Implement mock console

[WIP] [proxy] Migrate to async
proxy chart staging values update for labels (#1202 )
2026-08-16 11:08:30 +00:00 · 2022-02-09 14:30:01 -05:00 · 2022-02-08 05:43:32 +03:00 · 2022-02-01 13:31:05 +03:00 · 2022-01-28 20:40:50 +03:00 · 2022-01-28 19:56:28 +03:00
108 changed files with 6227 additions and 3619 deletions
@@ -1,28 +1,28 @@
 version: 2.1
 executors:
-  zenith-build-executor:
+  zenith-xlarge-executor:
    resource_class: xlarge
    docker:
-      - image: cimg/rust:1.56.1
+      # NB: when changed, do not forget to update rust image tag in all Dockerfiles
-  zenith-python-executor:
+      - image: zimg/rust:1.56
  zenith-executor:
    docker:
-      - image: cimg/python:3.7.10  # Oldest available 3.7 with Ubuntu 20.04 (for GLIBC and Rust) at CirlceCI
+      - image: zimg/rust:1.56
 jobs:
  check-codestyle-rust:
-    executor: zenith-build-executor
+    executor: zenith-xlarge-executor
    steps:
      - checkout
      - run:
          name: rustfmt
          when: always
-          command: |
+          command: cargo fmt --all -- --check
            cargo fmt --all -- --check
  # A job to build postgres
  build-postgres:
-    executor: zenith-build-executor
+    executor: zenith-xlarge-executor
    parameters:
      build_type:
        type: enum
@@ -37,8 +37,7 @@ jobs:
        # Note this works even though the submodule hasn't been checkout out yet.
      - run:
          name: Get postgres cache key
-          command: |
+          command: git rev-parse HEAD:vendor/postgres > /tmp/cache-key-postgres
            git rev-parse HEAD:vendor/postgres > /tmp/cache-key-postgres
      - restore_cache:
          name: Restore postgres cache
@@ -46,15 +45,6 @@ jobs:
            # Restore ONLY if the rev key matches exactly
            - v04-postgres-cache-<< parameters.build_type >>-{{ checksum "/tmp/cache-key-postgres" }}
        # FIXME We could cache our own docker container, instead of installing packages every time.
      - run:
          name: apt install dependencies
          command: |
            if [ ! -e tmp_install/bin/postgres ]; then
              sudo apt update
              sudo apt install build-essential libreadline-dev zlib1g-dev flex bison libseccomp-dev
            fi
        # Build postgres if the restore_cache didn't find a build.
        # `make` can't figure out whether the cache is valid, since
        # it only compares file timestamps.
@@ -64,7 +54,7 @@ jobs:
            if [ ! -e tmp_install/bin/postgres ]; then
              # "depth 1" saves some time by not cloning the whole repo
              git submodule update --init --depth 1
-              make postgres -j8
+              mold -run make postgres -j$(nproc)
            fi
      - save_cache:
@@ -75,7 +65,7 @@ jobs:
  # A job to build zenith rust code
  build-zenith:
-    executor: zenith-build-executor
+    executor: zenith-xlarge-executor
    parameters:
      build_type:
        type: enum
@@ -83,12 +73,6 @@ jobs:
    environment:
      BUILD_TYPE: << parameters.build_type >>
    steps:
      - run:
          name: apt install dependencies
          command: |
            sudo apt update
            sudo apt install libssl-dev clang
        # Checkout the git repo (without submodules)
      - checkout
@@ -126,7 +110,7 @@ jobs:
            fi
            export CARGO_INCREMENTAL=0
-            "${cov_prefix[@]}" cargo build $CARGO_FLAGS --bins --tests
+            "${cov_prefix[@]}" mold -run cargo build $CARGO_FLAGS --bins --tests
      - save_cache:
          name: Save rust cache
@@ -210,6 +194,14 @@ jobs:
          command: |
            cp -a tmp_install /tmp/zenith/pg_install
      - run:
          name: Merge coverage data
          command: |
            # This will speed up workspace uploads
            if [[ $BUILD_TYPE == "debug" ]]; then
              scripts/coverage "--profraw-prefix=$CIRCLE_JOB" --dir=/tmp/zenith/coverage merge
            fi
        # Save the rust binaries and coverage data for other jobs in this workflow.
      - persist_to_workspace:
          root: /tmp/zenith
@@ -217,23 +209,30 @@ jobs:
            - "*"
  check-codestyle-python:
-    executor: zenith-python-executor
+    executor: zenith-executor
    steps:
      - checkout
      - restore_cache:
          keys:
            - v1-python-deps-{{ checksum "poetry.lock" }}
      - run:
          name: Install deps
-          command: pipenv --python 3.7 install --dev
+          command: ./scripts/pysync
      - save_cache:
          key: v1-python-deps-{{ checksum "poetry.lock" }}
          paths:
            - /home/circleci/.cache/pypoetry/virtualenvs
      - run:
          name: Run yapf to ensure code format
          when: always
-          command: pipenv run yapf --recursive --diff .
+          command: poetry run yapf --recursive --diff .
      - run:
          name: Run mypy to check types
          when: always
-          command: pipenv run mypy .
+          command: poetry run mypy .
  run-pytest:
-    executor: zenith-python-executor
+    executor: zenith-executor
    parameters:
      # pytest args to specify the tests to run.
      #
@@ -272,9 +271,16 @@ jobs:
          condition: << parameters.needs_postgres_source >>
          steps:
            - run: git submodule update --init --depth 1
      - restore_cache:
          keys:
            - v1-python-deps-{{ checksum "poetry.lock" }}
      - run:
          name: Install deps
-          command: pipenv --python 3.7 install
+          command: ./scripts/pysync
      - save_cache:
          key: v1-python-deps-{{ checksum "poetry.lock" }}
          paths:
            - /home/circleci/.cache/pypoetry/virtualenvs
      - run:
          name: Run pytest
          # pytest doesn't output test logs in real time, so CI job may fail with
@@ -326,7 +332,7 @@ jobs:
            # -n4 uses four processes to run tests via pytest-xdist
            # -s is not used to prevent pytest from capturing output, because tests are running
            # in parallel and logs are mixed between different tests
-            "${cov_prefix[@]}" pipenv run pytest \
+            "${cov_prefix[@]}" ./scripts/pytest \
              --junitxml=$TEST_OUTPUT/junit.xml \
              --tb=short \
              --verbose \
@@ -356,6 +362,13 @@ jobs:
      # The store_test_results step tells CircleCI where to find the junit.xml file.
      - store_test_results:
          path: /tmp/test_output
      - run:
          name: Merge coverage data
          command: |
            # This will speed up workspace uploads
            if [[ $BUILD_TYPE == "debug" ]]; then
              scripts/coverage "--profraw-prefix=$CIRCLE_JOB" --dir=/tmp/zenith/coverage merge
            fi
      # Save coverage data (if any)
      - persist_to_workspace:
          root: /tmp/zenith
@@ -363,7 +376,7 @@ jobs:
            - "*"
  coverage-report:
-    executor: zenith-build-executor
+    executor: zenith-xlarge-executor
    steps:
      - attach_workspace:
          at: /tmp/zenith
@@ -375,12 +388,6 @@ jobs:
            # there's no way to clean out old packages, so the cache grows every time something
            # changes.
            - v04-rust-cache-deps-debug-{{ checksum "Cargo.lock" }}
      - run:
          name: Install llvm-tools
          command: |
            # TODO: install a proper symbol demangler, e.g. rustfilt
            # TODO: we should embed this into a docker image
            rustup component add llvm-tools-preview
      - run:
          name: Build coverage report
          command: |
@@ -450,7 +457,7 @@ jobs:
          name: Build and push compute-tools Docker image
          command: |
            echo $DOCKER_PWD | docker login -u $DOCKER_LOGIN --password-stdin
-            docker build -t zenithdb/compute-tools:latest ./compute_tools/
+            docker build -t zenithdb/compute-tools:latest -f Dockerfile.compute-tools .
