp.into_raw_slice();
+            let res;
+            (buf, res) = self.write_at(buf, offset, ctx).await;
            match res {
                Ok(0) => {
                    return (
-                        restore(buf),
+                        Slice::into_inner(buf),
                        Err(Error::new(
                            std::io::ErrorKind::WriteZero,
                            "failed to write whole buffer",
@@ -669,33 +666,33 @@ impl VirtualFile {
                    offset += n as u64;
                }
                Err(e) if e.kind() == std::io::ErrorKind::Interrupted => {}
-                Err(e) => return (restore(buf), Err(e)),
+                Err(e) => return (Slice::into_inner(buf), Err(e)),
            }
        }
-        (restore(buf), Ok(()))
+        (Slice::into_inner(buf), Ok(()))
    }

-    /// Writes `buf` to the file at the current offset.
-    ///
-    /// Panics if there is an uninitialized range in `buf`, as that is most likely a bug in the caller.
-    pub async fn write_all<Buf: IoBuf + Send>(
+    /// Writes `buf.slice(0..buf.bytes_init())`.
+    /// Returns the IoBuf that is underlying the BoundedBuf `buf`.
+    /// I.e., the returned value's `bytes_init()` method returns something different than the `bytes_init()` that was passed in.
+    /// It's quite brittle and easy to mis-use, so, we return the size in the Ok() variant.
+    pub async fn write_all<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
        &mut self,
-        buf: FullSlice<Buf>,
+        buf: B,
        ctx: &RequestContext,
-    ) -> (FullSlice<Buf>, Result<usize, Error>) {
-        let buf = buf.into_raw_slice();
-        let bounds = buf.bounds();
-        let restore =
-            |buf: Slice<_>| FullSlice::must_new(Slice::from_buf_bounds(buf.into_inner(), bounds));
-        let nbytes = buf.len();
-        let mut buf = buf;
+    ) -> (B::Buf, Result<usize, Error>) {
+        let nbytes = buf.bytes_init();
+        if nbytes == 0 {
+            return (Slice::into_inner(buf.slice_full()), Ok(0));
+        }
+        let mut buf = buf.slice(0..nbytes);
        while !buf.is_empty() {
-            let (tmp, res) = self.write(FullSlice::must_new(buf), ctx).await;
-            buf = tmp.into_raw_slice();
+            let res;
+            (buf, res) = self.write(buf, ctx).await;
            match res {
                Ok(0) => {
                    return (
-                        restore(buf),
+                        Slice::into_inner(buf),
                        Err(Error::new(
                            std::io::ErrorKind::WriteZero,
                            "failed to write whole buffer",
@@ -706,17 +703,17 @@ impl VirtualFile {
                    buf = buf.slice(n..);
                }
                Err(ref e) if e.kind() == std::io::ErrorKind::Interrupted => {}
-                Err(e) => return (restore(buf), Err(e)),
+                Err(e) => return (Slice::into_inner(buf), Err(e)),
            }
        }
-        (restore(buf), Ok(nbytes))
+        (Slice::into_inner(buf), Ok(nbytes))
    }

    async fn write<B: IoBuf + Send>(
        &mut self,
-        buf: FullSlice<B>,
+        buf: Slice<B>,
        ctx: &RequestContext,
-    ) -> (FullSlice<B>, Result<usize, std::io::Error>) {
+    ) -> (Slice<B>, Result<usize, std::io::Error>) {
        let pos = self.pos;
        let (buf, res) = self.write_at(buf, pos, ctx).await;
        let n = match res {
@@ -757,24 +754,12 @@ impl VirtualFile {
        })
    }

