compute_ctl: log ttid everywhere.

Allows to easily find compute logs by timeline id, and now loki agent can make a label from it. pairs with https://github.com/neondatabase/cloud/pull/7259
tests: prettier subprocess output in test log (#5485 )
2026-01-24 13:50:37 +00:00 · 2023-10-06 01:02:41 +03:00 · 2023-10-05 20:15:55 +00:00 · 2023-10-05 18:48:01 +01:00 · 2023-10-05 14:02:20 +00:00 · 2023-10-05 11:22:05 +01:00
189 changed files with 13338 additions and 5966 deletions
--- a/.github/actions/allure-report-generate/action.yml
+++ b/.github/actions/allure-report-generate/action.yml
@@ -76,8 +76,8 @@ runs:
          rm -f ${ALLURE_ZIP}
        fi
      env:
-        ALLURE_VERSION: 2.23.1
-        ALLURE_ZIP_SHA256: 11141bfe727504b3fd80c0f9801eb317407fd0ac983ebb57e671f14bac4bcd86
+        ALLURE_VERSION: 2.24.0
+        ALLURE_ZIP_SHA256: 60b1d6ce65d9ef24b23cf9c2c19fd736a123487c38e54759f1ed1a7a77353c90

    # Potentially we could have several running build for the same key (for example, for the main branch), so we use improvised lock for this
    - name: Acquire lock
--- a/.github/workflows/approved-for-ci-run.yml
+++ b/.github/workflows/approved-for-ci-run.yml
@@ -16,21 +16,29 @@ on:
      # Actual magic happens here:
      - labeled

+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number }}
+
 env:
  GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
  PR_NUMBER: ${{ github.event.pull_request.number }}
  BRANCH: "ci-run/pr-${{ github.event.pull_request.number }}"

-permissions: write-all
+# No permission for GITHUB_TOKEN by default; the **minimal required** set of permissions should be granted in each job.
+permissions: {}

-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number }}
+defaults:
+  run:
+    shell: bash -euo pipefail {0}

 jobs:
  remove-label:
    # Remove `approved-for-ci-run` label if the workflow is triggered by changes in a PR.
    # The PR should be reviewed and labelled manually again.

+    permissions:
+      pull-requests: write # For `gh pr edit`
+
    if: |
      contains(fromJSON('["opened", "synchronize", "reopened", "closed"]'), github.event.action) &&
      contains(github.event.pull_request.labels.*.name, 'approved-for-ci-run')
@@ -43,6 +51,10 @@ jobs:
  create-or-update-pr-for-ci-run:
    # Create local PR for an `approved-for-ci-run` labelled PR to run CI pipeline in it.

+    permissions:
+      pull-requests: write # for `gh pr edit`
+      # For `git push` and `gh pr create` we use CI_ACCESS_TOKEN
+
    if: |
      github.event.action == 'labeled' &&
      contains(github.event.pull_request.labels.*.name, 'approved-for-ci-run')
@@ -75,7 +87,7 @@ jobs:
            Feel free to review/comment/discuss the original PR #${PR_NUMBER}.
          EOF

-          ALREADY_CREATED="$(gh pr --repo ${GITHUB_REPOSITORY} list --head ${HEAD} --base main --json number --jq '.[].number')"
+          ALREADY_CREATED="$(gh pr --repo ${GITHUB_REPOSITORY} list --head ${BRANCH} --base main --json number --jq '.[].number')"
          if [ -z "${ALREADY_CREATED}" ]; then
            gh pr --repo "${GITHUB_REPOSITORY}" create --title "CI run for PR #${PR_NUMBER}" \
                                                       --body-file "body.md" \
@@ -87,6 +99,10 @@ jobs:
  cleanup:
    # Close PRs and delete branchs if the original PR is closed.

+    permissions:
+      contents: write # for `--delete-branch` flag in `gh pr close`
+      pull-requests: write # for `gh pr close`
+
    if: |
      github.event.action == 'closed' &&
      github.event.pull_request.head.repo.full_name != github.repository
@@ -94,8 +110,9 @@ jobs:
    runs-on: ubuntu-latest

    steps:
-      - run: |
-          CLOSED="$(gh pr --repo ${GITHUB_REPOSITORY} list --head ${HEAD} --json 'closed' --jq '.[].closed')"
+      - name: Close PR and delete `ci-run/pr-${{ env.PR_NUMBER }}` branch
+        run: |
+          CLOSED="$(gh pr --repo ${GITHUB_REPOSITORY} list --head ${BRANCH} --json 'closed' --jq '.[].closed')"
          if [ "${CLOSED}" == "false" ]; then
            gh pr --repo "${GITHUB_REPOSITORY}" close "${BRANCH}" --delete-branch
          fi
--- a/.github/workflows/build_and_test.yml
+++ b/.github/workflows/build_and_test.yml
@@ -834,7 +834,7 @@ jobs:
      run:
        shell: sh -eu {0}
    env:
-      VM_BUILDER_VERSION: v0.17.5
+      VM_BUILDER_VERSION: v0.17.12

    steps:
      - name: Checkout
@@ -1091,7 +1091,10 @@ jobs:
          GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
        run: |
          if [[ "$GITHUB_REF_NAME" == "main" ]]; then
-            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}}
+            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=false
+
+            # TODO: move deployPreprodRegion to release (`"$GITHUB_REF_NAME" == "release"` block), once Staging support different compute tag prefixes for different regions
+            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=true
          elif [[ "$GITHUB_REF_NAME" == "release" ]]; then
            gh workflow --repo neondatabase/aws run deploy-prod.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f disclamerAcknowledged=true
          else
--- a/2
+++ b/2
@@ -5,7 +5,7 @@
 /libs/remote_storage/ @neondatabase/storage
 /libs/safekeeper_api/ @neondatabase/safekeepers
 /libs/vm_monitor/ @neondatabase/autoscaling @neondatabase/compute
-/pageserver/ @neondatabase/compute @neondatabase/storage
+/pageserver/ @neondatabase/storage
 /pgxn/ @neondatabase/compute
 /proxy/ @neondatabase/proxy
 /safekeeper/ @neondatabase/safekeepers
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -636,7 +636,7 @@ dependencies = [
 "sha1",
 "sync_wrapper",
 "tokio",
- "tokio-tungstenite 0.20.0",
+ "tokio-tungstenite",
 "tower",
 "tower-layer",
 "tower-service",
@@ -798,6 +798,22 @@ dependencies = [
 "either",
 ]

+[[package]]
+name = "camino"
+version = "1.1.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c59e92b5a388f549b863a7bea62612c09f24c8393560709a54558a9abdfb3b9c"
+
+[[package]]
+name = "camino-tempfile"
+version = "1.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d2ab15a83d13f75dbd86f082bdefd160b628476ef58d3b900a0ef74e001bb097"
+dependencies = [
+ "camino",
+ "tempfile",
+]
+
 [[package]]
 name = "cast"
 version = "0.3.0"
@@ -1053,6 +1069,7 @@ name = "control_plane"
 version = "0.1.0"
 dependencies = [
 "anyhow",
+ "camino",
 "clap",
 "comfy-table",
 "compute_api",
@@ -1780,18 +1797,9 @@ checksum = "95505c38b4572b2d910cecb0281560f54b440a19336cbbcb27bf6ce6adc6f5a8"

 [[package]]
 name = "hermit-abi"
-version = "0.2.6"
+version = "0.3.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ee512640fe35acbfb4bb779db6f0d80704c2cacfa2e39b601ef3e3f47d1ae4c7"
-dependencies = [
- "libc",
-]
-
-[[package]]
-name = "hermit-abi"
-version = "0.3.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fed44880c466736ef9a5c5b5facefb5ed0785676d0c02d612db14e54f0d84286"
+checksum = "d77f7ec81a6d05a3abb01ab6eb7590f6083d08449fe5a1c8b1e620283546ccb7"

 [[package]]
 name = "hex"
@@ -1941,15 +1949,15 @@ dependencies = [

 [[package]]
 name = "hyper-tungstenite"
-version = "0.9.0"
+version = "0.11.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "880b8b1c98a5ec2a505c7c90db6d3f6f1f480af5655d9c5b55facc9382a5a5b5"
+checksum = "7cc7dcb1ab67cd336f468a12491765672e61a3b6b148634dbfe2fe8acd3fe7d9"
 dependencies = [
 "hyper",
- "pin-project",
+ "pin-project-lite",
 "tokio",
- "tokio-tungstenite 0.18.0",
- "tungstenite 0.18.0",
+ "tokio-tungstenite",
+ "tungstenite",
 ]

 [[package]]
@@ -2053,7 +2061,7 @@ version = "1.0.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "eae7b9aee968036d54dce06cebaefd919e4472e753296daccd6d344e3e2df0c2"
 dependencies = [
- "hermit-abi 0.3.1",
+ "hermit-abi",
 "libc",
 "windows-sys 0.48.0",
 ]
@@ -2070,7 +2078,7 @@ version = "0.4.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "adcf93614601c8129ddf72e2d5633df827ba6551541c6d8c59520a371475be1f"
 dependencies = [
- "hermit-abi 0.3.1",
+ "hermit-abi",
 "io-lifetimes",
 "rustix 0.37.19",
 "windows-sys 0.48.0",
@@ -2444,11 +2452,11 @@ dependencies = [

 [[package]]
 name = "num_cpus"
-version = "1.15.0"
+version = "1.16.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0fac9e2da13b5eb447a6ce3d392f23a29d8694bff781bf03a16cd9ac8697593b"
+checksum = "4161fcb6d602d4d2081af7c3a45852d875a03dd337a6bfdd6e06407b61342a43"
 dependencies = [
- "hermit-abi 0.2.6",
+ "hermit-abi",
 "libc",
 ]

@@ -2650,6 +2658,7 @@ version = "0.1.0"
 dependencies = [
 "anyhow",
 "bytes",
+ "camino",
 "clap",
 "git-version",
 "pageserver",
@@ -2670,6 +2679,8 @@ dependencies = [
 "async-trait",
 "byteorder",
 "bytes",
+ "camino",
+ "camino-tempfile",
 "chrono",
 "clap",
 "close_fds",
@@ -2721,7 +2732,6 @@ dependencies = [
 "strum_macros",
 "svg_fmt",
 "sync_wrapper",
- "tempfile",
 "tenant_size_model",
 "thiserror",
 "tokio",
@@ -2908,9 +2918,9 @@ dependencies = [

 [[package]]
 name = "pin-project-lite"
-version = "0.2.9"
+version = "0.2.13"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e0a7ae3ac2f1173085d398531c705756c94a4c56843785df85a60c1a0afac116"
+checksum = "8afb450f006bf6385ca15ef45d71d2288452bc3683ce2e2cacc0d18e4be60b58"

 [[package]]
 name = "pin-utils"
@@ -3246,6 +3256,7 @@ dependencies = [
 "reqwest-tracing",
 "routerify",
 "rstest",
+ "rustc-hash",
 "rustls",
 "rustls-pemfile",
 "scopeguard",
@@ -3413,14 +3424,16 @@ dependencies = [
 "aws-sdk-s3",
 "aws-smithy-http",
 "aws-types",
+ "camino",
+ "camino-tempfile",
 "hyper",
 "metrics",
 "once_cell",
 "pin-project-lite",
+ "rand",
 "scopeguard",
 "serde",
 "serde_json",
- "tempfile",
 "test-context",
 "tokio",
 "tokio-util",
@@ -3772,6 +3785,8 @@ dependencies = [
 "async-trait",
 "byteorder",
 "bytes",
+ "camino",
+ "camino-tempfile",
 "chrono",
 "clap",
 "const_format",
@@ -3800,7 +3815,6 @@ dependencies = [
 "serde_with",
 "signal-hook",
 "storage_broker",
- "tempfile",
 "thiserror",
 "tokio",
 "tokio-io-timeout",
@@ -4641,18 +4655,6 @@ dependencies = [
 "xattr",
 ]

-[[package]]
-name = "tokio-tungstenite"
-version = "0.18.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "54319c93411147bced34cb5609a80e0a8e44c5999c93903a81cd866630ec0bfd"
-dependencies = [
- "futures-util",
- "log",
- "tokio",
- "tungstenite 0.18.0",
-]
-
 [[package]]
 name = "tokio-tungstenite"
 version = "0.20.0"
@@ -4662,7 +4664,7 @@ dependencies = [
 "futures-util",
 "log",
 "tokio",
- "tungstenite 0.20.0",
+ "tungstenite",
 ]

 [[package]]
@@ -4977,28 +4979,9 @@ checksum = "3528ecfd12c466c6f163363caf2d02a71161dd5e1cc6ae7b34207ea2d42d81ed"

 [[package]]
 name = "tungstenite"
-version = "0.18.0"
+version = "0.20.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "30ee6ab729cd4cf0fd55218530c4522ed30b7b6081752839b68fcec8d0960788"
-dependencies = [
- "base64 0.13.1",
- "byteorder",
- "bytes",
- "http",
- "httparse",
- "log",
- "rand",
- "sha1",
- "thiserror",
- "url",
- "utf-8",
-]
-
-[[package]]
-name = "tungstenite"
-version = "0.20.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e862a1c4128df0112ab625f55cd5c934bcb4312ba80b39ae4b4835a3fd58e649"
+checksum = "9e3dac10fd62eaf6617d3a904ae222845979aec67c615d1c842b4002c7666fb9"
 dependencies = [
 "byteorder",
 "bytes",
@@ -5130,6 +5113,8 @@ dependencies = [
 "bincode",
 "byteorder",
 "bytes",
+ "camino",
+ "camino-tempfile",
 "chrono",
 "const_format",
 "criterion",
@@ -5155,7 +5140,6 @@ dependencies = [
 "signal-hook",
 "strum",
 "strum_macros",
- "tempfile",
 "thiserror",
 "tokio",
 "tokio-stream",
@@ -5229,6 +5213,7 @@ name = "wal_craft"
 version = "0.1.0"
 dependencies = [
 "anyhow",
+ "camino-tempfile",
 "clap",
 "env_logger",
 "log",
@@ -5236,7 +5221,6 @@ dependencies = [
 "postgres",
 "postgres_ffi",
 "regex",
- "tempfile",
 "utils",
 "workspace_hack",
 ]
@@ -5648,6 +5632,7 @@ dependencies = [
 "tower",
 "tracing",
 "tracing-core",
+ "tungstenite",
 "url",
 "uuid",
 ]
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -51,6 +51,7 @@ bindgen = "0.65"
 bstr = "1.0"
 byteorder = "1.4"
 bytes = "1.0"
+camino = "1.1.6"
 cfg-if = "1.0.0"
 chrono = { version = "0.4", default-features = false, features = ["clock"] }
 clap = { version = "4.0", features = ["derive"] }
@@ -78,7 +79,7 @@ hostname = "0.3.1"
 humantime = "2.1"
 humantime-serde = "1.1.1"
 hyper = "0.14"
-hyper-tungstenite = "0.9"
+hyper-tungstenite = "0.11"
 inotify = "0.10.2"
 itertools = "0.10"
 jsonwebtoken = "8"
@@ -107,6 +108,7 @@ reqwest-middleware = "0.2.0"
 reqwest-retry = "0.2.2"
 routerify = "3"
 rpds = "0.13"
+rustc-hash = "1.1.0"
 rustls = "0.21"
 rustls-pemfile = "1"
 rustls-split = "0.3"
@@ -186,7 +188,7 @@ workspace_hack = { version = "0.1", path = "./workspace_hack/" }
 criterion = "0.5.1"
 rcgen = "0.11"
 rstest = "0.18"
-tempfile = "3.4"
+camino-tempfile = "1.0.2"
 tonic-build = "0.9"

 [patch.crates-io]
--- a/Dockerfile.compute-node
+++ b/Dockerfile.compute-node
@@ -589,8 +589,7 @@ RUN case "${PG_VERSION}" in \
    echo "${PG_EMBEDDING_CHECKSUM} pg_embedding.tar.gz" | sha256sum --check && \
    mkdir pg_embedding-src && cd pg_embedding-src && tar xvzf ../pg_embedding.tar.gz --strip-components=1 -C . && \
    make -j $(getconf _NPROCESSORS_ONLN) && \
-    make -j $(getconf _NPROCESSORS_ONLN) install && \
-    echo 'trusted = true' >> /usr/local/pgsql/share/extension/embedding.control
+    make -j $(getconf _NPROCESSORS_ONLN) install

 #########################################################################################
 #
@@ -616,11 +615,7 @@ RUN wget https://gitlab.com/dalibo/postgresql_anonymizer/-/archive/1.1.0/postgre
 #########################################################################################
 #
 # Layer "rust extensions"
-# This layer is used to build `pgx` deps
-#
-# FIXME: This needs to be updated to latest version of 'pgrx' (it was renamed from
-# 'pgx' to 'pgrx') for PostgreSQL 16. And that in turn requires bumping the pgx
-# dependency on all the rust extension that depend on it, too.
+# This layer is used to build `pgrx` deps
 #
 #########################################################################################
 FROM build-deps AS rust-extensions-build
@@ -636,22 +631,12 @@ USER nonroot
 WORKDIR /home/nonroot
 ARG PG_VERSION

-RUN case "${PG_VERSION}" in \
-      "v14" | "v15") \
-        ;; \
-      "v16") \
-        echo "TODO: Not yet supported for PostgreSQL 16. Need to update pgrx dependencies" && exit 0 \
-	;; \
-      *) \
-        echo "unexpected PostgreSQL version ${PG_VERSION}" && exit 1 \
-        ;; \
-    esac && \
-    curl -sSO https://static.rust-lang.org/rustup/dist/$(uname -m)-unknown-linux-gnu/rustup-init && \
+RUN curl -sSO https://static.rust-lang.org/rustup/dist/$(uname -m)-unknown-linux-gnu/rustup-init && \
    chmod +x rustup-init && \
    ./rustup-init -y --no-modify-path --profile minimal --default-toolchain stable && \
    rm rustup-init && \
-    cargo install --locked --version 0.7.3 cargo-pgx && \
-    /bin/bash -c 'cargo pgx init --pg${PG_VERSION:1}=/usr/local/pgsql/bin/pg_config'
+    cargo install --locked --version 0.10.2 cargo-pgrx && \
+    /bin/bash -c 'cargo pgrx init --pg${PG_VERSION:1}=/usr/local/pgsql/bin/pg_config'

 USER root

@@ -665,23 +650,11 @@ USER root
 FROM rust-extensions-build AS pg-jsonschema-pg-build
 ARG PG_VERSION

-# caeab60d70b2fd3ae421ec66466a3abbb37b7ee6 made on 06/03/2023
-# there is no release tag yet, but we need it due to the superuser fix in the control file, switch to git tag after release >= 0.1.5
-RUN case "${PG_VERSION}" in \
-      "v14" | "v15") \
-        ;; \
-      "v16") \
-        echo "TODO: Not yet supported for PostgreSQL 16. Need to update pgrx dependencies" && exit 0 \
-	;; \
-      *) \
-        echo "unexpected PostgreSQL version \"${PG_VERSION}\"" && exit 1 \
-        ;; \
-    esac && \
-    wget https://github.com/supabase/pg_jsonschema/archive/caeab60d70b2fd3ae421ec66466a3abbb37b7ee6.tar.gz -O pg_jsonschema.tar.gz && \
-    echo "54129ce2e7ee7a585648dbb4cef6d73f795d94fe72f248ac01119992518469a4 pg_jsonschema.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/supabase/pg_jsonschema/archive/refs/tags/v0.2.0.tar.gz -O pg_jsonschema.tar.gz && \
+    echo "9118fc508a6e231e7a39acaa6f066fcd79af17a5db757b47d2eefbe14f7794f0 pg_jsonschema.tar.gz" | sha256sum --check && \
    mkdir pg_jsonschema-src && cd pg_jsonschema-src && tar xvzf ../pg_jsonschema.tar.gz --strip-components=1 -C . && \
-    sed -i 's/pgx = "0.7.1"/pgx = { version = "0.7.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
-    cargo pgx install --release && \
+    sed -i 's/pgrx = "0.10.2"/pgrx = { version = "0.10.2", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
+    cargo pgrx install --release && \
    echo "trusted = true" >> /usr/local/pgsql/share/extension/pg_jsonschema.control

 #########################################################################################
@@ -694,26 +667,11 @@ RUN case "${PG_VERSION}" in \
 FROM rust-extensions-build AS pg-graphql-pg-build
 ARG PG_VERSION

-# b4988843647450a153439be367168ed09971af85 made on 22/02/2023 (from remove-pgx-contrib-spiext branch)
-# Currently pgx version bump to >= 0.7.2  causes "call to unsafe function" compliation errors in
-# pgx-contrib-spiext. There is a branch that removes that dependency, so use it. It is on the
-# same 1.1 version we've used before.
-RUN case "${PG_VERSION}" in \
-      "v14" | "v15") \
-        ;; \
-      "v16") \
-        echo "TODO: Not yet supported for PostgreSQL 16. Need to update pgrx dependencies" && exit 0 \
-	;; \
-      *) \
-        echo "unexpected PostgreSQL version" && exit 1 \
-        ;; \
-    esac && \
-    wget https://github.com/yrashk/pg_graphql/archive/b4988843647450a153439be367168ed09971af85.tar.gz -O pg_graphql.tar.gz && \
-    echo "0c7b0e746441b2ec24187d0e03555faf935c2159e2839bddd14df6dafbc8c9bd pg_graphql.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/supabase/pg_graphql/archive/refs/tags/v1.4.0.tar.gz -O pg_graphql.tar.gz && \
+    echo "bd8dc7230282b3efa9ae5baf053a54151ed0e66881c7c53750e2d0c765776edc pg_graphql.tar.gz" | sha256sum --check && \
    mkdir pg_graphql-src && cd pg_graphql-src && tar xvzf ../pg_graphql.tar.gz --strip-components=1 -C . && \
-    sed -i 's/pgx = "~0.7.1"/pgx = { version = "0.7.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
-    sed -i 's/pgx-tests = "~0.7.1"/pgx-tests = "0.7.3"/g' Cargo.toml && \
-    cargo pgx install --release && \
+    sed -i 's/pgrx = "=0.10.2"/pgrx = { version = "0.10.2", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
+    cargo pgrx install --release && \
    # it's needed to enable extension because it uses untrusted C language
    sed -i 's/superuser = false/superuser = true/g' /usr/local/pgsql/share/extension/pg_graphql.control && \
    echo "trusted = true" >> /usr/local/pgsql/share/extension/pg_graphql.control
@@ -728,21 +686,11 @@ RUN case "${PG_VERSION}" in \
 FROM rust-extensions-build AS pg-tiktoken-pg-build
 ARG PG_VERSION

-# 801f84f08c6881c8aa30f405fafbf00eec386a72 made on 10/03/2023
-RUN case "${PG_VERSION}" in \
-      "v14" | "v15") \
-        ;; \
-      "v16") \
-        echo "TODO: Not yet supported for PostgreSQL 16. Need to update pgrx dependencies" && exit 0 \
-	;; \
-      *) \
-        echo "unexpected PostgreSQL version" && exit 1 \
-        ;; \
-    esac && \
-    wget https://github.com/kelvich/pg_tiktoken/archive/801f84f08c6881c8aa30f405fafbf00eec386a72.tar.gz -O pg_tiktoken.tar.gz && \
-    echo "52f60ac800993a49aa8c609961842b611b6b1949717b69ce2ec9117117e16e4a pg_tiktoken.tar.gz" | sha256sum --check && \
+# 26806147b17b60763039c6a6878884c41a262318 made on 26/09/2023
+RUN wget https://github.com/kelvich/pg_tiktoken/archive/26806147b17b60763039c6a6878884c41a262318.tar.gz -O pg_tiktoken.tar.gz && \
+    echo "e64e55aaa38c259512d3e27c572da22c4637418cf124caba904cd50944e5004e pg_tiktoken.tar.gz" | sha256sum --check && \
    mkdir pg_tiktoken-src && cd pg_tiktoken-src && tar xvzf ../pg_tiktoken.tar.gz --strip-components=1 -C . && \
-    cargo pgx install --release && \
+    cargo pgrx install --release && \
    echo "trusted = true" >> /usr/local/pgsql/share/extension/pg_tiktoken.control

 #########################################################################################
@@ -755,21 +703,15 @@ RUN case "${PG_VERSION}" in \
 FROM rust-extensions-build AS pg-pgx-ulid-build
 ARG PG_VERSION

-RUN case "${PG_VERSION}" in \
-      "v14" | "v15") \
-        ;; \
-      "v16") \
-        echo "TODO: Not yet supported for PostgreSQL 16. Need to update pgrx dependencies" && exit 0 \
-	;; \
-      *) \
-        echo "unexpected PostgreSQL version" && exit 1 \
-        ;; \
-    esac && \
-    wget https://github.com/pksunkara/pgx_ulid/archive/refs/tags/v0.1.0.tar.gz -O pgx_ulid.tar.gz && \
-    echo "908b7358e6f846e87db508ae5349fb56a88ee6305519074b12f3d5b0ff09f791 pgx_ulid.tar.gz" | sha256sum --check && \
+RUN wget https://github.com/pksunkara/pgx_ulid/archive/refs/tags/v0.1.3.tar.gz -O pgx_ulid.tar.gz && \
+    echo "ee5db82945d2d9f2d15597a80cf32de9dca67b897f605beb830561705f12683c pgx_ulid.tar.gz" | sha256sum --check && \
    mkdir pgx_ulid-src && cd pgx_ulid-src && tar xvzf ../pgx_ulid.tar.gz --strip-components=1 -C . && \
-    sed -i 's/pgx        = "=0.7.3"/pgx = { version = "0.7.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
-    cargo pgx install --release && \
+    echo "******************* Apply a patch for Postgres 16 support; delete in the next release ******************" && \
+    wget https://github.com/pksunkara/pgx_ulid/commit/f84954cf63fc8c80d964ac970d9eceed3c791196.patch && \
+    patch -p1 < f84954cf63fc8c80d964ac970d9eceed3c791196.patch && \
+    echo "********************************************************************************************************" && \
+    sed -i 's/pgrx       = "=0.10.2"/pgrx = { version = "=0.10.2", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
+    cargo pgrx install --release && \
    echo "trusted = true" >> /usr/local/pgsql/share/extension/ulid.control

 #########################################################################################
--- a/15
+++ b/15
@@ -153,18 +153,6 @@ neon-pg-ext-%: postgres-%
 		-C $(POSTGRES_INSTALL_DIR)/build/neon-utils-$* \
 		-f $(ROOT_PROJECT_DIR)/pgxn/neon_utils/Makefile install

-# pg_embedding was temporarily released as hnsw from this repo, when we only
-# supported PostgreSQL 14 and 15
-neon-pg-ext-v14: neon-pg-ext-hnsw-v14
-neon-pg-ext-v15: neon-pg-ext-hnsw-v15
-
-neon-pg-ext-hnsw-%: postgres-headers-% postgres-%
-	+@echo "Compiling hnsw $*"
-	mkdir -p $(POSTGRES_INSTALL_DIR)/build/hnsw-$*
-	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config CFLAGS='$(PG_CFLAGS) $(COPT)' \
-		-C $(POSTGRES_INSTALL_DIR)/build/hnsw-$* \
-		-f $(ROOT_PROJECT_DIR)/pgxn/hnsw/Makefile install
-
 .PHONY: neon-pg-ext-clean-%
 neon-pg-ext-clean-%:
 	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config \
@@ -179,9 +167,6 @@ neon-pg-ext-clean-%:
 	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config \
 	-C $(POSTGRES_INSTALL_DIR)/build/neon-utils-$* \
 	-f $(ROOT_PROJECT_DIR)/pgxn/neon_utils/Makefile clean
-	$(MAKE) PG_CONFIG=$(POSTGRES_INSTALL_DIR)/$*/bin/pg_config \
-	-C $(POSTGRES_INSTALL_DIR)/build/hnsw-$* \
-	-f $(ROOT_PROJECT_DIR)/pgxn/hnsw/Makefile clean

