Proxy release 2025-03-13

## Problem
#11061 introduced code fetching previous releases. #11151 introduced jq
error handling, which has also been applied in #11061, but parenthesis
have been missed.

## Summary of changes
Add parenthesis around error handling code.

.github/PULL_REQUEST_TEMPLATE/release-pr.md

@@ -1,21 +0,0 @@
-## Release 202Y-MM-DD
-
-**NB: this PR must be merged only by 'Create a merge commit'!**
-
-### Checklist when preparing for release
-- [ ] Read or refresh [the release flow guide](https://www.notion.so/neondatabase/Release-general-flow-61f2e39fd45d4d14a70c7749604bd70b)
-- [ ] Ask in the [cloud Slack channel](https://neondb.slack.com/archives/C033A2WE6BZ) that you are going to rollout the release. Any blockers?
-- [ ] Does this release contain any db migrations? Destructive ones? What is the rollback plan?
-
-<!-- List everything that should be done **before** release, any issues / setting changes / etc -->
-
-### Checklist after release
-- [ ] Make sure instructions from PRs included in this release and labeled `manual_release_instructions` are executed (either by you or by people who wrote them).
-- [ ] Based on the merged commits write release notes and open a PR into `website` repo ([example](https://github.com/neondatabase/website/pull/219/files))
-- [ ] Check [#dev-production-stream](https://neondb.slack.com/archives/C03F5SM1N02) Slack channel
-- [ ] Check [stuck projects page](https://console.neon.tech/admin/projects?sort=last_active&order=desc&stuck=true)
-- [ ] Check [recent operation failures](https://console.neon.tech/admin/operations?action=create_timeline%2Cstart_compute%2Cstop_compute%2Csuspend_compute%2Capply_config%2Cdelete_timeline%2Cdelete_tenant%2Ccreate_branch%2Ccheck_availability&sort=updated_at&order=desc&had_retries=some)
-- [ ] Check [cloud SLO dashboard](https://neonprod.grafana.net/d/_oWcBMJ7k/cloud-slos?orgId=1)
-- [ ] Check [compute startup metrics dashboard](https://neonprod.grafana.net/d/5OkYJEmVz/compute-startup-time)
-
-<!-- List everything that should be done **after** release, any admin UI configuration / Grafana dashboard / alert changes / setting changes / etc -->

.github/actionlint.yml

@@ -33,3 +33,5 @@ config-variables:
   - NEON_PROD_AWS_ACCOUNT_ID
   - AWS_ECR_REGION
   - BENCHMARK_LARGE_OLTP_PROJECTID
+  - SLACK_ON_CALL_DEVPROD_STREAM
+  - SLACK_RUST_CHANNEL_ID

.github/scripts/generate_image_maps.py

@@ -1,14 +1,16 @@
 import itertools
 import json
 import os
+import sys
 
-build_tag = os.environ["BUILD_TAG"]
-branch = os.environ["BRANCH"]
-dev_acr = os.environ["DEV_ACR"]
-prod_acr = os.environ["PROD_ACR"]
-dev_aws = os.environ["DEV_AWS"]
-prod_aws = os.environ["PROD_AWS"]
-aws_region = os.environ["AWS_REGION"]
+source_tag = os.getenv("SOURCE_TAG")
+target_tag = os.getenv("TARGET_TAG")
+branch = os.getenv("BRANCH")
+dev_acr = os.getenv("DEV_ACR")
+prod_acr = os.getenv("PROD_ACR")
+dev_aws = os.getenv("DEV_AWS")
+prod_aws = os.getenv("PROD_AWS")
+aws_region = os.getenv("AWS_REGION")
 
 components = {
     "neon": ["neon"],
@@ -39,24 +41,23 @@ registries = {
 
 outputs: dict[str, dict[str, list[str]]] = {}
 
-target_tags = [build_tag, "latest"] if branch == "main" else [build_tag]
-target_stages = ["dev", "prod"] if branch.startswith("release") else ["dev"]
+target_tags = [target_tag, "latest"] if branch == "main" else [target_tag]
+target_stages = (
+    ["dev", "prod"] if branch in ["release", "release-proxy", "release-compute"] else ["dev"]
+)
 
 for component_name, component_images in components.items():
     for stage in target_stages:
-        outputs[f"{component_name}-{stage}"] = dict(
-            [
-                (
-                    f"docker.io/neondatabase/{component_image}:{build_tag}",
-                    [
-                        f"{combo[0]}/{component_image}:{combo[1]}"
-                        for combo in itertools.product(registries[stage], target_tags)
-                    ],
-                )
-                for component_image in component_images
+        outputs[f"{component_name}-{stage}"] = {
+            f"docker.io/neondatabase/{component_image}:{source_tag}": [
+                f"{registry}/{component_image}:{tag}"
+                for registry, tag in itertools.product(registries[stage], target_tags)
+                if not (registry == "docker.io/neondatabase" and tag == source_tag)
             ]
-        )
+            for component_image in component_images
+        }
 
-with open(os.environ["GITHUB_OUTPUT"], "a") as f:
+with open(os.getenv("GITHUB_OUTPUT", "/dev/null"), "a") as f:
     for key, value in outputs.items():
         f.write(f"{key}={json.dumps(value)}\n")
+        print(f"Image map for {key}:\n{json.dumps(value, indent=2)}\n\n", file=sys.stderr)

.github/scripts/lint-release-pr.sh

@@ -0,0 +1,110 @@
+#!/usr/bin/env bash
+
+set -euo pipefail
+
+DOCS_URL="https://docs.neon.build/overview/repositories/neon.html"
+
+message() {
+  if [[ -n "${GITHUB_PR_NUMBER:-}" ]]; then
+    gh pr comment --repo "${GITHUB_REPOSITORY}" "${GITHUB_PR_NUMBER}" --edit-last --body "$1" \
+      || gh pr comment --repo "${GITHUB_REPOSITORY}" "${GITHUB_PR_NUMBER}" --body "$1"
+  fi
+  echo "$1"
+}
+
+report_error() {
+  message "❌ $1
+  For more details, see the documentation: ${DOCS_URL}"
+
+  exit 1
+}
+
+case "$RELEASE_BRANCH" in
+  "release") COMPONENT="Storage" ;;
+  "release-proxy") COMPONENT="Proxy" ;;
+  "release-compute") COMPONENT="Compute" ;;
+  *)
+    report_error "Unknown release branch: ${RELEASE_BRANCH}"
+    ;;
+esac
+
+
+# Identify main and release branches
+MAIN_BRANCH="origin/main"
+REMOTE_RELEASE_BRANCH="origin/${RELEASE_BRANCH}"
+
+# Find merge base
+MERGE_BASE=$(git merge-base "${MAIN_BRANCH}" "${REMOTE_RELEASE_BRANCH}")
+echo "Merge base of ${MAIN_BRANCH} and ${RELEASE_BRANCH}: ${MERGE_BASE}"
+
+# Get the HEAD commit (last commit in PR, expected to be the merge commit)
+LAST_COMMIT=$(git rev-parse HEAD)
+
+MERGE_COMMIT_MESSAGE=$(git log -1 --format=%s "${LAST_COMMIT}")
+EXPECTED_MESSAGE_REGEX="^$COMPONENT release [0-9]{4}-[0-9]{2}-[0-9]{2}$"
+
+if ! [[ "${MERGE_COMMIT_MESSAGE}" =~ ${EXPECTED_MESSAGE_REGEX} ]]; then
+  report_error "Merge commit message does not match expected pattern: '<component> release YYYY-MM-DD'
+  Expected component: ${COMPONENT}
+  Found: '${MERGE_COMMIT_MESSAGE}'"
+fi
+echo "✅ Merge commit message is correctly formatted: '${MERGE_COMMIT_MESSAGE}'"
+
+LAST_COMMIT_PARENTS=$(git cat-file -p "${LAST_COMMIT}" | jq -sR '[capture("parent (?<parent>[0-9a-f]{40})"; "g") | .parent]')
+
+if [[ "$(echo "${LAST_COMMIT_PARENTS}" | jq 'length')" -ne 2 ]]; then
+  report_error "Last commit must be a merge commit with exactly two parents"
+fi
+
+EXPECTED_RELEASE_HEAD=$(git rev-parse "${REMOTE_RELEASE_BRANCH}")
+if echo "${LAST_COMMIT_PARENTS}" | jq -e --arg rel "${EXPECTED_RELEASE_HEAD}" 'index($rel) != null' > /dev/null; then
+  LINEAR_HEAD=$(echo "${LAST_COMMIT_PARENTS}" | jq -r '[.[] | select(. != $rel)][0]' --arg rel "${EXPECTED_RELEASE_HEAD}")
+else
+  report_error "Last commit must merge the release branch (${RELEASE_BRANCH})"
+fi
+echo "✅ Last commit correctly merges the previous commit and the release branch"
+echo "Top commit of linear history: ${LINEAR_HEAD}"
+
+MERGE_COMMIT_TREE=$(git rev-parse "${LAST_COMMIT}^{tree}")
+LINEAR_HEAD_TREE=$(git rev-parse "${LINEAR_HEAD}^{tree}")
+
+if [[ "${MERGE_COMMIT_TREE}" != "${LINEAR_HEAD_TREE}" ]]; then
+  report_error "Tree of merge commit (${MERGE_COMMIT_TREE}) does not match tree of linear history head (${LINEAR_HEAD_TREE})
+  This indicates that the merge of ${RELEASE_BRANCH} into this branch was not performed using the merge strategy 'ours'"
+fi
+echo "✅ Merge commit tree matches the linear history head"
+
+EXPECTED_PREVIOUS_COMMIT="${LINEAR_HEAD}"
+
+# Now traverse down the history, ensuring each commit has exactly one parent
+CURRENT_COMMIT="${EXPECTED_PREVIOUS_COMMIT}"
+while [[ "${CURRENT_COMMIT}" != "${MERGE_BASE}" && "${CURRENT_COMMIT}" != "${EXPECTED_RELEASE_HEAD}" ]]; do
+  CURRENT_COMMIT_PARENTS=$(git cat-file -p "${CURRENT_COMMIT}" | jq -sR '[capture("parent (?<parent>[0-9a-f]{40})"; "g") | .parent]')
+
+  if [[ "$(echo "${CURRENT_COMMIT_PARENTS}" | jq 'length')" -ne 1 ]]; then
+    report_error "Commit ${CURRENT_COMMIT} must have exactly one parent"
+  fi
+
+  NEXT_COMMIT=$(echo "${CURRENT_COMMIT_PARENTS}" | jq -r '.[0]')
+
+  if [[ "${NEXT_COMMIT}" == "${MERGE_BASE}" ]]; then
+    echo "✅ Reached merge base (${MERGE_BASE})"
+    PR_BASE="${MERGE_BASE}"
+  elif [[ "${NEXT_COMMIT}" == "${EXPECTED_RELEASE_HEAD}" ]]; then
+    echo "✅ Reached release branch (${EXPECTED_RELEASE_HEAD})"
+    PR_BASE="${EXPECTED_RELEASE_HEAD}"
+  elif [[ -z "${NEXT_COMMIT}" ]]; then
+    report_error "Unexpected end of commit history before reaching merge base"
+  fi
+
+  # Move to the next commit in the chain
+  CURRENT_COMMIT="${NEXT_COMMIT}"
+done
+
+echo "✅ All commits are properly ordered and linear"
+echo "✅ Release PR structure is valid"
+
+echo
+
+message "Commits that are part of this release:
+$(git log --oneline "${PR_BASE}..${LINEAR_HEAD}")"

.github/scripts/previous-releases.jq

@@ -17,6 +17,12 @@
 ({};
  .[$entry.component] |= (if . == null or $entry.version > .version then $entry else . end))
 
+# Ensure that each component exists, or fail
+| (["storage", "compute", "proxy"] - (keys)) as $missing
+| if ($missing | length) > 0 then
+    "Error: Found no release for \($missing | join(", "))!\n" | halt_error(1)
+  else . end
+
 # Convert the resulting object into an array of formatted strings
 | to_entries
 | map("\(.key)=\(.value.full)")

.github/workflows/_create-release-pr.yml

@@ -7,8 +7,8 @@ on:
         description: 'Component name'
         required: true
         type: string
-      release-branch:
-        description: 'Release branch'
+      source-branch:
+        description: 'Source branch'
         required: true
         type: string
     secrets:
@@ -30,17 +30,25 @@ jobs:
     steps:
     - uses: actions/checkout@v4
       with:
-        ref: main
+        ref: ${{ inputs.source-branch }}
+        fetch-depth: 0
 
     - name: Set variables
       id: vars
       env:
         COMPONENT_NAME: ${{ inputs.component-name }}
-        RELEASE_BRANCH: ${{ inputs.release-branch }}
+        RELEASE_BRANCH: >-
+          ${{
+            false
+            || inputs.component-name == 'Storage' && 'release'
+            || inputs.component-name == 'Proxy' && 'release-proxy'
+            || inputs.component-name == 'Compute' && 'release-compute'
+          }}
       run: |
         today=$(date +'%Y-%m-%d')
         echo "title=${COMPONENT_NAME} release ${today}" | tee -a ${GITHUB_OUTPUT}
         echo "rc-branch=rc/${RELEASE_BRANCH}/${today}"  | tee -a ${GITHUB_OUTPUT}
+        echo "release-branch=${RELEASE_BRANCH}"         | tee -a ${GITHUB_OUTPUT}
 
     - name: Configure git
       run: |
@@ -49,31 +57,36 @@ jobs:
 
     - name: Create RC branch
       env:
+        RELEASE_BRANCH: ${{ steps.vars.outputs.release-branch }}
         RC_BRANCH: ${{ steps.vars.outputs.rc-branch }}
         TITLE: ${{ steps.vars.outputs.title }}
       run: |
-        git checkout -b "${RC_BRANCH}"
+        git switch -c "${RC_BRANCH}"
 
-        # create an empty commit to distinguish workflow runs
-        # from other possible releases from the same commit
-        git commit --allow-empty -m "${TITLE}"
+        # Manually create a merge commit on the current branch, keeping the
+        # tree and setting the parents to the current HEAD and the HEAD of the
+        # release branch. This commit is what we'll fast-forward the release
+        # branch to when merging the release branch.
+        # For details on why, look at
+        # https://docs.neon.build/overview/repositories/neon.html#background-on-commit-history-of-release-prs
+        current_tree=$(git rev-parse 'HEAD^{tree}')
+        release_head=$(git rev-parse "origin/${RELEASE_BRANCH}")
+        current_head=$(git rev-parse HEAD)
+        merge_commit=$(git commit-tree -p "${current_head}" -p "${release_head}" -m "${TITLE}" "${current_tree}")
+
+        # Fast-forward the current branch to the newly created merge_commit
+        git merge --ff-only ${merge_commit}
 
         git push origin "${RC_BRANCH}"
 
-    - name: Create a PR into ${{ inputs.release-branch }}
+    - name: Create a PR into ${{ steps.vars.outputs.release-branch }}
       env:
         GH_TOKEN: ${{ secrets.ci-access-token }}
         RC_BRANCH: ${{ steps.vars.outputs.rc-branch }}
-        RELEASE_BRANCH: ${{ inputs.release-branch }}
+        RELEASE_BRANCH: ${{ steps.vars.outputs.release-branch }}
         TITLE: ${{ steps.vars.outputs.title }}
       run: |
-        cat << EOF > body.md
-          ## ${TITLE}
-
-          **Please merge this Pull Request using 'Create a merge commit' button**
-        EOF
-
         gh pr create --title "${TITLE}" \
-                     --body-file "body.md" \
+                     --body "" \
                      --head "${RC_BRANCH}" \
                      --base "${RELEASE_BRANCH}"

.github/workflows/_meta.yml

@@ -19,11 +19,18 @@ on:
         description: "Tag of the last compute release"
         value: ${{ jobs.tags.outputs.compute }}
       run-kind:
-        description: "The kind of run we're currently in. Will be one of `pr`, `push-main`, `storage-rc`, `storage-release`, `proxy-rc`, `proxy-release`, `compute-rc`, `compute-release` or `merge_queue`"
+        description: "The kind of run we're currently in. Will be one of `push-main`, `storage-release`, `compute-release`, `proxy-release`, `storage-rc-pr`, `compute-rc-pr`,  `proxy-rc-pr`, `pr`, or `workflow-dispatch`"
         value: ${{ jobs.tags.outputs.run-kind }}
+      release-pr-run-id:
+        description: "Only available if `run-kind in [storage-release, proxy-release, compute-release]`. Contains the run ID of the `Build and Test` workflow, assuming one with the current commit can be found."
+        value: ${{ jobs.tags.outputs.release-pr-run-id }}
 
 permissions: {}
 
+defaults:
+  run:
+    shell: bash -euo pipefail {0}
+
 jobs:
   tags:
     runs-on: ubuntu-22.04
@@ -33,6 +40,7 @@ jobs:
       proxy: ${{ steps.previous-releases.outputs.proxy }}
       storage: ${{ steps.previous-releases.outputs.storage }}
       run-kind: ${{ steps.run-kind.outputs.run-kind }}
+      release-pr-run-id: ${{ steps.release-pr-run-id.outputs.release-pr-run-id }}
     permissions:
       contents: read
     steps:
@@ -55,6 +63,7 @@ jobs:
               || (inputs.github-event-name == 'pull_request' && github.base_ref == 'release-compute') && 'compute-rc-pr'
               || (inputs.github-event-name == 'pull_request' && github.base_ref == 'release-proxy')   && 'proxy-rc-pr'
               || (inputs.github-event-name == 'pull_request')                                         && 'pr'
+              || (inputs.github-event-name == 'workflow_dispatch')                                    && 'workflow-dispatch'
               || 'unknown'
             }}
         run: |
@@ -82,9 +91,16 @@ jobs:
             echo "tag=release-compute-$(git rev-list --count HEAD)" | tee -a $GITHUB_OUTPUT
             ;;
           pr|storage-rc-pr|compute-rc-pr|proxy-rc-pr)
-            BUILD_AND_TEST_RUN_ID=$(gh run list -b $CURRENT_BRANCH -c $CURRENT_SHA -w 'Build and Test' -L 1 --json databaseId --jq '.[].databaseId')
+            BUILD_AND_TEST_RUN_ID=$(gh api --paginate \
+              -H "Accept: application/vnd.github+json" \
+              -H "X-GitHub-Api-Version: 2022-11-28" \
+              "/repos/${GITHUB_REPOSITORY}/actions/runs?head_sha=${CURRENT_SHA}&branch=${CURRENT_BRANCH}" \
+              | jq '[.workflow_runs[] | select(.name == "Build and Test")][0].id // ("Error: No matching workflow run found." | halt_error(1))')
             echo "tag=$BUILD_AND_TEST_RUN_ID" | tee -a $GITHUB_OUTPUT
             ;;
+          workflow-dispatch)
+            echo "tag=$GITHUB_RUN_ID" | tee -a $GITHUB_OUTPUT
+            ;;
           *)
             echo "Unexpected RUN_KIND ('${RUN_KIND}'), failing to assign build-tag!"
             exit 1
@@ -101,3 +117,13 @@ jobs:
             "/repos/${GITHUB_REPOSITORY}/releases" \
           | jq -f .github/scripts/previous-releases.jq -r \
           | tee -a "${GITHUB_OUTPUT}"
+
+      - name: Get the release PR run ID
+        id: release-pr-run-id
+        if: ${{ contains(fromJson('["storage-release", "compute-release", "proxy-release"]'), steps.run-kind.outputs.run-kind) }}
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          CURRENT_SHA: ${{ github.event.pull_request.head.sha || github.sha }}
+        run: |
+          RELEASE_PR_RUN_ID=$(gh api "/repos/${GITHUB_REPOSITORY}/actions/runs?head_sha=$CURRENT_SHA" | jq '[.workflow_runs[] | select(.name == "Build and Test") | select(.head_branch | test("^rc/release(-(proxy)|(compute))?/[0-9]{4}-[0-9]{2}-[0-9]{2}$"; "s"))] | first | .id // ("Faied to find Build and Test run from  RC PR!" | halt_error(1))')
+          echo "release-pr-run-id=$RELEASE_PR_RUN_ID" | tee -a $GITHUB_OUTPUT

.github/workflows/build_and_test.yml

@@ -476,7 +476,7 @@ jobs:
         (
           !github.event.pull_request.draft
           || contains( github.event.pull_request.labels.*.name, 'run-e2e-tests-in-draft')
-          || contains(fromJSON('["push-main", "storage-release", "proxy-release", "compute-release"]'), needs.meta.outputs.run-kind)
+          || needs.meta.outputs.run-kind == 'push-main'
         ) && !failure() && !cancelled()
       }}
     needs: [ check-permissions, push-neon-image-dev, push-compute-image-dev, meta ]
@@ -487,7 +487,7 @@ jobs:
 
   neon-image-arch:
     needs: [ check-permissions, build-build-tools-image, meta ]
-    if: ${{ contains(fromJSON('["push-main", "pr", "storage-release", "storage-rc-pr", "proxy-release", "proxy-rc-pr"]'), needs.meta.outputs.run-kind) }}
+    if: ${{ contains(fromJSON('["push-main", "pr", "storage-rc-pr", "proxy-rc-pr"]'), needs.meta.outputs.run-kind) }}
     strategy:
       matrix:
         arch: [ x64, arm64 ]
@@ -537,7 +537,7 @@ jobs:
 
   neon-image:
     needs: [ neon-image-arch, meta ]
-    if: ${{ contains(fromJSON('["push-main", "pr", "storage-release", "storage-rc-pr", "proxy-release", "proxy-rc-pr"]'), needs.meta.outputs.run-kind) }}
+    if: ${{ contains(fromJSON('["push-main", "pr", "storage-rc-pr", "proxy-rc-pr"]'), needs.meta.outputs.run-kind) }}
     runs-on: ubuntu-22.04
     permissions:
       id-token: write # aws-actions/configure-aws-credentials
@@ -559,7 +559,7 @@ jobs:
 
   compute-node-image-arch:
     needs: [ check-permissions, build-build-tools-image, meta ]
-    if: ${{ contains(fromJSON('["push-main", "pr", "compute-release", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
+    if: ${{ contains(fromJSON('["push-main", "pr", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
     permissions:
       id-token: write # aws-actions/configure-aws-credentials
       statuses: write
@@ -651,7 +651,7 @@ jobs:
 
   compute-node-image:
     needs: [ compute-node-image-arch, meta ]
-    if: ${{ contains(fromJSON('["push-main", "pr", "compute-release", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
+    if: ${{ contains(fromJSON('["push-main", "pr", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
     permissions:
       id-token: write # aws-actions/configure-aws-credentials
       statuses: write
@@ -694,7 +694,7 @@ jobs:
 
   vm-compute-node-image-arch:
     needs: [ check-permissions, meta, compute-node-image ]
-    if: ${{ contains(fromJSON('["push-main", "pr", "compute-release", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
+    if: ${{ contains(fromJSON('["push-main", "pr", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
     runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', matrix.arch == 'arm64' && 'large-arm64' || 'large')) }}
     strategy:
       fail-fast: false
@@ -747,7 +747,7 @@ jobs:
 
   vm-compute-node-image:
     needs: [ vm-compute-node-image-arch, meta ]
-    if: ${{ contains(fromJSON('["push-main", "pr", "compute-release", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
+    if: ${{ contains(fromJSON('["push-main", "pr", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
     runs-on: ubuntu-22.04
     strategy:
       matrix:
@@ -773,7 +773,12 @@ jobs:
   test-images:
     needs: [ check-permissions, meta, neon-image, compute-node-image ]
     # Depends on jobs that can get skipped
-    if: "!failure() && !cancelled()"
+    if: >-
+      ${{
+        !failure()
+        && !cancelled()
+        && contains(fromJSON('["push-main", "pr", "storage-rc-pr", "proxy-rc-pr", "compute-rc-pr"]'), needs.meta.outputs.run-kind)
+      }}
     strategy:
       fail-fast: false
       matrix:
@@ -800,7 +805,7 @@ jobs:
       # Ensure that we don't have bad versions.
       - name: Verify image versions
         shell: bash # ensure no set -e for better error messages
-        if: ${{ contains(fromJSON('["push-main", "pr", "storage-release", "storage-rc-pr", "proxy-release", "proxy-rc-pr"]'), needs.meta.outputs.run-kind) }}
+        if: ${{ contains(fromJSON('["push-main", "pr", "storage-rc-pr", "proxy-rc-pr"]'), needs.meta.outputs.run-kind) }}
         run: |
           pageserver_version=$(docker run --rm neondatabase/neon:${{ needs.meta.outputs.build-tag }} "/bin/sh" "-c" "/usr/local/bin/pageserver --version")
 
@@ -821,19 +826,19 @@ jobs:
         env:
           TAG: >-
             ${{
-              contains(fromJSON('["compute-release", "compute-rc-pr"]'), needs.meta.outputs.run-kind)
+              needs.meta.outputs.run-kind == 'compute-rc-pr'
               && needs.meta.outputs.previous-storage-release
               || needs.meta.outputs.build-tag
             }}
           COMPUTE_TAG: >-
             ${{
-              contains(fromJSON('["storage-release", "storage-rc-pr", "proxy-release", "proxy-rc-pr"]'), needs.meta.outputs.run-kind)
+              contains(fromJSON('["storage-rc-pr", "proxy-rc-pr"]'), needs.meta.outputs.run-kind)
               && needs.meta.outputs.previous-compute-release
               || needs.meta.outputs.build-tag
             }}
           TEST_EXTENSIONS_TAG: >-
             ${{
-              contains(fromJSON('["storage-release", "storage-rc-pr", "proxy-release", "proxy-rc-pr"]'), needs.meta.outputs.run-kind)
+              contains(fromJSON('["storage-rc-pr", "proxy-rc-pr"]'), needs.meta.outputs.run-kind)
               && 'latest'
               || needs.meta.outputs.build-tag
             }}
@@ -885,7 +890,13 @@ jobs:
         id: generate
         run: python3 .github/scripts/generate_image_maps.py
         env:
-          BUILD_TAG: "${{ needs.meta.outputs.build-tag }}"
+          SOURCE_TAG: >-
+            ${{
+              contains(fromJson('["storage-release", "compute-release", "proxy-release"]'), needs.meta.outputs.run-kind)
+              && needs.meta.outputs.release-pr-run-id
+              || needs.meta.outputs.build-tag
+            }}
+          TARGET_TAG: ${{ needs.meta.outputs.build-tag }}
           BRANCH: "${{ github.ref_name }}"
           DEV_ACR: "${{ vars.AZURE_DEV_REGISTRY_NAME }}"
           PROD_ACR: "${{ vars.AZURE_PROD_REGISTRY_NAME }}"
@@ -895,7 +906,7 @@ jobs:
 
   push-neon-image-dev:
     needs: [ meta, generate-image-maps, neon-image ]
-    if: ${{ contains(fromJSON('["push-main", "pr", "storage-release", "storage-rc-pr", "proxy-release", "proxy-rc-pr"]'), needs.meta.outputs.run-kind) }}
+    if: ${{ !failure() && !cancelled() && contains(fromJSON('["push-main", "pr", "storage-release", "storage-rc-pr", "proxy-release", "proxy-rc-pr"]'), needs.meta.outputs.run-kind) }}
     uses: ./.github/workflows/_push-to-container-registry.yml
     permissions:
       id-token: write  # Required for aws/azure login
@@ -913,7 +924,7 @@ jobs:
 
   push-compute-image-dev:
     needs: [ meta, generate-image-maps, vm-compute-node-image ]
-    if: ${{ contains(fromJSON('["push-main", "pr", "compute-release", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
+    if: ${{ !failure() && !cancelled() && contains(fromJSON('["push-main", "pr", "compute-release", "compute-rc-pr"]'), needs.meta.outputs.run-kind) }}
     uses: ./.github/workflows/_push-to-container-registry.yml
     permissions:
       id-token: write  # Required for aws/azure login
@@ -1175,7 +1186,7 @@ jobs:
               -f deployPgSniRouter=false \
               -f deployProxy=false \
               -f deployStorage=true \
-              -f deployStorageBroker=true \
+              -f deployStorageBroker=false \
               -f deployStorageController=true \
               -f branch=main \
               -f dockerTag=${{needs.meta.outputs.build-tag}} \
@@ -1183,7 +1194,7 @@ jobs:
 
             gh workflow --repo neondatabase/infra run deploy-prod.yml --ref main \
               -f deployStorage=true \
-              -f deployStorageBroker=true \
+              -f deployStorageBroker=false \
               -f deployStorageController=true \
               -f branch=main \
               -f dockerTag=${{needs.meta.outputs.build-tag}}
@@ -1231,11 +1242,11 @@ jobs:
           payload: |
             channel: ${{ vars.SLACK_STORAGE_CHANNEL_ID }}
             text: |
-              🔴 @oncall-storage: deploy job on release branch had unexpected status "${{ needs.deploy.result }}" <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>.
+              🔴 <!subteam^S06CJ87UMNY|@oncall-storage>: deploy job on release branch had unexpected status "${{ needs.deploy.result }}" <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>.
 
   # The job runs on `release` branch and copies compatibility data and Neon artifact from the last *release PR* to the latest directory
   promote-compatibility-data:
-    needs: [ deploy ]
+    needs: [ meta, deploy ]
     permissions:
       id-token: write # aws-actions/configure-aws-credentials
       statuses: write
@@ -1245,37 +1256,6 @@ jobs:
 
     runs-on: ubuntu-22.04
     steps:
-      - name: Fetch GITHUB_RUN_ID and COMMIT_SHA for the last merged release PR
-        id: fetch-last-release-pr-info
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          branch_name_and_pr_number=$(gh pr list \
-            --repo "${GITHUB_REPOSITORY}" \
-            --base release \
-            --state merged \
-            --limit 10 \
-            --json mergeCommit,headRefName,number \
-            --jq ".[] | select(.mergeCommit.oid==\"${GITHUB_SHA}\") | { branch_name: .headRefName, pr_number: .number }")
-          branch_name=$(echo "${branch_name_and_pr_number}" | jq -r '.branch_name')
-          pr_number=$(echo "${branch_name_and_pr_number}" | jq -r '.pr_number')
-
-          run_id=$(gh run list \
-            --repo "${GITHUB_REPOSITORY}" \
-            --workflow build_and_test.yml \
-            --branch "${branch_name}" \
-            --json databaseId \
-            --limit 1 \
-            --jq '.[].databaseId')
-
-          last_commit_sha=$(gh pr view "${pr_number}" \
-            --repo "${GITHUB_REPOSITORY}" \
-            --json commits \
-            --jq '.commits[-1].oid')
-
-          echo "run-id=${run_id}" | tee -a ${GITHUB_OUTPUT}
-          echo "commit-sha=${last_commit_sha}" | tee -a ${GITHUB_OUTPUT}
-
       - uses: aws-actions/configure-aws-credentials@v4
         with:
           aws-region: eu-central-1
@@ -1286,8 +1266,8 @@ jobs:
         env:
           BUCKET: neon-github-public-dev
           AWS_REGION: eu-central-1
-          COMMIT_SHA: ${{ steps.fetch-last-release-pr-info.outputs.commit-sha }}
-          RUN_ID: ${{ steps.fetch-last-release-pr-info.outputs.run-id }}
+          COMMIT_SHA: ${{ github.sha }}
+          RUN_ID: ${{ needs.meta.outputs.release-pr-run-id }}
         run: |
           old_prefix="artifacts/${COMMIT_SHA}/${RUN_ID}"
           new_prefix="artifacts/latest"
@@ -1376,5 +1356,5 @@ jobs:
           || needs.files-changed.result == 'skipped'
           || (needs.push-compute-image-dev.result == 'skipped' && contains(fromJSON('["push-main", "pr", "compute-release", "compute-rc-pr"]'), needs.meta.outputs.run-kind))
           || (needs.push-neon-image-dev.result == 'skipped' && contains(fromJSON('["push-main", "pr", "storage-release", "storage-rc-pr", "proxy-release", "proxy-rc-pr"]'), needs.meta.outputs.run-kind))
-          || needs.test-images.result == 'skipped'
+          || (needs.test-images.result == 'skipped' && contains(fromJSON('["push-main", "pr", "storage-rc-pr", "proxy-rc-pr", "compute-rc-pr"]'), needs.meta.outputs.run-kind))
           || (needs.trigger-custom-extensions-build-and-wait.result == 'skipped' && contains(fromJSON('["push-main", "pr", "compute-release", "compute-rc-pr"]'), needs.meta.outputs.run-kind))

.github/workflows/cargo-deny.yml

@@ -7,7 +7,7 @@ on:
         required: false
         type: string
   schedule:
-    - cron: '0 0 * * *'
+    - cron: '0 10 * * *'
 
 jobs:
   cargo-deny:
@@ -50,8 +50,9 @@ jobs:
           method: chat.postMessage
           token: ${{ secrets.SLACK_BOT_TOKEN }}
           payload: |
-            channel: ${{ vars.SLACK_CICD_CHANNEL_ID }}
+            channel: ${{ vars.SLACK_ON_CALL_DEVPROD_STREAM }}
             text: |
               Periodic cargo-deny on ${{ matrix.ref }}: ${{ job.status }}
               <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|GitHub Run>
-              Pinging @oncall-devprod.
+              Fixing the problem should be fairly straight forward from the logs. If not, <#${{ vars.SLACK_RUST_CHANNEL_ID }}> is there to help.
+              Pinging <!subteam^S0838JPSH32|@oncall-devprod>.

.github/workflows/fast-forward.yml

@@ -0,0 +1,36 @@
+name: Fast forward merge
+on:
+  pull_request:
+    types: [labeled]
+    branches:
+      - release
+      - release-proxy
+      - release-compute
+
+jobs:
+  fast-forward:
+    if: ${{ github.event.label.name == 'fast-forward' }}
+    runs-on: ubuntu-22.04
+
+    steps:
+      - name: Remove fast-forward label to PR
+        env:
+          GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
+        run: |
+          gh pr edit ${{ github.event.pull_request.number }} --repo "${GITHUB_REPOSITORY}" --remove-label "fast-forward"
+
+      - name: Fast forwarding
+        uses: sequoia-pgp/fast-forward@ea7628bedcb0b0b96e94383ada458d812fca4979
+        # See https://docs.github.com/en/graphql/reference/enums#mergestatestatus
+        if: ${{ github.event.pull_request.mergeable_state  == 'clean' }}
+        with:
+          merge: true
+          comment: on-error
+          github_token: ${{ secrets.CI_ACCESS_TOKEN }}
+
+      - name: Comment if mergeable_state is not clean
+        if: ${{ github.event.pull_request.mergeable_state  != 'clean' }}
+        run: |
+          gh pr comment ${{ github.event.pull_request.number }} \
+            --repo "${GITHUB_REPOSITORY}" \
+            --body "Not trying to forward pull-request, because \`mergeable_state\` is \`${{ github.event.pull_request.mergeable_state }}\`, not \`clean\`."