            docker push zenithdb/compute-tools:latest
      - run:
          name: Init postgres submodule
@@ -571,55 +578,6 @@ jobs:
                }
              }"
  #
  #
  # compute-tools jobs
  # TODO: unify with main build_and_test pipeline
  #
  #
  compute-tools-test:
    executor: zenith-build-executor
    working_directory: ~/repo/compute_tools
    steps:
      - checkout:
          path: ~/repo
      - restore_cache:
          name: Restore rust cache
          keys:
            # Require an exact match. While an out of date cache might speed up the build,
            # there's no way to clean out old packages, so the cache grows every time something
            # changes.
            - v03-rust-cache-deps-debug-{{ checksum "Cargo.lock" }}
      # Build the rust code, including test binaries
      - run:
          name: Rust build
          environment:
            CARGO_INCREMENTAL: 0
          command: cargo build --bins --tests
      - save_cache:
          name: Save rust cache
          key: v03-rust-cache-deps-debug-{{ checksum "Cargo.lock" }}
          paths:
            - ~/.cargo/registry
            - ~/.cargo/git
            - target
      # Run Rust formatting checks
      - run:
          name: cargo fmt check
          command: cargo fmt --all -- --check
      # Run Rust linter (clippy)
      - run:
          name: cargo clippy check
          command: cargo clippy --all --all-targets -- -Dwarnings -Drust-2018-idioms
      # Run Rust integration and unittests
      - run: cargo test
 workflows:
  build_and_test:
    jobs:
@@ -668,7 +626,6 @@ workflows:
          requires:
            # TODO: consider adding more
            - other-tests-debug
      - compute-tools-test
      - docker-image:
          # Context gives an ability to login
          context: Docker Hub
@@ -691,7 +648,6 @@ workflows:
          requires:
            - pg_regress-tests-release
            - other-tests-release
            - compute-tools-test
      - deploy-staging:
          # Context gives an ability to login
          context: Docker Hub
@@ -5,9 +5,23 @@ settings:
  authEndpoint: "https://console.stage.zenith.tech/authenticate_proxy_request/"
  uri: "https://console.stage.zenith.tech/psql_session/"
 # -- Additional labels for zenith-proxy pods
 podLabels:
  zenith_service: proxy
  zenith_env: staging
  zenith_region: us-east-1
  zenith_region_slug: virginia
 exposedService:
  annotations:
    service.beta.kubernetes.io/aws-load-balancer-type: external
    service.beta.kubernetes.io/aws-load-balancer-nlb-target-type: ip
    service.beta.kubernetes.io/aws-load-balancer-scheme: internet-facing
    external-dns.alpha.kubernetes.io/hostname: start.stage.zenith.tech
 metrics:
  enabled: true
  serviceMonitor:
    enabled: true
    selector:
      release: kube-prometheus-stack
@@ -36,20 +36,20 @@ jobs:
    # see https://github.com/actions/setup-python/issues/162
    # and probably https://github.com/actions/setup-python/issues/162#issuecomment-865387976 in particular
    # so the simplest solution to me is to use already installed system python and spin virtualenvs for job runs.
-    # there is Python 3.7.10 already installed on the machine so use it to install pipenv and then use pipenv's virtuealenvs
+    # there is Python 3.7.10 already installed on the machine so use it to install poetry and then use poetry's virtuealenvs
-    - name: Install pipenv & deps
+    - name: Install poetry & deps
      run: |
-        python3 -m pip install --upgrade pipenv wheel
+        python3 -m pip install --upgrade poetry wheel
-        # since pip/pipenv caches are reused there shouldn't be any troubles with install every time
+        # since pip/poetry caches are reused there shouldn't be any troubles with install every time
-        pipenv install
+        poetry install
    - name: Show versions
      run: |
        echo Python
        python3 --version
-        pipenv run python3 --version
+        poetry run python3 --version
        echo Pipenv
-        pipenv --version
+        poetry --version
        echo Pgbench
        $PG_BIN/pgbench --version
@@ -90,7 +90,7 @@ jobs:
        REMOTE_ENV: "1" # indicate to test harness that we do not have zenith binaries locally
      run: |
        mkdir -p perf-report-staging
-        pipenv run pytest test_runner/performance/ -v -m "remote_cluster" --skip-interfering-proc-check --out-dir perf-report-staging
+        ./scripts/pytest test_runner/performance/ -v -m "remote_cluster" --skip-interfering-proc-check --out-dir perf-report-staging
    - name: Submit result
      env:
@@ -2,12 +2,38 @@
 # It is not intended for manual editing.
 version = 3
 [[package]]
 name = "addr2line"
 version = "0.17.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b9ecd88a8c8378ca913a680cd98f0f13ac67383d35993f86c90a70e3f137816b"
 dependencies = [
 "gimli",
 ]
 [[package]]
 name = "adler"
 version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f26201604c87b1e01bd3d98f8d5d9a8fcbb815e8cedb41ffccbeb4bf593a35fe"
 [[package]]
 name = "ahash"
 version = "0.4.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "739f4a8db6605981345c5654f3a85b056ce52f37a39d34da03f25bf2151ea16e"
 [[package]]
 name = "ahash"
 version = "0.7.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "fcb51a0695d8f838b1ee009b3fbf66bda078cd64590202a864a8f3e8c4315c47"
 dependencies = [
 "getrandom",
 "once_cell",
 "version_check",
 ]
 [[package]]
 name = "aho-corasick"
 version = "0.7.18"
@@ -40,6 +66,9 @@ name = "anyhow"
 version = "1.0.44"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "61604a8f862e1d5c3229fdd78f8b02c68dcf73a4c4b05fd636d12240aaa242c1"
 dependencies = [
 "backtrace",
 ]
 [[package]]
 name = "async-compression"
@@ -57,9 +86,9 @@ dependencies = [
 [[package]]
 name = "async-trait"
-version = "0.1.51"
+version = "0.1.52"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "44318e776df68115a881de9a8fd1b9e53368d7a4a5ce4cc48517da3393233a5e"
+checksum = "061a7acccaa286c011ddc30970520b98fa40e00c9d644633fb26b5fc63a265e3"
 dependencies = [
 "proc-macro2",
 "quote",
@@ -149,6 +178,21 @@ dependencies = [
 "anyhow",
 ]
 [[package]]
 name = "backtrace"
 version = "0.3.63"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "321629d8ba6513061f26707241fa9bc89524ff1cd7a915a97ef0c62c666ce1b6"
 dependencies = [
 "addr2line",
 "cc",
 "cfg-if",
 "libc",
 "miniz_oxide",
 "object",
 "rustc-demangle",
 ]
 [[package]]
 name = "base64"
 version = "0.12.3"
@@ -331,6 +375,25 @@ dependencies = [
 "memchr",
 ]
 [[package]]
 name = "compute_tools"
 version = "0.1.0"
 dependencies = [
 "anyhow",
 "chrono",
 "clap",
 "env_logger",
 "hyper",
 "libc",
 "log",
 "postgres 0.19.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858)",
 "regex",
 "serde",
 "serde_json",
 "tar",
 "tokio",
 ]
 [[package]]
 name = "const_format"
 version = "0.2.22"
@@ -359,7 +422,7 @@ dependencies = [
 "lazy_static",
 "nix",
 "pageserver",
- "postgres",
+ "postgres 0.19.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "regex",
 "reqwest",
 "serde",
@@ -655,6 +718,12 @@ dependencies = [
 "wasi",
 ]
 [[package]]
 name = "gimli"
 version = "0.26.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "78cc372d058dcf6d5ecd98510e7fbc9e5aec4d21de70f65fea8fecebcd881bd4"
 [[package]]
 name = "git-version"
 version = "0.3.5"
@@ -714,7 +783,7 @@ version = "0.9.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d7afe4a420e3fe79967a00898cc1f4db7c8a49a9333a29f8a4bd76a253d5cd04"
 dependencies = [
- "ahash",
+ "ahash 0.4.7",
 ]
 [[package]]
@@ -722,6 +791,9 @@ name = "hashbrown"
 version = "0.11.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ab5ef0d4909ef3724cc8cce6ccc8572c5c817592e9285f5464f8e86f8bd3726e"
 dependencies = [
 "ahash 0.7.6",
 ]
 [[package]]
 name = "hermit-abi"
@@ -922,7 +994,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "afabcc15e437a6484fc4f12d0fd63068fe457bf93f1c148d3d9649c60b103f32"