 .PHONY: neon-pg-ext
 neon-pg-ext: \
--- a/README.md
+++ b/README.md
@@ -29,13 +29,13 @@ See developer documentation in [SUMMARY.md](/docs/SUMMARY.md) for more informati
 ```bash
 apt install build-essential libtool libreadline-dev zlib1g-dev flex bison libseccomp-dev \
 libssl-dev clang pkg-config libpq-dev cmake postgresql-client protobuf-compiler \
-libcurl4-openssl-dev openssl python-poetry lsof
+libcurl4-openssl-dev openssl python-poetry lsof libicu-dev
 ```
 * On Fedora, these packages are needed:
 ```bash
 dnf install flex bison readline-devel zlib-devel openssl-devel \
  libseccomp-devel perl clang cmake postgresql postgresql-contrib protobuf-compiler \
-  protobuf-devel libcurl-devel openssl poetry lsof
+  protobuf-devel libcurl-devel openssl poetry lsof libicu-devel
 ```
 * On Arch based systems, these packages are needed:
 ```bash
--- a/compute_tools/src/bin/compute_ctl.rs
+++ b/compute_tools/src/bin/compute_ctl.rs
@@ -44,7 +44,7 @@ use std::{thread, time::Duration};
 use anyhow::{Context, Result};
 use chrono::Utc;
 use clap::Arg;
-use tracing::{error, info};
+use tracing::{error, info, info_span};
 use url::Url;

 use compute_api::responses::ComputeStatus;
@@ -57,6 +57,7 @@ use compute_tools::logger::*;
 use compute_tools::monitor::launch_monitor;
 use compute_tools::params::*;
 use compute_tools::spec::*;
+use utils::id::TenantTimelineId;

 // this is an arbitrary build tag. Fine as a default / for testing purposes
 // in-case of not-set environment var
@@ -249,11 +250,20 @@ fn main() -> Result<()> {

    state.status = ComputeStatus::Init;
    compute.state_changed.notify_all();
+    let pspec = state.pspec.as_ref().expect("spec must be set");
+    let ttid = TenantTimelineId {
+        tenant_id: pspec.tenant_id,
+        timeline_id: pspec.timeline_id,
+    };
    drop(state);

+    // Log ttid everywhere for easier log identification (e.g. loki agent can
+    // create label on that).
+    let _guard = info_span!("", ttid = %ttid).entered();
+
    // Launch remaining service threads
-    let _monitor_handle = launch_monitor(&compute);
-    let _configurator_handle = launch_configurator(&compute);
+    let _monitor_handle = launch_monitor(&compute, ttid);
+    let _configurator_handle = launch_configurator(&compute, ttid);

    // Start Postgres
    let mut delay_exit = false;
--- a/compute_tools/src/compute.rs
+++ b/compute_tools/src/compute.rs
@@ -1039,7 +1039,7 @@ LIMIT 100",
        let remote_extensions = spec
            .remote_extensions
            .as_ref()
-            .ok_or(anyhow::anyhow!("Remote extensions are not configured",))?;
+            .ok_or(anyhow::anyhow!("Remote extensions are not configured"))?;

        info!("parse shared_preload_libraries from spec.cluster.settings");
        let mut libs_vec = Vec::new();
--- a/compute_tools/src/configurator.rs
+++ b/compute_tools/src/configurator.rs
@@ -4,11 +4,13 @@ use std::thread;
 use tracing::{error, info, instrument};

 use compute_api::responses::ComputeStatus;
+use utils::id::TenantTimelineId;

 use crate::compute::ComputeNode;

-#[instrument(skip_all)]
-fn configurator_main_loop(compute: &Arc<ComputeNode>) {
+// Log ttid everywhere
+#[instrument(name = "", fields(ttid = %ttid), skip_all)]
+fn configurator_main_loop(compute: &Arc<ComputeNode>, ttid: TenantTimelineId) {
    info!("waiting for reconfiguration requests");
    loop {
        let state = compute.state.lock().unwrap();
@@ -41,13 +43,16 @@ fn configurator_main_loop(compute: &Arc<ComputeNode>) {
    }
 }

-pub fn launch_configurator(compute: &Arc<ComputeNode>) -> thread::JoinHandle<()> {
+pub fn launch_configurator(
+    compute: &Arc<ComputeNode>,
+    ttid: TenantTimelineId,
+) -> thread::JoinHandle<()> {
    let compute = Arc::clone(compute);

    thread::Builder::new()
        .name("compute-configurator".into())
        .spawn(move || {
-            configurator_main_loop(&compute);
+            configurator_main_loop(&compute, ttid);
            info!("configurator thread is exited");
        })
        .expect("cannot launch configurator thread")
--- a/compute_tools/src/monitor.rs
+++ b/compute_tools/src/monitor.rs
@@ -1,29 +1,30 @@
 use std::sync::Arc;
-use std::{thread, time};
+use std::{thread, time::Duration};

 use chrono::{DateTime, Utc};
 use postgres::{Client, NoTls};
-use tracing::{debug, info};
+use tracing::{debug, info, instrument};
+use utils::id::TenantTimelineId;

 use crate::compute::ComputeNode;

-const MONITOR_CHECK_INTERVAL: u64 = 500; // milliseconds
+const MONITOR_CHECK_INTERVAL: Duration = Duration::from_millis(500);

 // Spin in a loop and figure out the last activity time in the Postgres.
 // Then update it in the shared state. This function never errors out.
 // XXX: the only expected panic is at `RwLock` unwrap().
-fn watch_compute_activity(compute: &ComputeNode) {
+#[instrument(name = "", fields(ttid = %ttid), skip_all)]
+fn watch_compute_activity(compute: &ComputeNode, ttid: TenantTimelineId) {
    // Suppose that `connstr` doesn't change
    let connstr = compute.connstr.as_str();
    // Define `client` outside of the loop to reuse existing connection if it's active.
    let mut client = Client::connect(connstr, NoTls);
-    let timeout = time::Duration::from_millis(MONITOR_CHECK_INTERVAL);

    info!("watching Postgres activity at {}", connstr);

    loop {
        // Should be outside of the write lock to allow others to read while we sleep.
-        thread::sleep(timeout);
+        thread::sleep(MONITOR_CHECK_INTERVAL);

        match &mut client {
            Ok(cli) => {
@@ -104,11 +105,11 @@ fn watch_compute_activity(compute: &ComputeNode) {
 }

 /// Launch a separate compute monitor thread and return its `JoinHandle`.
-pub fn launch_monitor(state: &Arc<ComputeNode>) -> thread::JoinHandle<()> {
+pub fn launch_monitor(state: &Arc<ComputeNode>, ttid: TenantTimelineId) -> thread::JoinHandle<()> {
    let state = Arc::clone(state);

    thread::Builder::new()
        .name("compute-monitor".into())
-        .spawn(move || watch_compute_activity(&state))
+        .spawn(move || watch_compute_activity(&state, ttid))
        .expect("cannot launch compute monitor thread")
 }
--- a/control_plane/Cargo.toml
+++ b/control_plane/Cargo.toml
@@ -6,6 +6,7 @@ license.workspace = true

 [dependencies]
 anyhow.workspace = true
+camino.workspace = true
 clap.workspace = true
 comfy-table.workspace = true
 git-version.workspace = true
--- a/control_plane/src/attachment_service.rs
+++ b/control_plane/src/attachment_service.rs
@@ -1,5 +1,6 @@
 use crate::{background_process, local_env::LocalEnv};
 use anyhow::anyhow;
+use camino::Utf8PathBuf;
 use serde::{Deserialize, Serialize};
 use serde_with::{serde_as, DisplayFromStr};
 use std::{path::PathBuf, process::Child};
@@ -47,8 +48,9 @@ impl AttachmentService {
        }
    }

-    fn pid_file(&self) -> PathBuf {
-        self.env.base_data_dir.join("attachment_service.pid")
+    fn pid_file(&self) -> Utf8PathBuf {
+        Utf8PathBuf::from_path_buf(self.env.base_data_dir.join("attachment_service.pid"))
+            .expect("non-Unicode path")
    }

    pub fn start(&self) -> anyhow::Result<Child> {
--- a/control_plane/src/background_process.rs
+++ b/control_plane/src/background_process.rs
@@ -16,12 +16,13 @@ use std::ffi::OsStr;
 use std::io::Write;
 use std::os::unix::prelude::AsRawFd;
 use std::os::unix::process::CommandExt;
-use std::path::{Path, PathBuf};
+use std::path::Path;
 use std::process::{Child, Command};
 use std::time::Duration;
 use std::{fs, io, thread};

 use anyhow::Context;
+use camino::{Utf8Path, Utf8PathBuf};
 use nix::errno::Errno;
 use nix::fcntl::{FcntlArg, FdFlag};
 use nix::sys::signal::{kill, Signal};
@@ -45,9 +46,9 @@ const NOTICE_AFTER_RETRIES: u64 = 50;
 /// it itself.
 pub enum InitialPidFile<'t> {
    /// Create a pidfile, to allow future CLI invocations to manipulate the process.
-    Create(&'t Path),
+    Create(&'t Utf8Path),
    /// The process will create the pidfile itself, need to wait for that event.
-    Expect(&'t Path),
+    Expect(&'t Utf8Path),
 }

 /// Start a background child process using the parameters given.
@@ -137,7 +138,11 @@ where
 }

 /// Stops the process, using the pid file given. Returns Ok also if the process is already not running.
-pub fn stop_process(immediate: bool, process_name: &str, pid_file: &Path) -> anyhow::Result<()> {
+pub fn stop_process(
+    immediate: bool,
+    process_name: &str,
+    pid_file: &Utf8Path,
+) -> anyhow::Result<()> {
    let pid = match pid_file::read(pid_file)
        .with_context(|| format!("read pid_file {pid_file:?}"))?
    {
@@ -252,9 +257,9 @@ fn fill_aws_secrets_vars(mut cmd: &mut Command) -> &mut Command {
 ///    will remain held until the cmd exits.
 fn pre_exec_create_pidfile<P>(cmd: &mut Command, path: P) -> &mut Command
 where
-    P: Into<PathBuf>,
+    P: Into<Utf8PathBuf>,
 {
-    let path: PathBuf = path.into();
+    let path: Utf8PathBuf = path.into();
    // SAFETY
    // pre_exec is marked unsafe because it runs between fork and exec.
    // Why is that dangerous in various ways?
@@ -311,7 +316,7 @@ where

 fn process_started<F>(
    pid: Pid,
-    pid_file_to_check: Option<&Path>,
+    pid_file_to_check: Option<&Utf8Path>,
    status_check: &F,
 ) -> anyhow::Result<bool>
 where
--- a/control_plane/src/bin/attachment_service.rs
+++ b/control_plane/src/bin/attachment_service.rs
@@ -223,6 +223,7 @@ async fn handle_attach_hook(mut req: Request<Body>) -> Result<Response<Body>, Ap
    if attach_req.pageserver_id.is_some() {
        tenant_state.generation += 1;
    }
+    tenant_state.pageserver = attach_req.pageserver_id;
    let generation = tenant_state.generation;

    locked.save().await.map_err(ApiError::InternalServerError)?;
--- a/control_plane/src/broker.rs
+++ b/control_plane/src/broker.rs
@@ -7,7 +7,7 @@
 //! ```
 use anyhow::Context;

-use std::path::PathBuf;
+use camino::Utf8PathBuf;

 use crate::{background_process, local_env};

@@ -30,7 +30,7 @@ pub fn start_broker_process(env: &local_env::LocalEnv) -> anyhow::Result<()> {
        || {
            let url = broker.client_url();
            let status_url = url.join("status").with_context(|| {
-                format!("Failed to append /status path to broker endpoint {url}",)
+                format!("Failed to append /status path to broker endpoint {url}")
            })?;
            let request = client
                .get(status_url)
@@ -50,6 +50,7 @@ pub fn stop_broker_process(env: &local_env::LocalEnv) -> anyhow::Result<()> {
    background_process::stop_process(true, "storage_broker", &storage_broker_pid_file_path(env))
 }

-fn storage_broker_pid_file_path(env: &local_env::LocalEnv) -> PathBuf {
-    env.base_data_dir.join("storage_broker.pid")
+fn storage_broker_pid_file_path(env: &local_env::LocalEnv) -> Utf8PathBuf {
+    Utf8PathBuf::from_path_buf(env.base_data_dir.join("storage_broker.pid"))
+        .expect("non-Unicode path")
 }
--- a/control_plane/src/pageserver.rs
+++ b/control_plane/src/pageserver.rs
@@ -14,6 +14,7 @@ use std::process::{Child, Command};
 use std::{io, result};

 use anyhow::{bail, Context};
+use camino::Utf8PathBuf;
 use pageserver_api::models::{self, TenantInfo, TimelineInfo};
 use postgres_backend::AuthType;
 use postgres_connection::{parse_host_port, PgConnectionConfig};
@@ -144,7 +145,7 @@ impl PageServerNode {
    pub fn initialize(&self, config_overrides: &[&str]) -> anyhow::Result<()> {
        // First, run `pageserver --init` and wait for it to write a config into FS and exit.
        self.pageserver_init(config_overrides)
-            .with_context(|| format!("Failed to run init for pageserver node {}", self.conf.id,))
+            .with_context(|| format!("Failed to run init for pageserver node {}", self.conf.id))
    }

    pub fn repo_path(&self) -> PathBuf {
@@ -154,8 +155,9 @@ impl PageServerNode {
    /// The pid file is created by the pageserver process, with its pid stored inside.
    /// Other pageservers cannot lock the same file and overwrite it for as long as the current
    /// pageserver runs. (Unless someone removes the file manually; never do that!)
-    fn pid_file(&self) -> PathBuf {
-        self.repo_path().join("pageserver.pid")
+    fn pid_file(&self) -> Utf8PathBuf {
+        Utf8PathBuf::from_path_buf(self.repo_path().join("pageserver.pid"))
+            .expect("non-Unicode path")
    }

    pub fn start(&self, config_overrides: &[&str]) -> anyhow::Result<Child> {
--- a/control_plane/src/safekeeper.rs
+++ b/control_plane/src/safekeeper.rs
@@ -11,6 +11,7 @@ use std::process::Child;
 use std::{io, result};

 use anyhow::Context;
+use camino::Utf8PathBuf;
 use postgres_connection::PgConnectionConfig;
 use reqwest::blocking::{Client, RequestBuilder, Response};
 use reqwest::{IntoUrl, Method};
@@ -97,8 +98,9 @@ impl SafekeeperNode {
        SafekeeperNode::datadir_path_by_id(&self.env, self.id)
    }

-    pub fn pid_file(&self) -> PathBuf {
-        self.datadir_path().join("safekeeper.pid")
+    pub fn pid_file(&self) -> Utf8PathBuf {
+        Utf8PathBuf::from_path_buf(self.datadir_path().join("safekeeper.pid"))
+            .expect("non-Unicode path")
    }

    pub fn start(&self, extra_opts: Vec<String>) -> anyhow::Result<Child> {
--- a/deny.toml
+++ b/deny.toml
@@ -23,7 +23,7 @@ vulnerability = "deny"
 unmaintained = "warn"
 yanked = "warn"
 notice = "warn"
-ignore = ["RUSTSEC-2023-0052"]
+ignore = []

 # This section is considered when running `cargo deny check licenses`
 # More documentation for the licenses section can be found here:
--- a/docs/rfcs/028-pageserver-migration.md
+++ b/docs/rfcs/028-pageserver-migration.md
@@ -0,0 +1,599 @@
+# Seamless tenant migration
+
+- Author: john@neon.tech
+- Created on 2023-08-11
+- Implemented on ..
+
+## Summary
+
+The preceding [generation numbers RFC](025-generation-numbers.md) may be thought of as "making tenant
+migration safe". Following that,
+this RFC is about how those migrations are to be done:
+
+1. Seamlessly (without interruption to client availability)
+2. Quickly (enabling faster operations)
+3. Efficiently (minimizing I/O and $ cost)
+
+These points are in priority order: if we have to sacrifice
+efficiency to make a migration seamless for clients, we will
+do so, etc.
+
+This is accomplished by introducing two high level changes:
+
+- A dual-attached state for tenants, used in a control-plane-orchestrated
+  migration procedure that preserves availability during a migration.
+- Warm secondary locations for tenants, where on-disk content is primed
+  for a fast migration of the tenant from its current attachment to this
+  secondary location.
+
+## Motivation
+
+Migrating tenants between pageservers is essential to operating a service
+at scale, in several contexts:
+
+1. Responding to a pageserver node failure by migrating tenants to other pageservers
+2. Balancing load and capacity across pageservers, for example when a user expands their
+   database and they need to migrate to a pageserver with more capacity.
+3. Restarting pageservers for upgrades and maintenance
+
+The current situation steps for migration are:
+
+- detach from old node; skip if old node is dead; (the [skip part is still WIP](https://github.com/neondatabase/cloud/issues/5426)).
+- attach to new node
+- re-configure endpoints to use the new node
+
+Once [generation numbers](025-generation-numbers.md) are implemented,
+the detach step is no longer critical for correctness. So, we can
+
+- attach to a new node,
+- re-configure endpoints to use the new node, and then
+- detach from the old node.
+
+However, this still does not meet our seamless/fast/efficient goals:
+
+- Not fast: The new node will have to download potentially large amounts
+  of data from S3, which may take many minutes.
+- Not seamless: If we attach to a new pageserver before detaching an old one,
+  the new one might delete some objects that interrupt availability of reads on the old one.
+- Not efficient: the old pageserver will continue uploading
+  S3 content during the migration that will never be read.
+
+The user expectations for availability are:
+
+- For planned maintenance, there should be zero availability
+  gap. This expectation is fulfilled by this RFC.
+- For unplanned changes (e.g. node failures), there should be
+  minimal availability gap. This RFC provides the _mechanism_
+  to fail over quickly, but does not provide the failure _detection_
+  nor failover _policy_.
+
+## Non Goals
+
+- Defining service tiers with different storage strategies: the same
+  level of HA & overhead will apply to all tenants. This doesn't rule out
+  adding such tiers in future.
+- Enabling pageserver failover in the absence of a control plane: the control
+  plane will remain the source of truth for what should be attached where.
+- Totally avoiding availability gaps on unplanned migrations during
+  a failure (we expect a small, bounded window of
+  read unavailability of very recent LSNs)
+- Workload balancing: this RFC defines the mechanism for moving tenants
+  around, not the higher level logic for deciding who goes where.
+- Defining all possible configuration flows for tenants: the migration process
+  defined in this RFC demonstrates the sufficiency of the pageserver API, but
+  is not the only kind of configuration change the control plane will ever do.
+  The APIs defined here should let the control plane move tenants around in
+  whatever way is needed while preserving data safety and read availability.
+
+## Impacted components
+
+Pageserver, control plane
+
+## Terminology
+
+- **Attachment**: a tenant is _attached_ to a pageserver if it has
+  been issued a generation number, and is running an instance of
+  the `Tenant` type, ingesting the WAL, and available to serve
+  page reads.
+- **Location**: locations are a superset of attachments. A location
+  is a combination of a tenant and a pageserver. We may _attach_ at a _location_.
+
+- **Secondary location**: a location which is not currently attached.
+- **Warm secondary location**: a location which is not currently attached, but is endeavoring to maintain a warm local cache of layers. We avoid calling this a _warm standby_ to avoid confusion with similar postgres features.
+
+## Implementation (high level)
+
+### Warm secondary locations
+
+To enable faster migrations, we will identify at least one _secondary location_
+for each tenant. This secondary location will keep a warm cache of layers
+for the tenant, so that if it is later attached, it can catch up with the
+latest LSN quickly: rather than downloading everything, it only has to replay
+the recent part of the WAL to advance from the remote_consistent_offset to the
+most recent LSN in the WAL.
+
+The control plane is responsible for selecting secondary locations, and
+calling into pageservers to configure tenants into a secondary mode at this
+new location, as well as attaching the tenant in its existing primary location.
+
+The attached pageserver for a tenant will publish a [layer heatmap](#layer-heatmap)
+to advise secondaries of which layers should be downloaded.
+
+### Location modes
+
+Currently, we consider a tenant to be in one of two states on a pageserver:
+
+- Attached: active `Tenant` object, and layers on local disk
+- Detached: no layers on local disk, no runtime state.
+
+We will extend this with finer-grained modes, whose purpose will become
+clear in later sections:
+
+- **AttachedSingle**: equivalent the existing attached state.
+- **AttachedMulti**: like AttachedSingle, holds an up to date generation, but
+  does not do deletions.
+- **AttachedStale**: like AttachedSingle, holds a stale generation,
+  do not do any remote storage operations.
+- **Secondary**: keep local state on disk, periodically update from S3.
+- **Detached**: equivalent to existing detached state.
+
+To control these finer grained states, a new pageserver API endpoint will be added.
+
+### Cutover procedure
+
+Define old location and new location as "Node A" and "Node B". Consider
+the case where both nodes are available, and Node B was previously configured
+as a secondary location for the tenant we are migrating.
+
+The cutover procedure is orchestrated by the control plane, calling into
+the pageservers' APIs:
+
+1. Call to Node A requesting it to flush to S3 and enter AttachedStale state
+2. Increment generation, and call to Node B requesting it to enter AttachedMulti
+   state with the new generation.
+3. Call to Node B, requesting it to download the latest hot layers from remote storage,
+   according to the latest heatmap flushed by Node A.
+4. Wait for Node B's WAL ingestion to catch up with node A's
+5. Update endpoints to use node B instead of node A
+6. Call to node B requesting it to enter state AttachedSingle.
+7. Call to node A requesting it to enter state Secondary
+
+The following table summarizes how the state of the system advances:
+
+|     Step      |     Node A     |     Node B     | Node used by endpoints |
+| :-----------: | :------------: | :------------: | :--------------------: |
+| 1 (_initial_) | AttachedSingle |   Secondary    |           A            |
+|       2       | AttachedStale  | AttachedMulti  |           A            |
+|       3       | AttachedStale  | AttachedMulti  |           A            |
+|       4       | AttachedStale  | AttachedMulti  |           A            |
+| 5 (_cutover_) | AttachedStale  | AttachedMulti  |           B            |
+|       6       | AttachedStale  | AttachedSingle |           B            |
+|  7 (_final_)  |   Secondary    | AttachedSingle |           B            |
+
+The procedure described for a clean handover from a live node to a secondary
+is also used for failure cases and for migrations to a location that is not
+configured as a secondary, by simply skipping irrelevant steps, as described in
+the following sections.
+
+#### Migration from an unresponsive node
+
+If node A is unavailable, then all calls into
+node A are skipped and we don't wait for B to catch up before
+switching updating the endpoints to use B.
+
+#### Migration to a location that is not a secondary
+
+If node B is initially in Detached state, the procedure is identical. Since Node B
+is coming from a Detached state rather than Secondary, the download of layers and
+catch up with WAL will take much longer.
+
+We might do this if:
+
+- Attached and secondary locations are both critically low on disk, and we need
+  to migrate to a third node with more resources available.
+- We are migrating a tenant which does not use secondary locations to save on cost.
+
+#### Permanent migration away from a node
+
+In the final step of the migration, we generally request the original node to enter a Secondary
+state. This is typical if we are doing a planned migration during maintenance, or to
+balance CPU/network load away from a node.
+
+One might also want to permanently migrate away: this can be done by simply removing the secondary
+location after the migration is complete, or as an optimization by substituting the Detached state
+for the Secondary state in the final step.
+
+#### Cutover diagram
+
+```mermaid
+sequenceDiagram
+participant CP as Control plane
+participant A as Node A
+participant B as Node B
+participant E as Endpoint
+
+CP->>A: PUT Flush & go to AttachedStale
+note right of A: A continues to ingest WAL
+CP->>B: PUT AttachedMulti
+CP->>B: PUT Download layers from latest heatmap
+note right of B: B downloads from S3
+loop Poll until download complete
+CP->>B: GET download status
+end
+activate B
+note right of B: B ingests WAL
+loop Poll until catch up
+CP->>B: GET visible WAL
+CP->>A: GET visible WAL
+end
+deactivate B
+CP->>E: Configure to use Node B
+E->>B: Connect for reads
+CP->>B: PUT AttachedSingle
+CP->>A: PUT Secondary
+```
+
+#### Cutover from an unavailable pageserver
+
+This case is far simpler: we may skip straight to our intended
+end state.
+
+```mermaid
+sequenceDiagram
+participant A as Node A
+participant CP as Control plane
+participant B as Node B
+participant E as Endpoint
+
+note right of A: Node A offline
+activate A
+CP->>B: PUT AttachedSingle
+CP->>E: Configure to use Node B
+E->>B: Connect for reads
+deactivate A
+```
+
+## Implementation (detail)
+
+### Purpose of AttachedMulti, AttachedStale
+
+#### AttachedMulti
+
+Ordinarily, an attached pageserver whose generation is the latest may delete
+layers at will (e.g. during compaction). If a previous generation pageserver
+is also still attached, and in use by endpoints, then this layer deletion could
+lead to a loss of availability for the endpoint when reading from the previous
+generation pageserver.
+
+The _AttachedMulti_ state simply disables deletions. These will be enqueued
+in `RemoteTimelineClient` until the control plane transitions the
+node into AttachedSingle, which unblocks deletions.  Other remote storage operations
+such as uploads are not blocked.
+
+AttachedMulti is not required for data safety, only to preserve availability
+on pageservers running with stale generations.
+
+A node enters AttachedMulti only when explicitly asked to by the control plane. It should
+only remain in this state for the duration of a migration.
+
+If a control plane bug leaves
+the node in AttachedMulti for a long time, then we must avoid unbounded memory use from enqueued
+deletions. This may be accomplished simply, by dropping enqueued deletions when some modest
+threshold of delayed deletions (e.g. 10k layers per tenant) is reached. As with all deletions,
+it is safe to skip them, and the leaked objects will be eventually cleaned up by scrub or
+by timeline deletion.
+
+During AttachedMulti, the Tenant is free to drop layers from local disk in response to
+disk pressure: only the deletion of remote layers is blocked.
+
+#### AttachedStale
+
+Currently, a pageserver with a stale generation number will continue to
+upload layers, but be prevented from completing deletions. This is safe, but inefficient: layers uploaded by this stale generation
+will not be read back by future generations of pageservers.
+
+The _AttachedStale_ state disables S3 uploads. The stale pageserver
+will continue to ingest the WAL and write layers to local disk, but not to
+do any uploads to S3.
+
+A node may enter AttachedStale in two ways:
+
+- Explicitly, when control plane calls into the node at the start of a migration.
+- Implicitly, when the node tries to validate some deletions and discovers
+  that its generation is stale.
+
+The AttachedStale state also disables sending consumption metrics from
+that location: it is interpreted as an indication that some other pageserver
+is already attached or is about to be attached, and that new pageserver will
+be responsible for sending consumption metrics.
+
+#### Disk Pressure & AttachedStale
+
+Over long periods of time, a tenant location in AttachedStale will accumulate data
+on local disk, as it cannot evict any layers written since it entered the
+AttachStale state. We rely on the control plane to revert the location to
+Secondary or Detached at the end of a migration.
+
+This scenario is particularly noteworthy when evacuating all tenants on a pageserver:
+since _all_ the attached tenants will go into AttachedStale, we will be doing no
+uploads at all, therefore ingested data will cause disk usage to increase continuously.
+Under nominal conditions, the available disk space on pageservers should be sufficient
+to complete the evacuation before this becomes a problem, but we must also handle
+the case where we hit a low disk situation while in this state.
+
+The concept of disk pressure already exists in the pageserver: the `disk_usage_eviction_task`
+touches each Tenant when it determines that a low-disk condition requires
+some layer eviction. Having selected layers for eviction, the eviction
+task calls `Timeline::evict_layers`.
+
+**Safety**: If evict_layers is called while in AttachedStale state, and some of the to-be-evicted
+layers are not yet uploaded to S3, then the block on uploads will be lifted. This
+will result in leaking some objects once a migration is complete, but will enable
+the node to manage its disk space properly: if a node is left with some tenants
+in AttachedStale indefinitely due to a network partition or control plane bug,
+these tenants will not cause a full disk condition.
+
+### Warm secondary updates
+
+#### Layer heatmap
+
+The secondary location's job is to serve reads **with the same quality of service as the original location
+was serving them around the time of a migration**. This does not mean the secondary
+location needs the whole set of layers: inactive layers that might soon
+be evicted on the attached pageserver need not be downloaded by the
+secondary. A totally idle tenant only needs to maintain enough on-disk
+state to enable a fast cold start (i.e. the most recent image layers are
+typically sufficient).
+
+To enable this, we introduce the concept of a _layer heatmap_, which
+acts as an advisory input to secondary locations to decide which
+layers to download from S3.
+
+#### Attached pageserver
+
+The attached pageserver, if in state AttachedSingle, periodically
+uploads a serialized heat map to S3. It may skip this if there
+is no change since the last time it uploaded (e.g. if the tenant
+is totally idle).
+
+Additionally, when the tenant is flushed to remote storage prior to a migration
+(the first step in [cutover procedure](#cutover-procedure)), 
+the heatmap is written out. This enables a future attached pageserver
+to get an up to date view when deciding which layers to download.
+
+#### Secondary location behavior
+
+Secondary warm locations run a simple loop, implemented separately from
+the main `Tenant` type, which represents attached tenants:
+
+- Download the layer heatmap
+- Select any "hot enough" layers to download, if there is sufficient
+  free disk space.
+- Download layers, if they were not previously evicted (see below)
+- Download the latest index_part.json
+- Check if any layers currently on disk are no longer referenced by
+  IndexPart & delete them
+
+Note that the heatmap is only advisory: if a secondary location has plenty
+of disk space, it may choose to retain layers that aren't referenced
+by the heatmap, as long as they are still referenced by the IndexPart. Conversely,
+if a node is very low on disk space, it might opt to raise the heat threshold required
+to both downloading a layer, until more disk space is available.
+
+#### Secondary locations & disk pressure
+
+Secondary locations are subject to eviction on disk pressure, just as
+attached locations are.  For eviction purposes, the access time of a
+layer in a secondary location will be the access time given in the heatmap,
+rather than the literal time at which the local layer file was accessed.
+
+The heatmap will indicate which layers are in local storage on the attached
+location.  The secondary will always attempt to get back to having that
+set of layers on disk, but to avoid flapping, it will remember the access
+time of the layer it was most recently asked to evict, and layers whose
+access time is below that will not be re-downloaded.
+
+The resulting behavior is that after a layer is evicted from a secondary
+location, it is only re-downloaded once the attached pageserver accesses
+the layer and uploads a heatmap reflecting that access time.  On a pageserver
+restart, the secondary location will attempt to download all layers in
+the heatmap again, if they are not on local disk.
+
+This behavior will be slightly different when secondary locations are
+used for "low energy tenants", but that is beyond the scope of this RFC.
+
+### Location configuration API
+
+Currently, the `/tenant/<tenant_id>/config` API defines various
+tunables like compaction settings, which apply to the tenant irrespective
+of which pageserver it is running on.
+
+A new "location config" structure will be introduced, which defines
+configuration which is per-tenant, but local to a particular pageserver,
+such as the attachment mode and whether it is a secondary.
+
+The pageserver will expose a new per-tenant API for setting
+the state: `/tenant/<tenant_id>/location/config`.
+
+Body content:
+
+```
+{
+  state: 'enum{Detached, Secondary, AttachedSingle, AttachedMulti, AttachedStale}',
+  generation: Option<u32>,
+  configuration: `Option<TenantConfig>`
+  flush: bool
+}
+```
+
+Existing `/attach` and `/detach` endpoint will have the same
+behavior as calling `/location/config` with `AttachedSingle` and `Detached`
+states respectively. These endpoints will be deprecated and later
+removed.
+
+The generation attribute is mandatory for entering `AttachedSingle` or
+`AttachedMulti`.
+
+The configuration attribute is mandatory when entering any state other
+than `Detached`. This configuration is the same as the body for
+the existing `/tenant/<tenant_id>/config` endpoint.
+
+The `flush` argument indicates whether the pageservers should flush
+to S3 before proceeding: this only has any effect if the node is
+currently in AttachedSingle or AttachedMulti. This is used
+during the first phase of migration, when transitioning the
+old pageserver to AttachedSingle.
+
+The `/re-attach` API response will be extended to include a `state` as
+well as a `generation`, enabling the pageserver to enter the
+correct state for each tenant on startup.
+
+### Database schema for locations
+
+A new table `ProjectLocation`:
+
+- pageserver_id: int
+- tenant_id: TenantId
+- generation: Option<int>
+- state: `enum(Secondary, AttachedSingle, AttachedMulti)`
+
+Notes:
+
+- It is legacy for a Project to have zero `ProjectLocation`s
+- The `pageserver` column in `Project` now means "to which pageserver should
+  endpoints connect", rather than simply which pageserver is attached.
+- The `generation` column in `Project` remains, and is incremented and used
+  to set the generation of `ProjectLocation` rows when they are set into
+  an attached state.
+- The `Detached` state is implicitly represented as the absence of
+  a `ProjectLocation`.
+
+### Executing migrations
+
+Migrations will be implemented as Go functions, within the
+existing `Operation` framework in the control plane. These
+operations are persistent, such that they will always keep
+trying until completion: this property is important to avoid
+leaving garbage behind on pageservers, such as AttachedStale
+locations.
+
+### Recovery from failures during migration
+
+During migration, the control plane may encounter failures of either
+the original or new pageserver, or both:
+
+- If the original fails, skip past waiting for the new pageserver
+  to catch up, and put it into AttachedSingle immediately.
+- If the new node fails, put the old pageserver into Secondary
+  and then back into AttachedSingle (this has the effect of
+  retaining on-disk state and granting it a fresh generation number).
+- If both nodes fail, keep trying until one of them is available
+  again.
+
+### Control plane -> Pageserver reconciliation
+
+A migration may be done while the old node is unavailable,
+in which case the old node may still be running in an AttachedStale
+state.
+
+In this case, it is undesirable to have the migration `Operation`
+stay alive until the old node eventually comes back online
+and can be cleaned up. To handle this, the control plane
+should run a background reconciliation process to compare
+a pageserver's attachments with the database, and clean up
+any that shouldn't be there any more.
+
+Note that there will be no work to do if the old node was really
+offline, as during startup it will call into `/re-attach` and
+be updated that way. The reconciliation will only be needed
+if the node was unavailable but still running.
+
+## Alternatives considered
+
+### Only enabling secondary locations for tenants on a higher service tier
+
+This will make sense in future, especially for tiny databases that may be
+downloaded from S3 in milliseconds when needed.
+
+However, it is not wise to do it immediately, because pageservers contain
+a mixture of higher and lower tier workloads. If we had 1 tenant with
+a secondary location and 9 without, then those other 9 tenants will do
+a lot of I/O as they try to recover from S3, which may degrade the
+service of the tenant which had a secondary location.
+
+Until we segregate tenant on different service tiers on different pageserver
+nodes, or implement & test QoS to ensure that tenants with secondaries are
+not harmed by tenants without, we should use the same failover approach
+for all the tenants.
+
+### Hot secondary locations (continuous WAL replay)
+
+Instead of secondary locations populating their caches from S3, we could
+have them consume the WAL from safekeepers. The downsides of this would be:
+
+- Double load on safekeepers, which are a less scalable service than S3
+- Secondary locations' on-disk state would end up subtly different to
+  the remote state, which would make synchronizing with S3 more complex/expensive
+  when going into attached state.
+
+The downside of only updating secondary locations from S3 is that we will
+have a delay during migration from replaying the LSN range between what's
+in S3 and what's in the pageserver. This range will be very small on
+planned migrations, as we have the old pageserver flush to S3 immediately
+before attaching the new pageserver. On unplanned migrations (old pageserver
+is unavailable), the range of LSNs to replay is bounded by the flush frequency
+on the old pageserver. However, the migration doesn't have to wait for the
+replay: it's just that not-yet-replayed LSNs will be unavailable for read
+until the new pageserver catches up.
+
+We expect that pageserver reads of the most recent LSNs will be relatively
+rare, as for an active endpoint those pages will usually still be in the postgres
+page cache: this leads us to prefer synchronizing from S3 on secondary
+locations, rather than consuming the WAL from safekeepers.
+
+### Cold secondary locations
+
+It is not functionally necessary to keep warm caches on secondary locations at all. However, if we do not, then
+we would experience a de-facto availability loss in unplanned migrations, as reads to the new node would take an extremely long time (many seconds, perhaps minutes).
+
+Warm caches on secondary locations are necessary to meet
+our availability goals.
+
+### Pageserver-granularity failover
+
+Instead of migrating tenants individually, we could have entire spare nodes,
+and on a node death, move all its work to one of these spares.
+
+This approach is avoided for several reasons:
+
+- we would still need fine-grained tenant migration for other
+  purposes such as balancing load
+- by sharing the spare capacity over many peers rather than one spare node,
+  these peers may use the capacity for other purposes, until it is needed
+  to handle migrated tenants. e.g. for keeping a deeper cache of their
+  attached tenants.
+
+### Readonly during migration
+
+We could simplify migrations by making both previous and new nodes go into a
+readonly state, then flush remote content from the previous node, then activate
+attachment on the secondary node.
+
+The downside to this approach is a potentially large gap in readability of
+recent LSNs while loading data onto the new node. To avoid this, it is worthwhile
+to incur the extra cost of double-replaying the WAL onto old and new nodes' local
+storage during a migration.
+
+### Peer-to-peer pageserver communication
+
+Rather than uploading the heatmap to S3, attached pageservers could make it
+available to peers.
+
+Currently, pageservers have no peer to peer communication, so adding this
+for heatmaps would incur significant overhead in deployment and configuration
+of the service, and ensuring that when a new pageserver is deployed, other
+pageservers are updated to be aware of it.
+
+As well as simplifying implementation, putting heatmaps in S3 will be useful
+for future analytics purposes -- gathering aggregated statistics on activity
+pattersn across many tenants may be done directly from data in S3.
--- a/libs/consumption_metrics/src/lib.rs
+++ b/libs/consumption_metrics/src/lib.rs
@@ -3,9 +3,9 @@
 //!
 use chrono::{DateTime, Utc};
 use rand::Rng;
-use serde::Serialize;
+use serde::{Deserialize, Serialize};

-#[derive(Serialize, Debug, Clone, Copy, Eq, PartialEq, Ord, PartialOrd)]
+#[derive(Serialize, serde::Deserialize, Debug, Clone, Copy, Eq, PartialEq, Ord, PartialOrd)]
 #[serde(tag = "type")]
 pub enum EventType {
    #[serde(rename = "absolute")]
@@ -27,7 +27,8 @@ impl EventType {
    }

    pub fn incremental_timerange(&self) -> Option<std::ops::Range<&DateTime<Utc>>> {
-        // these can most likely be thought of as Range or RangeFull
+        // these can most likely be thought of as Range or RangeFull, at least pageserver creates
+        // incremental ranges where the stop and next start are equal.
        use EventType::*;
        match self {
            Incremental {
@@ -41,15 +42,25 @@ impl EventType {
    pub fn is_incremental(&self) -> bool {
        matches!(self, EventType::Incremental { .. })
    }
+
+    /// Returns the absolute time, or for incremental ranges, the stop time.
+    pub fn recorded_at(&self) -> &DateTime<Utc> {
+        use EventType::*;
+
+        match self {
+            Absolute { time } => time,
+            Incremental { stop_time, .. } => stop_time,
+        }
+    }
 }

-#[derive(Serialize, Debug, Clone, Eq, PartialEq, Ord, PartialOrd)]
-pub struct Event<Extra> {
+#[derive(Serialize, Deserialize, Debug, Clone, Eq, PartialEq, Ord, PartialOrd)]
+pub struct Event<Extra, Metric> {
    #[serde(flatten)]
    #[serde(rename = "type")]
    pub kind: EventType,

-    pub metric: &'static str,
+    pub metric: Metric,
    pub idempotency_key: String,
    pub value: u64,

@@ -58,19 +69,45 @@ pub struct Event<Extra> {
 }

 pub fn idempotency_key(node_id: &str) -> String {
-    format!(
-        "{}-{}-{:04}",
-        Utc::now(),
-        node_id,
-        rand::thread_rng().gen_range(0..=9999)
-    )
+    IdempotencyKey::generate(node_id).to_string()
+}
+
+/// Downstream users will use these to detect upload retries.
+pub struct IdempotencyKey<'a> {
+    now: chrono::DateTime<Utc>,
+    node_id: &'a str,
+    nonce: u16,
+}
+
+impl std::fmt::Display for IdempotencyKey<'_> {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{}-{}-{:04}", self.now, self.node_id, self.nonce)
+    }
+}
+
+impl<'a> IdempotencyKey<'a> {
+    pub fn generate(node_id: &'a str) -> Self {
+        IdempotencyKey {
+            now: Utc::now(),
+            node_id,
+            nonce: rand::thread_rng().gen_range(0..=9999),
+        }
+    }
+
+    pub fn for_tests(now: DateTime<Utc>, node_id: &'a str, nonce: u16) -> Self {
+        IdempotencyKey {
+            now,
+            node_id,
+            nonce,
+        }
+    }
 }

 pub const CHUNK_SIZE: usize = 1000;

 // Just a wrapper around a slice of events
 // to serialize it as `{"events" : [ ] }
-#[derive(serde::Serialize)]
+#[derive(serde::Serialize, serde::Deserialize)]
 pub struct EventChunk<'a, T: Clone> {
    pub events: std::borrow::Cow<'a, [T]>,
 }
--- a/libs/pageserver_api/src/models.rs
+++ b/libs/pageserver_api/src/models.rs
@@ -10,6 +10,7 @@ use serde_with::{serde_as, DisplayFromStr};
 use strum_macros;
 use utils::{
    completion,
+    generation::Generation,
    history_buffer::HistoryBufferWithDropCounter,
    id::{NodeId, TenantId, TimelineId},
    lsn::Lsn,
@@ -218,6 +219,8 @@ impl std::ops::Deref for TenantCreateRequest {
    }
 }

+/// An alternative representation of `pageserver::tenant::TenantConf` with
+/// simpler types.
 #[derive(Serialize, Deserialize, Debug, Default)]
 pub struct TenantConfig {
    pub checkpoint_distance: Option<u64>,
@@ -243,6 +246,39 @@ pub struct TenantConfig {
    pub gc_feedback: Option<bool>,
 }

+/// A flattened analog of a `pagesever::tenant::LocationMode`, which
+/// lists out all possible states (and the virtual "Detached" state)
+/// in a flat form rather than using rust-style enums.
+#[derive(Serialize, Deserialize, Debug)]
+pub enum LocationConfigMode {
+    AttachedSingle,
+    AttachedMulti,
+    AttachedStale,
+    Secondary,
+    Detached,
+}
+
+#[derive(Serialize, Deserialize, Debug)]
+pub struct LocationConfigSecondary {
+    pub warm: bool,
+}
+
+/// An alternative representation of `pageserver::tenant::LocationConf`,
+/// for use in external-facing APIs.
+#[derive(Serialize, Deserialize, Debug)]
+pub struct LocationConfig {
+    pub mode: LocationConfigMode,
+    /// If attaching, in what generation?
+    #[serde(default)]
+    pub generation: Option<Generation>,
+    #[serde(default)]
+    pub secondary_conf: Option<LocationConfigSecondary>,
+
+    // If requesting mode `Secondary`, configuration for that.
+    // Custom storage configuration for the tenant, if any
+    pub tenant_conf: TenantConfig,
+}
+
 #[serde_as]
 #[derive(Serialize, Deserialize)]
 #[serde(transparent)]
@@ -253,6 +289,16 @@ pub struct StatusResponse {
    pub id: NodeId,
 }

+#[serde_as]
+#[derive(Serialize, Deserialize, Debug)]
+#[serde(deny_unknown_fields)]
+pub struct TenantLocationConfigRequest {
+    #[serde_as(as = "DisplayFromStr")]
+    pub tenant_id: TenantId,
+    #[serde(flatten)]
+    pub config: LocationConfig, // as we have a flattened field, we should reject all unknown fields in it
+}
+
 #[serde_as]
 #[derive(Serialize, Deserialize, Debug)]
 #[serde(deny_unknown_fields)]
@@ -363,8 +409,15 @@ pub struct TimelineInfo {
    pub latest_gc_cutoff_lsn: Lsn,
    #[serde_as(as = "DisplayFromStr")]
    pub disk_consistent_lsn: Lsn,
+
+    /// The LSN that we have succesfully uploaded to remote storage
    #[serde_as(as = "DisplayFromStr")]
    pub remote_consistent_lsn: Lsn,
+
+    /// The LSN that we are advertizing to safekeepers
+    #[serde_as(as = "DisplayFromStr")]
+    pub remote_consistent_lsn_visible: Lsn,
+
    pub current_logical_size: Option<u64>, // is None when timeline is Unloaded
    /// Sum of the size of all layer files.
    /// If a layer is present in both local FS and S3, it counts only once.
--- a/libs/postgres_ffi/src/pg_constants.rs
+++ b/libs/postgres_ffi/src/pg_constants.rs
@@ -137,9 +137,12 @@ pub const XLOG_HEAP_INSERT: u8 = 0x00;
 pub const XLOG_HEAP_DELETE: u8 = 0x10;
 pub const XLOG_HEAP_UPDATE: u8 = 0x20;
 pub const XLOG_HEAP_HOT_UPDATE: u8 = 0x40;
+pub const XLOG_HEAP_LOCK: u8 = 0x60;
 pub const XLOG_HEAP_INIT_PAGE: u8 = 0x80;
 pub const XLOG_HEAP2_VISIBLE: u8 = 0x40;
 pub const XLOG_HEAP2_MULTI_INSERT: u8 = 0x50;
+pub const XLOG_HEAP2_LOCK_UPDATED: u8 = 0x60;
+pub const XLH_LOCK_ALL_FROZEN_CLEARED: u8 = 0x01;
 pub const XLH_INSERT_ALL_FROZEN_SET: u8 = (1 << 5) as u8;
 pub const XLH_INSERT_ALL_VISIBLE_CLEARED: u8 = (1 << 0) as u8;
 pub const XLH_UPDATE_OLD_ALL_VISIBLE_CLEARED: u8 = (1 << 0) as u8;
--- a/libs/postgres_ffi/wal_craft/Cargo.toml
+++ b/libs/postgres_ffi/wal_craft/Cargo.toml
@@ -12,7 +12,7 @@ log.workspace = true
 once_cell.workspace = true
 postgres.workspace = true
 postgres_ffi.workspace = true
-tempfile.workspace = true
+camino-tempfile.workspace = true

 workspace_hack.workspace = true

--- a/libs/postgres_ffi/wal_craft/src/lib.rs
+++ b/libs/postgres_ffi/wal_craft/src/lib.rs
@@ -1,4 +1,5 @@
 use anyhow::{bail, ensure};
+use camino_tempfile::{tempdir, Utf8TempDir};
 use log::*;
 use postgres::types::PgLsn;
 use postgres::Client;
@@ -8,7 +9,6 @@ use std::cmp::Ordering;
 use std::path::{Path, PathBuf};
 use std::process::Command;
 use std::time::{Duration, Instant};
-use tempfile::{tempdir, TempDir};

 macro_rules! xlog_utils_test {
    ($version:ident) => {
@@ -33,7 +33,7 @@ pub struct Conf {

 pub struct PostgresServer {
    process: std::process::Child,
-    _unix_socket_dir: TempDir,
+    _unix_socket_dir: Utf8TempDir,
    client_config: postgres::Config,
 }

--- a/libs/remote_storage/Cargo.toml
+++ b/libs/remote_storage/Cargo.toml
@@ -13,6 +13,7 @@ aws-types.workspace = true
 aws-config.workspace = true
 aws-sdk-s3.workspace = true
 aws-credential-types.workspace = true
+camino.workspace = true
 hyper = { workspace = true, features = ["stream"] }
 serde.workspace = true
 serde_json.workspace = true
@@ -27,5 +28,6 @@ pin-project-lite.workspace = true
 workspace_hack.workspace = true

 [dev-dependencies]
-tempfile.workspace = true
+camino-tempfile.workspace = true
 test-context.workspace = true
+rand.workspace = true
--- a/libs/remote_storage/src/lib.rs
+++ b/libs/remote_storage/src/lib.rs
@@ -13,13 +13,14 @@ use std::{
    collections::HashMap,
    fmt::Debug,
    num::{NonZeroU32, NonZeroUsize},
-    path::{Path, PathBuf},
    pin::Pin,
    sync::Arc,
 };

 use anyhow::{bail, Context};
+use camino::{Utf8Path, Utf8PathBuf};

+use serde::{Deserialize, Serialize};
 use tokio::io;
 use toml_edit::Item;
 use tracing::info;
@@ -42,22 +43,44 @@ pub const DEFAULT_REMOTE_STORAGE_S3_CONCURRENCY_LIMIT: usize = 100;
 /// <https://docs.aws.amazon.com/AmazonS3/latest/API/API_ListObjectsV2.html#API_ListObjectsV2_RequestSyntax>
 pub const DEFAULT_MAX_KEYS_PER_LIST_RESPONSE: Option<i32> = None;

+/// As defined in S3 docs
+pub const MAX_KEYS_PER_DELETE: usize = 1000;
+
 const REMOTE_STORAGE_PREFIX_SEPARATOR: char = '/';

 /// Path on the remote storage, relative to some inner prefix.
 /// The prefix is an implementation detail, that allows representing local paths
 /// as the remote ones, stripping the local storage prefix away.
 #[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
-pub struct RemotePath(PathBuf);
+pub struct RemotePath(Utf8PathBuf);
+
+impl Serialize for RemotePath {
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: serde::Serializer,
+    {
+        serializer.collect_str(self)
+    }
+}
+
+impl<'de> Deserialize<'de> for RemotePath {
+    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
+    where
+        D: serde::Deserializer<'de>,
+    {
+        let str = String::deserialize(deserializer)?;
+        Ok(Self(Utf8PathBuf::from(&str)))
+    }
+}

 impl std::fmt::Display for RemotePath {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        write!(f, "{}", self.0.display())
+        std::fmt::Display::fmt(&self.0, f)
    }
 }

 impl RemotePath {
-    pub fn new(relative_path: &Path) -> anyhow::Result<Self> {
+    pub fn new(relative_path: &Utf8Path) -> anyhow::Result<Self> {
        anyhow::ensure!(
            relative_path.is_relative(),
            "Path {relative_path:?} is not relative"
@@ -66,27 +89,31 @@ impl RemotePath {
    }

    pub fn from_string(relative_path: &str) -> anyhow::Result<Self> {
-        Self::new(Path::new(relative_path))
+        Self::new(Utf8Path::new(relative_path))
    }

-    pub fn with_base(&self, base_path: &Path) -> PathBuf {
+    pub fn with_base(&self, base_path: &Utf8Path) -> Utf8PathBuf {
        base_path.join(&self.0)
    }

    pub fn object_name(&self) -> Option<&str> {
-        self.0.file_name().and_then(|os_str| os_str.to_str())
+        self.0.file_name()
    }

-    pub fn join(&self, segment: &Path) -> Self {
+    pub fn join(&self, segment: &Utf8Path) -> Self {
        Self(self.0.join(segment))
    }

-    pub fn get_path(&self) -> &PathBuf {
+    pub fn get_path(&self) -> &Utf8PathBuf {
        &self.0
    }

    pub fn extension(&self) -> Option<&str> {
-        self.0.extension()?.to_str()
+        self.0.extension()
+    }
+
+    pub fn strip_prefix(&self, p: &RemotePath) -> Result<&Utf8Path, std::path::StripPrefixError> {
+        self.0.strip_prefix(&p.0)
    }
 }

@@ -284,7 +311,7 @@ impl GenericRemoteStorage {
    pub fn from_config(storage_config: &RemoteStorageConfig) -> anyhow::Result<Self> {
        Ok(match &storage_config.storage {
            RemoteStorageKind::LocalFs(root) => {
-                info!("Using fs root '{}' as a remote storage", root.display());
+                info!("Using fs root '{root}' as a remote storage");
                Self::LocalFs(LocalFs::new(root.clone())?)
            }
            RemoteStorageKind::AwsS3(s3_config) => {
@@ -352,7 +379,7 @@ pub struct RemoteStorageConfig {
 pub enum RemoteStorageKind {
    /// Storage based on local file system.
    /// Specify a root folder to place all stored files into.
-    LocalFs(PathBuf),
+    LocalFs(Utf8PathBuf),
    /// AWS S3 based storage, storing all files in the S3 bucket
    /// specified by the config
    AwsS3(S3Config),
@@ -447,7 +474,7 @@ impl RemoteStorageConfig {
                concurrency_limit,
                max_keys_per_list_response,
            }),
-            (Some(local_path), None, None) => RemoteStorageKind::LocalFs(PathBuf::from(
+            (Some(local_path), None, None) => RemoteStorageKind::LocalFs(Utf8PathBuf::from(
                parse_toml_string("local_path", local_path)?,
            )),
            (Some(_), Some(_), _) => bail!("local_path and bucket_name are mutually exclusive"),
@@ -492,23 +519,23 @@ mod tests {

    #[test]
    fn test_object_name() {
-        let k = RemotePath::new(Path::new("a/b/c")).unwrap();
+        let k = RemotePath::new(Utf8Path::new("a/b/c")).unwrap();
        assert_eq!(k.object_name(), Some("c"));

-        let k = RemotePath::new(Path::new("a/b/c/")).unwrap();
+        let k = RemotePath::new(Utf8Path::new("a/b/c/")).unwrap();
        assert_eq!(k.object_name(), Some("c"));

-        let k = RemotePath::new(Path::new("a/")).unwrap();
+        let k = RemotePath::new(Utf8Path::new("a/")).unwrap();
        assert_eq!(k.object_name(), Some("a"));

        // XXX is it impossible to have an empty key?
-        let k = RemotePath::new(Path::new("")).unwrap();
+        let k = RemotePath::new(Utf8Path::new("")).unwrap();
        assert_eq!(k.object_name(), None);
    }

    #[test]
    fn rempte_path_cannot_be_created_from_absolute_ones() {
-        let err = RemotePath::new(Path::new("/")).expect_err("Should fail on absolute paths");
+        let err = RemotePath::new(Utf8Path::new("/")).expect_err("Should fail on absolute paths");
        assert_eq!(err.to_string(), "Path \"/\" is not relative");
    }
 }
--- a/libs/remote_storage/src/local_fs.rs
+++ b/libs/remote_storage/src/local_fs.rs
@@ -4,15 +4,10 @@
 //! This storage used in tests, but can also be used in cases when a certain persistent
 //! volume is mounted to the local FS.

-use std::{
-    borrow::Cow,
-    future::Future,
-    io::ErrorKind,
-    path::{Path, PathBuf},
-    pin::Pin,
-};
+use std::{borrow::Cow, future::Future, io::ErrorKind, pin::Pin};

 use anyhow::{bail, ensure, Context};
+use camino::{Utf8Path, Utf8PathBuf};
 use tokio::{
    fs,
    io::{self, AsyncReadExt, AsyncSeekExt, AsyncWriteExt},
@@ -28,20 +23,20 @@ const LOCAL_FS_TEMP_FILE_SUFFIX: &str = "___temp";

 #[derive(Debug, Clone)]
 pub struct LocalFs {
-    storage_root: PathBuf,
+    storage_root: Utf8PathBuf,
 }

 impl LocalFs {
    /// Attempts to create local FS storage, along with its root directory.
    /// Storage root will be created (if does not exist) and transformed into an absolute path (if passed as relative).
-    pub fn new(mut storage_root: PathBuf) -> anyhow::Result<Self> {
+    pub fn new(mut storage_root: Utf8PathBuf) -> anyhow::Result<Self> {
        if !storage_root.exists() {
            std::fs::create_dir_all(&storage_root).with_context(|| {
                format!("Failed to create all directories in the given root path {storage_root:?}")
            })?;
        }
        if !storage_root.is_absolute() {
-            storage_root = storage_root.canonicalize().with_context(|| {
+            storage_root = storage_root.canonicalize_utf8().with_context(|| {
                format!("Failed to represent path {storage_root:?} as an absolute path")
            })?;
        }
@@ -50,7 +45,7 @@ impl LocalFs {
    }

    // mirrors S3Bucket::s3_object_to_relative_path
-    fn local_file_to_relative_path(&self, key: PathBuf) -> RemotePath {
+    fn local_file_to_relative_path(&self, key: Utf8PathBuf) -> RemotePath {
        let relative_path = key
            .strip_prefix(&self.storage_root)
            .expect("relative path must contain storage_root as prefix");
@@ -59,22 +54,18 @@ impl LocalFs {

    async fn read_storage_metadata(
        &self,
-        file_path: &Path,
+        file_path: &Utf8Path,
    ) -> anyhow::Result<Option<StorageMetadata>> {
        let metadata_path = storage_metadata_path(file_path);
        if metadata_path.exists() && metadata_path.is_file() {
            let metadata_string = fs::read_to_string(&metadata_path).await.with_context(|| {
-                format!(
-                    "Failed to read metadata from the local storage at '{}'",
-                    metadata_path.display()
-                )
+                format!("Failed to read metadata from the local storage at '{metadata_path}'")
            })?;

            serde_json::from_str(&metadata_string)
                .with_context(|| {
                    format!(
-                        "Failed to deserialize metadata from the local storage at '{}'",
-                        metadata_path.display()
+                        "Failed to deserialize metadata from the local storage at '{metadata_path}'",
                    )
                })
                .map(|metadata| Some(StorageMetadata(metadata)))
@@ -171,25 +162,21 @@ impl RemoteStorage for LocalFs {
            }
        }

-        // Note that PathBuf starts_with only considers full path segments, but
+        // Note that Utf8PathBuf starts_with only considers full path segments, but
        // object prefixes are arbitrary strings, so we need the strings for doing
        // starts_with later.
-        let prefix = full_path.to_string_lossy();
+        let prefix = full_path.as_str();

        let mut files = vec![];
-        let mut directory_queue = vec![initial_dir.clone()];
+        let mut directory_queue = vec![initial_dir];
        while let Some(cur_folder) = directory_queue.pop() {
-            let mut entries = fs::read_dir(cur_folder.clone()).await?;
-            while let Some(entry) = entries.next_entry().await? {
-                let file_name: PathBuf = entry.file_name().into();
-                let full_file_name = cur_folder.clone().join(&file_name);
-                if full_file_name
-                    .to_str()
-                    .map(|s| s.starts_with(prefix.as_ref()))
-                    .unwrap_or(false)
-                {
+            let mut entries = cur_folder.read_dir_utf8()?;
+            while let Some(Ok(entry)) = entries.next() {
+                let file_name = entry.file_name();
+                let full_file_name = cur_folder.join(file_name);
+                if full_file_name.as_str().starts_with(prefix) {
                    let file_remote_path = self.local_file_to_relative_path(full_file_name.clone());
-                    files.push(file_remote_path.clone());
+                    files.push(file_remote_path);
                    if full_file_name.is_dir() {
                        directory_queue.push(full_file_name);
                    }
@@ -230,10 +217,7 @@ impl RemoteStorage for LocalFs {
                .open(&temp_file_path)
                .await
                .with_context(|| {
-                    format!(
-                        "Failed to open target fs destination at '{}'",
-                        target_file_path.display()
-                    )
+                    format!("Failed to open target fs destination at '{target_file_path}'")
                })?,
        );

@@ -244,8 +228,7 @@ impl RemoteStorage for LocalFs {
            .await
            .with_context(|| {
                format!(
-                    "Failed to upload file (write temp) to the local storage at '{}'",
-                    temp_file_path.display()
+                    "Failed to upload file (write temp) to the local storage at '{temp_file_path}'",
                )
            })?;

@@ -262,8 +245,7 @@ impl RemoteStorage for LocalFs {

        destination.flush().await.with_context(|| {
            format!(
-                "Failed to upload (flush temp) file to the local storage at '{}'",
-                temp_file_path.display()
+                "Failed to upload (flush temp) file to the local storage at '{temp_file_path}'",
            )
        })?;

@@ -271,8 +253,7 @@ impl RemoteStorage for LocalFs {
            .await
            .with_context(|| {
                format!(
-                    "Failed to upload (rename) file to the local storage at '{}'",
-                    target_file_path.display()
+                    "Failed to upload (rename) file to the local storage at '{target_file_path}'",
                )
            })?;

@@ -286,8 +267,7 @@ impl RemoteStorage for LocalFs {
            .await
            .with_context(|| {
                format!(
-                    "Failed to write metadata to the local storage at '{}'",
-                    storage_metadata_path.display()
+                    "Failed to write metadata to the local storage at '{storage_metadata_path}'",
                )
            })?;
        }
@@ -393,16 +373,16 @@ impl RemoteStorage for LocalFs {
    }
 }

-fn storage_metadata_path(original_path: &Path) -> PathBuf {
+fn storage_metadata_path(original_path: &Utf8Path) -> Utf8PathBuf {
    path_with_suffix_extension(original_path, "metadata")
 }

 fn get_all_files<'a, P>(
    directory_path: P,
    recursive: bool,
-) -> Pin<Box<dyn Future<Output = anyhow::Result<Vec<PathBuf>>> + Send + Sync + 'a>>
+) -> Pin<Box<dyn Future<Output = anyhow::Result<Vec<Utf8PathBuf>>> + Send + Sync + 'a>>
 where
-    P: AsRef<Path> + Send + Sync + 'a,
+    P: AsRef<Utf8Path> + Send + Sync + 'a,
 {
    Box::pin(async move {
        let directory_path = directory_path.as_ref();
@@ -412,7 +392,13 @@ where
                let mut dir_contents = fs::read_dir(directory_path).await?;
                while let Some(dir_entry) = dir_contents.next_entry().await? {
                    let file_type = dir_entry.file_type().await?;
-                    let entry_path = dir_entry.path();
+                    let entry_path =
+                        Utf8PathBuf::from_path_buf(dir_entry.path()).map_err(|pb| {
+                            anyhow::Error::msg(format!(
+                                "non-Unicode path: {}",
+                                pb.to_string_lossy()
+                            ))
+                        })?;
                    if file_type.is_symlink() {
                        debug!("{entry_path:?} is a symlink, skipping")
                    } else if file_type.is_dir() {
@@ -435,13 +421,10 @@ where
    })
 }

-async fn create_target_directory(target_file_path: &Path) -> anyhow::Result<()> {
+async fn create_target_directory(target_file_path: &Utf8Path) -> anyhow::Result<()> {
    let target_dir = match target_file_path.parent() {
        Some(parent_dir) => parent_dir,
-        None => bail!(
-            "File path '{}' has no parent directory",
-            target_file_path.display()
-        ),
+        None => bail!("File path '{target_file_path}' has no parent directory"),
    };
    if !target_dir.exists() {
        fs::create_dir_all(target_dir).await?;
@@ -449,13 +432,9 @@ async fn create_target_directory(target_file_path: &Path) -> anyhow::Result<()>
    Ok(())
 }

-fn file_exists(file_path: &Path) -> anyhow::Result<bool> {
+fn file_exists(file_path: &Utf8Path) -> anyhow::Result<bool> {
    if file_path.exists() {
-        ensure!(
-            file_path.is_file(),
-            "file path '{}' is not a file",
-            file_path.display()
-        );
+        ensure!(file_path.is_file(), "file path '{file_path}' is not a file");
        Ok(true)
    } else {
        Ok(false)
@@ -466,13 +445,13 @@ fn file_exists(file_path: &Path) -> anyhow::Result<bool> {
 mod fs_tests {
    use super::*;

+    use camino_tempfile::tempdir;
    use std::{collections::HashMap, io::Write};
-    use tempfile::tempdir;

    async fn read_and_assert_remote_file_contents(
        storage: &LocalFs,
        #[allow(clippy::ptr_arg)]
-        // have to use &PathBuf due to `storage.local_path` parameter requirements
+        // have to use &Utf8PathBuf due to `storage.local_path` parameter requirements
        remote_storage_path: &RemotePath,
        expected_metadata: Option<&StorageMetadata>,
    ) -> anyhow::Result<String> {
@@ -519,7 +498,7 @@ mod fs_tests {
    async fn upload_file_negatives() -> anyhow::Result<()> {
        let storage = create_storage()?;

-        let id = RemotePath::new(Path::new("dummy"))?;
+        let id = RemotePath::new(Utf8Path::new("dummy"))?;
        let content = std::io::Cursor::new(b"12345");

        // Check that you get an error if the size parameter doesn't match the actual
@@ -544,7 +523,8 @@ mod fs_tests {
    }

    fn create_storage() -> anyhow::Result<LocalFs> {
-        LocalFs::new(tempdir()?.path().to_owned())
+        let storage_root = tempdir()?.path().to_path_buf();
+        LocalFs::new(storage_root)
    }

    #[tokio::test]
@@ -561,7 +541,7 @@ mod fs_tests {
        );

        let non_existing_path = "somewhere/else";
-        match storage.download(&RemotePath::new(Path::new(non_existing_path))?).await {
+        match storage.download(&RemotePath::new(Utf8Path::new(non_existing_path))?).await {
            Err(DownloadError::NotFound) => {} // Should get NotFound for non existing keys
            other => panic!("Should get a NotFound error when downloading non-existing storage files, but got: {other:?}"),
        }
@@ -775,7 +755,7 @@ mod fs_tests {
    }

    async fn create_file_for_upload(
-        path: &Path,
+        path: &Utf8Path,
        contents: &str,
    ) -> anyhow::Result<(io::BufReader<fs::File>, usize)> {
        std::fs::create_dir_all(path.parent().unwrap())?;
--- a/libs/remote_storage/src/s3_bucket.rs
+++ b/libs/remote_storage/src/s3_bucket.rs
@@ -33,11 +33,10 @@ use tracing::debug;

 use super::StorageMetadata;
 use crate::{
-    Download, DownloadError, RemotePath, RemoteStorage, S3Config, REMOTE_STORAGE_PREFIX_SEPARATOR,
+    Download, DownloadError, RemotePath, RemoteStorage, S3Config, MAX_KEYS_PER_DELETE,
+    REMOTE_STORAGE_PREFIX_SEPARATOR,
 };

-const MAX_DELETE_OBJECTS_REQUEST_SIZE: usize = 1000;
-
 pub(super) mod metrics;

 use self::metrics::{AttemptOutcome, RequestKind};
@@ -48,10 +47,47 @@ pub struct S3Bucket {
    bucket_name: String,
    prefix_in_bucket: Option<String>,
    max_keys_per_list_response: Option<i32>,
+    concurrency_limiter: ConcurrencyLimiter,
+}
+
+struct ConcurrencyLimiter {
    // Every request to S3 can be throttled or cancelled, if a certain number of requests per second is exceeded.
    // Same goes to IAM, which is queried before every S3 request, if enabled. IAM has even lower RPS threshold.
    // The helps to ensure we don't exceed the thresholds.
-    concurrency_limiter: Arc<Semaphore>,
+    write: Arc<Semaphore>,
+    read: Arc<Semaphore>,
+}
+
+impl ConcurrencyLimiter {
+    fn for_kind(&self, kind: RequestKind) -> &Arc<Semaphore> {
+        match kind {
+            RequestKind::Get => &self.read,
+            RequestKind::Put => &self.write,
+            RequestKind::List => &self.read,
+            RequestKind::Delete => &self.write,
+        }
+    }
+
+    async fn acquire(
+        &self,
+        kind: RequestKind,
+    ) -> Result<tokio::sync::SemaphorePermit<'_>, tokio::sync::AcquireError> {
+        self.for_kind(kind).acquire().await
+    }
+
+    async fn acquire_owned(
+        &self,
+        kind: RequestKind,
+    ) -> Result<tokio::sync::OwnedSemaphorePermit, tokio::sync::AcquireError> {
+        Arc::clone(self.for_kind(kind)).acquire_owned().await
+    }
+
+    fn new(limit: usize) -> ConcurrencyLimiter {
+        Self {
+            read: Arc::new(Semaphore::new(limit)),
+            write: Arc::new(Semaphore::new(limit)),
+        }
+    }
 }

 #[derive(Default)]
@@ -118,7 +154,7 @@ impl S3Bucket {
            bucket_name: aws_config.bucket_name.clone(),
            max_keys_per_list_response: aws_config.max_keys_per_list_response,
            prefix_in_bucket,
-            concurrency_limiter: Arc::new(Semaphore::new(aws_config.concurrency_limit.get())),
+            concurrency_limiter: ConcurrencyLimiter::new(aws_config.concurrency_limit.get()),
        })
    }

@@ -144,12 +180,11 @@ impl S3Bucket {
        assert_eq!(std::path::MAIN_SEPARATOR, REMOTE_STORAGE_PREFIX_SEPARATOR);
        let path_string = path
            .get_path()
-            .to_string_lossy()
-            .trim_end_matches(REMOTE_STORAGE_PREFIX_SEPARATOR)
-            .to_string();
+            .as_str()
+            .trim_end_matches(REMOTE_STORAGE_PREFIX_SEPARATOR);
        match &self.prefix_in_bucket {
-            Some(prefix) => prefix.clone() + "/" + &path_string,
-            None => path_string,
+            Some(prefix) => prefix.clone() + "/" + path_string,
+            None => path_string.to_string(),
        }
    }

@@ -157,7 +192,7 @@ impl S3Bucket {
        let started_at = start_counting_cancelled_wait(kind);
        let permit = self
            .concurrency_limiter
-            .acquire()
+            .acquire(kind)
            .await
            .expect("semaphore is never closed");

@@ -173,8 +208,7 @@ impl S3Bucket {
        let started_at = start_counting_cancelled_wait(kind);
        let permit = self
            .concurrency_limiter
-            .clone()
-            .acquire_owned()
+            .acquire_owned(kind)
            .await
            .expect("semaphore is never closed");

@@ -500,7 +534,7 @@ impl RemoteStorage for S3Bucket {
            delete_objects.push(obj_id);
        }

-        for chunk in delete_objects.chunks(MAX_DELETE_OBJECTS_REQUEST_SIZE) {
+        for chunk in delete_objects.chunks(MAX_KEYS_PER_DELETE) {
            let started_at = start_measuring_requests(kind);

            let resp = self
@@ -566,8 +600,8 @@ fn start_measuring_requests(

 #[cfg(test)]
 mod tests {
+    use camino::Utf8Path;
    use std::num::NonZeroUsize;
-    use std::path::Path;

    use crate::{RemotePath, S3Bucket, S3Config};

@@ -576,7 +610,7 @@ mod tests {
        let all_paths = ["", "some/path", "some/path/"];
        let all_paths: Vec<RemotePath> = all_paths
            .iter()
-            .map(|x| RemotePath::new(Path::new(x)).expect("bad path"))
+            .map(|x| RemotePath::new(Utf8Path::new(x)).expect("bad path"))
            .collect();
        let prefixes = [
            None,
--- a/libs/remote_storage/tests/test_real_s3.rs
+++ b/libs/remote_storage/tests/test_real_s3.rs
@@ -2,11 +2,12 @@ use std::collections::HashSet;
 use std::env;
 use std::num::{NonZeroU32, NonZeroUsize};
 use std::ops::ControlFlow;
-use std::path::{Path, PathBuf};
+use std::path::PathBuf;
 use std::sync::Arc;
 use std::time::UNIX_EPOCH;

 use anyhow::Context;
+use camino::Utf8Path;
 use once_cell::sync::OnceCell;
 use remote_storage::{
    GenericRemoteStorage, RemotePath, RemoteStorageConfig, RemoteStorageKind, S3Config,
@@ -55,7 +56,7 @@ async fn s3_pagination_should_work(ctx: &mut MaybeEnabledS3WithTestBlobs) -> any
    let test_client = Arc::clone(&ctx.enabled.client);
    let expected_remote_prefixes = ctx.remote_prefixes.clone();

-    let base_prefix = RemotePath::new(Path::new(ctx.enabled.base_prefix))
+    let base_prefix = RemotePath::new(Utf8Path::new(ctx.enabled.base_prefix))
        .context("common_prefix construction")?;
    let root_remote_prefixes = test_client
        .list_prefixes(None)
@@ -108,7 +109,7 @@ async fn s3_list_files_works(ctx: &mut MaybeEnabledS3WithSimpleTestBlobs) -> any
    };
    let test_client = Arc::clone(&ctx.enabled.client);
    let base_prefix =
-        RemotePath::new(Path::new("folder1")).context("common_prefix construction")?;
+        RemotePath::new(Utf8Path::new("folder1")).context("common_prefix construction")?;
    let root_files = test_client
        .list_files(None)
        .await
@@ -129,9 +130,9 @@ async fn s3_list_files_works(ctx: &mut MaybeEnabledS3WithSimpleTestBlobs) -> any
    let trim_remote_blobs: HashSet<_> = ctx
        .remote_blobs
        .iter()
-        .map(|x| x.get_path().to_str().expect("must be valid name"))
+        .map(|x| x.get_path())
        .filter(|x| x.starts_with("folder1"))
-        .map(|x| RemotePath::new(Path::new(x)).expect("must be valid name"))
+        .map(|x| RemotePath::new(x).expect("must be valid path"))
        .collect();
    assert_eq!(
        nested_remote_files, trim_remote_blobs,
@@ -148,10 +149,9 @@ async fn s3_delete_non_exising_works(ctx: &mut MaybeEnabledS3) -> anyhow::Result
        MaybeEnabledS3::Disabled => return Ok(()),
    };