.github/workflows/lint-release-pr.yml

@@ -0,0 +1,24 @@
+name: Lint Release PR
+
+on:
+  pull_request:
+    branches:
+      - release
+      - release-proxy
+      - release-compute
+
+jobs:
+  lint-release-pr:
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Checkout PR branch
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0  # Fetch full history for git operations
+          ref: ${{ github.event.pull_request.head.ref }}
+
+      - name: Run lint script
+        env:
+          RELEASE_BRANCH: ${{ github.base_ref }}
+        run: |
+          ./.github/scripts/lint-release-pr.sh

.github/workflows/pre-merge-checks.yml

@@ -8,8 +8,6 @@ on:
       - .github/workflows/build-build-tools-image.yml
       - .github/workflows/pre-merge-checks.yml
   merge_group:
-    branches:
-      - main
 
 defaults:
   run:
@@ -19,11 +17,13 @@ defaults:
 permissions: {}
 
 jobs:
-  get-changed-files:
+  meta:
     runs-on: ubuntu-22.04
     outputs:
       python-changed: ${{ steps.python-src.outputs.any_changed }}
       rust-changed: ${{ steps.rust-src.outputs.any_changed }}
+      branch: ${{ steps.group-metadata.outputs.branch }}
+      pr-number: ${{ steps.group-metadata.outputs.pr-number }}
     steps:
       - uses: actions/checkout@v4
 
@@ -58,12 +58,20 @@ jobs:
           echo "${PYTHON_CHANGED_FILES}"
           echo "${RUST_CHANGED_FILES}"
 
+      - name: Merge group metadata
+        if: ${{ github.event_name == 'merge_group' }}
+        id: group-metadata
+        env:
+          MERGE_QUEUE_REF: ${{ github.event.merge_group.head_ref }}
+        run: |
+          echo $MERGE_QUEUE_REF | jq -Rr 'capture("refs/heads/gh-readonly-queue/(?<branch>.*)/pr-(?<pr_number>[0-9]+)-[0-9a-f]{40}") | ["branch=" + .branch, "pr-number=" + .pr_number] | .[]' | tee -a "${GITHUB_OUTPUT}"
+
   build-build-tools-image:
     if: |
       false
-      || needs.get-changed-files.outputs.python-changed == 'true'
-      || needs.get-changed-files.outputs.rust-changed == 'true'
-    needs: [ get-changed-files ]
+      || needs.meta.outputs.python-changed == 'true'
+      || needs.meta.outputs.rust-changed == 'true'
+    needs: [ meta ]
     uses: ./.github/workflows/build-build-tools-image.yml
     with:
       # Build only one combination to save time
@@ -72,8 +80,8 @@ jobs:
     secrets: inherit
 
   check-codestyle-python:
-    if: needs.get-changed-files.outputs.python-changed == 'true'
-    needs: [ get-changed-files, build-build-tools-image ]
+    if: needs.meta.outputs.python-changed == 'true'
+    needs: [ meta, build-build-tools-image ]
     uses: ./.github/workflows/_check-codestyle-python.yml
     with:
       # `-bookworm-x64` suffix should match the combination in `build-build-tools-image`
@@ -81,8 +89,8 @@ jobs:
     secrets: inherit
 
   check-codestyle-rust:
-    if: needs.get-changed-files.outputs.rust-changed == 'true'
-    needs: [ get-changed-files, build-build-tools-image ]
+    if: needs.meta.outputs.rust-changed == 'true'
+    needs: [ meta, build-build-tools-image ]
     uses: ./.github/workflows/_check-codestyle-rust.yml
     with:
       # `-bookworm-x64` suffix should match the combination in `build-build-tools-image`
@@ -101,7 +109,7 @@ jobs:
       statuses: write # for `github.repos.createCommitStatus(...)`
       contents: write
     needs:
-      - get-changed-files
+      - meta
       - check-codestyle-python
       - check-codestyle-rust
     runs-on: ubuntu-22.04
@@ -129,7 +137,20 @@ jobs:
         run: exit 1
         if: |
           false
-          || (needs.check-codestyle-python.result == 'skipped' && needs.get-changed-files.outputs.python-changed == 'true')
-          || (needs.check-codestyle-rust.result   == 'skipped' && needs.get-changed-files.outputs.rust-changed   == 'true')
+          || (github.event_name == 'merge_group' && needs.meta.outputs.branch != 'main')
+          || (needs.check-codestyle-python.result == 'skipped' && needs.meta.outputs.python-changed == 'true')
+          || (needs.check-codestyle-rust.result   == 'skipped' && needs.meta.outputs.rust-changed   == 'true')
           || contains(needs.*.result, 'failure')
           || contains(needs.*.result, 'cancelled')
+
+      - name: Add fast-forward label to PR to trigger fast-forward merge
+        if: >-
+          ${{
+            always()
+            && github.event_name == 'merge_group'
+            && contains(fromJson('["release", "release-proxy", "release-compute"]'), github.base_ref)
+          }}
+        env:
+          GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
+        run: >-
+          gh pr edit ${{ needs.meta.outputs.pr-number }} --repo "${GITHUB_REPOSITORY}" --add-label "fast-forward"

.github/workflows/release.yml

@@ -38,7 +38,7 @@ jobs:
     uses: ./.github/workflows/_create-release-pr.yml
     with:
       component-name: 'Storage'
-      release-branch: 'release'
+      source-branch: ${{ github.ref_name }}
     secrets:
       ci-access-token: ${{ secrets.CI_ACCESS_TOKEN }}
 
@@ -51,7 +51,7 @@ jobs:
     uses: ./.github/workflows/_create-release-pr.yml
     with:
       component-name: 'Proxy'
-      release-branch: 'release-proxy'
+      source-branch: ${{ github.ref_name }}
     secrets:
       ci-access-token: ${{ secrets.CI_ACCESS_TOKEN }}
 
@@ -64,6 +64,6 @@ jobs:
     uses: ./.github/workflows/_create-release-pr.yml
     with:
       component-name: 'Compute'
-      release-branch: 'release-compute'
+      source-branch: ${{ github.ref_name }}
     secrets:
       ci-access-token: ${{ secrets.CI_ACCESS_TOKEN }}

CODEOWNERS

@@ -3,7 +3,6 @@
 
 # DevProd & PerfCorr
 /.github/ @neondatabase/developer-productivity @neondatabase/performance-correctness
-/test_runner/	@neondatabase/performance-correctness
 
 # Compute
 /pgxn/ @neondatabase/compute

Cargo.lock

@@ -191,7 +191,7 @@ checksum = "965c2d33e53cb6b267e148a4cb0760bc01f4904c1cd4bb4002a085bb016d1490"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
  "synstructure",
 ]
 
@@ -203,7 +203,7 @@ checksum = "7b18050c2cd6fe86c3a76584ef5e0baf286d038cda203eb6223df2cc413565f7"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -272,7 +272,7 @@ checksum = "16e62a023e7c117e27523144c5d2459f4397fcc3cab0085af8e2224f643a0193"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -283,7 +283,7 @@ checksum = "b9ccdd8f2a161be9bd5c023df56f1b2a0bd1d83872ae53b71a84a12c9bf6e842"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -1021,7 +1021,7 @@ dependencies = [
  "regex",
  "rustc-hash 2.1.1",
  "shlex",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -1127,9 +1127,9 @@ checksum = "37b2a672a2cb129a2e41c10b1224bb368f9f37a2b16b612598138befd7b37eb5"
 
 [[package]]
 name = "cc"
-version = "1.1.30"
+version = "1.2.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b16803a61b81d9eabb7eae2588776c4c1e584b738ede45fdbb4c972cec1e9945"
+checksum = "be714c154be609ec7f5dad223a33bf1482fff90472de28f7362806e6d4832b8c"
 dependencies = [
  "jobserver",
  "libc",
@@ -1248,7 +1248,7 @@ dependencies = [
  "heck",
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -1703,7 +1703,7 @@ checksum = "f46882e17999c6cc590af592290432be3bce0428cb0d5f8b6715e4dc7b383eb3"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -1727,7 +1727,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "strsim 0.10.0",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -1738,7 +1738,7 @@ checksum = "29a358ff9f12ec09c3e61fef9b5a9902623a695a46a917b07f269bff1445611a"
 dependencies = [
  "darling_core",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -1888,7 +1888,7 @@ dependencies = [
  "dsl_auto_type",
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -1908,7 +1908,7 @@ version = "0.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "209c735641a413bc68c4923a9d6ad4bcb3ca306b794edaa7eb0b3228a99ffb25"
 dependencies = [
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -1937,7 +1937,7 @@ checksum = "487585f4d0c6655fe74905e2504d8ad6908e4db67f744eb140876906c2f3175d"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -1960,7 +1960,7 @@ dependencies = [
  "heck",
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -2105,7 +2105,7 @@ dependencies = [
  "darling",
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -2115,28 +2115,19 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "186e05a59d4c50738528153b83b0b0194d3a29507dfec16eccd4b342903397d0"
 dependencies = [
  "log",
-]
-
-[[package]]
-name = "env_logger"
-version = "0.10.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4cd405aab171cb85d6735e5c8d9db038c17d3ca007a4d2c25f337935c3d90580"
-dependencies = [
- "humantime",
- "is-terminal",
- "log",
  "regex",
- "termcolor",
 ]
 
 [[package]]
 name = "env_logger"
-version = "0.11.2"
+version = "0.11.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6c012a26a7f605efc424dd53697843a72be7dc86ad2d01f7814337794a12231d"
+checksum = "c3716d7a920fb4fac5d84e9d4bce8ceb321e9414b4409da61b07b75c1e3d0697"
 dependencies = [
+ "anstream",
+ "anstyle",
  "env_filter",
+ "jiff",
  "log",
 ]
 
@@ -2157,7 +2148,7 @@ checksum = "3bf679796c0322556351f287a51b49e48f7c4986e727b5dd78c972d30e2e16cc"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -2417,7 +2408,7 @@ checksum = "162ee34ebcb7c64a8abebc059ce0fee27c2262618d7b60ed8faf72fef13c3650"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -2530,7 +2521,7 @@ checksum = "53010ccb100b96a67bc32c0175f0ed1426b31b655d562898e57325f81c023ac0"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -2848,6 +2839,7 @@ dependencies = [
  "anyhow",
  "bytes",
  "fail",
+ "futures",
  "hyper 0.14.30",
  "itertools 0.10.5",
  "jemalloc_pprof",
@@ -2861,6 +2853,7 @@ dependencies = [
  "serde_path_to_error",
  "thiserror 1.0.69",
  "tokio",
+ "tokio-rustls 0.26.0",
  "tokio-stream",
  "tokio-util",
  "tracing",
@@ -3146,7 +3139,7 @@ checksum = "1ec89e9337638ecdc08744df490b221a7399bf8d164eb52a665454e60e075ad6"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -3239,7 +3232,7 @@ dependencies = [
  "crossbeam-channel",
  "crossbeam-utils",
  "dashmap 6.1.0",
- "env_logger 0.11.2",
+ "env_logger",
  "indexmap 2.0.1",
  "itoa",
  "log",
@@ -3252,11 +3245,11 @@ dependencies = [
 
 [[package]]
 name = "inotify"
-version = "0.11.0"
+version = "0.9.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f37dccff2791ab604f9babef0ba14fbe0be30bd368dc541e2b08d07c8aa908f3"
+checksum = "f8069d3ec154eb856955c1c0fbffefbf5f3c40a104ec912d4797314c1801abff"
 dependencies = [
- "bitflags 2.8.0",
+ "bitflags 1.3.2",
  "inotify-sys",
  "libc",
 ]
@@ -3362,6 +3355,30 @@ dependencies = [
  "tracing",
 ]
 
+[[package]]
+name = "jiff"
+version = "0.2.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d699bc6dfc879fb1bf9bdff0d4c56f0884fc6f0d0eb0fba397a6d00cd9a6b85e"
+dependencies = [
+ "jiff-static",
+ "log",
+ "portable-atomic",
+ "portable-atomic-util",
+ "serde",
+]
+
+[[package]]
+name = "jiff-static"
+version = "0.2.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8d16e75759ee0aa64c57a56acbf43916987b20c77373cb7e808979e02b93c9f9"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.100",
+]
+
 [[package]]
 name = "jobserver"
 version = "0.1.32"
@@ -3533,9 +3550,9 @@ dependencies = [
 
 [[package]]
 name = "log"
-version = "0.4.20"
+version = "0.4.26"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b5e6163cb8c49088c2c36f57875e58ccd8c87c7427f7fbd50ea6710b2f3f2e8f"
+checksum = "30bde2b3dc3671ae49d8e2e9f044c7c005836e7a023ee57cffa25ab82764bb9e"
 
 [[package]]
 name = "lru"
@@ -3616,7 +3633,7 @@ dependencies = [
  "heck",
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -3730,6 +3747,18 @@ dependencies = [
  "adler2",
 ]
 
+[[package]]
+name = "mio"
+version = "0.8.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a4a650543ca06a924e8b371db273b2756685faae30f8487da1b56505a8f78b0c"
+dependencies = [
+ "libc",
+ "log",
+ "wasi 0.11.0+wasi-snapshot-preview1",
+ "windows-sys 0.48.0",
+]
+
 [[package]]
 name = "mio"
 version = "1.0.3"
@@ -3737,7 +3766,6 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2886843bf800fba2e3377cff24abf6379b4c4d5c6681eaf9ea5b0d15090450bd"
 dependencies = [
  "libc",
- "log",
  "wasi 0.11.0+wasi-snapshot-preview1",
  "windows-sys 0.52.0",
 ]
@@ -3815,29 +3843,23 @@ checksum = "38bf9645c8b145698bb0b18a4637dcacbc421ea49bef2317e4fd8065a387cf21"
 
 [[package]]
 name = "notify"
-version = "8.0.0"
+version = "6.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2fee8403b3d66ac7b26aee6e40a897d85dc5ce26f44da36b8b73e987cc52e943"
+checksum = "6205bd8bb1e454ad2e27422015fb5e4f2bcc7e08fa8f27058670d208324a4d2d"
 dependencies = [
  "bitflags 2.8.0",
+ "crossbeam-channel",
  "filetime",
  "fsevent-sys",
  "inotify",
  "kqueue",
  "libc",
  "log",
- "mio",
- "notify-types",
+ "mio 0.8.11",
  "walkdir",
- "windows-sys 0.59.0",
+ "windows-sys 0.48.0",
 ]
 
-[[package]]
-name = "notify-types"
-version = "2.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5e0826a989adedc2a244799e823aece04662b66609d96af8dff7ac6df9a8925d"
-
 [[package]]
 name = "ntapi"
 version = "0.4.1"
@@ -4062,7 +4084,7 @@ dependencies = [
  "opentelemetry-http",
  "opentelemetry-proto",
  "opentelemetry_sdk",
- "prost",
+ "prost 0.13.3",
  "reqwest",
  "thiserror 1.0.69",
 ]
@@ -4075,7 +4097,7 @@ checksum = "a6e05acbfada5ec79023c85368af14abd0b307c015e9064d249b2a950ef459a6"
 dependencies = [
  "opentelemetry",
  "opentelemetry_sdk",
- "prost",
+ "prost 0.13.3",
  "tonic",
 ]
 
@@ -4189,6 +4211,7 @@ dependencies = [
  "pageserver_api",
  "pageserver_client",
  "rand 0.8.5",
+ "reqwest",
  "serde",
  "serde_json",
  "tokio",
@@ -4278,6 +4301,9 @@ dependencies = [
  "remote_storage",
  "reqwest",
  "rpds",
+ "rustls 0.23.18",
+ "rustls-pemfile 2.1.1",
+ "rustls-pki-types",
  "scopeguard",
  "send-future",
  "serde",
@@ -4296,11 +4322,13 @@ dependencies = [
  "tokio-epoll-uring",
  "tokio-io-timeout",
  "tokio-postgres",
+ "tokio-rustls 0.26.0",
  "tokio-stream",
  "tokio-tar",
  "tokio-util",
  "toml_edit",
  "tracing",
+ "tracing-utils",
  "url",
  "utils",
  "uuid",
@@ -4478,7 +4506,7 @@ dependencies = [
  "parquet",
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -4580,7 +4608,7 @@ checksum = "f6e859e6e5bd50440ab63c47e3ebabc90f26251f7c73c3d3e837b74a1cc3fa67"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -4676,6 +4704,15 @@ version = "1.10.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "280dc24453071f1b63954171985a0b0d30058d287960968b9b2aca264c8d4ee6"
 
+[[package]]
+name = "portable-atomic-util"
+version = "0.2.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d8a2f0d8d040d7848a709caf78912debcc3f33ee4b3cac47d73d1e1069e83507"
+dependencies = [
+ "portable-atomic",
+]
+
 [[package]]
 name = "postgres"
 version = "0.19.7"
@@ -4783,7 +4820,7 @@ dependencies = [
  "bytes",
  "crc32c",
  "criterion",
- "env_logger 0.10.2",
+ "env_logger",
  "log",
  "memoffset 0.9.0",
  "once_cell",
@@ -4830,8 +4867,10 @@ dependencies = [
  "nix 0.26.4",
  "once_cell",
  "parking_lot 0.12.1",
- "protobuf",
- "protobuf-codegen-pure",
+ "prost 0.12.6",
+ "prost-build 0.12.6",
+ "prost-derive 0.12.6",
+ "sha2",
  "smallvec",
  "symbolic-demangle",
  "tempfile",
@@ -4850,7 +4889,7 @@ dependencies = [
  "inferno 0.12.0",
  "num",
  "paste",
- "prost",
+ "prost 0.13.3",
 ]
 
 [[package]]
@@ -4880,7 +4919,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8d3928fb5db768cb86f891ff014f0144589297e3c6a1aba6ed7cecfdace270c7"
 dependencies = [
  "proc-macro2",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -4894,9 +4933,9 @@ dependencies = [
 
 [[package]]
 name = "proc-macro2"
-version = "1.0.92"
+version = "1.0.94"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "37d3544b3f2748c54e147655edb5025752e2303145b5aefb3c3ea2c78b973bb0"
+checksum = "a31971752e70b8b2686d7e46ec17fb38dad4051d94024c88df49b667caea9c84"
 dependencies = [
  "unicode-ident",
 ]
@@ -4943,6 +4982,16 @@ dependencies = [
  "thiserror 1.0.69",
 ]
 
+[[package]]
+name = "prost"
+version = "0.12.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "deb1435c188b76130da55f17a466d252ff7b1418b2ad3e037d127b94e3411f29"
+dependencies = [
+ "bytes",
+ "prost-derive 0.12.6",
+]
+
 [[package]]
 name = "prost"
 version = "0.13.3"
@@ -4950,7 +4999,28 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7b0487d90e047de87f984913713b85c601c05609aad5b0df4b4573fbf69aa13f"
 dependencies = [
  "bytes",
- "prost-derive",
+ "prost-derive 0.13.3",
+]
+
+[[package]]
+name = "prost-build"
+version = "0.12.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "22505a5c94da8e3b7c2996394d1c933236c4d743e81a410bcca4e6989fc066a4"
+dependencies = [
+ "bytes",
+ "heck",
+ "itertools 0.12.1",
+ "log",
+ "multimap",
+ "once_cell",
+ "petgraph",
+ "prettyplease",
+ "prost 0.12.6",
+ "prost-types 0.12.6",
+ "regex",
+ "syn 2.0.100",
+ "tempfile",
 ]
 
 [[package]]
@@ -4967,13 +5037,26 @@ dependencies = [
  "once_cell",
  "petgraph",
  "prettyplease",
- "prost",
- "prost-types",
+ "prost 0.13.3",
+ "prost-types 0.13.3",
  "regex",
- "syn 2.0.90",
+ "syn 2.0.100",
  "tempfile",
 ]
 
+[[package]]
+name = "prost-derive"
+version = "0.12.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "81bddcdb20abf9501610992b6759a4c888aef7d1a7247ef75e2404275ac24af1"
+dependencies = [
+ "anyhow",
+ "itertools 0.12.1",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.100",
+]
+
 [[package]]
 name = "prost-derive"
 version = "0.13.3"
@@ -4984,7 +5067,16 @@ dependencies = [
  "itertools 0.12.1",
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
+]
+
+[[package]]
+name = "prost-types"
+version = "0.12.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9091c90b0a32608e984ff2fa4091273cbdd755d54935c51d520887f4a1dbd5b0"
+dependencies = [
+ "prost 0.12.6",
 ]
 
 [[package]]
@@ -4993,32 +5085,7 @@ version = "0.13.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4759aa0d3a6232fb8dbdb97b61de2c20047c68aca932c7ed76da9d788508d670"
 dependencies = [
- "prost",
-]
-
-[[package]]
-name = "protobuf"
-version = "2.28.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "106dd99e98437432fed6519dedecfade6a06a73bb7b2a1e019fdd2bee5778d94"
-
-[[package]]
-name = "protobuf-codegen"
-version = "2.28.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "033460afb75cf755fcfc16dfaed20b86468082a2ea24e05ac35ab4a099a017d6"
-dependencies = [
- "protobuf",
-]
-
-[[package]]
-name = "protobuf-codegen-pure"
-version = "2.28.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "95a29399fc94bcd3eeaa951c715f7bea69409b2445356b00519740bcd6ddd865"
-dependencies = [
- "protobuf",
- "protobuf-codegen",
+ "prost 0.13.3",
 ]
 
 [[package]]
@@ -5047,7 +5114,7 @@ dependencies = [
  "consumption_metrics",
  "ecdsa 0.16.9",
  "ed25519-dalek",
- "env_logger 0.10.2",
+ "env_logger",
  "fallible-iterator",
  "flate2",
  "framed-websockets",
@@ -5184,9 +5251,9 @@ dependencies = [
 
 [[package]]
 name = "quote"
-version = "1.0.37"
+version = "1.0.39"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b5b9d34b8991d19d98081b46eacdd8eb58c6f2b201139f7c5f643cc155a633af"
+checksum = "c1f1914ce909e1658d9907913b4b91947430c7d9be598b15a1912935b8c04801"
 dependencies = [
  "proc-macro2",
 ]
@@ -5627,16 +5694,16 @@ dependencies = [
 
 [[package]]
 name = "ring"
-version = "0.17.6"
+version = "0.17.13"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "684d5e6e18f669ccebf64a92236bb7db9a34f07be010e3627368182027180866"
+checksum = "70ac5d832aa16abd7d1def883a8545280c20a60f523a370aa3a9617c2b8550ee"
 dependencies = [
  "cc",
+ "cfg-if",
  "getrandom 0.2.11",
  "libc",
- "spin",
  "untrusted",
- "windows-sys 0.48.0",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -5715,7 +5782,7 @@ dependencies = [
  "regex",
  "relative-path",
  "rustc_version",
- "syn 2.0.90",
+ "syn 2.0.100",
  "unicode-ident",
 ]
 
@@ -5878,9 +5945,9 @@ dependencies = [
 
 [[package]]
 name = "rustls-pki-types"
-version = "1.10.0"
+version = "1.11.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "16f1201b3c9a7ee8039bcadc17b7e605e2945b27eee7631788c1bd2b0643674b"
+checksum = "917ce264624a4b4db1c364dcc35bfca9ded014d0a958cd47ad3e960e988ea51c"
 
 [[package]]
 name = "rustls-webpki"
@@ -5930,7 +5997,7 @@ dependencies = [
  "crc32c",
  "criterion",
  "desim",
- "env_logger 0.10.2",
+ "env_logger",
  "fail",
  "futures",
  "hex",
@@ -6261,7 +6328,7 @@ checksum = "ad1e866f866923f252f05c889987993144fb74e722403468a4ebd70c3cd756c0"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -6343,7 +6410,7 @@ dependencies = [
  "darling",
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -6566,7 +6633,7 @@ dependencies = [
  "metrics",
  "once_cell",
  "parking_lot 0.12.1",
- "prost",
+ "prost 0.13.3",
  "rustls 0.23.18",
  "tokio",
  "tonic",
@@ -6584,6 +6651,7 @@ dependencies = [
  "bytes",
  "chrono",
  "clap",
+ "clashmap",
  "control_plane",
  "cron",
  "diesel",
@@ -6744,7 +6812,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "rustversion",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -6795,9 +6863,9 @@ dependencies = [
 
 [[package]]
 name = "syn"
-version = "2.0.90"
+version = "2.0.100"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "919d3b74a5dd0ccd15aeb8f93e7006bd9e14c295087c9896a110f490752bcf31"
+checksum = "b09a44accad81e1ba1cd74a32461ba89dee89095ba17b32f5d03683b1b1fc2a0"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -6827,7 +6895,7 @@ checksum = "c8af7666ab7b6390ab78131fb5b0fce11d6b7a6951602017c35fa82800708971"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -6878,15 +6946,6 @@ dependencies = [
  "serde_json",
 ]
 
-[[package]]
-name = "termcolor"
-version = "1.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "be55cf8942feac5c765c2c993422806843c9a9a45d4d5c407ad6dd2ea95eb9b6"
-dependencies = [
- "winapi-util",
-]
-
 [[package]]
 name = "test-context"
 version = "0.3.0"
@@ -6905,7 +6964,7 @@ checksum = "78ea17a2dc368aeca6f554343ced1b1e31f76d63683fa8016e5844bd7a5144a1"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -6934,7 +6993,7 @@ checksum = "4fee6c4efc90059e10f81e6d42c60a18f76588c3d74cb83a0b242a2b6c7504c1"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -6945,7 +7004,7 @@ checksum = "26afc1baea8a989337eeb52b6e72a039780ce45c3edfcc9c5b9d112feeb173c2"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -7085,7 +7144,7 @@ dependencies = [
  "backtrace",
  "bytes",
  "libc",
- "mio",
+ "mio 1.0.3",
  "parking_lot 0.12.1",
  "pin-project-lite",
  "signal-hook-registry",
@@ -7128,7 +7187,7 @@ checksum = "6e06d43f1345a3bcd39f6a56dbb7dcab2ba47e68e8ac134855e7e2bdbaf8cab8"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -7338,7 +7397,7 @@ dependencies = [
  "hyper-util",
  "percent-encoding",
  "pin-project",
- "prost",
+ "prost 0.13.3",
  "rustls-native-certs 0.8.0",
  "rustls-pemfile 2.1.1",
  "tokio",
@@ -7358,10 +7417,10 @@ checksum = "9557ce109ea773b399c9b9e5dca39294110b74f1f342cb347a80d1fce8c26a11"
 dependencies = [
  "prettyplease",
  "proc-macro2",
- "prost-build",
- "prost-types",
+ "prost-build 0.13.3",
+ "prost-types 0.13.3",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -7476,7 +7535,7 @@ checksum = "395ae124c09f9e6918a2310af6038fba074bcf474ac352496d5910dd59a2226d"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -7807,6 +7866,7 @@ dependencies = [
  "tracing",
  "tracing-error",
  "tracing-subscriber",
+ "tracing-utils",
  "walkdir",
 ]
 
@@ -7870,7 +7930,7 @@ dependencies = [
  "anyhow",
  "camino-tempfile",
  "clap",
- "env_logger 0.10.2",
+ "env_logger",
  "log",
  "postgres",
  "postgres_ffi",
@@ -7892,7 +7952,7 @@ dependencies = [
  "pageserver_api",
  "postgres_ffi",
  "pprof",
- "prost",
+ "prost 0.13.3",
  "remote_storage",
  "serde",
  "serde_json",
@@ -7975,7 +8035,7 @@ dependencies = [
  "once_cell",
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
  "wasm-bindgen-shared",
 ]
 
@@ -8009,7 +8069,7 @@ checksum = "e94f17b526d0a461a191c78ea52bbce64071ed5c04c9ffe424dcb38f74171bb7"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
  "wasm-bindgen-backend",
  "wasm-bindgen-shared",
 ]
@@ -8316,6 +8376,7 @@ name = "workspace_hack"
 version = "0.1.0"
 dependencies = [
  "ahash",
+ "anstream",
  "anyhow",
  "base64 0.13.1",
  "base64 0.21.7",
@@ -8332,6 +8393,8 @@ dependencies = [
  "digest",
  "displaydoc",
  "either",
+ "env_filter",
+ "env_logger",
  "fail",
  "form_urlencoded",
  "futures-channel",
@@ -8367,7 +8430,7 @@ dependencies = [
  "parquet",
  "prettyplease",
  "proc-macro2",
- "prost",
+ "prost 0.13.3",
  "quote",
  "rand 0.8.5",
  "regex",
@@ -8384,7 +8447,7 @@ dependencies = [
  "spki 0.7.3",
  "stable_deref_trait",
  "subtle",
- "syn 2.0.90",
+ "syn 2.0.100",
  "sync_wrapper 0.1.2",
  "tikv-jemalloc-ctl",
  "tikv-jemalloc-sys",
@@ -8501,7 +8564,7 @@ checksum = "2380878cad4ac9aac1e2435f3eb4020e8374b5f13c296cb75b4620ff8e229154"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
  "synstructure",
 ]
 
@@ -8523,7 +8586,7 @@ checksum = "b3c129550b3e6de3fd0ba67ba5c81818f9805e58b8d7fee80a3a59d2c9fc601a"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -8543,7 +8606,7 @@ checksum = "595eed982f7d355beb85837f651fa22e90b3c044842dc7f2c2842c086f295808"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
  "synstructure",
 ]
 
@@ -8565,7 +8628,7 @@ checksum = "ce36e65b0d2999d2aafac989fb249189a141aee1f53c612c1f37d72631959f69"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]
@@ -8587,7 +8650,7 @@ checksum = "6eafa6dfb17584ea3e2bd6e76e0cc15ad7af12b09abdd1ca55961bed9b1063c6"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.90",
+ "syn 2.0.100",
 ]
 
 [[package]]

Cargo.toml

@@ -126,7 +126,9 @@ measured = { version = "0.0.22", features=["lasso"] }
 measured-process = { version = "0.0.22" }
 memoffset = "0.9"
 nix = { version = "0.27", features = ["dir", "fs", "process", "socket", "signal", "poll"] }
-notify = "8.0.0"
+# Do not update to >= 7.0.0, at least. The update will have a significant impact
+# on compute startup metrics (start_postgres_ms), >= 25% degradation.
+notify = "6.0.0"
 num_cpus = "1.15"
 num-traits = "0.2.15"
 once_cell = "1.13"
@@ -139,7 +141,7 @@ parquet = { version = "53", default-features = false, features = ["zstd"] }
 parquet_derive = "53"
 pbkdf2 = { version = "0.12.1", features = ["simple", "std"] }
 pin-project-lite = "0.2"
-pprof = { version = "0.14", features = ["criterion", "flamegraph", "frame-pointer", "protobuf", "protobuf-codec"] }
+pprof = { version = "0.14", features = ["criterion", "flamegraph", "frame-pointer", "prost-codec"] }
 procfs = "0.16"
 prometheus = {version = "0.13", default-features=false, features = ["process"]} # removes protobuf dependency
 prost = "0.13"
@@ -155,6 +157,7 @@ rpds = "0.13"
 rustc-hash = "1.1.0"
 rustls = { version = "0.23.16", default-features = false }
 rustls-pemfile = "2"
+rustls-pki-types = "1.11"
 scopeguard = "1.1"
 sysinfo = "0.29.2"
 sd-notify = "0.4.1"
@@ -218,7 +221,7 @@ zerocopy = { version = "0.7", features = ["derive"] }
 json-structural-diff = { version = "0.2.0" }
 
 ## TODO replace this with tracing
-env_logger = "0.10"
+env_logger = "0.11"
 log = "0.4"
 
 ## Libraries from neondatabase/ git forks, ideally with changes to be upstreamed

compute_tools/src/compute.rs

@@ -645,9 +645,9 @@ impl ComputeNode {
         if pspec.spec.mode == ComputeMode::Primary {
             self.configure_as_primary(&compute_state)?;
 
-            let conf = self.get_conn_conf(None);
-            tokio::task::spawn_blocking(|| {
-                let res = get_installed_extensions(conf);
+            let conf = self.get_tokio_conn_conf(None);
+            tokio::task::spawn(async {
+                let res = get_installed_extensions(conf).await;
                 match res {
                     Ok(extensions) => {
                         info!(

compute_tools/src/extension_server.rs

@@ -202,8 +202,24 @@ pub async fn download_extension(
     // move contents of the libdir / sharedir in unzipped archive to the correct local paths
     for paths in [sharedir_paths, libdir_paths] {
         let (zip_dir, real_dir) = paths;
+
+        let dir = match std::fs::read_dir(&zip_dir) {
+            Ok(dir) => dir,
+            Err(e) => match e.kind() {
+                // In the event of a SQL-only extension, there would be nothing
+                // to move from the lib/ directory, so note that in the log and
+                // move on.
+                std::io::ErrorKind::NotFound => {
+                    info!("nothing to move from {}", zip_dir);
+                    continue;
+                }
+                _ => return Err(anyhow::anyhow!(e)),
+            },
+        };
+
         info!("mv {zip_dir:?}/*  {real_dir:?}");
-        for file in std::fs::read_dir(zip_dir)? {
+
+        for file in dir {
             let old_file = file?.path();
             let new_file =
                 Path::new(&real_dir).join(old_file.file_name().context("error parsing file")?);

compute_tools/src/http/middleware/mod.rs

@@ -1 +1,2 @@
 pub(in crate::http) mod authorize;
+pub(in crate::http) mod request_id;

compute_tools/src/http/middleware/request_id.rs

@@ -0,0 +1,16 @@
+use axum::{extract::Request, middleware::Next, response::Response};
+use uuid::Uuid;
+
+use crate::http::headers::X_REQUEST_ID;
+
+/// This middleware function allows compute_ctl to generate its own request ID
+/// if one isn't supplied. The control plane will always send one as a UUID. The
+/// neon Postgres extension on the other hand does not send one.
+pub async fn maybe_add_request_id_header(mut request: Request, next: Next) -> Response {
+    let headers = request.headers_mut();
+    if !headers.contains_key(X_REQUEST_ID) {
+        headers.append(X_REQUEST_ID, Uuid::new_v4().to_string().parse().unwrap());
+    }
+
+    next.run(request).await
+}

compute_tools/src/http/server.rs

@@ -5,9 +5,8 @@ use std::time::Duration;
 
 use anyhow::Result;
 use axum::Router;
-use axum::extract::Request;
-use axum::middleware::{self, Next};
-use axum::response::{IntoResponse, Response};
+use axum::middleware::{self};
+use axum::response::IntoResponse;
 use axum::routing::{get, post};
 use http::StatusCode;
 use jsonwebtoken::jwk::JwkSet;
@@ -17,8 +16,8 @@ use tower_http::{
     auth::AsyncRequireAuthorizationLayer, request_id::PropagateRequestIdLayer, trace::TraceLayer,
 };
 use tracing::{Span, error, info};
-use uuid::Uuid;
 
+use super::middleware::request_id::maybe_add_request_id_header;
 use super::{
     headers::X_REQUEST_ID,
     middleware::authorize::Authorize,
@@ -219,15 +218,3 @@ impl Server {
         tokio::spawn(self.serve(state));
     }
 }
-
-/// This middleware function allows compute_ctl to generate its own request ID
-/// if one isn't supplied. The control plane will always send one as a UUID. The
-/// neon Postgres extension on the other hand does not send one.
-async fn maybe_add_request_id_header(mut request: Request, next: Next) -> Response {
-    let headers = request.headers_mut();
-    if headers.get(X_REQUEST_ID).is_none() {
-        headers.append(X_REQUEST_ID, Uuid::new_v4().to_string().parse().unwrap());
-    }
-
-    next.run(request).await
-}

compute_tools/src/installed_extensions.rs

@@ -2,7 +2,7 @@ use std::collections::HashMap;
 
 use anyhow::Result;
 use compute_api::responses::{InstalledExtension, InstalledExtensions};
-use postgres::{Client, NoTls};
+use tokio_postgres::{Client, Config, NoTls};
 
 use crate::metrics::INSTALLED_EXTENSIONS;
 
@@ -10,7 +10,7 @@ use crate::metrics::INSTALLED_EXTENSIONS;
 /// and to make database listing query here more explicit.
 ///
 /// Limit the number of databases to 500 to avoid excessive load.
-fn list_dbs(client: &mut Client) -> Result<Vec<String>> {
+async fn list_dbs(client: &mut Client) -> Result<Vec<String>> {
     // `pg_database.datconnlimit = -2` means that the database is in the
     // invalid state
     let databases = client
@@ -20,7 +20,8 @@ fn list_dbs(client: &mut Client) -> Result<Vec<String>> {
                 AND datconnlimit <> - 2
                 LIMIT 500",
             &[],
-        )?
+        )
+        .await?
         .iter()
         .map(|row| {
             let db: String = row.get("datname");
@@ -36,20 +37,36 @@ fn list_dbs(client: &mut Client) -> Result<Vec<String>> {
 /// Same extension can be installed in multiple databases with different versions,
 /// so we report a separate metric (number of databases where it is installed)
 /// for each extension version.
-pub fn get_installed_extensions(mut conf: postgres::config::Config) -> Result<InstalledExtensions> {
+pub async fn get_installed_extensions(mut conf: Config) -> Result<InstalledExtensions> {
     conf.application_name("compute_ctl:get_installed_extensions");
-    let mut client = conf.connect(NoTls)?;
-    let databases: Vec<String> = list_dbs(&mut client)?;
+    let databases: Vec<String> = {
+        let (mut client, connection) = conf.connect(NoTls).await?;
+        tokio::spawn(async move {
+            if let Err(e) = connection.await {
+                eprintln!("connection error: {}", e);
+            }
+        });
+
+        list_dbs(&mut client).await?
+    };
 
     let mut extensions_map: HashMap<(String, String, String), InstalledExtension> = HashMap::new();
     for db in databases.iter() {
         conf.dbname(db);
-        let mut db_client = conf.connect(NoTls)?;
-        let extensions: Vec<(String, String, i32)> = db_client
+
+        let (client, connection) = conf.connect(NoTls).await?;
+        tokio::spawn(async move {
+            if let Err(e) = connection.await {
+                eprintln!("connection error: {}", e);
+            }
+        });
+
+        let extensions: Vec<(String, String, i32)> = client
             .query(
                 "SELECT extname, extversion, extowner::integer FROM pg_catalog.pg_extension",
                 &[],
-            )?
+            )
+            .await?
             .iter()
             .map(|row| {
                 (

compute_tools/src/logger.rs

@@ -24,7 +24,8 @@ pub async fn init_tracing_and_logging(default_log_level: &str) -> anyhow::Result
         .with_writer(std::io::stderr);
 
     // Initialize OpenTelemetry
-    let otlp_layer = tracing_utils::init_tracing("compute_ctl").await;
+    let otlp_layer =
+        tracing_utils::init_tracing("compute_ctl", tracing_utils::ExportConfig::default()).await;
 
     // Put it all together
     tracing_subscriber::registry()

compute_tools/src/sql/pre_drop_role_revoke_privileges.sql

@@ -1,8 +1,7 @@
-SET SESSION ROLE neon_superuser;
-
 DO ${outer_tag}$
 DECLARE
     schema TEXT;
+    grantor TEXT;
     revoke_query TEXT;
 BEGIN
     FOR schema IN
@@ -15,16 +14,25 @@ BEGIN
         -- ii) it's easy to add more schemas to the list if needed.
         WHERE schema_name IN ('public')
     LOOP
-        revoke_query := format(
-            'REVOKE ALL PRIVILEGES ON ALL TABLES IN SCHEMA %I FROM %I GRANTED BY neon_superuser;',
-            schema,
-            -- N.B. this has to be properly dollar-escaped with `pg_quote_dollar()`
-            {role_name}
-        );
+        FOR grantor IN EXECUTE
+            format(
+                'SELECT DISTINCT rtg.grantor FROM information_schema.role_table_grants AS rtg WHERE grantee = %s',
+                -- N.B. this has to be properly dollar-escaped with `pg_quote_dollar()`
+                quote_literal({role_name})
+            )
+        LOOP
+            EXECUTE format('SET LOCAL ROLE %I', grantor);
 
-        EXECUTE revoke_query;
+            revoke_query := format(
+                'REVOKE ALL PRIVILEGES ON ALL TABLES IN SCHEMA %I FROM %I GRANTED BY %I',
+                schema,
+                -- N.B. this has to be properly dollar-escaped with `pg_quote_dollar()`
+                {role_name},
+                grantor
+            );
+
+            EXECUTE revoke_query;
+        END LOOP;
     END LOOP;
 END;
 ${outer_tag}$;
-
-RESET ROLE;

control_plane/src/bin/neon_local.rs

@@ -36,7 +36,9 @@ use pageserver_api::config::{
 use pageserver_api::controller_api::{
     NodeAvailabilityWrapper, PlacementPolicy, TenantCreateRequest,
 };
-use pageserver_api::models::{ShardParameters, TimelineCreateRequest, TimelineInfo};
+use pageserver_api::models::{
+    ShardParameters, TenantConfigRequest, TimelineCreateRequest, TimelineInfo,
+};
 use pageserver_api::shard::{ShardCount, ShardStripeSize, TenantShardId};
 use postgres_backend::AuthType;
 use postgres_connection::parse_host_port;
@@ -963,6 +965,7 @@ fn handle_init(args: &InitCmdArgs) -> anyhow::Result<LocalEnv> {
                         id: pageserver_id,
                         listen_pg_addr: format!("127.0.0.1:{pg_port}"),
                         listen_http_addr: format!("127.0.0.1:{http_port}"),
+                        listen_https_addr: None,
                         pg_auth_type: AuthType::Trust,
                         http_auth_type: AuthType::Trust,
                         other: Default::default(),
@@ -977,6 +980,7 @@ fn handle_init(args: &InitCmdArgs) -> anyhow::Result<LocalEnv> {
             default_tenant_id: TenantId::from_array(std::array::from_fn(|_| 0)),
             storage_controller: None,
             control_plane_compute_hook_api: None,
+            generate_local_ssl_certs: false,
         }
     };
 
@@ -1127,12 +1131,16 @@ async fn handle_tenant(subcmd: &TenantCmd, env: &mut local_env::LocalEnv) -> any
             let tenant_id = get_tenant_id(args.tenant_id, env)?;
             let tenant_conf: HashMap<_, _> =
                 args.config.iter().flat_map(|c| c.split_once(':')).collect();
+            let config = PageServerNode::parse_config(tenant_conf)?;
 
-            pageserver
-                .tenant_config(tenant_id, tenant_conf)
+            let req = TenantConfigRequest { tenant_id, config };
+
+            let storage_controller = StorageController::from_env(env);
+            storage_controller
+                .set_tenant_config(&req)
                 .await
                 .with_context(|| format!("Tenant config failed for tenant with id {tenant_id}"))?;
-            println!("tenant {tenant_id} successfully configured on the pageserver");
+            println!("tenant {tenant_id} successfully configured via storcon");
         }
     }
     Ok(())

control_plane/src/local_env.rs

@@ -81,6 +81,10 @@ pub struct LocalEnv {
     // but deserialization into a generic toml object as `toml::Value::try_from` fails with an error.
     // https://toml.io/en/v1.0.0 does not contain a concept of "a table inside another table".
     pub branch_name_mappings: HashMap<String, Vec<(TenantId, TimelineId)>>,
+
+    /// Flag to generate SSL certificates for components that need it.
+    /// Also generates root CA certificate that is used to sign all other certificates.
+    pub generate_local_ssl_certs: bool,
 }
 
 /// On-disk state stored in `.neon/config`.
@@ -102,6 +106,10 @@ pub struct OnDiskConfig {
     pub control_plane_api: Option<Url>,
     pub control_plane_compute_hook_api: Option<Url>,
     branch_name_mappings: HashMap<String, Vec<(TenantId, TimelineId)>>,
+    // Note: skip serializing because in compat tests old storage controller fails
+    // to load new config file. May be removed after this field is in release branch.
+    #[serde(skip_serializing_if = "std::ops::Not::not")]
+    pub generate_local_ssl_certs: bool,
 }
 
 fn fail_if_pageservers_field_specified<'de, D>(_: D) -> Result<Vec<PageServerConf>, D::Error>
@@ -129,6 +137,7 @@ pub struct NeonLocalInitConf {
     pub safekeepers: Vec<SafekeeperConf>,
     pub control_plane_api: Option<Url>,
     pub control_plane_compute_hook_api: Option<Option<Url>>,
+    pub generate_local_ssl_certs: bool,
 }
 
 /// Broker config for cluster internal communication.
@@ -165,6 +174,11 @@ pub struct NeonStorageControllerConf {
 
     #[serde(with = "humantime_serde")]
     pub long_reconcile_threshold: Option<Duration>,
+
+    #[serde(default)]
+    pub use_https_pageserver_api: bool,
+
+    pub timelines_onto_safekeepers: bool,
 }
 
 impl NeonStorageControllerConf {
@@ -188,6 +202,8 @@ impl Default for NeonStorageControllerConf {
             max_secondary_lag_bytes: None,
             heartbeat_interval: Self::DEFAULT_HEARTBEAT_INTERVAL,
             long_reconcile_threshold: None,
+            use_https_pageserver_api: false,
+            timelines_onto_safekeepers: false,
         }
     }
 }
@@ -217,6 +233,7 @@ pub struct PageServerConf {
     pub id: NodeId,
     pub listen_pg_addr: String,
     pub listen_http_addr: String,
+    pub listen_https_addr: Option<String>,
     pub pg_auth_type: AuthType,
     pub http_auth_type: AuthType,
     pub no_sync: bool,
@@ -228,6 +245,7 @@ impl Default for PageServerConf {
             id: NodeId(0),
             listen_pg_addr: String::new(),
             listen_http_addr: String::new(),
+            listen_https_addr: None,
             pg_auth_type: AuthType::Trust,
             http_auth_type: AuthType::Trust,
             no_sync: false,
@@ -243,6 +261,7 @@ pub struct NeonLocalInitPageserverConf {
     pub id: NodeId,
     pub listen_pg_addr: String,
     pub listen_http_addr: String,
+    pub listen_https_addr: Option<String>,
     pub pg_auth_type: AuthType,
     pub http_auth_type: AuthType,
     #[serde(default, skip_serializing_if = "std::ops::Not::not")]
@@ -257,6 +276,7 @@ impl From<&NeonLocalInitPageserverConf> for PageServerConf {
             id,
             listen_pg_addr,
             listen_http_addr,
+            listen_https_addr,
             pg_auth_type,
             http_auth_type,
             no_sync,
@@ -266,6 +286,7 @@ impl From<&NeonLocalInitPageserverConf> for PageServerConf {
             id: *id,
             listen_pg_addr: listen_pg_addr.clone(),
             listen_http_addr: listen_http_addr.clone(),
+            listen_https_addr: listen_https_addr.clone(),
             pg_auth_type: *pg_auth_type,
             http_auth_type: *http_auth_type,
             no_sync: *no_sync,
@@ -410,6 +431,41 @@ impl LocalEnv {
         }
     }
 