 dependencies = [
 "base64 0.12.3",
- "pem",
+ "pem 0.8.3",
 "ring",
 "serde",
 "serde_json",
@@ -982,6 +1054,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "51b9bbe6c47d51fc3e1a9b945965946b4c44142ab8792c50835a980d362c2710"
 dependencies = [
 "cfg-if",
 "serde",
 ]
 [[package]]
@@ -1048,6 +1121,16 @@ version = "0.3.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2a60c7ce501c71e03a9c9c0d35b861413ae925bd979cc7a4e30d060069aaac8d"
 [[package]]
 name = "miniz_oxide"
 version = "0.4.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a92518e98c078586bc6c934028adcca4c92a53d6a958196de835170a01d84e4b"
 dependencies = [
 "adler",
 "autocfg",
 ]
 [[package]]
 name = "mio"
 version = "0.7.13"
@@ -1144,6 +1227,15 @@ dependencies = [
 "libc",
 ]
 [[package]]
 name = "object"
 version = "0.27.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "67ac1d3f9a1d3616fd9a60c8d74296f22406a238b6a72f5cc1e6f314df4ffbf9"
 dependencies = [
 "memchr",
 ]
 [[package]]
 name = "once_cell"
 version = "1.8.0"
@@ -1192,9 +1284,9 @@ dependencies = [
 "nix",
 "once_cell",
 "parking_lot",
- "postgres",
+ "postgres 0.19.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
- "postgres-protocol",
+ "postgres-protocol 0.6.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
- "postgres-types",
+ "postgres-types 0.2.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "postgres_ffi",
 "rand",
 "regex",
@@ -1208,7 +1300,7 @@ dependencies = [
 "tempfile",
 "thiserror",
 "tokio",
- "tokio-postgres",
+ "tokio-postgres 0.7.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "tokio-stream",
 "toml_edit",
 "tracing",
@@ -1261,6 +1353,15 @@ dependencies = [
 "regex",
 ]
 [[package]]
 name = "pem"
 version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e9a3b09a20e374558580a4914d3b7d89bd61b954a5a5e1dcbea98753addb1947"
 dependencies = [
 "base64 0.13.0",
 ]
 [[package]]
 name = "percent-encoding"
 version = "2.1.0"
@@ -1317,6 +1418,20 @@ version = "0.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"
 [[package]]
 name = "postgres"
 version = "0.19.1"
 source = "git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7#2949d98df52587d562986aad155dd4e889e408b7"
 dependencies = [
 "bytes",
 "fallible-iterator",
 "futures",
 "log",
 "postgres-protocol 0.6.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "tokio",
 "tokio-postgres 0.7.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 ]
 [[package]]
 name = "postgres"
 version = "0.19.1"
@@ -1326,9 +1441,27 @@ dependencies = [
 "fallible-iterator",
 "futures",
 "log",
- "postgres-protocol",
+ "postgres-protocol 0.6.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858)",
 "tokio",
- "tokio-postgres",
+ "tokio-postgres 0.7.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858)",
 ]
 [[package]]
 name = "postgres-protocol"
 version = "0.6.1"
 source = "git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7#2949d98df52587d562986aad155dd4e889e408b7"
 dependencies = [
 "base64 0.13.0",
 "byteorder",
 "bytes",
 "fallible-iterator",
 "hmac 0.10.1",
 "lazy_static",
 "md-5",
 "memchr",
 "rand",
 "sha2",
 "stringprep",
 ]
 [[package]]
@@ -1349,6 +1482,16 @@ dependencies = [
 "stringprep",
 ]
 [[package]]
 name = "postgres-types"
 version = "0.2.1"
 source = "git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7#2949d98df52587d562986aad155dd4e889e408b7"
 dependencies = [
 "bytes",
 "fallible-iterator",
 "postgres-protocol 0.6.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 ]
 [[package]]
 name = "postgres-types"
 version = "0.2.1"
@@ -1356,7 +1499,7 @@ source = "git+https://github.com/zenithdb/rust-postgres.git?rev=9eb0dbfbeb6a6c1b
 dependencies = [
 "bytes",
 "fallible-iterator",
- "postgres-protocol",
+ "postgres-protocol 0.6.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858)",
 ]
 [[package]]
@@ -1421,21 +1564,30 @@ name = "proxy"
 version = "0.1.0"
 dependencies = [
 "anyhow",
 "async-trait",
 "base64 0.13.0",
 "bytes",
 "clap",
 "futures",
 "hashbrown 0.11.2",
 "hex",
 "hyper",
 "lazy_static",
 "md5",
 "parking_lot",
 "pin-project-lite",
 "rand",
 "rcgen",
 "reqwest",
 "routerify",
 "rustls 0.19.1",
 "scopeguard",
 "serde",
 "serde_json",
 "tokio",
- "tokio-postgres",
+ "tokio-postgres 0.7.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "tokio-postgres-rustls",
 "tokio-rustls",
 "zenith_metrics",
 "zenith_utils",
 ]
@@ -1495,6 +1647,18 @@ dependencies = [
 "rand_core",
 ]
 [[package]]
 name = "rcgen"
 version = "0.8.14"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5911d1403f4143c9d56a702069d593e8d0f3fab880a85e103604d0893ea31ba7"
 dependencies = [
 "chrono",
 "pem 1.0.2",
 "ring",
 "yasna",
 ]
 [[package]]
 name = "redox_syscall"
 version = "0.2.10"
@@ -1653,6 +1817,12 @@ dependencies = [
 "url",
 ]
 [[package]]
 name = "rustc-demangle"
 version = "0.1.21"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7ef03e0a2b150c7a90d01faf6254c9c48a41e95fb2a8c2ac1c6f0d2b9aefc342"
 [[package]]
 name = "rustc-hash"
 version = "1.1.0"
@@ -2082,6 +2252,28 @@ dependencies = [
 "syn",
 ]
 [[package]]
 name = "tokio-postgres"
 version = "0.7.1"
 source = "git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7#2949d98df52587d562986aad155dd4e889e408b7"
 dependencies = [
 "async-trait",
 "byteorder",
 "bytes",
 "fallible-iterator",
 "futures",
 "log",
 "parking_lot",
 "percent-encoding",
 "phf",
 "pin-project-lite",
 "postgres-protocol 0.6.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "postgres-types 0.2.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "socket2",
 "tokio",
 "tokio-util",
 ]
 [[package]]
 name = "tokio-postgres"
 version = "0.7.1"
@@ -2097,13 +2289,28 @@ dependencies = [
 "percent-encoding",
 "phf",
 "pin-project-lite",
- "postgres-protocol",
+ "postgres-protocol 0.6.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858)",
- "postgres-types",
+ "postgres-types 0.2.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858)",
 "socket2",
 "tokio",
 "tokio-util",
 ]
 [[package]]
 name = "tokio-postgres-rustls"
 version = "0.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7bd8c37d8c23cb6ecdc32fc171bade4e9c7f1be65f693a17afbaad02091a0a19"
 dependencies = [
 "futures",
 "ring",
 "rustls 0.19.1",
 "tokio",
 "tokio-postgres 0.7.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "tokio-rustls",
 "webpki 0.21.4",
 ]
 [[package]]
 name = "tokio-rustls"
 version = "0.22.0"
@@ -2335,8 +2542,8 @@ dependencies = [
 "humantime",
 "hyper",
 "lazy_static",
- "postgres",
+ "postgres 0.19.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
- "postgres-protocol",
+ "postgres-protocol 0.6.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "postgres_ffi",
 "regex",
 "routerify",
@@ -2346,7 +2553,7 @@ dependencies = [
 "signal-hook",
 "tempfile",
 "tokio",
- "tokio-postgres",
+ "tokio-postgres 0.7.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "tracing",
 "walkdir",
 "workspace_hack",
@@ -2576,6 +2783,15 @@ version = "0.8.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d2d7d3948613f75c98fd9328cfdcc45acc4d360655289d0a7d4ec931392200a3"
 [[package]]
 name = "yasna"
 version = "0.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e262a29d0e61ccf2b6190d7050d4b237535fc76ce4c1210d9caa316f71dffa75"
 dependencies = [
 "chrono",
 ]
 [[package]]
 name = "zenith"
 version = "0.1.0"
@@ -2584,7 +2800,7 @@ dependencies = [
 "clap",
 "control_plane",
 "pageserver",
- "postgres",
+ "postgres 0.19.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "postgres_ffi",
 "serde_json",
 "walkeeper",
@@ -2617,7 +2833,9 @@ dependencies = [
 "jsonwebtoken",
 "lazy_static",
 "nix",
- "postgres",
+ "pin-project-lite",
 "postgres 0.19.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "postgres-protocol 0.6.1 (git+https://github.com/zenithdb/rust-postgres.git?rev=2949d98df52587d562986aad155dd4e889e408b7)",
 "rand",
 "routerify",
 "rustls 0.19.1",
@@ -1,5 +1,6 @@
 [workspace]