-    let path = RemotePath::new(&PathBuf::from(format!(
-        "{}/for_sure_there_is_nothing_there_really",
-        ctx.base_prefix,
-    )))
+    let path = RemotePath::new(Utf8Path::new(
+        format!("{}/for_sure_there_is_nothing_there_really", ctx.base_prefix).as_str(),
+    ))
    .with_context(|| "RemotePath conversion")?;

    ctx.client.delete(&path).await.expect("should succeed");
@@ -167,13 +167,13 @@ async fn s3_delete_objects_works(ctx: &mut MaybeEnabledS3) -> anyhow::Result<()>
        MaybeEnabledS3::Disabled => return Ok(()),
    };

-    let path1 = RemotePath::new(&PathBuf::from(format!("{}/path1", ctx.base_prefix,)))
+    let path1 = RemotePath::new(Utf8Path::new(format!("{}/path1", ctx.base_prefix).as_str()))
        .with_context(|| "RemotePath conversion")?;

-    let path2 = RemotePath::new(&PathBuf::from(format!("{}/path2", ctx.base_prefix,)))
+    let path2 = RemotePath::new(Utf8Path::new(format!("{}/path2", ctx.base_prefix).as_str()))
        .with_context(|| "RemotePath conversion")?;

-    let path3 = RemotePath::new(&PathBuf::from(format!("{}/path3", ctx.base_prefix,)))
+    let path3 = RemotePath::new(Utf8Path::new(format!("{}/path3", ctx.base_prefix).as_str()))
        .with_context(|| "RemotePath conversion")?;

    let data1 = "remote blob data1".as_bytes();
@@ -378,21 +378,30 @@ impl AsyncTestContext for MaybeEnabledS3WithSimpleTestBlobs {
 fn create_s3_client(
    max_keys_per_list_response: Option<i32>,
 ) -> anyhow::Result<Arc<GenericRemoteStorage>> {
+    use rand::Rng;
+
    let remote_storage_s3_bucket = env::var("REMOTE_STORAGE_S3_BUCKET")
        .context("`REMOTE_STORAGE_S3_BUCKET` env var is not set, but real S3 tests are enabled")?;
    let remote_storage_s3_region = env::var("REMOTE_STORAGE_S3_REGION")
        .context("`REMOTE_STORAGE_S3_REGION` env var is not set, but real S3 tests are enabled")?;
-    let random_prefix_part = std::time::SystemTime::now()
+
+    // due to how time works, we've had test runners use the same nanos as bucket prefixes.
+    // millis is just a debugging aid for easier finding the prefix later.
+    let millis = std::time::SystemTime::now()
        .duration_since(UNIX_EPOCH)
        .context("random s3 test prefix part calculation")?
-        .as_nanos();
+        .as_millis();
+
+    // because nanos can be the same for two threads so can millis, add randomness
+    let random = rand::thread_rng().gen::<u32>();
+
    let remote_storage_config = RemoteStorageConfig {
        max_concurrent_syncs: NonZeroUsize::new(100).unwrap(),
        max_sync_errors: NonZeroU32::new(5).unwrap(),
        storage: RemoteStorageKind::AwsS3(S3Config {
            bucket_name: remote_storage_s3_bucket,
            bucket_region: remote_storage_s3_region,
-            prefix_in_bucket: Some(format!("pagination_should_work_test_{random_prefix_part}/")),
+            prefix_in_bucket: Some(format!("test_{millis}_{random:08x}/")),
            endpoint: None,
            concurrency_limit: NonZeroUsize::new(100).unwrap(),
            max_keys_per_list_response,
@@ -418,10 +427,10 @@ async fn upload_s3_data(
    for i in 1..upload_tasks_count + 1 {
        let task_client = Arc::clone(client);
        upload_tasks.spawn(async move {
-            let prefix = PathBuf::from(format!("{base_prefix_str}/sub_prefix_{i}/"));
-            let blob_prefix = RemotePath::new(&prefix)
+            let prefix = format!("{base_prefix_str}/sub_prefix_{i}/");
+            let blob_prefix = RemotePath::new(Utf8Path::new(&prefix))
                .with_context(|| format!("{prefix:?} to RemotePath conversion"))?;
-            let blob_path = blob_prefix.join(Path::new(&format!("blob_{i}")));
+            let blob_path = blob_prefix.join(Utf8Path::new(&format!("blob_{i}")));
            debug!("Creating remote item {i} at path {blob_path:?}");

            let data = format!("remote blob data {i}").into_bytes();
@@ -503,8 +512,10 @@ async fn upload_simple_s3_data(
        let task_client = Arc::clone(client);
        upload_tasks.spawn(async move {
            let blob_path = PathBuf::from(format!("folder{}/blob_{}.txt", i / 7, i));
-            let blob_path = RemotePath::new(&blob_path)
-                .with_context(|| format!("{blob_path:?} to RemotePath conversion"))?;
+            let blob_path = RemotePath::new(
+                Utf8Path::from_path(blob_path.as_path()).expect("must be valid blob path"),
+            )
+            .with_context(|| format!("{blob_path:?} to RemotePath conversion"))?;
            debug!("Creating remote item {i} at path {blob_path:?}");

            let data = format!("remote blob data {i}").into_bytes();
--- a/libs/utils/Cargo.toml
+++ b/libs/utils/Cargo.toml
@@ -10,6 +10,7 @@ async-trait.workspace = true
 anyhow.workspace = true
 bincode.workspace = true
 bytes.workspace = true
+camino.workspace = true
 chrono.workspace = true
 heapless.workspace = true
 hex = { workspace = true, features = ["serde"] }
@@ -53,7 +54,7 @@ byteorder.workspace = true
 bytes.workspace = true
 criterion.workspace = true
 hex-literal.workspace = true
-tempfile.workspace = true
+camino-tempfile.workspace = true

 [[bench]]
 name = "benchmarks"
--- a/libs/utils/src/auth.rs
+++ b/libs/utils/src/auth.rs
@@ -2,9 +2,9 @@

 use serde;
 use std::fs;
-use std::path::Path;

 use anyhow::Result;
+use camino::Utf8Path;
 use jsonwebtoken::{
    decode, encode, Algorithm, DecodingKey, EncodingKey, Header, TokenData, Validation,
 };
@@ -65,7 +65,7 @@ impl JwtAuth {
        }
    }

-    pub fn from_key_path(key_path: &Path) -> Result<Self> {
+    pub fn from_key_path(key_path: &Utf8Path) -> Result<Self> {
        let public_key = fs::read(key_path)?;
        Ok(Self::new(DecodingKey::from_ed_pem(&public_key)?))
    }
--- a/libs/utils/src/crashsafe.rs
+++ b/libs/utils/src/crashsafe.rs
@@ -1,14 +1,14 @@
 use std::{
    borrow::Cow,
-    ffi::OsStr,
    fs::{self, File},
    io,
-    path::{Path, PathBuf},
 };

+use camino::{Utf8Path, Utf8PathBuf};
+
 /// Similar to [`std::fs::create_dir`], except we fsync the
 /// created directory and its parent.
-pub fn create_dir(path: impl AsRef<Path>) -> io::Result<()> {
+pub fn create_dir(path: impl AsRef<Utf8Path>) -> io::Result<()> {
    let path = path.as_ref();

    fs::create_dir(path)?;
@@ -18,7 +18,7 @@ pub fn create_dir(path: impl AsRef<Path>) -> io::Result<()> {

 /// Similar to [`std::fs::create_dir_all`], except we fsync all
 /// newly created directories and the pre-existing parent.
-pub fn create_dir_all(path: impl AsRef<Path>) -> io::Result<()> {
+pub fn create_dir_all(path: impl AsRef<Utf8Path>) -> io::Result<()> {
    let mut path = path.as_ref();

    let mut dirs_to_create = Vec::new();
@@ -30,7 +30,7 @@ pub fn create_dir_all(path: impl AsRef<Path>) -> io::Result<()> {
            Ok(_) => {
                return Err(io::Error::new(
                    io::ErrorKind::AlreadyExists,
-                    format!("non-directory found in path: {}", path.display()),
+                    format!("non-directory found in path: {path}"),
                ));
            }
            Err(ref e) if e.kind() == io::ErrorKind::NotFound => {}
@@ -44,7 +44,7 @@ pub fn create_dir_all(path: impl AsRef<Path>) -> io::Result<()> {
            None => {
                return Err(io::Error::new(
                    io::ErrorKind::InvalidInput,
-                    format!("can't find parent of path '{}'", path.display()).as_str(),
+                    format!("can't find parent of path '{path}'"),
                ));
            }
        }
@@ -70,21 +70,18 @@ pub fn create_dir_all(path: impl AsRef<Path>) -> io::Result<()> {

 /// Adds a suffix to the file(directory) name, either appending the suffix to the end of its extension,
 /// or if there's no extension, creates one and puts a suffix there.
-pub fn path_with_suffix_extension(original_path: impl AsRef<Path>, suffix: &str) -> PathBuf {
-    let new_extension = match original_path
-        .as_ref()
-        .extension()
-        .map(OsStr::to_string_lossy)
-    {
+pub fn path_with_suffix_extension(
+    original_path: impl AsRef<Utf8Path>,
+    suffix: &str,
+) -> Utf8PathBuf {
+    let new_extension = match original_path.as_ref().extension() {
        Some(extension) => Cow::Owned(format!("{extension}.{suffix}")),
        None => Cow::Borrowed(suffix),
    };
-    original_path
-        .as_ref()
-        .with_extension(new_extension.as_ref())
+    original_path.as_ref().with_extension(new_extension)
 }

-pub fn fsync_file_and_parent(file_path: &Path) -> io::Result<()> {
+pub fn fsync_file_and_parent(file_path: &Utf8Path) -> io::Result<()> {
    let parent = file_path.parent().ok_or_else(|| {
        io::Error::new(
            io::ErrorKind::Other,
@@ -97,7 +94,7 @@ pub fn fsync_file_and_parent(file_path: &Path) -> io::Result<()> {
    Ok(())
 }

-pub fn fsync(path: &Path) -> io::Result<()> {
+pub fn fsync(path: &Utf8Path) -> io::Result<()> {
    File::open(path)
        .map_err(|e| io::Error::new(e.kind(), format!("Failed to open the file {path:?}: {e}")))
        .and_then(|file| {
@@ -111,19 +108,18 @@ pub fn fsync(path: &Path) -> io::Result<()> {
        .map_err(|e| io::Error::new(e.kind(), format!("Failed to fsync file {path:?}: {e}")))
 }

-pub async fn fsync_async(path: impl AsRef<std::path::Path>) -> Result<(), std::io::Error> {
-    tokio::fs::File::open(path).await?.sync_all().await
+pub async fn fsync_async(path: impl AsRef<Utf8Path>) -> Result<(), std::io::Error> {
+    tokio::fs::File::open(path.as_ref()).await?.sync_all().await
 }

 #[cfg(test)]
 mod tests {
-    use tempfile::tempdir;

    use super::*;

    #[test]
    fn test_create_dir_fsyncd() {
-        let dir = tempdir().unwrap();
+        let dir = camino_tempfile::tempdir().unwrap();

        let existing_dir_path = dir.path();
        let err = create_dir(existing_dir_path).unwrap_err();
@@ -139,7 +135,7 @@ mod tests {

    #[test]
    fn test_create_dir_all_fsyncd() {
-        let dir = tempdir().unwrap();
+        let dir = camino_tempfile::tempdir().unwrap();

        let existing_dir_path = dir.path();
        create_dir_all(existing_dir_path).unwrap();
@@ -166,29 +162,29 @@ mod tests {

    #[test]
    fn test_path_with_suffix_extension() {
-        let p = PathBuf::from("/foo/bar");
+        let p = Utf8PathBuf::from("/foo/bar");
        assert_eq!(
-            &path_with_suffix_extension(p, "temp").to_string_lossy(),
+            &path_with_suffix_extension(p, "temp").to_string(),
            "/foo/bar.temp"
        );
-        let p = PathBuf::from("/foo/bar");
+        let p = Utf8PathBuf::from("/foo/bar");
        assert_eq!(
-            &path_with_suffix_extension(p, "temp.temp").to_string_lossy(),
+            &path_with_suffix_extension(p, "temp.temp").to_string(),
            "/foo/bar.temp.temp"
        );
-        let p = PathBuf::from("/foo/bar.baz");
+        let p = Utf8PathBuf::from("/foo/bar.baz");
        assert_eq!(
-            &path_with_suffix_extension(p, "temp.temp").to_string_lossy(),
+            &path_with_suffix_extension(p, "temp.temp").to_string(),
            "/foo/bar.baz.temp.temp"
        );
-        let p = PathBuf::from("/foo/bar.baz");
+        let p = Utf8PathBuf::from("/foo/bar.baz");
        assert_eq!(
-            &path_with_suffix_extension(p, ".temp").to_string_lossy(),
+            &path_with_suffix_extension(p, ".temp").to_string(),
            "/foo/bar.baz..temp"
        );
-        let p = PathBuf::from("/foo/bar/dir/");
+        let p = Utf8PathBuf::from("/foo/bar/dir/");
        assert_eq!(
-            &path_with_suffix_extension(p, ".temp").to_string_lossy(),
+            &path_with_suffix_extension(p, ".temp").to_string(),
            "/foo/bar/dir..temp"
        );
    }
--- a/libs/utils/src/fs_ext.rs
+++ b/libs/utils/src/fs_ext.rs
@@ -55,8 +55,6 @@ where

 #[cfg(test)]
 mod test {
-    use std::path::PathBuf;
-
    use crate::fs_ext::{is_directory_empty, list_dir};

    use super::ignore_absent_files;
@@ -65,7 +63,7 @@ mod test {
    fn is_empty_dir() {
        use super::PathExt;

-        let dir = tempfile::tempdir().unwrap();
+        let dir = camino_tempfile::tempdir().unwrap();
        let dir_path = dir.path();

        // test positive case
@@ -75,7 +73,7 @@ mod test {
        );

        // invoke on a file to ensure it returns an error
-        let file_path: PathBuf = dir_path.join("testfile");
+        let file_path = dir_path.join("testfile");
        let f = std::fs::File::create(&file_path).unwrap();
        drop(f);
        assert!(file_path.is_empty_dir().is_err());
@@ -87,7 +85,7 @@ mod test {

    #[tokio::test]
    async fn is_empty_dir_async() {
-        let dir = tempfile::tempdir().unwrap();
+        let dir = camino_tempfile::tempdir().unwrap();
        let dir_path = dir.path();

        // test positive case
@@ -97,7 +95,7 @@ mod test {
        );

        // invoke on a file to ensure it returns an error
-        let file_path: PathBuf = dir_path.join("testfile");
+        let file_path = dir_path.join("testfile");
        let f = std::fs::File::create(&file_path).unwrap();
        drop(f);
        assert!(is_directory_empty(&file_path).await.is_err());
@@ -109,10 +107,9 @@ mod test {

    #[test]
    fn ignore_absent_files_works() {
-        let dir = tempfile::tempdir().unwrap();
-        let dir_path = dir.path();
+        let dir = camino_tempfile::tempdir().unwrap();

-        let file_path: PathBuf = dir_path.join("testfile");
+        let file_path = dir.path().join("testfile");

        ignore_absent_files(|| std::fs::remove_file(&file_path)).expect("should execute normally");

@@ -126,17 +123,17 @@ mod test {

    #[tokio::test]
    async fn list_dir_works() {
-        let dir = tempfile::tempdir().unwrap();
+        let dir = camino_tempfile::tempdir().unwrap();
        let dir_path = dir.path();

        assert!(list_dir(dir_path).await.unwrap().is_empty());

-        let file_path: PathBuf = dir_path.join("testfile");
+        let file_path = dir_path.join("testfile");
        let _ = std::fs::File::create(&file_path).unwrap();

        assert_eq!(&list_dir(dir_path).await.unwrap(), &["testfile"]);

-        let another_dir_path: PathBuf = dir_path.join("testdir");
+        let another_dir_path = dir_path.join("testdir");
        std::fs::create_dir(another_dir_path).unwrap();

        let expected = &["testdir", "testfile"];
--- a/libs/utils/src/generation.rs
+++ b/libs/utils/src/generation.rs
@@ -89,6 +89,22 @@ impl Generation {
            Self::Broken => panic!("Attempted to use a broken generation"),
        }
    }
+
+    pub fn next(&self) -> Generation {
+        match self {
+            Self::Valid(n) => Self::Valid(*n + 1),
+            Self::None => Self::Valid(1),
+            Self::Broken => panic!("Attempted to use a broken generation"),
+        }
+    }
+
+    pub fn into(self) -> Option<u32> {
+        if let Self::Valid(v) = self {
+            Some(v)
+        } else {
+            None
+        }
+    }
 }

 impl Serialize for Generation {
--- a/libs/utils/src/http/error.rs
+++ b/libs/utils/src/http/error.rs
@@ -24,6 +24,12 @@ pub enum ApiError {
    #[error("Precondition failed: {0}")]
    PreconditionFailed(Box<str>),

+    #[error("Resource temporarily unavailable: {0}")]
+    ResourceUnavailable(String),
+
+    #[error("Shutting down")]
+    ShuttingDown,
+
    #[error(transparent)]
    InternalServerError(anyhow::Error),
 }
@@ -52,6 +58,14 @@ impl ApiError {
                self.to_string(),
                StatusCode::PRECONDITION_FAILED,
            ),
+            ApiError::ShuttingDown => HttpErrorBody::response_from_msg_and_status(
+                "Shutting down".to_string(),
+                StatusCode::SERVICE_UNAVAILABLE,
+            ),
+            ApiError::ResourceUnavailable(err) => HttpErrorBody::response_from_msg_and_status(
+                err.to_string(),
+                StatusCode::SERVICE_UNAVAILABLE,
+            ),
            ApiError::InternalServerError(err) => HttpErrorBody::response_from_msg_and_status(
                err.to_string(),
                StatusCode::INTERNAL_SERVER_ERROR,
--- a/libs/utils/src/id.rs
+++ b/libs/utils/src/id.rs
@@ -1,4 +1,3 @@
-use std::ffi::OsStr;
 use std::{fmt, str::FromStr};

 use anyhow::Context;
@@ -215,12 +214,11 @@ pub struct TimelineId(Id);

 id_newtype!(TimelineId);

-impl TryFrom<Option<&OsStr>> for TimelineId {
+impl TryFrom<Option<&str>> for TimelineId {
    type Error = anyhow::Error;

-    fn try_from(value: Option<&OsStr>) -> Result<Self, Self::Error> {
+    fn try_from(value: Option<&str>) -> Result<Self, Self::Error> {
        value
-            .and_then(OsStr::to_str)
            .unwrap_or_default()
            .parse::<TimelineId>()
            .with_context(|| format!("Could not parse timeline id from {:?}", value))
--- a/libs/utils/src/lock_file.rs
+++ b/libs/utils/src/lock_file.rs
@@ -11,10 +11,10 @@ use std::{
    io::{Read, Write},
    ops::Deref,
    os::unix::prelude::AsRawFd,
-    path::{Path, PathBuf},
 };

 use anyhow::Context;
+use camino::{Utf8Path, Utf8PathBuf};
 use nix::{errno::Errno::EAGAIN, fcntl};

 use crate::crashsafe;
@@ -23,7 +23,7 @@ use crate::crashsafe;
 /// Returned by [`create_exclusive`].
 #[must_use]
 pub struct UnwrittenLockFile {
-    path: PathBuf,
+    path: Utf8PathBuf,
    file: fs::File,
 }

@@ -60,7 +60,7 @@ impl UnwrittenLockFile {
 ///
 /// It is not an error if the file already exists.
 /// It is an error if the file is already locked.
-pub fn create_exclusive(lock_file_path: &Path) -> anyhow::Result<UnwrittenLockFile> {
+pub fn create_exclusive(lock_file_path: &Utf8Path) -> anyhow::Result<UnwrittenLockFile> {
    let lock_file = fs::OpenOptions::new()
        .create(true) // O_CREAT
        .write(true)
@@ -101,7 +101,7 @@ pub enum LockFileRead {
 /// Open & try to lock the lock file at the given `path`, returning a [handle][`LockFileRead`] to
 /// inspect its content. It is not an `Err(...)` if the file does not exist or is already locked.
 /// Check the [`LockFileRead`] variants for details.
-pub fn read_and_hold_lock_file(path: &Path) -> anyhow::Result<LockFileRead> {
+pub fn read_and_hold_lock_file(path: &Utf8Path) -> anyhow::Result<LockFileRead> {
    let res = fs::OpenOptions::new().read(true).open(path);
    let mut lock_file = match res {
        Ok(f) => f,
--- a/libs/utils/src/logging.rs
+++ b/libs/utils/src/logging.rs
@@ -216,6 +216,30 @@ impl std::fmt::Debug for PrettyLocation<'_, '_> {
    }
 }

+/// When you will store a secret but want to make sure it won't
+/// be accidentally logged, wrap it in a SecretString, whose Debug
+/// implementation does not expose the contents.
+#[derive(Clone, Eq, PartialEq)]
+pub struct SecretString(String);
+
+impl SecretString {
+    pub fn get_contents(&self) -> &str {
+        self.0.as_str()
+    }
+}
+
+impl From<String> for SecretString {
+    fn from(s: String) -> Self {
+        Self(s)
+    }
+}
+
+impl std::fmt::Debug for SecretString {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "[SECRET]")
+    }
+}
+
 #[cfg(test)]
 mod tests {
    use metrics::{core::Opts, IntCounterVec};
--- a/libs/utils/src/lsn.rs
+++ b/libs/utils/src/lsn.rs
@@ -1,9 +1,9 @@
 #![warn(missing_docs)]

+use camino::Utf8Path;
 use serde::{Deserialize, Serialize};
 use std::fmt;
 use std::ops::{Add, AddAssign};
-use std::path::Path;
 use std::str::FromStr;
 use std::sync::atomic::{AtomicU64, Ordering};

@@ -44,11 +44,9 @@ impl Lsn {
    /// Parse an LSN from a filename in the form `0000000000000000`
    pub fn from_filename<F>(filename: F) -> Result<Self, LsnParseError>
    where
-        F: AsRef<Path>,
+        F: AsRef<Utf8Path>,
    {
-        let filename: &Path = filename.as_ref();
-        let filename = filename.to_str().ok_or(LsnParseError)?;
-        Lsn::from_hex(filename)
+        Lsn::from_hex(filename.as_ref().as_str())
    }

    /// Parse an LSN from a string in the form `0000000000000000`
--- a/libs/utils/src/pid_file.rs
+++ b/libs/utils/src/pid_file.rs
@@ -49,9 +49,10 @@
 //! At this point, `B` and `C` are running, which is hazardous.
 //! Morale of the story: don't unlink pidfiles, ever.

-use std::{ops::Deref, path::Path};
+use std::ops::Deref;

 use anyhow::Context;
+use camino::Utf8Path;
 use nix::unistd::Pid;

 use crate::lock_file::{self, LockFileRead};
@@ -84,7 +85,7 @@ impl Deref for PidFileGuard {
 /// The claim ends as soon as the returned guard object is dropped.
 /// To maintain the claim for the remaining lifetime of the current process,
 /// use [`std::mem::forget`] or similar.
-pub fn claim_for_current_process(path: &Path) -> anyhow::Result<PidFileGuard> {
+pub fn claim_for_current_process(path: &Utf8Path) -> anyhow::Result<PidFileGuard> {
    let unwritten_lock_file = lock_file::create_exclusive(path).context("lock file")?;
    // if any of the next steps fail, we drop the file descriptor and thereby release the lock
    let guard = unwritten_lock_file
@@ -132,7 +133,7 @@ pub enum PidFileRead {
 ///
 /// On success, this function returns a [`PidFileRead`].
 /// Check its docs for a description of the meaning of its different variants.
-pub fn read(pidfile: &Path) -> anyhow::Result<PidFileRead> {
+pub fn read(pidfile: &Utf8Path) -> anyhow::Result<PidFileRead> {
    let res = lock_file::read_and_hold_lock_file(pidfile).context("read and hold pid file")?;
    let ret = match res {
        LockFileRead::NotExist => PidFileRead::NotExist,
--- a/libs/vm_monitor/src/cgroup.rs
+++ b/libs/vm_monitor/src/cgroup.rs
@@ -431,14 +431,14 @@ impl CgroupWatcher {
                            .context("failed to request upscale")?;

                        let memory_high =
-                            self.get_high_bytes().context("failed to get memory.high")?;
+                            self.get_memory_high_bytes().context("failed to get memory.high")?;
                        let new_high = memory_high + self.config.memory_high_increase_by_bytes;
                        info!(
                            current_high_bytes = memory_high,
                            new_high_bytes = new_high,
                            "updating memory.high"
                        );
-                        self.set_high_bytes(new_high)
+                        self.set_memory_high_bytes(new_high)
                            .context("failed to set memory.high")?;
                        last_memory_high_increase_at = Some(Instant::now());
                        continue;
@@ -556,21 +556,6 @@ impl CgroupWatcher {
    }
 }

-/// Represents a set of limits we apply to a cgroup to control memory usage.
-///
-/// Setting these values also affects the thresholds for receiving usage alerts.
-#[derive(Debug)]
-pub struct MemoryLimits {
-    high: u64,
-    max: u64,
-}
-
-impl MemoryLimits {
-    pub fn new(high: u64, max: u64) -> Self {
-        Self { max, high }
-    }
-}
-
 // Methods for manipulating the actual cgroup
 impl CgroupWatcher {
    /// Get a handle on the freezer subsystem.
@@ -631,55 +616,29 @@ impl CgroupWatcher {
    }

    /// Set cgroup memory.high threshold.
-    pub fn set_high_bytes(&self, bytes: u64) -> anyhow::Result<()> {
+    pub fn set_memory_high_bytes(&self, bytes: u64) -> anyhow::Result<()> {
+        self.set_memory_high_internal(MaxValue::Value(u64::min(bytes, i64::MAX as u64) as i64))
+    }
+
+    /// Set the cgroup's memory.high to 'max', disabling it.
+    pub fn unset_memory_high(&self) -> anyhow::Result<()> {
+        self.set_memory_high_internal(MaxValue::Max)
+    }
+
+    fn set_memory_high_internal(&self, value: MaxValue) -> anyhow::Result<()> {
        self.memory()
            .context("failed to get memory subsystem")?
            .set_mem(cgroups_rs::memory::SetMemory {
                low: None,
-                high: Some(MaxValue::Value(u64::min(bytes, i64::MAX as u64) as i64)),
+                high: Some(value),
                min: None,
                max: None,
            })
-            .context("failed to set memory.high")
-    }
-
-    /// Set cgroup memory.high and memory.max.
-    pub fn set_limits(&self, limits: &MemoryLimits) -> anyhow::Result<()> {
-        info!(
-            limits.high,
-            limits.max,
-            path = self.path(),
-            "writing new memory limits",
-        );
-        self.memory()
-            .context("failed to get memory subsystem while setting memory limits")?
-            .set_mem(cgroups_rs::memory::SetMemory {
-                min: None,
-                low: None,
-                high: Some(MaxValue::Value(
-                    u64::min(limits.high, i64::MAX as u64) as i64
-                )),
-                max: Some(MaxValue::Value(u64::min(limits.max, i64::MAX as u64) as i64)),
-            })
-            .context("failed to set memory limits")
-    }
-
-    /// Given some amount of available memory, set the desired cgroup memory limits
-    pub fn set_memory_limits(&mut self, available_memory: u64) -> anyhow::Result<()> {
-        let new_high = self.config.calculate_memory_high_value(available_memory);
-        let limits = MemoryLimits::new(new_high, available_memory);
-        info!(
-            path = self.path(),
-            memory = ?limits,
-            "setting cgroup memory",
-        );
-        self.set_limits(&limits)
-            .context("failed to set cgroup memory limits")?;
-        Ok(())
+            .map_err(anyhow::Error::from)
    }

    /// Get memory.high threshold.
-    pub fn get_high_bytes(&self) -> anyhow::Result<u64> {
+    pub fn get_memory_high_bytes(&self) -> anyhow::Result<u64> {
        let high = self
            .memory()
            .context("failed to get memory subsystem while getting memory statistics")?
--- a/libs/vm_monitor/src/runner.rs
+++ b/libs/vm_monitor/src/runner.rs
@@ -4,9 +4,9 @@
 //! This is the "Monitor" part of the monitor binary and is the main entrypoint for
 //! all functionality.

+use std::fmt::Debug;
 use std::sync::Arc;
 use std::time::{Duration, Instant};
-use std::{fmt::Debug, mem};

 use anyhow::{bail, Context};
 use axum::extract::ws::{Message, WebSocket};
@@ -16,7 +16,7 @@ use tokio::sync::mpsc;
 use tokio_util::sync::CancellationToken;
 use tracing::{error, info, warn};

-use crate::cgroup::{CgroupWatcher, MemoryLimits, Sequenced};
+use crate::cgroup::{CgroupWatcher, Sequenced};
 use crate::dispatcher::Dispatcher;
 use crate::filecache::{FileCacheConfig, FileCacheState};
 use crate::protocol::{InboundMsg, InboundMsgKind, OutboundMsg, OutboundMsgKind, Resources};
@@ -106,6 +106,43 @@ impl Runner {
            kill,
        };

+        // If we have both the cgroup and file cache integrations enabled, it's possible for
+        // temporary failures to result in cgroup throttling (from memory.high), that in turn makes
+        // it near-impossible to connect to the file cache (because it times out). Unfortunately,
+        // we *do* still want to determine the file cache size before setting the cgroup's
+        // memory.high, so it's not as simple as just swapping the order.
+        //
+        // Instead, the resolution here is that on vm-monitor startup (note: happens on each
+        // connection from autoscaler-agent, possibly multiple times per compute_ctl lifecycle), we
+        // temporarily unset memory.high, to allow any existing throttling to dissipate. It's a bit
+        // of a hacky solution, but helps with reliability.
+        if let Some(name) = &args.cgroup {
+            // Best not to set up cgroup stuff more than once, so we'll initialize cgroup state
+            // now, and then set limits later.
+            info!("initializing cgroup");
+
+            let (cgroup, cgroup_event_stream) = CgroupWatcher::new(name.clone(), requesting_send)
+                .context("failed to create cgroup manager")?;
+
+            info!("temporarily unsetting memory.high");
+
+            // Temporarily un-set cgroup memory.high; see above.
+            cgroup
+                .unset_memory_high()
+                .context("failed to unset memory.high")?;
+
+            let cgroup = Arc::new(cgroup);
+
+            let cgroup_clone = Arc::clone(&cgroup);
+            spawn_with_cancel(
+                token.clone(),
+                |_| error!("cgroup watcher terminated"),
+                async move { cgroup_clone.watch(notified_recv, cgroup_event_stream).await },
+            );
+
+            state.cgroup = Some(cgroup);
+        }
+
        let mut file_cache_reserved_bytes = 0;
        let mem = get_total_system_memory();

@@ -119,7 +156,7 @@ impl Runner {
                false => FileCacheConfig::default_in_memory(),
            };

-            let mut file_cache = FileCacheState::new(connstr, config, token.clone())
+            let mut file_cache = FileCacheState::new(connstr, config, token)
                .await
                .context("failed to create file cache")?;

@@ -152,35 +189,15 @@ impl Runner {
            state.filecache = Some(file_cache);
        }

-        if let Some(name) = &args.cgroup {
-            let (mut cgroup, cgroup_event_stream) =
-                CgroupWatcher::new(name.clone(), requesting_send)
-                    .context("failed to create cgroup manager")?;
-
+        if let Some(cgroup) = &state.cgroup {
            let available = mem - file_cache_reserved_bytes;
+            let value = cgroup.config.calculate_memory_high_value(available);
+
+            info!(value, "setting memory.high");

            cgroup
-                .set_memory_limits(available)
-                .context("failed to set cgroup memory limits")?;
-
-            let cgroup = Arc::new(cgroup);
-
-            // Some might call this . . . cgroup v2
-            let cgroup_clone = Arc::clone(&cgroup);
-
-            spawn_with_cancel(token, |_| error!("cgroup watcher terminated"), async move {
-                cgroup_clone.watch(notified_recv, cgroup_event_stream).await
-            });
-
-            state.cgroup = Some(cgroup);
-        } else {
-            // *NOTE*: We need to forget the sender so that its drop impl does not get ran.
-            // This allows us to poll it in `Monitor::run` regardless of whether we
-            // are managing a cgroup or not. If we don't forget it, all receives will
-            // immediately return an error because the sender is droped and it will
-            // claim all select! statements, effectively turning `Monitor::run` into
-            // `loop { fail to receive }`.
-            mem::forget(requesting_send);
+                .set_memory_high_bytes(value)
+                .context("failed to set cgroup memory.high")?;
        }

        Ok(state)
@@ -257,15 +274,11 @@ impl Runner {
                new_cgroup_mem_high = cgroup.config.calculate_memory_high_value(available_memory);
            }

-            let limits = MemoryLimits::new(
-                // new_cgroup_mem_high is initialized to 0 but it is guarancontextd to not be here
-                // since it is properly initialized in the previous cgroup if let block
-                new_cgroup_mem_high,
-                available_memory,
-            );
+            // new_cgroup_mem_high is initialized to 0 but it is guaranteed to not be here
+            // since it is properly initialized in the previous cgroup if let block
            cgroup
-                .set_limits(&limits)
-                .context("failed to set cgroup memory limits")?;
+                .set_memory_high_bytes(new_cgroup_mem_high)
+                .context("failed to set cgroup memory.high")?;

            let message = format!(
                "set cgroup memory.high to {} MiB, of new max {} MiB",
@@ -328,10 +341,9 @@ impl Runner {
                name = cgroup.path(),
                "updating cgroup memory.high",
            );
-            let limits = MemoryLimits::new(new_cgroup_mem_high, available_memory);
            cgroup
-                .set_limits(&limits)
-                .context("failed to set file cache size")?;
+                .set_memory_high_bytes(new_cgroup_mem_high)
+                .context("failed to set cgroup memory.high")?;
        }

        Ok(())
@@ -397,7 +409,7 @@ impl Runner {
                    }
                }
                // we need to propagate an upscale request
-                request = self.dispatcher.request_upscale_events.recv() => {
+                request = self.dispatcher.request_upscale_events.recv(), if self.cgroup.is_some() => {
                    if request.is_none() {
                        bail!("failed to listen for upscale event from cgroup")
                    }
--- a/pageserver/Cargo.toml
+++ b/pageserver/Cargo.toml
@@ -17,6 +17,8 @@ async-stream.workspace = true
 async-trait.workspace = true
 byteorder.workspace = true
 bytes.workspace = true
+camino.workspace = true
+camino-tempfile.workspace = true
 chrono = { workspace = true, features = ["serde"] }
 clap = { workspace = true, features = ["string"] }
 close_fds.workspace = true
@@ -84,7 +86,6 @@ strum_macros.workspace = true
 [dev-dependencies]
 criterion.workspace = true
 hex-literal.workspace = true
-tempfile.workspace = true
 tokio = { workspace = true, features = ["process", "sync", "fs", "rt", "io-util", "time", "test-util"] }

 [[bench]]
--- a/pageserver/benches/bench_walredo.rs
+++ b/pageserver/benches/bench_walredo.rs
@@ -25,7 +25,7 @@ fn redo_scenarios(c: &mut Criterion) {
    // input to the stderr.
    // utils::logging::init(utils::logging::LogFormat::Plain).unwrap();

-    let repo_dir = tempfile::tempdir_in(env!("CARGO_TARGET_TMPDIR")).unwrap();
+    let repo_dir = camino_tempfile::tempdir_in(env!("CARGO_TARGET_TMPDIR")).unwrap();

    let conf = PageServerConf::dummy_conf(repo_dir.path().to_path_buf());
    let conf = Box::leak(Box::new(conf));
--- a/pageserver/ctl/Cargo.toml
+++ b/pageserver/ctl/Cargo.toml
@@ -9,6 +9,7 @@ license.workspace = true
 [dependencies]
 anyhow.workspace = true
 bytes.workspace = true
+camino.workspace = true
 clap = { workspace = true, features = ["string"] }
 git-version.workspace = true
 pageserver = { path = ".." }
--- a/pageserver/ctl/src/layer_map_analyzer.rs
+++ b/pageserver/ctl/src/layer_map_analyzer.rs
@@ -3,11 +3,14 @@
 //! Currently it only analyzes holes, which are regions within the layer range that the layer contains no updates for. In the future it might do more analysis (maybe key quantiles?) but it should never return sensitive data.

 use anyhow::Result;
+use camino::{Utf8Path, Utf8PathBuf};
+use pageserver::context::{DownloadBehavior, RequestContext};
+use pageserver::task_mgr::TaskKind;
 use pageserver::tenant::{TENANTS_SEGMENT_NAME, TIMELINES_SEGMENT_NAME};
 use std::cmp::Ordering;
 use std::collections::BinaryHeap;
 use std::ops::Range;
-use std::{fs, path::Path, str};
+use std::{fs, str};

 use pageserver::page_cache::PAGE_SZ;
 use pageserver::repository::{Key, KEY_SIZE};
@@ -96,9 +99,9 @@ pub(crate) fn parse_filename(name: &str) -> Option<LayerFile> {
 }

 // Finds the max_holes largest holes, ignoring any that are smaller than MIN_HOLE_LENGTH"
-async fn get_holes(path: &Path, max_holes: usize) -> Result<Vec<Hole>> {
+async fn get_holes(path: &Utf8Path, max_holes: usize, ctx: &RequestContext) -> Result<Vec<Hole>> {
    let file = FileBlockReader::new(VirtualFile::open(path).await?);
-    let summary_blk = file.read_blk(0).await?;
+    let summary_blk = file.read_blk(0, ctx).await?;
    let actual_summary = Summary::des_prefix(summary_blk.as_ref())?;
    let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
        actual_summary.index_start_blk,
@@ -125,6 +128,7 @@ async fn get_holes(path: &Path, max_holes: usize) -> Result<Vec<Hole>> {
                prev_key = Some(curr.next());
                true
            },
+            ctx,
        )
        .await?;
    let mut holes = heap.into_vec();
@@ -135,6 +139,7 @@ async fn get_holes(path: &Path, max_holes: usize) -> Result<Vec<Hole>> {
 pub(crate) async fn main(cmd: &AnalyzeLayerMapCmd) -> Result<()> {
    let storage_path = &cmd.path;
    let max_holes = cmd.max_holes.unwrap_or(DEFAULT_MAX_HOLES);
+    let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);

    // Initialize virtual_file (file desriptor cache) and page cache which are needed to access layer persistent B-Tree.
    pageserver::virtual_file::init(10);
@@ -163,7 +168,9 @@ pub(crate) async fn main(cmd: &AnalyzeLayerMapCmd) -> Result<()> {
                    parse_filename(&layer.file_name().into_string().unwrap())
                {
                    if layer_file.is_delta {
-                        layer_file.holes = get_holes(&layer.path(), max_holes).await?;
+                        let layer_path =
+                            Utf8PathBuf::from_path_buf(layer.path()).expect("non-Unicode path");
+                        layer_file.holes = get_holes(&layer_path, max_holes, &ctx).await?;
                        n_deltas += 1;
                    }
                    layers.push(layer_file);
--- a/pageserver/ctl/src/layers.rs
+++ b/pageserver/ctl/src/layers.rs
@@ -1,7 +1,10 @@
 use std::path::{Path, PathBuf};

 use anyhow::Result;
+use camino::Utf8Path;
 use clap::Subcommand;
+use pageserver::context::{DownloadBehavior, RequestContext};
+use pageserver::task_mgr::TaskKind;
 use pageserver::tenant::block_io::BlockCursor;
 use pageserver::tenant::disk_btree::DiskBtreeReader;
 use pageserver::tenant::storage_layer::delta_layer::{BlobRef, Summary};
@@ -44,12 +47,12 @@ pub(crate) enum LayerCmd {
    },
 }

-async fn read_delta_file(path: impl AsRef<Path>) -> Result<()> {
-    let path = path.as_ref();
+async fn read_delta_file(path: impl AsRef<Path>, ctx: &RequestContext) -> Result<()> {
+    let path = Utf8Path::from_path(path.as_ref()).expect("non-Unicode path");
    virtual_file::init(10);
    page_cache::init(100);
    let file = FileBlockReader::new(VirtualFile::open(path).await?);
-    let summary_blk = file.read_blk(0).await?;
+    let summary_blk = file.read_blk(0, ctx).await?;
    let actual_summary = Summary::des_prefix(summary_blk.as_ref())?;
    let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
        actual_summary.index_start_blk,
@@ -67,11 +70,12 @@ async fn read_delta_file(path: impl AsRef<Path>) -> Result<()> {
                all.push((curr, BlobRef(value_offset)));
                true
            },
+            ctx,
        )
        .await?;
    let cursor = BlockCursor::new_fileblockreader(&file);
    for (k, v) in all {
-        let value = cursor.read_blob(v.pos()).await?;
+        let value = cursor.read_blob(v.pos(), ctx).await?;
        println!("key:{} value_len:{}", k, value.len());
    }
    // TODO(chi): special handling for last key?
@@ -79,6 +83,7 @@ async fn read_delta_file(path: impl AsRef<Path>) -> Result<()> {
 }

 pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
+    let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);
    match cmd {
        LayerCmd::List { path } => {
            for tenant in fs::read_dir(path.join(TENANTS_SEGMENT_NAME))? {
@@ -153,7 +158,7 @@ pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
                        );

                        if layer_file.is_delta {
-                            read_delta_file(layer.path()).await?;
+                            read_delta_file(layer.path(), &ctx).await?;
                        } else {
                            anyhow::bail!("not supported yet :(");
                        }
--- a/pageserver/ctl/src/main.rs
+++ b/pageserver/ctl/src/main.rs
@@ -8,6 +8,7 @@ mod draw_timeline_dir;
 mod layer_map_analyzer;
 mod layers;

+use camino::{Utf8Path, Utf8PathBuf};
 use clap::{Parser, Subcommand};
 use layers::LayerCmd;
 use pageserver::{
@@ -18,7 +19,6 @@ use pageserver::{
    virtual_file,
 };
 use postgres_ffi::ControlFileData;
-use std::path::{Path, PathBuf};
 use utils::{lsn::Lsn, project_git_version};

 project_git_version!(GIT_VERSION);
@@ -49,7 +49,7 @@ enum Commands {
 #[derive(Parser)]
 struct MetadataCmd {
    /// Input metadata file path
-    metadata_path: PathBuf,
+    metadata_path: Utf8PathBuf,
    /// Replace disk consistent Lsn
    disk_consistent_lsn: Option<Lsn>,
    /// Replace previous record Lsn
@@ -61,13 +61,13 @@ struct MetadataCmd {
 #[derive(Parser)]
 struct PrintLayerFileCmd {
    /// Pageserver data path
-    path: PathBuf,
+    path: Utf8PathBuf,
 }

 #[derive(Parser)]
 struct AnalyzeLayerMapCmd {
    /// Pageserver data path
-    path: PathBuf,
+    path: Utf8PathBuf,
    /// Max holes
    max_holes: Option<usize>,
 }
@@ -102,7 +102,7 @@ async fn main() -> anyhow::Result<()> {
    Ok(())
 }

-fn read_pg_control_file(control_file_path: &Path) -> anyhow::Result<()> {
+fn read_pg_control_file(control_file_path: &Utf8Path) -> anyhow::Result<()> {
    let control_file = ControlFileData::decode(&std::fs::read(control_file_path)?)?;
    println!("{control_file:?}");
    let control_file_initdb = Lsn(control_file.checkPoint);
@@ -114,7 +114,7 @@ fn read_pg_control_file(control_file_path: &Path) -> anyhow::Result<()> {
    Ok(())
 }

-async fn print_layerfile(path: &Path) -> anyhow::Result<()> {
+async fn print_layerfile(path: &Utf8Path) -> anyhow::Result<()> {
    // Basic initialization of things that don't change after startup
    virtual_file::init(10);
    page_cache::init(100);
--- a/pageserver/src/bin/pageserver.rs
+++ b/pageserver/src/bin/pageserver.rs
@@ -2,12 +2,14 @@

 use std::env::{var, VarError};
 use std::sync::Arc;
-use std::{env, ops::ControlFlow, path::Path, str::FromStr};
+use std::{env, ops::ControlFlow, str::FromStr};

 use anyhow::{anyhow, Context};
+use camino::Utf8Path;
 use clap::{Arg, ArgAction, Command};

 use metrics::launch_timestamp::{set_launch_timestamp_metric, LaunchTimestamp};
+use pageserver::control_plane_client::ControlPlaneClient;
 use pageserver::disk_usage_eviction_task::{self, launch_disk_usage_global_eviction_task};
 use pageserver::metrics::{STARTUP_DURATION, STARTUP_IS_LOADING};
 use pageserver::task_mgr::WALRECEIVER_RUNTIME;
@@ -20,6 +22,7 @@ use metrics::set_build_info_metric;
 use pageserver::{
    config::{defaults::*, PageServerConf},
    context::{DownloadBehavior, RequestContext},
+    deletion_queue::DeletionQueue,
    http, page_cache, page_service, task_mgr,
    task_mgr::TaskKind,
    task_mgr::{BACKGROUND_RUNTIME, COMPUTE_REQUEST_RUNTIME, MGMT_REQUEST_RUNTIME},
@@ -63,21 +66,17 @@ fn main() -> anyhow::Result<()> {

    let workdir = arg_matches
        .get_one::<String>("workdir")
-        .map(Path::new)
-        .unwrap_or_else(|| Path::new(".neon"));
+        .map(Utf8Path::new)
+        .unwrap_or_else(|| Utf8Path::new(".neon"));
    let workdir = workdir
-        .canonicalize()
-        .with_context(|| format!("Error opening workdir '{}'", workdir.display()))?;
+        .canonicalize_utf8()
+        .with_context(|| format!("Error opening workdir '{workdir}'"))?;

    let cfg_file_path = workdir.join("pageserver.toml");

    // Set CWD to workdir for non-daemon modes
-    env::set_current_dir(&workdir).with_context(|| {
-        format!(
-            "Failed to set application's current dir to '{}'",
-            workdir.display()
-        )
-    })?;
+    env::set_current_dir(&workdir)
+        .with_context(|| format!("Failed to set application's current dir to '{workdir}'"))?;

    let conf = match initialize_config(&cfg_file_path, arg_matches, &workdir)? {
        ControlFlow::Continue(conf) => conf,
@@ -113,12 +112,8 @@ fn main() -> anyhow::Result<()> {

    let tenants_path = conf.tenants_path();
    if !tenants_path.exists() {
-        utils::crashsafe::create_dir_all(conf.tenants_path()).with_context(|| {
-            format!(
-                "Failed to create tenants root dir at '{}'",
-                tenants_path.display()
-            )
-        })?;
+        utils::crashsafe::create_dir_all(conf.tenants_path())
+            .with_context(|| format!("Failed to create tenants root dir at '{tenants_path}'"))?;
    }

    // Initialize up failpoints support
@@ -135,9 +130,9 @@ fn main() -> anyhow::Result<()> {
 }

 fn initialize_config(
-    cfg_file_path: &Path,
+    cfg_file_path: &Utf8Path,
    arg_matches: clap::ArgMatches,
-    workdir: &Path,
+    workdir: &Utf8Path,
 ) -> anyhow::Result<ControlFlow<(), &'static PageServerConf>> {
    let init = arg_matches.get_flag("init");
    let update_config = init || arg_matches.get_flag("update-config");
@@ -145,33 +140,22 @@ fn initialize_config(
    let (mut toml, config_file_exists) = if cfg_file_path.is_file() {
        if init {
            anyhow::bail!(
-                "Config file '{}' already exists, cannot init it, use --update-config to update it",
-                cfg_file_path.display()
+                "Config file '{cfg_file_path}' already exists, cannot init it, use --update-config to update it",
            );
        }
        // Supplement the CLI arguments with the config file
-        let cfg_file_contents = std::fs::read_to_string(cfg_file_path).with_context(|| {
-            format!(
-                "Failed to read pageserver config at '{}'",
-                cfg_file_path.display()
-            )
-        })?;
+        let cfg_file_contents = std::fs::read_to_string(cfg_file_path)
+            .with_context(|| format!("Failed to read pageserver config at '{cfg_file_path}'"))?;
        (
            cfg_file_contents
                .parse::<toml_edit::Document>()
                .with_context(|| {
-                    format!(
-                        "Failed to parse '{}' as pageserver config",
-                        cfg_file_path.display()
-                    )
+                    format!("Failed to parse '{cfg_file_path}' as pageserver config")
                })?,
            true,
        )
    } else if cfg_file_path.exists() {
-        anyhow::bail!(
-            "Config file '{}' exists but is not a regular file",
-            cfg_file_path.display()
-        );
+        anyhow::bail!("Config file '{cfg_file_path}' exists but is not a regular file");
    } else {
        // We're initializing the tenant, so there's no config file yet
        (
@@ -190,7 +174,7 @@ fn initialize_config(

            for (key, item) in doc.iter() {
                if config_file_exists && update_config && key == "id" && toml.contains_key(key) {
-                    anyhow::bail!("Pageserver config file exists at '{}' and has node id already, it cannot be overridden", cfg_file_path.display());
+                    anyhow::bail!("Pageserver config file exists at '{cfg_file_path}' and has node id already, it cannot be overridden");
                }
                toml.insert(key, item.clone());
            }
@@ -202,18 +186,11 @@ fn initialize_config(
        .context("Failed to parse pageserver configuration")?;

    if update_config {
-        info!("Writing pageserver config to '{}'", cfg_file_path.display());
+        info!("Writing pageserver config to '{cfg_file_path}'");

-        std::fs::write(cfg_file_path, toml.to_string()).with_context(|| {
-            format!(
-                "Failed to write pageserver config to '{}'",
-                cfg_file_path.display()
-            )
-        })?;
-        info!(
-            "Config successfully written to '{}'",
-            cfg_file_path.display()
-        )
+        std::fs::write(cfg_file_path, toml.to_string())
+            .with_context(|| format!("Failed to write pageserver config to '{cfg_file_path}'"))?;
+        info!("Config successfully written to '{cfg_file_path}'")
    }

    Ok(if init {
@@ -346,9 +323,22 @@ fn start_pageserver(
        }
    };

+    // Top-level cancellation token for the process
+    let shutdown_pageserver = tokio_util::sync::CancellationToken::new();
+
    // Set up remote storage client
    let remote_storage = create_remote_storage_client(conf)?;

+    // Set up deletion queue
+    let (deletion_queue, deletion_workers) = DeletionQueue::new(
+        remote_storage.clone(),
+        ControlPlaneClient::new(conf, &shutdown_pageserver),
+        conf,
+    );
+    if let Some(deletion_workers) = deletion_workers {
+        deletion_workers.spawn_with(BACKGROUND_RUNTIME.handle());
+    }
+
    // Up to this point no significant I/O has been done: this should have been fast.  Record
    // duration prior to starting I/O intensive phase of startup.
    startup_checkpoint("initial", "Starting loading tenants");
@@ -379,13 +369,13 @@ fn start_pageserver(
    };

    // Scan the local 'tenants/' directory and start loading the tenants
-    let shutdown_pageserver = tokio_util::sync::CancellationToken::new();
-
+    let deletion_queue_client = deletion_queue.new_client();
    BACKGROUND_RUNTIME.block_on(mgr::init_tenant_mgr(
        conf,
        TenantSharedResources {
            broker_client: broker_client.clone(),
            remote_storage: remote_storage.clone(),
+            deletion_queue_client,
        },
        order,
        shutdown_pageserver.clone(),
@@ -481,9 +471,10 @@ fn start_pageserver(
            http::routes::State::new(
                conf,
                http_auth.clone(),
-                remote_storage,
+                remote_storage.clone(),
                broker_client.clone(),
                disk_usage_eviction_state,
+                deletion_queue.new_client(),
            )
            .context("Failed to initialize router state")?,
        );
@@ -518,6 +509,9 @@ fn start_pageserver(
            // creates a child context with the right DownloadBehavior.
            DownloadBehavior::Error,
        );
+
+        let local_disk_storage = conf.workdir.join("last_consumption_metrics.json");
+
        task_mgr::spawn(
            crate::BACKGROUND_RUNTIME.handle(),
            TaskKind::MetricsCollection,
@@ -544,6 +538,7 @@ fn start_pageserver(
                    conf.cached_metric_collection_interval,
                    conf.synthetic_size_calculation_interval,
                    conf.id,
+                    local_disk_storage,
                    metrics_ctx,
                )
                .instrument(info_span!("metrics_collection"))
@@ -607,7 +602,12 @@ fn start_pageserver(
            // Right now that tree doesn't reach very far, and `task_mgr` is used instead.
            // The plan is to change that over time.
            shutdown_pageserver.take();
-            BACKGROUND_RUNTIME.block_on(pageserver::shutdown_pageserver(0));
+            let bg_remote_storage = remote_storage.clone();
+            let bg_deletion_queue = deletion_queue.clone();
+            BACKGROUND_RUNTIME.block_on(pageserver::shutdown_pageserver(
+                bg_remote_storage.map(|_| bg_deletion_queue),
+                0,
+            ));
            unreachable!()
        }
    })
@@ -619,7 +619,7 @@ fn create_remote_storage_client(
    let config = if let Some(config) = &conf.remote_storage_config {
        config
    } else {
-        // No remote storage configured.
+        tracing::warn!("no remote storage configured, this is a deprecated configuration");
        return Ok(None);
    };

--- a/pageserver/src/config.rs
+++ b/pageserver/src/config.rs
@@ -11,17 +11,18 @@ use std::env;
 use storage_broker::Uri;
 use utils::crashsafe::path_with_suffix_extension;
 use utils::id::ConnectionId;
+use utils::logging::SecretString;

 use once_cell::sync::OnceCell;
 use reqwest::Url;
 use std::num::NonZeroUsize;
-use std::path::{Path, PathBuf};
 use std::str::FromStr;
 use std::sync::Arc;
 use std::time::Duration;
 use toml_edit;
 use toml_edit::{Document, Item};

+use camino::{Utf8Path, Utf8PathBuf};
 use postgres_backend::AuthType;
 use utils::{
    id::{NodeId, TenantId, TimelineId},
@@ -36,8 +37,8 @@ use crate::tenant::{
    TIMELINES_SEGMENT_NAME,
 };
 use crate::{
-    IGNORED_TENANT_FILE_NAME, METADATA_FILE_NAME, TENANT_CONFIG_NAME, TIMELINE_DELETE_MARK_SUFFIX,
-    TIMELINE_UNINIT_MARK_SUFFIX,
+    IGNORED_TENANT_FILE_NAME, METADATA_FILE_NAME, TENANT_CONFIG_NAME, TENANT_LOCATION_CONFIG_NAME,
+    TIMELINE_DELETE_MARK_SUFFIX, TIMELINE_UNINIT_MARK_SUFFIX,
 };

 pub mod defaults {
@@ -64,7 +65,7 @@ pub mod defaults {
        super::ConfigurableSemaphore::DEFAULT_INITIAL.get();

    pub const DEFAULT_METRIC_COLLECTION_INTERVAL: &str = "10 min";
-    pub const DEFAULT_CACHED_METRIC_COLLECTION_INTERVAL: &str = "1 hour";
+    pub const DEFAULT_CACHED_METRIC_COLLECTION_INTERVAL: &str = "0s";
    pub const DEFAULT_METRIC_COLLECTION_ENDPOINT: Option<reqwest::Url> = None;
    pub const DEFAULT_SYNTHETIC_SIZE_CALCULATION_INTERVAL: &str = "10 min";
    pub const DEFAULT_BACKGROUND_TASK_MAXIMUM_DELAY: &str = "10s";
@@ -152,9 +153,9 @@ pub struct PageServerConf {
    // that during unit testing, because the current directory is global
    // to the process but different unit tests work on different
    // repositories.
-    pub workdir: PathBuf,
+    pub workdir: Utf8PathBuf,

-    pub pg_distrib_dir: PathBuf,
+    pub pg_distrib_dir: Utf8PathBuf,

    // Authentication
    /// authentication method for the HTTP mgmt API
@@ -163,7 +164,7 @@ pub struct PageServerConf {
    pub pg_auth_type: AuthType,
    /// Path to a file containing public key for verifying JWT tokens.
    /// Used for both mgmt and compute auth, if enabled.
-    pub auth_validation_public_key_path: Option<PathBuf>,
+    pub auth_validation_public_key_path: Option<Utf8PathBuf>,

    pub remote_storage_config: Option<RemoteStorageConfig>,

@@ -207,6 +208,9 @@ pub struct PageServerConf {
    pub background_task_maximum_delay: Duration,

    pub control_plane_api: Option<Url>,
+
+    /// JWT token for use with the control plane API.
+    pub control_plane_api_token: Option<SecretString>,
 }

 /// We do not want to store this in a PageServerConf because the latter may be logged
@@ -249,15 +253,15 @@ struct PageServerConfigBuilder {
    page_cache_size: BuilderValue<usize>,
    max_file_descriptors: BuilderValue<usize>,

-    workdir: BuilderValue<PathBuf>,
+    workdir: BuilderValue<Utf8PathBuf>,

-    pg_distrib_dir: BuilderValue<PathBuf>,
+    pg_distrib_dir: BuilderValue<Utf8PathBuf>,

    http_auth_type: BuilderValue<AuthType>,
    pg_auth_type: BuilderValue<AuthType>,

    //
-    auth_validation_public_key_path: BuilderValue<Option<PathBuf>>,
+    auth_validation_public_key_path: BuilderValue<Option<Utf8PathBuf>>,
    remote_storage_config: BuilderValue<Option<RemoteStorageConfig>>,

    id: BuilderValue<NodeId>,
@@ -283,6 +287,7 @@ struct PageServerConfigBuilder {
    background_task_maximum_delay: BuilderValue<Duration>,

    control_plane_api: BuilderValue<Option<Url>>,
+    control_plane_api_token: BuilderValue<Option<SecretString>>,
 }

 impl Default for PageServerConfigBuilder {
@@ -300,10 +305,12 @@ impl Default for PageServerConfigBuilder {
            superuser: Set(DEFAULT_SUPERUSER.to_string()),
            page_cache_size: Set(DEFAULT_PAGE_CACHE_SIZE),
            max_file_descriptors: Set(DEFAULT_MAX_FILE_DESCRIPTORS),
-            workdir: Set(PathBuf::new()),
-            pg_distrib_dir: Set(env::current_dir()
-                .expect("cannot access current directory")
-                .join("pg_install")),
+            workdir: Set(Utf8PathBuf::new()),
+            pg_distrib_dir: Set(Utf8PathBuf::from_path_buf(
+                env::current_dir().expect("cannot access current directory"),
+            )
+            .expect("non-Unicode path")
+            .join("pg_install")),
            http_auth_type: Set(AuthType::Trust),
            pg_auth_type: Set(AuthType::Trust),
            auth_validation_public_key_path: Set(None),
@@ -347,6 +354,7 @@ impl Default for PageServerConfigBuilder {
            .unwrap()),

            control_plane_api: Set(None),
+            control_plane_api_token: Set(None),
        }
    }
 }
@@ -384,11 +392,11 @@ impl PageServerConfigBuilder {
        self.max_file_descriptors = BuilderValue::Set(max_file_descriptors)
    }

-    pub fn workdir(&mut self, workdir: PathBuf) {
+    pub fn workdir(&mut self, workdir: Utf8PathBuf) {
        self.workdir = BuilderValue::Set(workdir)
    }

-    pub fn pg_distrib_dir(&mut self, pg_distrib_dir: PathBuf) {
+    pub fn pg_distrib_dir(&mut self, pg_distrib_dir: Utf8PathBuf) {
        self.pg_distrib_dir = BuilderValue::Set(pg_distrib_dir)
    }

@@ -402,7 +410,7 @@ impl PageServerConfigBuilder {

    pub fn auth_validation_public_key_path(
        &mut self,
-        auth_validation_public_key_path: Option<PathBuf>,
+        auth_validation_public_key_path: Option<Utf8PathBuf>,
    ) {
        self.auth_validation_public_key_path = BuilderValue::Set(auth_validation_public_key_path)
    }
@@ -475,8 +483,12 @@ impl PageServerConfigBuilder {
        self.background_task_maximum_delay = BuilderValue::Set(delay);
    }

-    pub fn control_plane_api(&mut self, api: Url) {
-        self.control_plane_api = BuilderValue::Set(Some(api))
+    pub fn control_plane_api(&mut self, api: Option<Url>) {
+        self.control_plane_api = BuilderValue::Set(api)
+    }
+
+    pub fn control_plane_api_token(&mut self, token: Option<SecretString>) {
+        self.control_plane_api_token = BuilderValue::Set(token)
    }

    pub fn build(self) -> anyhow::Result<PageServerConf> {
@@ -567,6 +579,9 @@ impl PageServerConfigBuilder {
            control_plane_api: self
                .control_plane_api
                .ok_or(anyhow!("missing control_plane_api"))?,
+            control_plane_api_token: self
+                .control_plane_api_token
+                .ok_or(anyhow!("missing control_plane_api_token"))?,
        })
    }
 }
@@ -576,34 +591,63 @@ impl PageServerConf {
    // Repository paths, relative to workdir.
    //

-    pub fn tenants_path(&self) -> PathBuf {
+    pub fn tenants_path(&self) -> Utf8PathBuf {
        self.workdir.join(TENANTS_SEGMENT_NAME)
    }

-    pub fn tenant_path(&self, tenant_id: &TenantId) -> PathBuf {
+    pub fn deletion_prefix(&self) -> Utf8PathBuf {
+        self.workdir.join("deletion")
+    }
+
+    pub fn deletion_list_path(&self, sequence: u64) -> Utf8PathBuf {
+        // Encode a version in the filename, so that if we ever switch away from JSON we can
+        // increment this.
+        const VERSION: u8 = 1;
+
+        self.deletion_prefix()
+            .join(format!("{sequence:016x}-{VERSION:02x}.list"))
+    }
+
+    pub fn deletion_header_path(&self) -> Utf8PathBuf {
+        // Encode a version in the filename, so that if we ever switch away from JSON we can
+        // increment this.
+        const VERSION: u8 = 1;
+
+        self.deletion_prefix().join(format!("header-{VERSION:02x}"))
+    }
+
+    pub fn tenant_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
        self.tenants_path().join(tenant_id.to_string())
    }

-    pub fn tenant_attaching_mark_file_path(&self, tenant_id: &TenantId) -> PathBuf {
+    pub fn tenant_attaching_mark_file_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
        self.tenant_path(tenant_id)
            .join(TENANT_ATTACHING_MARKER_FILENAME)
    }

-    pub fn tenant_ignore_mark_file_path(&self, tenant_id: &TenantId) -> PathBuf {
+    pub fn tenant_ignore_mark_file_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
        self.tenant_path(tenant_id).join(IGNORED_TENANT_FILE_NAME)
    }

    /// Points to a place in pageserver's local directory,
    /// where certain tenant's tenantconf file should be located.
-    pub fn tenant_config_path(&self, tenant_id: &TenantId) -> PathBuf {
+    ///
+    /// Legacy: superseded by tenant_location_config_path.  Eventually
+    /// remove this function.
+    pub fn tenant_config_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
        self.tenant_path(tenant_id).join(TENANT_CONFIG_NAME)
    }

-    pub fn timelines_path(&self, tenant_id: &TenantId) -> PathBuf {
+    pub fn tenant_location_config_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
+        self.tenant_path(tenant_id)
+            .join(TENANT_LOCATION_CONFIG_NAME)
+    }
+
+    pub fn timelines_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
        self.tenant_path(tenant_id).join(TIMELINES_SEGMENT_NAME)
    }

-    pub fn timeline_path(&self, tenant_id: &TenantId, timeline_id: &TimelineId) -> PathBuf {
+    pub fn timeline_path(&self, tenant_id: &TenantId, timeline_id: &TimelineId) -> Utf8PathBuf {
        self.timelines_path(tenant_id).join(timeline_id.to_string())
    }

@@ -611,7 +655,7 @@ impl PageServerConf {
        &self,
        tenant_id: TenantId,
        timeline_id: TimelineId,
-    ) -> PathBuf {
+    ) -> Utf8PathBuf {
        path_with_suffix_extension(
            self.timeline_path(&tenant_id, &timeline_id),
            TIMELINE_UNINIT_MARK_SUFFIX,
@@ -622,19 +666,19 @@ impl PageServerConf {
        &self,
        tenant_id: TenantId,
        timeline_id: TimelineId,
-    ) -> PathBuf {
+    ) -> Utf8PathBuf {
        path_with_suffix_extension(
            self.timeline_path(&tenant_id, &timeline_id),
            TIMELINE_DELETE_MARK_SUFFIX,
        )
    }

-    pub fn tenant_deleted_mark_file_path(&self, tenant_id: &TenantId) -> PathBuf {
+    pub fn tenant_deleted_mark_file_path(&self, tenant_id: &TenantId) -> Utf8PathBuf {
        self.tenant_path(tenant_id)
            .join(TENANT_DELETED_MARKER_FILE_NAME)
    }

-    pub fn traces_path(&self) -> PathBuf {
+    pub fn traces_path(&self) -> Utf8PathBuf {
        self.workdir.join("traces")
    }

@@ -643,7 +687,7 @@ impl PageServerConf {
        tenant_id: &TenantId,
        timeline_id: &TimelineId,
        connection_id: &ConnectionId,
-    ) -> PathBuf {
+    ) -> Utf8PathBuf {
        self.traces_path()
            .join(tenant_id.to_string())
            .join(timeline_id.to_string())
@@ -652,20 +696,20 @@ impl PageServerConf {

    /// Points to a place in pageserver's local directory,
    /// where certain timeline's metadata file should be located.
-    pub fn metadata_path(&self, tenant_id: &TenantId, timeline_id: &TimelineId) -> PathBuf {
+    pub fn metadata_path(&self, tenant_id: &TenantId, timeline_id: &TimelineId) -> Utf8PathBuf {
        self.timeline_path(tenant_id, timeline_id)
            .join(METADATA_FILE_NAME)
    }

    /// Turns storage remote path of a file into its local path.
-    pub fn local_path(&self, remote_path: &RemotePath) -> PathBuf {
+    pub fn local_path(&self, remote_path: &RemotePath) -> Utf8PathBuf {
        remote_path.with_base(&self.workdir)
    }

    //
    // Postgres distribution paths
    //
-    pub fn pg_distrib_dir(&self, pg_version: u32) -> anyhow::Result<PathBuf> {
+    pub fn pg_distrib_dir(&self, pg_version: u32) -> anyhow::Result<Utf8PathBuf> {
        let path = self.pg_distrib_dir.clone();

        #[allow(clippy::manual_range_patterns)]
@@ -675,10 +719,10 @@ impl PageServerConf {
        }
    }

-    pub fn pg_bin_dir(&self, pg_version: u32) -> anyhow::Result<PathBuf> {
+    pub fn pg_bin_dir(&self, pg_version: u32) -> anyhow::Result<Utf8PathBuf> {
        Ok(self.pg_distrib_dir(pg_version)?.join("bin"))
    }
-    pub fn pg_lib_dir(&self, pg_version: u32) -> anyhow::Result<PathBuf> {
+    pub fn pg_lib_dir(&self, pg_version: u32) -> anyhow::Result<Utf8PathBuf> {
        Ok(self.pg_distrib_dir(pg_version)?.join("lib"))
    }

@@ -686,7 +730,7 @@ impl PageServerConf {
    /// validating the input and failing on errors.
    ///
    /// This leaves any options not present in the file in the built-in defaults.
-    pub fn parse_and_validate(toml: &Document, workdir: &Path) -> anyhow::Result<Self> {
+    pub fn parse_and_validate(toml: &Document, workdir: &Utf8Path) -> anyhow::Result<Self> {
        let mut builder = PageServerConfigBuilder::default();
        builder.workdir(workdir.to_owned());

@@ -705,10 +749,10 @@ impl PageServerConf {
                    builder.max_file_descriptors(parse_toml_u64(key, item)? as usize)
                }
                "pg_distrib_dir" => {
-                    builder.pg_distrib_dir(PathBuf::from(parse_toml_string(key, item)?))
+                    builder.pg_distrib_dir(Utf8PathBuf::from(parse_toml_string(key, item)?))
                }
                "auth_validation_public_key_path" => builder.auth_validation_public_key_path(Some(
-                    PathBuf::from(parse_toml_string(key, item)?),
+                    Utf8PathBuf::from(parse_toml_string(key, item)?),
                )),
                "http_auth_type" => builder.http_auth_type(parse_toml_from_str(key, item)?),
                "pg_auth_type" => builder.pg_auth_type(parse_toml_from_str(key, item)?),
@@ -747,7 +791,22 @@ impl PageServerConf {
                },
                "ondemand_download_behavior_treat_error_as_warn" => builder.ondemand_download_behavior_treat_error_as_warn(parse_toml_bool(key, item)?),
                "background_task_maximum_delay" => builder.background_task_maximum_delay(parse_toml_duration(key, item)?),
-                "control_plane_api" => builder.control_plane_api(parse_toml_string(key, item)?.parse().context("failed to parse control plane URL")?),
+                "control_plane_api" => {
+                    let parsed = parse_toml_string(key, item)?;
+                    if parsed.is_empty() {
+                        builder.control_plane_api(None)
+                    } else {
+                        builder.control_plane_api(Some(parsed.parse().context("failed to parse control plane URL")?))
+                    }
+                },
+                "control_plane_api_token" => {
+                    let parsed = parse_toml_string(key, item)?;
+                    if parsed.is_empty() {
+                        builder.control_plane_api_token(None)
+                    } else {
+                        builder.control_plane_api_token(Some(parsed.into()))
+                    }
+                },
                _ => bail!("unrecognized pageserver option '{key}'"),
            }
        }
@@ -761,8 +820,7 @@ impl PageServerConf {
            ensure!(
                auth_validation_public_key_path.exists(),
                format!(
-                    "Can't find auth_validation_public_key at '{}'",
-                    auth_validation_public_key_path.display()
+                    "Can't find auth_validation_public_key at '{auth_validation_public_key_path}'",
                )
            );
        }
@@ -878,12 +936,12 @@ impl PageServerConf {
    }

    #[cfg(test)]
-    pub fn test_repo_dir(test_name: &str) -> PathBuf {
-        PathBuf::from(format!("../tmp_check/test_{test_name}"))
+    pub fn test_repo_dir(test_name: &str) -> Utf8PathBuf {
+        Utf8PathBuf::from(format!("../tmp_check/test_{test_name}"))
    }

-    pub fn dummy_conf(repo_dir: PathBuf) -> Self {
-        let pg_distrib_dir = PathBuf::from(env!("CARGO_MANIFEST_DIR")).join("../pg_install");
+    pub fn dummy_conf(repo_dir: Utf8PathBuf) -> Self {
+        let pg_distrib_dir = Utf8PathBuf::from(env!("CARGO_MANIFEST_DIR")).join("../pg_install");

        PageServerConf {
            id: NodeId(0),
@@ -917,6 +975,7 @@ impl PageServerConf {
            ondemand_download_behavior_treat_error_as_warn: false,
            background_task_maximum_delay: Duration::ZERO,
            control_plane_api: None,
+            control_plane_api_token: None,
        }
    }
 }
@@ -1049,8 +1108,8 @@ mod tests {
        num::{NonZeroU32, NonZeroUsize},
    };

+    use camino_tempfile::{tempdir, Utf8TempDir};
    use remote_storage::{RemoteStorageKind, S3Config};
-    use tempfile::{tempdir, TempDir};
    use utils::serde_percent::Percent;

    use super::*;
@@ -1089,8 +1148,7 @@ background_task_maximum_delay = '334 s'
        let broker_endpoint = storage_broker::DEFAULT_ENDPOINT;
        // we have to create dummy values to overcome the validation errors
        let config_string = format!(
-            "pg_distrib_dir='{}'\nid=10\nbroker_endpoint = '{broker_endpoint}'",
-            pg_distrib_dir.display()
+            "pg_distrib_dir='{pg_distrib_dir}'\nid=10\nbroker_endpoint = '{broker_endpoint}'",
        );
        let toml = config_string.parse()?;

@@ -1140,7 +1198,8 @@ background_task_maximum_delay = '334 s'
                background_task_maximum_delay: humantime::parse_duration(
                    defaults::DEFAULT_BACKGROUND_TASK_MAXIMUM_DELAY
                )?,
-                control_plane_api: None
+                control_plane_api: None,
+                control_plane_api_token: None
            },
            "Correct defaults should be used when no config values are provided"
        );
@@ -1155,8 +1214,7 @@ background_task_maximum_delay = '334 s'
        let broker_endpoint = storage_broker::DEFAULT_ENDPOINT;

        let config_string = format!(
-            "{ALL_BASE_VALUES_TOML}pg_distrib_dir='{}'\nbroker_endpoint = '{broker_endpoint}'",
-            pg_distrib_dir.display()
+            "{ALL_BASE_VALUES_TOML}pg_distrib_dir='{pg_distrib_dir}'\nbroker_endpoint = '{broker_endpoint}'",
        );
        let toml = config_string.parse()?;

@@ -1196,7 +1254,8 @@ background_task_maximum_delay = '334 s'
                test_remote_failures: 0,
                ondemand_download_behavior_treat_error_as_warn: false,
                background_task_maximum_delay: Duration::from_secs(334),
-                control_plane_api: None
+                control_plane_api: None,
+                control_plane_api_token: None
            },
            "Should be able to parse all basic config values correctly"
        );
@@ -1215,23 +1274,18 @@ background_task_maximum_delay = '334 s'
        let identical_toml_declarations = &[
            format!(
                r#"[remote_storage]
-local_path = '{}'"#,
-                local_storage_path.display()
-            ),
-            format!(
-                "remote_storage={{local_path='{}'}}",
-                local_storage_path.display()
+local_path = '{local_storage_path}'"#,
            ),
+            format!("remote_storage={{local_path='{local_storage_path}'}}"),
        ];

        for remote_storage_config_str in identical_toml_declarations {
            let config_string = format!(
                r#"{ALL_BASE_VALUES_TOML}
-pg_distrib_dir='{}'
+pg_distrib_dir='{pg_distrib_dir}'
 broker_endpoint = '{broker_endpoint}'

 {remote_storage_config_str}"#,
-                pg_distrib_dir.display(),
            );

            let toml = config_string.parse()?;
@@ -1294,11 +1348,10 @@ concurrency_limit = {s3_concurrency_limit}"#
        for remote_storage_config_str in identical_toml_declarations {
            let config_string = format!(
                r#"{ALL_BASE_VALUES_TOML}
-pg_distrib_dir='{}'
+pg_distrib_dir='{pg_distrib_dir}'
 broker_endpoint = '{broker_endpoint}'

 {remote_storage_config_str}"#,
-                pg_distrib_dir.display(),
            );

            let toml = config_string.parse()?;
@@ -1340,12 +1393,11 @@ broker_endpoint = '{broker_endpoint}'

        let config_string = format!(
            r#"{ALL_BASE_VALUES_TOML}
-pg_distrib_dir='{}'
+pg_distrib_dir='{pg_distrib_dir}'
 broker_endpoint = '{broker_endpoint}'

 [tenant_config]
 trace_read_requests = {trace_read_requests}"#,
-            pg_distrib_dir.display(),
        );

        let toml = config_string.parse()?;
@@ -1365,7 +1417,7 @@ trace_read_requests = {trace_read_requests}"#,
        let (workdir, pg_distrib_dir) = prepare_fs(&tempdir)?;

        let pageserver_conf_toml = format!(
-            r#"pg_distrib_dir = "{}"
+            r#"pg_distrib_dir = "{pg_distrib_dir}"
 metric_collection_endpoint = "http://sample.url"
 metric_collection_interval = "10min"
 id = 222
@@ -1383,7 +1435,6 @@ kind = "LayerAccessThreshold"
 period = "20m"
 threshold = "20m"
 "#,
-            pg_distrib_dir.display(),
        );
        let toml: Document = pageserver_conf_toml.parse()?;
        let conf = PageServerConf::parse_and_validate(&toml, &workdir)?;
@@ -1424,7 +1475,7 @@ threshold = "20m"
        Ok(())
    }

-    fn prepare_fs(tempdir: &TempDir) -> anyhow::Result<(PathBuf, PathBuf)> {
+    fn prepare_fs(tempdir: &Utf8TempDir) -> anyhow::Result<(Utf8PathBuf, Utf8PathBuf)> {
        let tempdir_path = tempdir.path();

        let workdir = tempdir_path.join("workdir");
--- a/pageserver/src/consumption_metrics.rs
+++ b/pageserver/src/consumption_metrics.rs
@@ -1,188 +1,54 @@
-//!
 //! Periodically collect consumption metrics for all active tenants
 //! and push them to a HTTP endpoint.
-//! Cache metrics to send only the updated ones.
-//!
 use crate::context::{DownloadBehavior, RequestContext};
 use crate::task_mgr::{self, TaskKind, BACKGROUND_RUNTIME};
 use crate::tenant::{mgr, LogicalSizeCalculationCause};
-use anyhow;
-use chrono::{DateTime, Utc};
-use consumption_metrics::{idempotency_key, Event, EventChunk, EventType, CHUNK_SIZE};
+use camino::Utf8PathBuf;