+    pub fn ssl_ca_cert_path(&self) -> Option<PathBuf> {
+        if self.generate_local_ssl_certs {
+            Some(self.base_data_dir.join("rootCA.crt"))
+        } else {
+            None
+        }
+    }
+
+    pub fn ssl_ca_key_path(&self) -> Option<PathBuf> {
+        if self.generate_local_ssl_certs {
+            Some(self.base_data_dir.join("rootCA.key"))
+        } else {
+            None
+        }
+    }
+
+    pub fn generate_ssl_ca_cert(&self) -> anyhow::Result<()> {
+        let cert_path = self.ssl_ca_cert_path().unwrap();
+        let key_path = self.ssl_ca_key_path().unwrap();
+        if !fs::exists(cert_path.as_path())? {
+            generate_ssl_ca_cert(cert_path.as_path(), key_path.as_path())?;
+        }
+        Ok(())
+    }
+
+    pub fn generate_ssl_cert(&self, cert_path: &Path, key_path: &Path) -> anyhow::Result<()> {
+        self.generate_ssl_ca_cert()?;
+        generate_ssl_cert(
+            cert_path,
+            key_path,
+            self.ssl_ca_cert_path().unwrap().as_path(),
+            self.ssl_ca_key_path().unwrap().as_path(),
+        )
+    }
+
     /// 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)>> {
@@ -519,6 +575,7 @@ impl LocalEnv {
                 control_plane_api,
                 control_plane_compute_hook_api,
                 branch_name_mappings,
+                generate_local_ssl_certs,
             } = on_disk_config;
             LocalEnv {
                 base_data_dir: repopath.to_owned(),
@@ -533,6 +590,7 @@ impl LocalEnv {
                 control_plane_api: control_plane_api.unwrap(),
                 control_plane_compute_hook_api,
                 branch_name_mappings,
+                generate_local_ssl_certs,
             }
         };
 
@@ -568,6 +626,7 @@ impl LocalEnv {
                 struct PageserverConfigTomlSubset {
                     listen_pg_addr: String,
                     listen_http_addr: String,
+                    listen_https_addr: Option<String>,
                     pg_auth_type: AuthType,
                     http_auth_type: AuthType,
                     #[serde(default)]
@@ -592,6 +651,7 @@ impl LocalEnv {
                 let PageserverConfigTomlSubset {
                     listen_pg_addr,
                     listen_http_addr,
+                    listen_https_addr,
                     pg_auth_type,
                     http_auth_type,
                     no_sync,
@@ -609,6 +669,7 @@ impl LocalEnv {
                     },
                     listen_pg_addr,
                     listen_http_addr,
+                    listen_https_addr,
                     pg_auth_type,
                     http_auth_type,
                     no_sync,
@@ -636,6 +697,7 @@ impl LocalEnv {
                 control_plane_api: Some(self.control_plane_api.clone()),
                 control_plane_compute_hook_api: self.control_plane_compute_hook_api.clone(),
                 branch_name_mappings: self.branch_name_mappings.clone(),
+                generate_local_ssl_certs: self.generate_local_ssl_certs,
             },
         )
     }
@@ -718,6 +780,7 @@ impl LocalEnv {
             safekeepers,
             control_plane_api,
             control_plane_compute_hook_api,
+            generate_local_ssl_certs,
         } = conf;
 
         // Find postgres binaries.
@@ -766,8 +829,13 @@ impl LocalEnv {
             control_plane_api: control_plane_api.unwrap(),
             control_plane_compute_hook_api: control_plane_compute_hook_api.unwrap_or_default(),
             branch_name_mappings: Default::default(),
+            generate_local_ssl_certs,
         };
 
+        if generate_local_ssl_certs {
+            env.generate_ssl_ca_cert()?;
+        }
+
         // create endpoints dir
         fs::create_dir_all(env.endpoints_path())?;
 
@@ -851,3 +919,80 @@ fn generate_auth_keys(private_key_path: &Path, public_key_path: &Path) -> anyhow
     }
     Ok(())
 }
+
+fn generate_ssl_ca_cert(cert_path: &Path, key_path: &Path) -> anyhow::Result<()> {
+    // openssl req -x509 -newkey rsa:2048 -nodes -subj "/CN=Neon Local CA" -days 36500 \
+    // -out rootCA.crt -keyout rootCA.key
+    let keygen_output = Command::new("openssl")
+        .args([
+            "req", "-x509", "-newkey", "rsa:2048", "-nodes", "-days", "36500",
+        ])
+        .args(["-subj", "/CN=Neon Local CA"])
+        .args(["-out", cert_path.to_str().unwrap()])
+        .args(["-keyout", key_path.to_str().unwrap()])
+        .output()
+        .context("failed to generate CA certificate")?;
+    if !keygen_output.status.success() {
+        bail!(
+            "openssl failed: '{}'",
+            String::from_utf8_lossy(&keygen_output.stderr)
+        );
+    }
+    Ok(())
+}
+
+fn generate_ssl_cert(
+    cert_path: &Path,
+    key_path: &Path,
+    ca_cert_path: &Path,
+    ca_key_path: &Path,
+) -> anyhow::Result<()> {
+    // Generate Certificate Signing Request (CSR).
+    let mut csr_path = cert_path.to_path_buf();
+    csr_path.set_extension(".csr");
+
+    // openssl req -new -nodes -newkey rsa:2048 -keyout server.key -out server.csr \
+    // -subj "/CN=localhost" -addext "subjectAltName=DNS:localhost,IP:127.0.0.1"
+    let keygen_output = Command::new("openssl")
+        .args(["req", "-new", "-nodes"])
+        .args(["-newkey", "rsa:2048"])
+        .args(["-subj", "/CN=localhost"])
+        .args(["-addext", "subjectAltName=DNS:localhost,IP:127.0.0.1"])
+        .args(["-keyout", key_path.to_str().unwrap()])
+        .args(["-out", csr_path.to_str().unwrap()])
+        .output()
+        .context("failed to generate CSR")?;
+    if !keygen_output.status.success() {
+        bail!(
+            "openssl failed: '{}'",
+            String::from_utf8_lossy(&keygen_output.stderr)
+        );
+    }
+
+    // Sign CSR with CA key.
+    //
+    // openssl x509 -req -in server.csr -CA rootCA.crt -CAkey rootCA.key -CAcreateserial \
+    // -out server.crt -days 36500 -copy_extensions copyall
+    let keygen_output = Command::new("openssl")
+        .args(["x509", "-req"])
+        .args(["-in", csr_path.to_str().unwrap()])
+        .args(["-CA", ca_cert_path.to_str().unwrap()])
+        .args(["-CAkey", ca_key_path.to_str().unwrap()])
+        .arg("-CAcreateserial")
+        .args(["-out", cert_path.to_str().unwrap()])
+        .args(["-days", "36500"])
+        .args(["-copy_extensions", "copyall"])
+        .output()
+        .context("failed to sign CSR")?;
+    if !keygen_output.status.success() {
+        bail!(
+            "openssl failed: '{}'",
+            String::from_utf8_lossy(&keygen_output.stderr)
+        );
+    }
+
+    // Remove CSR file as it's not needed anymore.
+    fs::remove_file(csr_path)?;
+
+    Ok(())
+}

control_plane/src/pageserver.rs

@@ -21,6 +21,7 @@ use pageserver_api::shard::TenantShardId;
 use pageserver_client::mgmt_api;
 use postgres_backend::AuthType;
 use postgres_connection::{PgConnectionConfig, parse_host_port};
+use reqwest::Certificate;
 use utils::auth::{Claims, Scope};
 use utils::id::{NodeId, TenantId, TimelineId};
 use utils::lsn::Lsn;
@@ -49,12 +50,29 @@ impl PageServerNode {
         let (host, port) =
             parse_host_port(&conf.listen_pg_addr).expect("Unable to parse listen_pg_addr");
         let port = port.unwrap_or(5432);
+
+        let ssl_ca_cert = env.ssl_ca_cert_path().map(|ssl_ca_file| {
+            let buf = std::fs::read(ssl_ca_file).expect("SSL root CA file should exist");
+            Certificate::from_pem(&buf).expect("CA certificate should be valid")
+        });
+
+        let endpoint = if env.storage_controller.use_https_pageserver_api {
+            format!(
+                "https://{}",
+                conf.listen_https_addr.as_ref().expect(
+                    "listen https address should be specified if use_https_pageserver_api is on"
+                )
+            )
+        } else {
+            format!("http://{}", conf.listen_http_addr)
+        };
+
         Self {
             pg_connection_config: PgConnectionConfig::new_host_port(host, port),
             conf: conf.clone(),
             env: env.clone(),
             http_client: mgmt_api::Client::new(
-                format!("http://{}", conf.listen_http_addr),
+                endpoint,
                 {
                     match conf.http_auth_type {
                         AuthType::Trust => None,
@@ -65,7 +83,9 @@ impl PageServerNode {
                     }
                 }
                 .as_deref(),
-            ),
+                ssl_ca_cert,
+            )
+            .expect("Client constructs with no errors"),
         }
     }
 
@@ -220,6 +240,13 @@ impl PageServerNode {
             .context("write identity toml")?;
         drop(identity_toml);
 
+        if self.env.generate_local_ssl_certs {
+            self.env.generate_ssl_cert(
+                datadir.join("server.crt").as_path(),
+                datadir.join("server.key").as_path(),
+            )?;
+        }
+
         // TODO: invoke a TBD config-check command to validate that pageserver will start with the written config
 
         // Write metadata file, used by pageserver on startup to register itself with
@@ -230,6 +257,15 @@ impl PageServerNode {
             parse_host_port(&self.conf.listen_http_addr).expect("Unable to parse listen_http_addr");
         let http_port = http_port.unwrap_or(9898);
 
+        let https_port = match self.conf.listen_https_addr.as_ref() {
+            Some(https_addr) => {
+                let (_https_host, https_port) =
+                    parse_host_port(https_addr).expect("Unable to parse listen_https_addr");
+                Some(https_port.unwrap_or(9899))
+            }
+            None => None,
+        };
+
         // Intentionally hand-craft JSON: this acts as an implicit format compat test
         // in case the pageserver-side structure is edited, and reflects the real life
         // situation: the metadata is written by some other script.
@@ -240,6 +276,7 @@ impl PageServerNode {
                 postgres_port: self.pg_connection_config.port(),
                 http_host: "localhost".to_string(),
                 http_port,
+                https_port,
                 other: HashMap::from([(
                     "availability_zone_id".to_string(),
                     serde_json::json!(az_id),

control_plane/src/storage_controller.rs

@@ -12,13 +12,10 @@ use hyper0::Uri;
 use nix::unistd::Pid;
 use pageserver_api::controller_api::{
     NodeConfigureRequest, NodeDescribeResponse, NodeRegisterRequest, TenantCreateRequest,
-    TenantCreateResponse, TenantLocateResponse, TenantShardMigrateRequest,
-    TenantShardMigrateResponse,
+    TenantCreateResponse, TenantLocateResponse,
 };
-use pageserver_api::models::{
-    TenantShardSplitRequest, TenantShardSplitResponse, TimelineCreateRequest, TimelineInfo,
-};
-use pageserver_api::shard::{ShardStripeSize, TenantShardId};
+use pageserver_api::models::{TenantConfigRequest, TimelineCreateRequest, TimelineInfo};
+use pageserver_api::shard::TenantShardId;
 use pageserver_client::mgmt_api::ResponseErrorMessageExt;
 use postgres_backend::AuthType;
 use reqwest::Method;
@@ -537,6 +534,14 @@ impl StorageController {
             args.push("--start-as-candidate".to_string());
         }
 
+        if self.config.use_https_pageserver_api {
+            args.push("--use-https-pageserver-api".to_string());
+        }
+
+        if let Some(ssl_ca_file) = self.env.ssl_ca_cert_path() {
+            args.push(format!("--ssl-ca-file={}", ssl_ca_file.to_str().unwrap()));
+        }
+
         if let Some(private_key) = &self.private_key {
             let claims = Claims::new(None, Scope::PageServerApi);
             let jwt_token =
@@ -579,6 +584,10 @@ impl StorageController {
             self.env.base_data_dir.display()
         ));
 
+        if self.config.timelines_onto_safekeepers {
+            args.push("--timelines-onto-safekeepers".to_string());
+        }
+
         background_process::start_process(
             COMMAND,
             &instance_dir,
@@ -825,41 +834,6 @@ impl StorageController {
         .await
     }
 
-    #[instrument(skip(self))]
-    pub async fn tenant_migrate(
-        &self,
-        tenant_shard_id: TenantShardId,
-        node_id: NodeId,
-    ) -> anyhow::Result<TenantShardMigrateResponse> {
-        self.dispatch(
-            Method::PUT,
-            format!("control/v1/tenant/{tenant_shard_id}/migrate"),
-            Some(TenantShardMigrateRequest {
-                node_id,
-                migration_config: None,
-            }),
-        )
-        .await
-    }
-
-    #[instrument(skip(self), fields(%tenant_id, %new_shard_count))]
-    pub async fn tenant_split(
-        &self,
-        tenant_id: TenantId,
-        new_shard_count: u8,
-        new_stripe_size: Option<ShardStripeSize>,
-    ) -> anyhow::Result<TenantShardSplitResponse> {
-        self.dispatch(
-            Method::PUT,
-            format!("control/v1/tenant/{tenant_id}/shard_split"),
-            Some(TenantShardSplitRequest {
-                new_shard_count,
-                new_stripe_size,
-            }),
-        )
-        .await
-    }
-
     #[instrument(skip_all, fields(node_id=%req.node_id))]
     pub async fn node_register(&self, req: NodeRegisterRequest) -> anyhow::Result<()> {
         self.dispatch::<_, ()>(Method::POST, "control/v1/node".to_string(), Some(req))
@@ -904,4 +878,9 @@ impl StorageController {
         )
         .await
     }
+
+    pub async fn set_tenant_config(&self, req: &TenantConfigRequest) -> anyhow::Result<()> {
+        self.dispatch(Method::PUT, "v1/tenant/config".to_string(), Some(req))
+            .await
+    }
 }

control_plane/storcon_cli/src/main.rs

@@ -1,20 +1,21 @@
 use std::collections::{HashMap, HashSet};
+use std::path::PathBuf;
 use std::str::FromStr;
 use std::time::Duration;
 
 use clap::{Parser, Subcommand};
 use futures::StreamExt;
 use pageserver_api::controller_api::{
-    AvailabilityZone, NodeAvailabilityWrapper, NodeConfigureRequest, NodeDescribeResponse,
-    NodeRegisterRequest, NodeSchedulingPolicy, NodeShardResponse, PlacementPolicy,
-    SafekeeperDescribeResponse, SafekeeperSchedulingPolicyRequest, ShardSchedulingPolicy,
-    ShardsPreferredAzsRequest, ShardsPreferredAzsResponse, SkSchedulingPolicy, TenantCreateRequest,
-    TenantDescribeResponse, TenantPolicyRequest, TenantShardMigrateRequest,
-    TenantShardMigrateResponse,
+    AvailabilityZone, MigrationConfig, NodeAvailabilityWrapper, NodeConfigureRequest,
+    NodeDescribeResponse, NodeRegisterRequest, NodeSchedulingPolicy, NodeShardResponse,
+    PlacementPolicy, SafekeeperDescribeResponse, SafekeeperSchedulingPolicyRequest,
+    ShardSchedulingPolicy, ShardsPreferredAzsRequest, ShardsPreferredAzsResponse,
+    SkSchedulingPolicy, TenantCreateRequest, TenantDescribeResponse, TenantPolicyRequest,
+    TenantShardMigrateRequest, TenantShardMigrateResponse,
 };
 use pageserver_api::models::{
-    EvictionPolicy, EvictionPolicyLayerAccessThreshold, LocationConfigSecondary, ShardParameters,
-    TenantConfig, TenantConfigPatchRequest, TenantConfigRequest, TenantShardSplitRequest,
+    EvictionPolicy, EvictionPolicyLayerAccessThreshold, ShardParameters, TenantConfig,
+    TenantConfigPatchRequest, TenantConfigRequest, TenantShardSplitRequest,
     TenantShardSplitResponse,
 };
 use pageserver_api::shard::{ShardStripeSize, TenantShardId};
@@ -112,6 +113,15 @@ enum Command {
         tenant_shard_id: TenantShardId,
         #[arg(long)]
         node: NodeId,
+        #[arg(long, default_value_t = true, action = clap::ArgAction::Set)]
+        prewarm: bool,
+        #[arg(long, default_value_t = false, action = clap::ArgAction::Set)]
+        override_scheduler: bool,
+    },
+    /// Watch the location of a tenant shard evolve, e.g. while expecting it to migrate
+    TenantShardWatch {
+        #[arg(long)]
+        tenant_shard_id: TenantShardId,
     },
     /// Migrate the secondary location for a tenant shard to a specific pageserver.
     TenantShardMigrateSecondary {
@@ -148,12 +158,6 @@ enum Command {
         #[arg(long)]
         tenant_id: TenantId,
     },
-    /// For a tenant which hasn't been onboarded to the storage controller yet, add it in secondary
-    /// mode so that it can warm up content on a pageserver.
-    TenantWarmup {
-        #[arg(long)]
-        tenant_id: TenantId,
-    },
     TenantSetPreferredAz {
         #[arg(long)]
         tenant_id: TenantId,
@@ -269,6 +273,10 @@ struct Cli {
     /// a token with both scopes to use with this tool.
     jwt: Option<String>,
 
+    #[arg(long)]
+    /// Trusted root CA certificate to use in https APIs.
+    ssl_ca_file: Option<PathBuf>,
+
     #[command(subcommand)]
     command: Command,
 }
@@ -379,9 +387,17 @@ async fn main() -> anyhow::Result<()> {
 
     let storcon_client = Client::new(cli.api.clone(), cli.jwt.clone());
 
+    let ssl_ca_cert = match &cli.ssl_ca_file {
+        Some(ssl_ca_file) => {
+            let buf = tokio::fs::read(ssl_ca_file).await?;
+            Some(reqwest::Certificate::from_pem(&buf)?)
+        }
+        None => None,
+    };
+
     let mut trimmed = cli.api.to_string();
     trimmed.pop();
-    let vps_client = mgmt_api::Client::new(trimmed, cli.jwt.as_deref());
+    let vps_client = mgmt_api::Client::new(trimmed, cli.jwt.as_deref(), ssl_ca_cert)?;
 
     match cli.command {
         Command::NodeRegister {
@@ -619,19 +635,43 @@ async fn main() -> anyhow::Result<()> {
         Command::TenantShardMigrate {
             tenant_shard_id,
             node,
+            prewarm,
+            override_scheduler,
         } => {
-            let req = TenantShardMigrateRequest {
-                node_id: node,
-                migration_config: None,
+            let migration_config = MigrationConfig {
+                prewarm,
+                override_scheduler,
+                ..Default::default()
             };
 
-            storcon_client
+            let req = TenantShardMigrateRequest {
+                node_id: node,
+                origin_node_id: None,
+                migration_config,
+            };
+
+            match storcon_client
                 .dispatch::<TenantShardMigrateRequest, TenantShardMigrateResponse>(
                     Method::PUT,
                     format!("control/v1/tenant/{tenant_shard_id}/migrate"),
                     Some(req),
                 )
-                .await?;
+                .await
+            {
+                Err(mgmt_api::Error::ApiError(StatusCode::PRECONDITION_FAILED, msg)) => {
+                    anyhow::bail!(
+                        "Migration to {node} rejected, may require `--force` ({}) ",
+                        msg
+                    );
+                }
+                Err(e) => return Err(e.into()),
+                Ok(_) => {}
+            }
+
+            watch_tenant_shard(storcon_client, tenant_shard_id, Some(node)).await?;
+        }
+        Command::TenantShardWatch { tenant_shard_id } => {
+            watch_tenant_shard(storcon_client, tenant_shard_id, None).await?;
         }
         Command::TenantShardMigrateSecondary {
             tenant_shard_id,
@@ -639,7 +679,8 @@ async fn main() -> anyhow::Result<()> {
         } => {
             let req = TenantShardMigrateRequest {
                 node_id: node,
-                migration_config: None,
+                origin_node_id: None,
+                migration_config: MigrationConfig::default(),
             };
 
             storcon_client
@@ -824,94 +865,6 @@ async fn main() -> anyhow::Result<()> {
                 )
                 .await?;
         }
-        Command::TenantWarmup { tenant_id } => {
-            let describe_response = storcon_client
-                .dispatch::<(), TenantDescribeResponse>(
-                    Method::GET,
-                    format!("control/v1/tenant/{tenant_id}"),
-                    None,
-                )
-                .await;
-            match describe_response {
-                Ok(describe) => {
-                    if matches!(describe.policy, PlacementPolicy::Secondary) {
-                        // Fine: it's already known to controller in secondary mode: calling
-                        // again to put it into secondary mode won't cause problems.
-                    } else {
-                        anyhow::bail!("Tenant already present with policy {:?}", describe.policy);
-                    }
-                }
-                Err(mgmt_api::Error::ApiError(StatusCode::NOT_FOUND, _)) => {
-                    // Fine: this tenant isn't know to the storage controller yet.
-                }
-                Err(e) => {
-                    // Unexpected API error
-                    return Err(e.into());
-                }
-            }
-
-            vps_client
-                .location_config(
-                    TenantShardId::unsharded(tenant_id),
-                    pageserver_api::models::LocationConfig {
-                        mode: pageserver_api::models::LocationConfigMode::Secondary,
-                        generation: None,
-                        secondary_conf: Some(LocationConfigSecondary { warm: true }),
-                        shard_number: 0,
-                        shard_count: 0,
-                        shard_stripe_size: ShardParameters::DEFAULT_STRIPE_SIZE.0,
-                        tenant_conf: TenantConfig::default(),
-                    },
-                    None,
-                    true,
-                )
-                .await?;
-
-            let describe_response = storcon_client
-                .dispatch::<(), TenantDescribeResponse>(
-                    Method::GET,
-                    format!("control/v1/tenant/{tenant_id}"),
-                    None,
-                )
-                .await?;
-
-            let secondary_ps_id = describe_response
-                .shards
-                .first()
-                .unwrap()
-                .node_secondary
-                .first()
-                .unwrap();
-
-            println!("Tenant {tenant_id} warming up on pageserver {secondary_ps_id}");
-            loop {
-                let (status, progress) = vps_client
-                    .tenant_secondary_download(
-                        TenantShardId::unsharded(tenant_id),
-                        Some(Duration::from_secs(10)),
-                    )
-                    .await?;
-                println!(
-                    "Progress: {}/{} layers, {}/{} bytes",
-                    progress.layers_downloaded,
-                    progress.layers_total,
-                    progress.bytes_downloaded,
-                    progress.bytes_total
-                );
-                match status {
-                    StatusCode::OK => {
-                        println!("Download complete");
-                        break;
-                    }
-                    StatusCode::ACCEPTED => {
-                        // Loop
-                    }
-                    _ => {
-                        anyhow::bail!("Unexpected download status: {status}");
-                    }
-                }
-            }
-        }
         Command::TenantDrop { tenant_id, unclean } => {
             if !unclean {
                 anyhow::bail!(
@@ -1105,7 +1058,8 @@ async fn main() -> anyhow::Result<()> {
                                 format!("control/v1/tenant/{}/migrate", mv.tenant_shard_id),
                                 Some(TenantShardMigrateRequest {
                                     node_id: mv.to,
-                                    migration_config: None,
+                                    origin_node_id: Some(mv.from),
+                                    migration_config: MigrationConfig::default(),
                                 }),
                             )
                             .await
@@ -1284,3 +1238,68 @@ async fn main() -> anyhow::Result<()> {
 
     Ok(())
 }
+
+static WATCH_INTERVAL: Duration = Duration::from_secs(5);
+
+async fn watch_tenant_shard(
+    storcon_client: Client,
+    tenant_shard_id: TenantShardId,
+    until_migrated_to: Option<NodeId>,
+) -> anyhow::Result<()> {
+    if let Some(until_migrated_to) = until_migrated_to {
+        println!(
+            "Waiting for tenant shard {} to be migrated to node {}",
+            tenant_shard_id, until_migrated_to
+        );
+    }
+
+    loop {
+        let desc = storcon_client
+            .dispatch::<(), TenantDescribeResponse>(
+                Method::GET,
+                format!("control/v1/tenant/{}", tenant_shard_id.tenant_id),
+                None,
+            )
+            .await?;
+
+        // Output the current state of the tenant shard
+        let shard = desc
+            .shards
+            .iter()
+            .find(|s| s.tenant_shard_id == tenant_shard_id)
+            .ok_or(anyhow::anyhow!("Tenant shard not found"))?;
+        let summary = format!(
+            "attached: {} secondary: {} {}",
+            shard
+                .node_attached
+                .map(|n| format!("{}", n))
+                .unwrap_or("none".to_string()),
+            shard
+                .node_secondary
+                .iter()
+                .map(|n| n.to_string())
+                .collect::<Vec<_>>()
+                .join(","),
+            if shard.is_reconciling {
+                "(reconciler active)"
+            } else {
+                "(reconciler idle)"
+            }
+        );
+        println!("{}", summary);
+
+        // Maybe drop out if we finished migration
+        if let Some(until_migrated_to) = until_migrated_to {
+            if shard.node_attached == Some(until_migrated_to) && !shard.is_reconciling {
+                println!(
+                    "Tenant shard {} is now on node {}",
+                    tenant_shard_id, until_migrated_to
+                );
+                break;
+            }
+        }
+
+        tokio::time::sleep(WATCH_INTERVAL).await;
+    }
+    Ok(())
+}

deny.toml

@@ -27,6 +27,14 @@ yanked = "warn"
 id = "RUSTSEC-2023-0071"
 reason = "the marvin attack only affects private key decryption, not public key signature verification"
 
+[[advisories.ignore]]
+id = "RUSTSEC-2024-0436"
+reason = "The paste crate is a build-only dependency with no runtime components. It is unlikely to have any security impact."
+
+[[advisories.ignore]]
+id = "RUSTSEC-2025-0014"
+reason = "The humantime is widely used and is not easy to replace right now. It is unmaintained, but it has no known vulnerabilities to care about. #11179"
+
 # This section is considered when running `cargo deny check licenses`
 # More documentation for the licenses section can be found here:
 # https://embarkstudios.github.io/cargo-deny/checks/licenses/cfg.html

docs/storage_controller.md

@@ -101,7 +101,7 @@ changes such as a pageserver node becoming unavailable, or the tenant's shard co
 postgres clients to handle such changes, the storage controller calls an API hook when a tenant's pageserver
 location changes.
 
-The hook is configured using the storage controller's `--compute-hook-url` CLI option. If the hook requires
+The hook is configured using the storage controller's `--control-plane-url` CLI option. If the hook requires
 JWT auth, the token may be provided with `--control-plane-jwt-token`. The hook will be invoked with a `PUT` request.
 
 In the Neon cloud service, this hook is implemented by Neon's internal cloud control plane. In `neon_local` systems

libs/http-utils/Cargo.toml

@@ -8,6 +8,7 @@ license.workspace = true
 anyhow.workspace = true
 bytes.workspace = true
 fail.workspace = true
+futures.workspace = true
 hyper0.workspace = true
 itertools.workspace = true
 jemalloc_pprof.workspace = true
@@ -21,6 +22,7 @@ serde_path_to_error.workspace = true
 thiserror.workspace = true
 tracing.workspace = true
 tokio.workspace = true
+tokio-rustls.workspace = true
 tokio-util.workspace = true
 url.workspace = true
 uuid.workspace = true

libs/http-utils/src/endpoint.rs

@@ -399,12 +399,10 @@ pub async fn profile_cpu_handler(req: Request<Body>) -> Result<Response<Body>, A
     // Return the report in the requested format.
     match format {
         Format::Pprof => {
-            let mut body = Vec::new();
-            report
+            let body = report
                 .pprof()
                 .map_err(|err| ApiError::InternalServerError(err.into()))?
-                .write_to_vec(&mut body)
-                .map_err(|err| ApiError::InternalServerError(err.into()))?;
+                .encode_to_vec();
 
             Response::builder()
                 .status(200)

libs/http-utils/src/lib.rs

@@ -3,9 +3,10 @@ pub mod error;
 pub mod failpoints;
 pub mod json;
 pub mod request;
+pub mod server;
 
 extern crate hyper0 as hyper;
 
 /// Current fast way to apply simple http routing in various Neon binaries.
 /// Re-exported for sake of uniform approach, that could be later replaced with better alternatives, if needed.
-pub use routerify::{RouterBuilder, RouterService, ext::RequestExt};
+pub use routerify::{RequestServiceBuilder, RouterBuilder, RouterService, ext::RequestExt};

libs/http-utils/src/server.rs

@@ -0,0 +1,155 @@
+use std::{error::Error, sync::Arc};
+
+use futures::StreamExt;
+use futures::stream::FuturesUnordered;
+use hyper0::Body;
+use hyper0::server::conn::Http;
+use routerify::{RequestService, RequestServiceBuilder};
+use tokio::io::{AsyncRead, AsyncWrite};
+use tokio_rustls::TlsAcceptor;
+use tokio_util::sync::CancellationToken;
+use tracing::{error, info};
+
+use crate::error::ApiError;
+
+/// A simple HTTP server over hyper library.
+/// You may want to use it instead of [`hyper0::server::Server`] because:
+/// 1. hyper0's Server was removed from hyper v1.
+///    It's recommended to replace hyepr0's Server with a manual loop, which is done here.
+/// 2. hyper0's Server doesn't support TLS out of the box, and there is no way
+///    to support it efficiently with the Accept trait that hyper0's Server uses.
+///    That's one of the reasons why it was removed from v1.
+///    <https://github.com/hyperium/hyper/blob/115339d3df50f20c8717680aa35f48858e9a6205/docs/ROADMAP.md#higher-level-client-and-server-problems>
+pub struct Server {
+    request_service: Arc<RequestServiceBuilder<Body, ApiError>>,
+    listener: tokio::net::TcpListener,
+    tls_acceptor: Option<TlsAcceptor>,
+}
+
+impl Server {
+    pub fn new(
+        request_service: Arc<RequestServiceBuilder<Body, ApiError>>,
+        listener: std::net::TcpListener,
+        tls_acceptor: Option<TlsAcceptor>,
+    ) -> anyhow::Result<Self> {
+        // Note: caller of from_std is responsible for setting nonblocking mode.
+        listener.set_nonblocking(true)?;
+        let listener = tokio::net::TcpListener::from_std(listener)?;
+
+        Ok(Self {
+            request_service,
+            listener,
+            tls_acceptor,
+        })
+    }
+
+    pub async fn serve(self, cancel: CancellationToken) -> anyhow::Result<()> {
+        fn suppress_io_error(err: &std::io::Error) -> bool {
+            use std::io::ErrorKind::*;
+            matches!(err.kind(), ConnectionReset | ConnectionAborted | BrokenPipe)
+        }
+        fn suppress_hyper_error(err: &hyper0::Error) -> bool {
+            if err.is_incomplete_message() || err.is_closed() || err.is_timeout() {
+                return true;
+            }
+            if let Some(inner) = err.source() {
+                if let Some(io) = inner.downcast_ref::<std::io::Error>() {
+                    return suppress_io_error(io);
+                }
+            }
+            false
+        }
+
+        let mut connections = FuturesUnordered::new();
+        loop {
+            tokio::select! {
+                stream = self.listener.accept() => {
+                    let (tcp_stream, remote_addr) = match stream {
+                        Ok(stream) => stream,
+                        Err(err) => {
+                            if !suppress_io_error(&err) {
+                                info!("Failed to accept TCP connection: {err:#}");
+                            }
+                            continue;
+                        }
+                    };
+
+                    let service = self.request_service.build(remote_addr);
+                    let tls_acceptor = self.tls_acceptor.clone();
+                    let cancel = cancel.clone();
+
+                    connections.push(tokio::spawn(
+                        async move {
+                            match tls_acceptor {
+                                Some(tls_acceptor) => {
+                                    // Handle HTTPS connection.
+                                    let tls_stream = tokio::select! {
+                                        tls_stream = tls_acceptor.accept(tcp_stream) => tls_stream,
+                                        _ = cancel.cancelled() => return,
+                                    };
+                                    let tls_stream = match tls_stream {
+                                        Ok(tls_stream) => tls_stream,
+                                        Err(err) => {
+                                            if !suppress_io_error(&err) {
+                                                info!("Failed to accept TLS connection: {err:#}");
+                                            }
+                                            return;
+                                        }
+                                    };
+                                    if let Err(err) = Self::serve_connection(tls_stream, service, cancel).await {
+                                        if !suppress_hyper_error(&err) {
+                                            info!("Failed to serve HTTPS connection: {err:#}");
+                                        }
+                                    }
+                                }
+                                None => {
+                                    // Handle HTTP connection.
+                                    if let Err(err) = Self::serve_connection(tcp_stream, service, cancel).await {
+                                        if !suppress_hyper_error(&err) {
+                                            info!("Failed to serve HTTP connection: {err:#}");
+                                        }
+                                    }
+                                }
+                            };
+                        }));
+                 }
+                Some(conn) = connections.next() => {
+                    if let Err(err) = conn {
+                        error!("Connection panicked: {err:#}");
+                    }
+                }
+                _ = cancel.cancelled() => {
+                    // Wait for graceful shutdown of all connections.
+                    while let Some(conn) = connections.next().await {
+                        if let Err(err) = conn {
+                            error!("Connection panicked: {err:#}");
+                        }
+                    }
+                    break;
+                }
+            }
+        }
+        Ok(())
+    }
+
+    /// Serves HTTP connection with graceful shutdown.
+    async fn serve_connection<I>(
+        io: I,
+        service: RequestService<Body, ApiError>,
+        cancel: CancellationToken,
+    ) -> Result<(), hyper0::Error>
+    where
+        I: AsyncRead + AsyncWrite + Unpin + Send + 'static,
+    {
+        let mut conn = Http::new().serve_connection(io, service).with_upgrades();
+
+        tokio::select! {
+            res = &mut conn => res,
+            _ = cancel.cancelled() => {
+                Pin::new(&mut conn).graceful_shutdown();
+                // Note: connection should still be awaited for graceful shutdown to complete.
+                conn.await
+            }
+        }
+    }
+}

libs/pageserver_api/src/config.rs

@@ -35,6 +35,7 @@ pub struct NodeMetadata {
     pub postgres_port: u16,
     pub http_host: String,
     pub http_port: u16,
+    pub https_port: Option<u16>,
 
     // Deployment tools may write fields to the metadata file beyond what we
     // use in this type: this type intentionally only names fields that require.
@@ -57,6 +58,9 @@ pub struct ConfigToml {
     // types mapped 1:1 into the runtime PageServerConfig type
     pub listen_pg_addr: String,
     pub listen_http_addr: String,
+    pub listen_https_addr: Option<String>,
+    pub ssl_key_file: Utf8PathBuf,
+    pub ssl_cert_file: Utf8PathBuf,
     pub availability_zone: Option<String>,
     #[serde(with = "humantime_serde")]
     pub wait_lsn_timeout: Duration,
@@ -421,6 +425,9 @@ pub mod defaults {
 
     pub const DEFAULT_WAL_RECEIVER_PROTOCOL: utils::postgres_client::PostgresClientProtocol =
         utils::postgres_client::PostgresClientProtocol::Vanilla;
+
+    pub const DEFAULT_SSL_KEY_FILE: &str = "server.key";
+    pub const DEFAULT_SSL_CERT_FILE: &str = "server.crt";
 }
 
 impl Default for ConfigToml {
@@ -430,6 +437,9 @@ impl Default for ConfigToml {
         Self {
             listen_pg_addr: (DEFAULT_PG_LISTEN_ADDR.to_string()),
             listen_http_addr: (DEFAULT_HTTP_LISTEN_ADDR.to_string()),
+            listen_https_addr: (None),
+            ssl_key_file: Utf8PathBuf::from(DEFAULT_SSL_KEY_FILE),
+            ssl_cert_file: Utf8PathBuf::from(DEFAULT_SSL_CERT_FILE),
             availability_zone: (None),
             wait_lsn_timeout: (humantime::parse_duration(DEFAULT_WAIT_LSN_TIMEOUT)
                 .expect("cannot parse default wait lsn timeout")),

libs/pageserver_api/src/config/tests.rs

@@ -16,6 +16,30 @@ fn test_node_metadata_v1_backward_compatibilty() {
             postgres_port: 23,
             http_host: "localhost".to_string(),
             http_port: 42,
+            https_port: None,
+            other: HashMap::new(),
+        }
+    )
+}
+
+#[test]
+fn test_node_metadata_v2_backward_compatibilty() {
+    let v2 = serde_json::to_vec(&serde_json::json!({
+        "host": "localhost",
+        "port": 23,
+        "http_host": "localhost",
+        "http_port": 42,
+        "https_port": 123,
+    }));
+
+    assert_eq!(
+        serde_json::from_slice::<NodeMetadata>(&v2.unwrap()).unwrap(),
+        NodeMetadata {
+            postgres_host: "localhost".to_string(),
+            postgres_port: 23,
+            http_host: "localhost".to_string(),
+            http_port: 42,
+            https_port: Some(123),
             other: HashMap::new(),
         }
     )

libs/pageserver_api/src/controller_api.rs

@@ -182,20 +182,66 @@ pub struct TenantDescribeResponseShard {
 #[derive(Serialize, Deserialize, Debug)]
 pub struct TenantShardMigrateRequest {
     pub node_id: NodeId,
+
+    /// Optionally, callers may specify the node they are migrating _from_, and the server will
+    /// reject the request if the shard is no longer attached there: this enables writing safer
+    /// clients that don't risk fighting with some other movement of the shard.
     #[serde(default)]
-    pub migration_config: Option<MigrationConfig>,
+    pub origin_node_id: Option<NodeId>,
+
+    #[serde(default)]
+    pub migration_config: MigrationConfig,
 }
 
-#[derive(Serialize, Deserialize, Debug)]
+#[derive(Serialize, Deserialize, Debug, PartialEq, Eq)]
 pub struct MigrationConfig {
+    /// If true, the migration will be executed even if it is to a location with a sub-optimal scheduling
+    /// score: this is usually not what you want, and if you use this then you'll also need to set the
+    /// tenant's scheduling policy to Essential or Pause to avoid the optimiser reverting your migration.
+    ///
+    /// Default: false
+    #[serde(default)]
+    pub override_scheduler: bool,
+
+    /// If true, the migration will be done gracefully by creating a secondary location first and
+    /// waiting for it to warm up before cutting over.  If false, if there is no existing secondary
+    /// location at the destination, the tenant will be migrated immediately.  If the tenant's data
+    /// can't be downloaded within [`Self::secondary_warmup_timeout`], then the migration will go
+    /// ahead but run with a cold cache that can severely reduce performance until it warms up.
+    ///
+    /// When doing a graceful migration, the migration API returns as soon as it is started.
+    ///
+    /// Default: true
+    #[serde(default = "default_prewarm")]
+    pub prewarm: bool,
+
+    /// For non-prewarm migrations which will immediately enter a cutover to the new node: how long to wait
+    /// overall for secondary warmup before cutting over
     #[serde(default)]
     #[serde(with = "humantime_serde")]
     pub secondary_warmup_timeout: Option<Duration>,
+    /// For non-prewarm migrations which will immediately enter a cutover to the new node: how long to wait
+    /// within each secondary download poll call to pageserver.
     #[serde(default)]
     #[serde(with = "humantime_serde")]
     pub secondary_download_request_timeout: Option<Duration>,
 }
 
+fn default_prewarm() -> bool {
+    true
+}
+
+impl Default for MigrationConfig {
+    fn default() -> Self {
+        Self {
+            override_scheduler: false,
+            prewarm: default_prewarm(),
+            secondary_warmup_timeout: None,
+            secondary_download_request_timeout: None,
+        }
+    }
+}
+
 #[derive(Serialize, Clone, Debug)]
 #[serde(into = "NodeAvailabilityWrapper")]
 pub enum NodeAvailability {
@@ -443,6 +489,7 @@ pub struct SafekeeperDescribeResponse {
     pub host: String,
     pub port: i32,
     pub http_port: i32,
+    pub https_port: Option<i32>,
     pub availability_zone_id: String,
     pub scheduling_policy: SkSchedulingPolicy,
 }
@@ -487,4 +534,43 @@ mod test {
             err
         );
     }
+
+    /// Check that a minimal migrate request with no config results in the expected default settings
+    #[test]
+    fn test_migrate_request_decode_defaults() {
+        let json = r#"{
+            "node_id": 123
+        }"#;
+
+        let request: TenantShardMigrateRequest = serde_json::from_str(json).unwrap();
+        assert_eq!(request.node_id, NodeId(123));
+        assert_eq!(request.origin_node_id, None);
+        assert!(!request.migration_config.override_scheduler);
+        assert!(request.migration_config.prewarm);
+        assert_eq!(request.migration_config.secondary_warmup_timeout, None);
+        assert_eq!(
+            request.migration_config.secondary_download_request_timeout,
+            None
+        );
+    }
+
+    /// Check that a partially specified migration config results in the expected default settings
+    #[test]
+    fn test_migration_config_decode_defaults() {
+        // Specify just one field of the config
+        let json = r#"{
+        }"#;
+
+        let config: MigrationConfig = serde_json::from_str(json).unwrap();
+
+        // Check each field's expected default value
+        assert!(!config.override_scheduler);
+        assert!(config.prewarm);
+        assert_eq!(config.secondary_warmup_timeout, None);
+        assert_eq!(config.secondary_download_request_timeout, None);
+        assert_eq!(config.secondary_warmup_timeout, None);
+
+        // Consistency check that the Default impl agrees with our serde defaults
+        assert_eq!(MigrationConfig::default(), config);
+    }
 }

libs/pageserver_api/src/models.rs

@@ -274,6 +274,31 @@ pub struct TimelineCreateRequest {
     pub mode: TimelineCreateRequestMode,
 }
 