 members = [
    "compute_tools",
    "control_plane",
    "pageserver",
    "postgres_ffi",
@@ -15,3 +16,8 @@ members = [
 # This is useful for profiling and, to some extent, debug.
 # Besides, debug info should not affect the performance.
 debug = true
 # This is only needed for proxy's tests
 # TODO: we should probably fork tokio-postgres-rustls instead
 [patch.crates-io]
 tokio-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
@@ -2,8 +2,9 @@
 # Image with all the required dependencies to build https://github.com/zenithdb/zenith
 # and Postgres from https://github.com/zenithdb/postgres
 # Also includes some rust development and build tools.
 # NB: keep in sync with rust image version in .circle/config.yml
 #
-FROM rust:slim-buster
+FROM rust:1.56.1-slim-buster
 WORKDIR /zenith
 # Install postgres and zenith build dependencies
@@ -0,0 +1,14 @@
 # First transient image to build compute_tools binaries
 # NB: keep in sync with rust image version in .circle/config.yml
 FROM rust:1.56.1-slim-buster AS rust-build
 WORKDIR /zenith
 COPY . .
 RUN cargo build -p compute_tools --release
 # Final image that only has one binary
 FROM debian:buster-slim
 COPY --from=rust-build /zenith/target/release/zenith_ctl /usr/local/bin/zenith_ctl
@@ -1,30 +0,0 @@
 [[source]]
 url = "https://pypi.python.org/simple"
 verify_ssl = true
 name = "pypi"
 [packages]
 pytest = ">=6.0.0"
 typing-extensions = "*"
 pyjwt = {extras = ["crypto"], version = "*"}
 requests = "*"
 pytest-xdist = "*"
 asyncpg = "*"
 cached-property = "*"
 psycopg2-binary = "*"
 jinja2 = "*"
 [dev-packages]
 # Behavior may change slightly between versions. These are run continuously,
 # so we pin exact versions to avoid suprising breaks. Update if comfortable.
 yapf = "==0.31.0"
 mypy = "==0.910"
 # Non-pinned packages follow.
 pipenv = "*"
 flake8 = "*"
 types-requests = "*"
 types-psycopg2 = "*"
 [requires]
 # we need at least 3.7, but pipenv doesn't allow to say this directly
 python_version = "3"
@@ -1,652 +0,0 @@
 {
    "_meta": {
        "hash": {
            "sha256": "c309cb963a7b07ae3d30e9cbf08b495f77bdecc0e5356fc89d133c4fbcb65b2b"
        },
        "pipfile-spec": 6,
        "requires": {
            "python_version": "3"
        },
        "sources": [
            {
                "name": "pypi",
                "url": "https://pypi.python.org/simple",
                "verify_ssl": true
            }
        ]
    },
    "default": {
        "asyncpg": {
            "hashes": [
                "sha256:129d501f3d30616afd51eb8d3142ef51ba05374256bd5834cec3ef4956a9b317",
                "sha256:29ef6ae0a617fc13cc2ac5dc8e9b367bb83cba220614b437af9b67766f4b6b20",
                "sha256:41704c561d354bef01353835a7846e5606faabbeb846214dfcf666cf53319f18",
                "sha256:556b0e92e2b75dc028b3c4bc9bd5162ddf0053b856437cf1f04c97f9c6837d03",
                "sha256:8ff5073d4b654e34bd5eaadc01dc4d68b8a9609084d835acd364cd934190a08d",
                "sha256:a458fc69051fbb67d995fdda46d75a012b5d6200f91e17d23d4751482640ed4c",
                "sha256:a7095890c96ba36f9f668eb552bb020dddb44f8e73e932f8573efc613ee83843",
                "sha256:a738f4807c853623d3f93f0fea11f61be6b0e5ca16ea8aeb42c2c7ee742aa853",
                "sha256:c4fc0205fe4ddd5aeb3dfdc0f7bafd43411181e1f5650189608e5971cceacff1",
                "sha256:dd2fa063c3344823487d9ddccb40802f02622ddf8bf8a6cc53885ee7a2c1c0c6",
                "sha256:ddffcb85227bf39cd1bedd4603e0082b243cf3b14ced64dce506a15b05232b83",
                "sha256:e36c6806883786b19551bb70a4882561f31135dc8105a59662e0376cf5b2cbc5",
                "sha256:eed43abc6ccf1dc02e0d0efc06ce46a411362f3358847c6b0ec9a43426f91ece"
            ],
            "index": "pypi",
            "version": "==0.24.0"
        },
        "attrs": {
            "hashes": [
                "sha256:149e90d6d8ac20db7a955ad60cf0e6881a3f20d37096140088356da6c716b0b1",
                "sha256:ef6aaac3ca6cd92904cdd0d83f629a15f18053ec84e6432106f7a4d04ae4f5fb"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4'",
            "version": "==21.2.0"
        },
        "cached-property": {
            "hashes": [
                "sha256:9fa5755838eecbb2d234c3aa390bd80fbd3ac6b6869109bfc1b499f7bd89a130",
                "sha256:df4f613cf7ad9a588cc381aaf4a512d26265ecebd5eb9e1ba12f1319eb85a6a0"
            ],
            "index": "pypi",
            "version": "==1.5.2"
        },
        "certifi": {
            "hashes": [
                "sha256:78884e7c1d4b00ce3cea67b44566851c4343c120abd683433ce934a68ea58872",
                "sha256:d62a0163eb4c2344ac042ab2bdf75399a71a2d8c7d47eac2e2ee91b9d6339569"
            ],
            "version": "==2021.10.8"
        },
        "cffi": {
            "hashes": [
                "sha256:00c878c90cb53ccfaae6b8bc18ad05d2036553e6d9d1d9dbcf323bbe83854ca3",
                "sha256:0104fb5ae2391d46a4cb082abdd5c69ea4eab79d8d44eaaf79f1b1fd806ee4c2",
                "sha256:06c48159c1abed75c2e721b1715c379fa3200c7784271b3c46df01383b593636",
                "sha256:0808014eb713677ec1292301ea4c81ad277b6cdf2fdd90fd540af98c0b101d20",
                "sha256:10dffb601ccfb65262a27233ac273d552ddc4d8ae1bf93b21c94b8511bffe728",
                "sha256:14cd121ea63ecdae71efa69c15c5543a4b5fbcd0bbe2aad864baca0063cecf27",
                "sha256:17771976e82e9f94976180f76468546834d22a7cc404b17c22df2a2c81db0c66",
                "sha256:181dee03b1170ff1969489acf1c26533710231c58f95534e3edac87fff06c443",
                "sha256:23cfe892bd5dd8941608f93348c0737e369e51c100d03718f108bf1add7bd6d0",
                "sha256:263cc3d821c4ab2213cbe8cd8b355a7f72a8324577dc865ef98487c1aeee2bc7",
                "sha256:2756c88cbb94231c7a147402476be2c4df2f6078099a6f4a480d239a8817ae39",
                "sha256:27c219baf94952ae9d50ec19651a687b826792055353d07648a5695413e0c605",