+use consumption_metrics::EventType;
 use pageserver_api::models::TenantState;
 use reqwest::Url;
-use serde::Serialize;
-use serde_with::{serde_as, DisplayFromStr};
 use std::collections::HashMap;
 use std::sync::Arc;
 use std::time::{Duration, SystemTime};
 use tracing::*;
-use utils::id::{NodeId, TenantId, TimelineId};
-use utils::lsn::Lsn;
+use utils::id::NodeId;
+
+mod metrics;
+use metrics::MetricsKey;
+mod disk_cache;
+mod upload;

 const DEFAULT_HTTP_REPORTING_TIMEOUT: Duration = Duration::from_secs(60);

-#[serde_as]
-#[derive(Serialize, Debug, Clone, Copy)]
-struct Ids {
-    #[serde_as(as = "DisplayFromStr")]
-    tenant_id: TenantId,
-    #[serde_as(as = "Option<DisplayFromStr>")]
-    #[serde(skip_serializing_if = "Option::is_none")]
-    timeline_id: Option<TimelineId>,
-}
+/// Basically a key-value pair, but usually in a Vec except for [`Cache`].
+///
+/// This is as opposed to `consumption_metrics::Event` which is the externally communicated form.
+/// Difference is basically the missing idempotency key, which lives only for the duration of
+/// upload attempts.
+type RawMetric = (MetricsKey, (EventType, u64));

-/// Key that uniquely identifies the object, this metric describes.
-#[derive(Debug, Clone, PartialEq, Eq, Hash)]
-struct MetricsKey {
-    tenant_id: TenantId,
-    timeline_id: Option<TimelineId>,
-    metric: &'static str,
-}
-
-impl MetricsKey {
-    const fn absolute_values(self) -> AbsoluteValueFactory {
-        AbsoluteValueFactory(self)
-    }
-    const fn incremental_values(self) -> IncrementalValueFactory {
-        IncrementalValueFactory(self)
-    }
-}
-
-/// Helper type which each individual metric kind can return to produce only absolute values.
-struct AbsoluteValueFactory(MetricsKey);
-
-impl AbsoluteValueFactory {
-    fn at(self, time: DateTime<Utc>, val: u64) -> (MetricsKey, (EventType, u64)) {
-        let key = self.0;
-        (key, (EventType::Absolute { time }, val))
-    }
-}
-
-/// Helper type which each individual metric kind can return to produce only incremental values.
-struct IncrementalValueFactory(MetricsKey);
-
-impl IncrementalValueFactory {
-    #[allow(clippy::wrong_self_convention)]
-    fn from_previous_up_to(
-        self,
-        prev_end: DateTime<Utc>,
-        up_to: DateTime<Utc>,
-        val: u64,
-    ) -> (MetricsKey, (EventType, u64)) {
-        let key = self.0;
-        // cannot assert prev_end < up_to because these are realtime clock based
-        (
-            key,
-            (
-                EventType::Incremental {
-                    start_time: prev_end,
-                    stop_time: up_to,
-                },
-                val,
-            ),
-        )
-    }
-
-    fn key(&self) -> &MetricsKey {
-        &self.0
-    }
-}
-
-// the static part of a MetricsKey
-impl MetricsKey {
-    /// Absolute value of [`Timeline::get_last_record_lsn`].
-    ///
-    /// [`Timeline::get_last_record_lsn`]: crate::tenant::Timeline::get_last_record_lsn
-    const fn written_size(tenant_id: TenantId, timeline_id: TimelineId) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: Some(timeline_id),
-            metric: "written_size",
-        }
-        .absolute_values()
-    }
-
-    /// Values will be the difference of the latest [`MetricsKey::written_size`] to what we
-    /// previously sent, starting from the previously sent incremental time range ending at the
-    /// latest absolute measurement.
-    const fn written_size_delta(
-        tenant_id: TenantId,
-        timeline_id: TimelineId,
-    ) -> IncrementalValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: Some(timeline_id),
-            // the name here is correctly about data not size, because that is what is wanted by
-            // downstream pipeline
-            metric: "written_data_bytes_delta",
-        }
-        .incremental_values()
-    }
-
-    /// Exact [`Timeline::get_current_logical_size`].
-    ///
-    /// [`Timeline::get_current_logical_size`]: crate::tenant::Timeline::get_current_logical_size
-    const fn timeline_logical_size(
-        tenant_id: TenantId,
-        timeline_id: TimelineId,
-    ) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: Some(timeline_id),
-            metric: "timeline_logical_size",
-        }
-        .absolute_values()
-    }
-
-    /// [`Tenant::remote_size`]
-    ///
-    /// [`Tenant::remote_size`]: crate::tenant::Tenant::remote_size
-    const fn remote_storage_size(tenant_id: TenantId) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: None,
-            metric: "remote_storage_size",
-        }
-        .absolute_values()
-    }
-
-    /// Sum of [`Timeline::resident_physical_size`] for each `Tenant`.
-    ///
-    /// [`Timeline::resident_physical_size`]: crate::tenant::Timeline::resident_physical_size
-    const fn resident_size(tenant_id: TenantId) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: None,
-            metric: "resident_size",
-        }
-        .absolute_values()
-    }
-
-    /// [`Tenant::cached_synthetic_size`] as refreshed by [`calculate_synthetic_size_worker`].
-    ///
-    /// [`Tenant::cached_synthetic_size`]: crate::tenant::Tenant::cached_synthetic_size
-    const fn synthetic_size(tenant_id: TenantId) -> AbsoluteValueFactory {
-        MetricsKey {
-            tenant_id,
-            timeline_id: None,
-            metric: "synthetic_storage_size",
-        }
-        .absolute_values()
-    }
-}
+/// Caches the [`RawMetric`]s
+///
+/// In practice, during startup, last sent values are stored here to be used in calculating new
+/// ones. After successful uploading, the cached values are updated to cache. This used to be used
+/// for deduplication, but that is no longer needed.
+type Cache = HashMap<MetricsKey, (EventType, u64)>;

 /// Main thread that serves metrics collection
 pub async fn collect_metrics(
    metric_collection_endpoint: &Url,
    metric_collection_interval: Duration,
-    cached_metric_collection_interval: Duration,
+    _cached_metric_collection_interval: Duration,
    synthetic_size_calculation_interval: Duration,
    node_id: NodeId,
+    local_disk_storage: Utf8PathBuf,
    ctx: RequestContext,
 ) -> anyhow::Result<()> {
-    let mut ticker = tokio::time::interval(metric_collection_interval);
-    info!("starting collect_metrics");
+    if _cached_metric_collection_interval != Duration::ZERO {
+        tracing::warn!(
+            "cached_metric_collection_interval is no longer used, please set it to zero."
+        )
+    }

    // spin up background worker that caclulates tenant sizes
    let worker_ctx =
@@ -202,543 +68,216 @@ pub async fn collect_metrics(
        },
    );

+    let path: Arc<Utf8PathBuf> = Arc::new(local_disk_storage);
+
+    let cancel = task_mgr::shutdown_token();
+
+    let restore_and_reschedule = restore_and_reschedule(&path, metric_collection_interval);
+
+    let mut cached_metrics = tokio::select! {
+        _ = cancel.cancelled() => return Ok(()),
+        ret = restore_and_reschedule => ret,
+    };
+
    // define client here to reuse it for all requests
    let client = reqwest::ClientBuilder::new()
        .timeout(DEFAULT_HTTP_REPORTING_TIMEOUT)
        .build()
        .expect("Failed to create http client with timeout");
-    let mut cached_metrics = HashMap::new();
-    let mut prev_iteration_time: std::time::Instant = std::time::Instant::now();
-
-    loop {
-        tokio::select! {
-            _ = task_mgr::shutdown_watcher() => {
-                info!("collect_metrics received cancellation request");
-                return Ok(());
-            },
-            tick_at = ticker.tick() => {
-
-                // send cached metrics every cached_metric_collection_interval
-                let send_cached = prev_iteration_time.elapsed() >= cached_metric_collection_interval;
-
-                if send_cached {
-                    prev_iteration_time = std::time::Instant::now();
-                }
-
-                collect_metrics_iteration(&client, &mut cached_metrics, metric_collection_endpoint, node_id, &ctx, send_cached).await;
-
-                crate::tenant::tasks::warn_when_period_overrun(
-                    tick_at.elapsed(),
-                    metric_collection_interval,
-                    "consumption_metrics_collect_metrics",
-                );
-            }
-        }
-    }
-}
-
-/// One iteration of metrics collection
-///
-/// Gather per-tenant and per-timeline metrics and send them to the `metric_collection_endpoint`.
-/// Cache metrics to avoid sending the same metrics multiple times.
-///
-/// This function handles all errors internally
-/// and doesn't break iteration if just one tenant fails.
-///
-/// TODO
-/// - refactor this function (chunking+sending part) to reuse it in proxy module;
-async fn collect_metrics_iteration(
-    client: &reqwest::Client,
-    cached_metrics: &mut HashMap<MetricsKey, (EventType, u64)>,
-    metric_collection_endpoint: &reqwest::Url,
-    node_id: NodeId,
-    ctx: &RequestContext,
-    send_cached: bool,
-) {
-    let mut current_metrics: Vec<(MetricsKey, (EventType, u64))> = Vec::new();
-    trace!(
-        "starting collect_metrics_iteration. metric_collection_endpoint: {}",
-        metric_collection_endpoint
-    );
-
-    // get list of tenants
-    let tenants = match mgr::list_tenants().await {
-        Ok(tenants) => tenants,
-        Err(err) => {
-            error!("failed to list tenants: {:?}", err);
-            return;
-        }
-    };
-
-    // iterate through list of Active tenants and collect metrics
-    for (tenant_id, tenant_state) in tenants {
-        if tenant_state != TenantState::Active {
-            continue;
-        }
-
-        let tenant = match mgr::get_tenant(tenant_id, true).await {
-            Ok(tenant) => tenant,
-            Err(err) => {
-                // It is possible that tenant was deleted between
-                // `list_tenants` and `get_tenant`, so just warn about it.
-                warn!("failed to get tenant {tenant_id:?}: {err:?}");
-                continue;
-            }
-        };
-
-        let mut tenant_resident_size = 0;
-
-        // iterate through list of timelines in tenant
-        for timeline in tenant.list_timelines() {
-            // collect per-timeline metrics only for active timelines
-
-            let timeline_id = timeline.timeline_id;
-
-            match TimelineSnapshot::collect(&timeline, ctx) {
-                Ok(Some(snap)) => {
-                    snap.to_metrics(
-                        tenant_id,
-                        timeline_id,
-                        Utc::now(),
-                        &mut current_metrics,
-                        cached_metrics,
-                    );
-                }
-                Ok(None) => {}
-                Err(e) => {
-                    error!(
-                        "failed to get metrics values for tenant {tenant_id} timeline {}: {e:#?}",
-                        timeline.timeline_id
-                    );
-                    continue;
-                }
-            }
-
-            tenant_resident_size += timeline.resident_physical_size();
-        }
-
-        current_metrics
-            .push(MetricsKey::remote_storage_size(tenant_id).at(Utc::now(), tenant.remote_size()));
-
-        current_metrics
-            .push(MetricsKey::resident_size(tenant_id).at(Utc::now(), tenant_resident_size));
-
-        // Note that this metric is calculated in a separate bgworker
-        // Here we only use cached value, which may lag behind the real latest one
-        let synthetic_size = tenant.cached_synthetic_size();
-
-        if synthetic_size != 0 {
-            // only send non-zeroes because otherwise these show up as errors in logs
-            current_metrics
-                .push(MetricsKey::synthetic_size(tenant_id).at(Utc::now(), synthetic_size));
-        }
-    }
-
-    // Filter metrics, unless we want to send all metrics, including cached ones.
-    // See: https://github.com/neondatabase/neon/issues/3485
-    if !send_cached {
-        current_metrics.retain(|(curr_key, (kind, curr_val))| {
-            if kind.is_incremental() {
-                // incremental values (currently only written_size_delta) should not get any cache
-                // deduplication because they will be used by upstream for "is still alive."
-                true
-            } else {
-                match cached_metrics.get(curr_key) {
-                    Some((_, val)) => val != curr_val,
-                    None => true,
-                }
-            }
-        });
-    }
-
-    if current_metrics.is_empty() {
-        trace!("no new metrics to send");
-        return;
-    }
-
-    // Send metrics.
-    // Split into chunks of 1000 metrics to avoid exceeding the max request size
-    let chunks = current_metrics.chunks(CHUNK_SIZE);
-
-    let mut chunk_to_send: Vec<Event<Ids>> = Vec::with_capacity(CHUNK_SIZE);

    let node_id = node_id.to_string();

-    for chunk in chunks {
-        chunk_to_send.clear();
+    // reminder: ticker is ready immediatedly
+    let mut ticker = tokio::time::interval(metric_collection_interval);

-        // enrich metrics with type,timestamp and idempotency key before sending
-        chunk_to_send.extend(chunk.iter().map(|(curr_key, (when, curr_val))| Event {
-            kind: *when,
-            metric: curr_key.metric,
-            idempotency_key: idempotency_key(&node_id),
-            value: *curr_val,
-            extra: Ids {
-                tenant_id: curr_key.tenant_id,
-                timeline_id: curr_key.timeline_id,
-            },
-        }));
+    loop {
+        let tick_at = tokio::select! {
+            _ = cancel.cancelled() => return Ok(()),
+            tick_at = ticker.tick() => tick_at,
+        };

-        const MAX_RETRIES: u32 = 3;
+        // these are point in time, with variable "now"
+        let metrics = metrics::collect_all_metrics(&cached_metrics, &ctx).await;

-        for attempt in 0..MAX_RETRIES {
-            let res = client
-                .post(metric_collection_endpoint.clone())
-                .json(&EventChunk {
-                    events: (&chunk_to_send).into(),
-                })
-                .send()
-                .await;
+        if metrics.is_empty() {
+            continue;
+        }

-            match res {
-                Ok(res) => {
-                    if res.status().is_success() {
-                        // update cached metrics after they were sent successfully
-                        for (curr_key, curr_val) in chunk.iter() {
-                            cached_metrics.insert(curr_key.clone(), *curr_val);
-                        }
-                    } else {
-                        error!("metrics endpoint refused the sent metrics: {:?}", res);
-                        for metric in chunk_to_send
-                            .iter()
-                            .filter(|metric| metric.value > (1u64 << 40))
-                        {
-                            // Report if the metric value is suspiciously large
-                            error!("potentially abnormal metric value: {:?}", metric);
-                        }
-                    }
-                    break;
+        let metrics = Arc::new(metrics);
+
+        // why not race cancellation here? because we are one of the last tasks, and if we are
+        // already here, better to try to flush the new values.
+
+        let flush = async {
+            match disk_cache::flush_metrics_to_disk(&metrics, &path).await {
+                Ok(()) => {
+                    tracing::debug!("flushed metrics to disk");
                }
-                Err(err) if err.is_timeout() => {
-                    error!(attempt, "timeout sending metrics, retrying immediately");
-                    continue;
-                }
-                Err(err) => {
-                    error!(attempt, ?err, "failed to send metrics");
-                    break;
+                Err(e) => {
+                    // idea here is that if someone creates a directory as our path, then they
+                    // might notice it from the logs before shutdown and remove it
+                    tracing::error!("failed to persist metrics to {path:?}: {e:#}");
                }
            }
+        };
+
+        let upload = async {
+            let res = upload::upload_metrics(
+                &client,
+                metric_collection_endpoint,
+                &cancel,
+                &node_id,
+                &metrics,
+                &mut cached_metrics,
+            )
+            .await;
+            if let Err(e) = res {
+                // serialization error which should never happen
+                tracing::error!("failed to upload due to {e:#}");
+            }
+        };
+
+        // let these run concurrently
+        let (_, _) = tokio::join!(flush, upload);
+
+        crate::tenant::tasks::warn_when_period_overrun(
+            tick_at.elapsed(),
+            metric_collection_interval,
+            "consumption_metrics_collect_metrics",
+        );
+    }
+}
+
+/// Called on the first iteration in an attempt to join the metric uploading schedule from previous
+/// pageserver session. Pageserver is supposed to upload at intervals regardless of restarts.
+///
+/// Cancellation safe.
+async fn restore_and_reschedule(
+    path: &Arc<Utf8PathBuf>,
+    metric_collection_interval: Duration,
+) -> Cache {
+    let (cached, earlier_metric_at) = match disk_cache::read_metrics_from_disk(path.clone()).await {
+        Ok(found_some) => {
+            // there is no min needed because we write these sequentially in
+            // collect_all_metrics
+            let earlier_metric_at = found_some
+                .iter()
+                .map(|(_, (et, _))| et.recorded_at())
+                .copied()
+                .next();
+
+            let cached = found_some.into_iter().collect::<Cache>();
+
+            (cached, earlier_metric_at)
+        }
+        Err(e) => {
+            use std::io::{Error, ErrorKind};
+
+            let root = e.root_cause();
+            let maybe_ioerr = root.downcast_ref::<Error>();
+            let is_not_found = maybe_ioerr.is_some_and(|e| e.kind() == ErrorKind::NotFound);
+
+            if !is_not_found {
+                tracing::info!("failed to read any previous metrics from {path:?}: {e:#}");
+            }
+
+            (HashMap::new(), None)
+        }
+    };
+
+    if let Some(earlier_metric_at) = earlier_metric_at {
+        let earlier_metric_at: SystemTime = earlier_metric_at.into();
+
+        let error = reschedule(earlier_metric_at, metric_collection_interval).await;
+
+        if let Some(error) = error {
+            if error.as_secs() >= 60 {
+                tracing::info!(
+                    error_ms = error.as_millis(),
+                    "startup scheduling error due to restart"
+                )
+            }
        }
    }
+
+    cached
 }

-/// Internal type to make timeline metric production testable.
-///
-/// As this value type contains all of the information needed from a timeline to produce the
-/// metrics, it can easily be created with different values in test.
-struct TimelineSnapshot {
-    loaded_at: (Lsn, SystemTime),
-    last_record_lsn: Lsn,
-    current_exact_logical_size: Option<u64>,
-}
+async fn reschedule(
+    earlier_metric_at: SystemTime,
+    metric_collection_interval: Duration,
+) -> Option<Duration> {
+    let now = SystemTime::now();
+    match now.duration_since(earlier_metric_at) {
+        Ok(from_last_send) if from_last_send < metric_collection_interval => {
+            let sleep_for = metric_collection_interval - from_last_send;

-impl TimelineSnapshot {
-    /// Collect the metrics from an actual timeline.
-    ///
-    /// Fails currently only when [`Timeline::get_current_logical_size`] fails.
-    ///
-    /// [`Timeline::get_current_logical_size`]: crate::tenant::Timeline::get_current_logical_size
-    fn collect(
-        t: &Arc<crate::tenant::Timeline>,
-        ctx: &RequestContext,
-    ) -> anyhow::Result<Option<Self>> {
-        use anyhow::Context;
+            let deadline = std::time::Instant::now() + sleep_for;

-        if !t.is_active() {
-            // no collection for broken or stopping needed, we will still keep the cached values
-            // though at the caller.
-            Ok(None)
-        } else {
-            let loaded_at = t.loaded_at;
-            let last_record_lsn = t.get_last_record_lsn();
+            tokio::time::sleep_until(deadline.into()).await;

-            let current_exact_logical_size = {
-                let span = info_span!("collect_metrics_iteration", tenant_id = %t.tenant_id, timeline_id = %t.timeline_id);
-                let res = span
-                    .in_scope(|| t.get_current_logical_size(ctx))
-                    .context("get_current_logical_size");
-                match res? {
-                    // Only send timeline logical size when it is fully calculated.
-                    (size, is_exact) if is_exact => Some(size),
-                    (_, _) => None,
-                }
-            };
+            let now = std::time::Instant::now();

-            Ok(Some(TimelineSnapshot {
-                loaded_at,
-                last_record_lsn,
-                current_exact_logical_size,
-            }))
-        }
-    }
-
-    /// Produce the timeline consumption metrics into the `metrics` argument.
-    fn to_metrics(
-        &self,
-        tenant_id: TenantId,
-        timeline_id: TimelineId,
-        now: DateTime<Utc>,
-        metrics: &mut Vec<(MetricsKey, (EventType, u64))>,
-        cache: &HashMap<MetricsKey, (EventType, u64)>,
-    ) {
-        let timeline_written_size = u64::from(self.last_record_lsn);
-
-        let (key, written_size_now) =
-            MetricsKey::written_size(tenant_id, timeline_id).at(now, timeline_written_size);
-
-        // last_record_lsn can only go up, right now at least, TODO: #2592 or related
-        // features might change this.
-
-        let written_size_delta_key = MetricsKey::written_size_delta(tenant_id, timeline_id);
-
-        // use this when available, because in a stream of incremental values, it will be
-        // accurate where as when last_record_lsn stops moving, we will only cache the last
-        // one of those.
-        let last_stop_time = cache
-            .get(written_size_delta_key.key())
-            .map(|(until, _val)| {
-                until
-                    .incremental_timerange()
-                    .expect("never create EventType::Absolute for written_size_delta")
-                    .end
-            });
-
-        // by default, use the last sent written_size as the basis for
-        // calculating the delta. if we don't yet have one, use the load time value.
-        let prev = cache
-            .get(&key)
-            .map(|(prev_at, prev)| {
-                // use the prev time from our last incremental update, or default to latest
-                // absolute update on the first round.
-                let prev_at = prev_at
-                    .absolute_time()
-                    .expect("never create EventType::Incremental for written_size");
-                let prev_at = last_stop_time.unwrap_or(prev_at);
-                (*prev_at, *prev)
-            })
-            .unwrap_or_else(|| {
-                // if we don't have a previous point of comparison, compare to the load time
-                // lsn.
-                let (disk_consistent_lsn, loaded_at) = &self.loaded_at;
-                (DateTime::from(*loaded_at), disk_consistent_lsn.0)
-            });
-
-        // written_size_bytes_delta
-        metrics.extend(
-            if let Some(delta) = written_size_now.1.checked_sub(prev.1) {
-                let up_to = written_size_now
-                    .0
-                    .absolute_time()
-                    .expect("never create EventType::Incremental for written_size");
-                let key_value = written_size_delta_key.from_previous_up_to(prev.0, *up_to, delta);
-                Some(key_value)
+            // executor threads might be busy, add extra measurements
+            Some(if now < deadline {
+                deadline - now
            } else {
-                None
-            },
-        );
-
-        // written_size
-        metrics.push((key, written_size_now));
-
-        if let Some(size) = self.current_exact_logical_size {
-            metrics.push(MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, size));
+                now - deadline
+            })
+        }
+        Ok(from_last_send) => Some(from_last_send.saturating_sub(metric_collection_interval)),
+        Err(_) => {
+            tracing::warn!(
+                ?now,
+                ?earlier_metric_at,
+                "oldest recorded metric is in future; first values will come out with inconsistent timestamps"
+            );
+            earlier_metric_at.duration_since(now).ok()
        }
    }
 }

 /// Caclculate synthetic size for each active tenant
-pub async fn calculate_synthetic_size_worker(
+async fn calculate_synthetic_size_worker(
    synthetic_size_calculation_interval: Duration,
    ctx: &RequestContext,
 ) -> anyhow::Result<()> {
    info!("starting calculate_synthetic_size_worker");

+    // reminder: ticker is ready immediatedly
    let mut ticker = tokio::time::interval(synthetic_size_calculation_interval);
+    let cause = LogicalSizeCalculationCause::ConsumptionMetricsSyntheticSize;

    loop {
-        tokio::select! {
-            _ = task_mgr::shutdown_watcher() => {
-                return Ok(());
-            },
-            tick_at = ticker.tick() => {
+        let tick_at = tokio::select! {
+            _ = task_mgr::shutdown_watcher() => return Ok(()),
+            tick_at = ticker.tick() => tick_at,
+        };

-                let tenants = match mgr::list_tenants().await {
-                    Ok(tenants) => tenants,
-                    Err(e) => {
-                        warn!("cannot get tenant list: {e:#}");
-                        continue;
-                    }
-                };
-                // iterate through list of Active tenants and collect metrics
-                for (tenant_id, tenant_state) in tenants {
+        let tenants = match mgr::list_tenants().await {
+            Ok(tenants) => tenants,
+            Err(e) => {
+                warn!("cannot get tenant list: {e:#}");
+                continue;
+            }
+        };

-                    if tenant_state != TenantState::Active {
-                        continue;
-                    }
-
-                    if let Ok(tenant) = mgr::get_tenant(tenant_id, true).await
-                    {
-                        if let Err(e) = tenant.calculate_synthetic_size(
-                            LogicalSizeCalculationCause::ConsumptionMetricsSyntheticSize,
-                            ctx).await {
-                            error!("failed to calculate synthetic size for tenant {}: {}", tenant_id, e);
-                        }
-                    }
+        for (tenant_id, tenant_state) in tenants {
+            if tenant_state != TenantState::Active {
+                continue;
+            }

+            if let Ok(tenant) = mgr::get_tenant(tenant_id, true).await {
+                if let Err(e) = tenant.calculate_synthetic_size(cause, ctx).await {
+                    error!("failed to calculate synthetic size for tenant {tenant_id}: {e:#}");
                }
-
-                crate::tenant::tasks::warn_when_period_overrun(
-                    tick_at.elapsed(),
-                    synthetic_size_calculation_interval,
-                    "consumption_metrics_synthetic_size_worker",
-                );
            }
        }
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use std::collections::HashMap;
-
-    use std::time::SystemTime;
-    use utils::{
-        id::{TenantId, TimelineId},
-        lsn::Lsn,
-    };
-
-    use crate::consumption_metrics::MetricsKey;
-
-    use super::TimelineSnapshot;
-    use chrono::{DateTime, Utc};
-
-    #[test]
-    fn startup_collected_timeline_metrics_before_advancing() {
-        let tenant_id = TenantId::generate();
-        let timeline_id = TimelineId::generate();
-
-        let mut metrics = Vec::new();
-        let cache = HashMap::new();
-
-        let initdb_lsn = Lsn(0x10000);
-        let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
-
-        let snap = TimelineSnapshot {
-            loaded_at: (disk_consistent_lsn, SystemTime::now()),
-            last_record_lsn: disk_consistent_lsn,
-            current_exact_logical_size: Some(0x42000),
-        };
-
-        let now = DateTime::<Utc>::from(SystemTime::now());
-
-        snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
-
-        assert_eq!(
-            metrics,
-            &[
-                MetricsKey::written_size_delta(tenant_id, timeline_id).from_previous_up_to(
-                    snap.loaded_at.1.into(),
-                    now,
-                    0
-                ),
-                MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
-                MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
-            ]
-        );
-    }
-
-    #[test]
-    fn startup_collected_timeline_metrics_second_round() {
-        let tenant_id = TenantId::generate();
-        let timeline_id = TimelineId::generate();
-
-        let [now, before, init] = time_backwards();
-
-        let now = DateTime::<Utc>::from(now);
-        let before = DateTime::<Utc>::from(before);
-
-        let initdb_lsn = Lsn(0x10000);
-        let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
-
-        let mut metrics = Vec::new();
-        let cache = HashMap::from([
-            MetricsKey::written_size(tenant_id, timeline_id).at(before, disk_consistent_lsn.0)
-        ]);
-
-        let snap = TimelineSnapshot {
-            loaded_at: (disk_consistent_lsn, init),
-            last_record_lsn: disk_consistent_lsn,
-            current_exact_logical_size: Some(0x42000),
-        };
-
-        snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
-
-        assert_eq!(
-            metrics,
-            &[
-                MetricsKey::written_size_delta(tenant_id, timeline_id)
-                    .from_previous_up_to(before, now, 0),
-                MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
-                MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
-            ]
-        );
-    }
-
-    #[test]
-    fn startup_collected_timeline_metrics_nth_round_at_same_lsn() {
-        let tenant_id = TenantId::generate();
-        let timeline_id = TimelineId::generate();
-
-        let [now, just_before, before, init] = time_backwards();
-
-        let now = DateTime::<Utc>::from(now);
-        let just_before = DateTime::<Utc>::from(just_before);
-        let before = DateTime::<Utc>::from(before);
-
-        let initdb_lsn = Lsn(0x10000);
-        let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
-
-        let mut metrics = Vec::new();
-        let cache = HashMap::from([
-            // at t=before was the last time the last_record_lsn changed
-            MetricsKey::written_size(tenant_id, timeline_id).at(before, disk_consistent_lsn.0),
-            // end time of this event is used for the next ones
-            MetricsKey::written_size_delta(tenant_id, timeline_id).from_previous_up_to(
-                before,
-                just_before,
-                0,
-            ),
-        ]);
-
-        let snap = TimelineSnapshot {
-            loaded_at: (disk_consistent_lsn, init),
-            last_record_lsn: disk_consistent_lsn,
-            current_exact_logical_size: Some(0x42000),
-        };
-
-        snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
-
-        assert_eq!(
-            metrics,
-            &[
-                MetricsKey::written_size_delta(tenant_id, timeline_id).from_previous_up_to(
-                    just_before,
-                    now,
-                    0
-                ),
-                MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
-                MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
-            ]
-        );
-    }
-
-    fn time_backwards<const N: usize>() -> [std::time::SystemTime; N] {
-        let mut times = [std::time::SystemTime::UNIX_EPOCH; N];
-        times[0] = std::time::SystemTime::now();
-        for behind in 1..N {
-            times[behind] = times[0] - std::time::Duration::from_secs(behind as u64);
-        }
-
-        times
+
+        crate::tenant::tasks::warn_when_period_overrun(
+            tick_at.elapsed(),
+            synthetic_size_calculation_interval,
+            "consumption_metrics_synthetic_size_worker",
+        );
    }
 }
--- a/pageserver/src/consumption_metrics/disk_cache.rs
+++ b/pageserver/src/consumption_metrics/disk_cache.rs
@@ -0,0 +1,119 @@
+use anyhow::Context;
+use camino::{Utf8Path, Utf8PathBuf};
+use std::sync::Arc;
+
+use super::RawMetric;
+
+pub(super) async fn read_metrics_from_disk(
+    path: Arc<Utf8PathBuf>,
+) -> anyhow::Result<Vec<RawMetric>> {
+    // do not add context to each error, callsite will log with full path
+    let span = tracing::Span::current();
+    tokio::task::spawn_blocking(move || {
+        let _e = span.entered();
+
+        if let Some(parent) = path.parent() {
+            if let Err(e) = scan_and_delete_with_same_prefix(&path) {
+                tracing::info!("failed to cleanup temporary files in {parent:?}: {e:#}");
+            }
+        }
+
+        let mut file = std::fs::File::open(&*path)?;
+        let reader = std::io::BufReader::new(&mut file);
+        anyhow::Ok(serde_json::from_reader::<_, Vec<RawMetric>>(reader)?)
+    })
+    .await
+    .context("read metrics join error")
+    .and_then(|x| x)
+}
+
+fn scan_and_delete_with_same_prefix(path: &Utf8Path) -> std::io::Result<()> {
+    let it = std::fs::read_dir(path.parent().expect("caller checked"))?;
+
+    let prefix = path.file_name().expect("caller checked").to_string();
+
+    for entry in it {
+        let entry = entry?;
+        if !entry.metadata()?.is_file() {
+            continue;
+        }
+        let file_name = entry.file_name();
+
+        if path.file_name().unwrap() == file_name {
+            // do not remove our actual file
+            continue;
+        }
+
+        let file_name = file_name.to_string_lossy();
+
+        if !file_name.starts_with(&*prefix) {
+            continue;
+        }
+
+        let path = entry.path();
+
+        if let Err(e) = std::fs::remove_file(&path) {
+            tracing::warn!("cleaning up old tempfile {file_name:?} failed: {e:#}");
+        } else {
+            tracing::info!("cleaned up old tempfile {file_name:?}");
+        }
+    }
+
+    Ok(())
+}
+
+pub(super) async fn flush_metrics_to_disk(
+    current_metrics: &Arc<Vec<RawMetric>>,
+    path: &Arc<Utf8PathBuf>,
+) -> anyhow::Result<()> {
+    use std::io::Write;
+
+    anyhow::ensure!(path.parent().is_some(), "path must have parent: {path:?}");
+    anyhow::ensure!(
+        path.file_name().is_some(),
+        "path must have filename: {path:?}"
+    );
+
+    let span = tracing::Span::current();
+    tokio::task::spawn_blocking({
+        let current_metrics = current_metrics.clone();
+        let path = path.clone();
+        move || {
+            let _e = span.entered();
+
+            let parent = path.parent().expect("existence checked");
+            let file_name = path.file_name().expect("existence checked");
+            let mut tempfile = camino_tempfile::Builder::new()
+                .prefix(file_name)
+                .suffix(".tmp")
+                .tempfile_in(parent)?;
+
+            tracing::debug!("using tempfile {:?}", tempfile.path());
+
+            // write out all of the raw metrics, to be read out later on restart as cached values
+            {
+                let mut writer = std::io::BufWriter::new(&mut tempfile);
+                serde_json::to_writer(&mut writer, &*current_metrics)
+                    .context("serialize metrics")?;
+                writer
+                    .into_inner()
+                    .map_err(|_| anyhow::anyhow!("flushing metrics failed"))?;
+            }
+
+            tempfile.flush()?;
+            tempfile.as_file().sync_all()?;
+
+            fail::fail_point!("before-persist-last-metrics-collected");
+
+            drop(tempfile.persist(&*path).map_err(|e| e.error)?);
+
+            let f = std::fs::File::open(path.parent().unwrap())?;
+            f.sync_all()?;
+
+            anyhow::Ok(())
+        }
+    })
+    .await
+    .with_context(|| format!("write metrics to {path:?} join error"))
+    .and_then(|x| x.with_context(|| format!("write metrics to {path:?}")))
+}
--- a/pageserver/src/consumption_metrics/metrics.rs
+++ b/pageserver/src/consumption_metrics/metrics.rs
@@ -0,0 +1,455 @@
+use crate::context::RequestContext;
+use anyhow::Context;
+use chrono::{DateTime, Utc};
+use consumption_metrics::EventType;
+use futures::stream::StreamExt;
+use serde_with::serde_as;
+use std::{sync::Arc, time::SystemTime};
+use utils::{
+    id::{TenantId, TimelineId},
+    lsn::Lsn,
+};
+
+use super::{Cache, RawMetric};
+
+/// Name of the metric, used by `MetricsKey` factory methods and `deserialize_cached_events`
+/// instead of static str.
+// Do not rename any of these without first consulting with data team and partner
+// management.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, serde::Serialize, serde::Deserialize)]
+pub(super) enum Name {
+    /// Timeline last_record_lsn, absolute
+    #[serde(rename = "written_size")]
+    WrittenSize,
+    /// Timeline last_record_lsn, incremental
+    #[serde(rename = "written_data_bytes_delta")]
+    WrittenSizeDelta,
+    /// Timeline logical size
+    #[serde(rename = "timeline_logical_size")]
+    LogicalSize,
+    /// Tenant remote size
+    #[serde(rename = "remote_storage_size")]
+    RemoteSize,
+    /// Tenant resident size
+    #[serde(rename = "resident_size")]
+    ResidentSize,
+    /// Tenant synthetic size
+    #[serde(rename = "synthetic_storage_size")]
+    SyntheticSize,
+}
+
+/// Key that uniquely identifies the object this metric describes.
+///
+/// This is a denormalization done at the MetricsKey const methods; these should not be constructed
+/// elsewhere.
+#[serde_with::serde_as]
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, serde::Serialize, serde::Deserialize)]
+pub(crate) struct MetricsKey {
+    #[serde_as(as = "serde_with::DisplayFromStr")]
+    pub(super) tenant_id: TenantId,
+
+    #[serde_as(as = "Option<serde_with::DisplayFromStr>")]
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub(super) timeline_id: Option<TimelineId>,
+
+    pub(super) metric: Name,
+}
+
+impl MetricsKey {
+    const fn absolute_values(self) -> AbsoluteValueFactory {
+        AbsoluteValueFactory(self)
+    }
+    const fn incremental_values(self) -> IncrementalValueFactory {
+        IncrementalValueFactory(self)
+    }
+}
+
+/// Helper type which each individual metric kind can return to produce only absolute values.
+struct AbsoluteValueFactory(MetricsKey);
+
+impl AbsoluteValueFactory {
+    const fn at(self, time: DateTime<Utc>, val: u64) -> RawMetric {
+        let key = self.0;
+        (key, (EventType::Absolute { time }, val))
+    }
+
+    fn key(&self) -> &MetricsKey {
+        &self.0
+    }
+}
+
+/// Helper type which each individual metric kind can return to produce only incremental values.
+struct IncrementalValueFactory(MetricsKey);
+
+impl IncrementalValueFactory {
+    #[allow(clippy::wrong_self_convention)]
+    const fn from_until(
+        self,
+        prev_end: DateTime<Utc>,
+        up_to: DateTime<Utc>,
+        val: u64,
+    ) -> RawMetric {
+        let key = self.0;
+        // cannot assert prev_end < up_to because these are realtime clock based
+        let when = EventType::Incremental {
+            start_time: prev_end,
+            stop_time: up_to,
+        };
+        (key, (when, val))
+    }
+
+    fn key(&self) -> &MetricsKey {
+        &self.0
+    }
+}
+
+// the static part of a MetricsKey
+impl MetricsKey {
+    /// Absolute value of [`Timeline::get_last_record_lsn`].
+    ///
+    /// [`Timeline::get_last_record_lsn`]: crate::tenant::Timeline::get_last_record_lsn
+    const fn written_size(tenant_id: TenantId, timeline_id: TimelineId) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: Some(timeline_id),
+            metric: Name::WrittenSize,
+        }
+        .absolute_values()
+    }
+
+    /// Values will be the difference of the latest [`MetricsKey::written_size`] to what we
+    /// previously sent, starting from the previously sent incremental time range ending at the
+    /// latest absolute measurement.
+    const fn written_size_delta(
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+    ) -> IncrementalValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: Some(timeline_id),
+            metric: Name::WrittenSizeDelta,
+        }
+        .incremental_values()
+    }
+
+    /// Exact [`Timeline::get_current_logical_size`].
+    ///
+    /// [`Timeline::get_current_logical_size`]: crate::tenant::Timeline::get_current_logical_size
+    const fn timeline_logical_size(
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+    ) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: Some(timeline_id),
+            metric: Name::LogicalSize,
+        }
+        .absolute_values()
+    }
+
+    /// [`Tenant::remote_size`]
+    ///
+    /// [`Tenant::remote_size`]: crate::tenant::Tenant::remote_size
+    const fn remote_storage_size(tenant_id: TenantId) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: None,
+            metric: Name::RemoteSize,
+        }
+        .absolute_values()
+    }
+
+    /// Sum of [`Timeline::resident_physical_size`] for each `Tenant`.
+    ///
+    /// [`Timeline::resident_physical_size`]: crate::tenant::Timeline::resident_physical_size
+    const fn resident_size(tenant_id: TenantId) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: None,
+            metric: Name::ResidentSize,
+        }
+        .absolute_values()
+    }
+
+    /// [`Tenant::cached_synthetic_size`] as refreshed by [`calculate_synthetic_size_worker`].
+    ///
+    /// [`Tenant::cached_synthetic_size`]: crate::tenant::Tenant::cached_synthetic_size
+    /// [`calculate_synthetic_size_worker`]: super::calculate_synthetic_size_worker
+    const fn synthetic_size(tenant_id: TenantId) -> AbsoluteValueFactory {
+        MetricsKey {
+            tenant_id,
+            timeline_id: None,
+            metric: Name::SyntheticSize,
+        }
+        .absolute_values()
+    }
+}
+
+pub(super) async fn collect_all_metrics(
+    cached_metrics: &Cache,
+    ctx: &RequestContext,
+) -> Vec<RawMetric> {
+    use pageserver_api::models::TenantState;
+
+    let started_at = std::time::Instant::now();
+
+    let tenants = match crate::tenant::mgr::list_tenants().await {
+        Ok(tenants) => tenants,
+        Err(err) => {
+            tracing::error!("failed to list tenants: {:?}", err);
+            return vec![];
+        }
+    };
+
+    let tenants = futures::stream::iter(tenants).filter_map(|(id, state)| async move {
+        if state != TenantState::Active {
+            None
+        } else {
+            crate::tenant::mgr::get_tenant(id, true)
+                .await
+                .ok()
+                .map(|tenant| (id, tenant))
+        }
+    });
+
+    let res = collect(tenants, cached_metrics, ctx).await;
+
+    tracing::info!(
+        elapsed_ms = started_at.elapsed().as_millis(),
+        total = res.len(),
+        "collected metrics"
+    );
+
+    res
+}
+
+async fn collect<S>(tenants: S, cache: &Cache, ctx: &RequestContext) -> Vec<RawMetric>
+where
+    S: futures::stream::Stream<Item = (TenantId, Arc<crate::tenant::Tenant>)>,
+{
+    let mut current_metrics: Vec<RawMetric> = Vec::new();
+
+    let mut tenants = std::pin::pin!(tenants);
+
+    while let Some((tenant_id, tenant)) = tenants.next().await {
+        let mut tenant_resident_size = 0;
+
+        for timeline in tenant.list_timelines() {
+            let timeline_id = timeline.timeline_id;
+
+            match TimelineSnapshot::collect(&timeline, ctx) {
+                Ok(Some(snap)) => {
+                    snap.to_metrics(
+                        tenant_id,
+                        timeline_id,
+                        Utc::now(),
+                        &mut current_metrics,
+                        cache,
+                    );
+                }
+                Ok(None) => {}
+                Err(e) => {
+                    tracing::error!(
+                        "failed to get metrics values for tenant {tenant_id} timeline {}: {e:#?}",
+                        timeline.timeline_id
+                    );
+                    continue;
+                }
+            }
+
+            tenant_resident_size += timeline.resident_physical_size();
+        }
+
+        let snap = TenantSnapshot::collect(&tenant, tenant_resident_size);
+        snap.to_metrics(tenant_id, Utc::now(), cache, &mut current_metrics);
+    }
+
+    current_metrics
+}
+
+/// In-between abstraction to allow testing metrics without actual Tenants.
+struct TenantSnapshot {
+    resident_size: u64,
+    remote_size: u64,
+    synthetic_size: u64,
+}
+
+impl TenantSnapshot {
+    /// Collect tenant status to have metrics created out of it.
+    ///
+    /// `resident_size` is calculated of the timelines we had access to for other metrics, so we
+    /// cannot just list timelines here.
+    fn collect(t: &Arc<crate::tenant::Tenant>, resident_size: u64) -> Self {
+        TenantSnapshot {
+            resident_size,
+            remote_size: t.remote_size(),
+            // Note that this metric is calculated in a separate bgworker
+            // Here we only use cached value, which may lag behind the real latest one
+            synthetic_size: t.cached_synthetic_size(),
+        }
+    }
+
+    fn to_metrics(
+        &self,
+        tenant_id: TenantId,
+        now: DateTime<Utc>,
+        cached: &Cache,
+        metrics: &mut Vec<RawMetric>,
+    ) {
+        let remote_size = MetricsKey::remote_storage_size(tenant_id).at(now, self.remote_size);
+
+        let resident_size = MetricsKey::resident_size(tenant_id).at(now, self.resident_size);
+
+        let synthetic_size = {
+            let factory = MetricsKey::synthetic_size(tenant_id);
+            let mut synthetic_size = self.synthetic_size;
+
+            if synthetic_size == 0 {
+                if let Some((_, value)) = cached.get(factory.key()) {
+                    // use the latest value from previous session
+                    synthetic_size = *value;
+                }
+            }
+
+            if synthetic_size != 0 {
+                // only send non-zeroes because otherwise these show up as errors in logs
+                Some(factory.at(now, synthetic_size))
+            } else {
+                None
+            }
+        };
+
+        metrics.extend(
+            [Some(remote_size), Some(resident_size), synthetic_size]
+                .into_iter()
+                .flatten(),
+        );
+    }
+}
+
+/// Internal type to make timeline metric production testable.
+///
+/// As this value type contains all of the information needed from a timeline to produce the
+/// metrics, it can easily be created with different values in test.
+struct TimelineSnapshot {
+    loaded_at: (Lsn, SystemTime),
+    last_record_lsn: Lsn,
+    current_exact_logical_size: Option<u64>,
+}
+
+impl TimelineSnapshot {
+    /// Collect the metrics from an actual timeline.
+    ///
+    /// Fails currently only when [`Timeline::get_current_logical_size`] fails.
+    ///
+    /// [`Timeline::get_current_logical_size`]: crate::tenant::Timeline::get_current_logical_size
+    fn collect(
+        t: &Arc<crate::tenant::Timeline>,
+        ctx: &RequestContext,
+    ) -> anyhow::Result<Option<Self>> {
+        if !t.is_active() {
+            // no collection for broken or stopping needed, we will still keep the cached values
+            // though at the caller.
+            Ok(None)
+        } else {
+            let loaded_at = t.loaded_at;
+            let last_record_lsn = t.get_last_record_lsn();
+
+            let current_exact_logical_size = {
+                let span = tracing::info_span!("collect_metrics_iteration", tenant_id = %t.tenant_id, timeline_id = %t.timeline_id);
+                let res = span
+                    .in_scope(|| t.get_current_logical_size(ctx))
+                    .context("get_current_logical_size");
+                match res? {
+                    // Only send timeline logical size when it is fully calculated.
+                    (size, is_exact) if is_exact => Some(size),
+                    (_, _) => None,
+                }
+            };
+
+            Ok(Some(TimelineSnapshot {
+                loaded_at,
+                last_record_lsn,
+                current_exact_logical_size,
+            }))
+        }
+    }
+
+    /// Produce the timeline consumption metrics into the `metrics` argument.
+    fn to_metrics(
+        &self,
+        tenant_id: TenantId,
+        timeline_id: TimelineId,
+        now: DateTime<Utc>,
+        metrics: &mut Vec<RawMetric>,
+        cache: &Cache,
+    ) {
+        let timeline_written_size = u64::from(self.last_record_lsn);
+
+        let written_size_delta_key = MetricsKey::written_size_delta(tenant_id, timeline_id);
+
+        let last_stop_time = cache
+            .get(written_size_delta_key.key())
+            .map(|(until, _val)| {
+                until
+                    .incremental_timerange()
+                    .expect("never create EventType::Absolute for written_size_delta")
+                    .end
+            });
+
+        let (key, written_size_now) =
+            MetricsKey::written_size(tenant_id, timeline_id).at(now, timeline_written_size);
+
+        // by default, use the last sent written_size as the basis for
+        // calculating the delta. if we don't yet have one, use the load time value.
+        let prev = cache
+            .get(&key)
+            .map(|(prev_at, prev)| {
+                // use the prev time from our last incremental update, or default to latest
+                // absolute update on the first round.
+                let prev_at = prev_at
+                    .absolute_time()
+                    .expect("never create EventType::Incremental for written_size");
+                let prev_at = last_stop_time.unwrap_or(prev_at);
+                (*prev_at, *prev)
+            })
+            .unwrap_or_else(|| {
+                // if we don't have a previous point of comparison, compare to the load time
+                // lsn.
+                let (disk_consistent_lsn, loaded_at) = &self.loaded_at;
+                (DateTime::from(*loaded_at), disk_consistent_lsn.0)
+            });
+
+        let up_to = now;
+
+        if let Some(delta) = written_size_now.1.checked_sub(prev.1) {
+            let key_value = written_size_delta_key.from_until(prev.0, up_to, delta);
+            // written_size_delta
+            metrics.push(key_value);
+            // written_size
+            metrics.push((key, written_size_now));
+        } else {
+            // the cached value was ahead of us, report zero until we've caught up
+            metrics.push(written_size_delta_key.from_until(prev.0, up_to, 0));
+            // the cached value was ahead of us, report the same until we've caught up
+            metrics.push((key, (written_size_now.0, prev.1)));
+        }
+
+        {
+            let factory = MetricsKey::timeline_logical_size(tenant_id, timeline_id);
+            let current_or_previous = self
+                .current_exact_logical_size
+                .or_else(|| cache.get(factory.key()).map(|(_, val)| *val));
+
+            if let Some(size) = current_or_previous {
+                metrics.push(factory.at(now, size));
+            }
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests;
+
+#[cfg(test)]
+pub(crate) use tests::metric_examples;
--- a/pageserver/src/consumption_metrics/metrics/tests.rs
+++ b/pageserver/src/consumption_metrics/metrics/tests.rs
@@ -0,0 +1,297 @@
+use super::*;
+use std::collections::HashMap;
+use std::time::SystemTime;
+use utils::lsn::Lsn;
+
+#[test]
+fn startup_collected_timeline_metrics_before_advancing() {
+    let tenant_id = TenantId::generate();
+    let timeline_id = TimelineId::generate();
+
+    let mut metrics = Vec::new();
+    let cache = HashMap::new();
+
+    let initdb_lsn = Lsn(0x10000);
+    let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
+
+    let snap = TimelineSnapshot {
+        loaded_at: (disk_consistent_lsn, SystemTime::now()),
+        last_record_lsn: disk_consistent_lsn,
+        current_exact_logical_size: Some(0x42000),
+    };
+
+    let now = DateTime::<Utc>::from(SystemTime::now());
+
+    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
+
+    assert_eq!(
+        metrics,
+        &[
+            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(
+                snap.loaded_at.1.into(),
+                now,
+                0
+            ),
+            MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
+            MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
+        ]
+    );
+}
+
+#[test]
+fn startup_collected_timeline_metrics_second_round() {
+    let tenant_id = TenantId::generate();
+    let timeline_id = TimelineId::generate();
+
+    let [now, before, init] = time_backwards();
+
+    let now = DateTime::<Utc>::from(now);
+    let before = DateTime::<Utc>::from(before);
+
+    let initdb_lsn = Lsn(0x10000);
+    let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
+
+    let mut metrics = Vec::new();
+    let cache = HashMap::from([
+        MetricsKey::written_size(tenant_id, timeline_id).at(before, disk_consistent_lsn.0)
+    ]);
+
+    let snap = TimelineSnapshot {
+        loaded_at: (disk_consistent_lsn, init),
+        last_record_lsn: disk_consistent_lsn,
+        current_exact_logical_size: Some(0x42000),
+    };
+
+    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
+
+    assert_eq!(
+        metrics,
+        &[
+            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(before, now, 0),
+            MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
+            MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
+        ]
+    );
+}
+
+#[test]
+fn startup_collected_timeline_metrics_nth_round_at_same_lsn() {
+    let tenant_id = TenantId::generate();
+    let timeline_id = TimelineId::generate();
+
+    let [now, just_before, before, init] = time_backwards();
+
+    let now = DateTime::<Utc>::from(now);
+    let just_before = DateTime::<Utc>::from(just_before);
+    let before = DateTime::<Utc>::from(before);
+
+    let initdb_lsn = Lsn(0x10000);
+    let disk_consistent_lsn = Lsn(initdb_lsn.0 * 2);
+
+    let mut metrics = Vec::new();
+    let cache = HashMap::from([
+        // at t=before was the last time the last_record_lsn changed
+        MetricsKey::written_size(tenant_id, timeline_id).at(before, disk_consistent_lsn.0),
+        // end time of this event is used for the next ones
+        MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(before, just_before, 0),
+    ]);
+
+    let snap = TimelineSnapshot {
+        loaded_at: (disk_consistent_lsn, init),
+        last_record_lsn: disk_consistent_lsn,
+        current_exact_logical_size: Some(0x42000),
+    };
+
+    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
+
+    assert_eq!(
+        metrics,
+        &[
+            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(just_before, now, 0),
+            MetricsKey::written_size(tenant_id, timeline_id).at(now, disk_consistent_lsn.0),
+            MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0x42000)
+        ]
+    );
+}
+
+#[test]
+fn post_restart_written_sizes_with_rolled_back_last_record_lsn() {
+    // it can happen that we lose the inmemorylayer but have previously sent metrics and we
+    // should never go backwards
+
+    let tenant_id = TenantId::generate();
+    let timeline_id = TimelineId::generate();
+
+    let [later, now, at_restart] = time_backwards();
+
+    // FIXME: tests would be so much easier if we did not need to juggle back and forth
+    // SystemTime and DateTime::<Utc> ... Could do the conversion only at upload time?
+    let now = DateTime::<Utc>::from(now);
+    let later = DateTime::<Utc>::from(later);
+    let before_restart = at_restart - std::time::Duration::from_secs(5 * 60);
+    let way_before = before_restart - std::time::Duration::from_secs(10 * 60);
+    let before_restart = DateTime::<Utc>::from(before_restart);
+    let way_before = DateTime::<Utc>::from(way_before);
+
+    let snap = TimelineSnapshot {
+        loaded_at: (Lsn(50), at_restart),
+        last_record_lsn: Lsn(50),
+        current_exact_logical_size: None,
+    };
+
+    let mut cache = HashMap::from([
+        MetricsKey::written_size(tenant_id, timeline_id).at(before_restart, 100),
+        MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(
+            way_before,
+            before_restart,
+            // not taken into account, but the timestamps are important
+            999_999_999,
+        ),
+    ]);
+
+    let mut metrics = Vec::new();
+    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
+
+    assert_eq!(
+        metrics,
+        &[
+            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(
+                before_restart,
+                now,
+                0
+            ),
+            MetricsKey::written_size(tenant_id, timeline_id).at(now, 100),
+        ]
+    );
+
+    // now if we cache these metrics, and re-run while "still in recovery"
+    cache.extend(metrics.drain(..));
+
+    // "still in recovery", because our snapshot did not change
+    snap.to_metrics(tenant_id, timeline_id, later, &mut metrics, &cache);
+
+    assert_eq!(
+        metrics,
+        &[
+            MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(now, later, 0),
+            MetricsKey::written_size(tenant_id, timeline_id).at(later, 100),
+        ]
+    );
+}
+
+#[test]
+fn post_restart_current_exact_logical_size_uses_cached() {
+    let tenant_id = TenantId::generate();
+    let timeline_id = TimelineId::generate();
+
+    let [now, at_restart] = time_backwards();
+
+    let now = DateTime::<Utc>::from(now);
+    let before_restart = at_restart - std::time::Duration::from_secs(5 * 60);
+    let before_restart = DateTime::<Utc>::from(before_restart);
+
+    let snap = TimelineSnapshot {
+        loaded_at: (Lsn(50), at_restart),
+        last_record_lsn: Lsn(50),
+        current_exact_logical_size: None,
+    };
+
+    let cache = HashMap::from([
+        MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(before_restart, 100)
+    ]);
+
+    let mut metrics = Vec::new();
+    snap.to_metrics(tenant_id, timeline_id, now, &mut metrics, &cache);
+
+    metrics.retain(|(key, _)| key.metric == Name::LogicalSize);
+
+    assert_eq!(
+        metrics,
+        &[MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 100)]
+    );
+}
+
+#[test]
+fn post_restart_synthetic_size_uses_cached_if_available() {
+    let tenant_id = TenantId::generate();
+
+    let ts = TenantSnapshot {
+        resident_size: 1000,
+        remote_size: 1000,
+        // not yet calculated
+        synthetic_size: 0,
+    };
+
+    let now = SystemTime::now();
+    let before_restart = DateTime::<Utc>::from(now - std::time::Duration::from_secs(5 * 60));
+    let now = DateTime::<Utc>::from(now);
+
+    let cached = HashMap::from([MetricsKey::synthetic_size(tenant_id).at(before_restart, 1000)]);
+
+    let mut metrics = Vec::new();
+    ts.to_metrics(tenant_id, now, &cached, &mut metrics);
+
+    assert_eq!(
+        metrics,
+        &[
+            MetricsKey::remote_storage_size(tenant_id).at(now, 1000),
+            MetricsKey::resident_size(tenant_id).at(now, 1000),
+            MetricsKey::synthetic_size(tenant_id).at(now, 1000),
+        ]
+    );
+}
+
+#[test]
+fn post_restart_synthetic_size_is_not_sent_when_not_cached() {
+    let tenant_id = TenantId::generate();
+
+    let ts = TenantSnapshot {
+        resident_size: 1000,
+        remote_size: 1000,
+        // not yet calculated
+        synthetic_size: 0,
+    };
+
+    let now = SystemTime::now();
+    let now = DateTime::<Utc>::from(now);
+
+    let cached = HashMap::new();
+
+    let mut metrics = Vec::new();
+    ts.to_metrics(tenant_id, now, &cached, &mut metrics);
+
+    assert_eq!(
+        metrics,
+        &[
+            MetricsKey::remote_storage_size(tenant_id).at(now, 1000),
+            MetricsKey::resident_size(tenant_id).at(now, 1000),
+            // no synthetic size here
+        ]
+    );
+}
+
+fn time_backwards<const N: usize>() -> [std::time::SystemTime; N] {
+    let mut times = [std::time::SystemTime::UNIX_EPOCH; N];
+    times[0] = std::time::SystemTime::now();
+    for behind in 1..N {
+        times[behind] = times[0] - std::time::Duration::from_secs(behind as u64);
+    }
+
+    times
+}
+
+pub(crate) const fn metric_examples(
+    tenant_id: TenantId,
+    timeline_id: TimelineId,
+    now: DateTime<Utc>,
+    before: DateTime<Utc>,
+) -> [RawMetric; 6] {
+    [
+        MetricsKey::written_size(tenant_id, timeline_id).at(now, 0),
+        MetricsKey::written_size_delta(tenant_id, timeline_id).from_until(before, now, 0),
+        MetricsKey::timeline_logical_size(tenant_id, timeline_id).at(now, 0),
+        MetricsKey::remote_storage_size(tenant_id).at(now, 0),
+        MetricsKey::resident_size(tenant_id).at(now, 0),
+        MetricsKey::synthetic_size(tenant_id).at(now, 1),
+    ]
+}
--- a/pageserver/src/consumption_metrics/upload.rs
+++ b/pageserver/src/consumption_metrics/upload.rs
@@ -0,0 +1,443 @@
+use consumption_metrics::{Event, EventChunk, IdempotencyKey, CHUNK_SIZE};
+use serde_with::serde_as;
+use tokio_util::sync::CancellationToken;
+use tracing::Instrument;
+
+use super::{metrics::Name, Cache, MetricsKey, RawMetric};
+use utils::id::{TenantId, TimelineId};
+
+/// How the metrics from pageserver are identified.
+#[serde_with::serde_as]
+#[derive(serde::Serialize, serde::Deserialize, Debug, Clone, Copy, PartialEq)]
+struct Ids {
+    #[serde_as(as = "serde_with::DisplayFromStr")]
+    pub(super) tenant_id: TenantId,
+    #[serde_as(as = "Option<serde_with::DisplayFromStr>")]
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub(super) timeline_id: Option<TimelineId>,
+}
+
+#[tracing::instrument(skip_all, fields(metrics_total = %metrics.len()))]
+pub(super) async fn upload_metrics(
+    client: &reqwest::Client,
+    metric_collection_endpoint: &reqwest::Url,
+    cancel: &CancellationToken,
+    node_id: &str,
+    metrics: &[RawMetric],
+    cached_metrics: &mut Cache,
+) -> anyhow::Result<()> {
+    let mut uploaded = 0;
+    let mut failed = 0;
+
+    let started_at = std::time::Instant::now();
+
+    let mut iter = serialize_in_chunks(CHUNK_SIZE, metrics, node_id);
+
+    while let Some(res) = iter.next() {
+        let (chunk, body) = res?;
+
+        let event_bytes = body.len();
+
+        let is_last = iter.len() == 0;
+
+        let res = upload(client, metric_collection_endpoint, body, cancel, is_last)
+            .instrument(tracing::info_span!(
+                "upload",
+                %event_bytes,
+                uploaded,
+                total = metrics.len(),
+            ))
+            .await;
+
+        match res {
+            Ok(()) => {
+                for (curr_key, curr_val) in chunk {
+                    cached_metrics.insert(*curr_key, *curr_val);
+                }
+                uploaded += chunk.len();
+            }
+            Err(_) => {
+                // failure(s) have already been logged
+                //
+                // however this is an inconsistency: if we crash here, we will start with the
+                // values as uploaded. in practice, the rejections no longer happen.
+                failed += chunk.len();
+            }
+        }
+    }
+
+    let elapsed = started_at.elapsed();
+
+    tracing::info!(
+        uploaded,
+        failed,
+        elapsed_ms = elapsed.as_millis(),
+        "done sending metrics"
+    );
+
+    Ok(())
+}
+
+// The return type is quite ugly, but we gain testability in isolation
+fn serialize_in_chunks<'a, F>(
+    chunk_size: usize,
+    input: &'a [RawMetric],
+    factory: F,
+) -> impl ExactSizeIterator<Item = Result<(&'a [RawMetric], bytes::Bytes), serde_json::Error>> + 'a
+where
+    F: KeyGen<'a> + 'a,
+{
+    use bytes::BufMut;
+
+    struct Iter<'a, F> {
+        inner: std::slice::Chunks<'a, RawMetric>,
+        chunk_size: usize,
+
+        // write to a BytesMut so that we can cheaply clone the frozen Bytes for retries
+        buffer: bytes::BytesMut,
+        // chunk amount of events are reused to produce the serialized document
+        scratch: Vec<Event<Ids, Name>>,
+        factory: F,
+    }
+
+    impl<'a, F: KeyGen<'a>> Iterator for Iter<'a, F> {
+        type Item = Result<(&'a [RawMetric], bytes::Bytes), serde_json::Error>;
+
+        fn next(&mut self) -> Option<Self::Item> {
+            let chunk = self.inner.next()?;
+
+            if self.scratch.is_empty() {
+                // first round: create events with N strings
+                self.scratch.extend(
+                    chunk
+                        .iter()
+                        .map(|raw_metric| raw_metric.as_event(&self.factory.generate())),
+                );
+            } else {
+                // next rounds: update_in_place to reuse allocations
+                assert_eq!(self.scratch.len(), self.chunk_size);
+                self.scratch
+                    .iter_mut()
+                    .zip(chunk.iter())
+                    .for_each(|(slot, raw_metric)| {
+                        raw_metric.update_in_place(slot, &self.factory.generate())
+                    });
+            }
+
+            let res = serde_json::to_writer(
+                (&mut self.buffer).writer(),
+                &EventChunk {
+                    events: (&self.scratch[..chunk.len()]).into(),
+                },
+            );
+
+            match res {
+                Ok(()) => Some(Ok((chunk, self.buffer.split().freeze()))),
+                Err(e) => Some(Err(e)),
+            }
+        }
+
+        fn size_hint(&self) -> (usize, Option<usize>) {
+            self.inner.size_hint()
+        }
+    }
+
+    impl<'a, F: KeyGen<'a>> ExactSizeIterator for Iter<'a, F> {}
+
+    let buffer = bytes::BytesMut::new();
+    let inner = input.chunks(chunk_size);
+    let scratch = Vec::new();
+
+    Iter {
+        inner,
+        chunk_size,
+        buffer,
+        scratch,
+        factory,
+    }
+}
+
+trait RawMetricExt {
+    fn as_event(&self, key: &IdempotencyKey<'_>) -> Event<Ids, Name>;
+    fn update_in_place(&self, event: &mut Event<Ids, Name>, key: &IdempotencyKey<'_>);
+}
+
+impl RawMetricExt for RawMetric {
+    fn as_event(&self, key: &IdempotencyKey<'_>) -> Event<Ids, Name> {
+        let MetricsKey {
+            metric,
+            tenant_id,
+            timeline_id,
+        } = self.0;
+
+        let (kind, value) = self.1;
+
+        Event {
+            kind,
+            metric,
+            idempotency_key: key.to_string(),
+            value,
+            extra: Ids {
+                tenant_id,
+                timeline_id,
+            },
+        }
+    }
+
+    fn update_in_place(&self, event: &mut Event<Ids, Name>, key: &IdempotencyKey<'_>) {
+        use std::fmt::Write;
+
+        let MetricsKey {
+            metric,
+            tenant_id,
+            timeline_id,
+        } = self.0;
+
+        let (kind, value) = self.1;
+
+        *event = Event {
+            kind,
+            metric,
+            idempotency_key: {
+                event.idempotency_key.clear();
+                write!(event.idempotency_key, "{key}").unwrap();
+                std::mem::take(&mut event.idempotency_key)
+            },
+            value,
+            extra: Ids {
+                tenant_id,
+                timeline_id,
+            },
+        };
+    }
+}
+
+trait KeyGen<'a>: Copy {
+    fn generate(&self) -> IdempotencyKey<'a>;
+}
+
+impl<'a> KeyGen<'a> for &'a str {
+    fn generate(&self) -> IdempotencyKey<'a> {
+        IdempotencyKey::generate(self)
+    }
+}
+
+enum UploadError {
+    Rejected(reqwest::StatusCode),
+    Reqwest(reqwest::Error),
+    Cancelled,
+}
+
+impl std::fmt::Debug for UploadError {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        // use same impl because backoff::retry will log this using both
+        std::fmt::Display::fmt(self, f)
+    }
+}
+
+impl std::fmt::Display for UploadError {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        use UploadError::*;
+
+        match self {
+            Rejected(code) => write!(f, "server rejected the metrics with {code}"),
+            Reqwest(e) => write!(f, "request failed: {e}"),
+            Cancelled => write!(f, "cancelled"),
+        }
+    }
+}
+
+impl UploadError {
+    fn is_reject(&self) -> bool {
+        matches!(self, UploadError::Rejected(_))
+    }
+}
+
+// this is consumed by the test verifiers
+static LAST_IN_BATCH: reqwest::header::HeaderName =
+    reqwest::header::HeaderName::from_static("pageserver-metrics-last-upload-in-batch");
+
+async fn upload(
+    client: &reqwest::Client,
+    metric_collection_endpoint: &reqwest::Url,
+    body: bytes::Bytes,
+    cancel: &CancellationToken,
+    is_last: bool,
+) -> Result<(), UploadError> {
+    let warn_after = 3;
+    let max_attempts = 10;
+    let res = utils::backoff::retry(
+        move || {
+            let body = body.clone();
+            async move {
+                let res = client
+                    .post(metric_collection_endpoint.clone())
+                    .header(reqwest::header::CONTENT_TYPE, "application/json")
+                    .header(
+                        LAST_IN_BATCH.clone(),
+                        if is_last { "true" } else { "false" },
+                    )
+                    .body(body)
+                    .send()
+                    .await;
+
+                let res = res.and_then(|res| res.error_for_status());
+
+                // 10 redirects are normally allowed, so we don't need worry about 3xx
+                match res {
+                    Ok(_response) => Ok(()),
+                    Err(e) => {
+                        let status = e.status().filter(|s| s.is_client_error());
+                        if let Some(status) = status {
+                            // rejection used to be a thing when the server could reject a
+                            // whole batch of metrics if one metric was bad.
+                            Err(UploadError::Rejected(status))
+                        } else {
+                            Err(UploadError::Reqwest(e))
+                        }
+                    }
+                }
+            }
+        },
+        UploadError::is_reject,
+        warn_after,
+        max_attempts,
+        "upload consumption_metrics",
+        utils::backoff::Cancel::new(cancel.clone(), || UploadError::Cancelled),
+    )
+    .await;
+
+    match &res {
+        Ok(_) => {}
+        Err(e) if e.is_reject() => {
+            // permanent errors currently do not get logged by backoff::retry
+            // display alternate has no effect, but keeping it here for easier pattern matching.
+            tracing::error!("failed to upload metrics: {e:#}");
+        }
+        Err(_) => {
+            // these have been logged already
+        }
+    }
+
+    res
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use chrono::{DateTime, Utc};
+    use once_cell::sync::Lazy;
+
+    #[test]
+    fn chunked_serialization() {
+        let examples = metric_samples();
+        assert!(examples.len() > 1);
+
+        let factory = FixedGen::new(Utc::now(), "1", 42);
+
+        // need to use Event here because serde_json::Value uses default hashmap, not linked
+        // hashmap
+        #[derive(serde::Deserialize)]
+        struct EventChunk {
+            events: Vec<Event<Ids, Name>>,