+/// Storage controller specific extensions to [`TimelineInfo`].
+#[derive(Serialize, Deserialize, Clone)]
+pub struct TimelineCreateResponseStorcon {
+    #[serde(flatten)]
+    pub timeline_info: TimelineInfo,
+
+    pub safekeepers: Option<SafekeepersInfo>,
+}
+
+/// Safekeepers as returned in timeline creation request to storcon or pushed to
+/// cplane in the post migration hook.
+#[derive(Serialize, Deserialize, Clone)]
+pub struct SafekeepersInfo {
+    pub tenant_id: TenantId,
+    pub timeline_id: TimelineId,
+    pub generation: u32,
+    pub safekeepers: Vec<SafekeeperInfo>,
+}
+
+#[derive(Serialize, Deserialize, Clone)]
+pub struct SafekeeperInfo {
+    pub id: NodeId,
+    pub hostname: String,
+}
+
 #[derive(Serialize, Deserialize, Clone)]
 #[serde(untagged)]
 pub enum TimelineCreateRequestMode {
@@ -1146,6 +1171,15 @@ pub struct TimelineArchivalConfigRequest {
     pub state: TimelineArchivalState,
 }
 
+#[derive(Serialize, Deserialize, PartialEq, Eq, Clone)]
+pub struct TimelinePatchIndexPartRequest {
+    pub rel_size_migration: Option<RelSizeMigration>,
+    pub gc_compaction_last_completed_lsn: Option<Lsn>,
+    pub applied_gc_cutoff_lsn: Option<Lsn>,
+    #[serde(default)]
+    pub force_index_update: bool,
+}
+
 #[derive(Debug, Serialize, Deserialize, Clone)]
 pub struct TimelinesInfoAndOffloaded {
     pub timelines: Vec<TimelineInfo>,
@@ -1191,9 +1225,10 @@ pub struct TimelineInfo {
     pub last_record_lsn: Lsn,
     pub prev_record_lsn: Option<Lsn>,
 
-    /// Legacy field for compat with control plane.  Synonym of `min_readable_lsn`.
-    /// TODO: remove once control plane no longer reads it.
-    pub latest_gc_cutoff_lsn: Lsn,
+    /// Legacy field, retained for one version to enable old storage controller to
+    /// decode (it was a mandatory field).
+    #[serde(default, rename = "latest_gc_cutoff_lsn")]
+    pub _unused: Lsn,
 
     /// The LSN up to which GC has advanced: older data may still exist but it is not available for clients.
     /// This LSN is not suitable for deciding where to create branches etc: use [`TimelineInfo::min_readable_lsn`] instead,
@@ -1442,8 +1477,14 @@ pub struct TenantScanRemoteStorageResponse {
 #[derive(Serialize, Deserialize, Debug, Clone)]
 #[serde(rename_all = "snake_case")]
 pub enum TenantSorting {
+    /// Total size of layers on local disk for all timelines in a shard.
     ResidentSize,
+    /// The logical size of the largest timeline within a _tenant_ (not shard). Only tracked on
+    /// shard 0, contains the sum across all shards.
     MaxLogicalSize,
+    /// The logical size of the largest timeline within a _tenant_ (not shard), divided by number of
+    /// shards. Only tracked on shard 0, and estimates the per-shard logical size.
+    MaxLogicalSizePerShard,
 }
 
 impl Default for TenantSorting {
@@ -1473,14 +1514,20 @@ pub struct TopTenantShardsRequest {
 pub struct TopTenantShardItem {
     pub id: TenantShardId,
 
-    /// Total size of layers on local disk for all timelines in this tenant
+    /// Total size of layers on local disk for all timelines in this shard.
     pub resident_size: u64,
 
-    /// Total size of layers in remote storage for all timelines in this tenant
+    /// Total size of layers in remote storage for all timelines in this shard.
     pub physical_size: u64,
 
-    /// The largest logical size of a timeline within this tenant
+    /// The largest logical size of a timeline within this _tenant_ (not shard). This is only
+    /// tracked on shard 0, and contains the sum of the logical size across all shards.
     pub max_logical_size: u64,
+
+    /// The largest logical size of a timeline within this _tenant_ (not shard) divided by number of
+    /// shards. This is only tracked on shard 0, and is only an estimate as we divide it evenly by
+    /// shard count, rounded up.
+    pub max_logical_size_per_shard: u64,
 }
 
 #[derive(Serialize, Deserialize, Debug, Default)]

libs/pageserver_api/src/shard.rs

@@ -112,6 +112,16 @@ impl ShardIdentity {
         }
     }
 
+    /// An unsharded identity with the given stripe size (if non-zero). This is typically used to
+    /// carry over a stripe size for an unsharded tenant from persistent storage.
+    pub fn unsharded_with_stripe_size(stripe_size: ShardStripeSize) -> Self {
+        let mut shard_identity = Self::unsharded();
+        if stripe_size.0 > 0 {
+            shard_identity.stripe_size = stripe_size;
+        }
+        shard_identity
+    }
+
     /// A broken instance of this type is only used for `TenantState::Broken` tenants,
     /// which are constructed in code paths that don't have access to proper configuration.
     ///

libs/postgres_ffi/src/lib.rs

@@ -396,6 +396,14 @@ pub mod waldecoder {
             self.lsn + self.inputbuf.remaining() as u64
         }
 
+        /// Returns the LSN up to which the WAL decoder has processed.
+        ///
+        /// If [`Self::poll_decode`] returned a record, then this will return
+        /// the end LSN of said record.
+        pub fn lsn(&self) -> Lsn {
+            self.lsn
+        }
+
         pub fn feed_bytes(&mut self, buf: &[u8]) {
             self.inputbuf.extend_from_slice(buf);
         }

libs/proxy/postgres-types2/src/lib.rs

@@ -135,8 +135,8 @@ impl Type {
 pub enum Kind {
     /// A simple type like `VARCHAR` or `INTEGER`.
     Simple,
-    /// An enumerated type along with its variants.
-    Enum(Vec<String>),
+    /// An enumerated type.
+    Enum,
     /// A pseudo-type.
     Pseudo,
     /// An array type along with the type of its elements.
@@ -146,9 +146,9 @@ pub enum Kind {
     /// A multirange type along with the type of its elements.
     Multirange(Type),
     /// A domain type along with its underlying type.
-    Domain(Type),
-    /// A composite type along with information about its fields.
-    Composite(Vec<Field>),
+    Domain(Oid),
+    /// A composite type.
+    Composite(Oid),
 }
 
 /// Information about a field of a composite type.

libs/proxy/tokio-postgres2/src/client.rs

@@ -19,10 +19,10 @@ use crate::config::{Host, SslMode};
 use crate::connection::{Request, RequestMessages};
 use crate::query::RowStream;
 use crate::simple_query::SimpleQueryStream;
-use crate::types::{Oid, ToSql, Type};
+use crate::types::{Oid, Type};
 use crate::{
-    CancelToken, Error, ReadyForQueryStatus, Row, SimpleQueryMessage, Statement, Transaction,
-    TransactionBuilder, query, simple_query, slice_iter,
+    CancelToken, Error, ReadyForQueryStatus, SimpleQueryMessage, Statement, Transaction,
+    TransactionBuilder, query, simple_query,
 };
 
 pub struct Responses {
@@ -54,26 +54,18 @@ impl Responses {
 /// A cache of type info and prepared statements for fetching type info
 /// (corresponding to the queries in the [crate::prepare] module).
 #[derive(Default)]
-struct CachedTypeInfo {
+pub(crate) struct CachedTypeInfo {
     /// A statement for basic information for a type from its
     /// OID. Corresponds to [TYPEINFO_QUERY](crate::prepare::TYPEINFO_QUERY) (or its
     /// fallback).
-    typeinfo: Option<Statement>,
-    /// A statement for getting information for a composite type from its OID.
-    /// Corresponds to [TYPEINFO_QUERY](crate::prepare::TYPEINFO_COMPOSITE_QUERY).
-    typeinfo_composite: Option<Statement>,
-    /// A statement for getting information for a composite type from its OID.
-    /// Corresponds to [TYPEINFO_QUERY](crate::prepare::TYPEINFO_COMPOSITE_QUERY) (or
-    /// its fallback).
-    typeinfo_enum: Option<Statement>,
+    pub(crate) typeinfo: Option<Statement>,
 
     /// Cache of types already looked up.
-    types: HashMap<Oid, Type>,
+    pub(crate) types: HashMap<Oid, Type>,
 }
 
 pub struct InnerClient {
     sender: mpsc::UnboundedSender<Request>,
-    cached_typeinfo: Mutex<CachedTypeInfo>,
 
     /// A buffer to use when writing out postgres commands.
     buffer: Mutex<BytesMut>,
@@ -91,38 +83,6 @@ impl InnerClient {
         })
     }
 
-    pub fn typeinfo(&self) -> Option<Statement> {
-        self.cached_typeinfo.lock().typeinfo.clone()
-    }
-
-    pub fn set_typeinfo(&self, statement: &Statement) {
-        self.cached_typeinfo.lock().typeinfo = Some(statement.clone());
-    }
-
-    pub fn typeinfo_composite(&self) -> Option<Statement> {
-        self.cached_typeinfo.lock().typeinfo_composite.clone()
-    }
-
-    pub fn set_typeinfo_composite(&self, statement: &Statement) {
-        self.cached_typeinfo.lock().typeinfo_composite = Some(statement.clone());
-    }
-
-    pub fn typeinfo_enum(&self) -> Option<Statement> {
-        self.cached_typeinfo.lock().typeinfo_enum.clone()
-    }
-
-    pub fn set_typeinfo_enum(&self, statement: &Statement) {
-        self.cached_typeinfo.lock().typeinfo_enum = Some(statement.clone());
-    }
-
-    pub fn type_(&self, oid: Oid) -> Option<Type> {
-        self.cached_typeinfo.lock().types.get(&oid).cloned()
-    }
-
-    pub fn set_type(&self, oid: Oid, type_: &Type) {
-        self.cached_typeinfo.lock().types.insert(oid, type_.clone());
-    }
-
     /// Call the given function with a buffer to be used when writing out
     /// postgres commands.
     pub fn with_buf<F, R>(&self, f: F) -> R
@@ -142,7 +102,6 @@ pub struct SocketConfig {
     pub host: Host,
     pub port: u16,
     pub connect_timeout: Option<Duration>,
-    // pub keepalive: Option<KeepaliveConfig>,
 }
 
 /// An asynchronous PostgreSQL client.
@@ -151,6 +110,7 @@ pub struct SocketConfig {
 /// through this client object.
 pub struct Client {
     inner: Arc<InnerClient>,
+    cached_typeinfo: CachedTypeInfo,
 
     socket_config: SocketConfig,
     ssl_mode: SslMode,
@@ -169,9 +129,9 @@ impl Client {
         Client {
             inner: Arc::new(InnerClient {
                 sender,
-                cached_typeinfo: Default::default(),
                 buffer: Default::default(),
             }),
+            cached_typeinfo: Default::default(),
 
             socket_config,
             ssl_mode,
@@ -189,55 +149,6 @@ impl Client {
         &self.inner
     }
 
-    /// Executes a statement, returning a vector of the resulting rows.
-    ///
-    /// A statement may contain parameters, specified by `$n`, where `n` is the index of the parameter of the list
-    /// provided, 1-indexed.
-    ///
-    /// The `statement` argument can either be a `Statement`, or a raw query string. If the same statement will be
-    /// repeatedly executed (perhaps with different query parameters), consider preparing the statement up front
-    /// with the `prepare` method.
-    ///
-    /// # Panics
-    ///
-    /// Panics if the number of parameters provided does not match the number expected.
-    pub async fn query(
-        &self,
-        statement: Statement,
-        params: &[&(dyn ToSql + Sync)],
-    ) -> Result<Vec<Row>, Error> {
-        self.query_raw(statement, slice_iter(params))
-            .await?
-            .try_collect()
-            .await
-    }
-
-    /// The maximally flexible version of [`query`].
-    ///
-    /// A statement may contain parameters, specified by `$n`, where `n` is the index of the parameter of the list
-    /// provided, 1-indexed.
-    ///
-    /// The `statement` argument can either be a `Statement`, or a raw query string. If the same statement will be
-    /// repeatedly executed (perhaps with different query parameters), consider preparing the statement up front
-    /// with the `prepare` method.
-    ///
-    /// # Panics
-    ///
-    /// Panics if the number of parameters provided does not match the number expected.
-    ///
-    /// [`query`]: #method.query
-    pub async fn query_raw<'a, I>(
-        &self,
-        statement: Statement,
-        params: I,
-    ) -> Result<RowStream, Error>
-    where
-        I: IntoIterator<Item = &'a (dyn ToSql + Sync)>,
-        I::IntoIter: ExactSizeIterator,
-    {
-        query::query(&self.inner, statement, params).await
-    }
-
     /// Pass text directly to the Postgres backend to allow it to sort out typing itself and
     /// to save a roundtrip
     pub async fn query_raw_txt<S, I>(&self, statement: &str, params: I) -> Result<RowStream, Error>
@@ -284,6 +195,14 @@ impl Client {
         simple_query::batch_execute(self.inner(), query).await
     }
 
+    pub async fn discard_all(&mut self) -> Result<ReadyForQueryStatus, Error> {
+        // clear the prepared statements that are about to be nuked from the postgres session
+
+        self.cached_typeinfo.typeinfo = None;
+
+        self.batch_execute("discard all").await
+    }
+
     /// Begins a new database transaction.
     ///
     /// The transaction will roll back by default - use the `commit` method to commit it.
@@ -347,8 +266,8 @@ impl Client {
     }
 
     /// Query for type information
-    pub async fn get_type(&self, oid: Oid) -> Result<Type, Error> {
-        crate::prepare::get_type(&self.inner, oid).await
+    pub(crate) async fn get_type_inner(&mut self, oid: Oid) -> Result<Type, Error> {
+        crate::prepare::get_type(&self.inner, &mut self.cached_typeinfo, oid).await
     }
 
     /// Determines if the connection to the server has already closed.

libs/proxy/tokio-postgres2/src/generic_client.rs

@@ -22,7 +22,7 @@ pub trait GenericClient: private::Sealed {
         I::IntoIter: ExactSizeIterator + Sync + Send;
 
     /// Query for type information
-    async fn get_type(&self, oid: Oid) -> Result<Type, Error>;
+    async fn get_type(&mut self, oid: Oid) -> Result<Type, Error>;
 }
 
 impl private::Sealed for Client {}
@@ -38,8 +38,8 @@ impl GenericClient for Client {
     }
 
     /// Query for type information
-    async fn get_type(&self, oid: Oid) -> Result<Type, Error> {
-        crate::prepare::get_type(self.inner(), oid).await
+    async fn get_type(&mut self, oid: Oid) -> Result<Type, Error> {
+        self.get_type_inner(oid).await
     }
 }
 
@@ -56,7 +56,7 @@ impl GenericClient for Transaction<'_> {
     }
 
     /// Query for type information
-    async fn get_type(&self, oid: Oid) -> Result<Type, Error> {
-        self.client().get_type(oid).await
+    async fn get_type(&mut self, oid: Oid) -> Result<Type, Error> {
+        self.client_mut().get_type(oid).await
     }
 }

libs/proxy/tokio-postgres2/src/prepare.rs

@@ -9,10 +9,10 @@ use log::debug;
 use postgres_protocol2::message::backend::Message;
 use postgres_protocol2::message::frontend;
 
-use crate::client::InnerClient;
+use crate::client::{CachedTypeInfo, InnerClient};
 use crate::codec::FrontendMessage;
 use crate::connection::RequestMessages;
-use crate::types::{Field, Kind, Oid, Type};
+use crate::types::{Kind, Oid, Type};
 use crate::{Column, Error, Statement, query, slice_iter};
 
 pub(crate) const TYPEINFO_QUERY: &str = "\
@@ -23,23 +23,7 @@ INNER JOIN pg_catalog.pg_namespace n ON t.typnamespace = n.oid
 WHERE t.oid = $1
 ";
 
-const TYPEINFO_ENUM_QUERY: &str = "\
-SELECT enumlabel
-FROM pg_catalog.pg_enum
-WHERE enumtypid = $1
-ORDER BY enumsortorder
-";
-
-pub(crate) const TYPEINFO_COMPOSITE_QUERY: &str = "\
-SELECT attname, atttypid
-FROM pg_catalog.pg_attribute
-WHERE attrelid = $1
-AND NOT attisdropped
-AND attnum > 0
-ORDER BY attnum
-";
-
-pub async fn prepare(
+async fn prepare_typecheck(
     client: &Arc<InnerClient>,
     name: &'static str,
     query: &str,
@@ -67,7 +51,7 @@ pub async fn prepare(
     let mut parameters = vec![];
     let mut it = parameter_description.parameters();
     while let Some(oid) = it.next().map_err(Error::parse)? {
-        let type_ = get_type(client, oid).await?;
+        let type_ = Type::from_oid(oid).ok_or_else(Error::unexpected_message)?;
         parameters.push(type_);
     }
 
@@ -75,7 +59,7 @@ pub async fn prepare(
     if let Some(row_description) = row_description {
         let mut it = row_description.fields();
         while let Some(field) = it.next().map_err(Error::parse)? {
-            let type_ = get_type(client, field.type_oid()).await?;
+            let type_ = Type::from_oid(field.type_oid()).ok_or_else(Error::unexpected_message)?;
             let column = Column::new(field.name().to_string(), type_, field);
             columns.push(column);
         }
@@ -84,15 +68,6 @@ pub async fn prepare(
     Ok(Statement::new(client, name, parameters, columns))
 }
 
-fn prepare_rec<'a>(
-    client: &'a Arc<InnerClient>,
-    name: &'static str,
-    query: &'a str,
-    types: &'a [Type],
-) -> Pin<Box<dyn Future<Output = Result<Statement, Error>> + 'a + Send>> {
-    Box::pin(prepare(client, name, query, types))
-}
-
 fn encode(client: &InnerClient, name: &str, query: &str, types: &[Type]) -> Result<Bytes, Error> {
     if types.is_empty() {
         debug!("preparing query {}: {}", name, query);
@@ -108,16 +83,20 @@ fn encode(client: &InnerClient, name: &str, query: &str, types: &[Type]) -> Resu
     })
 }
 
-pub async fn get_type(client: &Arc<InnerClient>, oid: Oid) -> Result<Type, Error> {
+pub async fn get_type(
+    client: &Arc<InnerClient>,
+    typecache: &mut CachedTypeInfo,
+    oid: Oid,
+) -> Result<Type, Error> {
     if let Some(type_) = Type::from_oid(oid) {
         return Ok(type_);
     }
 
-    if let Some(type_) = client.type_(oid) {
-        return Ok(type_);
-    }
+    if let Some(type_) = typecache.types.get(&oid) {
+        return Ok(type_.clone());
+    };
 
-    let stmt = typeinfo_statement(client).await?;
+    let stmt = typeinfo_statement(client, typecache).await?;
 
     let rows = query::query(client, stmt, slice_iter(&[&oid])).await?;
     pin_mut!(rows);
@@ -136,100 +115,48 @@ pub async fn get_type(client: &Arc<InnerClient>, oid: Oid) -> Result<Type, Error
     let relid: Oid = row.try_get(6)?;
 
     let kind = if type_ == b'e' as i8 {
-        let variants = get_enum_variants(client, oid).await?;
-        Kind::Enum(variants)
+        Kind::Enum
     } else if type_ == b'p' as i8 {
         Kind::Pseudo
     } else if basetype != 0 {
-        let type_ = get_type_rec(client, basetype).await?;
-        Kind::Domain(type_)
+        Kind::Domain(basetype)
     } else if elem_oid != 0 {
-        let type_ = get_type_rec(client, elem_oid).await?;
+        let type_ = get_type_rec(client, typecache, elem_oid).await?;
         Kind::Array(type_)
     } else if relid != 0 {
-        let fields = get_composite_fields(client, relid).await?;
-        Kind::Composite(fields)
+        Kind::Composite(relid)
     } else if let Some(rngsubtype) = rngsubtype {
-        let type_ = get_type_rec(client, rngsubtype).await?;
+        let type_ = get_type_rec(client, typecache, rngsubtype).await?;
         Kind::Range(type_)
     } else {
         Kind::Simple
     };
 
     let type_ = Type::new(name, oid, kind, schema);
-    client.set_type(oid, &type_);
+    typecache.types.insert(oid, type_.clone());
 
     Ok(type_)
 }
 
 fn get_type_rec<'a>(
     client: &'a Arc<InnerClient>,
+    typecache: &'a mut CachedTypeInfo,
     oid: Oid,
 ) -> Pin<Box<dyn Future<Output = Result<Type, Error>> + Send + 'a>> {
-    Box::pin(get_type(client, oid))
+    Box::pin(get_type(client, typecache, oid))
 }
 
-async fn typeinfo_statement(client: &Arc<InnerClient>) -> Result<Statement, Error> {
-    if let Some(stmt) = client.typeinfo() {
-        return Ok(stmt);
+async fn typeinfo_statement(
+    client: &Arc<InnerClient>,
+    typecache: &mut CachedTypeInfo,
+) -> Result<Statement, Error> {
+    if let Some(stmt) = &typecache.typeinfo {
+        return Ok(stmt.clone());
     }
 
     let typeinfo = "neon_proxy_typeinfo";
-    let stmt = prepare_rec(client, typeinfo, TYPEINFO_QUERY, &[]).await?;
+    let stmt = prepare_typecheck(client, typeinfo, TYPEINFO_QUERY, &[]).await?;
 
-    client.set_typeinfo(&stmt);
-    Ok(stmt)
-}
-
-async fn get_enum_variants(client: &Arc<InnerClient>, oid: Oid) -> Result<Vec<String>, Error> {
-    let stmt = typeinfo_enum_statement(client).await?;
-
-    query::query(client, stmt, slice_iter(&[&oid]))
-        .await?
-        .and_then(|row| async move { row.try_get(0) })
-        .try_collect()
-        .await
-}
-
-async fn typeinfo_enum_statement(client: &Arc<InnerClient>) -> Result<Statement, Error> {
-    if let Some(stmt) = client.typeinfo_enum() {
-        return Ok(stmt);
-    }
-
-    let typeinfo = "neon_proxy_typeinfo_enum";
-    let stmt = prepare_rec(client, typeinfo, TYPEINFO_ENUM_QUERY, &[]).await?;
-
-    client.set_typeinfo_enum(&stmt);
-    Ok(stmt)
-}
-
-async fn get_composite_fields(client: &Arc<InnerClient>, oid: Oid) -> Result<Vec<Field>, Error> {
-    let stmt = typeinfo_composite_statement(client).await?;
-
-    let rows = query::query(client, stmt, slice_iter(&[&oid]))
-        .await?
-        .try_collect::<Vec<_>>()
-        .await?;
-
-    let mut fields = vec![];
-    for row in rows {
-        let name = row.try_get(0)?;
-        let oid = row.try_get(1)?;
-        let type_ = get_type_rec(client, oid).await?;
-        fields.push(Field::new(name, type_));
-    }
-
-    Ok(fields)
-}
-
-async fn typeinfo_composite_statement(client: &Arc<InnerClient>) -> Result<Statement, Error> {
-    if let Some(stmt) = client.typeinfo_composite() {
-        return Ok(stmt);
-    }
-
-    let typeinfo = "neon_proxy_typeinfo_composite";
-    let stmt = prepare_rec(client, typeinfo, TYPEINFO_COMPOSITE_QUERY, &[]).await?;
-
-    client.set_typeinfo_composite(&stmt);
+    typecache.typeinfo = Some(stmt.clone());
     Ok(stmt)
 }

libs/proxy/tokio-postgres2/src/transaction.rs

@@ -72,4 +72,9 @@ impl<'a> Transaction<'a> {
     pub fn client(&self) -> &Client {
         self.client
     }
+
+    /// Returns a reference to the underlying `Client`.
+    pub fn client_mut(&mut self) -> &mut Client {
+        self.client
+    }
 }

libs/safekeeper_api/src/membership.rs

@@ -131,6 +131,14 @@ impl Configuration {
         }
     }
 
+    pub fn new(members: MemberSet) -> Self {
+        Configuration {
+            generation: INITIAL_GENERATION,
+            members,
+            new_members: None,
+        }
+    }
+
     /// Is `sk_id` member of the configuration?
     pub fn contains(&self, sk_id: NodeId) -> bool {
         self.members.contains(sk_id) || self.new_members.as_ref().is_some_and(|m| m.contains(sk_id))

libs/safekeeper_api/src/models.rs

@@ -18,7 +18,7 @@ pub struct SafekeeperStatus {
     pub id: NodeId,
 }
 
-#[derive(Serialize, Deserialize)]
+#[derive(Serialize, Deserialize, Clone)]
 pub struct TimelineCreateRequest {
     pub tenant_id: TenantId,
     pub timeline_id: TimelineId,
@@ -283,7 +283,7 @@ pub struct SafekeeperUtilization {
 }
 
 /// pull_timeline request body.
-#[derive(Debug, Deserialize, Serialize)]
+#[derive(Debug, Clone, Deserialize, Serialize)]
 pub struct PullTimelineRequest {
     pub tenant_id: TenantId,
     pub timeline_id: TimelineId,

libs/tracing-utils/src/lib.rs

@@ -21,7 +21,7 @@
 //!         .with_writer(std::io::stderr);
 //!
 //!     // Initialize OpenTelemetry. Exports tracing spans as OpenTelemetry traces
-//!     let otlp_layer = tracing_utils::init_tracing("my_application").await;
+//!     let otlp_layer = tracing_utils::init_tracing("my_application", tracing_utils::ExportConfig::default()).await;
 //!
 //!     // Put it all together
 //!     tracing_subscriber::registry()
@@ -38,8 +38,12 @@ pub mod http;
 
 use opentelemetry::KeyValue;
 use opentelemetry::trace::TracerProvider;
-use tracing::Subscriber;
+use opentelemetry_otlp::WithExportConfig;
+pub use opentelemetry_otlp::{ExportConfig, Protocol};
+use tracing::level_filters::LevelFilter;
+use tracing::{Dispatch, Subscriber};
 use tracing_subscriber::Layer;
+use tracing_subscriber::layer::SubscriberExt;
 use tracing_subscriber::registry::LookupSpan;
 
 /// Set up OpenTelemetry exporter, using configuration from environment variables.
@@ -69,19 +73,28 @@ use tracing_subscriber::registry::LookupSpan;
 ///
 /// This doesn't block, but is marked as 'async' to hint that this must be called in
 /// asynchronous execution context.
-pub async fn init_tracing<S>(service_name: &str) -> Option<impl Layer<S>>
+pub async fn init_tracing<S>(
+    service_name: &str,
+    export_config: ExportConfig,
+) -> Option<impl Layer<S>>
 where
     S: Subscriber + for<'span> LookupSpan<'span>,
 {
     if std::env::var("OTEL_SDK_DISABLED") == Ok("true".to_string()) {
         return None;
     };
-    Some(init_tracing_internal(service_name.to_string()))
+    Some(init_tracing_internal(
+        service_name.to_string(),
+        export_config,
+    ))
 }
 
 /// Like `init_tracing`, but creates a separate tokio Runtime for the tracing
 /// tasks.
-pub fn init_tracing_without_runtime<S>(service_name: &str) -> Option<impl Layer<S>>
+pub fn init_tracing_without_runtime<S>(
+    service_name: &str,
+    export_config: ExportConfig,
+) -> Option<impl Layer<S>>
 where
     S: Subscriber + for<'span> LookupSpan<'span>,
 {
@@ -112,16 +125,22 @@ where
     ));
     let _guard = runtime.enter();
 
-    Some(init_tracing_internal(service_name.to_string()))
+    Some(init_tracing_internal(
+        service_name.to_string(),
+        export_config,
+    ))
 }
 
-fn init_tracing_internal<S>(service_name: String) -> impl Layer<S>
+fn init_tracing_internal<S>(service_name: String, export_config: ExportConfig) -> impl Layer<S>
 where
     S: Subscriber + for<'span> LookupSpan<'span>,
 {
-    // Sets up exporter from the OTEL_EXPORTER_* environment variables.
+    // Sets up exporter from the provided [`ExportConfig`] parameter.
+    // If the endpoint is not specified, it is loaded from the
+    // OTEL_EXPORTER_OTLP_ENDPOINT environment variable.
     let exporter = opentelemetry_otlp::SpanExporter::builder()
         .with_http()
+        .with_export_config(export_config)
         .build()
         .expect("could not initialize opentelemetry exporter");
 
@@ -151,3 +170,51 @@ where
 pub fn shutdown_tracing() {
     opentelemetry::global::shutdown_tracer_provider();
 }
+
+pub enum OtelEnablement {
+    Disabled,
+    Enabled {
+        service_name: String,
+        export_config: ExportConfig,
+        runtime: &'static tokio::runtime::Runtime,
+    },
+}
+
+pub struct OtelGuard {
+    pub dispatch: Dispatch,
+}
+
+impl Drop for OtelGuard {
+    fn drop(&mut self) {
+        shutdown_tracing();
+    }
+}
+
+/// Initializes OTEL infrastructure for performance tracing according to the provided configuration
+///
+/// Performance tracing is handled by a different [`tracing::Subscriber`]. This functions returns
+/// an [`OtelGuard`] containing a [`tracing::Dispatch`] associated with a newly created subscriber.
+/// Applications should use this dispatch for their performance traces.
+///
+/// The lifetime of the guard should match taht of the application. On drop, it tears down the
+/// OTEL infra.
+pub fn init_performance_tracing(otel_enablement: OtelEnablement) -> Option<OtelGuard> {
+    let otel_subscriber = match otel_enablement {
+        OtelEnablement::Disabled => None,
+        OtelEnablement::Enabled {
+            service_name,
+            export_config,
+            runtime,
+        } => {
+            let otel_layer = runtime
+                .block_on(init_tracing(&service_name, export_config))
+                .with_filter(LevelFilter::INFO);
+            let otel_subscriber = tracing_subscriber::registry().with(otel_layer);
+            let otel_dispatch = Dispatch::new(otel_subscriber);
+
+            Some(otel_dispatch)
+        }
+    };
+
+    otel_subscriber.map(|dispatch| OtelGuard { dispatch })
+}

libs/utils/Cargo.toml

@@ -42,6 +42,7 @@ toml_edit = { workspace = true, features = ["serde"] }
 tracing.workspace = true
 tracing-error.workspace = true
 tracing-subscriber = { workspace = true, features = ["json", "registry"] }
+tracing-utils.workspace = true
 rand.workspace = true
 scopeguard.workspace = true
 strum.workspace = true

libs/utils/src/logging.rs

@@ -165,6 +165,7 @@ pub fn init(
         };
         log_layer.with_filter(rust_log_env_filter())
     });
+
     let r = r.with(
         TracingEventCountLayer(&TRACING_EVENT_COUNT_METRIC).with_filter(rust_log_env_filter()),
     );

pageserver/Cargo.toml

@@ -48,6 +48,9 @@ pprof.workspace = true
 rand.workspace = true
 range-set-blaze = { version = "0.1.16", features = ["alloc"] }
 regex.workspace = true
+rustls-pemfile.workspace = true
+rustls-pki-types.workspace = true
+rustls.workspace = true
 scopeguard.workspace = true
 send-future.workspace = true
 serde.workspace = true
@@ -62,10 +65,12 @@ tokio = { workspace = true, features = ["process", "sync", "fs", "rt", "io-util"
 tokio-epoll-uring.workspace = true
 tokio-io-timeout.workspace = true
 tokio-postgres.workspace = true
+tokio-rustls.workspace = true
 tokio-stream.workspace = true
 tokio-util.workspace = true
 toml_edit = { workspace = true, features = [ "serde" ] }
 tracing.workspace = true
+tracing-utils.workspace = true
 url.workspace = true
 walkdir.workspace = true
 metrics.workspace = true
@@ -116,6 +121,10 @@ harness = false
 name = "upload_queue"
 harness = false
 
+[[bench]]
+name = "bench_metrics"
+harness = false
+
 [[bin]]
 name = "test_helper_slow_client_reads"
 required-features = [ "testing" ]