                "sha256:2a23af14f408d53d5e6cd4e3d9a24ff9e05906ad574822a10563efcef137979a",
                "sha256:31fb708d9d7c3f49a60f04cf5b119aeefe5644daba1cd2a0fe389b674fd1de37",
                "sha256:3415c89f9204ee60cd09b235810be700e993e343a408693e80ce7f6a40108029",
                "sha256:3773c4d81e6e818df2efbc7dd77325ca0dcb688116050fb2b3011218eda36139",
                "sha256:3b96a311ac60a3f6be21d2572e46ce67f09abcf4d09344c49274eb9e0bf345fc",
                "sha256:3f7d084648d77af029acb79a0ff49a0ad7e9d09057a9bf46596dac9514dc07df",
                "sha256:41d45de54cd277a7878919867c0f08b0cf817605e4eb94093e7516505d3c8d14",
                "sha256:4238e6dab5d6a8ba812de994bbb0a79bddbdf80994e4ce802b6f6f3142fcc880",
                "sha256:45db3a33139e9c8f7c09234b5784a5e33d31fd6907800b316decad50af323ff2",
                "sha256:45e8636704eacc432a206ac7345a5d3d2c62d95a507ec70d62f23cd91770482a",
                "sha256:4958391dbd6249d7ad855b9ca88fae690783a6be9e86df65865058ed81fc860e",
                "sha256:4a306fa632e8f0928956a41fa8e1d6243c71e7eb59ffbd165fc0b41e316b2474",
                "sha256:57e9ac9ccc3101fac9d6014fba037473e4358ef4e89f8e181f8951a2c0162024",
                "sha256:59888172256cac5629e60e72e86598027aca6bf01fa2465bdb676d37636573e8",
                "sha256:5e069f72d497312b24fcc02073d70cb989045d1c91cbd53979366077959933e0",
                "sha256:64d4ec9f448dfe041705426000cc13e34e6e5bb13736e9fd62e34a0b0c41566e",
                "sha256:6dc2737a3674b3e344847c8686cf29e500584ccad76204efea14f451d4cc669a",
                "sha256:74fdfdbfdc48d3f47148976f49fab3251e550a8720bebc99bf1483f5bfb5db3e",
                "sha256:75e4024375654472cc27e91cbe9eaa08567f7fbdf822638be2814ce059f58032",
                "sha256:786902fb9ba7433aae840e0ed609f45c7bcd4e225ebb9c753aa39725bb3e6ad6",
                "sha256:8b6c2ea03845c9f501ed1313e78de148cd3f6cad741a75d43a29b43da27f2e1e",
                "sha256:91d77d2a782be4274da750752bb1650a97bfd8f291022b379bb8e01c66b4e96b",
                "sha256:91ec59c33514b7c7559a6acda53bbfe1b283949c34fe7440bcf917f96ac0723e",
                "sha256:920f0d66a896c2d99f0adbb391f990a84091179542c205fa53ce5787aff87954",
                "sha256:a5263e363c27b653a90078143adb3d076c1a748ec9ecc78ea2fb916f9b861962",
                "sha256:abb9a20a72ac4e0fdb50dae135ba5e77880518e742077ced47eb1499e29a443c",
                "sha256:c2051981a968d7de9dd2d7b87bcb9c939c74a34626a6e2f8181455dd49ed69e4",
                "sha256:c21c9e3896c23007803a875460fb786118f0cdd4434359577ea25eb556e34c55",
                "sha256:c2502a1a03b6312837279c8c1bd3ebedf6c12c4228ddbad40912d671ccc8a962",
                "sha256:d4d692a89c5cf08a8557fdeb329b82e7bf609aadfaed6c0d79f5a449a3c7c023",
                "sha256:da5db4e883f1ce37f55c667e5c0de439df76ac4cb55964655906306918e7363c",
                "sha256:e7022a66d9b55e93e1a845d8c9eba2a1bebd4966cd8bfc25d9cd07d515b33fa6",
                "sha256:ef1f279350da2c586a69d32fc8733092fd32cc8ac95139a00377841f59a3f8d8",
                "sha256:f54a64f8b0c8ff0b64d18aa76675262e1700f3995182267998c31ae974fbc382",
                "sha256:f5c7150ad32ba43a07c4479f40241756145a1f03b43480e058cfd862bf5041c7",
                "sha256:f6f824dc3bce0edab5f427efcfb1d63ee75b6fcb7282900ccaf925be84efb0fc",
                "sha256:fd8a250edc26254fe5b33be00402e6d287f562b6a5b2152dec302fa15bb3e997",
                "sha256:ffaa5c925128e29efbde7301d8ecaf35c8c60ffbcd6a1ffd3a552177c8e5e796"
            ],
            "version": "==1.15.0"
        },
        "charset-normalizer": {
            "hashes": [
                "sha256:e019de665e2bcf9c2b64e2e5aa025fa991da8720daa3c1138cadd2fd1856aed0",
                "sha256:f7af805c321bfa1ce6714c51f254e0d5bb5e5834039bc17db7ebe3a4cec9492b"
            ],
            "markers": "python_version >= '3'",
            "version": "==2.0.7"
        },
        "cryptography": {
            "hashes": [
                "sha256:07bb7fbfb5de0980590ddfc7f13081520def06dc9ed214000ad4372fb4e3c7f6",
                "sha256:18d90f4711bf63e2fb21e8c8e51ed8189438e6b35a6d996201ebd98a26abbbe6",
                "sha256:1ed82abf16df40a60942a8c211251ae72858b25b7421ce2497c2eb7a1cee817c",
                "sha256:22a38e96118a4ce3b97509443feace1d1011d0571fae81fc3ad35f25ba3ea999",
                "sha256:2d69645f535f4b2c722cfb07a8eab916265545b3475fdb34e0be2f4ee8b0b15e",
                "sha256:4a2d0e0acc20ede0f06ef7aa58546eee96d2592c00f450c9acb89c5879b61992",
                "sha256:54b2605e5475944e2213258e0ab8696f4f357a31371e538ef21e8d61c843c28d",
                "sha256:7075b304cd567694dc692ffc9747f3e9cb393cc4aa4fb7b9f3abd6f5c4e43588",
                "sha256:7b7ceeff114c31f285528ba8b390d3e9cfa2da17b56f11d366769a807f17cbaa",
                "sha256:7eba2cebca600a7806b893cb1d541a6e910afa87e97acf2021a22b32da1df52d",
                "sha256:928185a6d1ccdb816e883f56ebe92e975a262d31cc536429041921f8cb5a62fd",
                "sha256:9933f28f70d0517686bd7de36166dda42094eac49415459d9bdf5e7df3e0086d",
                "sha256:a688ebcd08250eab5bb5bca318cc05a8c66de5e4171a65ca51db6bd753ff8953",
                "sha256:abb5a361d2585bb95012a19ed9b2c8f412c5d723a9836418fab7aaa0243e67d2",
                "sha256:c10c797ac89c746e488d2ee92bd4abd593615694ee17b2500578b63cad6b93a8",
                "sha256:ced40344e811d6abba00295ced98c01aecf0c2de39481792d87af4fa58b7b4d6",
                "sha256:d57e0cdc1b44b6cdf8af1d01807db06886f10177469312fbde8f44ccbb284bc9",
                "sha256:d99915d6ab265c22873f1b4d6ea5ef462ef797b4140be4c9d8b179915e0985c6",
                "sha256:eb80e8a1f91e4b7ef8b33041591e6d89b2b8e122d787e87eeb2b08da71bb16ad",
                "sha256:ebeddd119f526bcf323a89f853afb12e225902a24d29b55fe18dd6fcb2838a76"
            ],
            "version": "==35.0.0"
        },
        "execnet": {