-    /// The function aborts the process if the error is fatal.
    async fn write_at<B: IoBuf + Send>(
        &self,
-        buf: FullSlice<B>,
+        buf: Slice<B>,
        offset: u64,
        _ctx: &RequestContext, /* TODO: use for metrics: https://github.com/neondatabase/neon/issues/6107 */
-    ) -> (FullSlice<B>, Result<usize, Error>) {
-        let (slice, result) = self.write_at_inner(buf, offset, _ctx).await;
-        let result = result.maybe_fatal_err("write_at");
-        (slice, result)
-    }
-
-    async fn write_at_inner<B: IoBuf + Send>(
-        &self,
-        buf: FullSlice<B>,
-        offset: u64,
-        _ctx: &RequestContext, /* TODO: use for metrics: https://github.com/neondatabase/neon/issues/6107 */
-    ) -> (FullSlice<B>, Result<usize, Error>) {
+    ) -> (Slice<B>, Result<usize, Error>) {
        let file_guard = match self.lock_file().await {
            Ok(file_guard) => file_guard,
            Err(e) => return (buf, Err(e)),
@@ -1108,11 +1093,11 @@ impl Drop for VirtualFile {

 impl OwnedAsyncWriter for VirtualFile {
    #[inline(always)]
-    async fn write_all<Buf: IoBuf + Send>(
+    async fn write_all<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
        &mut self,
-        buf: FullSlice<Buf>,
+        buf: B,
        ctx: &RequestContext,
-    ) -> std::io::Result<(usize, FullSlice<Buf>)> {
+    ) -> std::io::Result<(usize, B::Buf)> {
        let (buf, res) = VirtualFile::write_all(self, buf, ctx).await;
        res.map(move |v| (v, buf))
    }
@@ -1141,13 +1126,10 @@ impl OpenFiles {
 /// server startup.
 ///
 #[cfg(not(test))]
-pub fn init(num_slots: usize, engine: IoEngineKind, io_buffer_alignment: usize) {
+pub fn init(num_slots: usize, engine: IoEngineKind) {
    if OPEN_FILES.set(OpenFiles::new(num_slots)).is_err() {
        panic!("virtual_file::init called twice");
    }
-    if set_io_buffer_alignment(io_buffer_alignment).is_err() {
-        panic!("IO buffer alignment ({io_buffer_alignment}) is not a power of two");
-    }
    io_engine::init(engine);
    crate::metrics::virtual_file_descriptor_cache::SIZE_MAX.set(num_slots as u64);
 }
@@ -1171,69 +1153,13 @@ fn get_open_files() -> &'static OpenFiles {
    }
 }

-static IO_BUFFER_ALIGNMENT: AtomicUsize = AtomicUsize::new(DEFAULT_IO_BUFFER_ALIGNMENT);
-
-/// Returns true if `x` is zero or a power of two.
-fn is_zero_or_power_of_two(x: usize) -> bool {
-    (x == 0) || ((x & (x - 1)) == 0)
-}
-
-#[allow(unused)]
-pub(crate) fn set_io_buffer_alignment(align: usize) -> Result<(), usize> {
-    if is_zero_or_power_of_two(align) {
-        IO_BUFFER_ALIGNMENT.store(align, std::sync::atomic::Ordering::Relaxed);
-        Ok(())
-    } else {
-        Err(align)
-    }
-}
-
-/// Gets the io buffer alignment requirement. Returns 0 if there is no requirement specified.
-///
-/// This function should be used to check the raw config value.
-pub(crate) fn get_io_buffer_alignment_raw() -> usize {
-    let align = IO_BUFFER_ALIGNMENT.load(std::sync::atomic::Ordering::Relaxed);
-
-    if cfg!(test) {
-        let env_var_name = "NEON_PAGESERVER_UNIT_TEST_IO_BUFFER_ALIGNMENT";
-        if align == DEFAULT_IO_BUFFER_ALIGNMENT {
-            if let Some(test_align) = utils::env::var(env_var_name) {
-                if is_zero_or_power_of_two(test_align) {
-                    test_align
-                } else {
-                    panic!("IO buffer alignment ({test_align}) is not a power of two");
-                }
-            } else {
-                crate::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT
-            }
-        } else {
-            align
-        }
-    } else {
-        align
-    }
-}
-
-/// Gets the io buffer alignment requirement. Returns 1 if the alignment config is set to zero.
-///
-/// This function should be used for getting the actual alignment value to use.
-pub(crate) fn get_io_buffer_alignment() -> usize {
-    let align = get_io_buffer_alignment_raw();
-    if align == DEFAULT_IO_BUFFER_ALIGNMENT {
-        1
-    } else {
-        align
-    }
-}
-
 #[cfg(test)]
 mod tests {
    use crate::context::DownloadBehavior;
    use crate::task_mgr::TaskKind;