pageserver/benches/bench_metrics.rs

@@ -0,0 +1,366 @@
+use criterion::{BenchmarkId, Criterion, criterion_group, criterion_main};
+use utils::id::{TenantId, TimelineId};
+
+//
+// Demonstrates that repeat label values lookup is a multicore scalability bottleneck
+// that is worth avoiding.
+//
+criterion_group!(
+    label_values,
+    label_values::bench_naive_usage,
+    label_values::bench_cache_label_values_lookup
+);
+mod label_values {
+    use super::*;
+
+    pub fn bench_naive_usage(c: &mut Criterion) {
+        let mut g = c.benchmark_group("label_values__naive_usage");
+
+        for ntimelines in [1, 4, 8] {
+            g.bench_with_input(
+                BenchmarkId::new("ntimelines", ntimelines),
+                &ntimelines,
+                |b, ntimelines| {
+                    b.iter_custom(|iters| {
+                        let barrier = std::sync::Barrier::new(*ntimelines + 1);
+
+                        let timelines = (0..*ntimelines)
+                            .map(|_| {
+                                (
+                                    TenantId::generate().to_string(),
+                                    "0000".to_string(),
+                                    TimelineId::generate().to_string(),
+                                )
+                            })
+                            .collect::<Vec<_>>();
+
+                        let metric_vec = metrics::UIntGaugeVec::new(
+                            metrics::opts!("testmetric", "testhelp"),
+                            &["tenant_id", "shard_id", "timeline_id"],
+                        )
+                        .unwrap();
+
+                        std::thread::scope(|s| {
+                            for (tenant_id, shard_id, timeline_id) in &timelines {
+                                s.spawn(|| {
+                                    barrier.wait();
+                                    for _ in 0..iters {
+                                        metric_vec
+                                            .with_label_values(&[tenant_id, shard_id, timeline_id])
+                                            .inc();
+                                    }
+                                    barrier.wait();
+                                });
+                            }
+                            barrier.wait();
+                            let start = std::time::Instant::now();
+                            barrier.wait();
+                            start.elapsed()
+                        })
+                    })
+                },
+            );
+        }
+        g.finish();
+    }
+
+    pub fn bench_cache_label_values_lookup(c: &mut Criterion) {
+        let mut g = c.benchmark_group("label_values__cache_label_values_lookup");
+
+        for ntimelines in [1, 4, 8] {
+            g.bench_with_input(
+                BenchmarkId::new("ntimelines", ntimelines),
+                &ntimelines,
+                |b, ntimelines| {
+                    b.iter_custom(|iters| {
+                        let barrier = std::sync::Barrier::new(*ntimelines + 1);
+
+                        let timelines = (0..*ntimelines)
+                            .map(|_| {
+                                (
+                                    TenantId::generate().to_string(),
+                                    "0000".to_string(),
+                                    TimelineId::generate().to_string(),
+                                )
+                            })
+                            .collect::<Vec<_>>();
+
+                        let metric_vec = metrics::UIntGaugeVec::new(
+                            metrics::opts!("testmetric", "testhelp"),
+                            &["tenant_id", "shard_id", "timeline_id"],
+                        )
+                        .unwrap();
+
+                        std::thread::scope(|s| {
+                            for (tenant_id, shard_id, timeline_id) in &timelines {
+                                s.spawn(|| {
+                                    let metric = metric_vec.with_label_values(&[
+                                        tenant_id,
+                                        shard_id,
+                                        timeline_id,
+                                    ]);
+                                    barrier.wait();
+                                    for _ in 0..iters {
+                                        metric.inc();
+                                    }
+                                    barrier.wait();
+                                });
+                            }
+                            barrier.wait();
+                            let start = std::time::Instant::now();
+                            barrier.wait();
+                            start.elapsed()
+                        })
+                    })
+                },
+            );
+        }
+        g.finish();
+    }
+}
+
+//
+// Demonstrates that even a single metric can be a scalability bottleneck
+// if multiple threads in it concurrently but there's nothing we can do
+// about it without changing the metrics framework to use e.g. sharded counte atomics.
+//
+criterion_group!(
+    single_metric_multicore_scalability,
+    single_metric_multicore_scalability::bench,
+);
+mod single_metric_multicore_scalability {
+    use super::*;
+
+    pub fn bench(c: &mut Criterion) {
+        let mut g = c.benchmark_group("single_metric_multicore_scalability");
+
+        for nthreads in [1, 4, 8] {
+            g.bench_with_input(
+                BenchmarkId::new("nthreads", nthreads),
+                &nthreads,
+                |b, nthreads| {
+                    b.iter_custom(|iters| {
+                        let barrier = std::sync::Barrier::new(*nthreads + 1);
+
+                        let metric = metrics::UIntGauge::new("testmetric", "testhelp").unwrap();
+
+                        std::thread::scope(|s| {
+                            for _ in 0..*nthreads {
+                                s.spawn(|| {
+                                    barrier.wait();
+                                    for _ in 0..iters {
+                                        metric.inc();
+                                    }
+                                    barrier.wait();
+                                });
+                            }
+                            barrier.wait();
+                            let start = std::time::Instant::now();
+                            barrier.wait();
+                            start.elapsed()
+                        })
+                    })
+                },
+            );
+        }
+        g.finish();
+    }
+}
+
+//
+// Demonstrates that even if we cache label value, the propagation of such a cached metric value
+// by Clone'ing it is a scalability bottleneck.
+// The reason is that it's an Arc internally and thus there's contention on the reference count atomics.
+//
+// We can avoid that by having long-lived references per thread (= indirection).
+//
+criterion_group!(
+    propagation_of_cached_label_value,
+    propagation_of_cached_label_value::bench_naive,
+    propagation_of_cached_label_value::bench_long_lived_reference_per_thread,
+);
+mod propagation_of_cached_label_value {
+    use std::sync::Arc;
+
+    use super::*;
+
+    pub fn bench_naive(c: &mut Criterion) {
+        let mut g = c.benchmark_group("propagation_of_cached_label_value__naive");
+
+        for nthreads in [1, 4, 8] {
+            g.bench_with_input(
+                BenchmarkId::new("nthreads", nthreads),
+                &nthreads,
+                |b, nthreads| {
+                    b.iter_custom(|iters| {
+                        let barrier = std::sync::Barrier::new(*nthreads + 1);
+
+                        let metric = metrics::UIntGauge::new("testmetric", "testhelp").unwrap();
+
+                        std::thread::scope(|s| {
+                            for _ in 0..*nthreads {
+                                s.spawn(|| {
+                                    barrier.wait();
+                                    for _ in 0..iters {
+                                        // propagating the metric means we'd clone it into the child RequestContext
+                                        let propagated = metric.clone();
+                                        // simulate some work
+                                        criterion::black_box(propagated);
+                                    }
+                                    barrier.wait();
+                                });
+                            }
+                            barrier.wait();
+                            let start = std::time::Instant::now();
+                            barrier.wait();
+                            start.elapsed()
+                        })
+                    })
+                },
+            );
+        }
+        g.finish();
+    }
+
+    pub fn bench_long_lived_reference_per_thread(c: &mut Criterion) {
+        let mut g =
+            c.benchmark_group("propagation_of_cached_label_value__long_lived_reference_per_thread");
+
+        for nthreads in [1, 4, 8] {
+            g.bench_with_input(
+                BenchmarkId::new("nthreads", nthreads),
+                &nthreads,
+                |b, nthreads| {
+                    b.iter_custom(|iters| {
+                        let barrier = std::sync::Barrier::new(*nthreads + 1);
+
+                        let metric = metrics::UIntGauge::new("testmetric", "testhelp").unwrap();
+
+                        std::thread::scope(|s| {
+                            for _ in 0..*nthreads {
+                                s.spawn(|| {
+                                    // This is the technique.
+                                    let this_threads_metric_reference = Arc::new(metric.clone());
+
+                                    barrier.wait();
+                                    for _ in 0..iters {
+                                        // propagating the metric means we'd clone it into the child RequestContext
+                                        let propagated = Arc::clone(&this_threads_metric_reference);
+                                        // simulate some work (include the pointer chase!)
+                                        criterion::black_box(&*propagated);
+                                    }
+                                    barrier.wait();
+                                });
+                            }
+                            barrier.wait();
+                            let start = std::time::Instant::now();
+                            barrier.wait();
+                            start.elapsed()
+                        })
+                    })
+                },
+            );
+        }
+    }
+}
+
+criterion_main!(
+    label_values,
+    single_metric_multicore_scalability,
+    propagation_of_cached_label_value
+);
+
+/*
+RUST_BACKTRACE=full cargo bench --bench bench_metrics --  --discard-baseline --noplot
+
+Results on an im4gn.2xlarge instance
+
+label_values__naive_usage/ntimelines/1 time:   [178.71 ns 178.74 ns 178.76 ns]
+label_values__naive_usage/ntimelines/4 time:   [532.94 ns 539.59 ns 546.31 ns]
+label_values__naive_usage/ntimelines/8 time:   [1.1082 µs 1.1109 µs 1.1135 µs]
+label_values__cache_label_values_lookup/ntimelines/1 time:   [6.4116 ns 6.4119 ns 6.4123 ns]
+label_values__cache_label_values_lookup/ntimelines/4 time:   [6.3482 ns 6.3819 ns 6.4079 ns]
+label_values__cache_label_values_lookup/ntimelines/8 time:   [6.4213 ns 6.5279 ns 6.6293 ns]
+single_metric_multicore_scalability/nthreads/1 time:   [6.0102 ns 6.0104 ns 6.0106 ns]
+single_metric_multicore_scalability/nthreads/4 time:   [38.127 ns 38.275 ns 38.416 ns]
+single_metric_multicore_scalability/nthreads/8 time:   [73.698 ns 74.882 ns 75.864 ns]
+propagation_of_cached_label_value__naive/nthreads/1 time:   [14.424 ns 14.425 ns 14.426 ns]
+propagation_of_cached_label_value__naive/nthreads/4 time:   [100.71 ns 102.53 ns 104.35 ns]
+propagation_of_cached_label_value__naive/nthreads/8 time:   [211.50 ns 214.44 ns 216.87 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/1 time:   [14.135 ns 14.147 ns 14.160 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/4 time:   [14.243 ns 14.255 ns 14.268 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/8 time:   [14.470 ns 14.682 ns 14.895 ns]
+
+Results on an i3en.3xlarge instance
+
+label_values__naive_usage/ntimelines/1      time:   [117.32 ns 117.53 ns 117.74 ns]
+label_values__naive_usage/ntimelines/4      time:   [736.58 ns 741.12 ns 745.61 ns]
+label_values__naive_usage/ntimelines/8      time:   [1.4513 µs 1.4596 µs 1.4665 µs]
+label_values__cache_label_values_lookup/ntimelines/1      time:   [8.0964 ns 8.0979 ns 8.0995 ns]
+label_values__cache_label_values_lookup/ntimelines/4      time:   [8.1620 ns 8.2912 ns 8.4491 ns]
+label_values__cache_label_values_lookup/ntimelines/8      time:   [14.148 ns 14.237 ns 14.324 ns]
+single_metric_multicore_scalability/nthreads/1      time:   [8.0993 ns 8.1013 ns 8.1046 ns]
+single_metric_multicore_scalability/nthreads/4      time:   [80.039 ns 80.672 ns 81.297 ns]
+single_metric_multicore_scalability/nthreads/8      time:   [153.58 ns 154.23 ns 154.90 ns]
+propagation_of_cached_label_value__naive/nthreads/1     time:   [13.924 ns 13.926 ns 13.928 ns]
+propagation_of_cached_label_value__naive/nthreads/4     time:   [143.66 ns 145.27 ns 146.59 ns]
+propagation_of_cached_label_value__naive/nthreads/8     time:   [296.51 ns 297.90 ns 299.30 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/1     time:   [14.013 ns 14.149 ns 14.308 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/4     time:   [14.311 ns 14.625 ns 14.984 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/8     time:   [25.981 ns 26.227 ns 26.476 ns]
+
+Results on an Standard L16s v3 (16 vcpus, 128 GiB memory)  Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
+
+label_values__naive_usage/ntimelines/1      time:   [101.63 ns 101.84 ns 102.06 ns]
+label_values__naive_usage/ntimelines/4      time:   [417.55 ns 424.73 ns 432.63 ns]
+label_values__naive_usage/ntimelines/8      time:   [874.91 ns 889.51 ns 904.25 ns]
+label_values__cache_label_values_lookup/ntimelines/1      time:   [5.7724 ns 5.7760 ns 5.7804 ns]
+label_values__cache_label_values_lookup/ntimelines/4      time:   [7.8878 ns 7.9401 ns 8.0034 ns]
+label_values__cache_label_values_lookup/ntimelines/8      time:   [7.2621 ns 7.6354 ns 8.0337 ns]
+single_metric_multicore_scalability/nthreads/1      time:   [5.7710 ns 5.7744 ns 5.7785 ns]
+single_metric_multicore_scalability/nthreads/4      time:   [66.629 ns 66.994 ns 67.336 ns]
+single_metric_multicore_scalability/nthreads/8      time:   [130.85 ns 131.98 ns 132.91 ns]
+propagation_of_cached_label_value__naive/nthreads/1     time:   [11.540 ns 11.546 ns 11.553 ns]
+propagation_of_cached_label_value__naive/nthreads/4     time:   [131.22 ns 131.90 ns 132.56 ns]
+propagation_of_cached_label_value__naive/nthreads/8     time:   [260.99 ns 262.75 ns 264.26 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/1     time:   [11.544 ns 11.550 ns 11.557 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/4     time:   [11.568 ns 11.642 ns 11.763 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/8     time:   [13.416 ns 14.121 ns 14.886 ns
+
+Results on an M4 MAX MacBook Pro   Total Number of Cores:	14 (10 performance and 4 efficiency)
+
+label_values__naive_usage/ntimelines/1      time:   [52.711 ns 53.026 ns 53.381 ns]
+label_values__naive_usage/ntimelines/4      time:   [323.99 ns 330.40 ns 337.53 ns]
+label_values__naive_usage/ntimelines/8      time:   [1.1615 µs 1.1998 µs 1.2399 µs]
+label_values__cache_label_values_lookup/ntimelines/1      time:   [1.6635 ns 1.6715 ns 1.6809 ns]
+label_values__cache_label_values_lookup/ntimelines/4      time:   [1.7786 ns 1.7876 ns 1.8028 ns]
+label_values__cache_label_values_lookup/ntimelines/8      time:   [1.8195 ns 1.8371 ns 1.8665 ns]
+single_metric_multicore_scalability/nthreads/1      time:   [1.7764 ns 1.7909 ns 1.8079 ns]
+single_metric_multicore_scalability/nthreads/4      time:   [33.875 ns 34.868 ns 35.923 ns]
+single_metric_multicore_scalability/nthreads/8      time:   [226.85 ns 235.30 ns 244.18 ns]
+propagation_of_cached_label_value__naive/nthreads/1     time:   [3.4337 ns 3.4491 ns 3.4660 ns]
+propagation_of_cached_label_value__naive/nthreads/4     time:   [69.486 ns 71.937 ns 74.472 ns]
+propagation_of_cached_label_value__naive/nthreads/8     time:   [434.87 ns 456.47 ns 477.84 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/1     time:   [3.3767 ns 3.3974 ns 3.4220 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/4     time:   [3.6105 ns 4.2355 ns 5.1463 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/8     time:   [4.0889 ns 4.9714 ns 6.0779 ns]
+
+Results on a Hetzner AX102 AMD Ryzen 9 7950X3D 16-Core Processor
+
+label_values__naive_usage/ntimelines/1      time:   [64.510 ns 64.559 ns 64.610 ns]
+label_values__naive_usage/ntimelines/4      time:   [309.71 ns 326.09 ns 342.32 ns]
+label_values__naive_usage/ntimelines/8      time:   [776.92 ns 819.35 ns 856.93 ns]
+label_values__cache_label_values_lookup/ntimelines/1      time:   [1.2855 ns 1.2943 ns 1.3021 ns]
+label_values__cache_label_values_lookup/ntimelines/4      time:   [1.3865 ns 1.4139 ns 1.4441 ns]
+label_values__cache_label_values_lookup/ntimelines/8      time:   [1.5311 ns 1.5669 ns 1.6046 ns]
+single_metric_multicore_scalability/nthreads/1      time:   [1.1927 ns 1.1981 ns 1.2049 ns]
+single_metric_multicore_scalability/nthreads/4      time:   [24.346 ns 25.439 ns 26.634 ns]
+single_metric_multicore_scalability/nthreads/8      time:   [58.666 ns 60.137 ns 61.486 ns]
+propagation_of_cached_label_value__naive/nthreads/1     time:   [2.7067 ns 2.7238 ns 2.7402 ns]
+propagation_of_cached_label_value__naive/nthreads/4     time:   [62.723 ns 66.214 ns 69.787 ns]
+propagation_of_cached_label_value__naive/nthreads/8     time:   [164.24 ns 170.10 ns 175.68 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/1     time:   [2.2915 ns 2.2960 ns 2.3012 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/4     time:   [2.5726 ns 2.6158 ns 2.6624 ns]
+propagation_of_cached_label_value__long_lived_reference_per_thread/nthreads/8     time:   [2.7068 ns 2.8243 ns 2.9824 ns]
+
+*/

pageserver/client/src/mgmt_api.rs

@@ -7,7 +7,7 @@ use http_utils::error::HttpErrorBody;
 use pageserver_api::models::*;
 use pageserver_api::shard::TenantShardId;
 pub use reqwest::Body as ReqwestBody;
-use reqwest::{IntoUrl, Method, StatusCode};
+use reqwest::{Certificate, IntoUrl, Method, StatusCode};
 use utils::id::{TenantId, TimelineId};
 use utils::lsn::Lsn;
 
@@ -38,6 +38,9 @@ pub enum Error {
 
     #[error("Cancelled")]
     Cancelled,
+
+    #[error("create client: {0}{}", .0.source().map(|e| format!(": {e}")).unwrap_or_default())]
+    CreateClient(reqwest::Error),
 }
 
 pub type Result<T> = std::result::Result<T, Error>;
@@ -69,8 +72,17 @@ pub enum ForceAwaitLogicalSize {
 }
 
 impl Client {
-    pub fn new(mgmt_api_endpoint: String, jwt: Option<&str>) -> Self {
-        Self::from_client(reqwest::Client::new(), mgmt_api_endpoint, jwt)
+    pub fn new(
+        mgmt_api_endpoint: String,
+        jwt: Option<&str>,
+        ssl_ca_cert: Option<Certificate>,
+    ) -> Result<Self> {
+        let mut http_client = reqwest::Client::builder();
+        if let Some(ssl_ca_cert) = ssl_ca_cert {
+            http_client = http_client.add_root_certificate(ssl_ca_cert);
+        }
+        let http_client = http_client.build().map_err(Error::CreateClient)?;
+        Ok(Self::from_client(http_client, mgmt_api_endpoint, jwt))
     }
 
     pub fn from_client(
@@ -101,12 +113,10 @@ impl Client {
         debug_assert!(path.starts_with('/'));
         let uri = format!("{}{}", self.mgmt_api_endpoint, path);
 
-        let req = self.client.request(Method::GET, uri);
-        let req = if let Some(value) = &self.authorization_header {
-            req.header(reqwest::header::AUTHORIZATION, value)
-        } else {
-            req
-        };
+        let mut req = self.client.request(Method::GET, uri);
+        if let Some(value) = &self.authorization_header {
+            req = req.header(reqwest::header::AUTHORIZATION, value);
+        }
         req.send().await.map_err(Error::ReceiveBody)
     }
 

pageserver/compaction/tests/tests.rs

@@ -12,7 +12,7 @@ pub(crate) fn setup_logging() {
             logging::TracingErrorLayerEnablement::EnableWithRustLogFilter,
             logging::Output::Stdout,
         )
-        .expect("Failed to init test logging")
+        .expect("Failed to init test logging");
     });
 }
 

pageserver/pagebench/Cargo.toml

@@ -15,6 +15,7 @@ hdrhistogram.workspace = true
 humantime.workspace = true
 humantime-serde.workspace = true
 rand.workspace = true
+reqwest.workspace=true
 serde.workspace = true
 serde_json.workspace = true
 tracing.workspace = true

pageserver/pagebench/src/cmd/aux_files.rs

@@ -36,7 +36,8 @@ async fn main_impl(args: Args) -> anyhow::Result<()> {
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-    ));
+        None, // TODO: support ssl_ca_file for https APIs in pagebench.
+    )?);
 
     // discover targets
     let timelines: Vec<TenantTimelineId> = crate::util::cli::targets::discover(

pageserver/pagebench/src/cmd/basebackup.rs

@@ -77,7 +77,8 @@ async fn main_impl(
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-    ));
+        None, // TODO: support ssl_ca_file for https APIs in pagebench.
+    )?);
 
     // discover targets
     let timelines: Vec<TenantTimelineId> = crate::util::cli::targets::discover(

pageserver/pagebench/src/cmd/getpage_latest_lsn.rs

@@ -125,7 +125,8 @@ async fn main_impl(
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-    ));
+        None, // TODO: support ssl_ca_file for https APIs in pagebench.
+    )?);
 
     if let Some(engine_str) = &args.set_io_engine {
         mgmt_api_client.put_io_engine(engine_str).await?;

pageserver/pagebench/src/cmd/ondemand_download_churn.rs

@@ -83,7 +83,8 @@ async fn main_impl(args: Args) -> anyhow::Result<()> {
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-    ));
+        None, // TODO: support ssl_ca_file for https APIs in pagebench.
+    )?);
 
     if let Some(engine_str) = &args.set_io_engine {
         mgmt_api_client.put_io_engine(engine_str).await?;

pageserver/pagebench/src/cmd/trigger_initial_size_calculation.rs

@@ -40,7 +40,8 @@ async fn main_impl(args: Args) -> anyhow::Result<()> {
     let mgmt_api_client = Arc::new(pageserver_client::mgmt_api::Client::new(
         args.mgmt_api_endpoint.clone(),
         args.pageserver_jwt.as_deref(),
-    ));
+        None, // TODO: support ssl_ca_file for https APIs in pagebench.
+    )?);
 
     // discover targets
     let timelines: Vec<TenantTimelineId> = crate::util::cli::targets::discover(

pageserver/src/bin/pageserver.rs

@@ -25,11 +25,12 @@ use pageserver::task_mgr::{
 };
 use pageserver::tenant::{TenantSharedResources, mgr, secondary};
 use pageserver::{
-    CancellableTask, ConsumptionMetricsTasks, HttpEndpointListener, http, page_cache, page_service,
-    task_mgr, virtual_file,
+    CancellableTask, ConsumptionMetricsTasks, HttpEndpointListener, HttpsEndpointListener, http,
+    page_cache, page_service, task_mgr, virtual_file,
 };
 use postgres_backend::AuthType;
 use remote_storage::GenericRemoteStorage;
+use rustls_pki_types::{CertificateDer, PrivateKeyDer};
 use tokio::signal::unix::SignalKind;
 use tokio::time::Instant;
 use tokio_util::sync::CancellationToken;
@@ -110,6 +111,7 @@ fn main() -> anyhow::Result<()> {
     } else {
         TracingErrorLayerEnablement::Disabled
     };
+
     logging::init(
         conf.log_format,
         tracing_error_layer_enablement,
@@ -343,8 +345,15 @@ fn start_pageserver(
     info!("Starting pageserver http handler on {http_addr}");
     let http_listener = tcp_listener::bind(http_addr)?;
 
-    let pg_addr = &conf.listen_pg_addr;
+    let https_listener = match conf.listen_https_addr.as_ref() {
+        Some(https_addr) => {
+            info!("Starting pageserver https handler on {https_addr}");
+            Some(tcp_listener::bind(https_addr)?)
+        }
+        None => None,
+    };
 
+    let pg_addr = &conf.listen_pg_addr;
     info!("Starting pageserver pg protocol handler on {pg_addr}");
     let pageserver_listener = tcp_listener::bind(pg_addr)?;
 
@@ -575,9 +584,8 @@ fn start_pageserver(
 
     // Start up the service to handle HTTP mgmt API request. We created the
     // listener earlier already.
-    let http_endpoint_listener = {
+    let (http_endpoint_listener, https_endpoint_listener) = {
         let _rt_guard = MGMT_REQUEST_RUNTIME.enter(); // for hyper
-        let cancel = CancellationToken::new();
 
         let router_state = Arc::new(
             http::routes::State::new(
@@ -592,22 +600,51 @@ fn start_pageserver(
             )
             .context("Failed to initialize router state")?,
         );
+
         let router = http::make_router(router_state, launch_ts, http_auth.clone())?
             .build()
             .map_err(|err| anyhow!(err))?;
-        let service = http_utils::RouterService::new(router).unwrap();
-        let server = hyper0::Server::from_tcp(http_listener)?
-            .serve(service)
-            .with_graceful_shutdown({
-                let cancel = cancel.clone();
-                async move { cancel.clone().cancelled().await }
-            });
 
-        let task = MGMT_REQUEST_RUNTIME.spawn(task_mgr::exit_on_panic_or_error(
-            "http endpoint listener",
-            server,
-        ));
-        HttpEndpointListener(CancellableTask { task, cancel })
+        let service =
+            Arc::new(http_utils::RequestServiceBuilder::new(router).map_err(|err| anyhow!(err))?);
+
+        let http_task = {
+            let server =
+                http_utils::server::Server::new(Arc::clone(&service), http_listener, None)?;
+            let cancel = CancellationToken::new();
+
+            let task = MGMT_REQUEST_RUNTIME.spawn(task_mgr::exit_on_panic_or_error(
+                "http endpoint listener",
+                server.serve(cancel.clone()),
+            ));
+            HttpEndpointListener(CancellableTask { task, cancel })
+        };
+
+        let https_task = match https_listener {
+            Some(https_listener) => {
+                let certs = load_certs(&conf.ssl_cert_file)?;
+                let key = load_private_key(&conf.ssl_key_file)?;
+
+                let server_config = rustls::ServerConfig::builder()
+                    .with_no_client_auth()
+                    .with_single_cert(certs, key)?;
+
+                let tls_acceptor = tokio_rustls::TlsAcceptor::from(Arc::new(server_config));
+
+                let server =
+                    http_utils::server::Server::new(service, https_listener, Some(tls_acceptor))?;
+                let cancel = CancellationToken::new();
+
+                let task = MGMT_REQUEST_RUNTIME.spawn(task_mgr::exit_on_panic_or_error(
+                    "https endpoint listener",
+                    server.serve(cancel.clone()),
+                ));
+                Some(HttpsEndpointListener(CancellableTask { task, cancel }))
+            }
+            None => None,
+        };
+
+        (http_task, https_task)
     };
 
     let consumption_metrics_tasks = {
@@ -683,6 +720,7 @@ fn start_pageserver(
         shutdown_pageserver.cancel();
         pageserver::shutdown_pageserver(
             http_endpoint_listener,
+            https_endpoint_listener,
             page_service,
             consumption_metrics_tasks,
             disk_usage_eviction_task,
@@ -697,6 +735,25 @@ fn start_pageserver(
     })
 }
 
+fn load_certs(filename: &Utf8Path) -> std::io::Result<Vec<CertificateDer<'static>>> {
+    let file = std::fs::File::open(filename)?;
+    let mut reader = std::io::BufReader::new(file);
+
+    rustls_pemfile::certs(&mut reader).collect()
+}
+
+fn load_private_key(filename: &Utf8Path) -> anyhow::Result<PrivateKeyDer<'static>> {
+    let file = std::fs::File::open(filename)?;
+    let mut reader = std::io::BufReader::new(file);
+
+    let key = rustls_pemfile::private_key(&mut reader)?;
+
+    key.ok_or(anyhow::anyhow!(
+        "no private key found in {}",
+        filename.as_str(),
+    ))
+}
+
 async fn create_remote_storage_client(
     conf: &'static PageServerConf,
 ) -> anyhow::Result<GenericRemoteStorage> {

pageserver/src/config.rs

@@ -53,6 +53,11 @@ pub struct PageServerConf {
     pub listen_pg_addr: String,
     /// Example (default): 127.0.0.1:9898
     pub listen_http_addr: String,
+    /// Example: 127.0.0.1:9899
+    pub listen_https_addr: Option<String>,
+
+    pub ssl_key_file: Utf8PathBuf,
+    pub ssl_cert_file: Utf8PathBuf,
 
     /// Current availability zone. Used for traffic metrics.
     pub availability_zone: Option<String>,
@@ -317,6 +322,9 @@ impl PageServerConf {
         let pageserver_api::config::ConfigToml {
             listen_pg_addr,
             listen_http_addr,
+            listen_https_addr,
+            ssl_key_file,
+            ssl_cert_file,
             availability_zone,
             wait_lsn_timeout,
             wal_redo_timeout,
@@ -375,6 +383,9 @@ impl PageServerConf {
             // ------------------------------------------------------------
             listen_pg_addr,
             listen_http_addr,
+            listen_https_addr,
+            ssl_key_file,
+            ssl_cert_file,
             availability_zone,
             wait_lsn_timeout,
             wal_redo_timeout,
@@ -456,8 +467,8 @@ impl PageServerConf {
             no_sync: no_sync.unwrap_or(false),
             enable_read_path_debugging: enable_read_path_debugging.unwrap_or(false),
             validate_wal_contiguity: validate_wal_contiguity.unwrap_or(false),
-            load_previous_heatmap: load_previous_heatmap.unwrap_or(false),
-            generate_unarchival_heatmap: generate_unarchival_heatmap.unwrap_or(false),
+            load_previous_heatmap: load_previous_heatmap.unwrap_or(true),
+            generate_unarchival_heatmap: generate_unarchival_heatmap.unwrap_or(true),
         };
 
         // ------------------------------------------------------------

pageserver/src/context.rs

@@ -89,16 +89,112 @@
 //! [`RequestContext`] argument. Functions in the middle of the call chain
 //! only need to pass it on.
 
-use crate::task_mgr::TaskKind;
+use std::sync::Arc;
+
+use once_cell::sync::Lazy;
+use tracing::warn;
+use utils::{id::TimelineId, shard::TenantShardId};
+
+use crate::{
+    metrics::{StorageIoSizeMetrics, TimelineMetrics},
+    task_mgr::TaskKind,
+    tenant::Timeline,
+};
 
 // The main structure of this module, see module-level comment.
-#[derive(Debug)]
 pub struct RequestContext {
     task_kind: TaskKind,
     download_behavior: DownloadBehavior,
     access_stats_behavior: AccessStatsBehavior,
     page_content_kind: PageContentKind,
     read_path_debug: bool,
+    scope: Scope,
+}
+
+#[derive(Clone)]
+pub(crate) enum Scope {
+    Global {
+        io_size_metrics: &'static crate::metrics::StorageIoSizeMetrics,
+    },
+    SecondaryTenant {
+        io_size_metrics: &'static crate::metrics::StorageIoSizeMetrics,
+    },
+    SecondaryTimeline {
+        io_size_metrics: crate::metrics::StorageIoSizeMetrics,
+    },
+    Timeline {
+        // We wrap the `Arc<TimelineMetrics>`s inside another Arc to avoid child
+        // context creation contending for the ref counters of the Arc<TimelineMetrics>,
+        // which are shared among all tasks that operate on the timeline, especially
+        // concurrent page_service connections.
+        #[allow(clippy::redundant_allocation)]
+        arc_arc: Arc<Arc<TimelineMetrics>>,
+    },
+    #[cfg(test)]
+    UnitTest {
+        io_size_metrics: &'static crate::metrics::StorageIoSizeMetrics,
+    },
+}
+
+static GLOBAL_IO_SIZE_METRICS: Lazy<crate::metrics::StorageIoSizeMetrics> =
+    Lazy::new(|| crate::metrics::StorageIoSizeMetrics::new("*", "*", "*"));
+
+impl Scope {
+    pub(crate) fn new_global() -> Self {
+        Scope::Global {
+            io_size_metrics: &GLOBAL_IO_SIZE_METRICS,
+        }
+    }
+    /// NB: this allocates, so, use only at relatively long-lived roots, e.g., at start
+    /// of a compaction iteration.
+    pub(crate) fn new_timeline(timeline: &Timeline) -> Self {
+        Scope::Timeline {
+            arc_arc: Arc::new(Arc::clone(&timeline.metrics)),
+        }
+    }
+    pub(crate) fn new_page_service_pagestream(
+        timeline_handle: &crate::tenant::timeline::handle::Handle<
+            crate::page_service::TenantManagerTypes,
+        >,
+    ) -> Self {
+        Scope::Timeline {
+            arc_arc: Arc::clone(&timeline_handle.metrics),
+        }
+    }
+    pub(crate) fn new_secondary_timeline(
+        tenant_shard_id: &TenantShardId,
+        timeline_id: &TimelineId,
+    ) -> Self {
+        // TODO(https://github.com/neondatabase/neon/issues/11156): secondary timelines have no infrastructure for metrics lifecycle.
+
+        let tenant_id = tenant_shard_id.tenant_id.to_string();
+        let shard_id = tenant_shard_id.shard_slug().to_string();
+        let timeline_id = timeline_id.to_string();
+
+        let io_size_metrics =
+            crate::metrics::StorageIoSizeMetrics::new(&tenant_id, &shard_id, &timeline_id);
+        Scope::SecondaryTimeline { io_size_metrics }
+    }
+    pub(crate) fn new_secondary_tenant(_tenant_shard_id: &TenantShardId) -> Self {
+        // Before propagating metrics via RequestContext, the labels were inferred from file path.
+        // The only user of VirtualFile at tenant scope is the heatmap download & read.
+        // The inferred labels for the path of the heatmap file on local disk were that of the global metric (*,*,*).
+        // Thus, we do the same here, and extend that for anything secondary-tenant scoped.
+        //
+        // If we want to have (tenant_id, shard_id, '*') labels for secondary tenants in the future,
+        // we will need to think about the metric lifecycle, i.e., remove them during secondary tenant shutdown,
+        // like we do for attached timelines. (We don't have attached-tenant-scoped usage of VirtualFile
+        // at this point, so, we were able to completely side-step tenant-scoped stuff there).
+        Scope::SecondaryTenant {
+            io_size_metrics: &GLOBAL_IO_SIZE_METRICS,
+        }
+    }
+    #[cfg(test)]
+    pub(crate) fn new_unit_test() -> Self {
+        Scope::UnitTest {
+            io_size_metrics: &GLOBAL_IO_SIZE_METRICS,
+        }
+    }
 }
 
 /// The kind of access to the page cache.
@@ -157,6 +253,7 @@ impl RequestContextBuilder {
                 access_stats_behavior: AccessStatsBehavior::Update,
                 page_content_kind: PageContentKind::Unknown,
                 read_path_debug: false,
+                scope: Scope::new_global(),
             },
         }
     }
@@ -171,10 +268,16 @@ impl RequestContextBuilder {
                 access_stats_behavior: original.access_stats_behavior,
                 page_content_kind: original.page_content_kind,
                 read_path_debug: original.read_path_debug,
+                scope: original.scope.clone(),
             },
         }
     }
 
+    pub fn task_kind(mut self, k: TaskKind) -> Self {
+        self.inner.task_kind = k;
+        self
+    }
+
     /// Configure the DownloadBehavior of the context: whether to
     /// download missing layers, and/or warn on the download.
     pub fn download_behavior(mut self, b: DownloadBehavior) -> Self {
@@ -199,6 +302,11 @@ impl RequestContextBuilder {
         self
     }
 
+    pub(crate) fn scope(mut self, s: Scope) -> Self {
+        self.inner.scope = s;
+        self
+    }
+
     pub fn build(self) -> RequestContext {
         self.inner
     }
@@ -281,7 +389,50 @@ impl RequestContext {
     }
 
     fn child_impl(&self, task_kind: TaskKind, download_behavior: DownloadBehavior) -> Self {
-        Self::new(task_kind, download_behavior)
+        RequestContextBuilder::extend(self)
+            .task_kind(task_kind)
+            .download_behavior(download_behavior)
+            .build()
+    }
+
+    pub fn with_scope_timeline(&self, timeline: &Arc<Timeline>) -> Self {
+        RequestContextBuilder::extend(self)
+            .scope(Scope::new_timeline(timeline))
+            .build()
+    }
+
+    pub(crate) fn with_scope_page_service_pagestream(
+        &self,
+        timeline_handle: &crate::tenant::timeline::handle::Handle<
+            crate::page_service::TenantManagerTypes,
+        >,
+    ) -> Self {
+        RequestContextBuilder::extend(self)
+            .scope(Scope::new_page_service_pagestream(timeline_handle))
+            .build()
+    }
+
+    pub fn with_scope_secondary_timeline(
+        &self,
+        tenant_shard_id: &TenantShardId,
+        timeline_id: &TimelineId,
+    ) -> Self {
+        RequestContextBuilder::extend(self)
+            .scope(Scope::new_secondary_timeline(tenant_shard_id, timeline_id))
+            .build()
+    }
+
+    pub fn with_scope_secondary_tenant(&self, tenant_shard_id: &TenantShardId) -> Self {
+        RequestContextBuilder::extend(self)
+            .scope(Scope::new_secondary_tenant(tenant_shard_id))
+            .build()
+    }
+
+    #[cfg(test)]
+    pub fn with_scope_unit_test(&self) -> Self {
+        RequestContextBuilder::new(TaskKind::UnitTest)
+            .scope(Scope::new_unit_test())
+            .build()
     }
 
     pub fn task_kind(&self) -> TaskKind {
@@ -303,4 +454,38 @@ impl RequestContext {
     pub(crate) fn read_path_debug(&self) -> bool {
         self.read_path_debug
     }
+
+    pub(crate) fn io_size_metrics(&self) -> &StorageIoSizeMetrics {
+        match &self.scope {
+            Scope::Global { io_size_metrics } => {
+                let is_unit_test = cfg!(test);
+                let is_regress_test_build = cfg!(feature = "testing");
+                if is_unit_test || is_regress_test_build {
+                    panic!("all VirtualFile instances are timeline-scoped");
+                } else {
+                    use once_cell::sync::Lazy;
+                    use std::sync::Mutex;
+                    use std::time::Duration;
+                    use utils::rate_limit::RateLimit;
+                    static LIMIT: Lazy<Mutex<RateLimit>> =
+                        Lazy::new(|| Mutex::new(RateLimit::new(Duration::from_secs(1))));
+                    let mut guard = LIMIT.lock().unwrap();
+                    guard.call2(|rate_limit_stats| {
+                        warn!(
+                            %rate_limit_stats,
+                            backtrace=%std::backtrace::Backtrace::force_capture(),
+                            "all VirtualFile instances are timeline-scoped",
+                        );
+                    });
+
+                    io_size_metrics
+                }
+            }
+            Scope::Timeline { arc_arc } => &arc_arc.storage_io_size,
+            Scope::SecondaryTimeline { io_size_metrics } => io_size_metrics,
+            Scope::SecondaryTenant { io_size_metrics } => io_size_metrics,
+            #[cfg(test)]
+            Scope::UnitTest { io_size_metrics } => io_size_metrics,
+        }
+    }
 }

pageserver/src/controller_upcall_client.rs

@@ -181,7 +181,7 @@ impl ControlPlaneGenerationsApi for ControllerUpcallClient {
                         listen_pg_port: m.postgres_port,
                         listen_http_addr: m.http_host,
                         listen_http_port: m.http_port,
-                        listen_https_port: None, // TODO: Support https.
+                        listen_https_port: m.https_port,
                         availability_zone_id: az_id.expect("Checked above"),
                     })
                 }

pageserver/src/http/openapi_spec.yml

@@ -1079,7 +1079,6 @@ components:
         - last_record_lsn
         - disk_consistent_lsn
         - state
-        - latest_gc_cutoff_lsn
       properties:
         timeline_id:
           type: string
@@ -1123,9 +1122,6 @@ components:
         min_readable_lsn:
           type: string
           format: hex
-        latest_gc_cutoff_lsn:
-          type: string
-          format: hex
         applied_gc_cutoff_lsn:
           type: string
           format: hex

pageserver/src/http/routes.rs

@@ -37,7 +37,8 @@ use pageserver_api::models::{
     TenantShardSplitResponse, TenantSorting, TenantState, TenantWaitLsnRequest,
     TimelineArchivalConfigRequest, TimelineCreateRequest, TimelineCreateRequestMode,
     TimelineCreateRequestModeImportPgdata, TimelineGcRequest, TimelineInfo,
-    TimelinesInfoAndOffloaded, TopTenantShardItem, TopTenantShardsRequest, TopTenantShardsResponse,
+    TimelinePatchIndexPartRequest, TimelinesInfoAndOffloaded, TopTenantShardItem,
+    TopTenantShardsRequest, TopTenantShardsResponse,
 };
 use pageserver_api::shard::{ShardCount, TenantShardId};
 use remote_storage::{DownloadError, GenericRemoteStorage, TimeTravelError};
@@ -54,6 +55,7 @@ use utils::id::{TenantId, TimelineId};
 use utils::lsn::Lsn;
 
 use crate::config::PageServerConf;
+use crate::context;
 use crate::context::{DownloadBehavior, RequestContext, RequestContextBuilder};
 use crate::deletion_queue::DeletionQueueClient;
 use crate::pgdatadir_mapping::LsnForTimestamp;
@@ -63,12 +65,14 @@ use crate::tenant::mgr::{
     GetActiveTenantError, GetTenantError, TenantManager, TenantMapError, TenantMapInsertError,
     TenantSlot, TenantSlotError, TenantSlotUpsertError, TenantStateError, UpsertLocationError,
 };
+use crate::tenant::remote_timeline_client::index::GcCompactionState;
 use crate::tenant::remote_timeline_client::{
     download_index_part, list_remote_tenant_shards, list_remote_timelines,
 };
 use crate::tenant::secondary::SecondaryController;
 use crate::tenant::size::ModelInputs;
 use crate::tenant::storage_layer::{IoConcurrency, LayerAccessStatsReset, LayerName};
+use crate::tenant::timeline::detach_ancestor::DetachBehavior;
 use crate::tenant::timeline::offload::{OffloadError, offload_timeline};
 use crate::tenant::timeline::{
     CompactFlags, CompactOptions, CompactRequest, CompactionError, Timeline, WaitLsnTimeout,
@@ -457,10 +461,7 @@ async fn build_timeline_info_common(
         initdb_lsn,
         last_record_lsn,
         prev_record_lsn: Some(timeline.get_prev_record_lsn()),
-        // Externally, expose the lowest LSN that can be used to create a branch as the "GC cutoff", although internally
-        // we distinguish between the "planned" GC cutoff (PITR point) and the "latest" GC cutoff (where we
-        // actually trimmed data to), which can pass each other when PITR is changed.
-        latest_gc_cutoff_lsn: min_readable_lsn,
+        _unused: Default::default(), // Unused, for legacy decode only
         min_readable_lsn,
         applied_gc_cutoff_lsn: *timeline.get_applied_gc_cutoff_lsn(),
         current_logical_size: current_logical_size.size_dont_care_about_accuracy(),
@@ -858,6 +859,75 @@ async fn timeline_archival_config_handler(
     json_response(StatusCode::OK, ())
 }
 
+/// This API is used to patch the index part of a timeline. You must ensure such patches are safe to apply. Use this API as an emergency
+/// measure only.
+///
+/// Some examples of safe patches:
+/// - Increase the gc_cutoff and gc_compaction_cutoff to a larger value in case of a bug that didn't bump the cutoff and cause read errors.
+/// - Force set the index part to use reldir v2 (migrating/migrated).
+///
+/// Some examples of unsafe patches:
+/// - Force set the index part from v2 to v1 (legacy). This will cause the code path to ignore anything written to the new keyspace and cause
+///   errors.
+/// - Decrease the gc_cutoff without validating the data really exists. It will cause read errors in the background.
+async fn timeline_patch_index_part_handler(
+    mut request: Request<Body>,
+    _cancel: CancellationToken,
+) -> Result<Response<Body>, ApiError> {
+    let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
+    let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
+
+    let request_data: TimelinePatchIndexPartRequest = json_request(&mut request).await?;
+    check_permission(&request, None)?; // require global permission for this request
+    let state = get_state(&request);
+
+    async {
+        let timeline =
+            active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
+                .await?;
+
+        if let Some(rel_size_migration) = request_data.rel_size_migration {
+            timeline
+                .update_rel_size_v2_status(rel_size_migration)
+                .map_err(ApiError::InternalServerError)?;
+        }
+
+        if let Some(gc_compaction_last_completed_lsn) =
+            request_data.gc_compaction_last_completed_lsn
+        {
+            timeline
+                .update_gc_compaction_state(GcCompactionState {
+                    last_completed_lsn: gc_compaction_last_completed_lsn,
+                })
+                .map_err(ApiError::InternalServerError)?;
+        }
+
+        if let Some(applied_gc_cutoff_lsn) = request_data.applied_gc_cutoff_lsn {
+            {
+                let guard = timeline.applied_gc_cutoff_lsn.lock_for_write();
+                guard.store_and_unlock(applied_gc_cutoff_lsn);
+            }
+        }
+
+        if request_data.force_index_update {
+            timeline
+                .remote_client
+                .force_schedule_index_upload()
+                .context("force schedule index upload")
+                .map_err(ApiError::InternalServerError)?;
+        }
+
+        Ok::<_, ApiError>(())
+    }
+    .instrument(info_span!("timeline_patch_index_part",
+                tenant_id = %tenant_shard_id.tenant_id,
+                shard_id = %tenant_shard_id.shard_slug(),
+                %timeline_id))
+    .await?;
+
+    json_response(StatusCode::OK, ())
+}
+
 async fn timeline_detail_handler(
     request: Request<Body>,
     _cancel: CancellationToken,
@@ -882,12 +952,13 @@ async fn timeline_detail_handler(
         tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
 
         let timeline = tenant.get_timeline(timeline_id, false)?;
+        let ctx = &ctx.with_scope_timeline(&timeline);
 
         let timeline_info = build_timeline_info(
             &timeline,
             include_non_incremental_logical_size.unwrap_or(false),
             force_await_initial_logical_size.unwrap_or(false),
-            &ctx,
+            ctx,
         )
         .await
         .context("get local timeline info")
@@ -931,7 +1002,8 @@ async fn get_lsn_by_timestamp_handler(
     let timeline =
         active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
             .await?;
-    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
+    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download)
+        .with_scope_timeline(&timeline);
     let result = timeline
         .find_lsn_for_timestamp(timestamp_pg, &cancel, &ctx)
         .await?;
@@ -1003,7 +1075,8 @@ async fn get_timestamp_of_lsn_handler(
     let timeline =
         active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
             .await?;
-    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
+    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download)
+        .with_scope_timeline(&timeline);
     let result = timeline.get_timestamp_for_lsn(lsn, &ctx).await?;
 
     match result {
@@ -1358,7 +1431,8 @@ async fn timeline_layer_scan_disposable_keys(
         active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
             .await?;
 
-    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
+    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download)
+        .with_scope_timeline(&timeline);
 
     let guard = timeline.layers.read().await;
     let Some(layer) = guard.try_get_from_key(&layer_name.clone().into()) else {
@@ -1444,7 +1518,8 @@ async fn timeline_download_heatmap_layers_handler(
     let timeline =
         active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
             .await?;
-    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
+    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download)
+        .with_scope_timeline(&timeline);
 
     let max_concurrency = get_config(&request)
         .remote_storage_config
@@ -1492,7 +1567,8 @@ async fn layer_download_handler(
     let timeline =
         active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
             .await?;
-    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
+    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download)
+        .with_scope_timeline(&timeline);
     let downloaded = timeline
         .download_layer(&layer_name, &ctx)
         .await
@@ -2228,8 +2304,8 @@ async fn timeline_compact_handler(
         .unwrap_or(false);
 
     async {
-        let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
         let timeline = active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id).await?;
+        let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download).with_scope_timeline(&timeline);
         if scheduled {
             let tenant = state
                 .tenant_manager
@@ -2336,8 +2412,8 @@ async fn timeline_checkpoint_handler(
         parse_query_param::<_, bool>(&request, "wait_until_uploaded")?.unwrap_or(false);
 
     async {
-        let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
         let timeline = active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id).await?;
+        let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download).with_scope_timeline(&timeline);
         if wait_until_flushed {
             timeline.freeze_and_flush().await
         } else {
@@ -2392,7 +2468,8 @@ async fn timeline_download_remote_layers_handler_post(
     let timeline =
         active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
             .await?;
-    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
+    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download)
+        .with_scope_timeline(&timeline);
     match timeline.spawn_download_all_remote_layers(body, &ctx).await {
         Ok(st) => json_response(StatusCode::ACCEPTED, st),
         Err(st) => json_response(StatusCode::CONFLICT, st),
@@ -2429,6 +2506,8 @@ async fn timeline_detach_ancestor_handler(
     let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
     check_permission(&request, Some(tenant_shard_id.tenant_id))?;
     let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
+    let behavior: Option<DetachBehavior> = parse_query_param(&request, "detach_behavior")?;
+    let behavior = behavior.unwrap_or_default();
 
     let span = tracing::info_span!("detach_ancestor", tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), %timeline_id);
 