            "hashes": [
                "sha256:8f694f3ba9cc92cab508b152dcfe322153975c29bda272e2fd7f3f00f36e47c5",
                "sha256:a295f7cc774947aac58dde7fdc85f4aa00c42adf5d8f5468fc630c1acf30a142"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4'",
            "version": "==1.9.0"
        },
        "idna": {
            "hashes": [
                "sha256:84d9dd047ffa80596e0f246e2eab0b391788b0503584e8945f2368256d2735ff",
                "sha256:9d643ff0a55b762d5cdb124b8eaa99c66322e2157b69160bc32796e824360e6d"
            ],
            "markers": "python_version >= '3'",
            "version": "==3.3"
        },
        "importlib-metadata": {
            "hashes": [
                "sha256:b618b6d2d5ffa2f16add5697cf57a46c76a56229b0ed1c438322e4e95645bd15",
                "sha256:f284b3e11256ad1e5d03ab86bb2ccd6f5339688ff17a4d797a0fe7df326f23b1"
            ],
            "markers": "python_version < '3.8'",
            "version": "==4.8.1"
        },
        "iniconfig": {
            "hashes": [
                "sha256:011e24c64b7f47f6ebd835bb12a743f2fbe9a26d4cecaa7f53bc4f35ee9da8b3",
                "sha256:bc3af051d7d14b2ee5ef9969666def0cd1a000e121eaea580d4a313df4b37f32"
            ],
            "version": "==1.1.1"
        },
        "jinja2": {
            "hashes": [
                "sha256:827a0e32839ab1600d4eb1c4c33ec5a8edfbc5cb42dafa13b81f182f97784b45",
                "sha256:8569982d3f0889eed11dd620c706d39b60c36d6d25843961f33f77fb6bc6b20c"
            ],
            "index": "pypi",
            "version": "==3.0.2"
        },
        "markupsafe": {
            "hashes": [
                "sha256:01a9b8ea66f1658938f65b93a85ebe8bc016e6769611be228d797c9d998dd298",
                "sha256:023cb26ec21ece8dc3907c0e8320058b2e0cb3c55cf9564da612bc325bed5e64",
                "sha256:0446679737af14f45767963a1a9ef7620189912317d095f2d9ffa183a4d25d2b",
                "sha256:04635854b943835a6ea959e948d19dcd311762c5c0c6e1f0e16ee57022669194",
                "sha256:0717a7390a68be14b8c793ba258e075c6f4ca819f15edfc2a3a027c823718567",
                "sha256:0955295dd5eec6cb6cc2fe1698f4c6d84af2e92de33fbcac4111913cd100a6ff",
                "sha256:0d4b31cc67ab36e3392bbf3862cfbadac3db12bdd8b02a2731f509ed5b829724",
                "sha256:10f82115e21dc0dfec9ab5c0223652f7197feb168c940f3ef61563fc2d6beb74",
                "sha256:168cd0a3642de83558a5153c8bd34f175a9a6e7f6dc6384b9655d2697312a646",
                "sha256:1d609f577dc6e1aa17d746f8bd3c31aa4d258f4070d61b2aa5c4166c1539de35",
                "sha256:1f2ade76b9903f39aa442b4aadd2177decb66525062db244b35d71d0ee8599b6",
                "sha256:20dca64a3ef2d6e4d5d615a3fd418ad3bde77a47ec8a23d984a12b5b4c74491a",
                "sha256:2a7d351cbd8cfeb19ca00de495e224dea7e7d919659c2841bbb7f420ad03e2d6",
                "sha256:2d7d807855b419fc2ed3e631034685db6079889a1f01d5d9dac950f764da3dad",
                "sha256:2ef54abee730b502252bcdf31b10dacb0a416229b72c18b19e24a4509f273d26",
                "sha256:36bc903cbb393720fad60fc28c10de6acf10dc6cc883f3e24ee4012371399a38",
                "sha256:37205cac2a79194e3750b0af2a5720d95f786a55ce7df90c3af697bfa100eaac",
                "sha256:3c112550557578c26af18a1ccc9e090bfe03832ae994343cfdacd287db6a6ae7",
                "sha256:3dd007d54ee88b46be476e293f48c85048603f5f516008bee124ddd891398ed6",
                "sha256:4296f2b1ce8c86a6aea78613c34bb1a672ea0e3de9c6ba08a960efe0b0a09047",
                "sha256:47ab1e7b91c098ab893b828deafa1203de86d0bc6ab587b160f78fe6c4011f75",
                "sha256:49e3ceeabbfb9d66c3aef5af3a60cc43b85c33df25ce03d0031a608b0a8b2e3f",
                "sha256:4dc8f9fb58f7364b63fd9f85013b780ef83c11857ae79f2feda41e270468dd9b",
                "sha256:4efca8f86c54b22348a5467704e3fec767b2db12fc39c6d963168ab1d3fc9135",
                "sha256:53edb4da6925ad13c07b6d26c2a852bd81e364f95301c66e930ab2aef5b5ddd8",
                "sha256:5855f8438a7d1d458206a2466bf82b0f104a3724bf96a1c781ab731e4201731a",
                "sha256:594c67807fb16238b30c44bdf74f36c02cdf22d1c8cda91ef8a0ed8dabf5620a",
                "sha256:5b6d930f030f8ed98e3e6c98ffa0652bdb82601e7a016ec2ab5d7ff23baa78d1",
                "sha256:5bb28c636d87e840583ee3adeb78172efc47c8b26127267f54a9c0ec251d41a9",
                "sha256:60bf42e36abfaf9aff1f50f52644b336d4f0a3fd6d8a60ca0d054ac9f713a864",
                "sha256:611d1ad9a4288cf3e3c16014564df047fe08410e628f89805e475368bd304914",
                "sha256:6300b8454aa6930a24b9618fbb54b5a68135092bc666f7b06901f897fa5c2fee",
                "sha256:63f3268ba69ace99cab4e3e3b5840b03340efed0948ab8f78d2fd87ee5442a4f",
                "sha256:6557b31b5e2c9ddf0de32a691f2312a32f77cd7681d8af66c2692efdbef84c18",
                "sha256:693ce3f9e70a6cf7d2fb9e6c9d8b204b6b39897a2c4a1aa65728d5ac97dcc1d8",
                "sha256:6a7fae0dd14cf60ad5ff42baa2e95727c3d81ded453457771d02b7d2b3f9c0c2",
                "sha256:6c4ca60fa24e85fe25b912b01e62cb969d69a23a5d5867682dd3e80b5b02581d",
                "sha256:6fcf051089389abe060c9cd7caa212c707e58153afa2c649f00346ce6d260f1b",
                "sha256:7d91275b0245b1da4d4cfa07e0faedd5b0812efc15b702576d103293e252af1b",
                "sha256:89c687013cb1cd489a0f0ac24febe8c7a666e6e221b783e53ac50ebf68e45d86",
                "sha256:8d206346619592c6200148b01a2142798c989edcb9c896f9ac9722a99d4e77e6",
                "sha256:905fec760bd2fa1388bb5b489ee8ee5f7291d692638ea5f67982d968366bef9f",
                "sha256:97383d78eb34da7e1fa37dd273c20ad4320929af65d156e35a5e2d89566d9dfb",
                "sha256:984d76483eb32f1bcb536dc27e4ad56bba4baa70be32fa87152832cdd9db0833",
                "sha256:99df47edb6bda1249d3e80fdabb1dab8c08ef3975f69aed437cb69d0a5de1e28",
                "sha256:9f02365d4e99430a12647f09b6cc8bab61a6564363f313126f775eb4f6ef798e",