    use super::*;
-    use owned_buffers_io::io_buf_ext::IoBufExt;
-    use owned_buffers_io::slice::SliceMutExt;
+    use owned_buffers_io::slice::SliceExt;
    use rand::seq::SliceRandom;
    use rand::thread_rng;
    use rand::Rng;
@@ -1267,9 +1193,9 @@ mod tests {
                }
            }
        }
-        async fn write_all_at<Buf: IoBuf + Send>(
+        async fn write_all_at<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
            &self,
-            buf: FullSlice<Buf>,
+            buf: B,
            offset: u64,
            ctx: &RequestContext,
        ) -> Result<(), Error> {
@@ -1278,7 +1204,13 @@ mod tests {
                    let (_buf, res) = file.write_all_at(buf, offset, ctx).await;
                    res
                }
-                MaybeVirtualFile::File(file) => file.write_all_at(&buf[..], offset),
+                MaybeVirtualFile::File(file) => {
+                    let buf_len = buf.bytes_init();
+                    if buf_len == 0 {
+                        return Ok(());
+                    }
+                    file.write_all_at(&buf.slice(0..buf_len), offset)
+                }
            }
        }
        async fn seek(&mut self, pos: SeekFrom) -> Result<u64, Error> {
@@ -1287,9 +1219,9 @@ mod tests {
                MaybeVirtualFile::File(file) => file.seek(pos),
            }
        }
-        async fn write_all<Buf: IoBuf + Send>(
+        async fn write_all<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
            &mut self,
-            buf: FullSlice<Buf>,
+            buf: B,
            ctx: &RequestContext,
        ) -> Result<(), Error> {
            match self {
@@ -1297,7 +1229,13 @@ mod tests {
                    let (_buf, res) = file.write_all(buf, ctx).await;
                    res.map(|_| ())
                }
-                MaybeVirtualFile::File(file) => file.write_all(&buf[..]),
+                MaybeVirtualFile::File(file) => {
+                    let buf_len = buf.bytes_init();
+                    if buf_len == 0 {
+                        return Ok(());
+                    }
+                    file.write_all(&buf.slice(0..buf_len))
+                }
            }
        }

@@ -1409,9 +1347,7 @@ mod tests {
            &ctx,
        )
        .await?;
-        file_a
-            .write_all(b"foobar".to_vec().slice_len(), &ctx)
-            .await?;
+        file_a.write_all(b"foobar".to_vec(), &ctx).await?;

        // cannot read from a file opened in write-only mode
        let _ = file_a.read_string(&ctx).await.unwrap_err();
@@ -1420,10 +1356,7 @@ mod tests {
        let mut file_a = A::open(path_a, OpenOptions::new().read(true).to_owned(), &ctx).await?;

        // cannot write to a file opened in read-only mode
-        let _ = file_a
-            .write_all(b"bar".to_vec().slice_len(), &ctx)
-            .await
-            .unwrap_err();
+        let _ = file_a.write_all(b"bar".to_vec(), &ctx).await.unwrap_err();

        // Try simple read
        assert_eq!("foobar", file_a.read_string(&ctx).await?);
@@ -1466,12 +1399,8 @@ mod tests {
            &ctx,
        )
        .await?;
-        file_b
-            .write_all_at(b"BAR".to_vec().slice_len(), 3, &ctx)
-            .await?;
-        file_b
-            .write_all_at(b"FOO".to_vec().slice_len(), 0, &ctx)
-            .await?;
+        file_b.write_all_at(b"BAR".to_vec(), 3, &ctx).await?;
+        file_b.write_all_at(b"FOO".to_vec(), 0, &ctx).await?;

        assert_eq!(file_b.read_string_at(2, 3, &ctx).await?, "OBA");