@@ -2475,9 +2554,10 @@ async fn timeline_detach_ancestor_handler(
         tracing::info!("all timeline upload queues are drained");
 
         let timeline = tenant.get_timeline(timeline_id, true)?;
+        let ctx = &ctx.with_scope_timeline(&timeline);
 
         let progress = timeline
-            .prepare_to_detach_from_ancestor(&tenant, options, ctx)
+            .prepare_to_detach_from_ancestor(&tenant, options, behavior, ctx)
             .await?;
 
         // uncomment to allow early as possible Tenant::drop
@@ -2492,6 +2572,7 @@ async fn timeline_detach_ancestor_handler(
                         tenant_shard_id,
                         timeline_id,
                         prepared,
+                        behavior,
                         attempt,
                         ctx,
                     )
@@ -2581,8 +2662,9 @@ async fn getpage_at_lsn_handler_inner(
     async {
         let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
         // Enable read path debugging
-        let ctx = RequestContextBuilder::extend(&ctx).read_path_debug(true).build();
         let timeline = active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id).await?;
+        let ctx = RequestContextBuilder::extend(&ctx).read_path_debug(true)
+        .scope(context::Scope::new_timeline(&timeline)).build();
 
         // Use last_record_lsn if no lsn is provided
         let lsn = lsn.unwrap_or_else(|| timeline.get_last_record_lsn());
@@ -2616,8 +2698,8 @@ async fn timeline_collect_keyspace(
     let at_lsn: Option<Lsn> = parse_query_param(&request, "at_lsn")?;
 
     async {
-        let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
         let timeline = active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id).await?;
+        let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download).with_scope_timeline(&timeline);
         let at_lsn = at_lsn.unwrap_or_else(|| timeline.get_last_record_lsn());
         let (dense_ks, sparse_ks) = timeline
             .collect_keyspace(at_lsn, &ctx)
@@ -3142,6 +3224,7 @@ async fn post_top_tenants(
         match order_by {
             TenantSorting::ResidentSize => sizes.resident_size,
             TenantSorting::MaxLogicalSize => sizes.max_logical_size,
+            TenantSorting::MaxLogicalSizePerShard => sizes.max_logical_size_per_shard,
         }
     }
 
@@ -3254,7 +3337,7 @@ async fn put_tenant_timeline_import_basebackup(
 
         tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
 
-        let timeline = tenant
+        let (timeline, timeline_ctx) = tenant
             .create_empty_timeline(timeline_id, base_lsn, pg_version, &ctx)
             .map_err(ApiError::InternalServerError)
             .await?;
@@ -3273,7 +3356,13 @@ async fn put_tenant_timeline_import_basebackup(
         info!("importing basebackup");
 
         timeline
-            .import_basebackup_from_tar(tenant.clone(), &mut body, base_lsn, broker_client, &ctx)
+            .import_basebackup_from_tar(
+                tenant.clone(),
+                &mut body,
+                base_lsn,
+                broker_client,
+                &timeline_ctx,
+            )
             .await
             .map_err(ApiError::InternalServerError)?;
 
@@ -3313,6 +3402,7 @@ async fn put_tenant_timeline_import_wal(
         let state = get_state(&request);
 
         let timeline = active_timeline_of_active_tenant(&state.tenant_manager, TenantShardId::unsharded(tenant_id), timeline_id).await?;
+        let ctx = RequestContextBuilder::extend(&ctx).scope(context::Scope::new_timeline(&timeline)).build();
 
         let mut body = StreamReader::new(request.into_body().map(|res| {
             res.map_err(|error| {
@@ -3629,6 +3719,10 @@ pub fn make_router(
             "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/get_timestamp_of_lsn",
             |r| api_handler(r, get_timestamp_of_lsn_handler),
         )
+        .post(
+            "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/patch_index_part",
+            |r| api_handler(r, timeline_patch_index_part_handler),
+        )
         .post(
             "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/lsn_lease",
             |r| api_handler(r, lsn_lease_handler),

pageserver/src/lib.rs

@@ -64,6 +64,7 @@ pub struct CancellableTask {
     pub cancel: CancellationToken,
 }
 pub struct HttpEndpointListener(pub CancellableTask);
+pub struct HttpsEndpointListener(pub CancellableTask);
 pub struct ConsumptionMetricsTasks(pub CancellableTask);
 pub struct DiskUsageEvictionTask(pub CancellableTask);
 impl CancellableTask {
@@ -77,6 +78,7 @@ impl CancellableTask {
 #[allow(clippy::too_many_arguments)]
 pub async fn shutdown_pageserver(
     http_listener: HttpEndpointListener,
+    https_listener: Option<HttpsEndpointListener>,
     page_service: page_service::Listener,
     consumption_metrics_worker: ConsumptionMetricsTasks,
     disk_usage_eviction_task: Option<DiskUsageEvictionTask>,
@@ -213,6 +215,15 @@ pub async fn shutdown_pageserver(
     )
     .await;
 
+    if let Some(https_listener) = https_listener {
+        timed(
+            https_listener.0.shutdown(),
+            "shutdown https",
+            Duration::from_secs(1),
+        )
+        .await;
+    }
+
     // Shut down the HTTP endpoint last, so that you can still check the server's
     // status while it's shutting down.
     // FIXME: We should probably stop accepting commands like attach/detach earlier.

pageserver/src/metrics.rs

@@ -1227,11 +1227,24 @@ impl StorageIoTime {
 
 pub(crate) static STORAGE_IO_TIME_METRIC: Lazy<StorageIoTime> = Lazy::new(StorageIoTime::new);
 
-const STORAGE_IO_SIZE_OPERATIONS: &[&str] = &["read", "write"];
+#[derive(Clone, Copy)]
+#[repr(usize)]
+enum StorageIoSizeOperation {
+    Read,
+    Write,
+}
+
+impl StorageIoSizeOperation {
+    const VARIANTS: &'static [&'static str] = &["read", "write"];
+
+    fn as_str(&self) -> &'static str {
+        Self::VARIANTS[*self as usize]
+    }
+}
 
 // Needed for the https://neonprod.grafana.net/d/5uK9tHL4k/picking-tenant-for-relocation?orgId=1
-pub(crate) static STORAGE_IO_SIZE: Lazy<IntGaugeVec> = Lazy::new(|| {
-    register_int_gauge_vec!(
+static STORAGE_IO_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
+    register_uint_gauge_vec!(
         "pageserver_io_operations_bytes_total",
         "Total amount of bytes read/written in IO operations",
         &["operation", "tenant_id", "shard_id", "timeline_id"]
@@ -1239,6 +1252,34 @@ pub(crate) static STORAGE_IO_SIZE: Lazy<IntGaugeVec> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
+#[derive(Clone, Debug)]
+pub(crate) struct StorageIoSizeMetrics {
+    pub read: UIntGauge,
+    pub write: UIntGauge,
+}
+
+impl StorageIoSizeMetrics {
+    pub(crate) fn new(tenant_id: &str, shard_id: &str, timeline_id: &str) -> Self {
+        let read = STORAGE_IO_SIZE
+            .get_metric_with_label_values(&[
+                StorageIoSizeOperation::Read.as_str(),
+                tenant_id,
+                shard_id,
+                timeline_id,
+            ])
+            .unwrap();
+        let write = STORAGE_IO_SIZE
+            .get_metric_with_label_values(&[
+                StorageIoSizeOperation::Write.as_str(),
+                tenant_id,
+                shard_id,
+                timeline_id,
+            ])
+            .unwrap();
+        Self { read, write }
+    }
+}
+
 #[cfg(not(test))]
 pub(crate) mod virtual_file_descriptor_cache {
     use super::*;
@@ -2821,6 +2862,7 @@ pub(crate) struct TimelineMetrics {
     /// Number of valid LSN leases.
     pub valid_lsn_lease_count_gauge: UIntGauge,
     pub wal_records_received: IntCounter,
+    pub storage_io_size: StorageIoSizeMetrics,
     shutdown: std::sync::atomic::AtomicBool,
 }
 
@@ -2956,6 +2998,8 @@ impl TimelineMetrics {
             .get_metric_with_label_values(&[&tenant_id, &shard_id, &timeline_id])
             .unwrap();
 
+        let storage_io_size = StorageIoSizeMetrics::new(&tenant_id, &shard_id, &timeline_id);
+
         TimelineMetrics {
             tenant_id,
             shard_id,
@@ -2985,6 +3029,7 @@ impl TimelineMetrics {
             evictions_with_low_residence_duration: std::sync::RwLock::new(
                 evictions_with_low_residence_duration,
             ),
+            storage_io_size,
             valid_lsn_lease_count_gauge,
             wal_records_received,
             shutdown: std::sync::atomic::AtomicBool::default(),
@@ -3175,7 +3220,7 @@ impl TimelineMetrics {
             ]);
         }
 
-        for op in STORAGE_IO_SIZE_OPERATIONS {
+        for op in StorageIoSizeOperation::VARIANTS {
             let _ = STORAGE_IO_SIZE.remove_label_values(&[op, tenant_id, shard_id, timeline_id]);
         }
 

pageserver/src/page_service.rs

@@ -56,6 +56,7 @@ use crate::config::PageServerConf;
 use crate::context::{DownloadBehavior, RequestContext};
 use crate::metrics::{
     self, COMPUTE_COMMANDS_COUNTERS, ComputeCommandKind, LIVE_CONNECTIONS, SmgrOpTimer,
+    TimelineMetrics,
 };
 use crate::pgdatadir_mapping::Version;
 use crate::span::{
@@ -423,6 +424,9 @@ impl timeline::handle::Types for TenantManagerTypes {
 
 pub(crate) struct TenantManagerCacheItem {
     pub(crate) timeline: Arc<Timeline>,
+    // allow() for cheap propagation through RequestContext inside a task
+    #[allow(clippy::redundant_allocation)]
+    pub(crate) metrics: Arc<Arc<TimelineMetrics>>,
     #[allow(dead_code)] // we store it to keep the gate open
     pub(crate) gate_guard: GateGuard,
 }
@@ -506,8 +510,11 @@ impl timeline::handle::TenantManager<TenantManagerTypes> for TenantManagerWrappe
             }
         };
 
+        let metrics = Arc::new(Arc::clone(&timeline.metrics));
+
         Ok(TenantManagerCacheItem {
             timeline,
+            metrics,
             gate_guard,
         })
     }
@@ -1238,6 +1245,14 @@ impl PageServerHandler {
         ),
         QueryError,
     > {
+        macro_rules! upgrade_handle_and_set_context {
+            ($shard:ident) => {{
+                let weak_handle = &$shard;
+                let handle = weak_handle.upgrade()?;
+                let ctx = ctx.with_scope_page_service_pagestream(&handle);
+                (handle, ctx)
+            }};
+        }
         Ok(match batch {
             BatchedFeMessage::Exists {
                 span,
@@ -1246,9 +1261,10 @@ impl PageServerHandler {
                 req,
             } => {
                 fail::fail_point!("ps::handle-pagerequest-message::exists");
+                let (shard, ctx) = upgrade_handle_and_set_context!(shard);
                 (
                     vec![
-                        self.handle_get_rel_exists_request(&*shard.upgrade()?, &req, ctx)
+                        self.handle_get_rel_exists_request(&shard, &req, &ctx)
                             .instrument(span.clone())
                             .await
                             .map(|msg| (msg, timer))
@@ -1264,9 +1280,10 @@ impl PageServerHandler {
                 req,
             } => {
                 fail::fail_point!("ps::handle-pagerequest-message::nblocks");
+                let (shard, ctx) = upgrade_handle_and_set_context!(shard);
                 (
                     vec![
-                        self.handle_get_nblocks_request(&*shard.upgrade()?, &req, ctx)
+                        self.handle_get_nblocks_request(&shard, &req, &ctx)
                             .instrument(span.clone())
                             .await
                             .map(|msg| (msg, timer))
@@ -1282,17 +1299,18 @@ impl PageServerHandler {
                 pages,
             } => {
                 fail::fail_point!("ps::handle-pagerequest-message::getpage");
+                let (shard, ctx) = upgrade_handle_and_set_context!(shard);
                 (
                     {
                         let npages = pages.len();
                         trace!(npages, "handling getpage request");
                         let res = self
                             .handle_get_page_at_lsn_request_batched(
-                                &*shard.upgrade()?,
+                                &shard,
                                 effective_request_lsn,
                                 pages,
                                 io_concurrency,
-                                ctx,
+                                &ctx,
                             )
                             .instrument(span.clone())
                             .await;
@@ -1309,9 +1327,10 @@ impl PageServerHandler {
                 req,
             } => {
                 fail::fail_point!("ps::handle-pagerequest-message::dbsize");
+                let (shard, ctx) = upgrade_handle_and_set_context!(shard);
                 (
                     vec![
-                        self.handle_db_size_request(&*shard.upgrade()?, &req, ctx)
+                        self.handle_db_size_request(&shard, &req, &ctx)
                             .instrument(span.clone())
                             .await
                             .map(|msg| (msg, timer))
@@ -1327,9 +1346,10 @@ impl PageServerHandler {
                 req,
             } => {
                 fail::fail_point!("ps::handle-pagerequest-message::slrusegment");
+                let (shard, ctx) = upgrade_handle_and_set_context!(shard);
                 (
                     vec![
-                        self.handle_get_slru_segment_request(&*shard.upgrade()?, &req, ctx)
+                        self.handle_get_slru_segment_request(&shard, &req, &ctx)
                             .instrument(span.clone())
                             .await
                             .map(|msg| (msg, timer))
@@ -1345,12 +1365,13 @@ impl PageServerHandler {
                 requests,
             } => {
                 fail::fail_point!("ps::handle-pagerequest-message::test");
+                let (shard, ctx) = upgrade_handle_and_set_context!(shard);
                 (
                     {
                         let npages = requests.len();
                         trace!(npages, "handling getpage request");
                         let res = self
-                            .handle_test_request_batch(&*shard.upgrade()?, requests, ctx)
+                            .handle_test_request_batch(&shard, requests, &ctx)
                             .instrument(span.clone())
                             .await;
                         assert_eq!(res.len(), npages);
@@ -2126,6 +2147,7 @@ impl PageServerHandler {
             .get(tenant_id, timeline_id, ShardSelector::Zero)
             .await?;
         set_tracing_field_shard_id(&timeline);
+        let ctx = ctx.with_scope_timeline(&timeline);
 
         if timeline.is_archived() == Some(true) {
             tracing::info!(
@@ -2143,7 +2165,7 @@ impl PageServerHandler {
                     lsn,
                     crate::tenant::timeline::WaitLsnWaiter::PageService,
                     crate::tenant::timeline::WaitLsnTimeout::Default,
-                    ctx,
+                    &ctx,
                 )
                 .await?;
             timeline
@@ -2169,7 +2191,7 @@ impl PageServerHandler {
                 prev_lsn,
                 full_backup,
                 replica,
-                ctx,
+                &ctx,
             )
             .await
             .map_err(map_basebackup_error)?;
@@ -2192,7 +2214,7 @@ impl PageServerHandler {
                     prev_lsn,
                     full_backup,
                     replica,
-                    ctx,
+                    &ctx,
                 )
                 .await
                 .map_err(map_basebackup_error)?;
@@ -2209,7 +2231,7 @@ impl PageServerHandler {
                     prev_lsn,
                     full_backup,
                     replica,
-                    ctx,
+                    &ctx,
                 )
                 .await
                 .map_err(map_basebackup_error)?;

pageserver/src/pgdatadir_mapping.rs

@@ -2758,7 +2758,7 @@ mod tests {
             TimelineId::from_array(hex!("11223344556677881122334455667788"));
 
         let (tenant, ctx) = harness.load().await;
-        let tline = tenant
+        let (tline, ctx) = tenant
             .create_empty_timeline(TIMELINE_ID, Lsn(0x10), DEFAULT_PG_VERSION, &ctx)
             .await?;
         let tline = tline.raw_timeline().unwrap();

pageserver/src/tenant.rs

@@ -77,6 +77,8 @@ use self::timeline::{
     EvictionTaskTenantState, GcCutoffs, TimelineDeleteProgress, TimelineResources, WaitLsnError,
 };
 use crate::config::PageServerConf;
+use crate::context;
+use crate::context::RequestContextBuilder;
 use crate::context::{DownloadBehavior, RequestContext};
 use crate::deletion_queue::{DeletionQueueClient, DeletionQueueError};
 use crate::l0_flush::L0FlushGlobalState;
@@ -1114,7 +1116,7 @@ impl Tenant {
             }
         };
 
-        let timeline = self.create_timeline_struct(
+        let (timeline, timeline_ctx) = self.create_timeline_struct(
             timeline_id,
             &metadata,
             previous_heatmap,
@@ -1124,6 +1126,7 @@ impl Tenant {
             idempotency.clone(),
             index_part.gc_compaction.clone(),
             index_part.rel_size_migration.clone(),
+            ctx,
         )?;
         let disk_consistent_lsn = timeline.get_disk_consistent_lsn();
         anyhow::ensure!(
@@ -1257,7 +1260,7 @@ impl Tenant {
                         match activate {
                             ActivateTimelineArgs::Yes { broker_client } => {
                                 info!("activating timeline after reload from pgdata import task");
-                                timeline.activate(self.clone(), broker_client, None, ctx);
+                                timeline.activate(self.clone(), broker_client, None, &timeline_ctx);
                             }
                             ActivateTimelineArgs::No => (),
                         }
@@ -1765,6 +1768,7 @@ impl Tenant {
                         import_pgdata,
                         ActivateTimelineArgs::No,
                         guard,
+                        ctx.detached_child(TaskKind::ImportPgdata, DownloadBehavior::Warn),
                     ));
                 }
             }
@@ -1782,6 +1786,7 @@ impl Tenant {
                 timeline_id,
                 &index_part.metadata,
                 remote_timeline_client,
+                ctx,
             )
             .instrument(tracing::info_span!("timeline_delete", %timeline_id))
             .await
@@ -2219,7 +2224,7 @@ impl Tenant {
                 self.clone(),
                 broker_client.clone(),
                 background_jobs_can_start,
-                &ctx,
+                &ctx.with_scope_timeline(&timeline),
             );
         }
 
@@ -2416,8 +2421,8 @@ impl Tenant {
         new_timeline_id: TimelineId,
         initdb_lsn: Lsn,
         pg_version: u32,
-        _ctx: &RequestContext,
-    ) -> anyhow::Result<UninitializedTimeline> {
+        ctx: &RequestContext,
+    ) -> anyhow::Result<(UninitializedTimeline, RequestContext)> {
         anyhow::ensure!(
             self.is_active(),
             "Cannot create empty timelines on inactive tenant"
@@ -2452,6 +2457,7 @@ impl Tenant {
             initdb_lsn,
             None,
             None,
+            ctx,
         )
         .await
     }
@@ -2469,7 +2475,7 @@ impl Tenant {
         pg_version: u32,
         ctx: &RequestContext,
     ) -> anyhow::Result<Arc<Timeline>> {
-        let uninit_tl = self
+        let (uninit_tl, ctx) = self
             .create_empty_timeline(new_timeline_id, initdb_lsn, pg_version, ctx)
             .await?;
         let tline = uninit_tl.raw_timeline().expect("we just created it");
@@ -2481,7 +2487,7 @@ impl Tenant {
             .init_empty_test_timeline()
             .context("init_empty_test_timeline")?;
         modification
-            .commit(ctx)
+            .commit(&ctx)
             .await
             .context("commit init_empty_test_timeline modification")?;
 
@@ -2699,7 +2705,12 @@ impl Tenant {
         // doing stuff before the IndexPart is durable in S3, which is done by the previous section.
         let activated_timeline = match result {
             CreateTimelineResult::Created(timeline) => {
-                timeline.activate(self.clone(), broker_client, None, ctx);
+                timeline.activate(
+                    self.clone(),
+                    broker_client,
+                    None,
+                    &ctx.with_scope_timeline(&timeline),
+                );
                 timeline
             }
             CreateTimelineResult::Idempotent(timeline) => {
@@ -2761,10 +2772,9 @@ impl Tenant {
             }
         };
 
-        let mut uninit_timeline = {
+        let (mut uninit_timeline, timeline_ctx) = {
             let this = &self;
             let initdb_lsn = Lsn(0);
-            let _ctx = ctx;
             async move {
                 let new_metadata = TimelineMetadata::new(
                     // Initialize disk_consistent LSN to 0, The caller must import some data to
@@ -2784,6 +2794,7 @@ impl Tenant {
                     initdb_lsn,
                     None,
                     None,
+                    ctx,
                 )
                 .await
             }
@@ -2813,6 +2824,7 @@ impl Tenant {
             index_part,
             activate,
             timeline_create_guard,
+            timeline_ctx.detached_child(TaskKind::ImportPgdata, DownloadBehavior::Warn),
         ));
 
         // NB: the timeline doesn't exist in self.timelines at this point
@@ -2826,6 +2838,7 @@ impl Tenant {
         index_part: import_pgdata::index_part_format::Root,
         activate: ActivateTimelineArgs,
         timeline_create_guard: TimelineCreateGuard,
+        ctx: RequestContext,
     ) {
         debug_assert_current_span_has_tenant_and_timeline_id();
         info!("starting");
@@ -2837,6 +2850,7 @@ impl Tenant {
                 index_part,
                 activate,
                 timeline_create_guard,
+                ctx,
             )
             .await;
         if let Err(err) = &res {
@@ -2852,9 +2866,8 @@ impl Tenant {
         index_part: import_pgdata::index_part_format::Root,
         activate: ActivateTimelineArgs,
         timeline_create_guard: TimelineCreateGuard,
+        ctx: RequestContext,
     ) -> Result<(), anyhow::Error> {
-        let ctx = RequestContext::new(TaskKind::ImportPgdata, DownloadBehavior::Warn);
-
         info!("importing pgdata");
         import_pgdata::doit(&timeline, index_part, &ctx, self.cancel.clone())
             .await
@@ -3063,6 +3076,7 @@ impl Tenant {
 
             let mut has_pending_l0 = false;
             for timeline in compact_l0 {
+                let ctx = &ctx.with_scope_timeline(&timeline);
                 let outcome = timeline
                     .compact(cancel, CompactFlags::OnlyL0Compaction.into(), ctx)
                     .instrument(info_span!("compact_timeline", timeline_id = %timeline.timeline_id))
@@ -3096,6 +3110,7 @@ impl Tenant {
             if !timeline.is_active() {
                 continue;
             }
+            let ctx = &ctx.with_scope_timeline(&timeline);
 
             let mut outcome = timeline
                 .compact(cancel, EnumSet::default(), ctx)
@@ -3321,7 +3336,7 @@ impl Tenant {
                     self.clone(),
                     broker_client.clone(),
                     background_jobs_can_start,
-                    ctx,
+                    &ctx.with_scope_timeline(timeline),
                 );
                 activated_timelines += 1;
             }
@@ -3827,6 +3842,7 @@ impl Tenant {
             resident_size: 0,
             physical_size: 0,
             max_logical_size: 0,
+            max_logical_size_per_shard: 0,
         };
 
         for timeline in self.timelines.lock().unwrap().values() {
@@ -3843,6 +3859,10 @@ impl Tenant {
             );
         }
 
+        result.max_logical_size_per_shard = result
+            .max_logical_size
+            .div_ceil(self.tenant_shard_id.shard_count.count() as u64);
+
         result
     }
 }
@@ -4136,7 +4156,8 @@ impl Tenant {
         create_idempotency: CreateTimelineIdempotency,
         gc_compaction_state: Option<GcCompactionState>,
         rel_size_v2_status: Option<RelSizeMigration>,
-    ) -> anyhow::Result<Arc<Timeline>> {
+        ctx: &RequestContext,
+    ) -> anyhow::Result<(Arc<Timeline>, RequestContext)> {
         let state = match cause {
             CreateTimelineCause::Load => {
                 let ancestor_id = new_metadata.ancestor_timeline();
@@ -4172,7 +4193,11 @@ impl Tenant {
             self.cancel.child_token(),
         );
 
-        Ok(timeline)
+        let timeline_ctx = RequestContextBuilder::extend(ctx)
+            .scope(context::Scope::new_timeline(&timeline))
+            .build();
+
+        Ok((timeline, timeline_ctx))
     }
 
     /// [`Tenant::shutdown`] must be called before dropping the returned [`Tenant`] object
@@ -4588,6 +4613,7 @@ impl Tenant {
         // Ensures all timelines use the same start time when computing the time cutoff.
         let now_ts_for_pitr_calc = SystemTime::now();
         for timeline in timelines.iter() {
+            let ctx = &ctx.with_scope_timeline(timeline);
             let cutoff = timeline
                 .get_last_record_lsn()
                 .checked_sub(horizon)
@@ -4761,7 +4787,7 @@ impl Tenant {
         src_timeline: &Arc<Timeline>,
         dst_id: TimelineId,
         start_lsn: Option<Lsn>,
-        _ctx: &RequestContext,
+        ctx: &RequestContext,
     ) -> Result<CreateTimelineResult, CreateTimelineError> {
         let src_id = src_timeline.timeline_id;
 
@@ -4864,7 +4890,7 @@ impl Tenant {
             src_timeline.pg_version,
         );
 
-        let uninitialized_timeline = self
+        let (uninitialized_timeline, _timeline_ctx) = self
             .prepare_new_timeline(
                 dst_id,
                 &metadata,
@@ -4872,6 +4898,7 @@ impl Tenant {
                 start_lsn + 1,
                 Some(Arc::clone(src_timeline)),
                 Some(src_timeline.get_rel_size_v2_status()),
+                ctx,
             )
             .await?;
 
@@ -5138,7 +5165,7 @@ impl Tenant {
             pgdata_lsn,
             pg_version,
         );
-        let mut raw_timeline = self
+        let (mut raw_timeline, timeline_ctx) = self
             .prepare_new_timeline(
                 timeline_id,
                 &new_metadata,
@@ -5146,6 +5173,7 @@ impl Tenant {
                 pgdata_lsn,
                 None,
                 None,
+                ctx,
             )
             .await?;
 
@@ -5156,7 +5184,7 @@ impl Tenant {
                     &unfinished_timeline,
                     &pgdata_path,
                     pgdata_lsn,
-                    ctx,
+                    &timeline_ctx,
                 )
                 .await
                 .with_context(|| {
@@ -5217,6 +5245,7 @@ impl Tenant {
     /// An empty layer map is initialized, and new data and WAL can be imported starting
     /// at 'disk_consistent_lsn'. After any initial data has been imported, call
     /// `finish_creation` to insert the Timeline into the timelines map.
+    #[allow(clippy::too_many_arguments)]
     async fn prepare_new_timeline<'a>(
         &'a self,
         new_timeline_id: TimelineId,
@@ -5225,7 +5254,8 @@ impl Tenant {
         start_lsn: Lsn,
         ancestor: Option<Arc<Timeline>>,
         rel_size_v2_status: Option<RelSizeMigration>,
-    ) -> anyhow::Result<UninitializedTimeline<'a>> {
+        ctx: &RequestContext,
+    ) -> anyhow::Result<(UninitializedTimeline<'a>, RequestContext)> {
         let tenant_shard_id = self.tenant_shard_id;
 
         let resources = self.build_timeline_resources(new_timeline_id);
@@ -5233,7 +5263,7 @@ impl Tenant {
             .remote_client
             .init_upload_queue_for_empty_remote(new_metadata, rel_size_v2_status.clone())?;
 
-        let timeline_struct = self
+        let (timeline_struct, timeline_ctx) = self
             .create_timeline_struct(
                 new_timeline_id,
                 new_metadata,
@@ -5244,6 +5274,7 @@ impl Tenant {
                 create_guard.idempotency.clone(),
                 None,
                 rel_size_v2_status,
+                ctx,
             )
             .context("Failed to create timeline data structure")?;
 
@@ -5264,10 +5295,13 @@ impl Tenant {
             "Successfully created initial files for timeline {tenant_shard_id}/{new_timeline_id}"
         );
 
-        Ok(UninitializedTimeline::new(
-            self,
-            new_timeline_id,
-            Some((timeline_struct, create_guard)),
+        Ok((
+            UninitializedTimeline::new(
+                self,
+                new_timeline_id,
+                Some((timeline_struct, create_guard)),
+            ),
+            timeline_ctx,
         ))
     }
 
@@ -5720,7 +5754,7 @@ pub(crate) mod harness {
                 logging::TracingErrorLayerEnablement::EnableWithRustLogFilter,
                 logging::Output::Stdout,
             )
-            .expect("Failed to init test logging")
+            .expect("Failed to init test logging");
         });
     }
 
@@ -5802,7 +5836,8 @@ pub(crate) mod harness {
         }
 
         pub(crate) async fn load(&self) -> (Arc<Tenant>, RequestContext) {
-            let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+            let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error)
+                .with_scope_unit_test();
             (
                 self.do_try_load(&ctx)
                     .await
@@ -6825,7 +6860,7 @@ mod tests {
 
         let (tenant, ctx) = harness.load().await;
         let io_concurrency = IoConcurrency::spawn_for_test();
-        let tline = tenant
+        let (tline, ctx) = tenant
             .create_empty_timeline(TIMELINE_ID, Lsn(0), DEFAULT_PG_VERSION, &ctx)
             .await?;
         let tline = tline.raw_timeline().unwrap();
@@ -7447,7 +7482,7 @@ mod tests {
             .await;
 
         let initdb_lsn = Lsn(0x20);
-        let utline = tenant
+        let (utline, ctx) = tenant
             .create_empty_timeline(TIMELINE_ID, initdb_lsn, DEFAULT_PG_VERSION, &ctx)
             .await?;
         let tline = utline.raw_timeline().unwrap();
@@ -7514,7 +7549,7 @@ mod tests {
         let harness = TenantHarness::create(name).await?;
         {
             let (tenant, ctx) = harness.load().await;
-            let tline = tenant
+            let (tline, _ctx) = tenant
                 .create_empty_timeline(TIMELINE_ID, Lsn(0), DEFAULT_PG_VERSION, &ctx)
                 .await?;
             // Leave the timeline ID in [`Tenant::timelines_creating`] to exclude attempting to create it again

pageserver/src/tenant/blob_io.rs

@@ -471,7 +471,8 @@ pub(crate) mod tests {
         blobs: &[Vec<u8>],
         compression: bool,
     ) -> Result<(), Error> {
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
         let (_temp_dir, pathbuf, offsets) =
             write_maybe_compressed::<BUFFERED>(blobs, compression, &ctx).await?;
 

pageserver/src/tenant/config.rs

@@ -219,7 +219,11 @@ impl LocationConf {
         };
 
         let shard = if conf.shard_count == 0 {
-            ShardIdentity::unsharded()
+            // NB: carry over the persisted stripe size instead of using the default. This doesn't
+            // matter for most practical purposes, since unsharded tenants don't use the stripe
+            // size, but can cause inconsistencies between storcon and Pageserver and cause manual
+            // splits without `new_stripe_size` to use an unintended stripe size.
+            ShardIdentity::unsharded_with_stripe_size(ShardStripeSize(conf.shard_stripe_size))
         } else {
             ShardIdentity::new(
                 ShardNumber(conf.shard_number),

pageserver/src/tenant/disk_btree.rs

@@ -32,8 +32,7 @@ use hex;
 use thiserror::Error;
 use tracing::error;
 
-use crate::context::{DownloadBehavior, RequestContext};
-use crate::task_mgr::TaskKind;
+use crate::context::RequestContext;
 use crate::tenant::block_io::{BlockReader, BlockWriter};
 
 // The maximum size of a value stored in the B-tree. 5 bytes is enough currently.
@@ -477,16 +476,15 @@ where
     }
 
     #[allow(dead_code)]
-    pub async fn dump(&self) -> Result<()> {
+    pub async fn dump(&self, ctx: &RequestContext) -> Result<()> {
         let mut stack = Vec::new();
-        let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);
 
         stack.push((self.root_blk, String::new(), 0, 0, 0));
 
         let block_cursor = self.reader.block_cursor();
 
         while let Some((blknum, path, depth, child_idx, key_off)) = stack.pop() {
-            let blk = block_cursor.read_blk(self.start_blk + blknum, &ctx).await?;
+            let blk = block_cursor.read_blk(self.start_blk + blknum, ctx).await?;
             let buf: &[u8] = blk.as_ref();
             let node = OnDiskNode::<L>::deparse(buf)?;
 
@@ -835,6 +833,8 @@ pub(crate) mod tests {
     use rand::Rng;
 
     use super::*;
+    use crate::context::DownloadBehavior;
+    use crate::task_mgr::TaskKind;
     use crate::tenant::block_io::{BlockCursor, BlockLease, BlockReaderRef};
 
     #[derive(Clone, Default)]
@@ -869,7 +869,8 @@ pub(crate) mod tests {
         let mut disk = TestDisk::new();
         let mut writer = DiskBtreeBuilder::<_, 6>::new(&mut disk);
 
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
 
         let all_keys: Vec<&[u8; 6]> = vec![
             b"xaaaaa", b"xaaaba", b"xaaaca", b"xabaaa", b"xababa", b"xabaca", b"xabada", b"xabadb",
@@ -887,7 +888,7 @@ pub(crate) mod tests {
 
         let reader = DiskBtreeReader::new(0, root_offset, disk);
 
-        reader.dump().await?;
+        reader.dump(&ctx).await?;
 
         // Test the `get` function on all the keys.
         for (key, val) in all_data.iter() {
@@ -979,7 +980,8 @@ pub(crate) mod tests {
     async fn lots_of_keys() -> Result<()> {
         let mut disk = TestDisk::new();
         let mut writer = DiskBtreeBuilder::<_, 8>::new(&mut disk);
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
 
         const NUM_KEYS: u64 = 1000;
 
@@ -997,7 +999,7 @@ pub(crate) mod tests {
 
         let reader = DiskBtreeReader::new(0, root_offset, disk);
 
-        reader.dump().await?;
+        reader.dump(&ctx).await?;
 
         use std::sync::Mutex;
 
@@ -1167,7 +1169,8 @@ pub(crate) mod tests {
         // Build a tree from it
         let mut disk = TestDisk::new();
         let mut writer = DiskBtreeBuilder::<_, 26>::new(&mut disk);
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
 
         for (key, val) in disk_btree_test_data::TEST_DATA {
             writer.append(&key, val)?;
@@ -1198,7 +1201,7 @@ pub(crate) mod tests {
             .await?;
         assert_eq!(count, disk_btree_test_data::TEST_DATA.len());
 
-        reader.dump().await?;
+        reader.dump(&ctx).await?;
 
         Ok(())
     }

pageserver/src/tenant/ephemeral_file.rs

@@ -9,7 +9,7 @@ use camino::Utf8PathBuf;
 use num_traits::Num;
 use pageserver_api::shard::TenantShardId;
 use tokio_epoll_uring::{BoundedBuf, Slice};
-use tracing::error;
+use tracing::{error, info_span};
 use utils::id::TimelineId;
 
 use crate::assert_u64_eq_usize::{U64IsUsize, UsizeIsU64};
@@ -76,6 +76,7 @@ impl EphemeralFile {
                 || IoBufferMut::with_capacity(TAIL_SZ),
                 gate.enter()?,
                 ctx,
+                info_span!(parent: None, "ephemeral_file_buffered_writer", tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), timeline_id=%timeline_id, path = %filename),
             ),
             _gate_guard: gate.enter()?,
         })
@@ -351,7 +352,8 @@ mod tests {
         let timeline_id = TimelineId::from_str("22000000000000000000000000000000").unwrap();
         fs::create_dir_all(conf.timeline_path(&tenant_shard_id, &timeline_id))?;
 
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
 
         Ok((conf, tenant_shard_id, timeline_id, ctx))
     }

pageserver/src/tenant/metadata.rs

@@ -300,9 +300,8 @@ impl TimelineMetadata {
 
     /// Returns true if anything was changed
     pub fn detach_from_ancestor(&mut self, branchpoint: &(TimelineId, Lsn)) {
-        if let Some(ancestor) = self.body.ancestor_timeline {
-            assert_eq!(ancestor, branchpoint.0);
-        }
+        // Detaching from ancestor now doesn't always detach directly to the direct ancestor, but we
+        // ensure the LSN is the same. So we don't check the timeline ID.
         if self.body.ancestor_lsn != Lsn(0) {
             assert_eq!(self.body.ancestor_lsn, branchpoint.1);
         }

pageserver/src/tenant/mgr.rs

@@ -1914,6 +1914,7 @@ impl TenantManager {
         tenant_shard_id: TenantShardId,
         timeline_id: TimelineId,
         prepared: PreparedTimelineDetach,
+        behavior: detach_ancestor::DetachBehavior,
         mut attempt: detach_ancestor::Attempt,
         ctx: &RequestContext,
     ) -> Result<HashSet<TimelineId>, detach_ancestor::Error> {
@@ -1957,7 +1958,14 @@ impl TenantManager {
             .map_err(Error::NotFound)?;
 
         let resp = timeline
-            .detach_from_ancestor_and_reparent(&tenant, prepared, ctx)
+            .detach_from_ancestor_and_reparent(
+                &tenant,
+                prepared,
+                attempt.ancestor_timeline_id,
+                attempt.ancestor_lsn,
+                behavior,
+                ctx,
+            )
             .await?;
 
         let mut slot_guard = slot_guard;

pageserver/src/tenant/remote_timeline_client.rs

@@ -954,6 +954,14 @@ impl RemoteTimelineClient {
         Ok(())
     }
 
+    /// Only used in the `patch_index_part` HTTP API to force trigger an index upload.
+    pub fn force_schedule_index_upload(self: &Arc<Self>) -> Result<(), NotInitialized> {
+        let mut guard = self.upload_queue.lock().unwrap();
+        let upload_queue = guard.initialized_mut()?;
+        self.schedule_index_upload(upload_queue);
+        Ok(())
+    }
+
     /// Launch an index-file upload operation in the background (internal function)
     fn schedule_index_upload(self: &Arc<Self>, upload_queue: &mut UploadQueueInitialized) {
         let disk_consistent_lsn = upload_queue.dirty.metadata.disk_consistent_lsn();

pageserver/src/tenant/remote_timeline_client/download.rs

@@ -229,6 +229,7 @@ async fn download_object(
                     || IoBufferMut::with_capacity(super::BUFFER_SIZE),
                     gate.enter().map_err(|_| DownloadError::Cancelled)?,
                     ctx,
+                    tracing::info_span!(parent: None, "download_object_buffered_writer", %dst_path),
                 );
 
                 // TODO: use vectored write (writev) once supported by tokio-epoll-uring.

pageserver/src/tenant/secondary/downloader.rs

@@ -491,7 +491,10 @@ impl JobGenerator<PendingDownload, RunningDownload, CompleteDownload, DownloadCo
         let remote_storage = self.remote_storage.clone();
         let conf = self.tenant_manager.get_conf();
         let tenant_shard_id = *secondary_state.get_tenant_shard_id();
-        let download_ctx = self.root_ctx.attached_child();
+        let download_ctx = self
+            .root_ctx
+            .attached_child()
+            .with_scope_secondary_tenant(&tenant_shard_id);
         (RunningDownload { barrier }, Box::pin(async move {
             let _completion = completion;
 
@@ -771,6 +774,7 @@ impl<'a> TenantDownloader<'a> {
 
         // Download the layers in the heatmap
         for timeline in heatmap.timelines {
+            let ctx = &ctx.with_scope_secondary_timeline(tenant_shard_id, &timeline.timeline_id);
             let timeline_state = timeline_states
                 .remove(&timeline.timeline_id)
                 .expect("Just populated above");

pageserver/src/tenant/size.rs

@@ -474,7 +474,7 @@ async fn fill_logical_sizes(
             if cached_size.is_none() {
                 let timeline = Arc::clone(timeline_hash.get(&timeline_id).unwrap());
                 let parallel_size_calcs = Arc::clone(limit);
-                let ctx = ctx.attached_child();
+                let ctx = ctx.attached_child().with_scope_timeline(&timeline);
                 joinset.spawn(
                     calculate_logical_size(parallel_size_calcs, timeline, lsn, cause, ctx)
                         .in_current_span(),

pageserver/src/tenant/storage_layer/delta_layer.rs

@@ -1334,7 +1334,7 @@ impl DeltaLayerInner {
             block_reader,
         );
 