                "sha256:a30e67a65b53ea0a5e62fe23682cfe22712e01f453b95233b25502f7c61cb415",
                "sha256:ab3ef638ace319fa26553db0624c4699e31a28bb2a835c5faca8f8acf6a5a902",
                "sha256:aca6377c0cb8a8253e493c6b451565ac77e98c2951c45f913e0b52facdcff83f",
                "sha256:add36cb2dbb8b736611303cd3bfcee00afd96471b09cda130da3581cbdc56a6d",
                "sha256:b2f4bf27480f5e5e8ce285a8c8fd176c0b03e93dcc6646477d4630e83440c6a9",
                "sha256:b7f2d075102dc8c794cbde1947378051c4e5180d52d276987b8d28a3bd58c17d",
                "sha256:baa1a4e8f868845af802979fcdbf0bb11f94f1cb7ced4c4b8a351bb60d108145",
                "sha256:be98f628055368795d818ebf93da628541e10b75b41c559fdf36d104c5787066",
                "sha256:bf5d821ffabf0ef3533c39c518f3357b171a1651c1ff6827325e4489b0e46c3c",
                "sha256:c47adbc92fc1bb2b3274c4b3a43ae0e4573d9fbff4f54cd484555edbf030baf1",
                "sha256:cdfba22ea2f0029c9261a4bd07e830a8da012291fbe44dc794e488b6c9bb353a",
                "sha256:d6c7ebd4e944c85e2c3421e612a7057a2f48d478d79e61800d81468a8d842207",
                "sha256:d7f9850398e85aba693bb640262d3611788b1f29a79f0c93c565694658f4071f",
                "sha256:d8446c54dc28c01e5a2dbac5a25f071f6653e6e40f3a8818e8b45d790fe6ef53",
                "sha256:deb993cacb280823246a026e3b2d81c493c53de6acfd5e6bfe31ab3402bb37dd",
                "sha256:e0f138900af21926a02425cf736db95be9f4af72ba1bb21453432a07f6082134",
                "sha256:e9936f0b261d4df76ad22f8fee3ae83b60d7c3e871292cd42f40b81b70afae85",
                "sha256:f0567c4dc99f264f49fe27da5f735f414c4e7e7dd850cfd8e69f0862d7c74ea9",
                "sha256:f5653a225f31e113b152e56f154ccbe59eeb1c7487b39b9d9f9cdb58e6c79dc5",
                "sha256:f826e31d18b516f653fe296d967d700fddad5901ae07c622bb3705955e1faa94",
                "sha256:f8ba0e8349a38d3001fae7eadded3f6606f0da5d748ee53cc1dab1d6527b9509",
                "sha256:f9081981fe268bd86831e5c75f7de206ef275defcb82bc70740ae6dc507aee51",
                "sha256:fa130dd50c57d53368c9d59395cb5526eda596d3ffe36666cd81a44d56e48872"
            ],
            "markers": "python_version >= '3.6'",
            "version": "==2.0.1"
        },
        "packaging": {
            "hashes": [
                "sha256:096d689d78ca690e4cd8a89568ba06d07ca097e3306a4381635073ca91479966",
                "sha256:14317396d1e8cdb122989b916fa2c7e9ca8e2be9e8060a6eff75b6b7b4d8a7e0"
            ],
            "markers": "python_version >= '3.6'",
            "version": "==21.2"
        },
        "pluggy": {
            "hashes": [
                "sha256:4224373bacce55f955a878bf9cfa763c1e360858e330072059e10bad68531159",
                "sha256:74134bbf457f031a36d68416e1509f34bd5ccc019f0bcc952c7b909d06b37bd3"
            ],
            "markers": "python_version >= '3.6'",
            "version": "==1.0.0"
        },
        "psycopg2-binary": {
            "hashes": [
                "sha256:0b7dae87f0b729922e06f85f667de7bf16455d411971b2043bbd9577af9d1975",
                "sha256:0f2e04bd2a2ab54fa44ee67fe2d002bb90cee1c0f1cc0ebc3148af7b02034cbd",
                "sha256:123c3fb684e9abfc47218d3784c7b4c47c8587951ea4dd5bc38b6636ac57f616",
                "sha256:1473c0215b0613dd938db54a653f68251a45a78b05f6fc21af4326f40e8360a2",
                "sha256:14db1752acdd2187d99cb2ca0a1a6dfe57fc65c3281e0f20e597aac8d2a5bd90",
                "sha256:1e3a362790edc0a365385b1ac4cc0acc429a0c0d662d829a50b6ce743ae61b5a",
                "sha256:1e85b74cbbb3056e3656f1cc4781294df03383127a8114cbc6531e8b8367bf1e",
                "sha256:20f1ab44d8c352074e2d7ca67dc00843067788791be373e67a0911998787ce7d",
                "sha256:24b0b6688b9f31a911f2361fe818492650795c9e5d3a1bc647acbd7440142a4f",
                "sha256:2f62c207d1740b0bde5c4e949f857b044818f734a3d57f1d0d0edc65050532ed",
                "sha256:3242b9619de955ab44581a03a64bdd7d5e470cc4183e8fcadd85ab9d3756ce7a",
                "sha256:35c4310f8febe41f442d3c65066ca93cccefd75013df3d8c736c5b93ec288140",
                "sha256:4235f9d5ddcab0b8dbd723dca56ea2922b485ea00e1dafacf33b0c7e840b3d32",
                "sha256:542875f62bc56e91c6eac05a0deadeae20e1730be4c6334d8f04c944fcd99759",
                "sha256:5ced67f1e34e1a450cdb48eb53ca73b60aa0af21c46b9b35ac3e581cf9f00e31",
                "sha256:661509f51531ec125e52357a489ea3806640d0ca37d9dada461ffc69ee1e7b6e",
                "sha256:7360647ea04db2e7dff1648d1da825c8cf68dc5fbd80b8fb5b3ee9f068dcd21a",
                "sha256:736b8797b58febabb85494142c627bd182b50d2a7ec65322983e71065ad3034c",
                "sha256:8c13d72ed6af7fd2c8acbd95661cf9477f94e381fce0792c04981a8283b52917",
                "sha256:988b47ac70d204aed01589ed342303da7c4d84b56c2f4c4b8b00deda123372bf",
                "sha256:995fc41ebda5a7a663a254a1dcac52638c3e847f48307b5416ee373da15075d7",
                "sha256:a36c7eb6152ba5467fb264d73844877be8b0847874d4822b7cf2d3c0cb8cdcb0",
                "sha256:aed4a9a7e3221b3e252c39d0bf794c438dc5453bc2963e8befe9d4cd324dff72",
                "sha256:aef9aee84ec78af51107181d02fe8773b100b01c5dfde351184ad9223eab3698",
                "sha256:b0221ca5a9837e040ebf61f48899926b5783668b7807419e4adae8175a31f773",
                "sha256:b4d7679a08fea64573c969f6994a2631908bb2c0e69a7235648642f3d2e39a68",
                "sha256:c250a7ec489b652c892e4f0a5d122cc14c3780f9f643e1a326754aedf82d9a76",
                "sha256:ca86db5b561b894f9e5f115d6a159fff2a2570a652e07889d8a383b5fae66eb4",
                "sha256:cfc523edecddaef56f6740d7de1ce24a2fdf94fd5e704091856a201872e37f9f",
                "sha256:d92272c7c16e105788efe2cfa5d680f07e34e0c29b03c1908f8636f55d5f915a",
                "sha256:da113b70f6ec40e7d81b43d1b139b9db6a05727ab8be1ee559f3a69854a69d34",
                "sha256:f6fac64a38f6768e7bc7b035b9e10d8a538a9fadce06b983fb3e6fa55ac5f5ce",