+        }
+
+        let correct = serialize_in_chunks(examples.len(), &examples, factory)
+            .map(|res| res.unwrap().1)
+            .flat_map(|body| serde_json::from_slice::<EventChunk>(&body).unwrap().events)
+            .collect::<Vec<_>>();
+
+        for chunk_size in 1..examples.len() {
+            let actual = serialize_in_chunks(chunk_size, &examples, factory)
+                .map(|res| res.unwrap().1)
+                .flat_map(|body| serde_json::from_slice::<EventChunk>(&body).unwrap().events)
+                .collect::<Vec<_>>();
+
+            // if these are equal, it means that multi-chunking version works as well
+            assert_eq!(correct, actual);
+        }
+    }
+
+    #[derive(Clone, Copy)]
+    struct FixedGen<'a>(chrono::DateTime<chrono::Utc>, &'a str, u16);
+
+    impl<'a> FixedGen<'a> {
+        fn new(now: chrono::DateTime<chrono::Utc>, node_id: &'a str, nonce: u16) -> Self {
+            FixedGen(now, node_id, nonce)
+        }
+    }
+
+    impl<'a> KeyGen<'a> for FixedGen<'a> {
+        fn generate(&self) -> IdempotencyKey<'a> {
+            IdempotencyKey::for_tests(self.0, self.1, self.2)
+        }
+    }
+
+    static SAMPLES_NOW: Lazy<DateTime<Utc>> = Lazy::new(|| {
+        DateTime::parse_from_rfc3339("2023-09-15T00:00:00.123456789Z")
+            .unwrap()
+            .into()
+    });
+
+    #[test]
+    fn metric_image_stability() {
+        // it is important that these strings stay as they are
+
+        let examples = [
+            (
+                line!(),
+                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"written_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000","timeline_id":"ffffffffffffffffffffffffffffffff"}"#,
+            ),
+            (
+                line!(),
+                r#"{"type":"incremental","start_time":"2023-09-14T00:00:00.123456789Z","stop_time":"2023-09-15T00:00:00.123456789Z","metric":"written_data_bytes_delta","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000","timeline_id":"ffffffffffffffffffffffffffffffff"}"#,
+            ),
+            (
+                line!(),
+                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"timeline_logical_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000","timeline_id":"ffffffffffffffffffffffffffffffff"}"#,
+            ),
+            (
+                line!(),
+                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"remote_storage_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000"}"#,
+            ),
+            (
+                line!(),
+                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"resident_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":0,"tenant_id":"00000000000000000000000000000000"}"#,
+            ),
+            (
+                line!(),
+                r#"{"type":"absolute","time":"2023-09-15T00:00:00.123456789Z","metric":"synthetic_storage_size","idempotency_key":"2023-09-15 00:00:00.123456789 UTC-1-0000","value":1,"tenant_id":"00000000000000000000000000000000"}"#,
+            ),
+        ];
+
+        let idempotency_key = consumption_metrics::IdempotencyKey::for_tests(*SAMPLES_NOW, "1", 0);
+        let examples = examples.into_iter().zip(metric_samples());
+
+        for ((line, expected), (key, (kind, value))) in examples {
+            let e = consumption_metrics::Event {
+                kind,
+                metric: key.metric,
+                idempotency_key: idempotency_key.to_string(),
+                value,
+                extra: Ids {
+                    tenant_id: key.tenant_id,
+                    timeline_id: key.timeline_id,
+                },
+            };
+            let actual = serde_json::to_string(&e).unwrap();
+            assert_eq!(expected, actual, "example for {kind:?} from line {line}");
+        }
+    }
+
+    fn metric_samples() -> [RawMetric; 6] {
+        let tenant_id = TenantId::from_array([0; 16]);
+        let timeline_id = TimelineId::from_array([0xff; 16]);
+
+        let before = DateTime::parse_from_rfc3339("2023-09-14T00:00:00.123456789Z")
+            .unwrap()
+            .into();
+        let [now, before] = [*SAMPLES_NOW, before];
+
+        super::super::metrics::metric_examples(tenant_id, timeline_id, now, before)
+    }
+}
--- a/pageserver/src/context.rs
+++ b/pageserver/src/context.rs
@@ -94,6 +94,18 @@ pub struct RequestContext {
    task_kind: TaskKind,
    download_behavior: DownloadBehavior,
    access_stats_behavior: AccessStatsBehavior,
+    page_content_kind: PageContentKind,
+}
+
+/// The kind of access to the page cache.
+#[derive(Clone, Copy, PartialEq, Eq, Debug, enum_map::Enum, strum_macros::IntoStaticStr)]
+pub enum PageContentKind {
+    Unknown,
+    DeltaLayerBtreeNode,
+    DeltaLayerValue,
+    ImageLayerBtreeNode,
+    ImageLayerValue,
+    InMemoryLayer,
 }

 /// Desired behavior if the operation requires an on-demand download
@@ -137,6 +149,7 @@ impl RequestContextBuilder {
                task_kind,
                download_behavior: DownloadBehavior::Download,
                access_stats_behavior: AccessStatsBehavior::Update,
+                page_content_kind: PageContentKind::Unknown,
            },
        }
    }
@@ -149,6 +162,7 @@ impl RequestContextBuilder {
                task_kind: original.task_kind,
                download_behavior: original.download_behavior,
                access_stats_behavior: original.access_stats_behavior,
+                page_content_kind: original.page_content_kind,
            },
        }
    }
@@ -167,6 +181,11 @@ impl RequestContextBuilder {
        self
    }

+    pub(crate) fn page_content_kind(mut self, k: PageContentKind) -> Self {
+        self.inner.page_content_kind = k;
+        self
+    }
+
    pub fn build(self) -> RequestContext {
        self.inner
    }
@@ -263,4 +282,8 @@ impl RequestContext {
    pub(crate) fn access_stats_behavior(&self) -> AccessStatsBehavior {
        self.access_stats_behavior
    }
+
+    pub(crate) fn page_content_kind(&self) -> PageContentKind {
+        self.page_content_kind
+    }
 }
--- a/pageserver/src/control_plane_client.rs
+++ b/pageserver/src/control_plane_client.rs
@@ -1,7 +1,9 @@
 use std::collections::HashMap;

-use hyper::StatusCode;
-use pageserver_api::control_api::{ReAttachRequest, ReAttachResponse};
+use pageserver_api::control_api::{
+    ReAttachRequest, ReAttachResponse, ValidateRequest, ValidateRequestTenant, ValidateResponse,
+};
+use serde::{de::DeserializeOwned, Serialize};
 use tokio_util::sync::CancellationToken;
 use url::Url;
 use utils::{
@@ -12,25 +14,34 @@ use utils::{

 use crate::config::PageServerConf;

-// Backoffs when control plane requests do not succeed: compromise between reducing load
-// on control plane, and retrying frequently when we are blocked on a control plane
-// response to make progress.
-const BACKOFF_INCREMENT: f64 = 0.1;
-const BACKOFF_MAX: f64 = 10.0;
-
 /// The Pageserver's client for using the control plane API: this is a small subset
 /// of the overall control plane API, for dealing with generations (see docs/rfcs/025-generation-numbers.md)
-pub(crate) struct ControlPlaneClient {
+pub struct ControlPlaneClient {
    http_client: reqwest::Client,
    base_url: Url,
    node_id: NodeId,
    cancel: CancellationToken,
 }

+/// Represent operations which internally retry on all errors other than
+/// cancellation token firing: the only way they can fail is ShuttingDown.
+pub enum RetryForeverError {
+    ShuttingDown,
+}
+
+#[async_trait::async_trait]
+pub trait ControlPlaneGenerationsApi {
+    async fn re_attach(&self) -> Result<HashMap<TenantId, Generation>, RetryForeverError>;
+    async fn validate(
+        &self,
+        tenants: Vec<(TenantId, Generation)>,
+    ) -> Result<HashMap<TenantId, bool>, RetryForeverError>;
+}
+
 impl ControlPlaneClient {
    /// A None return value indicates that the input `conf` object does not have control
    /// plane API enabled.
-    pub(crate) fn new(conf: &'static PageServerConf, cancel: &CancellationToken) -> Option<Self> {
+    pub fn new(conf: &'static PageServerConf, cancel: &CancellationToken) -> Option<Self> {
        let mut url = match conf.control_plane_api.as_ref() {
            Some(u) => u.clone(),
            None => return None,
@@ -42,39 +53,78 @@ impl ControlPlaneClient {
            segs.pop_if_empty().push("");
        }

-        let client = reqwest::ClientBuilder::new()
-            .build()
-            .expect("Failed to construct http client");
+        let mut client = reqwest::ClientBuilder::new();
+
+        if let Some(jwt) = &conf.control_plane_api_token {
+            let mut headers = hyper::HeaderMap::new();
+            headers.insert("Authorization", jwt.get_contents().parse().unwrap());
+            client = client.default_headers(headers);
+        }

        Some(Self {
-            http_client: client,
+            http_client: client.build().expect("Failed to construct HTTP client"),
            base_url: url,
            node_id: conf.id,
            cancel: cancel.clone(),
        })
    }

-    async fn try_re_attach(
+    async fn retry_http_forever<R, T>(
        &self,
-        url: Url,
-        request: &ReAttachRequest,
-    ) -> anyhow::Result<ReAttachResponse> {
-        match self.http_client.post(url).json(request).send().await {
-            Err(e) => Err(anyhow::Error::from(e)),
-            Ok(r) => {
-                if r.status() == StatusCode::OK {
-                    r.json::<ReAttachResponse>()
-                        .await
-                        .map_err(anyhow::Error::from)
-                } else {
-                    Err(anyhow::anyhow!("Unexpected status {}", r.status()))
-                }
+        url: &url::Url,
+        request: R,
+    ) -> Result<T, RetryForeverError>
+    where
+        R: Serialize,
+        T: DeserializeOwned,
+    {
+        #[derive(thiserror::Error, Debug)]
+        enum RemoteAttemptError {
+            #[error("shutdown")]
+            Shutdown,
+            #[error("remote: {0}")]
+            Remote(reqwest::Error),
+        }
+
+        match backoff::retry(
+            || async {
+                let response = self
+                    .http_client
+                    .post(url.clone())
+                    .json(&request)
+                    .send()
+                    .await
+                    .map_err(RemoteAttemptError::Remote)?;
+
+                response
+                    .error_for_status_ref()
+                    .map_err(RemoteAttemptError::Remote)?;
+                response
+                    .json::<T>()
+                    .await
+                    .map_err(RemoteAttemptError::Remote)
+            },
+            |_| false,
+            3,
+            u32::MAX,
+            "calling control plane generation validation API",
+            backoff::Cancel::new(self.cancel.clone(), || RemoteAttemptError::Shutdown),
+        )
+        .await
+        {
+            Err(RemoteAttemptError::Shutdown) => Err(RetryForeverError::ShuttingDown),
+            Err(RemoteAttemptError::Remote(_)) => {
+                panic!("We retry forever, this should never be reached");
            }
+            Ok(r) => Ok(r),
        }
    }
+}

-    /// Block until we get a successful response
-    pub(crate) async fn re_attach(&self) -> anyhow::Result<HashMap<TenantId, Generation>> {
+#[async_trait::async_trait]
+impl ControlPlaneGenerationsApi for ControlPlaneClient {
+    /// Block until we get a successful response, or error out if we are shut down
+    async fn re_attach(&self) -> Result<HashMap<TenantId, Generation>, RetryForeverError> {
        let re_attach_path = self
            .base_url
            .join("re-attach")
@@ -83,37 +133,51 @@ impl ControlPlaneClient {
            node_id: self.node_id,
        };

-        let mut attempt = 0;
-        loop {
-            let result = self.try_re_attach(re_attach_path.clone(), &request).await;
-            match result {
-                Ok(res) => {
-                    tracing::info!(
-                        "Received re-attach response with {} tenants",
-                        res.tenants.len()
-                    );
+        fail::fail_point!("control-plane-client-re-attach");

-                    return Ok(res
-                        .tenants
-                        .into_iter()
-                        .map(|t| (t.id, Generation::new(t.generation)))
-                        .collect::<HashMap<_, _>>());
-                }
-                Err(e) => {
-                    tracing::error!("Error re-attaching tenants, retrying: {e:#}");
-                    backoff::exponential_backoff(
-                        attempt,
-                        BACKOFF_INCREMENT,
-                        BACKOFF_MAX,
-                        &self.cancel,
-                    )
-                    .await;
-                    if self.cancel.is_cancelled() {
-                        return Err(anyhow::anyhow!("Shutting down"));
-                    }
-                    attempt += 1;
-                }
-            }
-        }
+        let response: ReAttachResponse = self.retry_http_forever(&re_attach_path, request).await?;
+        tracing::info!(
+            "Received re-attach response with {} tenants",
+            response.tenants.len()
+        );
+
+        Ok(response
+            .tenants
+            .into_iter()
+            .map(|t| (t.id, Generation::new(t.generation)))
+            .collect::<HashMap<_, _>>())
+    }
+
+    /// Block until we get a successful response, or error out if we are shut down
+    async fn validate(
+        &self,
+        tenants: Vec<(TenantId, Generation)>,
+    ) -> Result<HashMap<TenantId, bool>, RetryForeverError> {
+        let re_attach_path = self
+            .base_url
+            .join("validate")
+            .expect("Failed to build validate path");
+
+        let request = ValidateRequest {
+            tenants: tenants
+                .into_iter()
+                .map(|(id, gen)| ValidateRequestTenant {
+                    id,
+                    gen: gen
+                        .into()
+                        .expect("Generation should always be valid for a Tenant doing deletions"),
+                })
+                .collect(),
+        };
+
+        fail::fail_point!("control-plane-client-validate");
+
+        let response: ValidateResponse = self.retry_http_forever(&re_attach_path, request).await?;
+
+        Ok(response
+            .tenants
+            .into_iter()
+            .map(|rt| (rt.id, rt.valid))
+            .collect())
    }
 }
--- a/pageserver/src/deletion_queue.rs
+++ b/pageserver/src/deletion_queue.rs
--- a/pageserver/src/deletion_queue/deleter.rs
+++ b/pageserver/src/deletion_queue/deleter.rs
@@ -0,0 +1,156 @@
+//! The deleter is the final stage in the deletion queue.  It accumulates remote
+//! paths to delete, and periodically executes them in batches of up to 1000
+//! using the DeleteObjects request.
+//!
+//! Its purpose is to increase efficiency of remote storage I/O by issuing a smaller
+//! number of full-sized DeleteObjects requests, rather than a larger number of
+//! smaller requests.
+
+use remote_storage::GenericRemoteStorage;
+use remote_storage::RemotePath;
+use remote_storage::MAX_KEYS_PER_DELETE;
+use std::time::Duration;
+use tokio_util::sync::CancellationToken;
+use tracing::info;
+use tracing::warn;
+
+use crate::metrics;
+
+use super::DeletionQueueError;
+use super::FlushOp;
+
+const AUTOFLUSH_INTERVAL: Duration = Duration::from_secs(10);
+
+pub(super) enum DeleterMessage {
+    Delete(Vec<RemotePath>),
+    Flush(FlushOp),
+}
+
+/// Non-persistent deletion queue, for coalescing multiple object deletes into
+/// larger DeleteObjects requests.
+pub(super) struct Deleter {
+    // Accumulate up to 1000 keys for the next deletion operation
+    accumulator: Vec<RemotePath>,
+
+    rx: tokio::sync::mpsc::Receiver<DeleterMessage>,
+
+    cancel: CancellationToken,
+    remote_storage: GenericRemoteStorage,
+}
+
+impl Deleter {
+    pub(super) fn new(
+        remote_storage: GenericRemoteStorage,
+        rx: tokio::sync::mpsc::Receiver<DeleterMessage>,
+        cancel: CancellationToken,
+    ) -> Self {
+        Self {
+            remote_storage,
+            rx,
+            cancel,
+            accumulator: Vec::new(),
+        }
+    }
+
+    /// Wrap the remote `delete_objects` with a failpoint
+    async fn remote_delete(&self) -> Result<(), anyhow::Error> {
+        fail::fail_point!("deletion-queue-before-execute", |_| {
+            info!("Skipping execution, failpoint set");
+            metrics::DELETION_QUEUE
+                .remote_errors
+                .with_label_values(&["failpoint"])
+                .inc();
+            Err(anyhow::anyhow!("failpoint hit"))
+        });
+
+        self.remote_storage.delete_objects(&self.accumulator).await
+    }
+
+    /// Block until everything in accumulator has been executed
+    async fn flush(&mut self) -> Result<(), DeletionQueueError> {
+        while !self.accumulator.is_empty() && !self.cancel.is_cancelled() {
+            match self.remote_delete().await {
+                Ok(()) => {
+                    // Note: we assume that the remote storage layer returns Ok(()) if some
+                    // or all of the deleted objects were already gone.
+                    metrics::DELETION_QUEUE
+                        .keys_executed
+                        .inc_by(self.accumulator.len() as u64);
+                    info!(
+                        "Executed deletion batch {}..{}",
+                        self.accumulator
+                            .first()
+                            .expect("accumulator should be non-empty"),
+                        self.accumulator
+                            .last()
+                            .expect("accumulator should be non-empty"),
+                    );
+                    self.accumulator.clear();
+                }
+                Err(e) => {
+                    warn!("DeleteObjects request failed: {e:#}, will retry");
+                    metrics::DELETION_QUEUE
+                        .remote_errors
+                        .with_label_values(&["execute"])
+                        .inc();
+                }
+            };
+        }
+        if self.cancel.is_cancelled() {
+            // Expose an error because we may not have actually flushed everything
+            Err(DeletionQueueError::ShuttingDown)
+        } else {
+            Ok(())
+        }
+    }
+
+    pub(super) async fn background(&mut self) -> Result<(), DeletionQueueError> {
+        self.accumulator.reserve(MAX_KEYS_PER_DELETE);
+
+        loop {
+            if self.cancel.is_cancelled() {
+                return Err(DeletionQueueError::ShuttingDown);
+            }
+
+            let msg = match tokio::time::timeout(AUTOFLUSH_INTERVAL, self.rx.recv()).await {
+                Ok(Some(m)) => m,
+                Ok(None) => {
+                    // All queue senders closed
+                    info!("Shutting down");
+                    return Err(DeletionQueueError::ShuttingDown);
+                }
+                Err(_) => {
+                    // Timeout, we hit deadline to execute whatever we have in hand.  These functions will
+                    // return immediately if no work is pending
+                    self.flush().await?;
+
+                    continue;
+                }
+            };
+
+            match msg {
+                DeleterMessage::Delete(mut list) => {
+                    while !list.is_empty() || self.accumulator.len() == MAX_KEYS_PER_DELETE {
+                        if self.accumulator.len() == MAX_KEYS_PER_DELETE {
+                            self.flush().await?;
+                            // If we have received this number of keys, proceed with attempting to execute
+                            assert_eq!(self.accumulator.len(), 0);
+                        }
+
+                        let available_slots = MAX_KEYS_PER_DELETE - self.accumulator.len();
+                        let take_count = std::cmp::min(available_slots, list.len());
+                        for path in list.drain(list.len() - take_count..) {
+                            self.accumulator.push(path);
+                        }
+                    }
+                }
+                DeleterMessage::Flush(flush_op) => {
+                    // If flush() errors, we drop the flush_op and the caller will get
+                    // an error recv()'ing their oneshot channel.
+                    self.flush().await?;
+                    flush_op.notify();
+                }
+            }
+        }
+    }
+}
--- a/pageserver/src/deletion_queue/list_writer.rs
+++ b/pageserver/src/deletion_queue/list_writer.rs
@@ -0,0 +1,478 @@
+//! The list writer is the first stage in the deletion queue.  It accumulates
+//! layers to delete, and periodically writes out these layers into a persistent
+//! DeletionList.
+//!
+//! The purpose of writing DeletionLists is to decouple the decision to
+//! delete an object from the validation required to execute it: even if
+//! validation is not possible, e.g. due to a control plane outage, we can
+//! still persist our intent to delete an object, in a way that would
+//! survive a restart.
+//!
+//! DeletionLists are passed onwards to the Validator.
+
+use super::DeletionHeader;
+use super::DeletionList;
+use super::FlushOp;
+use super::ValidatorQueueMessage;
+
+use std::collections::HashMap;
+use std::fs::create_dir_all;
+use std::time::Duration;
+
+use regex::Regex;
+use remote_storage::RemotePath;
+use tokio_util::sync::CancellationToken;
+use tracing::debug;
+use tracing::info;
+use tracing::warn;
+use utils::generation::Generation;
+use utils::id::TenantId;
+use utils::id::TimelineId;
+
+use crate::config::PageServerConf;
+use crate::deletion_queue::TEMP_SUFFIX;
+use crate::metrics;
+use crate::tenant::remote_timeline_client::remote_layer_path;
+use crate::tenant::storage_layer::LayerFileName;
+
+// The number of keys in a DeletionList before we will proactively persist it
+// (without reaching a flush deadline).  This aims to deliver objects of the order
+// of magnitude 1MB when we are under heavy delete load.
+const DELETION_LIST_TARGET_SIZE: usize = 16384;
+
+// Ordinarily, we only flush to DeletionList periodically, to bound the window during
+// which we might leak objects from not flushing a DeletionList after
+// the objects are already unlinked from timeline metadata.
+const FRONTEND_DEFAULT_TIMEOUT: Duration = Duration::from_millis(10000);
+
+// If someone is waiting for a flush to DeletionList, only delay a little to accumulate
+// more objects before doing the flush.
+const FRONTEND_FLUSHING_TIMEOUT: Duration = Duration::from_millis(100);
+
+#[derive(Debug)]
+pub(super) struct DeletionOp {
+    pub(super) tenant_id: TenantId,
+    pub(super) timeline_id: TimelineId,
+    // `layers` and `objects` are both just lists of objects.  `layers` is used if you do not
+    // have a config object handy to project it to a remote key, and need the consuming worker
+    // to do it for you.
+    pub(super) layers: Vec<(LayerFileName, Generation)>,
+    pub(super) objects: Vec<RemotePath>,
+
+    /// The _current_ generation of the Tenant attachment in which we are enqueuing
+    /// this deletion.
+    pub(super) generation: Generation,
+}
+
+#[derive(Debug)]
+pub(super) struct RecoverOp {
+    pub(super) attached_tenants: HashMap<TenantId, Generation>,
+}
+
+#[derive(Debug)]
+pub(super) enum ListWriterQueueMessage {
+    Delete(DeletionOp),
+    // Wait until all prior deletions make it into a persistent DeletionList
+    Flush(FlushOp),
+    // Wait until all prior deletions have been executed (i.e. objects are actually deleted)
+    FlushExecute(FlushOp),
+    // Call once after re-attaching to control plane, to notify the deletion queue about
+    // latest attached generations & load any saved deletion lists from disk.
+    Recover(RecoverOp),
+}
+
+pub(super) struct ListWriter {
+    conf: &'static PageServerConf,
+
+    // Incoming frontend requests to delete some keys
+    rx: tokio::sync::mpsc::Receiver<ListWriterQueueMessage>,
+
+    // Outbound requests to the backend to execute deletion lists we have composed.
+    tx: tokio::sync::mpsc::Sender<ValidatorQueueMessage>,
+
+    // The list we are currently building, contains a buffer of keys to delete
+    // and our next sequence number
+    pending: DeletionList,
+
+    // These FlushOps should notify the next time we flush
+    pending_flushes: Vec<FlushOp>,
+
+    // Worker loop is torn down when this fires.
+    cancel: CancellationToken,
+
+    // Safety guard to do recovery exactly once
+    recovered: bool,
+}
+
+impl ListWriter {
+    // Initially DeletionHeader.validated_sequence is zero.  The place we start our
+    // sequence numbers must be higher than that.
+    const BASE_SEQUENCE: u64 = 1;
+
+    pub(super) fn new(
+        conf: &'static PageServerConf,
+        rx: tokio::sync::mpsc::Receiver<ListWriterQueueMessage>,
+        tx: tokio::sync::mpsc::Sender<ValidatorQueueMessage>,
+        cancel: CancellationToken,
+    ) -> Self {
+        Self {
+            pending: DeletionList::new(Self::BASE_SEQUENCE),
+            conf,
+            rx,
+            tx,
+            pending_flushes: Vec::new(),
+            cancel,
+            recovered: false,
+        }
+    }
+
+    /// Try to flush `list` to persistent storage
+    ///
+    /// This does not return errors, because on failure to flush we do not lose
+    /// any state: flushing will be retried implicitly on the next deadline
+    async fn flush(&mut self) {
+        if self.pending.is_empty() {
+            for f in self.pending_flushes.drain(..) {
+                f.notify();
+            }
+            return;
+        }
+
+        match self.pending.save(self.conf).await {
+            Ok(_) => {
+                info!(sequence = self.pending.sequence, "Stored deletion list");
+
+                for f in self.pending_flushes.drain(..) {
+                    f.notify();
+                }
+
+                // Take the list we've accumulated, replace it with a fresh list for the next sequence
+                let next_list = DeletionList::new(self.pending.sequence + 1);
+                let list = std::mem::replace(&mut self.pending, next_list);
+
+                if let Err(e) = self.tx.send(ValidatorQueueMessage::Delete(list)).await {
+                    // This is allowed to fail: it will only happen if the backend worker is shut down,
+                    // so we can just drop this on the floor.
+                    info!("Deletion list dropped, this is normal during shutdown ({e:#})");
+                }
+            }
+            Err(e) => {
+                metrics::DELETION_QUEUE.unexpected_errors.inc();
+                warn!(
+                    sequence = self.pending.sequence,
+                    "Failed to write deletion list, will retry later ({e:#})"
+                );
+            }
+        }
+    }
+
+    /// Load the header, to learn the sequence number up to which deletions
+    /// have been validated.  We will apply validated=true to DeletionLists
+    /// <= this sequence when loading them.
+    ///
+    /// It is not an error for the header to not exist: we return None, and
+    /// the caller should act as if validated_sequence is 0
+    async fn load_validated_sequence(&self) -> Result<Option<u64>, anyhow::Error> {
+        let header_path = self.conf.deletion_header_path();
+        match tokio::fs::read(&header_path).await {
+            Ok(header_bytes) => {
+                match serde_json::from_slice::<DeletionHeader>(&header_bytes) {
+                    Ok(h) => Ok(Some(h.validated_sequence)),
+                    Err(e) => {
+                        warn!(
+                            "Failed to deserialize deletion header, ignoring {header_path}: {e:#}",
+                        );
+                        // This should never happen unless we make a mistake with our serialization.
+                        // Ignoring a deletion header is not consequential for correctnes because all deletions
+                        // are ultimately allowed to fail: worst case we leak some objects for the scrubber to clean up.
+                        metrics::DELETION_QUEUE.unexpected_errors.inc();
+                        Ok(None)
+                    }
+                }
+            }
+            Err(e) => {
+                if e.kind() == std::io::ErrorKind::NotFound {
+                    debug!("Deletion header {header_path} not found, first start?");
+                    Ok(None)
+                } else {
+                    Err(anyhow::anyhow!(e))
+                }
+            }
+        }
+    }
+
+    async fn recover(
+        &mut self,
+        attached_tenants: HashMap<TenantId, Generation>,
+    ) -> Result<(), anyhow::Error> {
+        debug!(
+            "recovering with {} attached tenants",
+            attached_tenants.len()
+        );
+
+        // Load the header
+        let validated_sequence = self.load_validated_sequence().await?.unwrap_or(0);
+
+        self.pending.sequence = validated_sequence + 1;
+
+        let deletion_directory = self.conf.deletion_prefix();
+        let mut dir = match tokio::fs::read_dir(&deletion_directory).await {
+            Ok(d) => d,
+            Err(e) => {
+                warn!("Failed to open deletion list directory {deletion_directory}: {e:#}");
+
+                // Give up: if we can't read the deletion list directory, we probably can't
+                // write lists into it later, so the queue won't work.
+                return Err(e.into());
+            }
+        };
+
+        let list_name_pattern =
+            Regex::new("(?<sequence>[a-zA-Z0-9]{16})-(?<version>[a-zA-Z0-9]{2}).list").unwrap();
+
+        let header_path = self.conf.deletion_header_path();
+        let mut seqs: Vec<u64> = Vec::new();
+        while let Some(dentry) = dir.next_entry().await? {
+            let file_name = dentry.file_name();
+            let dentry_str = file_name.to_string_lossy();
+
+            if file_name == header_path.file_name().unwrap_or("") {
+                // Don't try and parse the header's name like a list
+                continue;
+            }
+
+            if dentry_str.ends_with(TEMP_SUFFIX) {
+                info!("Cleaning up temporary file {dentry_str}");
+                let absolute_path =
+                    deletion_directory.join(dentry.file_name().to_str().expect("non-Unicode path"));
+                if let Err(e) = tokio::fs::remove_file(&absolute_path).await {
+                    // Non-fatal error: we will just leave the file behind but not
+                    // try and load it.
+                    warn!("Failed to clean up temporary file {absolute_path}: {e:#}");
+                }
+
+                continue;
+            }
+
+            let file_name = dentry.file_name().to_owned();
+            let basename = file_name.to_string_lossy();
+            let seq_part = if let Some(m) = list_name_pattern.captures(&basename) {
+                m.name("sequence")
+                    .expect("Non optional group should be present")
+                    .as_str()
+            } else {
+                warn!("Unexpected key in deletion queue: {basename}");
+                metrics::DELETION_QUEUE.unexpected_errors.inc();
+                continue;
+            };
+
+            let seq: u64 = match u64::from_str_radix(seq_part, 16) {
+                Ok(s) => s,
+                Err(e) => {
+                    warn!("Malformed key '{basename}': {e}");
+                    metrics::DELETION_QUEUE.unexpected_errors.inc();
+                    continue;
+                }
+            };
+            seqs.push(seq);
+        }
+        seqs.sort();
+
+        // Start our next deletion list from after the last location validated by
+        // previous process lifetime, or after the last location found (it is updated
+        // below after enumerating the deletion lists)
+        self.pending.sequence = validated_sequence + 1;
+        if let Some(max_list_seq) = seqs.last() {
+            self.pending.sequence = std::cmp::max(self.pending.sequence, max_list_seq + 1);
+        }
+
+        for s in seqs {
+            let list_path = self.conf.deletion_list_path(s);
+
+            let list_bytes = tokio::fs::read(&list_path).await?;
+
+            let mut deletion_list = match serde_json::from_slice::<DeletionList>(&list_bytes) {
+                Ok(l) => l,
+                Err(e) => {
+                    // Drop the list on the floor: any objects it referenced will be left behind
+                    // for scrubbing to clean up.  This should never happen unless we have a serialization bug.
+                    warn!(sequence = s, "Failed to deserialize deletion list: {e}");
+                    metrics::DELETION_QUEUE.unexpected_errors.inc();
+                    continue;
+                }
+            };
+
+            if deletion_list.sequence <= validated_sequence {
+                // If the deletion list falls below valid_seq, we may assume that it was
+                // already validated the last time this pageserver ran.  Otherwise, we still
+                // load it, as it may still contain content valid in this generation.
+                deletion_list.validated = true;
+            } else {
+                // Special case optimization: if a tenant is still attached, and no other
+                // generation was issued to another node in the interval while we restarted,
+                // then we may treat deletion lists from the previous generation as if they
+                // belong to our currently attached generation, and proceed to validate & execute.
+                for (tenant_id, tenant_list) in &mut deletion_list.tenants {
+                    if let Some(attached_gen) = attached_tenants.get(tenant_id) {
+                        if attached_gen.previous() == tenant_list.generation {
+                            tenant_list.generation = *attached_gen;
+                        }
+                    }
+                }
+            }
+
+            info!(
+                validated = deletion_list.validated,
+                sequence = deletion_list.sequence,
+                "Recovered deletion list"
+            );
+
+            // We will drop out of recovery if this fails: it indicates that we are shutting down
+            // or the backend has panicked
+            metrics::DELETION_QUEUE
+                .keys_submitted
+                .inc_by(deletion_list.len() as u64);
+            self.tx
+                .send(ValidatorQueueMessage::Delete(deletion_list))
+                .await?;
+        }
+
+        info!(next_sequence = self.pending.sequence, "Replay complete");
+
+        Ok(())
+    }
+
+    /// This is the front-end ingest, where we bundle up deletion requests into DeletionList
+    /// and write them out, for later validation by the backend and execution by the executor.
+    pub(super) async fn background(&mut self) {
+        info!("Started deletion frontend worker");
+
+        // Synchronous, but we only do it once per process lifetime so it's tolerable
+        if let Err(e) = create_dir_all(&self.conf.deletion_prefix()) {
+            tracing::error!(
+                "Failed to create deletion list directory {}, deletions will not be executed ({e})",
+                self.conf.deletion_prefix(),
+            );
+            metrics::DELETION_QUEUE.unexpected_errors.inc();
+            return;
+        }
+
+        while !self.cancel.is_cancelled() {
+            let timeout = if self.pending_flushes.is_empty() {
+                FRONTEND_DEFAULT_TIMEOUT
+            } else {
+                FRONTEND_FLUSHING_TIMEOUT
+            };
+
+            let msg = match tokio::time::timeout(timeout, self.rx.recv()).await {
+                Ok(Some(msg)) => msg,
+                Ok(None) => {
+                    // Queue sender destroyed, shutting down
+                    break;
+                }
+                Err(_) => {
+                    // Hit deadline, flush.
+                    self.flush().await;
+                    continue;
+                }
+            };
+
+            match msg {
+                ListWriterQueueMessage::Delete(op) => {
+                    assert!(
+                        self.recovered,
+                        "Cannot process deletions before recovery.  This is a bug."
+                    );
+
+                    debug!(
+                        "Delete: ingesting {} layers, {} other objects",
+                        op.layers.len(),
+                        op.objects.len()
+                    );
+
+                    let mut layer_paths = Vec::new();
+                    for (layer, generation) in op.layers {
+                        layer_paths.push(remote_layer_path(
+                            &op.tenant_id,
+                            &op.timeline_id,
+                            &layer,
+                            generation,
+                        ));
+                    }
+                    layer_paths.extend(op.objects);
+
+                    if !self.pending.push(
+                        &op.tenant_id,
+                        &op.timeline_id,
+                        op.generation,
+                        &mut layer_paths,
+                    ) {
+                        self.flush().await;
+                        let retry_succeeded = self.pending.push(
+                            &op.tenant_id,
+                            &op.timeline_id,
+                            op.generation,
+                            &mut layer_paths,
+                        );
+                        if !retry_succeeded {
+                            // Unexpected: after we flush, we should have
+                            // drained self.pending, so a conflict on
+                            // generation numbers should be impossible.
+                            tracing::error!(
+                                "Failed to enqueue deletions, leaking objects.  This is a bug."
+                            );
+                            metrics::DELETION_QUEUE.unexpected_errors.inc();
+                        }
+                    }
+                }
+                ListWriterQueueMessage::Flush(op) => {
+                    if self.pending.is_empty() {
+                        // Execute immediately
+                        debug!("Flush: No pending objects, flushing immediately");
+                        op.notify()
+                    } else {
+                        // Execute next time we flush
+                        debug!("Flush: adding to pending flush list for next deadline flush");
+                        self.pending_flushes.push(op);
+                    }
+                }
+                ListWriterQueueMessage::FlushExecute(op) => {
+                    debug!("FlushExecute: passing through to backend");
+                    // We do not flush to a deletion list here: the client sends a Flush before the FlushExecute
+                    if let Err(e) = self.tx.send(ValidatorQueueMessage::Flush(op)).await {
+                        info!("Can't flush, shutting down ({e})");
+                        // Caller will get error when their oneshot sender was dropped.
+                    }
+                }
+                ListWriterQueueMessage::Recover(op) => {
+                    if self.recovered {
+                        tracing::error!(
+                            "Deletion queue recovery called more than once.  This is a bug."
+                        );
+                        metrics::DELETION_QUEUE.unexpected_errors.inc();
+                        // Non-fatal: although this is a bug, since we did recovery at least once we may proceed.
+                        continue;
+                    }
+
+                    if let Err(e) = self.recover(op.attached_tenants).await {
+                        // This should only happen in truly unrecoverable cases, like the recovery finding that the backend
+                        // queue receiver has been dropped, or something is critically broken with
+                        // the local filesystem holding deletion lists.
+                        info!(
+                            "Deletion queue recover aborted, deletion queue will not proceed ({e})"
+                        );
+                        metrics::DELETION_QUEUE.unexpected_errors.inc();
+                        return;
+                    } else {
+                        self.recovered = true;
+                    }
+                }
+            }
+
+            if self.pending.len() > DELETION_LIST_TARGET_SIZE || !self.pending_flushes.is_empty() {
+                self.flush().await;
+            }
+        }
+        info!("Deletion queue shut down.");
+    }
+}
--- a/pageserver/src/deletion_queue/validator.rs
+++ b/pageserver/src/deletion_queue/validator.rs
@@ -0,0 +1,416 @@
+//! The validator is responsible for validating DeletionLists for execution,
+//! based on whethe the generation in the DeletionList is still the latest
+//! generation for a tenant.
+//!
+//! The purpose of validation is to ensure split-brain safety in the cluster
+//! of pageservers: a deletion may only be executed if the tenant generation
+//! that originated it is still current.  See docs/rfcs/025-generation-numbers.md
+//! The purpose of accumulating lists before validating them is to reduce load
+//! on the control plane API by issuing fewer, larger requests.
+//!
+//! In addition to validating DeletionLists, the validator validates updates to remote_consistent_lsn
+//! for timelines: these are logically deletions because the safekeepers use remote_consistent_lsn
+//! to decide when old
+//!
+//! Deletions are passed onward to the Deleter.
+
+use std::collections::HashMap;
+use std::sync::Arc;
+use std::time::Duration;
+
+use camino::Utf8PathBuf;
+use tokio_util::sync::CancellationToken;
+use tracing::debug;
+use tracing::info;
+use tracing::warn;
+
+use crate::config::PageServerConf;
+use crate::control_plane_client::ControlPlaneGenerationsApi;
+use crate::control_plane_client::RetryForeverError;
+use crate::metrics;
+
+use super::deleter::DeleterMessage;
+use super::DeletionHeader;
+use super::DeletionList;
+use super::DeletionQueueError;
+use super::FlushOp;
+use super::VisibleLsnUpdates;
+
+// After this length of time, do any validation work that is pending,
+// even if we haven't accumulated many keys to delete.
+//
+// This also causes updates to remote_consistent_lsn to be validated, even
+// if there were no deletions enqueued.
+const AUTOFLUSH_INTERVAL: Duration = Duration::from_secs(10);
+
+// If we have received this number of keys, proceed with attempting to execute
+const AUTOFLUSH_KEY_COUNT: usize = 16384;
+
+#[derive(Debug)]
+pub(super) enum ValidatorQueueMessage {
+    Delete(DeletionList),
+    Flush(FlushOp),
+}
+pub(super) struct Validator<C>
+where
+    C: ControlPlaneGenerationsApi,
+{
+    conf: &'static PageServerConf,
+    rx: tokio::sync::mpsc::Receiver<ValidatorQueueMessage>,
+    tx: tokio::sync::mpsc::Sender<DeleterMessage>,
+
+    // Client for calling into control plane API for validation of deletes
+    control_plane_client: Option<C>,
+
+    // DeletionLists which are waiting generation validation.  Not safe to
+    // execute until [`validate`] has processed them.
+    pending_lists: Vec<DeletionList>,
+
+    // DeletionLists which have passed validation and are ready to execute.
+    validated_lists: Vec<DeletionList>,
+
+    // Sum of all the lengths of lists in pending_lists
+    pending_key_count: usize,
+
+    // Lsn validation state: we read projected LSNs and write back visible LSNs
+    // after validation.  This is the LSN equivalent of `pending_validation_lists`:
+    // it is drained in [`validate`]
+    lsn_table: Arc<std::sync::RwLock<VisibleLsnUpdates>>,
+
+    // If we failed to rewrite a deletion list due to local filesystem I/O failure,
+    // we must remember that and refuse to advance our persistent validated sequence
+    // number past the failure.
+    list_write_failed: Option<u64>,
+
+    cancel: CancellationToken,
+}
+
+impl<C> Validator<C>
+where
+    C: ControlPlaneGenerationsApi,
+{
+    pub(super) fn new(
+        conf: &'static PageServerConf,
+        rx: tokio::sync::mpsc::Receiver<ValidatorQueueMessage>,
+        tx: tokio::sync::mpsc::Sender<DeleterMessage>,
+        control_plane_client: Option<C>,
+        lsn_table: Arc<std::sync::RwLock<VisibleLsnUpdates>>,
+        cancel: CancellationToken,
+    ) -> Self {
+        Self {
+            conf,
+            rx,
+            tx,
+            control_plane_client,
+            lsn_table,
+            pending_lists: Vec::new(),
+            validated_lists: Vec::new(),
+            pending_key_count: 0,
+            list_write_failed: None,
+            cancel,
+        }
+    }
+    /// Process any outstanding validations of generations of pending LSN updates or pending
+    /// DeletionLists.
+    ///
+    /// Valid LSN updates propagate back to Timelines immediately, valid DeletionLists
+    /// go into the queue of ready-to-execute lists.
+    async fn validate(&mut self) -> Result<(), DeletionQueueError> {
+        let mut tenant_generations = HashMap::new();
+        for list in &self.pending_lists {
+            for (tenant_id, tenant_list) in &list.tenants {
+                // Note: DeletionLists are in logical time order, so generation always
+                // goes up.  By doing a simple insert() we will always end up with
+                // the latest generation seen for a tenant.
+                tenant_generations.insert(*tenant_id, tenant_list.generation);
+            }
+        }
+
+        let pending_lsn_updates = {
+            let mut lsn_table = self.lsn_table.write().expect("Lock should not be poisoned");
+            std::mem::take(&mut *lsn_table)
+        };
+        for (tenant_id, update) in &pending_lsn_updates.tenants {
+            let entry = tenant_generations
+                .entry(*tenant_id)
+                .or_insert(update.generation);
+            if update.generation > *entry {
+                *entry = update.generation;
+            }
+        }
+
+        if tenant_generations.is_empty() {
+            // No work to do
+            return Ok(());
+        }
+
+        let tenants_valid = if let Some(control_plane_client) = &self.control_plane_client {
+            match control_plane_client
+                .validate(tenant_generations.iter().map(|(k, v)| (*k, *v)).collect())
+                .await
+            {
+                Ok(tenants) => tenants,
+                Err(RetryForeverError::ShuttingDown) => {
+                    // The only way a validation call returns an error is when the cancellation token fires
+                    return Err(DeletionQueueError::ShuttingDown);
+                }
+            }
+        } else {
+            // Control plane API disabled.  In legacy mode we consider everything valid.
+            tenant_generations.keys().map(|k| (*k, true)).collect()
+        };
+
+        let mut validated_sequence: Option<u64> = None;
+
+        // Apply the validation results to the pending LSN updates
+        for (tenant_id, tenant_lsn_state) in pending_lsn_updates.tenants {
+            let validated_generation = tenant_generations
+                .get(&tenant_id)
+                .expect("Map was built from the same keys we're reading");
+
+            let valid = tenants_valid
+                .get(&tenant_id)
+                .copied()
+                // If the tenant was missing from the validation response, it has been deleted.
+                // The Timeline that requested the LSN update is probably already torn down,
+                // or will be torn down soon.  In this case, drop the update by setting valid=false.
+                .unwrap_or(false);
+
+            if valid && *validated_generation == tenant_lsn_state.generation {
+                for (_timeline_id, pending_lsn) in tenant_lsn_state.timelines {
+                    pending_lsn.result_slot.store(pending_lsn.projected);
+                }
+            } else {
+                // If we failed validation, then do not apply any of the projected updates
+                warn!("Dropped remote consistent LSN updates for tenant {tenant_id} in stale generation {:?}", tenant_lsn_state.generation);
+                metrics::DELETION_QUEUE.dropped_lsn_updates.inc();
+            }
+        }
+
+        // Apply the validation results to the pending deletion lists
+        for list in &mut self.pending_lists {
+            // Filter the list based on whether the server responded valid: true.
+            // If a tenant is omitted in the response, it has been deleted, and we should
+            // proceed with deletion.
+            let mut mutated = false;
+            list.tenants.retain(|tenant_id, tenant| {
+                let validated_generation = tenant_generations
+                    .get(tenant_id)
+                    .expect("Map was built from the same keys we're reading");
+
+                // If the tenant was missing from the validation response, it has been deleted.
+                // This means that a deletion is valid, but also redundant since the tenant's
+                // objects should have already been deleted.  Treat it as invalid to drop the
+                // redundant deletion.
+                let valid = tenants_valid.get(tenant_id).copied().unwrap_or(false);
+
+                // A list is valid if it comes from the current _or previous_ generation.
+                // - The previous generation case is permitted due to how we store deletion lists locally:
+                // if we see the immediately previous generation in a locally stored deletion list,
+                // it proves that this node's disk was used for both current & previous generations,
+                // and therefore no other node was involved in between: the two generations may be
+                // logically treated as the same.
+                // - In that previous generation case, we rewrote it to the current generation
+                // in recover(), so the comparison here is simply an equality.
+
+                let this_list_valid = valid
+                    && (tenant.generation == *validated_generation);
+
+                if !this_list_valid {
+                    warn!("Dropping stale deletions for tenant {tenant_id} in generation {:?}, objects may be leaked", tenant.generation);
+                    metrics::DELETION_QUEUE.keys_dropped.inc_by(tenant.len() as u64);
+                    mutated = true;
+                } else {
+                    metrics::DELETION_QUEUE.keys_validated.inc_by(tenant.len() as u64);
+                }
+                this_list_valid
+            });
+            list.validated = true;
+
+            if mutated {
+                // Save the deletion list if we had to make changes due to stale generations.  The
+                // saved list is valid for execution.
+                if let Err(e) = list.save(self.conf).await {
+                    // Highly unexpected.  Could happen if e.g. disk full.
+                    // If we didn't save the trimmed list, it is _not_ valid to execute.
+                    warn!("Failed to save modified deletion list {list}: {e:#}");
+                    metrics::DELETION_QUEUE.unexpected_errors.inc();
+
+                    // Rather than have a complex retry process, just drop it and leak the objects,
+                    // scrubber will clean up eventually.
+                    list.tenants.clear(); // Result is a valid-but-empty list, which is a no-op for execution.
+
+                    // We must remember this failure, to prevent later writing out a header that
+                    // would imply the unwritable list was valid on disk.
+                    if self.list_write_failed.is_none() {
+                        self.list_write_failed = Some(list.sequence);
+                    }
+                }
+            }
+
+            validated_sequence = Some(list.sequence);
+        }
+
+        if let Some(validated_sequence) = validated_sequence {
+            if let Some(list_write_failed) = self.list_write_failed {
+                // Rare error case: we failed to write out a deletion list to excise invalid
+                // entries, so we cannot advance the header's valid sequence number past that point.
+                //
+                // In this state we will continue to validate, execute and delete deletion lists,
+                // we just cannot update the header.  It should be noticed and fixed by a human due to
+                // the nonzero value of our unexpected_errors metric.
+                warn!(
+                    sequence_number = list_write_failed,
+                    "Cannot write header because writing a deletion list failed earlier",
+                );
+            } else {
+                // Write the queue header to record how far validation progressed.  This avoids having
+                // to rewrite each DeletionList to set validated=true in it.
+                let header = DeletionHeader::new(validated_sequence);
+
+                // Drop result because the validated_sequence is an optimization.  If we fail to save it,
+                // then restart, we will drop some deletion lists, creating work for scrubber.
+                // The save() function logs a warning on error.
+                if let Err(e) = header.save(self.conf).await {
+                    warn!("Failed to write deletion queue header: {e:#}");
+                    metrics::DELETION_QUEUE.unexpected_errors.inc();
+                }
+            }
+        }
+
+        // Transfer the validated lists to the validated queue, for eventual execution
+        self.validated_lists.append(&mut self.pending_lists);
+
+        Ok(())
+    }
+
+    async fn cleanup_lists(&mut self, list_paths: Vec<Utf8PathBuf>) {
+        for list_path in list_paths {
+            debug!("Removing deletion list {list_path}");
+
+            if let Err(e) = tokio::fs::remove_file(&list_path).await {
+                // Unexpected: we should have permissions and nothing else should
+                // be touching these files.  We will leave the file behind.  Subsequent
+                // pageservers will try and load it again: hopefully whatever storage
+                // issue (probably permissions) has been fixed by then.
+                tracing::error!("Failed to delete {list_path}: {e:#}");
+                metrics::DELETION_QUEUE.unexpected_errors.inc();
+                break;
+            }
+        }
+    }
+
+    async fn flush(&mut self) -> Result<(), DeletionQueueError> {
+        tracing::debug!("Flushing with {} pending lists", self.pending_lists.len());
+
+        // Issue any required generation validation calls to the control plane
+        self.validate().await?;
+
+        // After successful validation, nothing is pending: any lists that
+        // made it through validation will be in validated_lists.
+        assert!(self.pending_lists.is_empty());
+        self.pending_key_count = 0;
+
+        tracing::debug!(
+            "Validation complete, have {} validated lists",
+            self.validated_lists.len()
+        );
+
+        // Return quickly if we have no validated lists to execute.  This avoids flushing the
+        // executor when an idle backend hits its autoflush interval
+        if self.validated_lists.is_empty() {
+            return Ok(());
+        }
+
+        // Drain `validated_lists` into the executor
+        let mut executing_lists = Vec::new();
+        for list in self.validated_lists.drain(..) {
+            let list_path = self.conf.deletion_list_path(list.sequence);
+            let objects = list.into_remote_paths();
+            self.tx
+                .send(DeleterMessage::Delete(objects))
+                .await
+                .map_err(|_| DeletionQueueError::ShuttingDown)?;
+            executing_lists.push(list_path);
+        }
+
+        self.flush_executor().await?;
+
+        // Erase the deletion lists whose keys have all be deleted from remote storage
+        self.cleanup_lists(executing_lists).await;
+
+        Ok(())
+    }
+
+    async fn flush_executor(&mut self) -> Result<(), DeletionQueueError> {
+        // Flush the executor, so that all the keys referenced by these deletion lists
+        // are actually removed from remote storage.  This is a precondition to deleting
+        // the deletion lists themselves.
+        let (flush_op, rx) = FlushOp::new();
+        self.tx
+            .send(DeleterMessage::Flush(flush_op))
+            .await
+            .map_err(|_| DeletionQueueError::ShuttingDown)?;
+
+        rx.await.map_err(|_| DeletionQueueError::ShuttingDown)
+    }
+
+    pub(super) async fn background(&mut self) {
+        tracing::info!("Started deletion backend worker");
+
+        while !self.cancel.is_cancelled() {
+            let msg = match tokio::time::timeout(AUTOFLUSH_INTERVAL, self.rx.recv()).await {
+                Ok(Some(m)) => m,
+                Ok(None) => {
+                    // All queue senders closed
+                    info!("Shutting down");
+                    break;
+                }
+                Err(_) => {
+                    // Timeout, we hit deadline to execute whatever we have in hand.  These functions will
+                    // return immediately if no work is pending.
+                    match self.flush().await {
+                        Ok(()) => {}
+                        Err(DeletionQueueError::ShuttingDown) => {
+                            // If we are shutting down, then auto-flush can safely be skipped
+                        }
+                    }
+
+                    continue;
+                }
+            };
+
+            match msg {
+                ValidatorQueueMessage::Delete(list) => {
+                    if list.validated {
+                        // A pre-validated list may only be seen during recovery, if we are recovering
+                        // a DeletionList whose on-disk state has validated=true
+                        self.validated_lists.push(list)
+                    } else {
+                        self.pending_key_count += list.len();
+                        self.pending_lists.push(list);
+                    }
+
+                    if self.pending_key_count > AUTOFLUSH_KEY_COUNT {
+                        match self.flush().await {
+                            Ok(()) => {}
+                            Err(DeletionQueueError::ShuttingDown) => {
+                                // If we are shutting down, then auto-flush can safely be skipped
+                            }
+                        }
+                    }
+                }
+                ValidatorQueueMessage::Flush(op) => {
+                    match self.flush().await {
+                        Ok(()) => {
+                            op.notify();
+                        }
+                        Err(DeletionQueueError::ShuttingDown) => {
+                            // If we fail due to shutting down, we will just drop `op` to propagate that status.
+                        }
+                    }
+                }
+            }
+        }
+    }
+}
--- a/pageserver/src/disk_usage_eviction_task.rs
+++ b/pageserver/src/disk_usage_eviction_task.rs
@@ -43,12 +43,12 @@

 use std::{
    collections::HashMap,
-    path::Path,
    sync::Arc,
    time::{Duration, SystemTime},
 };

 use anyhow::Context;
+use camino::Utf8Path;
 use remote_storage::GenericRemoteStorage;
 use serde::{Deserialize, Serialize};
 use tokio::time::Instant;
@@ -122,7 +122,7 @@ async fn disk_usage_eviction_task(
    state: &State,
    task_config: &DiskUsageEvictionTaskConfig,
    storage: GenericRemoteStorage,
-    tenants_dir: &Path,
+    tenants_dir: &Utf8Path,
    cancel: CancellationToken,
 ) {
    scopeguard::defer! {
@@ -184,7 +184,7 @@ async fn disk_usage_eviction_task_iteration(
    state: &State,
    task_config: &DiskUsageEvictionTaskConfig,
    storage: &GenericRemoteStorage,
-    tenants_dir: &Path,
+    tenants_dir: &Utf8Path,
    cancel: &CancellationToken,
 ) -> anyhow::Result<()> {
    let usage_pre = filesystem_level_usage::get(tenants_dir, task_config)
@@ -620,9 +620,8 @@ impl std::ops::Deref for TimelineKey {
 }

 mod filesystem_level_usage {
-    use std::path::Path;
-
    use anyhow::Context;
+    use camino::Utf8Path;

    use crate::statvfs::Statvfs;

@@ -664,7 +663,7 @@ mod filesystem_level_usage {
    }

    pub fn get<'a>(
-        tenants_dir: &Path,
+        tenants_dir: &Utf8Path,
        config: &'a DiskUsageEvictionTaskConfig,
    ) -> anyhow::Result<Usage<'a>> {
        let mock_config = {
--- a/pageserver/src/http/openapi_spec.yml
+++ b/pageserver/src/http/openapi_spec.yml
@@ -1093,6 +1093,9 @@ components:
        remote_consistent_lsn:
          type: string
          format: hex
+        remote_consistent_lsn_visible:
+          type: string
+          format: hex
        ancestor_timeline_id:
          type: string
          format: hex
--- a/pageserver/src/http/routes.rs
+++ b/pageserver/src/http/routes.rs
@@ -5,11 +5,13 @@ use std::collections::HashMap;
 use std::sync::Arc;

 use anyhow::{anyhow, Context, Result};
+use futures::TryFutureExt;
 use hyper::StatusCode;
 use hyper::{Body, Request, Response, Uri};
 use metrics::launch_timestamp::LaunchTimestamp;
 use pageserver_api::models::{
-    DownloadRemoteLayersTaskSpawnRequest, TenantAttachRequest, TenantLoadRequest,
+    DownloadRemoteLayersTaskSpawnRequest, LocationConfigMode, TenantAttachRequest,
+    TenantLoadRequest, TenantLocationConfigRequest,
 };
 use remote_storage::GenericRemoteStorage;
 use tenant_size_model::{SizeResult, StorageModel};
@@ -24,17 +26,18 @@ use super::models::{
    TimelineCreateRequest, TimelineGcRequest, TimelineInfo,
 };
 use crate::context::{DownloadBehavior, RequestContext};
+use crate::deletion_queue::DeletionQueueClient;
 use crate::metrics::{StorageTimeOperation, STORAGE_TIME_GLOBAL};
 use crate::pgdatadir_mapping::LsnForTimestamp;
 use crate::task_mgr::TaskKind;
-use crate::tenant::config::TenantConfOpt;
+use crate::tenant::config::{LocationConf, TenantConfOpt};
 use crate::tenant::mgr::{
    GetTenantError, SetNewTenantConfigError, TenantMapInsertError, TenantStateError,
 };
 use crate::tenant::size::ModelInputs;
 use crate::tenant::storage_layer::LayerAccessStatsReset;
 use crate::tenant::timeline::Timeline;
-use crate::tenant::{LogicalSizeCalculationCause, PageReconstructError};
+use crate::tenant::{LogicalSizeCalculationCause, PageReconstructError, TenantSharedResources};
 use crate::{config::PageServerConf, tenant::mgr};
 use crate::{disk_usage_eviction_task, tenant};
 use utils::{
@@ -61,6 +64,7 @@ pub struct State {
    remote_storage: Option<GenericRemoteStorage>,
    broker_client: storage_broker::BrokerClientChannel,
    disk_usage_eviction_state: Arc<disk_usage_eviction_task::State>,
+    deletion_queue_client: DeletionQueueClient,
 }

 impl State {
@@ -70,6 +74,7 @@ impl State {
        remote_storage: Option<GenericRemoteStorage>,
        broker_client: storage_broker::BrokerClientChannel,
        disk_usage_eviction_state: Arc<disk_usage_eviction_task::State>,
+        deletion_queue_client: DeletionQueueClient,
    ) -> anyhow::Result<Self> {
        let allowlist_routes = ["/v1/status", "/v1/doc", "/swagger.yml"]
            .iter()
@@ -82,8 +87,17 @@ impl State {
            remote_storage,
            broker_client,
            disk_usage_eviction_state,
+            deletion_queue_client,
        })
    }
+
+    fn tenant_resources(&self) -> TenantSharedResources {
+        TenantSharedResources {
+            broker_client: self.broker_client.clone(),
+            remote_storage: self.remote_storage.clone(),
+            deletion_queue_client: self.deletion_queue_client.clone(),
+        }
+    }
 }

 #[inline(always)]
@@ -119,7 +133,7 @@ impl From<PageReconstructError> for ApiError {
                ApiError::InternalServerError(anyhow::anyhow!("request was cancelled"))
            }
            PageReconstructError::AncestorStopping(_) => {
-                ApiError::InternalServerError(anyhow::Error::new(pre))
+                ApiError::ResourceUnavailable(format!("{pre}"))
            }
            PageReconstructError::WalRedo(pre) => {
                ApiError::InternalServerError(anyhow::Error::new(pre))
@@ -132,12 +146,15 @@ impl From<TenantMapInsertError> for ApiError {
    fn from(tmie: TenantMapInsertError) -> ApiError {
        match tmie {
            TenantMapInsertError::StillInitializing | TenantMapInsertError::ShuttingDown => {
-                ApiError::InternalServerError(anyhow::Error::new(tmie))
+                ApiError::ResourceUnavailable(format!("{tmie}"))
            }
            TenantMapInsertError::TenantAlreadyExists(id, state) => {
                ApiError::Conflict(format!("tenant {id} already exists, state: {state:?}"))
            }
-            TenantMapInsertError::Closure(e) => ApiError::InternalServerError(e),
+            TenantMapInsertError::TenantExistsSecondary(id) => {
+                ApiError::Conflict(format!("tenant {id} already exists as secondary"))
+            }
+            TenantMapInsertError::Other(e) => ApiError::InternalServerError(e),
        }
    }
 }
@@ -146,6 +163,12 @@ impl From<TenantStateError> for ApiError {
    fn from(tse: TenantStateError) -> ApiError {
        match tse {
            TenantStateError::NotFound(tid) => ApiError::NotFound(anyhow!("tenant {}", tid).into()),
+            TenantStateError::NotActive(_) => {
+                ApiError::ResourceUnavailable("Tenant not yet active".into())
+            }
+            TenantStateError::IsStopping(_) => {
+                ApiError::ResourceUnavailable("Tenant is stopping".into())
+            }
            _ => ApiError::InternalServerError(anyhow::Error::new(tse)),
        }
    }
@@ -155,14 +178,17 @@ impl From<GetTenantError> for ApiError {
    fn from(tse: GetTenantError) -> ApiError {
        match tse {
            GetTenantError::NotFound(tid) => ApiError::NotFound(anyhow!("tenant {}", tid).into()),
-            e @ GetTenantError::NotActive(_) => {
+            GetTenantError::Broken(reason) => {
+                ApiError::InternalServerError(anyhow!("tenant is broken: {}", reason))
+            }
+            GetTenantError::NotActive(_) => {
                // Why is this not `ApiError::NotFound`?
                // Because we must be careful to never return 404 for a tenant if it does
                // in fact exist locally. If we did, the caller could draw the conclusion
                // that it can attach the tenant to another PS and we'd be in split-brain.
                //
                // (We can produce this variant only in `mgr::get_tenant(..., active=true)` calls).
-                ApiError::InternalServerError(anyhow::Error::new(e))
+                ApiError::ResourceUnavailable("Tenant not yet active".into())
            }
        }
    }
@@ -283,7 +309,12 @@ async fn build_timeline_info_common(
    };
    let current_physical_size = Some(timeline.layer_size_sum().await);
    let state = timeline.current_state();
-    let remote_consistent_lsn = timeline.get_remote_consistent_lsn().unwrap_or(Lsn(0));
+    let remote_consistent_lsn_projected = timeline
+        .get_remote_consistent_lsn_projected()
+        .unwrap_or(Lsn(0));
+    let remote_consistent_lsn_visible = timeline
+        .get_remote_consistent_lsn_visible()
+        .unwrap_or(Lsn(0));

    let walreceiver_status = timeline.walreceiver_status();

@@ -293,7 +324,8 @@ async fn build_timeline_info_common(
        ancestor_timeline_id,
        ancestor_lsn,
        disk_consistent_lsn: timeline.get_disk_consistent_lsn(),
-        remote_consistent_lsn,
+        remote_consistent_lsn: remote_consistent_lsn_projected,
+        remote_consistent_lsn_visible,
        last_record_lsn,
        prev_record_lsn: Some(timeline.get_prev_record_lsn()),
        latest_gc_cutoff_lsn: *timeline.get_latest_gc_cutoff_lsn(),
@@ -492,24 +524,23 @@ async fn tenant_attach_handler(

    let generation = get_request_generation(state, maybe_body.as_ref().and_then(|r| r.generation))?;

-    if let Some(remote_storage) = &state.remote_storage {
-        mgr::attach_tenant(
-            state.conf,
-            tenant_id,
-            generation,
-            tenant_conf,
-            state.broker_client.clone(),
-            remote_storage.clone(),
-            &ctx,
-        )
-        .instrument(info_span!("tenant_attach", %tenant_id))
-        .await?;
-    } else {
+    if state.remote_storage.is_none() {
        return Err(ApiError::BadRequest(anyhow!(
            "attach_tenant is not possible because pageserver was configured without remote storage"
        )));
    }

+    mgr::attach_tenant(
+        state.conf,
+        tenant_id,
+        generation,
+        tenant_conf,
+        state.tenant_resources(),
+        &ctx,
+    )
+    .instrument(info_span!("tenant_attach", %tenant_id))
+    .await?;
+
    json_response(StatusCode::ACCEPTED, ())
 }

@@ -570,6 +601,7 @@ async fn tenant_load_handler(
        generation,
        state.broker_client.clone(),
        state.remote_storage.clone(),
+        state.deletion_queue_client.clone(),
        &ctx,
    )
    .instrument(info_span!("load", %tenant_id))
@@ -603,8 +635,9 @@ async fn tenant_list_handler(
    let response_data = mgr::list_tenants()
        .instrument(info_span!("tenant_list"))
        .await
-        .map_err(anyhow::Error::new)
-        .map_err(ApiError::InternalServerError)?
+        .map_err(|_| {
+            ApiError::ResourceUnavailable("Tenant map is initializing or shutting down".to_string())
+        })?
        .iter()
        .map(|(id, state)| TenantInfo {
            id: *id,
@@ -911,8 +944,7 @@ async fn tenant_create_handler(
        tenant_conf,
        target_tenant_id,
        generation,
-        state.broker_client.clone(),
-        state.remote_storage.clone(),
+        state.tenant_resources(),
        &ctx,
    )
    .instrument(info_span!("tenant_create", tenant_id = %target_tenant_id))
@@ -983,6 +1015,48 @@ async fn update_tenant_config_handler(
    json_response(StatusCode::OK, ())
 }

+async fn put_tenant_location_config_handler(
+    mut request: Request<Body>,
+    _cancel: CancellationToken,
+) -> Result<Response<Body>, ApiError> {
+    let request_data: TenantLocationConfigRequest = json_request(&mut request).await?;
+    let tenant_id = request_data.tenant_id;
+    check_permission(&request, Some(tenant_id))?;
+
+    let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Warn);
+    let state = get_state(&request);
+    let conf = state.conf;
+
+    // The `Detached` state is special, it doesn't upsert a tenant, it removes
+    // its local disk content and drops it from memory.
+    if let LocationConfigMode::Detached = request_data.config.mode {
+        mgr::detach_tenant(conf, tenant_id, true)
+            .instrument(info_span!("tenant_detach", %tenant_id))
+            .await?;
+        return json_response(StatusCode::OK, ());
+    }
+
+    let location_conf =
+        LocationConf::try_from(&request_data.config).map_err(ApiError::BadRequest)?;
+
+    mgr::upsert_location(
+        state.conf,
+        tenant_id,
+        location_conf,
+        state.broker_client.clone(),
+        state.remote_storage.clone(),
+        state.deletion_queue_client.clone(),
+        &ctx,
+    )
+    .await
+    // TODO: badrequest assumes the caller was asking for something unreasonable, but in
+    // principle we might have hit something like concurrent API calls to the same tenant,
+    // which is not a 400 but a 409.
+    .map_err(ApiError::BadRequest)?;
+
+    json_response(StatusCode::OK, ())
+}
+
 /// Testing helper to transition a tenant to [`crate::tenant::TenantState::Broken`].
 async fn handle_tenant_break(
    r: Request<Body>,
@@ -1129,6 +1203,39 @@ async fn timeline_download_remote_layers_handler_get(
    json_response(StatusCode::OK, info)
 }

+async fn deletion_queue_flush(
+    r: Request<Body>,