--- a/pageserver/src/virtual_file/io_engine.rs
+++ b/pageserver/src/virtual_file/io_engine.rs
@@ -12,7 +12,7 @@
 #[cfg(target_os = "linux")]
 pub(super) mod tokio_epoll_uring_ext;

-use tokio_epoll_uring::IoBuf;
+use tokio_epoll_uring::{IoBuf, Slice};
 use tracing::Instrument;

 pub(crate) use super::api::IoEngineKind;
@@ -107,10 +107,7 @@ use std::{
    sync::atomic::{AtomicU8, Ordering},
 };

-use super::{
-    owned_buffers_io::{io_buf_ext::FullSlice, slice::SliceMutExt},
-    FileGuard, Metadata,
-};
+use super::{owned_buffers_io::slice::SliceExt, FileGuard, Metadata};

 #[cfg(target_os = "linux")]
 fn epoll_uring_error_to_std(e: tokio_epoll_uring::Error<std::io::Error>) -> std::io::Error {
@@ -209,8 +206,8 @@ impl IoEngine {
        &self,
        file_guard: FileGuard,
        offset: u64,
-        buf: FullSlice<B>,
-    ) -> ((FileGuard, FullSlice<B>), std::io::Result<usize>) {
+        buf: Slice<B>,
+    ) -> ((FileGuard, Slice<B>), std::io::Result<usize>) {
        match self {
            IoEngine::NotSet => panic!("not initialized"),
            IoEngine::StdFs => {
@@ -220,12 +217,8 @@ impl IoEngine {
            #[cfg(target_os = "linux")]
            IoEngine::TokioEpollUring => {
                let system = tokio_epoll_uring_ext::thread_local_system().await;
-                let ((file_guard, slice), res) =
-                    system.write(file_guard, offset, buf.into_raw_slice()).await;
-                (
-                    (file_guard, FullSlice::must_new(slice)),
-                    res.map_err(epoll_uring_error_to_std),
-                )
+                let (resources, res) = system.write(file_guard, offset, buf).await;
+                (resources, res.map_err(epoll_uring_error_to_std))
            }
        }
    }
--- a/pageserver/src/virtual_file/owned_buffers_io/io_buf_ext.rs
+++ b/pageserver/src/virtual_file/owned_buffers_io/io_buf_ext.rs
@@ -1,78 +0,0 @@
-//! See [`FullSlice`].
-
-use bytes::{Bytes, BytesMut};
-use std::ops::{Deref, Range};
-use tokio_epoll_uring::{BoundedBuf, IoBuf, Slice};
-
-/// The true owned equivalent for Rust [`slice`]. Use this for the write path.
-///
-/// Unlike [`tokio_epoll_uring::Slice`], which we unfortunately inherited from `tokio-uring`,
-/// [`FullSlice`] is guaranteed to have all its bytes initialized. This means that
-/// [`<FullSlice as Deref<Target = [u8]>>::len`] is equal to [`Slice::bytes_init`] and [`Slice::bytes_total`].
-///
-pub struct FullSlice<B> {
-    slice: Slice<B>,
-}
-
-impl<B> FullSlice<B>
-where
-    B: IoBuf,
-{
-    pub(crate) fn must_new(slice: Slice<B>) -> Self {
-        assert_eq!(slice.bytes_init(), slice.bytes_total());
-        FullSlice { slice }
-    }
-    pub(crate) fn into_raw_slice(self) -> Slice<B> {
-        let FullSlice { slice: s } = self;
-        s
-    }
-}
-
-impl<B> Deref for FullSlice<B>
-where
-    B: IoBuf,
-{
-    type Target = [u8];
-
-    fn deref(&self) -> &[u8] {
-        let rust_slice = &self.slice[..];
-        assert_eq!(rust_slice.len(), self.slice.bytes_init());
-        assert_eq!(rust_slice.len(), self.slice.bytes_total());
-        rust_slice
-    }
-}
-
-pub(crate) trait IoBufExt {
-    /// Get a [`FullSlice`] for the entire buffer, i.e., `self[..]` or `self[0..self.len()]`.
-    fn slice_len(self) -> FullSlice<Self>
-    where
-        Self: Sized;
-}
-
-macro_rules! impl_io_buf_ext {
-    ($T:ty) => {
-        impl IoBufExt for $T {
-            #[inline(always)]
-            fn slice_len(self) -> FullSlice<Self> {
-                let len = self.len();
-                let s = if len == 0 {
-                    // `BoundedBuf::slice(0..len)` or `BoundedBuf::slice(..)` has an incorrect assertion,
-                    // causing a panic if len == 0.
-                    // The Slice::from_buf_bounds has the correct assertion (<= instead of <).
-                    // => https://github.com/neondatabase/tokio-epoll-uring/issues/46
-                    let slice = self.slice_full();
-                    let mut bounds: Range<_> = slice.bounds();
-                    bounds.end = bounds.start;
-                    Slice::from_buf_bounds(slice.into_inner(), bounds)
-                } else {
-                    self.slice(0..len)
-                };
-                FullSlice::must_new(s)
-            }
-        }
-    };
-}
-
-impl_io_buf_ext!(Bytes);
-impl_io_buf_ext!(BytesMut);
-impl_io_buf_ext!(Vec<u8>);
--- a/pageserver/src/virtual_file/owned_buffers_io/slice.rs
+++ b/pageserver/src/virtual_file/owned_buffers_io/slice.rs
@@ -3,14 +3,14 @@ use tokio_epoll_uring::BoundedBufMut;
 use tokio_epoll_uring::IoBufMut;
 use tokio_epoll_uring::Slice;