-        tree_reader.dump().await?;
+        tree_reader.dump(ctx).await?;
 
         let keys = self.index_entries(ctx).await?;
 
@@ -1972,6 +1972,7 @@ pub(crate) mod test {
             .create_test_timeline(TimelineId::generate(), Lsn(0x10), 14, ctx)
             .await
             .unwrap();
+        let ctx = &ctx.with_scope_timeline(&timeline);
 
         let initdb_layer = timeline
             .layers

pageserver/src/tenant/storage_layer/image_layer.rs

@@ -199,7 +199,7 @@ impl ImageLayerInner {
             block_reader,
         );
 
-        tree_reader.dump().await?;
+        tree_reader.dump(ctx).await?;
 
         tree_reader
             .visit(

pageserver/src/tenant/storage_layer/layer/tests.rs

@@ -8,7 +8,6 @@ use utils::id::TimelineId;
 use super::failpoints::{Failpoint, FailpointKind};
 use super::*;
 use crate::context::DownloadBehavior;
-use crate::task_mgr::TaskKind;
 use crate::tenant::harness::{TenantHarness, test_img};
 use crate::tenant::storage_layer::{IoConcurrency, LayerVisibilityHint};
 
@@ -27,11 +26,9 @@ async fn smoke_test() {
     let h = TenantHarness::create("smoke_test").await.unwrap();
     let span = h.span();
     let download_span = span.in_scope(|| tracing::info_span!("downloading", timeline_id = 1));
-    let (tenant, _) = h.load().await;
+    let (tenant, ctx) = h.load().await;
     let io_concurrency = IoConcurrency::spawn_for_test();
 
-    let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Download);
-
     let image_layers = vec![(
         Lsn(0x40),
         vec![(
@@ -56,6 +53,7 @@ async fn smoke_test() {
         )
         .await
         .unwrap();
+    let ctx = &ctx.with_scope_timeline(&timeline);
 
     // Grab one of the timeline's layers to exercise in the test, and the other layer that is just
     // there to avoid the timeline being illegally empty
@@ -94,7 +92,7 @@ async fn smoke_test() {
                 controlfile_keyspace.clone(),
                 Lsn(0x10)..Lsn(0x11),
                 &mut data,
-                &ctx,
+                ctx,
             )
             .await
             .unwrap();
@@ -129,7 +127,7 @@ async fn smoke_test() {
                 controlfile_keyspace.clone(),
                 Lsn(0x10)..Lsn(0x11),
                 &mut data,
-                &ctx,
+                ctx,
             )
             .instrument(download_span.clone())
             .await
@@ -179,7 +177,7 @@ async fn smoke_test() {
 
     // plain downloading is rarely needed
     layer
-        .download_and_keep_resident(&ctx)
+        .download_and_keep_resident(ctx)
         .instrument(download_span)
         .await
         .unwrap();
@@ -341,6 +339,7 @@ fn read_wins_pending_eviction() {
             .create_test_timeline(TimelineId::generate(), Lsn(0x10), 14, &ctx)
             .await
             .unwrap();
+        let ctx = ctx.with_scope_timeline(&timeline);
 
         let layer = {
             let mut layers = {
@@ -473,6 +472,7 @@ fn multiple_pending_evictions_scenario(name: &'static str, in_order: bool) {
             .create_test_timeline(TimelineId::generate(), Lsn(0x10), 14, &ctx)
             .await
             .unwrap();
+        let ctx = ctx.with_scope_timeline(&timeline);
 
         let layer = {
             let mut layers = {
@@ -642,12 +642,12 @@ async fn cancelled_get_or_maybe_download_does_not_cancel_eviction() {
         .create_test_timeline(TimelineId::generate(), Lsn(0x10), 14, &ctx)
         .await
         .unwrap();
+    let ctx = ctx.with_scope_timeline(&timeline);
 
     // This test does downloads
     let ctx = RequestContextBuilder::extend(&ctx)
         .download_behavior(DownloadBehavior::Download)
         .build();
-
     let layer = {
         let mut layers = {
             let layers = timeline.layers.read().await;
@@ -727,6 +727,7 @@ async fn evict_and_wait_does_not_wait_for_download() {
         .create_test_timeline(TimelineId::generate(), Lsn(0x10), 14, &ctx)
         .await
         .unwrap();
+    let ctx = ctx.with_scope_timeline(&timeline);
 
     // This test does downloads
     let ctx = RequestContextBuilder::extend(&ctx)

pageserver/src/tenant/timeline.rs

@@ -287,7 +287,7 @@ pub struct Timeline {
     // The LSN of gc-compaction that was last applied to this timeline.
     gc_compaction_state: ArcSwap<Option<GcCompactionState>>,
 
-    pub(super) metrics: TimelineMetrics,
+    pub(crate) metrics: Arc<TimelineMetrics>,
 
     // `Timeline` doesn't write these metrics itself, but it manages the lifetime.  Code
     // in `crate::page_service` writes these metrics.
@@ -2685,14 +2685,14 @@ impl Timeline {
         }
 
         Arc::new_cyclic(|myself| {
-            let metrics = TimelineMetrics::new(
+            let metrics = Arc::new(TimelineMetrics::new(
                 &tenant_shard_id,
                 &timeline_id,
                 crate::metrics::EvictionsWithLowResidenceDurationBuilder::new(
                     "mtime",
                     evictions_low_residence_duration_metric_threshold,
                 ),
-            );
+            ));
             let aux_file_metrics = metrics.aux_file_size_gauge.clone();
 
             let mut result = Timeline {
@@ -2876,7 +2876,7 @@ impl Timeline {
             "layer flush task",
             async move {
                 let _guard = guard;
-                let background_ctx = RequestContext::todo_child(TaskKind::LayerFlushTask, DownloadBehavior::Error);
+                let background_ctx = RequestContext::todo_child(TaskKind::LayerFlushTask, DownloadBehavior::Error).with_scope_timeline(&self_clone);
                 self_clone.flush_loop(layer_flush_start_rx, &background_ctx).await;
                 let mut flush_loop_state = self_clone.flush_loop_state.lock().unwrap();
                 assert!(matches!(*flush_loop_state, FlushLoopState::Running{..}));
@@ -5388,9 +5388,10 @@ impl Timeline {
         self: &Arc<Timeline>,
         tenant: &crate::tenant::Tenant,
         options: detach_ancestor::Options,
+        behavior: detach_ancestor::DetachBehavior,
         ctx: &RequestContext,
     ) -> Result<detach_ancestor::Progress, detach_ancestor::Error> {
-        detach_ancestor::prepare(self, tenant, options, ctx).await
+        detach_ancestor::prepare(self, tenant, behavior, options, ctx).await
     }
 
     /// Second step of detach from ancestor; detaches the `self` from it's current ancestor and
@@ -5406,9 +5407,21 @@ impl Timeline {
         self: &Arc<Timeline>,
         tenant: &crate::tenant::Tenant,
         prepared: detach_ancestor::PreparedTimelineDetach,
+        ancestor_timeline_id: TimelineId,
+        ancestor_lsn: Lsn,
+        behavior: detach_ancestor::DetachBehavior,
         ctx: &RequestContext,
     ) -> Result<detach_ancestor::DetachingAndReparenting, detach_ancestor::Error> {
-        detach_ancestor::detach_and_reparent(self, tenant, prepared, ctx).await
+        detach_ancestor::detach_and_reparent(
+            self,
+            tenant,
+            prepared,
+            ancestor_timeline_id,
+            ancestor_lsn,
+            behavior,
+            ctx,
+        )
+        .await
     }
 
     /// Final step which unblocks the GC.
@@ -7127,6 +7140,7 @@ mod tests {
             )
             .await
             .unwrap();
+        let ctx = &ctx.with_scope_timeline(&timeline);
 
         // Layer visibility is an input to heatmap generation, so refresh it first
         timeline.update_layer_visibility().await.unwrap();
@@ -7192,7 +7206,7 @@ mod tests {
 
             eprintln!("Downloading {layer} and re-generating heatmap");
 
-            let ctx = &RequestContextBuilder::extend(&ctx)
+            let ctx = &RequestContextBuilder::extend(ctx)
                 .download_behavior(crate::context::DownloadBehavior::Download)
                 .build();
 

pageserver/src/tenant/timeline/compaction.rs

@@ -393,6 +393,9 @@ impl GcCompactionQueue {
                 if job.dry_run {
                     flags |= CompactFlags::DryRun;
                 }
+                if options.flags.contains(CompactFlags::NoYield) {
+                    flags |= CompactFlags::NoYield;
+                }
                 let options = CompactOptions {
                     flags,
                     sub_compaction: false,
@@ -1091,7 +1094,7 @@ impl Timeline {
         let latest_gc_cutoff = self.get_applied_gc_cutoff_lsn();
 
         tracing::info!(
-            "latest_gc_cutoff: {}, pitr cutoff {}",
+            "starting shard ancestor compaction, latest_gc_cutoff: {}, pitr cutoff {}",
             *latest_gc_cutoff,
             self.gc_info.read().unwrap().cutoffs.time
         );
@@ -1120,6 +1123,7 @@ impl Timeline {
                     // Expensive, exhaustive check of keys in this layer: this guards against ShardedRange's calculations being
                     // wrong.  If ShardedRange claims the local page count is zero, then no keys in this layer
                     // should be !is_key_disposable()
+                    // TODO: exclude sparse keyspace from this check, otherwise it will infinitely loop.
                     let range = layer_desc.get_key_range();
                     let mut key = range.start;
                     while key < range.end {
@@ -2616,6 +2620,7 @@ impl Timeline {
     ) -> Result<CompactionOutcome, CompactionError> {
         let sub_compaction = options.sub_compaction;
         let job = GcCompactJob::from_compact_options(options.clone());
+        let no_yield = options.flags.contains(CompactFlags::NoYield);
         if sub_compaction {
             info!(
                 "running enhanced gc bottom-most compaction with sub-compaction, splitting compaction jobs"
@@ -2630,14 +2635,15 @@ impl Timeline {
                     idx + 1,
                     jobs_len
                 );
-                self.compact_with_gc_inner(cancel, job, ctx).await?;
+                self.compact_with_gc_inner(cancel, job, ctx, no_yield)
+                    .await?;
             }
             if jobs_len == 0 {
                 info!("no jobs to run, skipping gc bottom-most compaction");
             }
             return Ok(CompactionOutcome::Done);
         }
-        self.compact_with_gc_inner(cancel, job, ctx).await
+        self.compact_with_gc_inner(cancel, job, ctx, no_yield).await
     }
 
     async fn compact_with_gc_inner(
@@ -2645,6 +2651,7 @@ impl Timeline {
         cancel: &CancellationToken,
         job: GcCompactJob,
         ctx: &RequestContext,
+        no_yield: bool,
     ) -> Result<CompactionOutcome, CompactionError> {
         // Block other compaction/GC tasks from running for now. GC-compaction could run along
         // with legacy compaction tasks in the future. Always ensure the lock order is compaction -> gc.
@@ -2914,14 +2921,18 @@ impl Timeline {
             if cancel.is_cancelled() {
                 return Err(CompactionError::ShuttingDown);
             }
-            let should_yield = self
-                .l0_compaction_trigger
-                .notified()
-                .now_or_never()
-                .is_some();
-            if should_yield {
-                tracing::info!("preempt gc-compaction when downloading layers: too many L0 layers");
-                return Ok(CompactionOutcome::YieldForL0);
+            if !no_yield {
+                let should_yield = self
+                    .l0_compaction_trigger
+                    .notified()
+                    .now_or_never()
+                    .is_some();
+                if should_yield {
+                    tracing::info!(
+                        "preempt gc-compaction when downloading layers: too many L0 layers"
+                    );
+                    return Ok(CompactionOutcome::YieldForL0);
+                }
             }
             let resident_layer = layer
                 .download_and_keep_resident(ctx)
@@ -3054,16 +3065,21 @@ impl Timeline {
             if cancel.is_cancelled() {
                 return Err(CompactionError::ShuttingDown);
             }
-            keys_processed += 1;
-            if keys_processed % 1000 == 0 {
-                let should_yield = self
-                    .l0_compaction_trigger
-                    .notified()
-                    .now_or_never()
-                    .is_some();
-                if should_yield {
-                    tracing::info!("preempt gc-compaction in the main loop: too many L0 layers");
-                    return Ok(CompactionOutcome::YieldForL0);
+
+            if !no_yield {
+                keys_processed += 1;
+                if keys_processed % 1000 == 0 {
+                    let should_yield = self
+                        .l0_compaction_trigger
+                        .notified()
+                        .now_or_never()
+                        .is_some();
+                    if should_yield {
+                        tracing::info!(
+                            "preempt gc-compaction in the main loop: too many L0 layers"
+                        );
+                        return Ok(CompactionOutcome::YieldForL0);
+                    }
                 }
             }
             if self.shard_identity.is_key_disposable(&key) {

pageserver/src/tenant/timeline/delete.rs

@@ -11,6 +11,7 @@ use utils::id::TimelineId;
 use utils::{crashsafe, fs_ext, pausable_failpoint};
 
 use crate::config::PageServerConf;
+use crate::context::RequestContext;
 use crate::task_mgr::{self, TaskKind};
 use crate::tenant::metadata::TimelineMetadata;
 use crate::tenant::remote_timeline_client::{
@@ -291,10 +292,11 @@ impl DeleteTimelineFlow {
         timeline_id: TimelineId,
         local_metadata: &TimelineMetadata,
         remote_client: RemoteTimelineClient,
+        ctx: &RequestContext,
     ) -> anyhow::Result<()> {
         // Note: here we even skip populating layer map. Timeline is essentially uninitialized.
         // RemoteTimelineClient is the only functioning part.
-        let timeline = tenant
+        let (timeline, _timeline_ctx) = tenant
             .create_timeline_struct(
                 timeline_id,
                 local_metadata,
@@ -307,6 +309,7 @@ impl DeleteTimelineFlow {
                 crate::tenant::CreateTimelineIdempotency::FailWithConflict, // doesn't matter what we put here
                 None, // doesn't matter what we put here
                 None, // doesn't matter what we put here
+                ctx,
             )
             .context("create_timeline_struct")?;
 

pageserver/src/tenant/timeline/detach_ancestor.rs

@@ -32,6 +32,9 @@ pub(crate) enum Error {
     #[error("too many ancestors")]
     TooManyAncestors,
 
+    #[error("ancestor is not empty")]
+    AncestorNotEmpty,
+
     #[error("shutting down, please retry later")]
     ShuttingDown,
 
@@ -89,7 +92,9 @@ impl From<Error> for ApiError {
     fn from(value: Error) -> Self {
         match value {
             Error::NoAncestor => ApiError::Conflict(value.to_string()),
-            Error::TooManyAncestors => ApiError::BadRequest(anyhow::anyhow!("{value}")),
+            Error::TooManyAncestors | Error::AncestorNotEmpty => {
+                ApiError::BadRequest(anyhow::anyhow!("{value}"))
+            }
             Error::ShuttingDown => ApiError::ShuttingDown,
             Error::Archived(_) => ApiError::BadRequest(anyhow::anyhow!("{value}")),
             Error::OtherTimelineDetachOngoing(_) | Error::FailedToReparentAll => {
@@ -127,13 +132,37 @@ pub(crate) struct PreparedTimelineDetach {
     layers: Vec<Layer>,
 }
 
-/// TODO: this should be part of PageserverConf because we cannot easily modify cplane arguments.
+// TODO: this should be part of PageserverConf because we cannot easily modify cplane arguments.
 #[derive(Debug)]
 pub(crate) struct Options {
     pub(crate) rewrite_concurrency: std::num::NonZeroUsize,
     pub(crate) copy_concurrency: std::num::NonZeroUsize,
 }
 
+/// Controls the detach ancestor behavior.
+/// - When set to `NoAncestorAndReparent`, we will only detach a branch if its ancestor is a root branch. It will automatically reparent any children of the ancestor before and at the branch point.
+/// - When set to `MultiLevelAndNoReparent`, we will detach a branch from multiple levels of ancestors, and no reparenting will happen at all.
+#[derive(Debug, Clone, Copy, Default)]
+pub enum DetachBehavior {
+    #[default]
+    NoAncestorAndReparent,
+    MultiLevelAndNoReparent,
+}
+
+impl std::str::FromStr for DetachBehavior {
+    type Err = &'static str;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        match s {
+            "no_ancestor_and_reparent" => Ok(DetachBehavior::NoAncestorAndReparent),
+            "multi_level_and_no_reparent" => Ok(DetachBehavior::MultiLevelAndNoReparent),
+            "v1" => Ok(DetachBehavior::NoAncestorAndReparent),
+            "v2" => Ok(DetachBehavior::MultiLevelAndNoReparent),
+            _ => Err("cannot parse detach behavior"),
+        }
+    }
+}
+
 impl Default for Options {
     fn default() -> Self {
         Self {
@@ -147,7 +176,8 @@ impl Default for Options {
 #[derive(Debug)]
 pub(crate) struct Attempt {
     pub(crate) timeline_id: TimelineId,
-
+    pub(crate) ancestor_timeline_id: TimelineId,
+    pub(crate) ancestor_lsn: Lsn,
     _guard: completion::Completion,
     gate_entered: Option<utils::sync::gate::GateGuard>,
 }
@@ -167,25 +197,30 @@ impl Attempt {
 pub(super) async fn prepare(
     detached: &Arc<Timeline>,
     tenant: &Tenant,
+    behavior: DetachBehavior,
     options: Options,
     ctx: &RequestContext,
 ) -> Result<Progress, Error> {
     use Error::*;
 
-    let Some((ancestor, ancestor_lsn)) = detached
+    let Some((mut ancestor, mut ancestor_lsn)) = detached
         .ancestor_timeline
         .as_ref()
         .map(|tl| (tl.clone(), detached.ancestor_lsn))
     else {
+        let ancestor_id;
+        let ancestor_lsn;
         let still_in_progress = {
             let accessor = detached.remote_client.initialized_upload_queue()?;
 
             // we are safe to inspect the latest uploaded, because we can only witness this after
             // restart is complete and ancestor is no more.
             let latest = accessor.latest_uploaded_index_part();
-            if latest.lineage.detached_previous_ancestor().is_none() {
+            let Some((id, lsn)) = latest.lineage.detached_previous_ancestor() else {
                 return Err(NoAncestor);
             };
+            ancestor_id = id;
+            ancestor_lsn = lsn;
 
             latest
                 .gc_blocking
@@ -196,7 +231,8 @@ pub(super) async fn prepare(
         if still_in_progress {
             // gc is still blocked, we can still reparent and complete.
             // we are safe to reparent remaining, because they were locked in in the beginning.
-            let attempt = continue_with_blocked_gc(detached, tenant).await?;
+            let attempt =
+                continue_with_blocked_gc(detached, tenant, ancestor_id, ancestor_lsn).await?;
 
             // because the ancestor of detached is already set to none, we have published all
             // of the layers, so we are still "prepared."
@@ -224,13 +260,34 @@ pub(super) async fn prepare(
 
     check_no_archived_children_of_ancestor(tenant, detached, &ancestor, ancestor_lsn)?;
 
-    if ancestor.ancestor_timeline.is_some() {
+    if let DetachBehavior::MultiLevelAndNoReparent = behavior {
+        // If the ancestor has an ancestor, we might be able to fast-path detach it if the current ancestor does not have any data written/used by the detaching timeline.
+        while let Some(ancestor_of_ancestor) = ancestor.ancestor_timeline.clone() {
+            if ancestor_lsn != ancestor.ancestor_lsn {
+                // non-technical requirement; we could flatten still if ancestor LSN does not match but that needs
+                // us to copy and cut more layers.
+                return Err(AncestorNotEmpty);
+            }
+            // Use the ancestor of the ancestor as the new ancestor (only when the ancestor LSNs are the same)
+            ancestor_lsn = ancestor.ancestor_lsn; // Get the LSN first before resetting the `ancestor` variable
+            ancestor = ancestor_of_ancestor;
+            // TODO: do we still need to check if we don't want to reparent?
+            check_no_archived_children_of_ancestor(tenant, detached, &ancestor, ancestor_lsn)?;
+        }
+    } else if ancestor.ancestor_timeline.is_some() {
         // non-technical requirement; we could flatten N ancestors just as easily but we chose
         // not to, at least initially
         return Err(TooManyAncestors);
     }
 
-    let attempt = start_new_attempt(detached, tenant).await?;
+    tracing::info!(
+        "attempt to detach the timeline from the ancestor: {}@{}, behavior={:?}",
+        ancestor.timeline_id,
+        ancestor_lsn,
+        behavior
+    );
+
+    let attempt = start_new_attempt(detached, tenant, ancestor.timeline_id, ancestor_lsn).await?;
 
     utils::pausable_failpoint!("timeline-detach-ancestor::before_starting_after_locking-pausable");
 
@@ -450,8 +507,13 @@ pub(super) async fn prepare(
     Ok(Progress::Prepared(attempt, prepared))
 }
 
-async fn start_new_attempt(detached: &Timeline, tenant: &Tenant) -> Result<Attempt, Error> {
-    let attempt = obtain_exclusive_attempt(detached, tenant)?;
+async fn start_new_attempt(
+    detached: &Timeline,
+    tenant: &Tenant,
+    ancestor_timeline_id: TimelineId,
+    ancestor_lsn: Lsn,
+) -> Result<Attempt, Error> {
+    let attempt = obtain_exclusive_attempt(detached, tenant, ancestor_timeline_id, ancestor_lsn)?;
 
     // insert the block in the index_part.json, if not already there.
     let _dont_care = tenant
@@ -466,13 +528,23 @@ async fn start_new_attempt(detached: &Timeline, tenant: &Tenant) -> Result<Attem
     Ok(attempt)
 }
 
-async fn continue_with_blocked_gc(detached: &Timeline, tenant: &Tenant) -> Result<Attempt, Error> {
+async fn continue_with_blocked_gc(
+    detached: &Timeline,
+    tenant: &Tenant,
+    ancestor_timeline_id: TimelineId,
+    ancestor_lsn: Lsn,
+) -> Result<Attempt, Error> {
     // FIXME: it would be nice to confirm that there is an in-memory version, since we've just
     // verified there is a persistent one?
-    obtain_exclusive_attempt(detached, tenant)
+    obtain_exclusive_attempt(detached, tenant, ancestor_timeline_id, ancestor_lsn)
 }
 
-fn obtain_exclusive_attempt(detached: &Timeline, tenant: &Tenant) -> Result<Attempt, Error> {
+fn obtain_exclusive_attempt(
+    detached: &Timeline,
+    tenant: &Tenant,
+    ancestor_timeline_id: TimelineId,
+    ancestor_lsn: Lsn,
+) -> Result<Attempt, Error> {
     use Error::{OtherTimelineDetachOngoing, ShuttingDown};
 
     // ensure we are the only active attempt for this tenant
@@ -493,6 +565,8 @@ fn obtain_exclusive_attempt(detached: &Timeline, tenant: &Tenant) -> Result<Atte
 
     Ok(Attempt {
         timeline_id: detached.timeline_id,
+        ancestor_timeline_id,
+        ancestor_lsn,
         _guard: guard,
         gate_entered: Some(_gate_entered),
     })
@@ -795,6 +869,9 @@ pub(super) async fn detach_and_reparent(
     detached: &Arc<Timeline>,
     tenant: &Tenant,
     prepared: PreparedTimelineDetach,
+    ancestor_timeline_id: TimelineId,
+    ancestor_lsn: Lsn,
+    behavior: DetachBehavior,
     _ctx: &RequestContext,
 ) -> Result<DetachingAndReparenting, Error> {
     let PreparedTimelineDetach { layers } = prepared;
@@ -822,7 +899,30 @@ pub(super) async fn detach_and_reparent(
         "cannot (detach? reparent)? complete if the operation is not still ongoing"
     );
 
-    let ancestor = match (detached.ancestor_timeline.as_ref(), recorded_branchpoint) {
+    let ancestor_to_detach = match detached.ancestor_timeline.as_ref() {
+        Some(mut ancestor) => {
+            while ancestor.timeline_id != ancestor_timeline_id {
+                match ancestor.ancestor_timeline.as_ref() {
+                    Some(found) => {
+                        if ancestor_lsn != ancestor.ancestor_lsn {
+                            return Err(Error::DetachReparent(anyhow::anyhow!(
+                                "cannot find the ancestor timeline to detach from: wrong ancestor lsn"
+                            )));
+                        }
+                        ancestor = found;
+                    }
+                    None => {
+                        return Err(Error::DetachReparent(anyhow::anyhow!(
+                            "cannot find the ancestor timeline to detach from"
+                        )));
+                    }
+                }
+            }
+            Some(ancestor)
+        }
+        None => None,
+    };
+    let ancestor = match (ancestor_to_detach, recorded_branchpoint) {
         (Some(ancestor), None) => {
             assert!(
                 !layers.is_empty(),
@@ -895,6 +995,11 @@ pub(super) async fn detach_and_reparent(
         Ancestor::Detached(ancestor, ancestor_lsn) => (ancestor, ancestor_lsn, false),
     };
 
+    if let DetachBehavior::MultiLevelAndNoReparent = behavior {
+        // Do not reparent if the user requests to behave so.
+        return Ok(DetachingAndReparenting::Reparented(HashSet::new()));
+    }
+
     let mut tasks = tokio::task::JoinSet::new();
 
     // Returns a single permit semaphore which will be used to make one reparenting succeed,
@@ -1032,6 +1137,11 @@ pub(super) async fn complete(
 }
 
 /// Query against a locked `Tenant::timelines`.
+///
+/// A timeline is reparentable if:
+///
+/// - It is not the timeline being detached.
+/// - It has the same ancestor as the timeline being detached. Note that the ancestor might not be the direct ancestor.
 fn reparentable_timelines<'a, I>(
     timelines: I,
     detached: &'a Arc<Timeline>,

pageserver/src/tenant/timeline/eviction_task.rs

@@ -93,7 +93,8 @@ impl Timeline {
             }
         }
 
-        let ctx = RequestContext::new(TaskKind::Eviction, DownloadBehavior::Warn);
+        let ctx = RequestContext::new(TaskKind::Eviction, DownloadBehavior::Warn)
+            .with_scope_timeline(&self);
         loop {
             let policy = self.get_eviction_policy();
             let cf = self

pageserver/src/tenant/vectored_blob_io.rs

@@ -961,7 +961,8 @@ mod tests {
     }
 
     async fn round_trip_test_compressed(blobs: &[Vec<u8>], compression: bool) -> Result<(), Error> {
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
         let (_temp_dir, pathbuf, offsets) =
             write_maybe_compressed::<true>(blobs, compression, &ctx).await?;
 

pageserver/src/virtual_file.rs

@@ -26,15 +26,14 @@ use owned_buffers_io::io_buf_aligned::{IoBufAligned, IoBufAlignedMut};
 use owned_buffers_io::io_buf_ext::FullSlice;
 use pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
 pub use pageserver_api::models::virtual_file as api;
-use pageserver_api::shard::TenantShardId;
 use tokio::sync::{RwLock, RwLockReadGuard, RwLockWriteGuard};
 use tokio::time::Instant;
 use tokio_epoll_uring::{BoundedBuf, IoBuf, IoBufMut, Slice};
 
+use crate::assert_u64_eq_usize::UsizeIsU64;
 use crate::context::RequestContext;
-use crate::metrics::{STORAGE_IO_SIZE, STORAGE_IO_TIME_METRIC, StorageIoOperation};
+use crate::metrics::{STORAGE_IO_TIME_METRIC, StorageIoOperation};
 use crate::page_cache::{PAGE_SZ, PageWriteGuard};
-use crate::tenant::TENANTS_SEGMENT_NAME;
 pub(crate) mod io_engine;
 pub use io_engine::{
     FeatureTestResult as IoEngineFeatureTestResult, feature_test as io_engine_feature_test,
@@ -121,7 +120,7 @@ impl VirtualFile {
     pub async fn open_with_options<P: AsRef<Utf8Path>>(
         path: P,
         open_options: &OpenOptions,
-        ctx: &RequestContext, /* TODO: carry a pointer to the metrics in the RequestContext instead of the parsing https://github.com/neondatabase/neon/issues/6107 */
+        ctx: &RequestContext,
     ) -> Result<Self, std::io::Error> {
         let inner = VirtualFileInner::open_with_options(path, open_options, ctx).await?;
         Ok(VirtualFile {
@@ -133,7 +132,7 @@ impl VirtualFile {
     pub async fn open_with_options_v2<P: AsRef<Utf8Path>>(
         path: P,
         open_options: &OpenOptions,
-        ctx: &RequestContext, /* TODO: carry a pointer to the metrics in the RequestContext instead of the parsing https://github.com/neondatabase/neon/issues/6107 */
+        ctx: &RequestContext,
     ) -> Result<Self, std::io::Error> {
         let file = match get_io_mode() {
             IoMode::Buffered => {
@@ -304,13 +303,6 @@ pub struct VirtualFileInner {
     /// storing it here.
     pub path: Utf8PathBuf,
     open_options: OpenOptions,
-
-    // These are strings becase we only use them for metrics, and those expect strings.
-    // It makes no sense for us to constantly turn the `TimelineId` and `TenantId` into
-    // strings.
-    tenant_id: String,
-    shard_id: String,
-    timeline_id: String,
 }
 
 #[derive(Debug, PartialEq, Clone, Copy)]
@@ -592,36 +584,16 @@ impl VirtualFileInner {
     pub async fn open_with_options<P: AsRef<Utf8Path>>(
         path: P,
         open_options: &OpenOptions,
-        _ctx: &RequestContext, /* TODO: carry a pointer to the metrics in the RequestContext instead of the parsing https://github.com/neondatabase/neon/issues/6107 */
+        _ctx: &RequestContext,
     ) -> Result<VirtualFileInner, std::io::Error> {
-        let path_ref = path.as_ref();
-        let path_str = path_ref.to_string();
-        let parts = path_str.split('/').collect::<Vec<&str>>();
-        let (tenant_id, shard_id, timeline_id) =
-            if parts.len() > 5 && parts[parts.len() - 5] == TENANTS_SEGMENT_NAME {
-                let tenant_shard_part = parts[parts.len() - 4];
-                let (tenant_id, shard_id) = match tenant_shard_part.parse::<TenantShardId>() {
-                    Ok(tenant_shard_id) => (
-                        tenant_shard_id.tenant_id.to_string(),
-                        format!("{}", tenant_shard_id.shard_slug()),
-                    ),
-                    Err(_) => {
-                        // Malformed path: this ID is just for observability, so tolerate it
-                        // and pass through
-                        (tenant_shard_part.to_string(), "*".to_string())
-                    }
-                };
-                (tenant_id, shard_id, parts[parts.len() - 2].to_string())
-            } else {
-                ("*".to_string(), "*".to_string(), "*".to_string())
-            };
+        let path = path.as_ref();
         let (handle, mut slot_guard) = get_open_files().find_victim_slot().await;
 
         // NB: there is also StorageIoOperation::OpenAfterReplace which is for the case
         // where our caller doesn't get to use the returned VirtualFile before its
         // slot gets re-used by someone else.
         let file = observe_duration!(StorageIoOperation::Open, {
-            open_options.open(path_ref.as_std_path()).await?
+            open_options.open(path.as_std_path()).await?
         });
 
         // Strip all options other than read and write.
@@ -637,11 +609,8 @@ impl VirtualFileInner {
         let vfile = VirtualFileInner {
             handle: RwLock::new(handle),
             pos: 0,
-            path: path_ref.to_path_buf(),
+            path: path.to_owned(),
             open_options: reopen_options,
-            tenant_id,
-            shard_id,
-            timeline_id,
         };
 
         // TODO: Under pressure, it's likely the slot will get re-used and
@@ -944,7 +913,7 @@ impl VirtualFileInner {
         &self,
         buf: tokio_epoll_uring::Slice<Buf>,
         offset: u64,
-        _ctx: &RequestContext, /* TODO: use for metrics: https://github.com/neondatabase/neon/issues/6107 */
+        ctx: &RequestContext,
     ) -> (tokio_epoll_uring::Slice<Buf>, Result<usize, Error>)
     where
         Buf: tokio_epoll_uring::IoBufMut + Send,
@@ -962,14 +931,7 @@ impl VirtualFileInner {
             let ((_file_guard, buf), res) = io_engine::get().read_at(file_guard, offset, buf).await;
             let res = res.maybe_fatal_err("io_engine read_at inside VirtualFileInner::read_at");
             if let Ok(size) = res {
-                STORAGE_IO_SIZE
-                    .with_label_values(&[
-                        "read",
-                        &self.tenant_id,
-                        &self.shard_id,
-                        &self.timeline_id,
-                    ])
-                    .add(size as i64);
+                ctx.io_size_metrics().read.add(size.into_u64());
             }
             (buf, res)
         })
@@ -980,9 +942,9 @@ impl VirtualFileInner {
         &self,
         buf: FullSlice<B>,
         offset: u64,
-        _ctx: &RequestContext, /* TODO: use for metrics: https://github.com/neondatabase/neon/issues/6107 */
+        ctx: &RequestContext,
     ) -> (FullSlice<B>, Result<usize, Error>) {
-        let (slice, result) = self.write_at_inner(buf, offset, _ctx).await;
+        let (slice, result) = self.write_at_inner(buf, offset, ctx).await;
         let result = result.maybe_fatal_err("write_at");
         (slice, result)
     }
@@ -991,7 +953,7 @@ impl VirtualFileInner {
         &self,
         buf: FullSlice<B>,
         offset: u64,
-        _ctx: &RequestContext, /* TODO: use for metrics: https://github.com/neondatabase/neon/issues/6107 */
+        ctx: &RequestContext,
     ) -> (FullSlice<B>, Result<usize, Error>) {
         let file_guard = match self.lock_file().await {
             Ok(file_guard) => file_guard,
@@ -1001,14 +963,7 @@ impl VirtualFileInner {
             let ((_file_guard, buf), result) =
                 io_engine::get().write_at(file_guard, offset, buf).await;
             if let Ok(size) = result {
-                STORAGE_IO_SIZE
-                    .with_label_values(&[
-                        "write",
-                        &self.tenant_id,
-                        &self.shard_id,
-                        &self.timeline_id,
-                    ])
-                    .add(size as i64);
+                ctx.io_size_metrics().write.add(size.into_u64());
             }
             (buf, result)
         })
@@ -1344,9 +1299,8 @@ impl OwnedAsyncWriter for VirtualFile {
         buf: FullSlice<Buf>,
         offset: u64,
         ctx: &RequestContext,
-    ) -> std::io::Result<FullSlice<Buf>> {
-        let (buf, res) = VirtualFile::write_all_at(self, buf, offset, ctx).await;
-        res.map(|_| buf)
+    ) -> (FullSlice<Buf>, std::io::Result<()>) {
+        VirtualFile::write_all_at(self, buf, offset, ctx).await
     }
 }
 
@@ -1593,7 +1547,8 @@ mod tests {
     where
         A: Adapter,
     {
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
         let testdir = crate::config::PageServerConf::test_repo_dir(testname);
         std::fs::create_dir_all(&testdir)?;
 
@@ -1720,7 +1675,8 @@ mod tests {
         const THREADS: usize = 100;
         const SAMPLE: [u8; SIZE] = [0xADu8; SIZE];
 
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
         let testdir = crate::config::PageServerConf::test_repo_dir("vfile_concurrency");
         std::fs::create_dir_all(&testdir)?;
 
@@ -1779,7 +1735,8 @@ mod tests {
 
     #[tokio::test]
     async fn test_atomic_overwrite_basic() {
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
         let testdir = crate::config::PageServerConf::test_repo_dir("test_atomic_overwrite_basic");
         std::fs::create_dir_all(&testdir).unwrap();
 
@@ -1807,7 +1764,8 @@ mod tests {
 
     #[tokio::test]
     async fn test_atomic_overwrite_preexisting_tmp() {
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+        let ctx =
+            RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error).with_scope_unit_test();
         let testdir =
             crate::config::PageServerConf::test_repo_dir("test_atomic_overwrite_preexisting_tmp");
         std::fs::create_dir_all(&testdir).unwrap();

pageserver/src/virtual_file/owned_buffers_io/write.rs

@@ -31,7 +31,7 @@ pub trait OwnedAsyncWriter {
         buf: FullSlice<Buf>,
         offset: u64,
         ctx: &RequestContext,
-    ) -> impl std::future::Future<Output = std::io::Result<FullSlice<Buf>>> + Send;
+    ) -> impl std::future::Future<Output = (FullSlice<Buf>, std::io::Result<()>)> + Send;
 }
 
 /// A wrapper aorund an [`OwnedAsyncWriter`] that uses a [`Buffer`] to batch
@@ -66,6 +66,7 @@ where
         buf_new: impl Fn() -> B,
         gate_guard: utils::sync::gate::GateGuard,
         ctx: &RequestContext,
+        flush_task_span: tracing::Span,
     ) -> Self {
         Self {
             writer: writer.clone(),
@@ -75,6 +76,7 @@ where
                 buf_new(),
                 gate_guard,
                 ctx.attached_child(),
+                flush_task_span,
             ),
             bytes_submitted: 0,
         }
@@ -269,12 +271,12 @@ mod tests {
             buf: FullSlice<Buf>,
             offset: u64,
             _: &RequestContext,
-        ) -> std::io::Result<FullSlice<Buf>> {
+        ) -> (FullSlice<Buf>, std::io::Result<()>) {
             self.writes
                 .lock()
                 .unwrap()
                 .push((Vec::from(&buf[..]), offset));
-            Ok(buf)
+            (buf, Ok(()))
         }
     }
 
@@ -293,6 +295,7 @@ mod tests {
             || IoBufferMut::with_capacity(2),
             gate.enter()?,
             ctx,
+            tracing::Span::none(),
         );
 
         writer.write_buffered_borrowed(b"abc", ctx).await?;

pageserver/src/virtual_file/owned_buffers_io/write/flush.rs

@@ -1,9 +1,14 @@
+use std::ops::ControlFlow;
 use std::sync::Arc;
 
+use once_cell::sync::Lazy;
+use tokio_util::sync::CancellationToken;
+use tracing::{Instrument, info, info_span, warn};
 use utils::sync::duplex;
 
 use super::{Buffer, CheapCloneForRead, OwnedAsyncWriter};
 use crate::context::RequestContext;
+use crate::virtual_file::MaybeFatalIo;
 use crate::virtual_file::owned_buffers_io::io_buf_aligned::IoBufAligned;
 use crate::virtual_file::owned_buffers_io::io_buf_ext::FullSlice;
 
@@ -118,6 +123,7 @@ where
         buf: B,
         gate_guard: utils::sync::gate::GateGuard,
         ctx: RequestContext,
+        span: tracing::Span,
     ) -> Self
     where
         B: Buffer<IoBuf = Buf> + Send + 'static,
@@ -125,11 +131,14 @@ where
         // It is fine to buffer up to only 1 message. We only 1 message in-flight at a time.
         let (front, back) = duplex::mpsc::channel(1);
 
-        let join_handle = tokio::spawn(async move {
-            FlushBackgroundTask::new(back, file, gate_guard, ctx)
-                .run(buf.flush())
-                .await
-        });
+        let join_handle = tokio::spawn(
+            async move {
+                FlushBackgroundTask::new(back, file, gate_guard, ctx)
+                    .run(buf.flush())
+                    .await
+            }
+            .instrument(span),
+        );
 
         FlushHandle {
             inner: Some(FlushHandleInner {
@@ -236,6 +245,7 @@ where
     /// The passed in slice is immediately sent back to the flush handle through the duplex channel.
     async fn run(mut self, slice: FullSlice<Buf>) -> std::io::Result<Arc<W>> {
         // Sends the extra buffer back to the handle.
+        // TODO: can this ever await and or fail? I think not.
         self.channel.send(slice).await.map_err(|_| {
             std::io::Error::new(std::io::ErrorKind::BrokenPipe, "flush handle closed early")
         })?;
@@ -251,10 +261,47 @@ where
             }
 