+    cancel: CancellationToken,
+) -> Result<Response<Body>, ApiError> {
+    let state = get_state(&r);
+
+    if state.remote_storage.is_none() {
+        // Nothing to do if remote storage is disabled.
+        return json_response(StatusCode::OK, ());
+    }
+
+    let execute = parse_query_param(&r, "execute")?.unwrap_or(false);
+
+    let flush = async {
+        if execute {
+            state.deletion_queue_client.flush_execute().await
+        } else {
+            state.deletion_queue_client.flush().await
+        }
+    }
+    // DeletionQueueError's only case is shutting down.
+    .map_err(|_| ApiError::ShuttingDown);
+
+    tokio::select! {
+        res = flush => {
+            res.map(|()| json_response(StatusCode::OK, ()))?
+        }
+        _ = cancel.cancelled() => {
+            Err(ApiError::ShuttingDown)
+        }
+    }
+}
+
 async fn active_timeline_of_active_tenant(
    tenant_id: TenantId,
    timeline_id: TimelineId,
@@ -1403,6 +1510,9 @@ pub fn make_router(
        .get("/v1/tenant/:tenant_id/config", |r| {
            api_handler(r, get_tenant_config_handler)
        })
+        .put("/v1/tenant/:tenant_id/location_config", |r| {
+            api_handler(r, put_tenant_location_config_handler)
+        })
        .get("/v1/tenant/:tenant_id/timeline", |r| {
            api_handler(r, timeline_list_handler)
        })
@@ -1463,6 +1573,9 @@ pub fn make_router(
        .put("/v1/disk_usage_eviction/run", |r| {
            api_handler(r, disk_usage_eviction_run)
        })
+        .put("/v1/deletion_queue/flush", |r| {
+            api_handler(r, deletion_queue_flush)
+        })
        .put("/v1/tenant/:tenant_id/break", |r| {
            testing_api_handler("set tenant state to broken", r, handle_tenant_break)
        })
--- a/pageserver/src/import_datadir.rs
+++ b/pageserver/src/import_datadir.rs
@@ -6,6 +6,7 @@ use std::path::{Path, PathBuf};

 use anyhow::{bail, ensure, Context, Result};
 use bytes::Bytes;
+use camino::Utf8Path;
 use futures::StreamExt;
 use tokio::io::{AsyncRead, AsyncReadExt};
 use tokio_tar::Archive;
@@ -29,7 +30,7 @@ use postgres_ffi::{BLCKSZ, WAL_SEGMENT_SIZE};
 use utils::lsn::Lsn;

 // Returns checkpoint LSN from controlfile
-pub fn get_lsn_from_controlfile(path: &Path) -> Result<Lsn> {
+pub fn get_lsn_from_controlfile(path: &Utf8Path) -> Result<Lsn> {
    // Read control file to extract the LSN
    let controlfile_path = path.join("global").join("pg_control");
    let controlfile = ControlFileData::decode(&std::fs::read(controlfile_path)?)?;
@@ -46,7 +47,7 @@ pub fn get_lsn_from_controlfile(path: &Path) -> Result<Lsn> {
 /// cluster was not shut down cleanly.
 pub async fn import_timeline_from_postgres_datadir(
    tline: &Timeline,
-    pgdata_path: &Path,
+    pgdata_path: &Utf8Path,
    pgdata_lsn: Lsn,
    ctx: &RequestContext,
 ) -> Result<()> {
@@ -75,12 +76,12 @@ pub async fn import_timeline_from_postgres_datadir(
            {
                pg_control = Some(control_file);
            }
-            modification.flush().await?;
+            modification.flush(ctx).await?;
        }
    }

    // We're done importing all the data files.
-    modification.commit().await?;
+    modification.commit(ctx).await?;

    // We expect the Postgres server to be shut down cleanly.
    let pg_control = pg_control.context("pg_control file not found")?;
@@ -256,7 +257,7 @@ async fn import_slru(
 /// Scan PostgreSQL WAL files in given directory and load all records between
 /// 'startpoint' and 'endpoint' into the repository.
 async fn import_wal(
-    walpath: &Path,
+    walpath: &Utf8Path,
    tline: &Timeline,
    startpoint: Lsn,
    endpoint: Lsn,
@@ -359,7 +360,7 @@ pub async fn import_basebackup_from_tar(
                    // We found the pg_control file.
                    pg_control = Some(res);
                }
-                modification.flush().await?;
+                modification.flush(ctx).await?;
            }
            tokio_tar::EntryType::Directory => {
                debug!("directory {:?}", file_path);
@@ -377,7 +378,7 @@ pub async fn import_basebackup_from_tar(
    // sanity check: ensure that pg_control is loaded
    let _pg_control = pg_control.context("pg_control file not found")?;

-    modification.commit().await?;
+    modification.commit(ctx).await?;
    Ok(())
 }

--- a/pageserver/src/lib.rs
+++ b/pageserver/src/lib.rs
@@ -3,7 +3,8 @@ pub mod basebackup;
 pub mod config;
 pub mod consumption_metrics;
 pub mod context;
-mod control_plane_client;
+pub mod control_plane_client;
+pub mod deletion_queue;
 pub mod disk_usage_eviction_task;
 pub mod http;
 pub mod import_datadir;
@@ -24,9 +25,9 @@ pub mod walredo;

 pub mod failpoint_support;

-use std::path::Path;
-
 use crate::task_mgr::TaskKind;
+use camino::Utf8Path;
+use deletion_queue::DeletionQueue;
 use tracing::info;

 /// Current storage format version
@@ -48,8 +49,8 @@ static ZERO_PAGE: bytes::Bytes = bytes::Bytes::from_static(&[0u8; 8192]);

 pub use crate::metrics::preinitialize_metrics;

-#[tracing::instrument]
-pub async fn shutdown_pageserver(exit_code: i32) {
+#[tracing::instrument(skip_all, fields(%exit_code))]
+pub async fn shutdown_pageserver(deletion_queue: Option<DeletionQueue>, exit_code: i32) {
    use std::time::Duration;
    // Shut down the libpq endpoint task. This prevents new connections from
    // being accepted.
@@ -77,6 +78,11 @@ pub async fn shutdown_pageserver(exit_code: i32) {
    )
    .await;

+    // Best effort to persist any outstanding deletions, to avoid leaking objects
+    if let Some(mut deletion_queue) = deletion_queue {
+        deletion_queue.shutdown(Duration::from_secs(5)).await;
+    }
+
    // Shut down the HTTP endpoint last, so that you can still check the server's
    // status while it's shutting down.
    // FIXME: We should probably stop accepting commands like attach/detach earlier.
@@ -106,6 +112,10 @@ pub const METADATA_FILE_NAME: &str = "metadata";
 /// Full path: `tenants/<tenant_id>/config`.
 pub const TENANT_CONFIG_NAME: &str = "config";

+/// Per-tenant configuration file.
+/// Full path: `tenants/<tenant_id>/config`.
+pub const TENANT_LOCATION_CONFIG_NAME: &str = "config-v1";
+
 /// A suffix used for various temporary files. Any temporary files found in the
 /// data directory at pageserver startup can be automatically removed.
 pub const TEMP_FILE_SUFFIX: &str = "___temp";
@@ -125,25 +135,25 @@ pub const TIMELINE_DELETE_MARK_SUFFIX: &str = "___delete";
 /// Full path: `tenants/<tenant_id>/___ignored_tenant`.
 pub const IGNORED_TENANT_FILE_NAME: &str = "___ignored_tenant";

-pub fn is_temporary(path: &Path) -> bool {
+pub fn is_temporary(path: &Utf8Path) -> bool {
    match path.file_name() {
-        Some(name) => name.to_string_lossy().ends_with(TEMP_FILE_SUFFIX),
+        Some(name) => name.ends_with(TEMP_FILE_SUFFIX),
        None => false,
    }
 }

-fn ends_with_suffix(path: &Path, suffix: &str) -> bool {
+fn ends_with_suffix(path: &Utf8Path, suffix: &str) -> bool {
    match path.file_name() {
-        Some(name) => name.to_string_lossy().ends_with(suffix),
+        Some(name) => name.ends_with(suffix),
        None => false,
    }
 }

-pub fn is_uninit_mark(path: &Path) -> bool {
+pub fn is_uninit_mark(path: &Utf8Path) -> bool {
    ends_with_suffix(path, TIMELINE_UNINIT_MARK_SUFFIX)
 }

-pub fn is_delete_mark(path: &Path) -> bool {
+pub fn is_delete_mark(path: &Utf8Path) -> bool {
    ends_with_suffix(path, TIMELINE_DELETE_MARK_SUFFIX)
 }

--- a/pageserver/src/metrics.rs
+++ b/pageserver/src/metrics.rs
@@ -1,3 +1,4 @@
+use enum_map::EnumMap;
 use metrics::metric_vec_duration::DurationResultObserver;
 use metrics::{
    register_counter_vec, register_gauge_vec, register_histogram, register_histogram_vec,
@@ -93,15 +94,35 @@ pub(crate) static READ_NUM_FS_LAYERS: Lazy<Histogram> = Lazy::new(|| {
 });

 // Metrics collected on operations on the storage repository.
-pub(crate) static RECONSTRUCT_TIME: Lazy<Histogram> = Lazy::new(|| {
-    register_histogram!(
+
+pub(crate) struct ReconstructTimeMetrics {
+    ok: Histogram,
+    err: Histogram,
+}
+
+pub(crate) static RECONSTRUCT_TIME: Lazy<ReconstructTimeMetrics> = Lazy::new(|| {
+    let inner = register_histogram_vec!(
        "pageserver_getpage_reconstruct_seconds",
        "Time spent in reconstruct_value (reconstruct a page from deltas)",
+        &["result"],
        CRITICAL_OP_BUCKETS.into(),
    )
-    .expect("failed to define a metric")
+    .expect("failed to define a metric");
+    ReconstructTimeMetrics {
+        ok: inner.get_metric_with_label_values(&["ok"]).unwrap(),
+        err: inner.get_metric_with_label_values(&["err"]).unwrap(),
+    }
 });

+impl ReconstructTimeMetrics {
+    pub(crate) fn for_result<T, E>(&self, result: &Result<T, E>) -> &Histogram {
+        match result {
+            Ok(_) => &self.ok,
+            Err(_) => &self.err,
+        }
+    }
+}
+
 pub(crate) static MATERIALIZED_PAGE_CACHE_HIT_DIRECT: Lazy<IntCounter> = Lazy::new(|| {
    register_int_counter!(
        "pageserver_materialized_cache_hits_direct_total",
@@ -127,22 +148,24 @@ pub(crate) static MATERIALIZED_PAGE_CACHE_HIT: Lazy<IntCounter> = Lazy::new(|| {
    .expect("failed to define a metric")
 });

-pub struct PageCacheMetrics {
+pub struct PageCacheMetricsForTaskKind {
    pub read_accesses_materialized_page: IntCounter,
-    pub read_accesses_ephemeral: IntCounter,
    pub read_accesses_immutable: IntCounter,

-    pub read_hits_ephemeral: IntCounter,
    pub read_hits_immutable: IntCounter,
    pub read_hits_materialized_page_exact: IntCounter,
    pub read_hits_materialized_page_older_lsn: IntCounter,
 }

+pub struct PageCacheMetrics {
+    map: EnumMap<TaskKind, EnumMap<PageContentKind, PageCacheMetricsForTaskKind>>,
+}
+
 static PAGE_CACHE_READ_HITS: Lazy<IntCounterVec> = Lazy::new(|| {
    register_int_counter_vec!(
        "pageserver_page_cache_read_hits_total",
        "Number of read accesses to the page cache that hit",
-        &["key_kind", "hit_kind"]
+        &["task_kind", "key_kind", "content_kind", "hit_kind"]
    )
    .expect("failed to define a metric")
 });
@@ -151,55 +174,73 @@ static PAGE_CACHE_READ_ACCESSES: Lazy<IntCounterVec> = Lazy::new(|| {
    register_int_counter_vec!(
        "pageserver_page_cache_read_accesses_total",
        "Number of read accesses to the page cache",
-        &["key_kind"]
+        &["task_kind", "key_kind", "content_kind"]
    )
    .expect("failed to define a metric")
 });

 pub static PAGE_CACHE: Lazy<PageCacheMetrics> = Lazy::new(|| PageCacheMetrics {
-    read_accesses_materialized_page: {
-        PAGE_CACHE_READ_ACCESSES
-            .get_metric_with_label_values(&["materialized_page"])
-            .unwrap()
-    },
+    map: EnumMap::from_array(std::array::from_fn(|task_kind| {
+        let task_kind = <TaskKind as enum_map::Enum>::from_usize(task_kind);
+        let task_kind: &'static str = task_kind.into();
+        EnumMap::from_array(std::array::from_fn(|content_kind| {
+            let content_kind = <PageContentKind as enum_map::Enum>::from_usize(content_kind);
+            let content_kind: &'static str = content_kind.into();
+            PageCacheMetricsForTaskKind {
+                read_accesses_materialized_page: {
+                    PAGE_CACHE_READ_ACCESSES
+                        .get_metric_with_label_values(&[
+                            task_kind,
+                            "materialized_page",
+                            content_kind,
+                        ])
+                        .unwrap()
+                },

-    read_accesses_ephemeral: {
-        PAGE_CACHE_READ_ACCESSES
-            .get_metric_with_label_values(&["ephemeral"])
-            .unwrap()
-    },
+                read_accesses_immutable: {
+                    PAGE_CACHE_READ_ACCESSES
+                        .get_metric_with_label_values(&[task_kind, "immutable", content_kind])
+                        .unwrap()
+                },

-    read_accesses_immutable: {
-        PAGE_CACHE_READ_ACCESSES
-            .get_metric_with_label_values(&["immutable"])
-            .unwrap()
-    },
+                read_hits_immutable: {
+                    PAGE_CACHE_READ_HITS
+                        .get_metric_with_label_values(&[task_kind, "immutable", content_kind, "-"])
+                        .unwrap()
+                },

-    read_hits_ephemeral: {
-        PAGE_CACHE_READ_HITS
-            .get_metric_with_label_values(&["ephemeral", "-"])
-            .unwrap()
-    },
+                read_hits_materialized_page_exact: {
+                    PAGE_CACHE_READ_HITS
+                        .get_metric_with_label_values(&[
+                            task_kind,
+                            "materialized_page",
+                            content_kind,
+                            "exact",
+                        ])
+                        .unwrap()
+                },

-    read_hits_immutable: {
-        PAGE_CACHE_READ_HITS
-            .get_metric_with_label_values(&["immutable", "-"])
-            .unwrap()
-    },
-
-    read_hits_materialized_page_exact: {
-        PAGE_CACHE_READ_HITS
-            .get_metric_with_label_values(&["materialized_page", "exact"])
-            .unwrap()
-    },
-
-    read_hits_materialized_page_older_lsn: {
-        PAGE_CACHE_READ_HITS
-            .get_metric_with_label_values(&["materialized_page", "older_lsn"])
-            .unwrap()
-    },
+                read_hits_materialized_page_older_lsn: {
+                    PAGE_CACHE_READ_HITS
+                        .get_metric_with_label_values(&[
+                            task_kind,
+                            "materialized_page",
+                            content_kind,
+                            "older_lsn",
+                        ])
+                        .unwrap()
+                },
+            }
+        }))
+    })),
 });

+impl PageCacheMetrics {
+    pub(crate) fn for_ctx(&self, ctx: &RequestContext) -> &PageCacheMetricsForTaskKind {
+        &self.map[ctx.task_kind()][ctx.page_content_kind()]
+    }
+}
+
 pub struct PageCacheSizeMetrics {
    pub max_bytes: UIntGauge,

@@ -243,6 +284,46 @@ pub static PAGE_CACHE_SIZE: Lazy<PageCacheSizeMetrics> = Lazy::new(|| PageCacheS
    },
 });

+pub(crate) static PAGE_CACHE_ACQUIRE_PINNED_SLOT_TIME: Lazy<Histogram> = Lazy::new(|| {
+    register_histogram!(
+        "pageserver_page_cache_acquire_pinned_slot_seconds",
+        "Time spent acquiring a pinned slot in the page cache",
+        CRITICAL_OP_BUCKETS.into(),
+    )
+    .expect("failed to define a metric")
+});
+
+pub(crate) static PAGE_CACHE_FIND_VICTIMS_ITERS_TOTAL: Lazy<IntCounter> = Lazy::new(|| {
+    register_int_counter!(
+        "pageserver_page_cache_find_victim_iters_total",
+        "Counter for the number of iterations in the find_victim loop",
+    )
+    .expect("failed to define a metric")
+});
+
+static PAGE_CACHE_ERRORS: Lazy<IntCounterVec> = Lazy::new(|| {
+    register_int_counter_vec!(
+        "page_cache_errors_total",
+        "Number of timeouts while acquiring a pinned slot in the page cache",
+        &["error_kind"]
+    )
+    .expect("failed to define a metric")
+});
+
+#[derive(IntoStaticStr)]
+#[strum(serialize_all = "kebab_case")]
+pub(crate) enum PageCacheErrorKind {
+    AcquirePinnedSlotTimeout,
+    EvictIterLimit,
+}
+
+pub(crate) fn page_cache_errors_inc(error_kind: PageCacheErrorKind) {
+    PAGE_CACHE_ERRORS
+        .get_metric_with_label_values(&[error_kind.into()])
+        .unwrap()
+        .inc();
+}
+
 pub(crate) static WAIT_LSN_TIME: Lazy<Histogram> = Lazy::new(|| {
    register_histogram!(
        "pageserver_wait_lsn_seconds",
@@ -270,6 +351,14 @@ static RESIDENT_PHYSICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
    .expect("failed to define a metric")
 });

+pub(crate) static RESIDENT_PHYSICAL_SIZE_GLOBAL: Lazy<UIntGauge> = Lazy::new(|| {
+    register_uint_gauge!(
+        "pageserver_resident_physical_size_global",
+        "Like `pageserver_resident_physical_size`, but without tenant/timeline dimensions."
+    )
+    .expect("failed to define a metric")
+});
+
 static REMOTE_PHYSICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
    register_uint_gauge_vec!(
        "pageserver_remote_physical_size",
@@ -280,6 +369,14 @@ static REMOTE_PHYSICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
    .expect("failed to define a metric")
 });

+static REMOTE_PHYSICAL_SIZE_GLOBAL: Lazy<UIntGauge> = Lazy::new(|| {
+    register_uint_gauge!(
+        "pageserver_remote_physical_size_global",
+        "Like `pageserver_remote_physical_size`, but without tenant/timeline dimensions."
+    )
+    .expect("failed to define a metric")
+});
+
 pub(crate) static REMOTE_ONDEMAND_DOWNLOADED_LAYERS: Lazy<IntCounter> = Lazy::new(|| {
    register_int_counter!(
        "pageserver_remote_ondemand_downloaded_layers_total",
@@ -866,6 +963,61 @@ static REMOTE_TIMELINE_CLIENT_BYTES_FINISHED_COUNTER: Lazy<IntCounterVec> = Lazy
    .expect("failed to define a metric")
 });

+pub(crate) struct DeletionQueueMetrics {
+    pub(crate) keys_submitted: IntCounter,
+    pub(crate) keys_dropped: IntCounter,
+    pub(crate) keys_executed: IntCounter,
+    pub(crate) keys_validated: IntCounter,
+    pub(crate) dropped_lsn_updates: IntCounter,
+    pub(crate) unexpected_errors: IntCounter,
+    pub(crate) remote_errors: IntCounterVec,
+}
+pub(crate) static DELETION_QUEUE: Lazy<DeletionQueueMetrics> = Lazy::new(|| {
+    DeletionQueueMetrics{
+
+    keys_submitted: register_int_counter!(
+        "pageserver_deletion_queue_submitted_total",
+        "Number of objects submitted for deletion"
+    )
+    .expect("failed to define a metric"),
+
+    keys_dropped: register_int_counter!(
+        "pageserver_deletion_queue_dropped_total",
+        "Number of object deletions dropped due to stale generation."
+    )
+    .expect("failed to define a metric"),
+
+    keys_executed: register_int_counter!(
+        "pageserver_deletion_queue_executed_total",
+        "Number of objects deleted. Only includes objects that we actually deleted, sum with pageserver_deletion_queue_dropped_total for the total number of keys processed to completion"
+    )
+    .expect("failed to define a metric"),
+
+    keys_validated: register_int_counter!(
+        "pageserver_deletion_queue_validated_total",
+        "Number of keys validated for deletion.  Sum with pageserver_deletion_queue_dropped_total for the total number of keys that have passed through the validation stage."
+    )
+    .expect("failed to define a metric"),
+
+    dropped_lsn_updates: register_int_counter!(
+        "pageserver_deletion_queue_dropped_lsn_updates_total",
+        "Updates to remote_consistent_lsn dropped due to stale generation number."
+    )
+    .expect("failed to define a metric"),
+    unexpected_errors: register_int_counter!(
+        "pageserver_deletion_queue_unexpected_errors_total",
+        "Number of unexpected condiions that may stall the queue: any value above zero is unexpected."
+    )
+    .expect("failed to define a metric"),
+    remote_errors: register_int_counter_vec!(
+        "pageserver_deletion_queue_remote_errors_total",
+        "Retryable remote I/O errors while executing deletions, for example 503 responses to DeleteObjects",
+        &["op_kind"],
+    )
+    .expect("failed to define a metric")
+}
+});
+
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
 pub enum RemoteOpKind {
    Upload,
@@ -1140,7 +1292,7 @@ pub struct TimelineMetrics {
    pub load_layer_map_histo: StorageTimeMetrics,
    pub garbage_collect_histo: StorageTimeMetrics,
    pub last_record_gauge: IntGauge,
-    pub resident_physical_size_gauge: UIntGauge,
+    resident_physical_size_gauge: UIntGauge,
    /// copy of LayeredTimeline.current_logical_size
    pub current_logical_size_gauge: UIntGauge,
    pub num_persistent_files_created: IntCounter,
@@ -1218,10 +1370,29 @@ impl TimelineMetrics {
    }

    pub fn record_new_file_metrics(&self, sz: u64) {
-        self.resident_physical_size_gauge.add(sz);
+        self.resident_physical_size_add(sz);
        self.num_persistent_files_created.inc_by(1);
        self.persistent_bytes_written.inc_by(sz);
    }
+
+    pub fn resident_physical_size_sub(&self, sz: u64) {
+        self.resident_physical_size_gauge.sub(sz);
+        crate::metrics::RESIDENT_PHYSICAL_SIZE_GLOBAL.sub(sz);
+    }
+
+    pub fn resident_physical_size_add(&self, sz: u64) {
+        self.resident_physical_size_gauge.add(sz);
+        crate::metrics::RESIDENT_PHYSICAL_SIZE_GLOBAL.add(sz);
+    }
+
+    pub fn resident_physical_size_set(&self, sz: u64) {
+        self.resident_physical_size_gauge.set(sz);
+        crate::metrics::RESIDENT_PHYSICAL_SIZE_GLOBAL.set(sz);
+    }
+
+    pub fn resident_physical_size_get(&self) -> u64 {
+        self.resident_physical_size_gauge.get()
+    }
 }

 impl Drop for TimelineMetrics {
@@ -1229,7 +1400,10 @@ impl Drop for TimelineMetrics {
        let tenant_id = &self.tenant_id;
        let timeline_id = &self.timeline_id;
        let _ = LAST_RECORD_LSN.remove_label_values(&[tenant_id, timeline_id]);
-        let _ = RESIDENT_PHYSICAL_SIZE.remove_label_values(&[tenant_id, timeline_id]);
+        {
+            RESIDENT_PHYSICAL_SIZE_GLOBAL.sub(self.resident_physical_size_get());
+            let _ = RESIDENT_PHYSICAL_SIZE.remove_label_values(&[tenant_id, timeline_id]);
+        }
        let _ = CURRENT_LOGICAL_SIZE.remove_label_values(&[tenant_id, timeline_id]);
        let _ = NUM_PERSISTENT_FILES_CREATED.remove_label_values(&[tenant_id, timeline_id]);
        let _ = PERSISTENT_BYTES_WRITTEN.remove_label_values(&[tenant_id, timeline_id]);
@@ -1280,10 +1454,46 @@ use std::sync::{Arc, Mutex};
 use std::task::{Context, Poll};
 use std::time::{Duration, Instant};

+use crate::context::{PageContentKind, RequestContext};
+use crate::task_mgr::TaskKind;
+
+/// Maintain a per timeline gauge in addition to the global gauge.
+struct PerTimelineRemotePhysicalSizeGauge {
+    last_set: u64,
+    gauge: UIntGauge,
+}
+
+impl PerTimelineRemotePhysicalSizeGauge {
+    fn new(per_timeline_gauge: UIntGauge) -> Self {
+        Self {
+            last_set: per_timeline_gauge.get(),
+            gauge: per_timeline_gauge,
+        }
+    }
+    fn set(&mut self, sz: u64) {
+        self.gauge.set(sz);
+        if sz < self.last_set {
+            REMOTE_PHYSICAL_SIZE_GLOBAL.sub(self.last_set - sz);
+        } else {
+            REMOTE_PHYSICAL_SIZE_GLOBAL.add(sz - self.last_set);
+        };
+        self.last_set = sz;
+    }
+    fn get(&self) -> u64 {
+        self.gauge.get()
+    }
+}
+
+impl Drop for PerTimelineRemotePhysicalSizeGauge {
+    fn drop(&mut self) {
+        REMOTE_PHYSICAL_SIZE_GLOBAL.sub(self.last_set);
+    }
+}
+
 pub struct RemoteTimelineClientMetrics {
    tenant_id: String,
    timeline_id: String,
-    remote_physical_size_gauge: Mutex<Option<UIntGauge>>,
+    remote_physical_size_gauge: Mutex<Option<PerTimelineRemotePhysicalSizeGauge>>,
    calls_unfinished_gauge: Mutex<HashMap<(&'static str, &'static str), IntGauge>>,
    bytes_started_counter: Mutex<HashMap<(&'static str, &'static str), IntCounter>>,
    bytes_finished_counter: Mutex<HashMap<(&'static str, &'static str), IntCounter>>,
@@ -1301,18 +1511,24 @@ impl RemoteTimelineClientMetrics {
        }
    }

-    pub fn remote_physical_size_gauge(&self) -> UIntGauge {
+    pub(crate) fn remote_physical_size_set(&self, sz: u64) {
        let mut guard = self.remote_physical_size_gauge.lock().unwrap();
-        guard
-            .get_or_insert_with(|| {
+        let gauge = guard.get_or_insert_with(|| {
+            PerTimelineRemotePhysicalSizeGauge::new(
                REMOTE_PHYSICAL_SIZE
                    .get_metric_with_label_values(&[
                        &self.tenant_id.to_string(),
                        &self.timeline_id.to_string(),
                    ])
-                    .unwrap()
-            })
-            .clone()
+                    .unwrap(),
+            )
+        });
+        gauge.set(sz);
+    }
+
+    pub(crate) fn remote_physical_size_get(&self) -> u64 {
+        let guard = self.remote_physical_size_gauge.lock().unwrap();
+        guard.as_ref().map(|gauge| gauge.get()).unwrap_or(0)
    }

    pub fn remote_operation_time(
@@ -1651,6 +1867,9 @@ pub fn preinitialize_metrics() {
        Lazy::force(c);
    });

+    // Deletion queue stats
+    Lazy::force(&DELETION_QUEUE);
+
    // countervecs
    [&BACKGROUND_LOOP_PERIOD_OVERRUN_COUNT]
        .into_iter()
@@ -1664,7 +1883,6 @@ pub fn preinitialize_metrics() {
    // histograms
    [
        &READ_NUM_FS_LAYERS,
-        &RECONSTRUCT_TIME,
        &WAIT_LSN_TIME,
        &WAL_REDO_TIME,
        &WAL_REDO_WAIT_TIME,
@@ -1675,4 +1893,7 @@ pub fn preinitialize_metrics() {
    .for_each(|h| {
        Lazy::force(h);
    });
+
+    // Custom
+    Lazy::force(&RECONSTRUCT_TIME);
 }
--- a/pageserver/src/page_cache.rs
+++ b/pageserver/src/page_cache.rs
@@ -75,7 +75,11 @@
 use std::{
    collections::{hash_map::Entry, HashMap},
    convert::TryInto,
-    sync::atomic::{AtomicU64, AtomicU8, AtomicUsize, Ordering},
+    sync::{
+        atomic::{AtomicU64, AtomicU8, AtomicUsize, Ordering},
+        Arc, Weak,
+    },
+    time::Duration,
 };

 use anyhow::Context;
@@ -85,7 +89,7 @@ use utils::{
    lsn::Lsn,
 };

-use crate::{metrics::PageCacheSizeMetrics, repository::Key};
+use crate::{context::RequestContext, metrics::PageCacheSizeMetrics, repository::Key};

 static PAGE_CACHE: OnceCell<PageCache> = OnceCell::new();
 const TEST_PAGE_CACHE_SIZE: usize = 50;
@@ -165,6 +169,8 @@ struct Slot {

 struct SlotInner {
    key: Option<CacheKey>,
+    // for `coalesce_readers_permit`
+    permit: std::sync::Mutex<Weak<PinnedSlotsPermit>>,
    buf: &'static mut [u8; PAGE_SZ],
 }

@@ -207,6 +213,22 @@ impl Slot {
    }
 }

+impl SlotInner {
+    /// If there is aready a reader, drop our permit and share its permit, just like we share read access.
+    fn coalesce_readers_permit(&self, permit: PinnedSlotsPermit) -> Arc<PinnedSlotsPermit> {
+        let mut guard = self.permit.lock().unwrap();
+        if let Some(existing_permit) = guard.upgrade() {
+            drop(guard);
+            drop(permit);
+            existing_permit
+        } else {
+            let permit = Arc::new(permit);
+            *guard = Arc::downgrade(&permit);
+            permit
+        }
+    }
+}
+
 pub struct PageCache {
    /// This contains the mapping from the cache key to buffer slot that currently
    /// contains the page, if any.
@@ -224,6 +246,8 @@ pub struct PageCache {
    /// The actual buffers with their metadata.
    slots: Box<[Slot]>,

+    pinned_slots: Arc<tokio::sync::Semaphore>,
+
    /// Index of the next candidate to evict, for the Clock replacement algorithm.
    /// This is interpreted modulo the page cache size.
    next_evict_slot: AtomicUsize,
@@ -231,23 +255,28 @@ pub struct PageCache {
    size_metrics: &'static PageCacheSizeMetrics,
 }

+struct PinnedSlotsPermit(tokio::sync::OwnedSemaphorePermit);
+
 ///
 /// PageReadGuard is a "lease" on a buffer, for reading. The page is kept locked
 /// until the guard is dropped.
 ///
-pub struct PageReadGuard<'i>(tokio::sync::RwLockReadGuard<'i, SlotInner>);
+pub struct PageReadGuard<'i> {
+    _permit: Arc<PinnedSlotsPermit>,
+    slot_guard: tokio::sync::RwLockReadGuard<'i, SlotInner>,
+}

 impl std::ops::Deref for PageReadGuard<'_> {
    type Target = [u8; PAGE_SZ];

    fn deref(&self) -> &Self::Target {
-        self.0.buf
+        self.slot_guard.buf
    }
 }

 impl AsRef<[u8; PAGE_SZ]> for PageReadGuard<'_> {
    fn as_ref(&self) -> &[u8; PAGE_SZ] {
-        self.0.buf
+        self.slot_guard.buf
    }
 }

@@ -264,6 +293,8 @@ impl AsRef<[u8; PAGE_SZ]> for PageReadGuard<'_> {
 pub struct PageWriteGuard<'i> {
    inner: tokio::sync::RwLockWriteGuard<'i, SlotInner>,

+    _permit: PinnedSlotsPermit,
+
    // Are the page contents currently valid?
    // Used to mark pages as invalid that are assigned but not yet filled with data.
    valid: bool,
@@ -346,8 +377,14 @@ impl PageCache {
        timeline_id: TimelineId,
        key: &Key,
        lsn: Lsn,
+        ctx: &RequestContext,
    ) -> Option<(Lsn, PageReadGuard)> {
+        let Ok(permit) = self.try_get_pinned_slot_permit().await else {
+            return None;
+        };
+
        crate::metrics::PAGE_CACHE
+            .for_ctx(ctx)
            .read_accesses_materialized_page
            .inc();

@@ -360,7 +397,10 @@ impl PageCache {
            lsn,
        };

-        if let Some(guard) = self.try_lock_for_read(&mut cache_key).await {
+        if let Some(guard) = self
+            .try_lock_for_read(&mut cache_key, &mut Some(permit))
+            .await
+        {
            if let CacheKey::MaterializedPage {
                hash_key: _,
                lsn: available_lsn,
@@ -368,10 +408,12 @@ impl PageCache {
            {
                if available_lsn == lsn {
                    crate::metrics::PAGE_CACHE
+                        .for_ctx(ctx)
                        .read_hits_materialized_page_exact
                        .inc();
                } else {
                    crate::metrics::PAGE_CACHE
+                        .for_ctx(ctx)
                        .read_hits_materialized_page_older_lsn
                        .inc();
                }
@@ -426,10 +468,11 @@ impl PageCache {
        &self,
        file_id: FileId,
        blkno: u32,
+        ctx: &RequestContext,
    ) -> anyhow::Result<ReadBufResult> {
        let mut cache_key = CacheKey::ImmutableFilePage { file_id, blkno };

-        self.lock_for_read(&mut cache_key).await
+        self.lock_for_read(&mut cache_key, ctx).await
    }

    //
@@ -440,6 +483,29 @@ impl PageCache {
    // "mappings" after this section. But the routines in this section should
    // not require changes.

+    async fn try_get_pinned_slot_permit(&self) -> anyhow::Result<PinnedSlotsPermit> {
+        let timer = crate::metrics::PAGE_CACHE_ACQUIRE_PINNED_SLOT_TIME.start_timer();
+        match tokio::time::timeout(
+            // Choose small timeout, neon_smgr does its own retries.
+            // https://neondb.slack.com/archives/C04DGM6SMTM/p1694786876476869
+            Duration::from_secs(10),
+            Arc::clone(&self.pinned_slots).acquire_owned(),
+        )
+        .await
+        {
+            Ok(res) => Ok(PinnedSlotsPermit(
+                res.expect("this semaphore is never closed"),
+            )),
+            Err(_timeout) => {
+                timer.stop_and_discard();
+                crate::metrics::page_cache_errors_inc(
+                    crate::metrics::PageCacheErrorKind::AcquirePinnedSlotTimeout,
+                );
+                anyhow::bail!("timeout: there were page guards alive for all page cache slots")
+            }
+        }
+    }
+
    /// Look up a page in the cache.
    ///
    /// If the search criteria is not exact, *cache_key is updated with the key
@@ -449,7 +515,11 @@ impl PageCache {
    ///
    /// If no page is found, returns None and *cache_key is left unmodified.
    ///
-    async fn try_lock_for_read(&self, cache_key: &mut CacheKey) -> Option<PageReadGuard> {
+    async fn try_lock_for_read(
+        &self,
+        cache_key: &mut CacheKey,
+        permit: &mut Option<PinnedSlotsPermit>,
+    ) -> Option<PageReadGuard> {
        let cache_key_orig = cache_key.clone();
        if let Some(slot_idx) = self.search_mapping(cache_key) {
            // The page was found in the mapping. Lock the slot, and re-check
@@ -459,7 +529,10 @@ impl PageCache {
            let inner = slot.inner.read().await;
            if inner.key.as_ref() == Some(cache_key) {
                slot.inc_usage_count();
-                return Some(PageReadGuard(inner));
+                return Some(PageReadGuard {
+                    _permit: inner.coalesce_readers_permit(permit.take().unwrap()),
+                    slot_guard: inner,
+                });
            } else {
                // search_mapping might have modified the search key; restore it.
                *cache_key = cache_key_orig;
@@ -497,14 +570,22 @@ impl PageCache {
    /// }
    /// ```
    ///
-    async fn lock_for_read(&self, cache_key: &mut CacheKey) -> anyhow::Result<ReadBufResult> {
+    async fn lock_for_read(
+        &self,
+        cache_key: &mut CacheKey,
+        ctx: &RequestContext,
+    ) -> anyhow::Result<ReadBufResult> {
+        let mut permit = Some(self.try_get_pinned_slot_permit().await?);
+
        let (read_access, hit) = match cache_key {
            CacheKey::MaterializedPage { .. } => {
                unreachable!("Materialized pages use lookup_materialized_page")
            }
            CacheKey::ImmutableFilePage { .. } => (
-                &crate::metrics::PAGE_CACHE.read_accesses_immutable,
-                &crate::metrics::PAGE_CACHE.read_hits_immutable,
+                &crate::metrics::PAGE_CACHE
+                    .for_ctx(ctx)
+                    .read_accesses_immutable,
+                &crate::metrics::PAGE_CACHE.for_ctx(ctx).read_hits_immutable,
            ),
        };
        read_access.inc();
@@ -512,17 +593,21 @@ impl PageCache {
        let mut is_first_iteration = true;
        loop {
            // First check if the key already exists in the cache.
-            if let Some(read_guard) = self.try_lock_for_read(cache_key).await {
+            if let Some(read_guard) = self.try_lock_for_read(cache_key, &mut permit).await {
+                debug_assert!(permit.is_none());
                if is_first_iteration {
                    hit.inc();
                }
                return Ok(ReadBufResult::Found(read_guard));
            }
+            debug_assert!(permit.is_some());
            is_first_iteration = false;

            // Not found. Find a victim buffer
-            let (slot_idx, mut inner) =
-                self.find_victim().context("Failed to find evict victim")?;
+            let (slot_idx, mut inner) = self
+                .find_victim(permit.as_ref().unwrap())
+                .await
+                .context("Failed to find evict victim")?;

            // Insert mapping for this. At this point, we may find that another
            // thread did the same thing concurrently. In that case, we evicted
@@ -544,7 +629,16 @@ impl PageCache {
            inner.key = Some(cache_key.clone());
            slot.set_usage_count(1);

+            debug_assert!(
+                {
+                    let guard = inner.permit.lock().unwrap();
+                    guard.upgrade().is_none()
+                },
+                "we hold a write lock, so, no one else should have a permit"
+            );
+
            return Ok(ReadBufResult::NotFound(PageWriteGuard {
+                _permit: permit.take().unwrap(),
                inner,
                valid: false,
            }));
@@ -555,7 +649,11 @@ impl PageCache {
    /// found, returns None.
    ///
    /// When locking a page for writing, the search criteria is always "exact".
-    async fn try_lock_for_write(&self, cache_key: &CacheKey) -> Option<PageWriteGuard> {
+    async fn try_lock_for_write(
+        &self,
+        cache_key: &CacheKey,
+        permit: &mut Option<PinnedSlotsPermit>,
+    ) -> Option<PageWriteGuard> {
        if let Some(slot_idx) = self.search_mapping_for_write(cache_key) {
            // The page was found in the mapping. Lock the slot, and re-check
            // that it's still what we expected (because we don't released the mapping
@@ -564,7 +662,18 @@ impl PageCache {
            let inner = slot.inner.write().await;
            if inner.key.as_ref() == Some(cache_key) {
                slot.inc_usage_count();
-                return Some(PageWriteGuard { inner, valid: true });
+                debug_assert!(
+                    {
+                        let guard = inner.permit.lock().unwrap();
+                        guard.upgrade().is_none()
+                    },
+                    "we hold a write lock, so, no one else should have a permit"
+                );
+                return Some(PageWriteGuard {
+                    _permit: permit.take().unwrap(),
+                    inner,
+                    valid: true,
+                });
            }
        }
        None
@@ -575,15 +684,20 @@ impl PageCache {
    /// Similar to lock_for_read(), but the returned buffer is write-locked and
    /// may be modified by the caller even if it's already found in the cache.
    async fn lock_for_write(&self, cache_key: &CacheKey) -> anyhow::Result<WriteBufResult> {
+        let mut permit = Some(self.try_get_pinned_slot_permit().await?);
        loop {
            // First check if the key already exists in the cache.
-            if let Some(write_guard) = self.try_lock_for_write(cache_key).await {
+            if let Some(write_guard) = self.try_lock_for_write(cache_key, &mut permit).await {
+                debug_assert!(permit.is_none());
                return Ok(WriteBufResult::Found(write_guard));
            }
+            debug_assert!(permit.is_some());

            // Not found. Find a victim buffer
-            let (slot_idx, mut inner) =
-                self.find_victim().context("Failed to find evict victim")?;
+            let (slot_idx, mut inner) = self
+                .find_victim(permit.as_ref().unwrap())
+                .await
+                .context("Failed to find evict victim")?;

            // Insert mapping for this. At this point, we may find that another
            // thread did the same thing concurrently. In that case, we evicted
@@ -605,7 +719,16 @@ impl PageCache {
            inner.key = Some(cache_key.clone());
            slot.set_usage_count(1);

+            debug_assert!(
+                {
+                    let guard = inner.permit.lock().unwrap();
+                    guard.upgrade().is_none()
+                },
+                "we hold a write lock, so, no one else should have a permit"
+            );
+
            return Ok(WriteBufResult::NotFound(PageWriteGuard {
+                _permit: permit.take().unwrap(),
                inner,
                valid: false,
            }));
@@ -758,7 +881,10 @@ impl PageCache {
    /// Find a slot to evict.
    ///
    /// On return, the slot is empty and write-locked.
-    fn find_victim(&self) -> anyhow::Result<(usize, tokio::sync::RwLockWriteGuard<SlotInner>)> {
+    async fn find_victim(
+        &self,
+        _permit_witness: &PinnedSlotsPermit,
+    ) -> anyhow::Result<(usize, tokio::sync::RwLockWriteGuard<SlotInner>)> {
        let iter_limit = self.slots.len() * 10;
        let mut iters = 0;
        loop {
@@ -771,13 +897,40 @@ impl PageCache {
                let mut inner = match slot.inner.try_write() {
                    Ok(inner) => inner,
                    Err(_err) => {
-                        // If we have looped through the whole buffer pool 10 times
-                        // and still haven't found a victim buffer, something's wrong.
-                        // Maybe all the buffers were in locked. That could happen in
-                        // theory, if you have more threads holding buffers locked than
-                        // there are buffers in the pool. In practice, with a reasonably
-                        // large buffer pool it really shouldn't happen.
                        if iters > iter_limit {
+                            // NB: Even with the permits, there's no hard guarantee that we will find a slot with
+                            // any particular number of iterations: other threads might race ahead and acquire and
+                            // release pins just as we're scanning the array.
+                            //
+                            // Imagine that nslots is 2, and as starting point, usage_count==1 on all
+                            // slots. There are two threads running concurrently, A and B. A has just
+                            // acquired the permit from the semaphore.
+                            //
+                            //   A: Look at slot 1. Its usage_count == 1, so decrement it to zero, and continue the search
+                            //   B: Acquire permit.
+                            //   B: Look at slot 2, decrement its usage_count to zero and continue the search
+                            //   B: Look at slot 1. Its usage_count is zero, so pin it and bump up its usage_count to 1.
+                            //   B: Release pin and permit again
+                            //   B: Acquire permit.
+                            //   B: Look at slot 2. Its usage_count is zero, so pin it and bump up its usage_count to 1.
+                            //   B: Release pin and permit again
+                            //
+                            // Now we're back in the starting situation that both slots have
+                            // usage_count 1, but A has now been through one iteration of the
+                            // find_victim() loop. This can repeat indefinitely and on each
+                            // iteration, A's iteration count increases by one.
+                            //
+                            // So, even though the semaphore for the permits is fair, the victim search
+                            // itself happens in parallel and is not fair.
+                            // Hence even with a permit, a task can theoretically be starved.
+                            // To avoid this, we'd need tokio to give priority to tasks that are holding
+                            // permits for longer.
+                            // Note that just yielding to tokio during iteration without such
+                            // priority boosting is likely counter-productive. We'd just give more opportunities
+                            // for B to bump usage count, further starving A.
+                            crate::metrics::page_cache_errors_inc(
+                                crate::metrics::PageCacheErrorKind::EvictIterLimit,
+                            );
                            anyhow::bail!("exceeded evict iter limit");
                        }
                        continue;
@@ -788,6 +941,7 @@ impl PageCache {
                    self.remove_mapping(old_key);
                    inner.key = None;
                }
+                crate::metrics::PAGE_CACHE_FIND_VICTIMS_ITERS_TOTAL.inc_by(iters as u64);
                return Ok((slot_idx, inner));
            }
        }
@@ -815,7 +969,11 @@ impl PageCache {
                let buf: &mut [u8; PAGE_SZ] = chunk.try_into().unwrap();

                Slot {
-                    inner: tokio::sync::RwLock::new(SlotInner { key: None, buf }),
+                    inner: tokio::sync::RwLock::new(SlotInner {
+                        key: None,
+                        buf,
+                        permit: std::sync::Mutex::new(Weak::new()),
+                    }),
                    usage_count: AtomicU8::new(0),
                }
            })
@@ -827,6 +985,7 @@ impl PageCache {
            slots,
            next_evict_slot: AtomicUsize::new(0),
            size_metrics,
+            pinned_slots: Arc::new(tokio::sync::Semaphore::new(num_pages)),
        }
    }
 }
--- a/pageserver/src/page_service.rs
+++ b/pageserver/src/page_service.rs
@@ -412,31 +412,53 @@ impl PageServerHandler {
            // TODO: We could create a new per-request context here, with unique ID.
            // Currently we use the same per-timeline context for all requests

-            let response = match neon_fe_msg {
+            let (response, span) = match neon_fe_msg {
                PagestreamFeMessage::Exists(req) => {
                    let _timer = metrics.start_timer(metrics::SmgrQueryType::GetRelExists);
-                    self.handle_get_rel_exists_request(&timeline, &req, &ctx)
-                        .await
+                    let span = tracing::info_span!("handle_get_rel_exists_request", rel = %req.rel, req_lsn = %req.lsn);
+                    (
+                        self.handle_get_rel_exists_request(&timeline, &req, &ctx)
+                            .instrument(span.clone())
+                            .await,
+                        span,
+                    )
                }
                PagestreamFeMessage::Nblocks(req) => {
                    let _timer = metrics.start_timer(metrics::SmgrQueryType::GetRelSize);
-                    self.handle_get_nblocks_request(&timeline, &req, &ctx).await
+                    let span = tracing::info_span!("handle_get_nblocks_request", rel = %req.rel, req_lsn = %req.lsn);
+                    (
+                        self.handle_get_nblocks_request(&timeline, &req, &ctx)
+                            .instrument(span.clone())
+                            .await,
+                        span,
+                    )
                }
                PagestreamFeMessage::GetPage(req) => {
                    let _timer = metrics.start_timer(metrics::SmgrQueryType::GetPageAtLsn);
-                    self.handle_get_page_at_lsn_request(&timeline, &req, &ctx)
-                        .await
+                    let span = tracing::info_span!("handle_get_page_at_lsn_request", rel = %req.rel, blkno = %req.blkno, req_lsn = %req.lsn);
+                    (
+                        self.handle_get_page_at_lsn_request(&timeline, &req, &ctx)
+                            .instrument(span.clone())
+                            .await,
+                        span,
+                    )
                }
                PagestreamFeMessage::DbSize(req) => {
                    let _timer = metrics.start_timer(metrics::SmgrQueryType::GetDbSize);
-                    self.handle_db_size_request(&timeline, &req, &ctx).await
+                    let span = tracing::info_span!("handle_db_size_request", dbnode = %req.dbnode, req_lsn = %req.lsn);
+                    (
+                        self.handle_db_size_request(&timeline, &req, &ctx)
+                            .instrument(span.clone())
+                            .await,
+                        span,
+                    )
                }
            };

            let response = response.unwrap_or_else(|e| {
                // print the all details to the log with {:#}, but for the client the
                // error message is enough
-                error!("error reading relation or page version: {:?}", e);
+                span.in_scope(|| error!("error reading relation or page version: {:#}", e));
                PagestreamBeMessage::Error(PagestreamErrorResponse {
                    message: e.to_string(),
                })
@@ -627,7 +649,6 @@ impl PageServerHandler {
        Ok(lsn)
    }

-    #[instrument(skip(self, timeline, req, ctx), fields(rel = %req.rel, req_lsn = %req.lsn))]
    async fn handle_get_rel_exists_request(
        &self,
        timeline: &Timeline,
@@ -648,7 +669,6 @@ impl PageServerHandler {
        }))
    }

-    #[instrument(skip(self, timeline, req, ctx), fields(rel = %req.rel, req_lsn = %req.lsn))]
    async fn handle_get_nblocks_request(
        &self,
        timeline: &Timeline,
@@ -667,7 +687,6 @@ impl PageServerHandler {
        }))
    }

-    #[instrument(skip(self, timeline, req, ctx), fields(dbnode = %req.dbnode, req_lsn = %req.lsn))]
    async fn handle_db_size_request(
        &self,
        timeline: &Timeline,
@@ -689,7 +708,6 @@ impl PageServerHandler {
        }))
    }

-    #[instrument(skip(self, timeline, req, ctx), fields(rel = %req.rel, blkno = %req.blkno, req_lsn = %req.lsn))]
    async fn handle_get_page_at_lsn_request(
        &self,
        timeline: &Timeline,
@@ -1265,7 +1283,10 @@ async fn get_active_tenant_with_timeout(
        Ok(tenant) => tenant,
        Err(e @ GetTenantError::NotFound(_)) => return Err(GetActiveTenantError::NotFound(e)),
        Err(GetTenantError::NotActive(_)) => {
-            unreachable!("we're calling get_tenant with active=false")
+            unreachable!("we're calling get_tenant with active_only=false")
+        }
+        Err(GetTenantError::Broken(_)) => {
+            unreachable!("we're calling get_tenant with active_only=false")
        }
    };
    let wait_time = Duration::from_secs(30);
--- a/pageserver/src/pgdatadir_mapping.rs
+++ b/pageserver/src/pgdatadir_mapping.rs
@@ -1138,7 +1138,7 @@ impl<'a> DatadirModification<'a> {
    /// retains all the metadata, but data pages are flushed. That's again OK
    /// for bulk import, where you are just loading data pages and won't try to
    /// modify the same pages twice.
-    pub async fn flush(&mut self) -> anyhow::Result<()> {
+    pub async fn flush(&mut self, ctx: &RequestContext) -> anyhow::Result<()> {
        // Unless we have accumulated a decent amount of changes, it's not worth it
        // to scan through the pending_updates list.
        let pending_nblocks = self.pending_nblocks;
@@ -1154,7 +1154,7 @@ impl<'a> DatadirModification<'a> {
            if is_rel_block_key(key) || is_slru_block_key(key) {
                // This bails out on first error without modifying pending_updates.
                // That's Ok, cf this function's doc comment.
-                writer.put(key, self.lsn, &value).await?;
+                writer.put(key, self.lsn, &value, ctx).await?;
            } else {
                retained_pending_updates.insert(key, value);
            }
@@ -1174,14 +1174,14 @@ impl<'a> DatadirModification<'a> {
    /// underlying timeline.
    /// All the modifications in this atomic update are stamped by the specified LSN.
    ///
-    pub async fn commit(&mut self) -> anyhow::Result<()> {
+    pub async fn commit(&mut self, ctx: &RequestContext) -> anyhow::Result<()> {
        let writer = self.tline.writer().await;
        let lsn = self.lsn;
        let pending_nblocks = self.pending_nblocks;
        self.pending_nblocks = 0;

        for (key, value) in self.pending_updates.drain() {
-            writer.put(key, lsn, &value).await?;
+            writer.put(key, lsn, &value, ctx).await?;
        }
        for key_range in self.pending_deletions.drain(..) {
            writer.delete(key_range, lsn).await?;
--- a/pageserver/src/repository.rs
+++ b/pageserver/src/repository.rs
@@ -37,7 +37,7 @@ impl Key {
            | self.field6 as i128
    }

-    pub fn from_i128(x: i128) -> Self {
+    pub const fn from_i128(x: i128) -> Self {
        Key {
            field1: ((x >> 120) & 0xf) as u8,
            field2: ((x >> 104) & 0xFFFF) as u32,
--- a/pageserver/src/statvfs.rs
+++ b/pageserver/src/statvfs.rs
@@ -1,6 +1,6 @@
 //! Wrapper around nix::sys::statvfs::Statvfs that allows for mocking.

-use std::path::Path;
+use camino::Utf8Path;

 pub enum Statvfs {
    Real(nix::sys::statvfs::Statvfs),
@@ -12,11 +12,13 @@ pub enum Statvfs {
 // Sincce it should only be a problem on > 2TiB disks, let's ignore
 // the problem for now and upcast to u64.
 impl Statvfs {
-    pub fn get(tenants_dir: &Path, mocked: Option<&mock::Behavior>) -> nix::Result<Self> {
+    pub fn get(tenants_dir: &Utf8Path, mocked: Option<&mock::Behavior>) -> nix::Result<Self> {
        if let Some(mocked) = mocked {
            Ok(Statvfs::Mock(mock::get(tenants_dir, mocked)?))
        } else {
-            Ok(Statvfs::Real(nix::sys::statvfs::statvfs(tenants_dir)?))
+            Ok(Statvfs::Real(nix::sys::statvfs::statvfs(
+                tenants_dir.as_std_path(),
+            )?))
        }
    }

@@ -55,8 +57,8 @@ impl Statvfs {

 pub mod mock {
    use anyhow::Context;
+    use camino::Utf8Path;
    use regex::Regex;
-    use std::path::Path;
    use tracing::log::info;

    #[derive(Debug, Clone, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
@@ -86,7 +88,7 @@ pub mod mock {
        }
    }

-    pub fn get(tenants_dir: &Path, behavior: &Behavior) -> nix::Result<Statvfs> {
+    pub fn get(tenants_dir: &Utf8Path, behavior: &Behavior) -> nix::Result<Statvfs> {
        info!("running mocked statvfs");

        match behavior {
@@ -119,7 +121,7 @@ pub mod mock {
        }
    }

-    fn walk_dir_disk_usage(path: &Path, name_filter: Option<&Regex>) -> anyhow::Result<u64> {
+    fn walk_dir_disk_usage(path: &Utf8Path, name_filter: Option<&Regex>) -> anyhow::Result<u64> {
        let mut total = 0;
        for entry in walkdir::WalkDir::new(path) {
            let entry = entry?;
--- a/pageserver/src/task_mgr.rs
+++ b/pageserver/src/task_mgr.rs
@@ -187,6 +187,7 @@ task_local! {
    Debug,
    // NB: enumset::EnumSetType derives PartialEq, Eq, Clone, Copy
    enumset::EnumSetType,
+    enum_map::Enum,
    serde::Serialize,
    serde::Deserialize,
    strum_macros::IntoStaticStr,
@@ -455,7 +456,7 @@ async fn task_finish(
    }

    if shutdown_process {
-        shutdown_pageserver(1).await;
+        shutdown_pageserver(None, 1).await;
    }
 }

--- a/pageserver/src/tenant.rs
+++ b/pageserver/src/tenant.rs
--- a/pageserver/src/tenant/blob_io.rs
+++ b/pageserver/src/tenant/blob_io.rs
@@ -11,6 +11,7 @@
 //! len <  128: 0XXXXXXX
 //! len >= 128: 1XXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX
 //!
+use crate::context::RequestContext;
 use crate::page_cache::PAGE_SZ;
 use crate::tenant::block_io::BlockCursor;
 use crate::virtual_file::VirtualFile;
@@ -19,9 +20,13 @@ use std::io::{Error, ErrorKind};

 impl<'a> BlockCursor<'a> {
    /// Read a blob into a new buffer.
-    pub async fn read_blob(&self, offset: u64) -> Result<Vec<u8>, std::io::Error> {
+    pub async fn read_blob(
+        &self,
+        offset: u64,
+        ctx: &RequestContext,
+    ) -> Result<Vec<u8>, std::io::Error> {
        let mut buf = Vec::new();
-        self.read_blob_into_buf(offset, &mut buf).await?;
+        self.read_blob_into_buf(offset, &mut buf, ctx).await?;
        Ok(buf)
    }
    /// Read blob into the given buffer. Any previous contents in the buffer
@@ -30,11 +35,12 @@ impl<'a> BlockCursor<'a> {
        &self,
        offset: u64,
        dstbuf: &mut Vec<u8>,
+        ctx: &RequestContext,
    ) -> Result<(), std::io::Error> {
        let mut blknum = (offset / PAGE_SZ as u64) as u32;
        let mut off = (offset % PAGE_SZ as u64) as usize;

-        let mut buf = self.read_blk(blknum).await?;
+        let mut buf = self.read_blk(blknum, ctx).await?;

        // peek at the first byte, to determine if it's a 1- or 4-byte length
        let first_len_byte = buf[off];
@@ -50,7 +56,7 @@ impl<'a> BlockCursor<'a> {
                // it is split across two pages
                len_buf[..thislen].copy_from_slice(&buf[off..PAGE_SZ]);
                blknum += 1;
-                buf = self.read_blk(blknum).await?;
+                buf = self.read_blk(blknum, ctx).await?;
                len_buf[thislen..].copy_from_slice(&buf[0..4 - thislen]);
                off = 4 - thislen;
            } else {
@@ -71,7 +77,7 @@ impl<'a> BlockCursor<'a> {
            if page_remain == 0 {
                // continue on next page
                blknum += 1;
-                buf = self.read_blk(blknum).await?;
+                buf = self.read_blk(blknum, ctx).await?;
                off = 0;
                page_remain = PAGE_SZ;
            }
@@ -228,17 +234,18 @@ impl BlobWriter<false> {
 #[cfg(test)]
 mod tests {
    use super::*;
-    use crate::tenant::block_io::BlockReaderRef;
+    use crate::{context::DownloadBehavior, task_mgr::TaskKind, tenant::block_io::BlockReaderRef};
    use rand::{Rng, SeedableRng};

    async fn round_trip_test<const BUFFERED: bool>(blobs: &[Vec<u8>]) -> Result<(), Error> {
-        let temp_dir = tempfile::tempdir()?;
-        let path = temp_dir.path().join("file");
+        let temp_dir = camino_tempfile::tempdir()?;
+        let pathbuf = temp_dir.path().join("file");
+        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);

        // Write part (in block to drop the file)
        let mut offsets = Vec::new();
        {
-            let file = VirtualFile::create(&path).await?;
+            let file = VirtualFile::create(pathbuf.as_path()).await?;
            let mut wtr = BlobWriter::<BUFFERED>::new(file, 0);
            for blob in blobs.iter() {
                let offs = wtr.write_blob(blob).await?;
@@ -251,11 +258,11 @@ mod tests {
            wtr.flush_buffer().await?;
        }

-        let file = VirtualFile::open(&path).await?;
+        let file = VirtualFile::open(pathbuf.as_path()).await?;
        let rdr = BlockReaderRef::VirtualFile(&file);
        let rdr = BlockCursor::new(rdr);
        for (idx, (blob, offset)) in blobs.iter().zip(offsets.iter()).enumerate() {
-            let blob_read = rdr.read_blob(*offset).await?;
+            let blob_read = rdr.read_blob(*offset, &ctx).await?;
            assert_eq!(
                blob, &blob_read,
                "mismatch for idx={idx} at offset={offset}"
--- a/pageserver/src/tenant/block_io.rs
+++ b/pageserver/src/tenant/block_io.rs
@@ -4,6 +4,7 @@

 use super::ephemeral_file::EphemeralFile;
 use super::storage_layer::delta_layer::{Adapter, DeltaLayerInner};
+use crate::context::RequestContext;
 use crate::page_cache::{self, PageReadGuard, ReadBufResult, PAGE_SZ};
 use crate::virtual_file::VirtualFile;
 use bytes::Bytes;
@@ -82,12 +83,16 @@ pub(crate) enum BlockReaderRef<'a> {

 impl<'a> BlockReaderRef<'a> {
    #[inline(always)]
-    async fn read_blk(&self, blknum: u32) -> Result<BlockLease, std::io::Error> {
+    async fn read_blk(
+        &self,
+        blknum: u32,
+        ctx: &RequestContext,
+    ) -> Result<BlockLease, std::io::Error> {
        use BlockReaderRef::*;
        match self {
-            FileBlockReader(r) => r.read_blk(blknum).await,
-            EphemeralFile(r) => r.read_blk(blknum).await,
-            Adapter(r) => r.read_blk(blknum).await,
+            FileBlockReader(r) => r.read_blk(blknum, ctx).await,
+            EphemeralFile(r) => r.read_blk(blknum, ctx).await,
+            Adapter(r) => r.read_blk(blknum, ctx).await,
            #[cfg(test)]
            TestDisk(r) => r.read_blk(blknum),
            #[cfg(test)]
@@ -105,11 +110,13 @@ impl<'a> BlockReaderRef<'a> {
 ///
 /// ```no_run
 /// # use pageserver::tenant::block_io::{BlockReader, FileBlockReader};
+/// # use pageserver::context::RequestContext;
 /// # let reader: FileBlockReader = unimplemented!("stub");
+/// # let ctx: RequestContext = unimplemented!("stub");
 /// let cursor = reader.block_cursor();
-/// let buf = cursor.read_blk(1);
+/// let buf = cursor.read_blk(1, &ctx);
 /// // do stuff with 'buf'
-/// let buf = cursor.read_blk(2);
+/// let buf = cursor.read_blk(2, &ctx);
 /// // do stuff with 'buf'
 /// ```
 ///
@@ -134,8 +141,12 @@ impl<'a> BlockCursor<'a> {
    /// access to the contents of the page. (For the page cache, the
    /// lease object represents a lock on the buffer.)
    #[inline(always)]
-    pub async fn read_blk(&self, blknum: u32) -> Result<BlockLease, std::io::Error> {
-        self.reader.read_blk(blknum).await
+    pub async fn read_blk(
+        &self,
+        blknum: u32,
+        ctx: &RequestContext,
+    ) -> Result<BlockLease, std::io::Error> {
+        self.reader.read_blk(blknum, ctx).await
    }
 }

@@ -169,11 +180,15 @@ impl FileBlockReader {
    /// Returns a "lease" object that can be used to
    /// access to the contents of the page. (For the page cache, the
    /// lease object represents a lock on the buffer.)
-    pub async fn read_blk(&self, blknum: u32) -> Result<BlockLease, std::io::Error> {
+    pub async fn read_blk(
+        &self,
+        blknum: u32,
+        ctx: &RequestContext,
+    ) -> Result<BlockLease, std::io::Error> {
        let cache = page_cache::get();
        loop {
            match cache
-                .read_immutable_buf(self.file_id, blknum)
+                .read_immutable_buf(self.file_id, blknum, ctx)
                .await
                .map_err(|e| {
                    std::io::Error::new(
--- a/pageserver/src/tenant/config.rs
+++ b/pageserver/src/tenant/config.rs
@@ -13,6 +13,7 @@ use pageserver_api::models;
 use serde::{Deserialize, Serialize};
 use std::num::NonZeroU64;
 use std::time::Duration;
+use utils::generation::Generation;

 pub mod defaults {
    // FIXME: This current value is very low. I would imagine something like 1 GB or 10 GB
@@ -44,7 +45,211 @@ pub mod defaults {
    pub const DEFAULT_EVICTIONS_LOW_RESIDENCE_DURATION_METRIC_THRESHOLD: &str = "24 hour";
 }

-/// Per-tenant configuration options
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
+pub(crate) enum AttachmentMode {
+    /// Our generation is current as far as we know, and as far as we know we are the only attached
+    /// pageserver.  This is the "normal" attachment mode.
+    Single,
+    /// Our generation number is current as far as we know, but we are advised that another
+    /// pageserver is still attached, and therefore to avoid executing deletions.   This is
+    /// the attachment mode of a pagesever that is the destination of a migration.
+    Multi,
+    /// Our generation number is superseded, or about to be superseded.  We are advised
+    /// to avoid remote storage writes if possible, and to avoid sending billing data.  This
+    /// is the attachment mode of a pageserver that is the origin of a migration.
+    Stale,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
+pub(crate) struct AttachedLocationConfig {
+    pub(crate) generation: Generation,
+    pub(crate) attach_mode: AttachmentMode,
+    // TODO: add a flag to override AttachmentMode's policies under
+    // disk pressure (i.e. unblock uploads under disk pressure in Stale
+    // state, unblock deletions after timeout in Multi state)
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
+pub(crate) struct SecondaryLocationConfig {
+    /// If true, keep the local cache warm by polling remote storage
+    pub(crate) warm: bool,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
+pub(crate) enum LocationMode {
+    Attached(AttachedLocationConfig),
+    Secondary(SecondaryLocationConfig),
+}
+
+/// Per-tenant, per-pageserver configuration.  All pageservers use the same TenantConf,
+/// but have distinct LocationConf.
+#[derive(Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub(crate) struct LocationConf {
+    /// The location-specific part of the configuration, describes the operating
+    /// mode of this pageserver for this tenant.
+    pub(crate) mode: LocationMode,
+    /// The pan-cluster tenant configuration, the same on all locations
+    pub(crate) tenant_conf: TenantConfOpt,
+}
+
+impl std::fmt::Debug for LocationConf {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match &self.mode {
+            LocationMode::Attached(conf) => {
+                write!(
+                    f,
+                    "Attached {:?}, gen={:?}",
+                    conf.attach_mode, conf.generation
+                )
+            }
+            LocationMode::Secondary(conf) => {
+                write!(f, "Secondary, warm={}", conf.warm)
+            }
+        }
+    }
+}
+
+impl AttachedLocationConfig {
+    /// Consult attachment mode to determine whether we are currently permitted
+    /// to delete layers.  This is only advisory, not required for data safety.
+    /// See [`AttachmentMode`] for more context.
+    pub(crate) fn may_delete_layers_hint(&self) -> bool {
+        // TODO: add an override for disk pressure in AttachedLocationConfig,
+        // and respect it here.
+        match &self.attach_mode {
+            AttachmentMode::Single => true,
+            AttachmentMode::Multi | AttachmentMode::Stale => {
+                // In Multi mode we avoid doing deletions because some other
+                // attached pageserver might get 404 while trying to read
+                // a layer we delete which is still referenced in their metadata.
+                //
+                // In Stale mode, we avoid doing deletions because we expect
+                // that they would ultimately fail validation in the deletion
+                // queue due to our stale generation.
+                false
+            }
+        }
+    }
+
+    /// Whether we are currently hinted that it is worthwhile to upload layers.
+    /// This is only advisory, not required for data safety.
+    /// See [`AttachmentMode`] for more context.
+    pub(crate) fn may_upload_layers_hint(&self) -> bool {
+        // TODO: add an override for disk pressure in AttachedLocationConfig,
+        // and respect it here.
+        match &self.attach_mode {
+            AttachmentMode::Single | AttachmentMode::Multi => true,
+            AttachmentMode::Stale => {
+                // In Stale mode, we avoid doing uploads because we expect that
+                // our replacement pageserver will already have started its own
+                // IndexPart that will never reference layers we upload: it is
+                // wasteful.
+                false
+            }
+        }
+    }
+}
+
+impl LocationConf {
+    /// For use when loading from a legacy configuration: presence of a tenant
+    /// implies it is in AttachmentMode::Single, which used to be the only
+    /// possible state.  This function should eventually be removed.
+    pub(crate) fn attached_single(tenant_conf: TenantConfOpt, generation: Generation) -> Self {
+        Self {
+            mode: LocationMode::Attached(AttachedLocationConfig {
+                generation,
+                attach_mode: AttachmentMode::Single,
+            }),
+            tenant_conf,
+        }
+    }
+
+    /// For use when attaching/re-attaching: update the generation stored in this
+    /// structure.  If we were in a secondary state, promote to attached (posession
+    /// of a fresh generation implies this).
+    pub(crate) fn attach_in_generation(&mut self, generation: Generation) {
+        match &mut self.mode {
+            LocationMode::Attached(attach_conf) => {
+                attach_conf.generation = generation;
+            }
+            LocationMode::Secondary(_) => {
+                // We are promoted to attached by the control plane's re-attach response
+                self.mode = LocationMode::Attached(AttachedLocationConfig {
+                    generation,
+                    attach_mode: AttachmentMode::Single,
+                })
+            }
+        }
+    }
+
+    pub(crate) fn try_from(conf: &'_ models::LocationConfig) -> anyhow::Result<Self> {
+        let tenant_conf = TenantConfOpt::try_from(&conf.tenant_conf)?;
+
+        fn get_generation(conf: &'_ models::LocationConfig) -> Result<Generation, anyhow::Error> {
+            conf.generation
+                .ok_or_else(|| anyhow::anyhow!("Generation must be set when attaching"))
+        }
+
+        let mode = match &conf.mode {
+            models::LocationConfigMode::AttachedMulti => {
+                LocationMode::Attached(AttachedLocationConfig {
+                    generation: get_generation(conf)?,
+                    attach_mode: AttachmentMode::Multi,
+                })
+            }
+            models::LocationConfigMode::AttachedSingle => {
+                LocationMode::Attached(AttachedLocationConfig {
+                    generation: get_generation(conf)?,
+                    attach_mode: AttachmentMode::Single,
+                })
+            }
+            models::LocationConfigMode::AttachedStale => {
+                LocationMode::Attached(AttachedLocationConfig {
+                    generation: get_generation(conf)?,
+                    attach_mode: AttachmentMode::Stale,
+                })
+            }
+            models::LocationConfigMode::Secondary => {
+                anyhow::ensure!(conf.generation.is_none());
+
+                let warm = conf
+                    .secondary_conf
+                    .as_ref()
+                    .map(|c| c.warm)
+                    .unwrap_or(false);
+                LocationMode::Secondary(SecondaryLocationConfig { warm })
+            }
+            models::LocationConfigMode::Detached => {
+                // Should not have been called: API code should translate this mode
+                // into a detach rather than trying to decode it as a LocationConf
+                return Err(anyhow::anyhow!("Cannot decode a Detached configuration"));
+            }
+        };
+
+        Ok(Self { mode, tenant_conf })
+    }
+}
+
+impl Default for LocationConf {
+    // TODO: this should be removed once tenant loading can guarantee that we are never
+    // loading from a directory without a configuration.
+    // => tech debt since https://github.com/neondatabase/neon/issues/1555
+    fn default() -> Self {
+        Self {
+            mode: LocationMode::Attached(AttachedLocationConfig {
+                generation: Generation::none(),
+                attach_mode: AttachmentMode::Single,
+            }),
+            tenant_conf: TenantConfOpt::default(),
+        }
+    }
+}
+
+/// A tenant's calcuated configuration, which is the result of merging a
+/// tenant's TenantConfOpt with the global TenantConf from PageServerConf.
+///
+/// For storing and transmitting individual tenant's configuration, see
+/// TenantConfOpt.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
 pub struct TenantConf {
    // Flush out an inmemory layer, if it's holding WAL older than this
--- a/pageserver/src/tenant/delete.rs
+++ b/pageserver/src/tenant/delete.rs
@@ -1,9 +1,7 @@
-use std::{
-    path::{Path, PathBuf},
-    sync::Arc,
-};
+use std::sync::Arc;

 use anyhow::Context;
+use camino::{Utf8Path, Utf8PathBuf};
 use pageserver_api::models::TenantState;
 use remote_storage::{DownloadError, GenericRemoteStorage, RemotePath};
 use tokio::sync::OwnedMutexGuard;
@@ -62,7 +60,7 @@ fn remote_tenant_delete_mark_path(
        .context("Failed to strip workdir prefix")
        .and_then(RemotePath::new)
        .context("tenant path")?;
-    Ok(tenant_remote_path.join(Path::new("deleted")))
+    Ok(tenant_remote_path.join(Utf8Path::new("deleted")))
 }

 async fn create_remote_delete_mark(
@@ -148,7 +146,7 @@ async fn schedule_ordered_timeline_deletions(
    Ok(already_running_deletions)
 }

-async fn ensure_timelines_dir_empty(timelines_path: &Path) -> Result<(), DeleteTenantError> {
+async fn ensure_timelines_dir_empty(timelines_path: &Utf8Path) -> Result<(), DeleteTenantError> {
    // Assert timelines dir is empty.
    if !fs_ext::is_directory_empty(timelines_path).await? {
        // Display first 10 items in directory
@@ -188,20 +186,18 @@ async fn cleanup_remaining_fs_traces(
    conf: &PageServerConf,
    tenant_id: &TenantId,
 ) -> Result<(), DeleteTenantError> {
-    let rm = |p: PathBuf, is_dir: bool| async move {
+    let rm = |p: Utf8PathBuf, is_dir: bool| async move {
        if is_dir {
            tokio::fs::remove_dir(&p).await
        } else {
            tokio::fs::remove_file(&p).await
        }
        .or_else(fs_ext::ignore_not_found)
-        .with_context(|| {
-            let to_display = p.display();
-            format!("failed to delete {to_display}")
-        })
+        .with_context(|| format!("failed to delete {p}"))
    };

    rm(conf.tenant_config_path(tenant_id), false).await?;
+    rm(conf.tenant_location_config_path(tenant_id), false).await?;

    fail::fail_point!("tenant-delete-before-remove-timelines-dir", |_| {
        Err(anyhow::anyhow!(
--- a/pageserver/src/tenant/disk_btree.rs
+++ b/pageserver/src/tenant/disk_btree.rs
@@ -26,7 +26,11 @@ use std::{cmp::Ordering, io, result};
 use thiserror::Error;
 use tracing::error;

-use crate::tenant::block_io::{BlockReader, BlockWriter};
+use crate::{
+    context::{DownloadBehavior, RequestContext},
+    task_mgr::TaskKind,
+    tenant::block_io::{BlockReader, BlockWriter},
+};

 // The maximum size of a value stored in the B-tree. 5 bytes is enough currently.
 pub const VALUE_SZ: usize = 5;
@@ -231,14 +235,19 @@ where
    ///
    /// Read the value for given key. Returns the value, or None if it doesn't exist.
    ///
-    pub async fn get(&self, search_key: &[u8; L]) -> Result<Option<u64>> {
+    pub async fn get(&self, search_key: &[u8; L], ctx: &RequestContext) -> Result<Option<u64>> {
        let mut result: Option<u64> = None;
-        self.visit(search_key, VisitDirection::Forwards, |key, value| {
-            if key == search_key {
-                result = Some(value);
-            }
-            false
-        })
+        self.visit(
+            search_key,
+            VisitDirection::Forwards,
+            |key, value| {
+                if key == search_key {