-pub(crate) trait SliceMutExt {
+pub(crate) trait SliceExt {
    /// Get a `&mut[0..self.bytes_total()`] slice, for when you need to do borrow-based IO.
    ///
    /// See the test case `test_slice_full_zeroed` for the difference to just doing `&slice[..]`
    fn as_mut_rust_slice_full_zeroed(&mut self) -> &mut [u8];
 }

-impl<B> SliceMutExt for Slice<B>
+impl<B> SliceExt for Slice<B>
 where
    B: IoBufMut,
 {
--- a/pageserver/src/virtual_file/owned_buffers_io/util/size_tracking_writer.rs
+++ b/pageserver/src/virtual_file/owned_buffers_io/util/size_tracking_writer.rs
@@ -1,8 +1,5 @@
-use crate::{
-    context::RequestContext,
-    virtual_file::owned_buffers_io::{io_buf_ext::FullSlice, write::OwnedAsyncWriter},
-};
-use tokio_epoll_uring::IoBuf;
+use crate::{context::RequestContext, virtual_file::owned_buffers_io::write::OwnedAsyncWriter};
+use tokio_epoll_uring::{BoundedBuf, IoBuf};

 pub struct Writer<W> {
    dst: W,
@@ -38,11 +35,11 @@ where
    W: OwnedAsyncWriter,
 {
    #[inline(always)]
-    async fn write_all<Buf: IoBuf + Send>(
+    async fn write_all<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
        &mut self,
-        buf: FullSlice<Buf>,
+        buf: B,
        ctx: &RequestContext,
-    ) -> std::io::Result<(usize, FullSlice<Buf>)> {
+    ) -> std::io::Result<(usize, B::Buf)> {
        let (nwritten, buf) = self.dst.write_all(buf, ctx).await?;
        self.bytes_amount += u64::try_from(nwritten).unwrap();
        Ok((nwritten, buf))
--- a/Show More
+++ b/Show More