             // Write slice to disk at `offset`.
-            let slice = self
-                .writer
-                .write_all_at(request.slice, request.offset, &self.ctx)
-                .await?;
+            //
+            // Error handling happens according to the current policy of crashing
+            // on fatal IO errors and retrying in place otherwise (deeming all other errors retryable).
+            // (The upper layers of the Pageserver write path are not equipped to retry write errors
+            //  becasuse they often deallocate the buffers that were already written).
+            //
+            // TODO: cancellation sensitiity.
+            // Without it, if we hit a bug where retrying is never successful,
+            // then we can't shut down the timeline/tenant/pageserver cleanly because
+            // layers of the Pageserver write path are holding the gate open for EphemeralFile.
+            //
+            // TODO: use utils::backoff::retry once async closures are actually usable
+            //
+            let mut slice_storage = Some(request.slice);
+            for attempt in 1.. {
+                let result = async {
+                    if attempt > 1 {
+                        info!("retrying flush");
+                    }
+                    let slice = slice_storage.take().expect(
+                        "likely previous invocation of this future didn't get polled to completion",
+                    );
+                    let (slice, res) = self.writer.write_all_at(slice, request.offset, &self.ctx).await;
+                    slice_storage = Some(slice);
+                    let res = res.maybe_fatal_err("owned_buffers_io flush");
+                    let Err(err) = res else {
+                        return ControlFlow::Break(());
+                    };
+                    warn!(%err, "error flushing buffered writer buffer to disk, retrying after backoff");
+                    static NO_CANCELLATION: Lazy<CancellationToken> = Lazy::new(CancellationToken::new);
+                    utils::backoff::exponential_backoff(attempt, 1.0, 10.0, &NO_CANCELLATION).await;
+                    ControlFlow::Continue(())
+                }
+                .instrument(info_span!("flush_attempt", %attempt))
+                .await;
+                match result {
+                    ControlFlow::Break(()) => break,
+                    ControlFlow::Continue(()) => continue,
+                }
+            }
+            let slice = slice_storage.expect("loop must have run at least once");
 
             #[cfg(test)]
             {

pgxn/neon/file_cache.c

@@ -1195,9 +1195,11 @@ lfc_writev(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 				state = GET_STATE(entry, chunk_offs + i);
 				if (state == PENDING) {
 					SET_STATE(entry, chunk_offs + i, REQUESTED);
-				} else if (state != REQUESTED) {
+				} else if (state == UNAVAILABLE) {
 					SET_STATE(entry, chunk_offs + i, PENDING);
 					break;
+				} else if (state == AVAILABLE) {
+					break;
 				}
 				if (!sleeping)
 				{

pgxn/neon/libpagestore.c

@@ -16,6 +16,8 @@
 
 #include <math.h>
 
+#include "libpq-int.h"
+
 #include "access/xlog.h"
 #include "common/hashfn.h"
 #include "fmgr.h"
@@ -815,9 +817,10 @@ retry:
 			get_socket_stats(PQsocket(pageserver_conn), &sndbuf, &recvbuf);
 
 			neon_shard_log(shard_no, LOG,
-						   "no response received from pageserver for %0.3f s, still waiting (sent " UINT64_FORMAT " requests, received " UINT64_FORMAT " responses) (socket sndbuf=%d recvbuf=%d)",
+						   "no response received from pageserver for %0.3f s, still waiting (sent " UINT64_FORMAT " requests, received " UINT64_FORMAT " responses) (socket sndbuf=%d recvbuf=%d) (conn start=%d end=%d)",
 						   INSTR_TIME_GET_DOUBLE(since_start),
-						   shard->nrequests_sent, shard->nresponses_received, sndbuf, recvbuf);
+						   shard->nrequests_sent, shard->nresponses_received, sndbuf, recvbuf,
+				           pageserver_conn->inStart, pageserver_conn->inEnd);
 			shard->receive_last_log_time = now;
 			shard->receive_logged = true;
 		}

pgxn/neon/neon.c

@@ -457,6 +457,15 @@ _PG_init(void)
 							PGC_SIGHUP,
 							0,
 							NULL, NULL, NULL);
+	DefineCustomBoolVariable(
+							"neon.disable_wal_prevlink_checks",
+							"Disable validation of prev link in WAL records",
+							NULL,
+							&disable_wal_prev_lsn_checks,
+							false,
+							PGC_SIGHUP,
+							0,
+							NULL, NULL, NULL);
 
 	DefineCustomBoolVariable(
 							"neon.allow_replica_misconfig",

pgxn/neon/neon.h

@@ -23,6 +23,7 @@ extern char *wal_acceptors_list;
 extern int	wal_acceptor_reconnect_timeout;
 extern int	wal_acceptor_connection_timeout;
 extern int	readahead_getpage_pull_timeout_ms;
+extern bool	disable_wal_prev_lsn_checks;
 
 #if PG_MAJORVERSION_NUM >= 17
 extern uint32		WAIT_EVENT_NEON_LFC_MAINTENANCE;

pgxn/neon/pagestore_smgr.c

@@ -76,6 +76,10 @@
 #include "access/xlogrecovery.h"
 #endif
 
+#if PG_VERSION_NUM < 160000
+typedef PGAlignedBlock PGIOAlignedBlock;
+#endif
+
 /*
  * If DEBUG_COMPARE_LOCAL is defined, we pass through all the SMGR API
  * calls to md.c, and *also* do the calls to the Page Server. On every
@@ -1803,7 +1807,7 @@ static XLogRecPtr
 log_newpage_copy(NRelFileInfo * rinfo, ForkNumber forkNum, BlockNumber blkno,
 				 Page page, bool page_std)
 {
-	PGAlignedBlock copied_buffer;
+	PGIOAlignedBlock copied_buffer;
 
 	memcpy(copied_buffer.data, page, BLCKSZ);
 	return log_newpage(rinfo, forkNum, blkno, copied_buffer.data, page_std);
@@ -1820,7 +1824,7 @@ static XLogRecPtr
 log_newpages_copy(NRelFileInfo * rinfo, ForkNumber forkNum, BlockNumber blkno,
 				  BlockNumber nblocks, Page *pages, bool page_std)
 {
-	PGAlignedBlock copied_buffer[XLR_MAX_BLOCK_ID];
+	PGIOAlignedBlock copied_buffer[XLR_MAX_BLOCK_ID];
 	BlockNumber	blknos[XLR_MAX_BLOCK_ID];
 	Page		pageptrs[XLR_MAX_BLOCK_ID];
 	int			nregistered = 0;
@@ -1858,7 +1862,7 @@ log_newpages_copy(NRelFileInfo * rinfo, ForkNumber forkNum, BlockNumber blkno,
 static bool
 PageIsEmptyHeapPage(char *buffer)
 {
-	PGAlignedBlock empty_page;
+	PGIOAlignedBlock empty_page;
 
 	PageInit((Page) empty_page.data, BLCKSZ, 0);
 
@@ -2847,7 +2851,7 @@ static void
 neon_zeroextend(SMgrRelation reln, ForkNumber forkNum, BlockNumber blocknum,
 				int nblocks, bool skipFsync)
 {
-	const PGAlignedBlock buffer = {0};
+	const PGIOAlignedBlock buffer = {0};
 	int			remblocks = nblocks;
 	XLogRecPtr	lsn = 0;
 
@@ -3389,15 +3393,16 @@ neon_read(SMgrRelation reln, ForkNumber forkNum, BlockNumber blkno, void *buffer
 	if (forkNum == MAIN_FORKNUM && IS_LOCAL_REL(reln))
 	{
 		char		pageserver_masked[BLCKSZ];
-		char		mdbuf[BLCKSZ];
-		char		mdbuf_masked[BLCKSZ];
+		PGIOAlignedBlock mdbuf;
+		PGIOAlignedBlock mdbuf_masked;
+		XLogRecPtr  request_lsn = request_lsns.request_lsn;
 
-		mdread(reln, forkNum, blkno, mdbuf);
+		mdread(reln, forkNum, blkno, mdbuf.data);
 
 		memcpy(pageserver_masked, buffer, BLCKSZ);
-		memcpy(mdbuf_masked, mdbuf, BLCKSZ);
+		memcpy(mdbuf_masked.data, mdbuf.data, BLCKSZ);
 
-		if (PageIsNew((Page) mdbuf))
+		if (PageIsNew((Page) mdbuf.data))
 		{
 			if (!PageIsNew((Page) pageserver_masked))
 			{
@@ -3416,41 +3421,41 @@ neon_read(SMgrRelation reln, ForkNumber forkNum, BlockNumber blkno, void *buffer
 				 RelFileInfoFmt(InfoFromSMgrRel(reln)),
 				 forkNum,
 				 (uint32) (request_lsn >> 32), (uint32) request_lsn,
-				 hexdump_page(mdbuf));
+				 hexdump_page(mdbuf.data));
 		}
-		else if (PageGetSpecialSize(mdbuf) == 0)
+		else if (PageGetSpecialSize(mdbuf.data) == 0)
 		{
 			/* assume heap */
-			RmgrTable[RM_HEAP_ID].rm_mask(mdbuf_masked, blkno);
+			RmgrTable[RM_HEAP_ID].rm_mask(mdbuf_masked.data, blkno);
 			RmgrTable[RM_HEAP_ID].rm_mask(pageserver_masked, blkno);
 
-			if (memcmp(mdbuf_masked, pageserver_masked, BLCKSZ) != 0)
+			if (memcmp(mdbuf_masked.data, pageserver_masked, BLCKSZ) != 0)
 			{
 				neon_log(PANIC, "heap buffers differ at blk %u in rel %u/%u/%u fork %u (request LSN %X/%08X):\n------ MD ------\n%s\n------ Page Server ------\n%s\n",
 					 blkno,
 					 RelFileInfoFmt(InfoFromSMgrRel(reln)),
 					 forkNum,
 					 (uint32) (request_lsn >> 32), (uint32) request_lsn,
-					 hexdump_page(mdbuf_masked),
+					 hexdump_page(mdbuf_masked.data),
 					 hexdump_page(pageserver_masked));
 			}
 		}
-		else if (PageGetSpecialSize(mdbuf) == MAXALIGN(sizeof(BTPageOpaqueData)))
+		else if (PageGetSpecialSize(mdbuf.data) == MAXALIGN(sizeof(BTPageOpaqueData)))
 		{
-			if (((BTPageOpaqueData *) PageGetSpecialPointer(mdbuf))->btpo_cycleid < MAX_BT_CYCLE_ID)
+			if (((BTPageOpaqueData *) PageGetSpecialPointer(mdbuf.data))->btpo_cycleid < MAX_BT_CYCLE_ID)
 			{
 				/* assume btree */
-				RmgrTable[RM_BTREE_ID].rm_mask(mdbuf_masked, blkno);
+				RmgrTable[RM_BTREE_ID].rm_mask(mdbuf_masked.data, blkno);
 				RmgrTable[RM_BTREE_ID].rm_mask(pageserver_masked, blkno);
 
-				if (memcmp(mdbuf_masked, pageserver_masked, BLCKSZ) != 0)
+				if (memcmp(mdbuf_masked.data, pageserver_masked, BLCKSZ) != 0)
 				{
 					neon_log(PANIC, "btree buffers differ at blk %u in rel %u/%u/%u fork %u (request LSN %X/%08X):\n------ MD ------\n%s\n------ Page Server ------\n%s\n",
 						 blkno,
 						 RelFileInfoFmt(InfoFromSMgrRel(reln)),
 						 forkNum,
 						 (uint32) (request_lsn >> 32), (uint32) request_lsn,
-						 hexdump_page(mdbuf_masked),
+						 hexdump_page(mdbuf_masked.data),
 						 hexdump_page(pageserver_masked));
 				}
 			}
@@ -3542,77 +3547,85 @@ neon_readv(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
 	prefetch_pump_state(false);
 
 #ifdef DEBUG_COMPARE_LOCAL
-	if (forkNum == MAIN_FORKNUM && IS_LOCAL_REL(reln))
+	if (forknum == MAIN_FORKNUM && IS_LOCAL_REL(reln))
 	{
 		char		pageserver_masked[BLCKSZ];
-		char		mdbuf[BLCKSZ];
-		char		mdbuf_masked[BLCKSZ];
+		PGIOAlignedBlock mdbuf;
+		PGIOAlignedBlock mdbuf_masked;
+		XLogRecPtr  request_lsn = request_lsns->request_lsn;
 
 		for (int i = 0; i < nblocks; i++)
 		{
+			BlockNumber blkno = blocknum + i;
+			if (!BITMAP_ISSET(read, i))
+				continue;
+
 #if PG_MAJORVERSION_NUM >= 17
-			mdreadv(reln, forkNum, blkno + i, &mdbuf, 1);
+			{
+				void* mdbuffers[1] = { mdbuf.data };
+				mdreadv(reln, forknum, blkno, mdbuffers, 1);
+			}
 #else
-			mdread(reln, forkNum, blkno + i, mdbuf);
+			mdread(reln, forknum, blkno, mdbuf.data);
 #endif
 
-			memcpy(pageserver_masked, buffer, BLCKSZ);
-			memcpy(mdbuf_masked, mdbuf, BLCKSZ);
+			memcpy(pageserver_masked, buffers[i], BLCKSZ);
+			memcpy(mdbuf_masked.data, mdbuf.data, BLCKSZ);
 
-			if (PageIsNew((Page) mdbuf))
+			if (PageIsNew((Page) mdbuf.data))
 			{
 				if (!PageIsNew((Page) pageserver_masked))
 				{
 					neon_log(PANIC, "page is new in MD but not in Page Server at blk %u in rel %u/%u/%u fork %u (request LSN %X/%08X):\n%s\n",
 						 blkno,
 						 RelFileInfoFmt(InfoFromSMgrRel(reln)),
-						 forkNum,
+						 forknum,
 						 (uint32) (request_lsn >> 32), (uint32) request_lsn,
-						 hexdump_page(buffer));
+						 hexdump_page(buffers[i]));
 				}
 			}
-			else if (PageIsNew((Page) buffer))
+			else if (PageIsNew((Page) buffers[i]))
 			{
 				neon_log(PANIC, "page is new in Page Server but not in MD at blk %u in rel %u/%u/%u fork %u (request LSN %X/%08X):\n%s\n",
 					 blkno,
 					 RelFileInfoFmt(InfoFromSMgrRel(reln)),
-					 forkNum,
+					 forknum,
 					 (uint32) (request_lsn >> 32), (uint32) request_lsn,
-					 hexdump_page(mdbuf));
+					 hexdump_page(mdbuf.data));
 			}
-			else if (PageGetSpecialSize(mdbuf) == 0)
+			else if (PageGetSpecialSize(mdbuf.data) == 0)
 			{
 				/* assume heap */
-				RmgrTable[RM_HEAP_ID].rm_mask(mdbuf_masked, blkno);
+				RmgrTable[RM_HEAP_ID].rm_mask(mdbuf_masked.data, blkno);
 				RmgrTable[RM_HEAP_ID].rm_mask(pageserver_masked, blkno);
 
-				if (memcmp(mdbuf_masked, pageserver_masked, BLCKSZ) != 0)
+				if (memcmp(mdbuf_masked.data, pageserver_masked, BLCKSZ) != 0)
 				{
 					neon_log(PANIC, "heap buffers differ at blk %u in rel %u/%u/%u fork %u (request LSN %X/%08X):\n------ MD ------\n%s\n------ Page Server ------\n%s\n",
 						 blkno,
 						 RelFileInfoFmt(InfoFromSMgrRel(reln)),
-						 forkNum,
+						 forknum,
 						 (uint32) (request_lsn >> 32), (uint32) request_lsn,
-						 hexdump_page(mdbuf_masked),
+						 hexdump_page(mdbuf_masked.data),
 						 hexdump_page(pageserver_masked));
 				}
 			}
-			else if (PageGetSpecialSize(mdbuf) == MAXALIGN(sizeof(BTPageOpaqueData)))
+			else if (PageGetSpecialSize(mdbuf.data) == MAXALIGN(sizeof(BTPageOpaqueData)))
 			{
-				if (((BTPageOpaqueData *) PageGetSpecialPointer(mdbuf))->btpo_cycleid < MAX_BT_CYCLE_ID)
+				if (((BTPageOpaqueData *) PageGetSpecialPointer(mdbuf.data))->btpo_cycleid < MAX_BT_CYCLE_ID)
 				{
 					/* assume btree */
-					RmgrTable[RM_BTREE_ID].rm_mask(mdbuf_masked, blkno);
+					RmgrTable[RM_BTREE_ID].rm_mask(mdbuf_masked.data, blkno);
 					RmgrTable[RM_BTREE_ID].rm_mask(pageserver_masked, blkno);
 	
-					if (memcmp(mdbuf_masked, pageserver_masked, BLCKSZ) != 0)
+					if (memcmp(mdbuf_masked.data, pageserver_masked, BLCKSZ) != 0)
 					{
 						neon_log(PANIC, "btree buffers differ at blk %u in rel %u/%u/%u fork %u (request LSN %X/%08X):\n------ MD ------\n%s\n------ Page Server ------\n%s\n",
 							 blkno,
 							 RelFileInfoFmt(InfoFromSMgrRel(reln)),
-							 forkNum,
+							 forknum,
 							 (uint32) (request_lsn >> 32), (uint32) request_lsn,
-							 hexdump_page(mdbuf_masked),
+							 hexdump_page(mdbuf_masked.data),
 							 hexdump_page(pageserver_masked));
 					}
 				}
@@ -3664,6 +3677,7 @@ neon_write(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum, const vo
 	switch (reln->smgr_relpersistence)
 	{
 		case 0:
+#ifndef DEBUG_COMPARE_LOCAL
 			/* This is a bit tricky. Check if the relation exists locally */
 			if (mdexists(reln, forknum))
 			{
@@ -3682,6 +3696,7 @@ neon_write(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum, const vo
 				 */
 				return;
 			}
+#endif
 			break;
 
 		case RELPERSISTENCE_PERMANENT:
@@ -3732,6 +3747,7 @@ neon_writev(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
 	switch (reln->smgr_relpersistence)
 	{
 		case 0:
+#ifndef DEBUG_COMPARE_LOCAL
 			/* This is a bit tricky. Check if the relation exists locally */
 			if (mdexists(reln, forknum))
 			{
@@ -3747,6 +3763,7 @@ neon_writev(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
 				 */
 				return;
 			}
+#endif
 			break;
 
 		case RELPERSISTENCE_PERMANENT:
@@ -3768,7 +3785,7 @@ neon_writev(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
 
 #ifdef DEBUG_COMPARE_LOCAL
 	if (IS_LOCAL_REL(reln))
-		mdwritev(reln, forknum, blocknum, &buffer, 1, skipFsync);
+		mdwritev(reln, forknum, blkno, buffers, nblocks, skipFsync);
 #endif
 }
 

pgxn/neon/walproposer.c

@@ -57,10 +57,11 @@ static void SendProposerGreeting(Safekeeper *sk);
 static void RecvAcceptorGreeting(Safekeeper *sk);
 static void SendVoteRequest(Safekeeper *sk);
 static void RecvVoteResponse(Safekeeper *sk);
+static bool VotesCollected(WalProposer *wp);
 static void HandleElectedProposer(WalProposer *wp);
 static term_t GetHighestTerm(TermHistory *th);
-static term_t GetEpoch(Safekeeper *sk);
-static void DetermineEpochStartLsn(WalProposer *wp);
+static term_t GetLastLogTerm(Safekeeper *sk);
+static void ProcessPropStartPos(WalProposer *wp);
 static void SendProposerElected(Safekeeper *sk);
 static void StartStreaming(Safekeeper *sk);
 static void SendMessageToNode(Safekeeper *sk);
@@ -97,6 +98,7 @@ WalProposerCreate(WalProposerConfig *config, walproposer_api api)
 	wp = palloc0(sizeof(WalProposer));
 	wp->config = config;
 	wp->api = api;
+	wp->state = WPS_COLLECTING_TERMS;
 
 	wp_log(LOG, "neon.safekeepers=%s", wp->config->safekeepers_list);
 
@@ -518,7 +520,7 @@ AdvancePollState(Safekeeper *sk, uint32 events)
 			 * nodes are transferred from SS_VOTING to sending actual vote
 			 * requests.
 			 */
-		case SS_VOTING:
+		case SS_WAIT_VOTING:
 			wp_log(WARNING, "EOF from node %s:%s in %s state", sk->host,
 				   sk->port, FormatSafekeeperState(sk));
 			ResetConnection(sk);
@@ -547,7 +549,7 @@ AdvancePollState(Safekeeper *sk, uint32 events)
 			/*
 			 * Idle state for waiting votes from quorum.
 			 */
-		case SS_IDLE:
+		case SS_WAIT_ELECTED:
 			wp_log(WARNING, "EOF from node %s:%s in %s state", sk->host,
 				   sk->port, FormatSafekeeperState(sk));
 			ResetConnection(sk);
@@ -721,6 +723,15 @@ SendProposerGreeting(Safekeeper *sk)
 	BlockingWrite(sk, sk->outbuf.data, sk->outbuf.len, SS_HANDSHAKE_RECV);
 }
 
+/*
+ * Have we received greeting from enough (quorum) safekeepers to start voting?
+ */
+static bool
+TermsCollected(WalProposer *wp)
+{
+	return wp->n_connected >= wp->quorum;
+}
+
 static void
 RecvAcceptorGreeting(Safekeeper *sk)
 {
@@ -754,7 +765,7 @@ RecvAcceptorGreeting(Safekeeper *sk)
 	}
 
 	/* Protocol is all good, move to voting. */
-	sk->state = SS_VOTING;
+	sk->state = SS_WAIT_VOTING;
 
 	/*
 	 * Note: it would be better to track the counter on per safekeeper basis,
@@ -762,17 +773,18 @@ RecvAcceptorGreeting(Safekeeper *sk)
 	 * as is for now.
 	 */
 	++wp->n_connected;
-	if (wp->n_connected <= wp->quorum)
+	if (wp->state == WPS_COLLECTING_TERMS)
 	{
 		/* We're still collecting terms from the majority. */
 		wp->propTerm = Max(sk->greetResponse.term, wp->propTerm);
 
 		/* Quorum is acquried, prepare the vote request. */
-		if (wp->n_connected == wp->quorum)
+		if (TermsCollected(wp))
 		{
 			wp->propTerm++;
 			wp_log(LOG, "proposer connected to quorum (%d) safekeepers, propTerm=" INT64_FORMAT, wp->quorum, wp->propTerm);
 
+			wp->state = WPS_CAMPAIGN;
 			wp->voteRequest.pam.tag = 'v';
 			wp->voteRequest.generation = wp->mconf.generation;
 			wp->voteRequest.term = wp->propTerm;
@@ -787,12 +799,10 @@ RecvAcceptorGreeting(Safekeeper *sk)
 	}
 
 	/*
-	 * Check if we have quorum. If there aren't enough safekeepers, wait and
-	 * do nothing. We'll eventually get a task when the election starts.
-	 *
-	 * If we do have quorum, we can start an election.
+	 * If we have quorum, start (or just send vote request to newly connected
+	 * node) election, otherwise wait until we have more greetings.
 	 */
-	if (wp->n_connected < wp->quorum)
+	if (wp->state == WPS_COLLECTING_TERMS)
 	{
 		/*
 		 * SS_VOTING is an idle state; read-ready indicates the connection
@@ -807,11 +817,7 @@ RecvAcceptorGreeting(Safekeeper *sk)
 		 */
 		for (int j = 0; j < wp->n_safekeepers; j++)
 		{
-			/*
-			 * Remember: SS_VOTING indicates that the safekeeper is
-			 * participating in voting, but hasn't sent anything yet.
-			 */
-			if (wp->safekeeper[j].state == SS_VOTING)
+			if (wp->safekeeper[j].state == SS_WAIT_VOTING)
 				SendVoteRequest(&wp->safekeeper[j]);
 		}
 	}
@@ -838,6 +844,8 @@ RecvVoteResponse(Safekeeper *sk)
 {
 	WalProposer *wp = sk->wp;
 
+	Assert(wp->state >= WPS_CAMPAIGN);
+
 	sk->voteResponse.apm.tag = 'v';
 	if (!AsyncReadMessage(sk, (AcceptorProposerMessage *) &sk->voteResponse))
 		return;
@@ -856,7 +864,7 @@ RecvVoteResponse(Safekeeper *sk)
 	 * we are not elected yet and thus need the vote.
 	 */
 	if ((!sk->voteResponse.voteGiven) &&
-		(sk->voteResponse.term > wp->propTerm || wp->n_votes < wp->quorum))
+		(sk->voteResponse.term > wp->propTerm || wp->state == WPS_CAMPAIGN))
 	{
 		wp_log(FATAL, "WAL acceptor %s:%s with term " INT64_FORMAT " rejects our connection request with term " INT64_FORMAT "",
 			   sk->host, sk->port,
@@ -864,38 +872,83 @@ RecvVoteResponse(Safekeeper *sk)
 	}
 	Assert(sk->voteResponse.term == wp->propTerm);
 
-	/* Handshake completed, do we have quorum? */
+	/* ready for elected message */
+	sk->state = SS_WAIT_ELECTED;
+
 	wp->n_votes++;
-	if (wp->n_votes < wp->quorum)
+	/* Are we already elected? */
+	if (wp->state == WPS_CAMPAIGN)
 	{
-		sk->state = SS_IDLE;	/* can't do much yet, no quorum */
-	}
-	else if (wp->n_votes > wp->quorum)
-	{
-		/* already elected, start streaming */
-		SendProposerElected(sk);
+		/* no; check if this vote makes us elected */
+		if (VotesCollected(wp))
+		{
+			wp->state = WPS_ELECTED;
+			HandleElectedProposer(wp);
+		}
+		else
+		{
+			/* can't do much yet, no quorum */
+			return;
+		}
 	}
 	else
 	{
-		sk->state = SS_IDLE;
-		/* Idle state waits for read-ready events */
-		wp->api.update_event_set(sk, WL_SOCKET_READABLE);
-
-		HandleElectedProposer(sk->wp);
+		Assert(wp->state == WPS_ELECTED);
+		/* send elected only to this sk */
+		SendProposerElected(sk);
 	}
 }
 
+/*
+ * Checks if enough votes has been collected to get elected and if that's the
+ * case finds the highest vote, setting donor, donorLastLogTerm,
+ * propTermStartLsn fields. Also sets truncateLsn.
+ */
+static bool
+VotesCollected(WalProposer *wp)
+{
+	int			n_ready = 0;
+
+	/* assumed to be called only when not elected yet */
+	Assert(wp->state == WPS_CAMPAIGN);
+
+	wp->propTermStartLsn = InvalidXLogRecPtr;
+	wp->donorLastLogTerm = 0;
+	wp->truncateLsn = InvalidXLogRecPtr;
+
+	for (int i = 0; i < wp->n_safekeepers; i++)
+	{
+		if (wp->safekeeper[i].state == SS_WAIT_ELECTED)
+		{
+			n_ready++;
+
+			if (GetLastLogTerm(&wp->safekeeper[i]) > wp->donorLastLogTerm ||
+				(GetLastLogTerm(&wp->safekeeper[i]) == wp->donorLastLogTerm &&
+				 wp->safekeeper[i].voteResponse.flushLsn > wp->propTermStartLsn))
+			{
+				wp->donorLastLogTerm = GetLastLogTerm(&wp->safekeeper[i]);
+				wp->propTermStartLsn = wp->safekeeper[i].voteResponse.flushLsn;
+				wp->donor = i;
+			}
+			wp->truncateLsn = Max(wp->safekeeper[i].voteResponse.truncateLsn, wp->truncateLsn);
+		}
+	}
+
+	return n_ready >= wp->quorum;
+}
+
 /*
  * Called once a majority of acceptors have voted for us and current proposer
  * has been elected.
  *
- * Sends ProposerElected message to all acceptors in SS_IDLE state and starts
+ * Sends ProposerElected message to all acceptors in SS_WAIT_ELECTED state and starts
  * replication from walsender.
  */
 static void
 HandleElectedProposer(WalProposer *wp)
 {
-	DetermineEpochStartLsn(wp);
+	ProcessPropStartPos(wp);
+	Assert(wp->propTermStartLsn != InvalidXLogRecPtr);
 
 	/*
 	 * Synchronously download WAL from the most advanced safekeeper. We do
@@ -907,40 +960,24 @@ HandleElectedProposer(WalProposer *wp)
 		wp_log(FATAL, "failed to download WAL for logical replicaiton");
 	}
 
-	/*
-	 * Zero propEpochStartLsn means majority of safekeepers doesn't have any
-	 * WAL, timeline was just created. Compute bumps it to basebackup LSN,
-	 * otherwise we must be sync-safekeepers and we have nothing to do then.
-	 *
-	 * Proceeding is not only pointless but harmful, because we'd give
-	 * safekeepers term history starting with 0/0. These hacks will go away
-	 * once we disable implicit timeline creation on safekeepers and create it
-	 * with non zero LSN from the start.
-	 */
-	if (wp->propEpochStartLsn == InvalidXLogRecPtr)
-	{
-		Assert(wp->config->syncSafekeepers);
-		wp_log(LOG, "elected with zero propEpochStartLsn in sync-safekeepers, exiting");
-		wp->api.finish_sync_safekeepers(wp, wp->propEpochStartLsn);
-	}
-
-	if (wp->truncateLsn == wp->propEpochStartLsn && wp->config->syncSafekeepers)
+	if (wp->truncateLsn == wp->propTermStartLsn && wp->config->syncSafekeepers)
 	{
 		/* Sync is not needed: just exit */
-		wp->api.finish_sync_safekeepers(wp, wp->propEpochStartLsn);
+		wp->api.finish_sync_safekeepers(wp, wp->propTermStartLsn);
 		/* unreachable */
 	}
 
 	for (int i = 0; i < wp->n_safekeepers; i++)
 	{
-		if (wp->safekeeper[i].state == SS_IDLE)
+		if (wp->safekeeper[i].state == SS_WAIT_ELECTED)
 			SendProposerElected(&wp->safekeeper[i]);
 	}
 
 	/*
 	 * The proposer has been elected, and there will be no quorum waiting
-	 * after this point. There will be no safekeeper with state SS_IDLE also,
-	 * because that state is used only for quorum waiting.
+	 * after this point. There will be no safekeeper with state
+	 * SS_WAIT_ELECTED also, because that state is used only for quorum
+	 * waiting.
 	 */
 
 	if (wp->config->syncSafekeepers)
@@ -957,7 +994,7 @@ HandleElectedProposer(WalProposer *wp)
 		return;
 	}
 
-	wp->api.start_streaming(wp, wp->propEpochStartLsn);
+	wp->api.start_streaming(wp, wp->propTermStartLsn);
 	/* Should not return here */
 }
 
@@ -970,7 +1007,7 @@ GetHighestTerm(TermHistory *th)
 
 /* safekeeper's epoch is the term of the highest entry in the log */
 static term_t
-GetEpoch(Safekeeper *sk)
+GetLastLogTerm(Safekeeper *sk)
 {
 	return GetHighestTerm(&sk->voteResponse.termHistory);
 }
@@ -991,72 +1028,52 @@ SkipXLogPageHeader(WalProposer *wp, XLogRecPtr lsn)
 }
 
 /*
- * Called after majority of acceptors gave votes, it calculates the most
- * advanced safekeeper (who will be the donor) and epochStartLsn -- LSN since
- * which we'll write WAL in our term.
- *
- * Sets truncateLsn along the way (though it is not of much use at this point --
- * only for skipping recovery).
+ * Called after quorum gave votes and proposer starting position (highest vote
+ * term + flush LSN) -- is determined (VotesCollected true), this function
+ * adopts it: pushes LSN to shmem, sets wp term history, verifies that the
+ * basebackup matches.
  */
 static void
-DetermineEpochStartLsn(WalProposer *wp)
+ProcessPropStartPos(WalProposer *wp)
 {
 	TermHistory *dth;
-	int			n_ready = 0;
 	WalproposerShmemState *walprop_shared;
 
-	wp->propEpochStartLsn = InvalidXLogRecPtr;
-	wp->donorEpoch = 0;
-	wp->truncateLsn = InvalidXLogRecPtr;
-
-	for (int i = 0; i < wp->n_safekeepers; i++)
-	{
-		if (wp->safekeeper[i].state == SS_IDLE)
-		{
-			n_ready++;
-
-			if (GetEpoch(&wp->safekeeper[i]) > wp->donorEpoch ||
-				(GetEpoch(&wp->safekeeper[i]) == wp->donorEpoch &&
-				 wp->safekeeper[i].voteResponse.flushLsn > wp->propEpochStartLsn))
-			{
-				wp->donorEpoch = GetEpoch(&wp->safekeeper[i]);
-				wp->propEpochStartLsn = wp->safekeeper[i].voteResponse.flushLsn;
-				wp->donor = i;
-			}
-			wp->truncateLsn = Max(wp->safekeeper[i].voteResponse.truncateLsn, wp->truncateLsn);
-		}
-	}
-
-	if (n_ready < wp->quorum)
-	{
-		/*
-		 * This is a rare case that can be triggered if safekeeper has voted
-		 * and disconnected. In this case, its state will not be SS_IDLE and
-		 * its vote cannot be used, because we clean up `voteResponse` in
-		 * `ShutdownConnection`.
-		 */
-		wp_log(FATAL, "missing majority of votes, collected %d, expected %d, got %d", wp->n_votes, wp->quorum, n_ready);
-	}
+	/* must have collected votes */
+	Assert(wp->state == WPS_ELECTED);
 
 	/*
-	 * If propEpochStartLsn is 0, it means flushLsn is 0 everywhere, we are
-	 * bootstrapping and nothing was committed yet. Start streaming then from
-	 * the basebackup LSN.
+	 * If propTermStartLsn is 0, it means flushLsn is 0 everywhere, we are
+	 * bootstrapping and nothing was committed yet. Start streaming from the
+	 * basebackup LSN then.
+	 *
+	 * In case of sync-safekeepers just exit: proceeding is not only pointless
+	 * but harmful, because we'd give safekeepers term history starting with
+	 * 0/0. These hacks will go away once we disable implicit timeline
+	 * creation on safekeepers and create it with non zero LSN from the start.
 	 */
-	if (wp->propEpochStartLsn == InvalidXLogRecPtr && !wp->config->syncSafekeepers)
+	if (wp->propTermStartLsn == InvalidXLogRecPtr)
 	{
-		wp->propEpochStartLsn = wp->truncateLsn = wp->api.get_redo_start_lsn(wp);
-		wp_log(LOG, "bumped epochStartLsn to the first record %X/%X", LSN_FORMAT_ARGS(wp->propEpochStartLsn));
+		if (!wp->config->syncSafekeepers)
+		{
+			wp->propTermStartLsn = wp->truncateLsn = wp->api.get_redo_start_lsn(wp);
+			wp_log(LOG, "bumped epochStartLsn to the first record %X/%X", LSN_FORMAT_ARGS(wp->propTermStartLsn));
+		}
+		else
+		{
+			wp_log(LOG, "elected with zero propTermStartLsn in sync-safekeepers, exiting");
+			wp->api.finish_sync_safekeepers(wp, wp->propTermStartLsn);
+		}
 	}
-	pg_atomic_write_u64(&wp->api.get_shmem_state(wp)->propEpochStartLsn, wp->propEpochStartLsn);
+	pg_atomic_write_u64(&wp->api.get_shmem_state(wp)->propEpochStartLsn, wp->propTermStartLsn);
 
 	Assert(wp->truncateLsn != InvalidXLogRecPtr || wp->config->syncSafekeepers);
 
 	/*
-	 * We will be generating WAL since propEpochStartLsn, so we should set
+	 * We will be generating WAL since propTermStartLsn, so we should set
 	 * availableLsn to mark this LSN as the latest available position.
 	 */
-	wp->availableLsn = wp->propEpochStartLsn;
+	wp->availableLsn = wp->propTermStartLsn;
 
 	/*
 	 * Proposer's term history is the donor's + its own entry.
@@ -1067,12 +1084,12 @@ DetermineEpochStartLsn(WalProposer *wp)
 	if (dth->n_entries > 0)
 		memcpy(wp->propTermHistory.entries, dth->entries, sizeof(TermSwitchEntry) * dth->n_entries);
 	wp->propTermHistory.entries[wp->propTermHistory.n_entries - 1].term = wp->propTerm;
-	wp->propTermHistory.entries[wp->propTermHistory.n_entries - 1].lsn = wp->propEpochStartLsn;
+	wp->propTermHistory.entries[wp->propTermHistory.n_entries - 1].lsn = wp->propTermStartLsn;
 
 	wp_log(LOG, "got votes from majority (%d) of nodes, term " UINT64_FORMAT ", epochStartLsn %X/%X, donor %s:%s, truncate_lsn %X/%X",
 		   wp->quorum,
 		   wp->propTerm,
-		   LSN_FORMAT_ARGS(wp->propEpochStartLsn),
+		   LSN_FORMAT_ARGS(wp->propTermStartLsn),
 		   wp->safekeeper[wp->donor].host, wp->safekeeper[wp->donor].port,
 		   LSN_FORMAT_ARGS(wp->truncateLsn));
 
@@ -1090,7 +1107,7 @@ DetermineEpochStartLsn(WalProposer *wp)
 		 * Safekeepers don't skip header as they need continious stream of
 		 * data, so correct LSN for comparison.
 		 */
-		if (SkipXLogPageHeader(wp, wp->propEpochStartLsn) != wp->api.get_redo_start_lsn(wp))
+		if (SkipXLogPageHeader(wp, wp->propTermStartLsn) != wp->api.get_redo_start_lsn(wp))
 		{
 			/*
 			 * However, allow to proceed if last_log_term on the node which
@@ -1111,8 +1128,8 @@ DetermineEpochStartLsn(WalProposer *wp)
 				 */
 				disable_core_dump();
 				wp_log(PANIC,
-					   "collected propEpochStartLsn %X/%X, but basebackup LSN %X/%X",
-					   LSN_FORMAT_ARGS(wp->propEpochStartLsn),
+					   "collected propTermStartLsn %X/%X, but basebackup LSN %X/%X",
+					   LSN_FORMAT_ARGS(wp->propTermStartLsn),
 					   LSN_FORMAT_ARGS(wp->api.get_redo_start_lsn(wp)));
 			}
 		}
@@ -1623,7 +1640,7 @@ GetAcknowledgedByQuorumWALPosition(WalProposer *wp)
 		 * Like in Raft, we aren't allowed to commit entries from previous
 		 * terms, so ignore reported LSN until it gets to epochStartLsn.
 		 */
-		responses[i] = wp->safekeeper[i].appendResponse.flushLsn >= wp->propEpochStartLsn ? wp->safekeeper[i].appendResponse.flushLsn : 0;
+		responses[i] = wp->safekeeper[i].appendResponse.flushLsn >= wp->propTermStartLsn ? wp->safekeeper[i].appendResponse.flushLsn : 0;
 	}
 	qsort(responses, wp->n_safekeepers, sizeof(XLogRecPtr), CompareLsn);
 
@@ -1656,10 +1673,10 @@ UpdateDonorShmem(WalProposer *wp)
 	 * about its position immediately after election before any feedbacks are
 	 * sent.
 	 */
-	if (wp->safekeeper[wp->donor].state >= SS_IDLE)
+	if (wp->safekeeper[wp->donor].state >= SS_WAIT_ELECTED)
 	{
 		donor = &wp->safekeeper[wp->donor];
-		donor_lsn = wp->propEpochStartLsn;
+		donor_lsn = wp->propTermStartLsn;
 	}
 
 	/*
@@ -1748,7 +1765,7 @@ HandleSafekeeperResponse(WalProposer *wp, Safekeeper *fromsk)
 		for (int i = 0; i < wp->n_safekeepers; i++)
 		{
 			Safekeeper *sk = &wp->safekeeper[i];
-			bool		synced = sk->appendResponse.commitLsn >= wp->propEpochStartLsn;
+			bool		synced = sk->appendResponse.commitLsn >= wp->propTermStartLsn;
 
 			/* alive safekeeper which is not synced yet; wait for it */
 			if (sk->state != SS_OFFLINE && !synced)
@@ -1772,7 +1789,7 @@ HandleSafekeeperResponse(WalProposer *wp, Safekeeper *fromsk)
 			 */
 			BroadcastAppendRequest(wp);
 
-			wp->api.finish_sync_safekeepers(wp, wp->propEpochStartLsn);
+			wp->api.finish_sync_safekeepers(wp, wp->propTermStartLsn);
 			/* unreachable */
 		}
 	}
@@ -2378,7 +2395,7 @@ FormatSafekeeperState(Safekeeper *sk)
 		case SS_HANDSHAKE_RECV:
 			return_val = "handshake (receiving)";
 			break;
-		case SS_VOTING:
+		case SS_WAIT_VOTING:
 			return_val = "voting";
 			break;
 		case SS_WAIT_VERDICT:
@@ -2387,7 +2404,7 @@ FormatSafekeeperState(Safekeeper *sk)
 		case SS_SEND_ELECTED_FLUSH:
 			return_val = "send-announcement-flush";
 			break;
-		case SS_IDLE:
+		case SS_WAIT_ELECTED:
 			return_val = "idle";
 			break;
 		case SS_ACTIVE:
@@ -2476,8 +2493,8 @@ SafekeeperStateDesiredEvents(Safekeeper *sk, uint32 *sk_events, uint32 *nwr_even
 			 * Idle states use read-readiness as a sign that the connection
 			 * has been disconnected.
 			 */
-		case SS_VOTING:
-		case SS_IDLE:
+		case SS_WAIT_VOTING:
+		case SS_WAIT_ELECTED:
 			*sk_events = WL_SOCKET_READABLE;
 			return;