+                    result = Some(value);
+                }
+                false
+            },
+            ctx,
+        )
        .await?;
        Ok(result)
    }
@@ -253,6 +262,7 @@ where
        search_key: &[u8; L],
        dir: VisitDirection,
        mut visitor: V,
+        ctx: &RequestContext,
    ) -> Result<bool>
    where
        V: FnMut(&[u8], u64) -> bool,
@@ -262,7 +272,9 @@ where
        let block_cursor = self.reader.block_cursor();
        while let Some((node_blknum, opt_iter)) = stack.pop() {
            // Locate the node.
-            let node_buf = block_cursor.read_blk(self.start_blk + node_blknum).await?;
+            let node_buf = block_cursor
+                .read_blk(self.start_blk + node_blknum, ctx)
+                .await?;

            let node = OnDiskNode::deparse(node_buf.as_ref())?;
            let prefix_len = node.prefix_len as usize;
@@ -351,13 +363,14 @@ where
    #[allow(dead_code)]
    pub async fn dump(&self) -> 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).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)?;

@@ -688,6 +701,8 @@ impl<const L: usize> BuildNode<L> {
 #[cfg(test)]
 pub(crate) mod tests {
    use super::*;
+    use crate::context::DownloadBehavior;
+    use crate::task_mgr::TaskKind;
    use crate::tenant::block_io::{BlockCursor, BlockLease, BlockReaderRef};
    use rand::Rng;
    use std::collections::BTreeMap;
@@ -725,6 +740,8 @@ 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 all_keys: Vec<&[u8; 6]> = vec![
            b"xaaaaa", b"xaaaba", b"xaaaca", b"xabaaa", b"xababa", b"xabaca", b"xabada", b"xabadb",
        ];
@@ -745,12 +762,12 @@ pub(crate) mod tests {

        // Test the `get` function on all the keys.
        for (key, val) in all_data.iter() {
-            assert_eq!(reader.get(key).await?, Some(*val));
+            assert_eq!(reader.get(key, &ctx).await?, Some(*val));
        }
        // And on some keys that don't exist
-        assert_eq!(reader.get(b"aaaaaa").await?, None);
-        assert_eq!(reader.get(b"zzzzzz").await?, None);
-        assert_eq!(reader.get(b"xaaabx").await?, None);
+        assert_eq!(reader.get(b"aaaaaa", &ctx).await?, None);
+        assert_eq!(reader.get(b"zzzzzz", &ctx).await?, None);
+        assert_eq!(reader.get(b"xaaabx", &ctx).await?, None);

        // Test search with `visit` function
        let search_key = b"xabaaa";
@@ -762,10 +779,15 @@ pub(crate) mod tests {

        let mut data = Vec::new();
        reader
-            .visit(search_key, VisitDirection::Forwards, |key, value| {
-                data.push((key.to_vec(), value));
-                true
-            })
+            .visit(
+                search_key,
+                VisitDirection::Forwards,
+                |key, value| {
+                    data.push((key.to_vec(), value));
+                    true
+                },
+                &ctx,
+            )
            .await?;
        assert_eq!(data, expected);

@@ -778,18 +800,28 @@ pub(crate) mod tests {
        expected.reverse();
        let mut data = Vec::new();
        reader
-            .visit(search_key, VisitDirection::Backwards, |key, value| {
-                data.push((key.to_vec(), value));
-                true
-            })
+            .visit(
+                search_key,
+                VisitDirection::Backwards,
+                |key, value| {
+                    data.push((key.to_vec(), value));
+                    true
+                },
+                &ctx,
+            )
            .await?;
        assert_eq!(data, expected);

        // Backward scan where nothing matches
        reader
-            .visit(b"aaaaaa", VisitDirection::Backwards, |key, value| {
-                panic!("found unexpected key {}: {}", hex::encode(key), value);
-            })
+            .visit(
+                b"aaaaaa",
+                VisitDirection::Backwards,
+                |key, value| {
+                    panic!("found unexpected key {}: {}", hex::encode(key), value);
+                },
+                &ctx,
+            )
            .await?;

        // Full scan
@@ -799,10 +831,15 @@ pub(crate) mod tests {
            .collect();
        let mut data = Vec::new();
        reader
-            .visit(&[0u8; 6], VisitDirection::Forwards, |key, value| {
-                data.push((key.to_vec(), value));
-                true
-            })
+            .visit(
+                &[0u8; 6],
+                VisitDirection::Forwards,
+                |key, value| {
+                    data.push((key.to_vec(), value));
+                    true
+                },
+                &ctx,
+            )
            .await?;
        assert_eq!(data, expected);

@@ -813,6 +850,7 @@ 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);

        const NUM_KEYS: u64 = 1000;

@@ -851,14 +889,14 @@ pub(crate) mod tests {
        for search_key_int in 0..(NUM_KEYS * 2 + 10) {
            let search_key = u64::to_be_bytes(search_key_int);
            assert_eq!(
-                reader.get(&search_key).await?,
+                reader.get(&search_key, &ctx).await?,
                all_data.get(&search_key_int).cloned()
            );

            // Test a forward scan starting with this key
            result.lock().unwrap().clear();
            reader
-                .visit(&search_key, VisitDirection::Forwards, take_ten)
+                .visit(&search_key, VisitDirection::Forwards, take_ten, &ctx)
                .await?;
            let expected = all_data
                .range(search_key_int..)
@@ -870,7 +908,7 @@ pub(crate) mod tests {
            // And a backwards scan
            result.lock().unwrap().clear();
            reader
-                .visit(&search_key, VisitDirection::Backwards, take_ten)
+                .visit(&search_key, VisitDirection::Backwards, take_ten, &ctx)
                .await?;
            let expected = all_data
                .range(..=search_key_int)
@@ -886,7 +924,7 @@ pub(crate) mod tests {
        limit.store(usize::MAX, Ordering::Relaxed);
        result.lock().unwrap().clear();
        reader
-            .visit(&search_key, VisitDirection::Forwards, take_ten)
+            .visit(&search_key, VisitDirection::Forwards, take_ten, &ctx)
            .await?;
        let expected = all_data
            .iter()
@@ -899,7 +937,7 @@ pub(crate) mod tests {
        limit.store(usize::MAX, Ordering::Relaxed);
        result.lock().unwrap().clear();
        reader
-            .visit(&search_key, VisitDirection::Backwards, take_ten)
+            .visit(&search_key, VisitDirection::Backwards, take_ten, &ctx)
            .await?;
        let expected = all_data
            .iter()
@@ -913,6 +951,8 @@ pub(crate) mod tests {

    #[tokio::test]
    async fn random_data() -> Result<()> {
+        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+
        // Generate random keys with exponential distribution, to
        // exercise the prefix compression
        const NUM_KEYS: usize = 100000;
@@ -939,22 +979,24 @@ pub(crate) mod tests {
        // Test get() operation on all the keys
        for (&key, &val) in all_data.iter() {
            let search_key = u128::to_be_bytes(key);
-            assert_eq!(reader.get(&search_key).await?, Some(val));
+            assert_eq!(reader.get(&search_key, &ctx).await?, Some(val));
        }

        // Test get() operations on random keys, most of which will not exist
        for _ in 0..100000 {
            let key_int = rand::thread_rng().gen::<u128>();
            let search_key = u128::to_be_bytes(key_int);
-            assert!(reader.get(&search_key).await? == all_data.get(&key_int).cloned());
+            assert!(reader.get(&search_key, &ctx).await? == all_data.get(&key_int).cloned());
        }

        // Test boundary cases
        assert!(
-            reader.get(&u128::to_be_bytes(u128::MIN)).await? == all_data.get(&u128::MIN).cloned()
+            reader.get(&u128::to_be_bytes(u128::MIN), &ctx).await?
+                == all_data.get(&u128::MIN).cloned()
        );
        assert!(
-            reader.get(&u128::to_be_bytes(u128::MAX)).await? == all_data.get(&u128::MAX).cloned()
+            reader.get(&u128::to_be_bytes(u128::MAX), &ctx).await?
+                == all_data.get(&u128::MAX).cloned()
        );

        Ok(())
@@ -985,6 +1027,7 @@ 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);

        for (key, val) in disk_btree_test_data::TEST_DATA {
            writer.append(&key, val)?;
@@ -997,16 +1040,21 @@ pub(crate) mod tests {

        // Test get() operation on all the keys
        for (key, val) in disk_btree_test_data::TEST_DATA {
-            assert_eq!(reader.get(&key).await?, Some(val));
+            assert_eq!(reader.get(&key, &ctx).await?, Some(val));
        }

        // Test full scan
        let mut count = 0;
        reader
-            .visit(&[0u8; 26], VisitDirection::Forwards, |_key, _value| {
-                count += 1;
-                true
-            })
+            .visit(
+                &[0u8; 26],
+                VisitDirection::Forwards,
+                |_key, _value| {
+                    count += 1;
+                    true
+                },
+                &ctx,
+            )
            .await?;
        assert_eq!(count, disk_btree_test_data::TEST_DATA.len());

--- a/pageserver/src/tenant/ephemeral_file.rs
+++ b/pageserver/src/tenant/ephemeral_file.rs
@@ -2,14 +2,15 @@
 //! used to keep in-memory layers spilled on disk.

 use crate::config::PageServerConf;
+use crate::context::RequestContext;
 use crate::page_cache::{self, PAGE_SZ};
 use crate::tenant::block_io::{BlockCursor, BlockLease, BlockReader};
 use crate::virtual_file::VirtualFile;
+use camino::Utf8PathBuf;
 use std::cmp::min;
 use std::fs::OpenOptions;
 use std::io::{self, ErrorKind};
 use std::ops::DerefMut;
-use std::path::PathBuf;
 use std::sync::atomic::AtomicU64;
 use tracing::*;
 use utils::id::{TenantId, TimelineId};
@@ -39,7 +40,9 @@ impl EphemeralFile {

        let filename = conf
            .timeline_path(&tenant_id, &timeline_id)
-            .join(PathBuf::from(format!("ephemeral-{filename_disambiguator}")));
+            .join(Utf8PathBuf::from(format!(
+                "ephemeral-{filename_disambiguator}"
+            )));

        let file = VirtualFile::open_with_options(
            &filename,
@@ -61,13 +64,17 @@ impl EphemeralFile {
        self.len
    }

-    pub(crate) async fn read_blk(&self, blknum: u32) -> Result<BlockLease, io::Error> {
+    pub(crate) async fn read_blk(
+        &self,
+        blknum: u32,
+        ctx: &RequestContext,
+    ) -> Result<BlockLease, io::Error> {
        let flushed_blknums = 0..self.len / PAGE_SZ as u64;
        if flushed_blknums.contains(&(blknum as u64)) {
            let cache = page_cache::get();
            loop {
                match cache
-                    .read_immutable_buf(self.page_cache_file_id, blknum)
+                    .read_immutable_buf(self.page_cache_file_id, blknum, ctx)
                    .await
                    .map_err(|e| {
                        std::io::Error::new(
@@ -75,9 +82,7 @@ impl EphemeralFile {
                            // order path before error because error is anyhow::Error => might have many contexts
                            format!(
                                "ephemeral file: read immutable page #{}: {}: {:#}",
-                                blknum,
-                                self.file.path.display(),
-                                e,
+                                blknum, self.file.path, e,
                            ),
                        )
                    })? {
@@ -103,7 +108,11 @@ impl EphemeralFile {
        }
    }

-    pub(crate) async fn write_blob(&mut self, srcbuf: &[u8]) -> Result<u64, io::Error> {
+    pub(crate) async fn write_blob(
+        &mut self,
+        srcbuf: &[u8],
+        ctx: &RequestContext,
+    ) -> Result<u64, io::Error> {
        struct Writer<'a> {
            ephemeral_file: &'a mut EphemeralFile,
            /// The block to which the next [`push_bytes`] will write.
@@ -120,7 +129,11 @@ impl EphemeralFile {
                })
            }
            #[inline(always)]
-            async fn push_bytes(&mut self, src: &[u8]) -> Result<(), io::Error> {
+            async fn push_bytes(
+                &mut self,
+                src: &[u8],
+                ctx: &RequestContext,
+            ) -> Result<(), io::Error> {
                let mut src_remaining = src;
                while !src_remaining.is_empty() {
                    let dst_remaining = &mut self.ephemeral_file.mutable_tail[self.off..];
@@ -146,6 +159,7 @@ impl EphemeralFile {
                                    .read_immutable_buf(
                                        self.ephemeral_file.page_cache_file_id,
                                        self.blknum,
+                                        ctx,
                                    )
                                    .await
                                {
@@ -181,7 +195,7 @@ impl EphemeralFile {
                                        "ephemeral_file: write_blob: write-back full tail blk #{}: {:#}: {}",
                                        self.blknum,
                                        e,
-                                        self.ephemeral_file.file.path.display(),
+                                        self.ephemeral_file.file.path,
                                    ),
                                ));
                            }
@@ -199,15 +213,15 @@ impl EphemeralFile {
        if srcbuf.len() < 0x80 {
            // short one-byte length header
            let len_buf = [srcbuf.len() as u8];
-            writer.push_bytes(&len_buf).await?;
+            writer.push_bytes(&len_buf, ctx).await?;
        } else {
            let mut len_buf = u32::to_be_bytes(srcbuf.len() as u32);
            len_buf[0] |= 0x80;
-            writer.push_bytes(&len_buf).await?;
+            writer.push_bytes(&len_buf, ctx).await?;
        }

        // Write the payload
-        writer.push_bytes(srcbuf).await?;
+        writer.push_bytes(srcbuf, ctx).await?;

        if srcbuf.len() < 0x80 {
            self.len += 1;
@@ -244,8 +258,7 @@ impl Drop for EphemeralFile {
                // not found files might also be related to https://github.com/neondatabase/neon/issues/2442
                error!(
                    "could not remove ephemeral file '{}': {}",
-                    self.file.path.display(),
-                    e
+                    self.file.path, e
                );
            }
        }
@@ -261,6 +274,8 @@ impl BlockReader for EphemeralFile {
 #[cfg(test)]
 mod tests {
    use super::*;
+    use crate::context::DownloadBehavior;
+    use crate::task_mgr::TaskKind;
    use crate::tenant::block_io::{BlockCursor, BlockReaderRef};
    use rand::{thread_rng, RngCore};
    use std::fs;
@@ -268,7 +283,15 @@ mod tests {

    fn harness(
        test_name: &str,
-    ) -> Result<(&'static PageServerConf, TenantId, TimelineId), io::Error> {
+    ) -> Result<
+        (
+            &'static PageServerConf,
+            TenantId,
+            TimelineId,
+            RequestContext,
+        ),
+        io::Error,
+    > {
        let repo_dir = PageServerConf::test_repo_dir(test_name);
        let _ = fs::remove_dir_all(&repo_dir);
        let conf = PageServerConf::dummy_conf(repo_dir);
@@ -280,46 +303,57 @@ mod tests {
        let timeline_id = TimelineId::from_str("22000000000000000000000000000000").unwrap();
        fs::create_dir_all(conf.timeline_path(&tenant_id, &timeline_id))?;

-        Ok((conf, tenant_id, timeline_id))
+        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
+
+        Ok((conf, tenant_id, timeline_id, ctx))
    }

    #[tokio::test]
    async fn test_ephemeral_blobs() -> Result<(), io::Error> {
-        let (conf, tenant_id, timeline_id) = harness("ephemeral_blobs")?;
+        let (conf, tenant_id, timeline_id, ctx) = harness("ephemeral_blobs")?;

        let mut file = EphemeralFile::create(conf, tenant_id, timeline_id).await?;

-        let pos_foo = file.write_blob(b"foo").await?;
+        let pos_foo = file.write_blob(b"foo", &ctx).await?;
        assert_eq!(
            b"foo",
-            file.block_cursor().read_blob(pos_foo).await?.as_slice()
+            file.block_cursor()
+                .read_blob(pos_foo, &ctx)
+                .await?
+                .as_slice()
        );
-        let pos_bar = file.write_blob(b"bar").await?;
+        let pos_bar = file.write_blob(b"bar", &ctx).await?;
        assert_eq!(
            b"foo",
-            file.block_cursor().read_blob(pos_foo).await?.as_slice()
+            file.block_cursor()
+                .read_blob(pos_foo, &ctx)
+                .await?
+                .as_slice()
        );
        assert_eq!(
            b"bar",
-            file.block_cursor().read_blob(pos_bar).await?.as_slice()
+            file.block_cursor()
+                .read_blob(pos_bar, &ctx)
+                .await?
+                .as_slice()
        );

        let mut blobs = Vec::new();
        for i in 0..10000 {
            let data = Vec::from(format!("blob{}", i).as_bytes());
-            let pos = file.write_blob(&data).await?;
+            let pos = file.write_blob(&data, &ctx).await?;
            blobs.push((pos, data));
        }
        // also test with a large blobs
        for i in 0..100 {
            let data = format!("blob{}", i).as_bytes().repeat(100);
-            let pos = file.write_blob(&data).await?;
+            let pos = file.write_blob(&data, &ctx).await?;
            blobs.push((pos, data));
        }

        let cursor = BlockCursor::new(BlockReaderRef::EphemeralFile(&file));
        for (pos, expected) in blobs {
-            let actual = cursor.read_blob(pos).await?;
+            let actual = cursor.read_blob(pos, &ctx).await?;
            assert_eq!(actual, expected);
        }

@@ -327,8 +361,8 @@ mod tests {
        let mut large_data = Vec::new();
        large_data.resize(20000, 0);
        thread_rng().fill_bytes(&mut large_data);
-        let pos_large = file.write_blob(&large_data).await?;
-        let result = file.block_cursor().read_blob(pos_large).await?;
+        let pos_large = file.write_blob(&large_data, &ctx).await?;
+        let result = file.block_cursor().read_blob(pos_large, &ctx).await?;
        assert_eq!(result, large_data);

        Ok(())
--- a/pageserver/src/tenant/mgr.rs
+++ b/pageserver/src/tenant/mgr.rs
@@ -1,10 +1,9 @@
 //! This module acts as a switchboard to access different repositories managed by this
 //! page server.

+use camino::{Utf8Path, Utf8PathBuf};
 use rand::{distributions::Alphanumeric, Rng};
 use std::collections::{hash_map, HashMap};
-use std::ffi::OsStr;
-use std::path::{Path, PathBuf};
 use std::sync::Arc;
 use tokio::fs;

@@ -20,11 +19,16 @@ use utils::crashsafe;

 use crate::config::PageServerConf;
 use crate::context::{DownloadBehavior, RequestContext};
-use crate::control_plane_client::ControlPlaneClient;
+use crate::control_plane_client::{
+    ControlPlaneClient, ControlPlaneGenerationsApi, RetryForeverError,
+};
+use crate::deletion_queue::DeletionQueueClient;
 use crate::task_mgr::{self, TaskKind};
-use crate::tenant::config::TenantConfOpt;
+use crate::tenant::config::{LocationConf, LocationMode, TenantConfOpt};
 use crate::tenant::delete::DeleteTenantFlow;
-use crate::tenant::{create_tenant_files, CreateTenantFilesMode, Tenant, TenantState};
+use crate::tenant::{
+    create_tenant_files, AttachedTenantConf, CreateTenantFilesMode, Tenant, TenantState,
+};
 use crate::{InitializationOrder, IGNORED_TENANT_FILE_NAME, TEMP_FILE_SUFFIX};

 use utils::crashsafe::path_with_suffix_extension;
@@ -36,6 +40,39 @@ use super::delete::DeleteTenantError;
 use super::timeline::delete::DeleteTimelineFlow;
 use super::TenantSharedResources;

+/// For a tenant that appears in TenantsMap, it may either be
+/// - `Attached`: has a full Tenant object, is elegible to service
+///    reads and ingest WAL.
+/// - `Secondary`: is only keeping a local cache warm.
+///
+/// Secondary is a totally distinct state rather than being a mode of a `Tenant`, because
+/// that way we avoid having to carefully switch a tenant's ingestion etc on and off during
+/// its lifetime, and we can preserve some important safety invariants like `Tenant` always
+/// having a properly acquired generation (Secondary doesn't need a generation)
+#[derive(Clone)]
+pub enum TenantSlot {
+    Attached(Arc<Tenant>),
+    Secondary,
+}
+
+impl TenantSlot {
+    /// Return the `Tenant` in this slot if attached, else None
+    fn get_attached(&self) -> Option<&Arc<Tenant>> {
+        match self {
+            Self::Attached(t) => Some(t),
+            Self::Secondary => None,
+        }
+    }
+
+    /// Consume self and return the `Tenant` that was in this slot if attached, else None
+    fn into_attached(self) -> Option<Arc<Tenant>> {
+        match self {
+            Self::Attached(t) => Some(t),
+            Self::Secondary => None,
+        }
+    }
+}
+
 /// The tenants known to the pageserver.
 /// The enum variants are used to distinguish the different states that the pageserver can be in.
 pub(crate) enum TenantsMap {
@@ -43,14 +80,27 @@ pub(crate) enum TenantsMap {
    Initializing,
    /// [`init_tenant_mgr`] is done, all on-disk tenants have been loaded.
    /// New tenants can be added using [`tenant_map_insert`].
-    Open(HashMap<TenantId, Arc<Tenant>>),
+    Open(HashMap<TenantId, TenantSlot>),
    /// The pageserver has entered shutdown mode via [`shutdown_all_tenants`].
    /// Existing tenants are still accessible, but no new tenants can be created.
-    ShuttingDown(HashMap<TenantId, Arc<Tenant>>),
+    ShuttingDown(HashMap<TenantId, TenantSlot>),
 }

 impl TenantsMap {
+    /// Convenience function for typical usage, where we want to get a `Tenant` object, for
+    /// working with attached tenants.  If the TenantId is in the map but in Secondary state,
+    /// None is returned.
    pub(crate) fn get(&self, tenant_id: &TenantId) -> Option<&Arc<Tenant>> {
+        match self {
+            TenantsMap::Initializing => None,
+            TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => {
+                m.get(tenant_id).and_then(TenantSlot::get_attached)
+            }
+        }
+    }
+
+    /// Get the contents of the map at this tenant ID, even if it is in secondary state.
+    pub(crate) fn get_slot(&self, tenant_id: &TenantId) -> Option<&TenantSlot> {
        match self {
            TenantsMap::Initializing => None,
            TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => m.get(tenant_id),
@@ -59,7 +109,9 @@ impl TenantsMap {
    pub(crate) fn remove(&mut self, tenant_id: &TenantId) -> Option<Arc<Tenant>> {
        match self {
            TenantsMap::Initializing => None,
-            TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => m.remove(tenant_id),
+            TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => {
+                m.remove(tenant_id).and_then(TenantSlot::into_attached)
+            }
        }
    }
 }
@@ -70,12 +122,12 @@ impl TenantsMap {
 ///
 /// This is pageserver-specific, as it relies on future processes after a crash to check
 /// for TEMP_FILE_SUFFIX when loading things.
-async fn safe_remove_tenant_dir_all(path: impl AsRef<Path>) -> std::io::Result<()> {
+async fn safe_remove_tenant_dir_all(path: impl AsRef<Utf8Path>) -> std::io::Result<()> {
    let tmp_path = safe_rename_tenant_dir(path).await?;
    fs::remove_dir_all(tmp_path).await
 }

-async fn safe_rename_tenant_dir(path: impl AsRef<Path>) -> std::io::Result<PathBuf> {
+async fn safe_rename_tenant_dir(path: impl AsRef<Utf8Path>) -> std::io::Result<Utf8PathBuf> {
    let parent = path
        .as_ref()
        .parent()
@@ -92,7 +144,7 @@ async fn safe_rename_tenant_dir(path: impl AsRef<Path>) -> std::io::Result<PathB
        .collect::<String>()
        + TEMP_FILE_SUFFIX;
    let tmp_path = path_with_suffix_extension(&path, &rand_suffix);
-    fs::rename(&path, &tmp_path).await?;
+    fs::rename(path.as_ref(), &tmp_path).await?;
    fs::File::open(parent).await?.sync_all().await?;
    Ok(tmp_path)
 }
@@ -116,35 +168,52 @@ pub async fn init_tenant_mgr(

    // If we are configured to use the control plane API, then it is the source of truth for what tenants to load.
    let tenant_generations = if let Some(client) = ControlPlaneClient::new(conf, &cancel) {
-        Some(client.re_attach().await?)
+        let result = match client.re_attach().await {
+            Ok(tenants) => tenants,
+            Err(RetryForeverError::ShuttingDown) => {
+                anyhow::bail!("Shut down while waiting for control plane re-attach response")
+            }
+        };
+
+        // The deletion queue needs to know about the startup attachment state to decide which (if any) stored
+        // deletion list entries may still be valid.  We provide that by pushing a recovery operation into
+        // the queue. Sequential processing of te queue ensures that recovery is done before any new tenant deletions
+        // are processed, even though we don't block on recovery completing here.
+        //
+        // Must only do this if remote storage is enabled, otherwise deletion queue
+        // is not running and channel push will fail.
+        if resources.remote_storage.is_some() {
+            resources
+                .deletion_queue_client
+                .recover(result.clone())
+                .await?;
+        }
+
+        Some(result)
    } else {
        info!("Control plane API not configured, tenant generations are disabled");
        None
    };

-    let mut dir_entries = fs::read_dir(&tenants_dir)
-        .await
+    let mut dir_entries = tenants_dir
+        .read_dir_utf8()
        .with_context(|| format!("Failed to list tenants dir {tenants_dir:?}"))?;

    let ctx = RequestContext::todo_child(TaskKind::Startup, DownloadBehavior::Warn);

    loop {
-        match dir_entries.next_entry().await {
-            Ok(None) => break,
-            Ok(Some(dir_entry)) => {
-                let tenant_dir_path = dir_entry.path();
+        match dir_entries.next() {
+            None => break,
+            Some(Ok(dir_entry)) => {
+                let tenant_dir_path = dir_entry.path().to_path_buf();
                if crate::is_temporary(&tenant_dir_path) {
-                    info!(
-                        "Found temporary tenant directory, removing: {}",
-                        tenant_dir_path.display()
-                    );
+                    info!("Found temporary tenant directory, removing: {tenant_dir_path}");
                    // No need to use safe_remove_tenant_dir_all because this is already
                    // a temporary path
                    if let Err(e) = fs::remove_dir_all(&tenant_dir_path).await {
                        error!(
                            "Failed to remove temporary directory '{}': {:?}",
-                            tenant_dir_path.display(),
-                            e
+                            tenant_dir_path, e
                        );
                    }
                } else {
@@ -159,7 +228,7 @@ pub async fn init_tenant_mgr(
                        if let Err(e) = fs::remove_dir(&tenant_dir_path).await {
                            error!(
                                "Failed to remove empty tenant directory '{}': {e:#}",
-                                tenant_dir_path.display()
+                                tenant_dir_path
                            )
                        }
                        continue;
@@ -173,7 +242,6 @@ pub async fn init_tenant_mgr(

                    let tenant_id = match tenant_dir_path
                        .file_name()
-                        .and_then(OsStr::to_str)
                        .unwrap_or_default()
                        .parse::<TenantId>()
                    {
@@ -181,26 +249,65 @@ pub async fn init_tenant_mgr(
                        Err(_) => {
                            warn!(
                                "Invalid tenant path (garbage in our repo directory?): {}",
-                                tenant_dir_path.display()
+                                tenant_dir_path
                            );
                            continue;
                        }
                    };

+                    // Try loading the location configuration
+                    let mut location_conf = match Tenant::load_tenant_config(conf, &tenant_id)
+                        .context("load tenant config")
+                    {
+                        Ok(c) => c,
+                        Err(e) => {
+                            warn!("Marking tenant broken, failed to {e:#}");
+
+                            tenants.insert(
+                                tenant_id,
+                                TenantSlot::Attached(Tenant::create_broken_tenant(
+                                    conf,
+                                    tenant_id,
+                                    "error loading tenant location configuration".to_string(),
+                                )),
+                            );
+
+                            continue;
+                        }
+                    };
+
                    let generation = if let Some(generations) = &tenant_generations {
                        // We have a generation map: treat it as the authority for whether
                        // this tenant is really attached.
                        if let Some(gen) = generations.get(&tenant_id) {
                            *gen
                        } else {
-                            info!("Detaching tenant {tenant_id}, control plane omitted it in re-attach response");
-                            if let Err(e) = safe_remove_tenant_dir_all(&tenant_dir_path).await {
-                                error!(
-                                    "Failed to remove detached tenant directory '{}': {:?}",
-                                    tenant_dir_path.display(),
-                                    e
-                                );
-                            }
+                            match &location_conf.mode {
+                                LocationMode::Secondary(_) => {
+                                    // We do not require the control plane's permission for secondary mode
+                                    // tenants, because they do no remote writes and hence require no
+                                    // generation number
+                                    info!("Loaded tenant {tenant_id} in secondary mode");
+                                    tenants.insert(tenant_id, TenantSlot::Secondary);
+                                }
+                                LocationMode::Attached(_) => {
+                                    // TODO: augment re-attach API to enable the control plane to
+                                    // instruct us about secondary attachments.  That way, instead of throwing
+                                    // away local state, we can gracefully fall back to secondary here, if the control
+                                    // plane tells us so.
+                                    // (https://github.com/neondatabase/neon/issues/5377)
+                                    info!("Detaching tenant {tenant_id}, control plane omitted it in re-attach response");
+                                    if let Err(e) =
+                                        safe_remove_tenant_dir_all(&tenant_dir_path).await
+                                    {
+                                        error!(
+                                            "Failed to remove detached tenant directory '{}': {:?}",
+                                            tenant_dir_path, e
+                                        );
+                                    }
+                                }
+                            };
+
                            continue;
                        }
                    } else {
@@ -208,23 +315,28 @@ pub async fn init_tenant_mgr(
                        // on local disk may activate
                        info!(
                            "Starting tenant {} in legacy mode, no generation",
-                            tenant_dir_path.display()
+                            tenant_dir_path
                        );
                        Generation::none()
                    };

+                    // Presence of a generation number implies attachment: attach the tenant
+                    // if it wasn't already, and apply the generation number.
+                    location_conf.attach_in_generation(generation);
+                    Tenant::persist_tenant_config(conf, &tenant_id, &location_conf).await?;
+
                    match schedule_local_tenant_processing(
                        conf,
                        tenant_id,
                        &tenant_dir_path,
-                        generation,
+                        AttachedTenantConf::try_from(location_conf)?,
                        resources.clone(),
                        Some(init_order.clone()),
                        &TENANTS,
                        &ctx,
                    ) {
                        Ok(tenant) => {
-                            tenants.insert(tenant.tenant_id(), tenant);
+                            tenants.insert(tenant.tenant_id(), TenantSlot::Attached(tenant));
                        }
                        Err(e) => {
                            error!("Failed to collect tenant files from dir {tenants_dir:?} for entry {dir_entry:?}, reason: {e:#}");
@@ -232,7 +344,7 @@ pub async fn init_tenant_mgr(
                    }
                }
            }
-            Err(e) => {
+            Some(Err(e)) => {
                // On error, print it, but continue with the other tenants. If we error out
                // here, the pageserver startup fails altogether, causing outage for *all*
                // tenants. That seems worse.
@@ -255,8 +367,8 @@ pub async fn init_tenant_mgr(
 pub(crate) fn schedule_local_tenant_processing(
    conf: &'static PageServerConf,
    tenant_id: TenantId,
-    tenant_path: &Path,
-    generation: Generation,
+    tenant_path: &Utf8Path,
+    location_conf: AttachedTenantConf,
    resources: TenantSharedResources,
    init_order: Option<InitializationOrder>,
    tenants: &'static tokio::sync::RwLock<TenantsMap>,
@@ -285,35 +397,33 @@ pub(crate) fn schedule_local_tenant_processing(

    let tenant = if conf.tenant_attaching_mark_file_path(&tenant_id).exists() {
        info!("tenant {tenant_id} has attaching mark file, resuming its attach operation");
-        if let Some(remote_storage) = resources.remote_storage {
-            match Tenant::spawn_attach(
-                conf,
-                tenant_id,
-                generation,
-                resources.broker_client,
-                tenants,
-                remote_storage,
-                ctx,
-            ) {
-                Ok(tenant) => tenant,
-                Err(e) => {
-                    error!("Failed to spawn_attach tenant {tenant_id}, reason: {e:#}");
-                    Tenant::create_broken_tenant(conf, tenant_id, format!("{e:#}"))
-                }
-            }
-        } else {
+        if resources.remote_storage.is_none() {
            warn!("tenant {tenant_id} has attaching mark file, but pageserver has no remote storage configured");
            Tenant::create_broken_tenant(
                conf,
                tenant_id,
                "attaching mark file present but no remote storage configured".to_string(),
            )
+        } else {
+            match Tenant::spawn_attach(conf, tenant_id, resources, location_conf, tenants, ctx) {
+                Ok(tenant) => tenant,
+                Err(e) => {
+                    error!("Failed to spawn_attach tenant {tenant_id}, reason: {e:#}");
+                    Tenant::create_broken_tenant(conf, tenant_id, format!("{e:#}"))
+                }
+            }
        }
    } else {
        info!("tenant {tenant_id} is assumed to be loadable, starting load operation");
        // Start loading the tenant into memory. It will initially be in Loading state.
        Tenant::spawn_load(
-            conf, tenant_id, generation, resources, init_order, tenants, ctx,
+            conf,
+            tenant_id,
+            location_conf,
+            resources,
+            init_order,
+            tenants,
+            ctx,
        )
    };
    Ok(tenant)
@@ -369,7 +479,16 @@ async fn shutdown_all_tenants0(tenants: &tokio::sync::RwLock<TenantsMap>) {

                let res = {
                    let (_guard, shutdown_progress) = completion::channel();
-                    tenant.shutdown(shutdown_progress, freeze_and_flush).await
+                    match tenant {
+                        TenantSlot::Attached(t) => {
+                            t.shutdown(shutdown_progress, freeze_and_flush).await
+                        }
+                        TenantSlot::Secondary => {
+                            // TODO: once secondary mode downloads are implemented,
+                            // ensure they have all stopped before we reach this point.
+                            Ok(())
+                        }
+                    }
                };

                if let Err(other_progress) = res {
@@ -438,25 +557,23 @@ pub async fn create_tenant(
    tenant_conf: TenantConfOpt,
    tenant_id: TenantId,
    generation: Generation,
-    broker_client: storage_broker::BrokerClientChannel,
-    remote_storage: Option<GenericRemoteStorage>,
+    resources: TenantSharedResources,
    ctx: &RequestContext,
 ) -> Result<Arc<Tenant>, TenantMapInsertError> {
    tenant_map_insert(tenant_id, || async {
+
+        let location_conf = LocationConf::attached_single(tenant_conf, generation);
+
        // We're holding the tenants lock in write mode while doing local IO.
        // If this section ever becomes contentious, introduce a new `TenantState::Creating`
        // and do the work in that state.
-        let tenant_directory = super::create_tenant_files(conf, tenant_conf, &tenant_id, CreateTenantFilesMode::Create).await?;
+        let tenant_directory = super::create_tenant_files(conf, &location_conf, &tenant_id, CreateTenantFilesMode::Create).await?;
        // TODO: tenant directory remains on disk if we bail out from here on.
        //       See https://github.com/neondatabase/neon/issues/4233

-        let tenant_resources = TenantSharedResources {
-            broker_client,
-            remote_storage,
-        };
        let created_tenant =
            schedule_local_tenant_processing(conf, tenant_id, &tenant_directory,
-                generation, tenant_resources, None, &TENANTS, ctx)?;
+                AttachedTenantConf::try_from(location_conf)?, resources, None, &TENANTS, ctx)?;
        // TODO: tenant object & its background loops remain, untracked in tenant map, if we fail here.
        //      See https://github.com/neondatabase/neon/issues/4233

@@ -485,20 +602,137 @@ pub async fn set_new_tenant_config(
    info!("configuring tenant {tenant_id}");
    let tenant = get_tenant(tenant_id, true).await?;

-    let tenant_config_path = conf.tenant_config_path(&tenant_id);
-    Tenant::persist_tenant_config(&tenant_id, &tenant_config_path, new_tenant_conf)
+    // This is a legacy API that only operates on attached tenants: the preferred
+    // API to use is the location_config/ endpoint, which lets the caller provide
+    // the full LocationConf.
+    let location_conf = LocationConf::attached_single(new_tenant_conf, tenant.generation);
+
+    Tenant::persist_tenant_config(conf, &tenant_id, &location_conf)
        .await
        .map_err(SetNewTenantConfigError::Persist)?;
    tenant.set_new_tenant_config(new_tenant_conf);
    Ok(())
 }

+#[instrument(skip_all, fields(tenant_id, new_location_config))]
+pub(crate) async fn upsert_location(
+    conf: &'static PageServerConf,
+    tenant_id: TenantId,
+    new_location_config: LocationConf,
+    broker_client: storage_broker::BrokerClientChannel,
+    remote_storage: Option<GenericRemoteStorage>,
+    deletion_queue_client: DeletionQueueClient,
+    ctx: &RequestContext,
+) -> Result<(), anyhow::Error> {
+    info!("configuring tenant location {tenant_id} to state {new_location_config:?}");
+
+    let mut existing_tenant = match get_tenant(tenant_id, false).await {
+        Ok(t) => Some(t),
+        Err(GetTenantError::NotFound(_)) => None,
+        Err(e) => anyhow::bail!(e),
+    };
+
+    // If we need to shut down a Tenant, do that first
+    let shutdown_tenant = match (&new_location_config.mode, &existing_tenant) {
+        (LocationMode::Secondary(_), Some(t)) => Some(t),
+        (LocationMode::Attached(attach_conf), Some(t)) => {
+            if attach_conf.generation != t.generation {
+                Some(t)
+            } else {
+                None
+            }
+        }
+        _ => None,
+    };
+
+    // TODO: currently we risk concurrent operations interfering with the tenant
+    // while we await shutdown, but we also should not hold the TenantsMap lock
+    // across the whole operation.  Before we start using this function in production,
+    // a follow-on change will revise how concurrency is handled in TenantsMap.
+    // (https://github.com/neondatabase/neon/issues/5378)
+
+    if let Some(tenant) = shutdown_tenant {
+        let (_guard, progress) = utils::completion::channel();
+        info!("Shutting down attached tenant");
+        match tenant.shutdown(progress, false).await {
+            Ok(()) => {}
+            Err(barrier) => {
+                info!("Shutdown already in progress, waiting for it to complete");
+                barrier.wait().await;
+            }
+        }
+        existing_tenant = None;
+    }
+
+    if let Some(tenant) = existing_tenant {
+        // Update the existing tenant
+        Tenant::persist_tenant_config(conf, &tenant_id, &new_location_config)
+            .await
+            .map_err(SetNewTenantConfigError::Persist)?;
+        tenant.set_new_location_config(AttachedTenantConf::try_from(new_location_config)?);
+    } else {
+        // Upsert a fresh TenantSlot into TenantsMap.  Do it within the map write lock,
+        // and re-check that the state of anything we are replacing is as expected.
+        tenant_map_upsert_slot(tenant_id, |old_value| async move {
+            if let Some(TenantSlot::Attached(t)) = old_value {
+                if !matches!(t.current_state(), TenantState::Stopping { .. }) {
+                    anyhow::bail!("Tenant state changed during location configuration update");
+                }
+            }
+
+            let new_slot = match &new_location_config.mode {
+                LocationMode::Secondary(_) => TenantSlot::Secondary,
+                LocationMode::Attached(_attach_config) => {
+                    // Do a schedule_local_tenant_processing
+                    // FIXME: should avoid doing this disk I/O inside the TenantsMap lock,
+                    // we have the same problem in load_tenant/attach_tenant.  Probably
+                    // need a lock in TenantSlot to fix this.
+                    Tenant::persist_tenant_config(conf, &tenant_id, &new_location_config)
+                        .await
+                        .map_err(SetNewTenantConfigError::Persist)?;
+                    let tenant_path = conf.tenant_path(&tenant_id);
+                    let resources = TenantSharedResources {
+                        broker_client,
+                        remote_storage,
+                        deletion_queue_client,
+                    };
+                    let new_tenant = schedule_local_tenant_processing(
+                        conf,
+                        tenant_id,
+                        &tenant_path,
+                        AttachedTenantConf::try_from(new_location_config)?,
+                        resources,
+                        None,
+                        &TENANTS,
+                        ctx,
+                    )
+                    .with_context(|| {
+                        format!("Failed to schedule tenant processing in path {tenant_path:?}")
+                    })?;
+
+                    TenantSlot::Attached(new_tenant)
+                }
+            };
+
+            Ok(new_slot)
+        })
+        .await?;
+    }
+
+    Ok(())
+}
+
 #[derive(Debug, thiserror::Error)]
 pub enum GetTenantError {
    #[error("Tenant {0} not found")]
    NotFound(TenantId),
    #[error("Tenant {0} is not active")]
    NotActive(TenantId),
+    /// Broken is logically a subset of NotActive, but a distinct error is useful as
+    /// NotActive is usually a retryable state for API purposes, whereas Broken
+    /// is a stuck error state
+    #[error("Tenant is broken: {0}")]
+    Broken(String),
 }

 /// Gets the tenant from the in-memory data, erroring if it's absent or is not fitting to the query.
@@ -513,10 +747,20 @@ pub async fn get_tenant(
    let tenant = m
        .get(&tenant_id)
        .ok_or(GetTenantError::NotFound(tenant_id))?;
-    if active_only && !tenant.is_active() {
-        Err(GetTenantError::NotActive(tenant_id))
-    } else {
-        Ok(Arc::clone(tenant))
+
+    match tenant.current_state() {
+        TenantState::Broken {
+            reason,
+            backtrace: _,
+        } if active_only => Err(GetTenantError::Broken(reason)),
+        TenantState::Active => Ok(Arc::clone(tenant)),
+        _ => {
+            if active_only {
+                Err(GetTenantError::NotActive(tenant_id))
+            } else {
+                Ok(Arc::clone(tenant))
+            }
+        }
    }
 }

@@ -589,7 +833,7 @@ async fn detach_tenant0(
    tenants: &tokio::sync::RwLock<TenantsMap>,
    tenant_id: TenantId,
    detach_ignored: bool,
-) -> Result<PathBuf, TenantStateError> {
+) -> Result<Utf8PathBuf, TenantStateError> {
    let tenant_dir_rename_operation = |tenant_id_to_clean| async move {
        let local_tenant_directory = conf.tenant_path(&tenant_id_to_clean);
        safe_rename_tenant_dir(&local_tenant_directory)
@@ -622,6 +866,7 @@ pub async fn load_tenant(
    generation: Generation,
    broker_client: storage_broker::BrokerClientChannel,
    remote_storage: Option<GenericRemoteStorage>,
+    deletion_queue_client: DeletionQueueClient,
    ctx: &RequestContext,
 ) -> Result<(), TenantMapInsertError> {
    tenant_map_insert(tenant_id, || async {
@@ -635,8 +880,14 @@ pub async fn load_tenant(
        let resources = TenantSharedResources {
            broker_client,
            remote_storage,
+            deletion_queue_client
        };
-        let new_tenant = schedule_local_tenant_processing(conf, tenant_id, &tenant_path, generation, resources, None,  &TENANTS, ctx)
+
+        let mut location_conf = Tenant::load_tenant_config(conf, &tenant_id).map_err( TenantMapInsertError::Other)?;
+        location_conf.attach_in_generation(generation);
+        Tenant::persist_tenant_config(conf, &tenant_id, &location_conf).await?;
+
+        let new_tenant = schedule_local_tenant_processing(conf, tenant_id, &tenant_path, AttachedTenantConf::try_from(location_conf)?, resources, None,  &TENANTS, ctx)
            .with_context(|| {
                format!("Failed to schedule tenant processing in path {tenant_path:?}")
            })?;
@@ -689,7 +940,10 @@ pub async fn list_tenants() -> Result<Vec<(TenantId, TenantState)>, TenantMapLis
        TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => m,
    };
    Ok(m.iter()
-        .map(|(id, tenant)| (*id, tenant.current_state()))
+        .filter_map(|(id, tenant)| match tenant {
+            TenantSlot::Attached(tenant) => Some((*id, tenant.current_state())),
+            TenantSlot::Secondary => None,
+        })
        .collect())
 }

@@ -702,12 +956,12 @@ pub async fn attach_tenant(
    tenant_id: TenantId,
    generation: Generation,
    tenant_conf: TenantConfOpt,
-    broker_client: storage_broker::BrokerClientChannel,
-    remote_storage: GenericRemoteStorage,
+    resources: TenantSharedResources,
    ctx: &RequestContext,
 ) -> Result<(), TenantMapInsertError> {
    tenant_map_insert(tenant_id, || async {
-        let tenant_dir = create_tenant_files(conf, tenant_conf, &tenant_id, CreateTenantFilesMode::Attach).await?;
+        let location_conf = LocationConf::attached_single(tenant_conf, generation);
+        let tenant_dir = create_tenant_files(conf, &location_conf, &tenant_id, CreateTenantFilesMode::Attach).await?;
        // TODO: tenant directory remains on disk if we bail out from here on.
        //       See https://github.com/neondatabase/neon/issues/4233

@@ -718,11 +972,7 @@ pub async fn attach_tenant(
            .context("check for attach marker file existence")?;
        anyhow::ensure!(marker_file_exists, "create_tenant_files should have created the attach marker file");

-        let resources = TenantSharedResources {
-            broker_client,
-            remote_storage: Some(remote_storage),
-        };
-        let attached_tenant = schedule_local_tenant_processing(conf, tenant_id, &tenant_dir, generation, resources, None, &TENANTS, ctx)?;
+        let attached_tenant = schedule_local_tenant_processing(conf, tenant_id, &tenant_dir, AttachedTenantConf::try_from(location_conf)?, resources, None, &TENANTS, ctx)?;
        // TODO: tenant object & its background loops remain, untracked in tenant map, if we fail here.
        //      See https://github.com/neondatabase/neon/issues/4233

@@ -745,8 +995,10 @@ pub enum TenantMapInsertError {
    ShuttingDown,
    #[error("tenant {0} already exists, state: {1:?}")]
    TenantAlreadyExists(TenantId, TenantState),
+    #[error("tenant {0} already exists in secondary state")]
+    TenantExistsSecondary(TenantId),
    #[error(transparent)]
-    Closure(#[from] anyhow::Error),
+    Other(#[from] anyhow::Error),
 }

 /// Give the given closure access to the tenants map entry for the given `tenant_id`, iff that
@@ -770,20 +1022,47 @@ where
        TenantsMap::Open(m) => m,
    };
    match m.entry(tenant_id) {
-        hash_map::Entry::Occupied(e) => Err(TenantMapInsertError::TenantAlreadyExists(
-            tenant_id,
-            e.get().current_state(),
-        )),
+        hash_map::Entry::Occupied(e) => match e.get() {
+            TenantSlot::Attached(t) => Err(TenantMapInsertError::TenantAlreadyExists(
+                tenant_id,
+                t.current_state(),
+            )),
+            TenantSlot::Secondary => Err(TenantMapInsertError::TenantExistsSecondary(tenant_id)),
+        },
        hash_map::Entry::Vacant(v) => match insert_fn().await {
            Ok(tenant) => {
-                v.insert(tenant.clone());
+                v.insert(TenantSlot::Attached(tenant.clone()));
                Ok(tenant)
            }
-            Err(e) => Err(TenantMapInsertError::Closure(e)),
+            Err(e) => Err(TenantMapInsertError::Other(e)),
        },
    }
 }

+async fn tenant_map_upsert_slot<'a, F, R>(
+    tenant_id: TenantId,
+    upsert_fn: F,
+) -> Result<(), TenantMapInsertError>
+where
+    F: FnOnce(Option<TenantSlot>) -> R,
+    R: std::future::Future<Output = anyhow::Result<TenantSlot>>,
+{
+    let mut guard = TENANTS.write().await;
+    let m = match &mut *guard {
+        TenantsMap::Initializing => return Err(TenantMapInsertError::StillInitializing),
+        TenantsMap::ShuttingDown(_) => return Err(TenantMapInsertError::ShuttingDown),
+        TenantsMap::Open(m) => m,
+    };
+
+    match upsert_fn(m.remove(&tenant_id)).await {
+        Ok(upsert_val) => {
+            m.insert(tenant_id, upsert_val);
+            Ok(())
+        }
+        Err(e) => Err(TenantMapInsertError::Other(e)),
+    }
+}
+
 /// Stops and removes the tenant from memory, if it's not [`TenantState::Stopping`] already, bails otherwise.
 /// Allows to remove other tenant resources manually, via `tenant_cleanup`.
 /// If the cleanup fails, tenant will stay in memory in [`TenantState::Broken`] state, and another removal
@@ -803,28 +1082,40 @@ where
    // tenant-wde cleanup operations may take some time (removing the entire tenant directory), we want to
    // avoid holding the lock for the entire process.
    let tenant = {
-        tenants
+        match tenants
            .write()
            .await
-            .get(&tenant_id)
-            .cloned()
+            .get_slot(&tenant_id)
            .ok_or(TenantStateError::NotFound(tenant_id))?
+        {
+            TenantSlot::Attached(t) => Some(t.clone()),
+            TenantSlot::Secondary => None,
+        }
    };

    // allow pageserver shutdown to await for our completion
    let (_guard, progress) = completion::channel();

-    // whenever we remove a tenant from memory, we don't want to flush and wait for upload
-    let freeze_and_flush = false;
+    // If the tenant was attached, shut it down gracefully.  For secondary
+    // locations this part is not necessary
+    match tenant {
+        Some(attached_tenant) => {
+            // whenever we remove a tenant from memory, we don't want to flush and wait for upload
+            let freeze_and_flush = false;

-    // shutdown is sure to transition tenant to stopping, and wait for all tasks to complete, so
-    // that we can continue safely to cleanup.
-    match tenant.shutdown(progress, freeze_and_flush).await {
-        Ok(()) => {}
-        Err(_other) => {
-            // if pageserver shutdown or other detach/ignore is already ongoing, we don't want to
-            // wait for it but return an error right away because these are distinct requests.
-            return Err(TenantStateError::IsStopping(tenant_id));
+            // shutdown is sure to transition tenant to stopping, and wait for all tasks to complete, so
+            // that we can continue safely to cleanup.
+            match attached_tenant.shutdown(progress, freeze_and_flush).await {
+                Ok(()) => {}
+                Err(_other) => {
+                    // if pageserver shutdown or other detach/ignore is already ongoing, we don't want to
+                    // wait for it but return an error right away because these are distinct requests.
+                    return Err(TenantStateError::IsStopping(tenant_id));
+                }
+            }
+        }
+        None => {
+            // Nothing to wait on when not attached, proceed.
        }
    }

@@ -915,6 +1206,8 @@ mod tests {
    use std::sync::Arc;
    use tracing::{info_span, Instrument};

+    use crate::tenant::mgr::TenantSlot;
+
    use super::{super::harness::TenantHarness, TenantsMap};

    #[tokio::test(start_paused = true)]
@@ -936,7 +1229,7 @@ mod tests {
        // tenant harness configures the logging and we cannot escape it
        let _e = info_span!("testing", tenant_id = %id).entered();

-        let tenants = HashMap::from([(id, t.clone())]);
+        let tenants = HashMap::from([(id, TenantSlot::Attached(t.clone()))]);
        let tenants = Arc::new(tokio::sync::RwLock::new(TenantsMap::Open(tenants)));

        let (until_cleanup_completed, can_complete_cleanup) = utils::completion::channel();
--- a/pageserver/src/tenant/par_fsync.rs
+++ b/pageserver/src/tenant/par_fsync.rs
@@ -1,16 +1,17 @@
 use std::{
    io,
-    path::{Path, PathBuf},
    sync::atomic::{AtomicUsize, Ordering},
 };

-fn fsync_path(path: &Path) -> io::Result<()> {
+use camino::{Utf8Path, Utf8PathBuf};
+
+fn fsync_path(path: &Utf8Path) -> io::Result<()> {
    // TODO use VirtualFile::fsync_all once we fully go async.
    let file = std::fs::File::open(path)?;
    file.sync_all()
 }

-fn parallel_worker(paths: &[PathBuf], next_path_idx: &AtomicUsize) -> io::Result<()> {
+fn parallel_worker(paths: &[Utf8PathBuf], next_path_idx: &AtomicUsize) -> io::Result<()> {
    while let Some(path) = paths.get(next_path_idx.fetch_add(1, Ordering::Relaxed)) {
        fsync_path(path)?;
    }
@@ -18,7 +19,7 @@ fn parallel_worker(paths: &[PathBuf], next_path_idx: &AtomicUsize) -> io::Result
    Ok(())
 }

-fn fsync_in_thread_pool(paths: &[PathBuf]) -> io::Result<()> {
+fn fsync_in_thread_pool(paths: &[Utf8PathBuf]) -> io::Result<()> {
    // TODO: remove this function in favor of `par_fsync_async` once we asyncify everything.

    /// Use at most this number of threads.
@@ -47,7 +48,7 @@ fn fsync_in_thread_pool(paths: &[PathBuf]) -> io::Result<()> {
 }

 /// Parallel fsync all files. Can be used in non-async context as it is using rayon thread pool.
-pub fn par_fsync(paths: &[PathBuf]) -> io::Result<()> {
+pub fn par_fsync(paths: &[Utf8PathBuf]) -> io::Result<()> {
    if paths.len() == 1 {
        fsync_path(&paths[0])?;
        return Ok(());
@@ -58,7 +59,7 @@ pub fn par_fsync(paths: &[PathBuf]) -> io::Result<()> {

 /// Parallel fsync asynchronously. If number of files are less than PARALLEL_PATH_THRESHOLD, fsync is done in the current
 /// execution thread. Otherwise, we will spawn_blocking and run it in tokio.
-pub async fn par_fsync_async(paths: &[PathBuf]) -> io::Result<()> {
+pub async fn par_fsync_async(paths: &[Utf8PathBuf]) -> io::Result<()> {
    const MAX_CONCURRENT_FSYNC: usize = 64;
    let mut next = paths.iter().peekable();
    let mut js = tokio::task::JoinSet::new();
--- a/pageserver/src/tenant/remote_timeline_client.rs
+++ b/pageserver/src/tenant/remote_timeline_client.rs
@@ -116,8 +116,12 @@
 //! # Completion
 //!
 //! Once an operation has completed, we update
-//! [`UploadQueueInitialized::last_uploaded_consistent_lsn`] which indicates
-//! to safekeepers that they can delete the WAL up to that LSN.
+//! [`UploadQueueInitialized::projected_remote_consistent_lsn`] immediately,
+//! and submit a request through the DeletionQueue to update
+//! [`UploadQueueInitialized::visible_remote_consistent_lsn`] after it has
+//! validated that our generation is not stale.  It is this visible value
+//! that is advertized to safekeepers as a signal that that they can
+//! delete the WAL up to that LSN.
 //!
 //! The [`RemoteTimelineClient::wait_completion`] method can be used to wait
 //! for all pending operations to complete. It does not prevent more
@@ -200,12 +204,12 @@
 //! [`Tenant::timeline_init_and_sync`]: super::Tenant::timeline_init_and_sync
 //! [`Timeline::load_layer_map`]: super::Timeline::load_layer_map

-mod delete;
 mod download;
 pub mod index;
 mod upload;

 use anyhow::Context;
+use camino::Utf8Path;
 use chrono::{NaiveDateTime, Utc};
 // re-export these
 pub use download::{is_temp_download_file, list_remote_timelines};
@@ -216,7 +220,6 @@ use utils::backoff::{
 };

 use std::collections::{HashMap, VecDeque};
-use std::path::{Path, PathBuf};
 use std::sync::atomic::{AtomicU32, Ordering};
 use std::sync::{Arc, Mutex};

@@ -226,6 +229,7 @@ use tracing::{debug, error, info, instrument, warn};
 use tracing::{info_span, Instrument};
 use utils::lsn::Lsn;

+use crate::deletion_queue::DeletionQueueClient;
 use crate::metrics::{
    MeasureRemoteOp, RemoteOpFileKind, RemoteOpKind, RemoteTimelineClientMetrics,
    RemoteTimelineClientMetricsCallTrackSize, REMOTE_ONDEMAND_DOWNLOADED_BYTES,
@@ -324,6 +328,8 @@ pub struct RemoteTimelineClient {
    metrics: Arc<RemoteTimelineClientMetrics>,

    storage_impl: GenericRemoteStorage,
+
+    deletion_queue_client: DeletionQueueClient,
 }

 impl RemoteTimelineClient {
@@ -335,6 +341,7 @@ impl RemoteTimelineClient {
    ///
    pub fn new(
        remote_storage: GenericRemoteStorage,
+        deletion_queue_client: DeletionQueueClient,
        conf: &'static PageServerConf,
        tenant_id: TenantId,
        timeline_id: TimelineId,
@@ -352,6 +359,7 @@ impl RemoteTimelineClient {
            timeline_id,
            generation,
            storage_impl: remote_storage,
+            deletion_queue_client,
            upload_queue: Mutex::new(UploadQueue::Uninitialized),
            metrics: Arc::new(RemoteTimelineClientMetrics::new(&tenant_id, &timeline_id)),
        }
@@ -413,13 +421,24 @@ impl RemoteTimelineClient {
        Ok(())
    }

-    pub fn last_uploaded_consistent_lsn(&self) -> Option<Lsn> {
-        match &*self.upload_queue.lock().unwrap() {
+    pub fn remote_consistent_lsn_projected(&self) -> Option<Lsn> {
+        match &mut *self.upload_queue.lock().unwrap() {
            UploadQueue::Uninitialized => None,
-            UploadQueue::Initialized(q) => Some(q.last_uploaded_consistent_lsn),
-            UploadQueue::Stopped(q) => {
-                Some(q.upload_queue_for_deletion.last_uploaded_consistent_lsn)
-            }
+            UploadQueue::Initialized(q) => q.get_last_remote_consistent_lsn_projected(),
+            UploadQueue::Stopped(q) => q
+                .upload_queue_for_deletion
+                .get_last_remote_consistent_lsn_projected(),
+        }
+    }
+
+    pub fn remote_consistent_lsn_visible(&self) -> Option<Lsn> {
+        match &mut *self.upload_queue.lock().unwrap() {
+            UploadQueue::Uninitialized => None,
+            UploadQueue::Initialized(q) => Some(q.get_last_remote_consistent_lsn_visible()),
+            UploadQueue::Stopped(q) => Some(
+                q.upload_queue_for_deletion
+                    .get_last_remote_consistent_lsn_visible(),
+            ),
        }
    }

@@ -434,11 +453,11 @@ impl RemoteTimelineClient {
        } else {
            0
        };
-        self.metrics.remote_physical_size_gauge().set(size);
+        self.metrics.remote_physical_size_set(size);
    }

    pub fn get_remote_physical_size(&self) -> u64 {
-        self.metrics.remote_physical_size_gauge().get()
+        self.metrics.remote_physical_size_get()
    }

    //
@@ -643,7 +662,7 @@ impl RemoteTimelineClient {
    /// successfully.
    pub fn schedule_layer_file_deletion(
        self: &Arc<Self>,
-        names: &[LayerFileName],
+        names: Vec<LayerFileName>,
    ) -> anyhow::Result<()> {
        let mut guard = self.upload_queue.lock().unwrap();
        let upload_queue = guard.initialized_mut()?;
@@ -663,10 +682,10 @@ impl RemoteTimelineClient {
            // Decorate our list of names with each name's generation, dropping
            // makes that are unexpectedly missing from our metadata.
            let with_generations: Vec<_> = names
-                .iter()
+                .into_iter()
                .filter_map(|name| {
                    // Remove from latest_files, learning the file's remote generation in the process
-                    let meta = upload_queue.latest_files.remove(name);
+                    let meta = upload_queue.latest_files.remove(&name);

                    if let Some(meta) = meta {
                        upload_queue.latest_files_changes_since_metadata_upload_scheduled += 1;
@@ -688,19 +707,17 @@ impl RemoteTimelineClient {
                self.schedule_index_upload(upload_queue, metadata);
            }

-            // schedule the actual deletions
-            for (name, generation) in with_generations {
-                let op = UploadOp::Delete(Delete {
-                    file_kind: RemoteOpFileKind::Layer,
-                    layer_file_name: name.clone(),
-                    scheduled_from_timeline_delete: false,
-                    generation,
-                });
-                self.calls_unfinished_metric_begin(&op);
-                upload_queue.queued_operations.push_back(op);
-                info!("scheduled layer file deletion {name}");
+            for (name, gen) in &with_generations {
+                info!("scheduling deletion of layer {}{}", name, gen.get_suffix());
            }

+            // schedule the actual deletions
+            let op = UploadOp::Delete(Delete {
+                layers: with_generations,
+            });
+            self.calls_unfinished_metric_begin(&op);
+            upload_queue.queued_operations.push_back(op);
+
            // Launch the tasks immediately, if possible
            self.launch_queued_tasks(upload_queue);
        };
@@ -833,9 +850,7 @@ impl RemoteTimelineClient {
    pub(crate) async fn delete_all(self: &Arc<Self>) -> anyhow::Result<()> {
        debug_assert_current_span_has_tenant_and_timeline_id();

-        let (mut receiver, deletions_queued) = {
-            let mut deletions_queued = 0;
-
+        let layers: Vec<RemotePath> = {
            let mut locked = self.upload_queue.lock().unwrap();
            let stopped = locked.stopped_mut()?;

@@ -847,42 +862,30 @@ impl RemoteTimelineClient {

            stopped
                .upload_queue_for_deletion
-                .queued_operations
-                .reserve(stopped.upload_queue_for_deletion.latest_files.len());
-
-            // schedule the actual deletions
-            for (name, meta) in &stopped.upload_queue_for_deletion.latest_files {
-                let op = UploadOp::Delete(Delete {
-                    file_kind: RemoteOpFileKind::Layer,
-                    layer_file_name: name.clone(),
-                    scheduled_from_timeline_delete: true,
-                    generation: meta.generation,
-                });
-
-                self.calls_unfinished_metric_begin(&op);
-                stopped
-                    .upload_queue_for_deletion
-                    .queued_operations
-                    .push_back(op);
-
-                info!("scheduled layer file deletion {name}");
-                deletions_queued += 1;
-            }
-
-            self.launch_queued_tasks(&mut stopped.upload_queue_for_deletion);
-
-            (
-                self.schedule_barrier(&mut stopped.upload_queue_for_deletion),
-                deletions_queued,
-            )
+                .latest_files
+                .drain()
+                .map(|(file_name, meta)| {
+                    remote_layer_path(
+                        &self.tenant_id,
+                        &self.timeline_id,
+                        &file_name,
+                        meta.generation,
+                    )
+                })
+                .collect()
        };

-        receiver.changed().await.context("upload queue shut down")?;
+        let layer_deletion_count = layers.len();
+        self.deletion_queue_client.push_immediate(layers).await?;

        // Do not delete index part yet, it is needed for possible retry. If we remove it first
        // and retry will arrive to different pageserver there wont be any traces of it on remote storage
        let timeline_storage_path = remote_timeline_path(&self.tenant_id, &self.timeline_id);

+        // Execute all pending deletions, so that when we proceed to do a list_prefixes below, we aren't
+        // taking the burden of listing all the layers that we already know we should delete.
+        self.deletion_queue_client.flush_immediate().await?;
+
        let remaining = backoff::retry(
            || async {
                self.storage_impl
@@ -910,17 +913,9 @@ impl RemoteTimelineClient {
            })
            .collect();

+        let not_referenced_count = remaining.len();
        if !remaining.is_empty() {
-            backoff::retry(
-                || async { self.storage_impl.delete_objects(&remaining).await },
-                |_e| false,
-                FAILED_UPLOAD_WARN_THRESHOLD,
-                FAILED_REMOTE_OP_RETRIES,
-                "delete_objects",
-                backoff::Cancel::new(shutdown_token(), || anyhow::anyhow!("Cancelled!")),
-            )
-            .await
-            .context("delete_objects")?;
+            self.deletion_queue_client.push_immediate(remaining).await?;
        }

        fail::fail_point!("timeline-delete-before-index-delete", |_| {
@@ -929,20 +924,16 @@ impl RemoteTimelineClient {
            ))?
        });

-        let index_file_path = timeline_storage_path.join(Path::new(IndexPart::FILE_NAME));
+        let index_file_path = timeline_storage_path.join(Utf8Path::new(IndexPart::FILE_NAME));

-        debug!("deleting index part");
+        debug!("enqueuing index part deletion");
+        self.deletion_queue_client
+            .push_immediate([index_file_path].to_vec())
+            .await?;

-        backoff::retry(
-            || async { self.storage_impl.delete(&index_file_path).await },
-            |_e| false,
-            FAILED_UPLOAD_WARN_THRESHOLD,
-            FAILED_REMOTE_OP_RETRIES,
-            "delete_index",
-            backoff::Cancel::new(shutdown_token(), || anyhow::anyhow!("Cancelled")),
-        )
-        .await
-        .context("delete_index")?;
+        // Timeline deletion is rare and we have probably emitted a reasonably number of objects: wait
+        // for a flush to a persistent deletion list so that we may be sure deletion will occur.
+        self.deletion_queue_client.flush_immediate().await?;

        fail::fail_point!("timeline-delete-after-index-delete", |_| {
            Err(anyhow::anyhow!(
@@ -950,7 +941,7 @@ impl RemoteTimelineClient {
            ))?
        });

-        info!(prefix=%timeline_storage_path, referenced=deletions_queued, not_referenced=%remaining.len(), "done deleting in timeline prefix, including index_part.json");
+        info!(prefix=%timeline_storage_path, referenced=layer_deletion_count, not_referenced=%not_referenced_count, "done deleting in timeline prefix, including index_part.json");

        Ok(())
    }
@@ -1140,21 +1131,16 @@ impl RemoteTimelineClient {
                    }
                    res
                }
-                UploadOp::Delete(delete) => {
-                    let path = &self
-                        .conf
-                        .timeline_path(&self.tenant_id, &self.timeline_id)
-                        .join(delete.layer_file_name.file_name());
-                    delete::delete_layer(self.conf, &self.storage_impl, path, delete.generation)
-                        .measure_remote_op(
-                            self.tenant_id,
-                            self.timeline_id,
-                            delete.file_kind,
-                            RemoteOpKind::Delete,
-                            Arc::clone(&self.metrics),
-                        )
-                        .await
-                }
+                UploadOp::Delete(delete) => self
+                    .deletion_queue_client
+                    .push_layers(
+                        self.tenant_id,
+                        self.timeline_id,
+                        self.generation,
+                        delete.layers.clone(),
+                    )
+                    .await
+                    .map_err(|e| anyhow::anyhow!(e)),
                UploadOp::Barrier(_) => {
                    // unreachable. Barrier operations are handled synchronously in
                    // launch_queued_tasks
@@ -1210,18 +1196,12 @@ impl RemoteTimelineClient {
        }

        // The task has completed successfully. Remove it from the in-progress list.
-        {
+        let lsn_update = {
            let mut upload_queue_guard = self.upload_queue.lock().unwrap();
            let upload_queue = match upload_queue_guard.deref_mut() {
                UploadQueue::Uninitialized => panic!("callers are responsible for ensuring this is only called on an initialized queue"),
-                UploadQueue::Stopped(stopped) => {
-                    // Special care is needed for deletions, if it was an earlier deletion (not scheduled from deletion)
-                    // then stop() took care of it so we just return.
-                    // For deletions that come from delete_all we still want to maintain metrics, launch following tasks, etc.
-                    match &task.op {
-                        UploadOp::Delete(delete) if delete.scheduled_from_timeline_delete => Some(&mut stopped.upload_queue_for_deletion),
-                        _ => None
-                    }
+                UploadQueue::Stopped(_stopped) => {
+                    None
                },
                UploadQueue::Initialized(qi) => { Some(qi) }
            };
@@ -1236,23 +1216,51 @@ impl RemoteTimelineClient {

            upload_queue.inprogress_tasks.remove(&task.task_id);

-            match task.op {
+            let lsn_update = match task.op {
                UploadOp::UploadLayer(_, _) => {
                    upload_queue.num_inprogress_layer_uploads -= 1;
+                    None
                }
                UploadOp::UploadMetadata(_, lsn) => {
                    upload_queue.num_inprogress_metadata_uploads -= 1;
-                    upload_queue.last_uploaded_consistent_lsn = lsn; // XXX monotonicity check?
+                    // XXX monotonicity check?
+
+                    upload_queue.projected_remote_consistent_lsn = Some(lsn);
+                    if self.generation.is_none() {
+                        // Legacy mode: skip validating generation
+                        upload_queue.visible_remote_consistent_lsn.store(lsn);
+                        None
+                    } else {
+                        Some((lsn, upload_queue.visible_remote_consistent_lsn.clone()))
+                    }
                }
                UploadOp::Delete(_) => {
                    upload_queue.num_inprogress_deletions -= 1;
+                    None
                }
                UploadOp::Barrier(_) => unreachable!(),
            };

            // Launch any queued tasks that were unblocked by this one.
            self.launch_queued_tasks(upload_queue);
+            lsn_update
+        };
+
+        if let Some((lsn, slot)) = lsn_update {
+            // Updates to the remote_consistent_lsn we advertise to pageservers
+            // are all routed through the DeletionQueue, to enforce important
+            // data safety guarantees (see docs/rfcs/025-generation-numbers.md)
+            self.deletion_queue_client
+                .update_remote_consistent_lsn(
+                    self.tenant_id,
+                    self.timeline_id,
+                    self.generation,
+                    lsn,
+                    slot,
+                )
+                .await;
        }
+
        self.calls_unfinished_metric_end(&task.op);
    }

@@ -1278,8 +1286,8 @@ impl RemoteTimelineClient {
                    reason: "metadata uploads are tiny",
                },
            ),
-            UploadOp::Delete(delete) => (
-                delete.file_kind,
+            UploadOp::Delete(_delete) => (
+                RemoteOpFileKind::Layer,
                RemoteOpKind::Delete,
                DontTrackSize {
                    reason: "should we track deletes? positive or negative sign?",
@@ -1341,7 +1349,10 @@ impl RemoteTimelineClient {
                        latest_files: initialized.latest_files.clone(),
                        latest_files_changes_since_metadata_upload_scheduled: 0,
                        latest_metadata: initialized.latest_metadata.clone(),
-                        last_uploaded_consistent_lsn: initialized.last_uploaded_consistent_lsn,
+                        projected_remote_consistent_lsn: None,
+                        visible_remote_consistent_lsn: initialized
+                            .visible_remote_consistent_lsn
+                            .clone(),
                        num_inprogress_layer_uploads: 0,
                        num_inprogress_metadata_uploads: 0,
                        num_inprogress_deletions: 0,
@@ -1398,20 +1409,20 @@ pub fn remote_timelines_path(tenant_id: &TenantId) -> RemotePath {
 }

 pub fn remote_timeline_path(tenant_id: &TenantId, timeline_id: &TimelineId) -> RemotePath {
-    remote_timelines_path(tenant_id).join(&PathBuf::from(timeline_id.to_string()))
+    remote_timelines_path(tenant_id).join(Utf8Path::new(&timeline_id.to_string()))
 }

 pub fn remote_layer_path(
    tenant_id: &TenantId,
    timeline_id: &TimelineId,
    layer_file_name: &LayerFileName,
-    layer_meta: &LayerFileMetadata,
+    generation: Generation,
 ) -> RemotePath {
    // Generation-aware key format
    let path = format!(
        "tenants/{tenant_id}/{TIMELINES_SEGMENT_NAME}/{timeline_id}/{0}{1}",
        layer_file_name.file_name(),
-        layer_meta.generation.get_suffix()
+        generation.get_suffix()
    );

    RemotePath::from_string(&path).expect("Failed to construct path")
@@ -1441,14 +1452,7 @@ pub(crate) fn parse_remote_index_path(path: RemotePath) -> Option<Generation> {
        }
    };

-    let file_name_str = match file_name.to_str() {
-        Some(s) => s,
-        None => {
-            tracing::warn!("Malformed index key {:?}", path);
-            return None;
-        }
-    };
-    match file_name_str.split_once('-') {
+    match file_name.split_once('-') {
        Some((_, gen_suffix)) => Generation::parse_suffix(gen_suffix),
        None => None,
    }
@@ -1460,20 +1464,16 @@ pub(crate) fn parse_remote_index_path(path: RemotePath) -> Option<Generation> {
 /// Errors if the path provided does not start from pageserver's workdir.
 pub fn remote_path(
    conf: &PageServerConf,
-    local_path: &Path,
+    local_path: &Utf8Path,
    generation: Generation,
 ) -> anyhow::Result<RemotePath> {
    let stripped = local_path
        .strip_prefix(&conf.workdir)
        .context("Failed to strip workdir prefix")?;

-    let suffixed = format!(
-        "{0}{1}",
-        stripped.to_string_lossy(),
-        generation.get_suffix()
-    );
+    let suffixed = format!("{0}{1}", stripped, generation.get_suffix());

-    RemotePath::new(&PathBuf::from(suffixed)).with_context(|| {
+    RemotePath::new(Utf8Path::new(&suffixed)).with_context(|| {
        format!(
            "to resolve remote part of path {:?} for base {:?}",
            local_path, conf.workdir
@@ -1493,7 +1493,7 @@ mod tests {
        DEFAULT_PG_VERSION,
    };

-    use std::{collections::HashSet, path::Path};
+    use std::collections::HashSet;
    use utils::lsn::Lsn;

    pub(super) fn dummy_contents(name: &str) -> Vec<u8> {
@@ -1527,7 +1527,7 @@ mod tests {
        assert_eq!(avec, bvec);
    }

-    fn assert_remote_files(expected: &[&str], remote_path: &Path, generation: Generation) {
+    fn assert_remote_files(expected: &[&str], remote_path: &Utf8Path, generation: Generation) {
        let mut expected: Vec<String> = expected
            .iter()
            .map(|x| format!("{}{}", x, generation.get_suffix()))
@@ -1554,7 +1554,6 @@ mod tests {

    impl TestSetup {
        async fn new(test_name: &str) -> anyhow::Result<Self> {
-            // Use a current-thread runtime in the test
            let test_name = Box::leak(Box::new(format!("remote_timeline_client__{test_name}")));
            let harness = TenantHarness::create(test_name)?;
            let (tenant, ctx) = harness.load().await;
@@ -1580,6 +1579,7 @@ mod tests {
                timeline_id: TIMELINE_ID,
                generation,
                storage_impl: self.harness.remote_storage.clone(),
+                deletion_queue_client: self.harness.deletion_queue.new_client(),
                upload_queue: Mutex::new(UploadQueue::Uninitialized),
                metrics: Arc::new(RemoteTimelineClientMetrics::new(
                    &self.harness.tenant_id,
@@ -1646,12 +1646,12 @@ mod tests {

        let timeline_path = harness.timeline_path(&TIMELINE_ID);

-        println!("workdir: {}", harness.conf.workdir.display());
+        println!("workdir: {}", harness.conf.workdir);

        let remote_timeline_dir = harness
            .remote_fs_dir
            .join(timeline_path.strip_prefix(&harness.conf.workdir).unwrap());
-        println!("remote_timeline_dir: {}", remote_timeline_dir.display());
+        println!("remote_timeline_dir: {remote_timeline_dir}");

        let generation = harness.generation;

@@ -1749,7 +1749,7 @@ mod tests {
            )
            .unwrap();
        client
-            .schedule_layer_file_deletion(&[layer_file_name_1.clone()])
+            .schedule_layer_file_deletion([layer_file_name_1.clone()].to_vec())
            .unwrap();
        {
            let mut guard = client.upload_queue.lock().unwrap();
@@ -1775,6 +1775,7 @@ mod tests {

        // Finish them
        client.wait_completion().await.unwrap();
+        harness.deletion_queue.pump().await;

        assert_remote_files(
            &[
@@ -1897,7 +1898,7 @@ mod tests {
        let index_path = test_state.harness.remote_fs_dir.join(
            remote_index_path(&test_state.harness.tenant_id, &TIMELINE_ID, generation).get_path(),
        );
-        eprintln!("Writing {}", index_path.display());
+        eprintln!("Writing {index_path}");
        std::fs::write(&index_path, index_part_bytes).unwrap();
        example_index_part
    }
--- a/pageserver/src/tenant/remote_timeline_client/delete.rs
+++ b/pageserver/src/tenant/remote_timeline_client/delete.rs
@@ -1,34 +0,0 @@
-//! Helper functions to delete files from remote storage with a RemoteStorage
-use anyhow::Context;
-use std::path::Path;
-use tracing::debug;
-
-use remote_storage::GenericRemoteStorage;
-
-use crate::{
-    config::PageServerConf,
-    tenant::{remote_timeline_client::remote_path, Generation},
-};
-
-pub(super) async fn delete_layer<'a>(
-    conf: &'static PageServerConf,
-    storage: &'a GenericRemoteStorage,
-    local_layer_path: &'a Path,
-    generation: Generation,
-) -> anyhow::Result<()> {
-    fail::fail_point!("before-delete-layer", |_| {
-        anyhow::bail!("failpoint before-delete-layer")
-    });
-    debug!("Deleting layer from remote storage: {local_layer_path:?}",);
-
-    let path_to_delete = remote_path(conf, local_layer_path, generation)?;
-
-    // We don't want to print an error if the delete failed if the file has
-    // already been deleted. Thankfully, in this situation S3 already
-    // does not yield an error. While OS-provided local file system APIs do yield
-    // errors, we avoid them in the `LocalFs` wrapper.
-    storage
-        .delete(&path_to_delete)
-        .await
-        .with_context(|| format!("delete remote layer from storage at {path_to_delete:?}"))
-}
--- a/pageserver/src/tenant/remote_timeline_client/download.rs
+++ b/pageserver/src/tenant/remote_timeline_client/download.rs
@@ -5,10 +5,10 @@

 use std::collections::HashSet;
 use std::future::Future;
-use std::path::Path;
 use std::time::Duration;

 use anyhow::{anyhow, Context};
+use camino::Utf8Path;
 use tokio::fs;
 use tokio::io::AsyncWriteExt;
 use tokio_util::sync::CancellationToken;
@@ -50,7 +50,12 @@ pub async fn download_layer_file<'a>(
        .timeline_path(&tenant_id, &timeline_id)
        .join(layer_file_name.file_name());

-    let remote_path = remote_layer_path(&tenant_id, &timeline_id, layer_file_name, layer_metadata);
+    let remote_path = remote_layer_path(
+        &tenant_id,
+        &timeline_id,
+        layer_file_name,
+        layer_metadata.generation,
+    );

    // Perform a rename inspired by durable_rename from file_utils.c.
    // The sequence:
@@ -69,12 +74,7 @@ pub async fn download_layer_file<'a>(
            // TODO: this doesn't use the cached fd for some reason?
            let mut destination_file = fs::File::create(&temp_file_path)
                .await
-                .with_context(|| {
-                    format!(
-                        "create a destination file for layer '{}'",
-                        temp_file_path.display()
-                    )
-                })
+                .with_context(|| format!("create a destination file for layer '{temp_file_path}'"))
                .map_err(DownloadError::Other)?;
            let mut download = storage
                .download(&remote_path)
@@ -116,7 +116,7 @@ pub async fn download_layer_file<'a>(
    destination_file
        .flush()
        .await
-        .with_context(|| format!("flush source file at {}", temp_file_path.display()))
+        .with_context(|| format!("flush source file at {temp_file_path}"))
        .map_err(DownloadError::Other)?;

    let expected = layer_metadata.file_size();
@@ -130,12 +130,7 @@ pub async fn download_layer_file<'a>(
    destination_file
        .sync_all()
        .await
-        .with_context(|| {
-            format!(
-                "failed to fsync source file at {}",
-                temp_file_path.display()
-            )
-        })
+        .with_context(|| format!("failed to fsync source file at {temp_file_path}"))
        .map_err(DownloadError::Other)?;
    drop(destination_file);

@@ -147,27 +142,23 @@ pub async fn download_layer_file<'a>(

    fs::rename(&temp_file_path, &local_path)
        .await
-        .with_context(|| format!("rename download layer file to {}", local_path.display(),))
+        .with_context(|| format!("rename download layer file to {local_path}"))
        .map_err(DownloadError::Other)?;

    crashsafe::fsync_async(&local_path)
        .await
-        .with_context(|| format!("fsync layer file {}", local_path.display(),))
+        .with_context(|| format!("fsync layer file {local_path}"))
        .map_err(DownloadError::Other)?;

-    tracing::debug!("download complete: {}", local_path.display());
+    tracing::debug!("download complete: {local_path}");

    Ok(bytes_amount)
 }

 const TEMP_DOWNLOAD_EXTENSION: &str = "temp_download";

-pub fn is_temp_download_file(path: &Path) -> bool {
-    let extension = path.extension().map(|pname| {
-        pname
-            .to_str()
-            .expect("paths passed to this function must be valid Rust strings")
-    });
+pub fn is_temp_download_file(path: &Utf8Path) -> bool {
+    let extension = path.extension();
    match extension {
        Some(TEMP_DOWNLOAD_EXTENSION) => true,
        Some(_) => false,
--- a/pageserver/src/tenant/remote_timeline_client/upload.rs
+++ b/pageserver/src/tenant/remote_timeline_client/upload.rs
@@ -1,8 +1,9 @@
 //! Helper functions to upload files to remote storage with a RemoteStorage

 use anyhow::{bail, Context};
+use camino::Utf8Path;
 use fail::fail_point;
-use std::{io::ErrorKind, path::Path};
+use std::io::ErrorKind;
 use tokio::fs;

 use super::Generation;
@@ -50,7 +51,7 @@ pub(super) async fn upload_index_part<'a>(
 pub(super) async fn upload_timeline_layer<'a>(
    conf: &'static PageServerConf,
    storage: &'a GenericRemoteStorage,
-    source_path: &'a Path,
+    source_path: &'a Utf8Path,
    known_metadata: &'a LayerFileMetadata,
    generation: Generation,
 ) -> anyhow::Result<()> {
@@ -68,7 +69,7 @@ pub(super) async fn upload_timeline_layer<'a>(
            // upload. However, a nonexistent file can also be indicative of
            // something worse, like when a file is scheduled for upload before
            // it has been written to disk yet.
-            info!(path = %source_path.display(), "File to upload doesn't exist. Likely the file has been deleted and an upload is not required any more.");
+            info!(path = %source_path, "File to upload doesn't exist. Likely the file has been deleted and an upload is not required any more.");
            return Ok(());
        }
        Err(e) => {
@@ -93,7 +94,7 @@ pub(super) async fn upload_timeline_layer<'a>(
    storage
        .upload(source_file, fs_size, &storage_path, None)
        .await
-        .with_context(|| format!("upload layer from local path '{}'", source_path.display()))?;
+        .with_context(|| format!("upload layer from local path '{source_path}'"))?;

    Ok(())
 }
--- a/pageserver/src/tenant/storage_layer.rs
+++ b/pageserver/src/tenant/storage_layer.rs
@@ -14,6 +14,7 @@ use crate::task_mgr::TaskKind;
 use crate::walrecord::NeonWalRecord;
 use anyhow::Result;
 use bytes::Bytes;
+use camino::Utf8PathBuf;
 use enum_map::EnumMap;
 use enumset::EnumSet;
 use once_cell::sync::Lazy;
@@ -22,7 +23,6 @@ use pageserver_api::models::{
    HistoricLayerInfo, LayerResidenceEvent, LayerResidenceEventReason, LayerResidenceStatus,
 };
 use std::ops::Range;
-use std::path::PathBuf;
 use std::sync::{Arc, Mutex};
 use std::time::{Duration, SystemTime, UNIX_EPOCH};
 use tracing::warn;
@@ -378,7 +378,7 @@ pub trait PersistentLayer: Layer + AsLayerDesc {

    // Path to the layer file in the local filesystem.
    // `None` for `RemoteLayer`.
-    fn local_path(&self) -> Option<PathBuf>;
+    fn local_path(&self) -> Option<Utf8PathBuf>;

    /// Permanently remove this layer from disk.
    fn delete_resident_layer_file(&self) -> Result<()>;
@@ -456,7 +456,7 @@ pub mod tests {
 /// config. In that case, we use the Path variant to hold the full path to the file on
 /// disk.
 enum PathOrConf {
-    Path(PathBuf),
+    Path(Utf8PathBuf),
    Conf(&'static PageServerConf),
 }

--- a/pageserver/src/tenant/storage_layer/delta_layer.rs
+++ b/pageserver/src/tenant/storage_layer/delta_layer.rs
@@ -28,7 +28,7 @@
 //! "values" part.
 //!
 use crate::config::PageServerConf;
-use crate::context::RequestContext;
+use crate::context::{PageContentKind, RequestContext, RequestContextBuilder};
 use crate::page_cache::PAGE_SZ;
 use crate::repository::{Key, Value, KEY_SIZE};
 use crate::tenant::blob_io::BlobWriter;
@@ -41,6 +41,7 @@ use crate::virtual_file::VirtualFile;
 use crate::{walrecord, TEMP_FILE_SUFFIX};
 use crate::{DELTA_FILE_MAGIC, STORAGE_FORMAT_VERSION};
 use anyhow::{bail, ensure, Context, Result};
+use camino::{Utf8Path, Utf8PathBuf};
 use pageserver_api::models::{HistoricLayerInfo, LayerAccessKind};
 use rand::{distributions::Alphanumeric, Rng};
 use serde::{Deserialize, Serialize};
@@ -48,7 +49,6 @@ use std::fs::{self, File};
 use std::io::SeekFrom;
 use std::ops::Range;
 use std::os::unix::fs::FileExt;
-use std::path::{Path, PathBuf};
 use std::sync::Arc;
 use tokio::sync::OnceCell;
 use tracing::*;
@@ -267,7 +267,7 @@ impl PersistentLayer for DeltaLayer {
        Some(self)
    }

-    fn local_path(&self) -> Option<PathBuf> {
+    fn local_path(&self) -> Option<Utf8PathBuf> {
        self.local_path()
    }

@@ -317,11 +317,11 @@ impl DeltaLayer {

        tree_reader.dump().await?;

-        let keys = DeltaLayerInner::load_keys(&inner).await?;
+        let keys = DeltaLayerInner::load_keys(&inner, ctx).await?;

        // A subroutine to dump a single blob
-        async fn dump_blob(val: ValueRef<'_>) -> Result<String> {
-            let buf = val.reader.read_blob(val.blob_ref.pos()).await?;
+        async fn dump_blob(val: ValueRef<'_>, ctx: &RequestContext) -> Result<String> {
+            let buf = val.reader.read_blob(val.blob_ref.pos(), ctx).await?;
            let val = Value::des(&buf)?;
            let desc = match val {
                Value::Image(img) => {
@@ -342,7 +342,7 @@ impl DeltaLayer {

        for entry in keys {
            let DeltaEntry { key, lsn, val, .. } = entry;
-            let desc = match dump_blob(val).await {
+            let desc = match dump_blob(val, ctx).await {
                Ok(desc) => desc,
                Err(err) => {
                    let err: anyhow::Error = err;
@@ -370,11 +370,11 @@ impl DeltaLayer {
            .load(LayerAccessKind::GetValueReconstructData, ctx)
            .await?;
        inner
-            .get_value_reconstruct_data(key, lsn_range, reconstruct_state)
+            .get_value_reconstruct_data(key, lsn_range, reconstruct_state, ctx)
            .await
    }

-    pub(crate) fn local_path(&self) -> Option<PathBuf> {
+    pub(crate) fn local_path(&self) -> Option<Utf8PathBuf> {
        Some(self.path())
    }

@@ -409,7 +409,7 @@ impl DeltaLayer {
        tenant_id: &TenantId,
        timeline_id: &TimelineId,
        fname: &DeltaFileName,
-    ) -> PathBuf {
+    ) -> Utf8PathBuf {
        match path_or_conf {
            PathOrConf::Path(path) => path.clone(),
            PathOrConf::Conf(conf) => conf
@@ -424,7 +424,7 @@ impl DeltaLayer {
        timeline_id: &TimelineId,
        key_start: Key,
        lsn_range: &Range<Lsn>,
-    ) -> PathBuf {
+    ) -> Utf8PathBuf {
        let rand_string: String = rand::thread_rng()
            .sample_iter(&Alphanumeric)
            .take(8)
@@ -453,12 +453,12 @@ impl DeltaLayer {
        self.access_stats.record_access(access_kind, ctx);
        // Quick exit if already loaded
        self.inner
-            .get_or_try_init(|| self.load_inner())
+            .get_or_try_init(|| self.load_inner(ctx))
            .await
-            .with_context(|| format!("Failed to load delta layer {}", self.path().display()))
+            .with_context(|| format!("Failed to load delta layer {}", self.path()))
    }

-    async fn load_inner(&self) -> Result<Arc<DeltaLayerInner>> {
+    async fn load_inner(&self, ctx: &RequestContext) -> Result<Arc<DeltaLayerInner>> {
        let path = self.path();

        let summary = match &self.path_or_conf {
@@ -466,12 +466,12 @@ impl DeltaLayer {
            PathOrConf::Path(_) => None,
        };

-        let loaded = DeltaLayerInner::load(&path, summary).await?;
+        let loaded = DeltaLayerInner::load(&path, summary, ctx).await?;

        if let PathOrConf::Path(ref path) = self.path_or_conf {
            // not production code

-            let actual_filename = path.file_name().unwrap().to_str().unwrap().to_owned();
+            let actual_filename = path.file_name().unwrap().to_owned();
            let expected_filename = self.filename().file_name();

            if actual_filename != expected_filename {
@@ -510,7 +510,7 @@ impl DeltaLayer {
    /// Create a DeltaLayer struct representing an existing file on disk.
    ///
    /// This variant is only used for debugging purposes, by the 'pagectl' binary.
-    pub fn new_for_path(path: &Path, file: File) -> Result<Self> {
+    pub fn new_for_path(path: &Utf8Path, file: File) -> Result<Self> {
        let mut summary_buf = Vec::new();
        summary_buf.resize(PAGE_SZ, 0);
        file.read_exact_at(&mut summary_buf, 0)?;
@@ -538,7 +538,7 @@ impl DeltaLayer {
        self.desc.delta_file_name()
    }
    /// Path to the layer file in pageserver workdir.
-    pub fn path(&self) -> PathBuf {
+    pub fn path(&self) -> Utf8PathBuf {
        Self::path_for(
            &self.path_or_conf,
            &self.desc.tenant_id,
@@ -554,7 +554,7 @@ impl DeltaLayer {
            .load(LayerAccessKind::KeyIter, ctx)
            .await
            .context("load delta layer keys")?;
-        DeltaLayerInner::load_keys(inner)
+        DeltaLayerInner::load_keys(inner, ctx)
            .await
            .context("Layer index is corrupted")
    }
@@ -573,7 +573,7 @@ impl DeltaLayer {
 ///
 struct DeltaLayerWriterInner {
    conf: &'static PageServerConf,
-    pub path: PathBuf,
+    pub path: Utf8PathBuf,
    timeline_id: TimelineId,
    tenant_id: TenantId,

@@ -711,7 +711,7 @@ impl DeltaLayerWriterInner {
        ensure!(
            metadata.len() <= S3_UPLOAD_LIMIT,
            "Created delta layer file at {} of size {} above limit {S3_UPLOAD_LIMIT}!",
-            file.path.display(),
+            file.path,
            metadata.len()
        );

@@ -748,7 +748,7 @@ impl DeltaLayerWriterInner {
        );
        std::fs::rename(self.path, &final_path)?;

-        trace!("created delta layer {}", final_path.display());
+        trace!("created delta layer {final_path}");

        Ok(layer)
    }
@@ -847,15 +847,16 @@ impl Drop for DeltaLayerWriter {

 impl DeltaLayerInner {
    pub(super) async fn load(
-        path: &std::path::Path,
+        path: &Utf8Path,
        summary: Option<Summary>,
+        ctx: &RequestContext,
    ) -> anyhow::Result<Self> {
        let file = VirtualFile::open(path)
            .await
-            .with_context(|| format!("Failed to open file '{}'", path.display()))?;
+            .with_context(|| format!("Failed to open file '{path}'"))?;
        let file = FileBlockReader::new(file);

-        let summary_blk = file.read_blk(0).await?;
+        let summary_blk = file.read_blk(0, ctx).await?;
        let actual_summary = Summary::des_prefix(summary_blk.as_ref())?;

        if let Some(mut expected_summary) = summary {
@@ -883,6 +884,7 @@ impl DeltaLayerInner {
        key: Key,
        lsn_range: Range<Lsn>,
        reconstruct_state: &mut ValueReconstructState,
+        ctx: &RequestContext,
    ) -> anyhow::Result<ValueReconstructResult> {
        let mut need_image = true;
        // Scan the page versions backwards, starting from `lsn`.
@@ -897,38 +899,46 @@ impl DeltaLayerInner {
        let mut offsets: Vec<(Lsn, u64)> = Vec::new();

        tree_reader
-            .visit(&search_key.0, VisitDirection::Backwards, |key, value| {
-                let blob_ref = BlobRef(value);
-                if key[..KEY_SIZE] != search_key.0[..KEY_SIZE] {
-                    return false;
-                }
-                let entry_lsn = DeltaKey::extract_lsn_from_buf(key);
-                if entry_lsn < lsn_range.start {
-                    return false;
-                }
-                offsets.push((entry_lsn, blob_ref.pos()));
+            .visit(
+                &search_key.0,
+                VisitDirection::Backwards,
+                |key, value| {
+                    let blob_ref = BlobRef(value);
+                    if key[..KEY_SIZE] != search_key.0[..KEY_SIZE] {
+                        return false;
+                    }
+                    let entry_lsn = DeltaKey::extract_lsn_from_buf(key);
+                    if entry_lsn < lsn_range.start {
+                        return false;
+                    }
+                    offsets.push((entry_lsn, blob_ref.pos()));

-                !blob_ref.will_init()
-            })
+                    !blob_ref.will_init()
+                },
+                &RequestContextBuilder::extend(ctx)
+                    .page_content_kind(PageContentKind::DeltaLayerBtreeNode)
+                    .build(),
+            )
            .await?;

+        let ctx = &RequestContextBuilder::extend(ctx)
+            .page_content_kind(PageContentKind::DeltaLayerValue)
+            .build();
+
        // Ok, 'offsets' now contains the offsets of all the entries we need to read
        let cursor = file.block_cursor();
        let mut buf = Vec::new();
        for (entry_lsn, pos) in offsets {
            cursor
-                .read_blob_into_buf(pos, &mut buf)
+                .read_blob_into_buf(pos, &mut buf, ctx)
                .await
                .with_context(|| {
-                    format!(
-                        "Failed to read blob from virtual file {}",
-                        file.file.path.display()
-                    )
+                    format!("Failed to read blob from virtual file {}", file.file.path)
                })?;
            let val = Value::des(&buf).with_context(|| {
                format!(
                    "Failed to deserialize file blob from virtual file {}",
-                    file.file.path.display()
+                    file.file.path
                )
            })?;
            match val {
@@ -958,9 +968,10 @@ impl DeltaLayerInner {
        }
    }

-    pub(super) async fn load_keys<T: AsRef<DeltaLayerInner> + Clone>(
-        this: &T,
-    ) -> Result<Vec<DeltaEntry<'_>>> {
+    pub(super) async fn load_keys<'a, 'b, T: AsRef<DeltaLayerInner> + Clone>(
+        this: &'a T,
+        ctx: &'b RequestContext,
+    ) -> Result<Vec<DeltaEntry<'a>>> {
        let dl = this.as_ref();
        let file = &dl.file;

@@ -997,6 +1008,9 @@ impl DeltaLayerInner {
                    all_keys.push(entry);
                    true
                },
+                &RequestContextBuilder::extend(ctx)
+                    .page_content_kind(PageContentKind::DeltaLayerBtreeNode)
+                    .build(),
            )
            .await?;
        if let Some(last) = all_keys.last_mut() {
@@ -1026,9 +1040,9 @@ pub struct ValueRef<'a> {

 impl<'a> ValueRef<'a> {
    /// Loads the value from disk
-    pub async fn load(&self) -> Result<Value> {
+    pub async fn load(&self, ctx: &RequestContext) -> Result<Value> {
        // theoretically we *could* record an access time for each, but it does not really matter
-        let buf = self.reader.read_blob(self.blob_ref.pos()).await?;
+        let buf = self.reader.read_blob(self.blob_ref.pos(), ctx).await?;
        let val = Value::des(&buf)?;
        Ok(val)
    }
@@ -1037,7 +1051,11 @@ impl<'a> ValueRef<'a> {
 pub(crate) struct Adapter<T>(T);

 impl<T: AsRef<DeltaLayerInner>> Adapter<T> {
-    pub(crate) async fn read_blk(&self, blknum: u32) -> Result<BlockLease, std::io::Error> {
-        self.0.as_ref().file.read_blk(blknum).await
+    pub(crate) async fn read_blk(
+        &self,
+        blknum: u32,
+        ctx: &RequestContext,
+    ) -> Result<BlockLease, std::io::Error> {
+        self.0.as_ref().file.read_blk(blknum, ctx).await
    }
 }
--- a/pageserver/src/tenant/storage_layer/image_layer.rs
+++ b/pageserver/src/tenant/storage_layer/image_layer.rs
@@ -24,7 +24,7 @@
 //! mapping from Key to an offset in the "values" part.  The
 //! actual page images are stored in the "values" part.
 use crate::config::PageServerConf;
-use crate::context::RequestContext;
+use crate::context::{PageContentKind, RequestContext, RequestContextBuilder};
 use crate::page_cache::PAGE_SZ;
 use crate::repository::{Key, KEY_SIZE};
 use crate::tenant::blob_io::BlobWriter;
@@ -37,6 +37,7 @@ use crate::virtual_file::VirtualFile;
 use crate::{IMAGE_FILE_MAGIC, STORAGE_FORMAT_VERSION, TEMP_FILE_SUFFIX};
 use anyhow::{bail, ensure, Context, Result};
 use bytes::Bytes;
+use camino::{Utf8Path, Utf8PathBuf};
 use hex;
 use pageserver_api::models::{HistoricLayerInfo, LayerAccessKind};
 use rand::{distributions::Alphanumeric, Rng};
@@ -45,7 +46,6 @@ use std::fs::{self, File};
 use std::io::SeekFrom;
 use std::ops::Range;
 use std::os::unix::prelude::FileExt;
-use std::path::{Path, PathBuf};
 use tokio::sync::OnceCell;
 use tracing::*;

@@ -195,7 +195,7 @@ impl AsLayerDesc for ImageLayer {
 }

 impl PersistentLayer for ImageLayer {
-    fn local_path(&self) -> Option<PathBuf> {
+    fn local_path(&self) -> Option<Utf8PathBuf> {
        self.local_path()
    }

@@ -237,10 +237,15 @@ impl ImageLayer {
        tree_reader.dump().await?;

        tree_reader
-            .visit(&[0u8; KEY_SIZE], VisitDirection::Forwards, |key, value| {
-                println!("key: {} offset {}", hex::encode(key), value);
-                true
-            })
+            .visit(
+                &[0u8; KEY_SIZE],
+                VisitDirection::Forwards,
+                |key, value| {
+                    println!("key: {} offset {}", hex::encode(key), value);
+                    true
+                },
+                ctx,
+            )
            .await?;

        Ok(())
@@ -261,13 +266,13 @@ impl ImageLayer {
            .load(LayerAccessKind::GetValueReconstructData, ctx)
            .await?;
        inner
-            .get_value_reconstruct_data(key, reconstruct_state)
+            .get_value_reconstruct_data(key, reconstruct_state, ctx)
            .await
            // FIXME: makes no sense to dump paths
-            .with_context(|| format!("read {}", self.path().display()))
+            .with_context(|| format!("read {}", self.path()))
    }

-    pub(crate) fn local_path(&self) -> Option<PathBuf> {
+    pub(crate) fn local_path(&self) -> Option<Utf8PathBuf> {
        Some(self.path())
    }

@@ -299,7 +304,7 @@ impl ImageLayer {
        timeline_id: TimelineId,
        tenant_id: TenantId,
        fname: &ImageFileName,
-    ) -> PathBuf {
+    ) -> Utf8PathBuf {
        match path_or_conf {
            PathOrConf::Path(path) => path.to_path_buf(),
            PathOrConf::Conf(conf) => conf
@@ -313,7 +318,7 @@ impl ImageLayer {
        timeline_id: TimelineId,
        tenant_id: TenantId,
        fname: &ImageFileName,
-    ) -> PathBuf {
+    ) -> Utf8PathBuf {
        let rand_string: String = rand::thread_rng()
            .sample_iter(&Alphanumeric)
            .take(8)
@@ -335,12 +340,12 @@ impl ImageLayer {
    ) -> Result<&ImageLayerInner> {
        self.access_stats.record_access(access_kind, ctx);
        self.inner
-            .get_or_try_init(|| self.load_inner())
+            .get_or_try_init(|| self.load_inner(ctx))
            .await
-            .with_context(|| format!("Failed to load image layer {}", self.path().display()))
+            .with_context(|| format!("Failed to load image layer {}", self.path()))
    }

-    async fn load_inner(&self) -> Result<ImageLayerInner> {
+    async fn load_inner(&self, ctx: &RequestContext) -> Result<ImageLayerInner> {
        let path = self.path();

        let expected_summary = match &self.path_or_conf {
@@ -349,11 +354,12 @@ impl ImageLayer {
        };

        let loaded =
-            ImageLayerInner::load(&path, self.desc.image_layer_lsn(), expected_summary).await?;
+            ImageLayerInner::load(&path, self.desc.image_layer_lsn(), expected_summary, ctx)
+                .await?;

        if let PathOrConf::Path(ref path) = self.path_or_conf {
            // not production code
-            let actual_filename = path.file_name().unwrap().to_str().unwrap().to_owned();
+            let actual_filename = path.file_name().unwrap().to_owned();
            let expected_filename = self.filename().file_name();

            if actual_filename != expected_filename {
@@ -393,7 +399,7 @@ impl ImageLayer {
    /// Create an ImageLayer struct representing an existing file on disk.
    ///
    /// This variant is only used for debugging purposes, by the 'pagectl' binary.
-    pub fn new_for_path(path: &Path, file: File) -> Result<ImageLayer> {
+    pub fn new_for_path(path: &Utf8Path, file: File) -> Result<ImageLayer> {
        let mut summary_buf = Vec::new();
        summary_buf.resize(PAGE_SZ, 0);
        file.read_exact_at(&mut summary_buf, 0)?;
@@ -421,7 +427,7 @@ impl ImageLayer {
    }

    /// Path to the layer file in pageserver workdir.
-    pub fn path(&self) -> PathBuf {
+    pub fn path(&self) -> Utf8PathBuf {
        Self::path_for(
            &self.path_or_conf,
            self.desc.timeline_id,
@@ -433,15 +439,16 @@ impl ImageLayer {

 impl ImageLayerInner {
    pub(super) async fn load(
-        path: &std::path::Path,
+        path: &Utf8Path,
        lsn: Lsn,
        summary: Option<Summary>,
+        ctx: &RequestContext,
    ) -> anyhow::Result<Self> {
        let file = VirtualFile::open(path)
            .await
-            .with_context(|| format!("Failed to open file '{}'", path.display()))?;
+            .with_context(|| format!("Failed to open file '{}'", path))?;
        let file = FileBlockReader::new(file);
-        let summary_blk = file.read_blk(0).await?;
+        let summary_blk = file.read_blk(0, ctx).await?;
        let actual_summary = Summary::des_prefix(summary_blk.as_ref())?;

        if let Some(mut expected_summary) = summary {
@@ -470,16 +477,30 @@ impl ImageLayerInner {
        &self,
        key: Key,
        reconstruct_state: &mut ValueReconstructState,
+        ctx: &RequestContext,
    ) -> anyhow::Result<ValueReconstructResult> {
        let file = &self.file;
        let tree_reader = DiskBtreeReader::new(self.index_start_blk, self.index_root_blk, file);

        let mut keybuf: [u8; KEY_SIZE] = [0u8; KEY_SIZE];
        key.write_to_byte_slice(&mut keybuf);
-        if let Some(offset) = tree_reader.get(&keybuf).await? {
+        if let Some(offset) = tree_reader
+            .get(
+                &keybuf,
+                &RequestContextBuilder::extend(ctx)
+                    .page_content_kind(PageContentKind::ImageLayerBtreeNode)
+                    .build(),
+            )
+            .await?
+        {
            let blob = file
                .block_cursor()
-                .read_blob(offset)
+                .read_blob(
+                    offset,
+                    &RequestContextBuilder::extend(ctx)
+                        .page_content_kind(PageContentKind::ImageLayerValue)
+                        .build(),
+                )
                .await
                .with_context(|| format!("failed to read value from offset {}", offset))?;
            let value = Bytes::from(blob);
@@ -505,7 +526,7 @@ impl ImageLayerInner {
 ///
 struct ImageLayerWriterInner {
    conf: &'static PageServerConf,
-    path: PathBuf,
+    path: Utf8PathBuf,
    timeline_id: TimelineId,
    tenant_id: TenantId,
    key_range: Range<Key>,
@@ -537,7 +558,7 @@ impl ImageLayerWriterInner {
                lsn,
            },
        );
-        info!("new image layer {}", path.display());
+        info!("new image layer {path}");
        let mut file = VirtualFile::open_with_options(
            &path,
            std::fs::OpenOptions::new().write(true).create_new(true),
@@ -664,7 +685,7 @@ impl ImageLayerWriterInner {
        );
        std::fs::rename(self.path, final_path)?;

-        trace!("created image layer {}", layer.path().display());
+        trace!("created image layer {}", layer.path());

        Ok(layer)
    }
--- a/pageserver/src/tenant/storage_layer/inmemory_layer.rs
+++ b/pageserver/src/tenant/storage_layer/inmemory_layer.rs
@@ -5,7 +5,7 @@
 //! its position in the file, is kept in memory, though.
 //!
 use crate::config::PageServerConf;
-use crate::context::RequestContext;
+use crate::context::{PageContentKind, RequestContext, RequestContextBuilder};
 use crate::repository::{Key, Value};
 use crate::tenant::block_io::BlockReader;
 use crate::tenant::ephemeral_file::EphemeralFile;
@@ -106,7 +106,7 @@ impl InMemoryLayer {
    /// debugging function to print out the contents of the layer
    ///
    /// this is likely completly unused
-    pub async fn dump(&self, verbose: bool, _ctx: &RequestContext) -> Result<()> {
+    pub async fn dump(&self, verbose: bool, ctx: &RequestContext) -> Result<()> {
        let inner = self.inner.read().await;

        let end_str = self.end_lsn_or_max();
@@ -125,7 +125,7 @@ impl InMemoryLayer {
        for (key, vec_map) in inner.index.iter() {
            for (lsn, pos) in vec_map.as_slice() {
                let mut desc = String::new();
-                cursor.read_blob_into_buf(*pos, &mut buf).await?;
+                cursor.read_blob_into_buf(*pos, &mut buf, ctx).await?;
                let val = Value::des(&buf);
                match val {
                    Ok(Value::Image(img)) => {
@@ -158,11 +158,15 @@ impl InMemoryLayer {
        key: Key,
        lsn_range: Range<Lsn>,
        reconstruct_state: &mut ValueReconstructState,
-        _ctx: &RequestContext,
+        ctx: &RequestContext,
    ) -> anyhow::Result<ValueReconstructResult> {
        ensure!(lsn_range.start >= self.start_lsn);
        let mut need_image = true;

+        let ctx = RequestContextBuilder::extend(ctx)
+            .page_content_kind(PageContentKind::InMemoryLayer)
+            .build();
+
        let inner = self.inner.read().await;

        let reader = inner.file.block_cursor();
@@ -171,7 +175,7 @@ impl InMemoryLayer {
        if let Some(vec_map) = inner.index.get(&key) {
            let slice = vec_map.slice_range(lsn_range);
            for (entry_lsn, pos) in slice.iter().rev() {
-                let buf = reader.read_blob(*pos).await?;
+                let buf = reader.read_blob(*pos, &ctx).await?;
                let value = Value::des(&buf)?;
                match value {
                    Value::Image(img) => {
@@ -263,7 +267,13 @@ impl InMemoryLayer {

    /// Common subroutine of the public put_wal_record() and put_page_image() functions.
    /// Adds the page version to the in-memory tree
-    pub async fn put_value(&self, key: Key, lsn: Lsn, val: &Value) -> Result<()> {
+    pub async fn put_value(
+        &self,
+        key: Key,
+        lsn: Lsn,
+        val: &Value,
+        ctx: &RequestContext,
+    ) -> Result<()> {
        trace!("put_value key {} at {}/{}", key, self.timeline_id, lsn);
        let inner: &mut _ = &mut *self.inner.write().await;
        self.assert_writable();
@@ -275,7 +285,15 @@ impl InMemoryLayer {
            let mut buf = smallvec::SmallVec::<[u8; 256]>::new();
            buf.clear();
            val.ser_into(&mut buf)?;
-            inner.file.write_blob(&buf).await?
+            inner
+                .file
+                .write_blob(
+                    &buf,
+                    &RequestContextBuilder::extend(ctx)
+                        .page_content_kind(PageContentKind::InMemoryLayer)
+                        .build(),
+                )
+                .await?
        };

        let vec_map = inner.index.entry(key).or_default();
@@ -313,7 +331,7 @@ impl InMemoryLayer {
    /// Write this frozen in-memory layer to disk.
    ///
    /// Returns a new delta layer with all the same data as this in-memory layer
-    pub(crate) async fn write_to_disk(&self) -> Result<DeltaLayer> {
+    pub(crate) async fn write_to_disk(&self, ctx: &RequestContext) -> Result<DeltaLayer> {
        // Grab the lock in read-mode. We hold it over the I/O, but because this
        // layer is not writeable anymore, no one should be trying to acquire the
        // write lock on it, so we shouldn't block anyone. There's one exception
@@ -343,11 +361,14 @@ impl InMemoryLayer {
        let mut keys: Vec<(&Key, &VecMap<Lsn, u64>)> = inner.index.iter().collect();
        keys.sort_by_key(|k| k.0);

+        let ctx = RequestContextBuilder::extend(ctx)
+            .page_content_kind(PageContentKind::InMemoryLayer)
+            .build();
        for (key, vec_map) in keys.iter() {
            let key = **key;
            // Write all page versions
            for (lsn, pos) in vec_map.as_slice() {
-                cursor.read_blob_into_buf(*pos, &mut buf).await?;
+                cursor.read_blob_into_buf(*pos, &mut buf, &ctx).await?;
                let will_init = Value::des(&buf)?.will_init();
                delta_layer_writer
                    .put_value_bytes(key, *lsn, &buf, will_init)
--- a/pageserver/src/tenant/storage_layer/remote_layer.rs
+++ b/pageserver/src/tenant/storage_layer/remote_layer.rs
@@ -8,9 +8,9 @@ use crate::tenant::remote_timeline_client::index::LayerFileMetadata;
 use crate::tenant::storage_layer::{Layer, ValueReconstructResult, ValueReconstructState};
 use crate::tenant::timeline::layer_manager::LayerManager;
 use anyhow::{bail, Result};
+use camino::Utf8PathBuf;
 use pageserver_api::models::HistoricLayerInfo;
 use std::ops::Range;
-use std::path::PathBuf;
 use std::sync::Arc;

 use utils::{
@@ -92,7 +92,7 @@ impl AsLayerDesc for RemoteLayer {
 }

 impl PersistentLayer for RemoteLayer {
-    fn local_path(&self) -> Option<PathBuf> {
+    fn local_path(&self) -> Option<Utf8PathBuf> {
        None
    }

--- a/pageserver/src/tenant/timeline.rs
+++ b/pageserver/src/tenant/timeline.rs
@@ -9,6 +9,7 @@ mod walreceiver;

 use anyhow::{anyhow, bail, ensure, Context, Result};
 use bytes::Bytes;
+use camino::{Utf8Path, Utf8PathBuf};
 use fail::fail_point;
 use futures::StreamExt;
 use itertools::Itertools;
@@ -29,7 +30,6 @@ use utils::id::TenantTimelineId;
 use std::cmp::{max, min, Ordering};
 use std::collections::{BinaryHeap, HashMap, HashSet};
 use std::ops::{Deref, Range};
-use std::path::{Path, PathBuf};
 use std::pin::pin;
 use std::sync::atomic::Ordering as AtomicOrdering;
 use std::sync::{Arc, Mutex, RwLock, Weak};
@@ -38,6 +38,7 @@ use std::time::{Duration, Instant, SystemTime};
 use crate::context::{
    AccessStatsBehavior, DownloadBehavior, RequestContext, RequestContextBuilder,
 };
+use crate::deletion_queue::DeletionQueueClient;
 use crate::tenant::remote_timeline_client::index::LayerFileMetadata;
 use crate::tenant::storage_layer::delta_layer::DeltaEntry;
 use crate::tenant::storage_layer::{
@@ -55,7 +56,7 @@ use crate::config::PageServerConf;
 use crate::keyspace::{KeyPartitioning, KeySpace, KeySpaceRandomAccum};
 use crate::metrics::{
    TimelineMetrics, MATERIALIZED_PAGE_CACHE_HIT, MATERIALIZED_PAGE_CACHE_HIT_DIRECT,
-    RECONSTRUCT_TIME, UNEXPECTED_ONDEMAND_DOWNLOADS,
+    UNEXPECTED_ONDEMAND_DOWNLOADS,
 };
 use crate::pgdatadir_mapping::LsnForTimestamp;
 use crate::pgdatadir_mapping::{is_rel_fsm_block_key, is_rel_vm_block_key};
@@ -90,12 +91,12 @@ use self::logical_size::LogicalSize;
 use self::walreceiver::{WalReceiver, WalReceiverConf};

 use super::config::TenantConf;
-use super::debug_assert_current_span_has_tenant_and_timeline_id;
 use super::remote_timeline_client::index::IndexPart;
 use super::remote_timeline_client::RemoteTimelineClient;
 use super::storage_layer::{
    AsLayerDesc, DeltaLayer, ImageLayer, LayerAccessStatsReset, PersistentLayerDesc,
 };
+use super::{debug_assert_current_span_has_tenant_and_timeline_id, AttachedTenantConf};

 #[derive(Debug, PartialEq, Eq, Clone, Copy)]
 pub(super) enum FlushLoopState {
@@ -143,11 +144,12 @@ fn drop_wlock<T>(rlock: tokio::sync::RwLockWriteGuard<'_, T>) {
 /// The outward-facing resources required to build a Timeline
 pub struct TimelineResources {
    pub remote_client: Option<RemoteTimelineClient>,
+    pub deletion_queue_client: DeletionQueueClient,
 }

 pub struct Timeline {
    conf: &'static PageServerConf,
-    tenant_conf: Arc<RwLock<TenantConfOpt>>,
+    tenant_conf: Arc<RwLock<AttachedTenantConf>>,

    myself: Weak<Self>,

@@ -156,6 +158,9 @@ pub struct Timeline {

    /// The generation of the tenant that instantiated us: this is used for safety when writing remote objects.
    /// Never changes for the lifetime of this [`Timeline`] object.
+    ///  
+    /// This duplicates the generation stored in LocationConf, but that structure is mutable:
+    /// this copy enforces the invariant that generatio doesn't change during a Tenant's lifetime.
    generation: Generation,

    pub pg_version: u32,
@@ -471,7 +476,7 @@ impl Timeline {
        // The cached image can be returned directly if there is no WAL between the cached image
        // and requested LSN. The cached image can also be used to reduce the amount of WAL needed
        // for redo.
-        let cached_page_img = match self.lookup_cached_page(&key, lsn).await {
+        let cached_page_img = match self.lookup_cached_page(&key, lsn, ctx).await {
            Some((cached_lsn, cached_img)) => {
                match cached_lsn.cmp(&lsn) {
                    Ordering::Less => {} // there might be WAL between cached_lsn and lsn, we need to check
@@ -494,13 +499,39 @@ impl Timeline {
        };

        let timer = crate::metrics::GET_RECONSTRUCT_DATA_TIME.start_timer();
-        self.get_reconstruct_data(key, lsn, &mut reconstruct_state, ctx)
+        let path = self
+            .get_reconstruct_data(key, lsn, &mut reconstruct_state, ctx)
            .await?;
        timer.stop_and_record();

-        RECONSTRUCT_TIME
-            .observe_closure_duration(|| self.reconstruct_value(key, lsn, reconstruct_state))
-            .await
+        let start = Instant::now();
+        let res = self.reconstruct_value(key, lsn, reconstruct_state).await;
+        let elapsed = start.elapsed();
+        crate::metrics::RECONSTRUCT_TIME
+            .for_result(&res)
+            .observe(elapsed.as_secs_f64());
+
+        if cfg!(feature = "testing") && res.is_err() {
+            // it can only be walredo issue
+            use std::fmt::Write;
+
+            let mut msg = String::new();
+
+            path.into_iter().for_each(|(res, cont_lsn, layer)| {
+                writeln!(
+                    msg,
+                    "- layer traversal: result {res:?}, cont_lsn {cont_lsn}, layer: {}",
+                    layer(),
+                )
+                .expect("string grows")
+            });
+
+            // this is to rule out or provide evidence that we could in some cases read a duplicate
+            // walrecord
+            tracing::info!("walredo failed, path:\n{msg}");
+        }
+
+        res
    }

    /// Get last or prev record separately. Same as get_last_record_rlsn().last/prev.
@@ -521,9 +552,23 @@ impl Timeline {
        self.disk_consistent_lsn.load()
    }

-    pub fn get_remote_consistent_lsn(&self) -> Option<Lsn> {
+    /// remote_consistent_lsn from the perspective of the tenant's current generation,
+    /// not validated with control plane yet.
+    /// See [`Self::get_remote_consistent_lsn_visible`].
+    pub fn get_remote_consistent_lsn_projected(&self) -> Option<Lsn> {
        if let Some(remote_client) = &self.remote_client {
-            remote_client.last_uploaded_consistent_lsn()
+            remote_client.remote_consistent_lsn_projected()
+        } else {
+            None
+        }
+    }
+
+    /// remote_consistent_lsn which the tenant is guaranteed not to go backward from,
+    /// i.e. a value of remote_consistent_lsn_projected which has undergone
+    /// generation validation in the deletion queue.
+    pub fn get_remote_consistent_lsn_visible(&self) -> Option<Lsn> {
+        if let Some(remote_client) = &self.remote_client {
+            remote_client.remote_consistent_lsn_visible()
        } else {
            None
        }
@@ -543,7 +588,7 @@ impl Timeline {
    }

    pub fn resident_physical_size(&self) -> u64 {
-        self.metrics.resident_physical_size_gauge.get()
+        self.metrics.resident_physical_size_get()
    }

    ///
@@ -1293,10 +1338,7 @@ impl Timeline {
        // will treat the file as a local layer again, count it towards resident size,
        // and it'll be like the layer removal never happened.
        // The bump in resident size is perhaps unexpected but overall a robust behavior.
-        self.metrics
-            .resident_physical_size_gauge
-            .sub(layer_file_size);
-
+        self.metrics.resident_physical_size_sub(layer_file_size);
        self.metrics.evictions.inc();

        if let Some(delta) = local_layer_residence_duration {
@@ -1339,42 +1381,42 @@ const REPARTITION_FREQ_IN_CHECKPOINT_DISTANCE: u64 = 10;
 // Private functions
 impl Timeline {
    fn get_checkpoint_distance(&self) -> u64 {
-        let tenant_conf = self.tenant_conf.read().unwrap();
+        let tenant_conf = self.tenant_conf.read().unwrap().tenant_conf;
        tenant_conf
            .checkpoint_distance
            .unwrap_or(self.conf.default_tenant_conf.checkpoint_distance)
    }

    fn get_checkpoint_timeout(&self) -> Duration {
-        let tenant_conf = self.tenant_conf.read().unwrap();
+        let tenant_conf = self.tenant_conf.read().unwrap().tenant_conf;
        tenant_conf
            .checkpoint_timeout
            .unwrap_or(self.conf.default_tenant_conf.checkpoint_timeout)
    }

    fn get_compaction_target_size(&self) -> u64 {
-        let tenant_conf = self.tenant_conf.read().unwrap();
+        let tenant_conf = self.tenant_conf.read().unwrap().tenant_conf;
        tenant_conf
            .compaction_target_size
            .unwrap_or(self.conf.default_tenant_conf.compaction_target_size)
    }

    fn get_compaction_threshold(&self) -> usize {
-        let tenant_conf = self.tenant_conf.read().unwrap();
+        let tenant_conf = self.tenant_conf.read().unwrap().tenant_conf;
        tenant_conf
            .compaction_threshold
            .unwrap_or(self.conf.default_tenant_conf.compaction_threshold)
    }

    fn get_image_creation_threshold(&self) -> usize {
-        let tenant_conf = self.tenant_conf.read().unwrap();
+        let tenant_conf = self.tenant_conf.read().unwrap().tenant_conf;
        tenant_conf
            .image_creation_threshold
            .unwrap_or(self.conf.default_tenant_conf.image_creation_threshold)
    }

    fn get_eviction_policy(&self) -> EvictionPolicy {
-        let tenant_conf = self.tenant_conf.read().unwrap();
+        let tenant_conf = self.tenant_conf.read().unwrap().tenant_conf;
        tenant_conf
            .eviction_policy
            .unwrap_or(self.conf.default_tenant_conf.eviction_policy)
@@ -1390,7 +1432,7 @@ impl Timeline {
    }

    fn get_gc_feedback(&self) -> bool {
-        let tenant_conf = self.tenant_conf.read().unwrap();
+        let tenant_conf = &self.tenant_conf.read().unwrap().tenant_conf;
        tenant_conf
            .gc_feedback
            .unwrap_or(self.conf.default_tenant_conf.gc_feedback)
@@ -1403,7 +1445,7 @@ impl Timeline {
        // The threshold is embedded in the metric. So, we need to update it.
        {
            let new_threshold = Self::get_evictions_low_residence_duration_metric_threshold(
-                &self.tenant_conf.read().unwrap(),
+                &self.tenant_conf.read().unwrap().tenant_conf,
                &self.conf.default_tenant_conf,
            );
            let tenant_id_str = self.tenant_id.to_string();
@@ -1422,7 +1464,7 @@ impl Timeline {
    #[allow(clippy::too_many_arguments)]
    pub(super) fn new(
        conf: &'static PageServerConf,
-        tenant_conf: Arc<RwLock<TenantConfOpt>>,
+        tenant_conf: Arc<RwLock<AttachedTenantConf>>,
        metadata: &TimelineMetadata,
        ancestor: Option<Arc<Timeline>>,
        timeline_id: TimelineId,
@@ -1445,7 +1487,7 @@ impl Timeline {

        let evictions_low_residence_duration_metric_threshold =
            Self::get_evictions_low_residence_duration_metric_threshold(
-                &tenant_conf_guard,
+                &tenant_conf_guard.tenant_conf,
                &conf.default_tenant_conf,
            );
        drop(tenant_conf_guard);
@@ -1610,12 +1652,15 @@ impl Timeline {

        let tenant_conf_guard = self.tenant_conf.read().unwrap();
        let wal_connect_timeout = tenant_conf_guard
+            .tenant_conf
            .walreceiver_connect_timeout
            .unwrap_or(self.conf.default_tenant_conf.walreceiver_connect_timeout);
        let lagging_wal_timeout = tenant_conf_guard
+            .tenant_conf
            .lagging_wal_timeout
            .unwrap_or(self.conf.default_tenant_conf.lagging_wal_timeout);
        let max_lsn_wal_lag = tenant_conf_guard
+            .tenant_conf
            .max_lsn_wal_lag
            .unwrap_or(self.conf.default_tenant_conf.max_lsn_wal_lag);
        drop(tenant_conf_guard);
@@ -1697,7 +1742,7 @@ impl Timeline {
                        Discovered::Temporary(name) => (name, "temporary timeline file"),
                        Discovered::TemporaryDownload(name) => (name, "temporary download"),
                    };
-                    path.push(name);
+                    path.push(Utf8Path::new(&name));
                    init::cleanup(&path, kind)?;
                    path.pop();
                }
@@ -1820,7 +1865,7 @@ impl Timeline {
            for (layer, m) in needs_upload {
                rtc.schedule_layer_file_upload(&layer.layer_desc().filename(), &m)?;
            }
-            rtc.schedule_layer_file_deletion(&needs_cleanup)?;
+            rtc.schedule_layer_file_deletion(needs_cleanup)?;
            rtc.schedule_index_upload_for_file_changes()?;
            // Tenant::create_timeline will wait for these uploads to happen before returning, or
            // on retry.
@@ -1830,9 +1875,7 @@ impl Timeline {
            "loaded layer map with {} layers at {}, total physical size: {}",
            num_layers, disk_consistent_lsn, total_physical_size
        );
-        self.metrics
-            .resident_physical_size_gauge
-            .set(total_physical_size);
+        self.metrics.resident_physical_size_set(total_physical_size);

        timer.stop_and_record();
        Ok(())
@@ -2180,10 +2223,10 @@ impl TraversalLayerExt for Arc<dyn PersistentLayer> {
        let timeline_id = self.layer_desc().timeline_id;
        match self.local_path() {
            Some(local_path) => {
-                debug_assert!(local_path.to_str().unwrap().contains(&format!("{}", timeline_id)),
+                debug_assert!(local_path.to_string().contains(&format!("{}", timeline_id)),
                    "need timeline ID to uniquely identify the layer when traversal crosses ancestor boundary",
                );
-                format!("{}", local_path.display())
+                format!("{local_path}")
            }
            None => {
                format!("remote {}/{self}", timeline_id)
@@ -2213,7 +2256,7 @@ impl Timeline {
        request_lsn: Lsn,
        reconstruct_state: &mut ValueReconstructState,
        ctx: &RequestContext,
-    ) -> Result<(), PageReconstructError> {
+    ) -> Result<Vec<TraversalPathItem>, PageReconstructError> {
        // Start from the current timeline.
        let mut timeline_owned;
        let mut timeline = self;
@@ -2244,12 +2287,12 @@ impl Timeline {
            // The function should have updated 'state'
            //info!("CALLED for {} at {}: {:?} with {} records, cached {}", key, cont_lsn, result, reconstruct_state.records.len(), cached_lsn);
            match result {
-                ValueReconstructResult::Complete => return Ok(()),
+                ValueReconstructResult::Complete => return Ok(traversal_path),
                ValueReconstructResult::Continue => {
                    // If we reached an earlier cached page image, we're done.
                    if cont_lsn == cached_lsn + 1 {
                        MATERIALIZED_PAGE_CACHE_HIT.inc_by(1);
-                        return Ok(());
+                        return Ok(traversal_path);
                    }
                    if prev_lsn <= cont_lsn {
                        // Didn't make any progress in last iteration. Error out to avoid
@@ -2518,13 +2561,18 @@ impl Timeline {
        }
    }

-    async fn lookup_cached_page(&self, key: &Key, lsn: Lsn) -> Option<(Lsn, Bytes)> {
+    async fn lookup_cached_page(
+        &self,
+        key: &Key,
+        lsn: Lsn,
+        ctx: &RequestContext,
+    ) -> Option<(Lsn, Bytes)> {
        let cache = page_cache::get();

        // FIXME: It's pointless to check the cache for things that are not 8kB pages.
        // We should look at the key to determine if it's a cacheable object
        let (lsn, read_guard) = cache
-            .lookup_materialized_page(self.tenant_id, self.timeline_id, key, lsn)
+            .lookup_materialized_page(self.tenant_id, self.timeline_id, key, lsn, ctx)
            .await?;
        let img = Bytes::from(read_guard.to_vec());
        Some((lsn, img))
@@ -2558,10 +2606,16 @@ impl Timeline {
        Ok(layer)
    }

-    async fn put_value(&self, key: Key, lsn: Lsn, val: &Value) -> anyhow::Result<()> {
+    async fn put_value(
+        &self,
+        key: Key,
+        lsn: Lsn,
+        val: &Value,
+        ctx: &RequestContext,
+    ) -> anyhow::Result<()> {
        //info!("PUT: key {} at {}", key, lsn);
        let layer = self.get_layer_for_write(lsn).await?;
-        layer.put_value(key, lsn, val).await?;
+        layer.put_value(key, lsn, val, ctx).await?;
        Ok(())
    }

@@ -2733,7 +2787,7 @@ impl Timeline {
                // Normal case, write out a L0 delta layer file.
                // `create_delta_layer` will not modify the layer map.
                // We will remove frozen layer and add delta layer in one atomic operation later.
-                let layer = self.create_delta_layer(&frozen_layer).await?;
+                let layer = self.create_delta_layer(&frozen_layer, ctx).await?;
                (
                    HashMap::from([(
                        layer.filename(),
@@ -2856,19 +2910,21 @@ impl Timeline {
    async fn create_delta_layer(
        self: &Arc<Self>,
        frozen_layer: &Arc<InMemoryLayer>,
+        ctx: &RequestContext,
    ) -> anyhow::Result<DeltaLayer> {
        let span = tracing::info_span!("blocking");
        let new_delta: DeltaLayer = tokio::task::spawn_blocking({
            let _g = span.entered();
            let self_clone = Arc::clone(self);
            let frozen_layer = Arc::clone(frozen_layer);
+            let ctx = ctx.attached_child();
            move || {
                // Write it out
                // Keep this inside `spawn_blocking` and `Handle::current`
                // as long as the write path is still sync and the read impl
                // is still not fully async. Otherwise executor threads would
                // be blocked.
-                let new_delta = Handle::current().block_on(frozen_layer.write_to_disk())?;
+                let new_delta = Handle::current().block_on(frozen_layer.write_to_disk(&ctx))?;
                let new_delta_path = new_delta.path();

                // Sync it to disk.
@@ -3574,7 +3630,7 @@ impl Timeline {
            key, lsn, ref val, ..
        } in all_values_iter
        {
-            let value = val.load().await?;
+            let value = val.load(ctx).await?;
            let same_key = prev_key.map_or(false, |prev_key| prev_key == key);
            // We need to check key boundaries once we reach next key or end of layer with the same key
            if !same_key || lsn == dup_end_lsn {
@@ -3672,6 +3728,11 @@ impl Timeline {
            });

            writer.as_mut().unwrap().put_value(key, lsn, value).await?;
+
+            if !new_layers.is_empty() {
+                fail_point!("after-timeline-compacted-first-L1");
+            }
+
            prev_key = Some(key);
        }
        if let Some(writer) = writer {
@@ -3693,7 +3754,7 @@ impl Timeline {
                    );
                }
            }
-            let mut layer_paths: Vec<PathBuf> = new_layers.iter().map(|l| l.path()).collect();
+            let mut layer_paths: Vec<Utf8PathBuf> = new_layers.iter().map(|l| l.path()).collect();

            // Fsync all the layer files and directory using multiple threads to
            // minimize latency.
@@ -3803,10 +3864,7 @@ impl Timeline {
            let new_delta_path = l.path();

            let metadata = new_delta_path.metadata().with_context(|| {
-                format!(
-                    "read file metadata for new created layer {}",
-                    new_delta_path.display()
-                )
+                format!("read file metadata for new created layer {new_delta_path}")
            })?;

            if let Some(remote_client) = &self.remote_client {
@@ -3829,6 +3887,7 @@ impl Timeline {
            );
            let l = l as Arc<dyn PersistentLayer>;
            if guard.contains(&l) {
+                tracing::error!(layer=%l, "duplicated L1 layer");
                duplicated_layers.insert(l.layer_desc().key());
            } else {
                if LayerMap::is_l0(l.layer_desc()) {
@@ -3862,7 +3921,7 @@ impl Timeline {

        // Also schedule the deletions in remote storage
        if let Some(remote_client) = &self.remote_client {
-            remote_client.schedule_layer_file_deletion(&layer_names_to_delete)?;
+            remote_client.schedule_layer_file_deletion(layer_names_to_delete)?;
        }

        Ok(())
@@ -4197,7 +4256,7 @@ impl Timeline {
            }

            if let Some(remote_client) = &self.remote_client {
-                remote_client.schedule_layer_file_deletion(&layer_names_to_delete)?;
+                remote_client.schedule_layer_file_deletion(layer_names_to_delete)?;
            }

            apply.flush();
@@ -4369,7 +4428,7 @@ impl Timeline {

                    // XXX the temp file is still around in Err() case
                    // and consumes space until we clean up upon pageserver restart.
-                    self_clone.metrics.resident_physical_size_gauge.add(*size);
+                    self_clone.metrics.resident_physical_size_add(*size);

                    // Download complete. Replace the RemoteLayer with the corresponding
                    // Delta- or ImageLayer in the layer map.
@@ -4699,8 +4758,14 @@ impl<'a> TimelineWriter<'a> {
    ///
    /// This will implicitly extend the relation, if the page is beyond the
    /// current end-of-file.
-    pub async fn put(&self, key: Key, lsn: Lsn, value: &Value) -> anyhow::Result<()> {
-        self.tl.put_value(key, lsn, value).await
+    pub async fn put(
+        &self,
+        key: Key,
+        lsn: Lsn,
+        value: &Value,
+        ctx: &RequestContext,
+    ) -> anyhow::Result<()> {
+        self.tl.put_value(key, lsn, value, ctx).await
    }

    pub async fn delete(&self, key_range: Range<Key>, lsn: Lsn) -> anyhow::Result<()> {
@@ -4734,11 +4799,10 @@ fn is_send() {

 /// Add a suffix to a layer file's name: .{num}.old
 /// Uses the first available num (starts at 0)
-fn rename_to_backup(path: &Path) -> anyhow::Result<()> {
+fn rename_to_backup(path: &Utf8Path) -> anyhow::Result<()> {
    let filename = path
        .file_name()
-        .ok_or_else(|| anyhow!("Path {} don't have a file name", path.display()))?
-        .to_string_lossy();
+        .ok_or_else(|| anyhow!("Path {path} don't have a file name"))?;
    let mut new_path = path.to_owned();

    for i in 0u32.. {
--- a/pageserver/src/tenant/timeline/delete.rs
+++ b/pageserver/src/tenant/timeline/delete.rs
@@ -14,6 +14,7 @@ use utils::{

 use crate::{
    config::PageServerConf,
+    deletion_queue::DeletionQueueClient,
    task_mgr::{self, TaskKind},
    tenant::{
        metadata::TimelineMetadata,
@@ -407,6 +408,7 @@ impl DeleteTimelineFlow {
        timeline_id: TimelineId,
        local_metadata: &TimelineMetadata,
        remote_client: Option<RemoteTimelineClient>,
+        deletion_queue_client: DeletionQueueClient,
        init_order: Option<&InitializationOrder>,
    ) -> anyhow::Result<()> {
        // Note: here we even skip populating layer map. Timeline is essentially uninitialized.
@@ -416,7 +418,10 @@ impl DeleteTimelineFlow {
                timeline_id,
                local_metadata,
                None, // Ancestor is not needed for deletion.
-                TimelineResources { remote_client },
+                TimelineResources {
+                    remote_client,
+                    deletion_queue_client,
+                },
                init_order,
                // Important. We dont pass ancestor above because it can be missing.
                // Thus we need to skip the validation here.
--- a/pageserver/src/tenant/timeline/init.rs
+++ b/pageserver/src/tenant/timeline/init.rs
@@ -12,7 +12,8 @@ use crate::{
    METADATA_FILE_NAME,
 };
 use anyhow::Context;
-use std::{collections::HashMap, ffi::OsString, path::Path, str::FromStr};
+use camino::Utf8Path;
+use std::{collections::HashMap, str::FromStr};
 use utils::lsn::Lsn;

 /// Identified files in the timeline directory.
@@ -20,46 +21,43 @@ pub(super) enum Discovered {
    /// The only one we care about
    Layer(LayerFileName, u64),
    /// Old ephmeral files from previous launches, should be removed
-    Ephemeral(OsString),
+    Ephemeral(String),
    /// Old temporary timeline files, unsure what these really are, should be removed
-    Temporary(OsString),
+    Temporary(String),
    /// Temporary on-demand download files, should be removed
-    TemporaryDownload(OsString),
+    TemporaryDownload(String),
    /// "metadata" file we persist locally and include in `index_part.json`
    Metadata,
    /// Backup file from previously future layers
    IgnoredBackup,
    /// Unrecognized, warn about these
-    Unknown(OsString),
+    Unknown(String),
 }

 /// Scans the timeline directory for interesting files.
-pub(super) fn scan_timeline_dir(path: &Path) -> anyhow::Result<Vec<Discovered>> {
+pub(super) fn scan_timeline_dir(path: &Utf8Path) -> anyhow::Result<Vec<Discovered>> {
    let mut ret = Vec::new();

-    for direntry in std::fs::read_dir(path)? {
+    for direntry in path.read_dir_utf8()? {
        let direntry = direntry?;
-        let direntry_path = direntry.path();
-        let file_name = direntry.file_name();
+        let file_name = direntry.file_name().to_string();

-        let fname = file_name.to_string_lossy();
-
-        let discovered = match LayerFileName::from_str(&fname) {
+        let discovered = match LayerFileName::from_str(&file_name) {
            Ok(file_name) => {
                let file_size = direntry.metadata()?.len();
                Discovered::Layer(file_name, file_size)
            }
            Err(_) => {
-                if fname == METADATA_FILE_NAME {
+                if file_name == METADATA_FILE_NAME {
                    Discovered::Metadata
-                } else if fname.ends_with(".old") {
+                } else if file_name.ends_with(".old") {
                    // ignore these
                    Discovered::IgnoredBackup
-                } else if remote_timeline_client::is_temp_download_file(&direntry_path) {
+                } else if remote_timeline_client::is_temp_download_file(direntry.path()) {
                    Discovered::TemporaryDownload(file_name)
-                } else if is_ephemeral_file(&fname) {
+                } else if is_ephemeral_file(&file_name) {
                    Discovered::Ephemeral(file_name)
-                } else if is_temporary(&direntry_path) {
+                } else if is_temporary(direntry.path()) {
                    Discovered::Temporary(file_name)
                } else {
                    Discovered::Unknown(file_name)
@@ -162,15 +160,14 @@ pub(super) fn reconcile(
        .collect::<Vec<_>>()
 }

-pub(super) fn cleanup(path: &Path, kind: &str) -> anyhow::Result<()> {
+pub(super) fn cleanup(path: &Utf8Path, kind: &str) -> anyhow::Result<()> {
    let file_name = path.file_name().expect("must be file path");
    tracing::debug!(kind, ?file_name, "cleaning up");
-    std::fs::remove_file(path)
-        .with_context(|| format!("failed to remove {kind} at {}", path.display()))
+    std::fs::remove_file(path).with_context(|| format!("failed to remove {kind} at {path}"))
 }

 pub(super) fn cleanup_local_file_for_remote(
-    path: &Path,
+    path: &Utf8Path,
    local: &LayerFileMetadata,
    remote: &LayerFileMetadata,
 ) -> anyhow::Result<()> {
@@ -182,8 +179,7 @@ pub(super) fn cleanup_local_file_for_remote(
    if let Err(err) = crate::tenant::timeline::rename_to_backup(path) {
        assert!(
            path.exists(),
-            "we would leave the local_layer without a file if this does not hold: {}",
-            path.display()
+            "we would leave the local_layer without a file if this does not hold: {path}",
        );
        Err(err)
    } else {
@@ -192,7 +188,7 @@ pub(super) fn cleanup_local_file_for_remote(
 }

 pub(super) fn cleanup_future_layer(
-    path: &Path,
+    path: &Utf8Path,
    name: &LayerFileName,
    disk_consistent_lsn: Lsn,
 ) -> anyhow::Result<()> {
--- a/pageserver/src/tenant/timeline/layer_manager.rs
+++ b/pageserver/src/tenant/timeline/layer_manager.rs
@@ -263,7 +263,7 @@ impl LayerManager {
        let desc = layer.layer_desc();
        if !layer.is_remote_layer() {
            layer.delete_resident_layer_file()?;
-            metrics.resident_physical_size_gauge.sub(desc.file_size);
+            metrics.resident_physical_size_sub(desc.file_size);
        }

        // TODO Removing from the bottom of the layer map is expensive.
--- a/pageserver/src/tenant/timeline/uninit.rs
+++ b/pageserver/src/tenant/timeline/uninit.rs
@@ -1,6 +1,7 @@
-use std::{collections::hash_map::Entry, fs, path::PathBuf, sync::Arc};
+use std::{collections::hash_map::Entry, fs, sync::Arc};

 use anyhow::Context;
+use camino::Utf8PathBuf;
 use tracing::{error, info, info_span, warn};
 use utils::{crashsafe, fs_ext, id::TimelineId, lsn::Lsn};

@@ -155,12 +156,12 @@ pub(crate) fn cleanup_timeline_directory(uninit_mark: TimelineUninitMark) {
 #[must_use]
 pub(crate) struct TimelineUninitMark {
    uninit_mark_deleted: bool,
-    uninit_mark_path: PathBuf,
-    pub(crate) timeline_path: PathBuf,
+    uninit_mark_path: Utf8PathBuf,
+    pub(crate) timeline_path: Utf8PathBuf,
 }

 impl TimelineUninitMark {
-    pub(crate) fn new(uninit_mark_path: PathBuf, timeline_path: PathBuf) -> Self {
+    pub(crate) fn new(uninit_mark_path: Utf8PathBuf, timeline_path: Utf8PathBuf) -> Self {
        Self {
            uninit_mark_deleted: false,
            uninit_mark_path,
@@ -197,14 +198,13 @@ impl Drop for TimelineUninitMark {
            if self.timeline_path.exists() {
                error!(
                    "Uninit mark {} is not removed, timeline {} stays uninitialized",
-                    self.uninit_mark_path.display(),
-                    self.timeline_path.display()
+                    self.uninit_mark_path, self.timeline_path
                )
            } else {
                // unblock later timeline creation attempts
                warn!(
                    "Removing intermediate uninit mark file {}",
-                    self.uninit_mark_path.display()
+                    self.uninit_mark_path
                );
                if let Err(e) = self.delete_mark_file_if_present() {
                    error!("Failed to remove the uninit mark file: {e}")
--- a/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs
+++ b/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs
@@ -370,8 +370,9 @@ pub(super) async fn handle_walreceiver_connection(
            })?;

        if let Some(last_lsn) = status_update {
-            let timeline_remote_consistent_lsn =
-                timeline.get_remote_consistent_lsn().unwrap_or(Lsn(0));
+            let timeline_remote_consistent_lsn = timeline
+                .get_remote_consistent_lsn_visible()
+                .unwrap_or(Lsn(0));

            // The last LSN we processed. It is not guaranteed to survive pageserver crash.
            let last_received_lsn = last_lsn;
--- a/Show More
+++ b/Show More