                "sha256:f8559617b1fcf59a9aedba2c9838b5b6aa211ffedecabca412b92a1ff75aac1a",
                "sha256:fbb42a541b1093385a2d8c7eec94d26d30437d0e77c1d25dae1dcc46741a385e"
            ],
            "index": "pypi",
            "version": "==2.9.1"
        },
        "py": {
            "hashes": [
                "sha256:21b81bda15b66ef5e1a777a21c4dcd9c20ad3efd0b3f817e7a809035269e1bd3",
                "sha256:3b80836aa6d1feeaa108e046da6423ab8f6ceda6468545ae8d02d9d58d18818a"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3'",
            "version": "==1.10.0"
        },
        "pycparser": {
            "hashes": [
                "sha256:2d475327684562c3a96cc71adf7dc8c4f0565175cf86b6d7a404ff4c771f15f0",
                "sha256:7582ad22678f0fcd81102833f60ef8d0e57288b6b5fb00323d101be910e35705"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3'",
            "version": "==2.20"
        },
        "pyjwt": {
            "extras": [
                "crypto"
            ],
            "hashes": [
                "sha256:b888b4d56f06f6dcd777210c334e69c737be74755d3e5e9ee3fe67dc18a0ee41",
                "sha256:e0c4bb8d9f0af0c7f5b1ec4c5036309617d03d56932877f2f7a0beeb5318322f"
            ],
            "index": "pypi",
            "version": "==2.3.0"
        },
        "pyparsing": {
            "hashes": [
                "sha256:c203ec8783bf771a155b207279b9bccb8dea02d8f0c9e5f8ead507bc3246ecc1",
                "sha256:ef9d7589ef3c200abe66653d3f1ab1033c3c419ae9b9bdb1240a85b024efc88b"
            ],
            "markers": "python_version >= '2.6' and python_version not in '3.0, 3.1, 3.2, 3.3'",
            "version": "==2.4.7"
        },
        "pytest": {
            "hashes": [
                "sha256:131b36680866a76e6781d13f101efb86cf674ebb9762eb70d3082b6f29889e89",
                "sha256:7310f8d27bc79ced999e760ca304d69f6ba6c6649c0b60fb0e04a4a77cacc134"
            ],
            "index": "pypi",
            "version": "==6.2.5"
        },
        "pytest-forked": {
            "hashes": [
                "sha256:6aa9ac7e00ad1a539c41bec6d21011332de671e938c7637378ec9710204e37ca",
                "sha256:dc4147784048e70ef5d437951728825a131b81714b398d5d52f17c7c144d8815"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4'",
            "version": "==1.3.0"
        },
        "pytest-xdist": {
            "hashes": [
                "sha256:7b61ebb46997a0820a263553179d6d1e25a8c50d8a8620cd1aa1e20e3be99168",
                "sha256:89b330316f7fc475f999c81b577c2b926c9569f3d397ae432c0c2e2496d61ff9"
            ],
            "index": "pypi",
            "version": "==2.4.0"
        },
        "requests": {
            "hashes": [
                "sha256:6c1246513ecd5ecd4528a0906f910e8f0f9c6b8ec72030dc9fd154dc1a6efd24",
                "sha256:b8aa58f8cf793ffd8782d3d8cb19e66ef36f7aba4353eec859e74678b01b07a7"
            ],
            "index": "pypi",
            "version": "==2.26.0"
        },
        "toml": {
            "hashes": [
                "sha256:806143ae5bfb6a3c6e736a764057db0e6a0e05e338b5630894a5f779cabb4f9b",
                "sha256:b3bda1d108d5dd99f4a20d24d9c348e91c4db7ab1b749200bded2f839ccbe68f"
            ],
            "markers": "python_version >= '2.6' and python_version not in '3.0, 3.1, 3.2, 3.3'",
            "version": "==0.10.2"
        },
        "typing-extensions": {
            "hashes": [
                "sha256:49f75d16ff11f1cd258e1b988ccff82a3ca5570217d7ad8c5f48205dd99a677e",
                "sha256:d8226d10bc02a29bcc81df19a26e56a9647f8b0a6d4a83924139f4a8b01f17b7",
                "sha256:f1d25edafde516b146ecd0613dabcc61409817af4766fbbcfb8d1ad4ec441a34"
            ],
            "index": "pypi",
            "version": "==3.10.0.2"
        },
        "urllib3": {
            "hashes": [
                "sha256:4987c65554f7a2dbf30c18fd48778ef124af6fab771a377103da0585e2336ece",
                "sha256:c4fdf4019605b6e5423637e01bc9fe4daef873709a7973e195ceba0a62bbc844"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' and python_version < '4'",
            "version": "==1.26.7"
        },
        "zipp": {
            "hashes": [
                "sha256:71c644c5369f4a6e07636f0aa966270449561fcea2e3d6747b8d23efaa9d7832",
                "sha256:9fe5ea21568a0a70e50f273397638d39b03353731e6cbbb3fd8502a33fec40bc"
            ],
            "markers": "python_version >= '3.6'",
            "version": "==3.6.0"
        }
    },
    "develop": {
        "backports.entry-points-selectable": {
            "hashes": [
                "sha256:988468260ec1c196dab6ae1149260e2f5472c9110334e5d51adcb77867361f6a",
                "sha256:a6d9a871cde5e15b4c4a53e3d43ba890cc6861ec1332c9c2428c92f977192acc"
            ],
            "markers": "python_version >= '2.7'",
            "version": "==1.1.0"
        },
        "certifi": {
            "hashes": [
                "sha256:78884e7c1d4b00ce3cea67b44566851c4343c120abd683433ce934a68ea58872",
                "sha256:d62a0163eb4c2344ac042ab2bdf75399a71a2d8c7d47eac2e2ee91b9d6339569"
            ],
            "version": "==2021.10.8"
        },
        "distlib": {
            "hashes": [
                "sha256:c8b54e8454e5bf6237cc84c20e8264c3e991e824ef27e8f1e81049867d861e31",
                "sha256:d982d0751ff6eaaab5e2ec8e691d949ee80eddf01a62eaa96ddb11531fe16b05"
            ],
            "version": "==0.3.3"
        },
        "filelock": {
            "hashes": [
                "sha256:7afc856f74fa7006a289fd10fa840e1eebd8bbff6bffb69c26c54a0512ea8cf8",
                "sha256:bb2a1c717df74c48a2d00ed625e5a66f8572a3a30baacb7657add1d7bac4097b"
            ],
            "markers": "python_version >= '3.6'",
            "version": "==3.3.2"
        },
        "flake8": {
            "hashes": [
                "sha256:479b1304f72536a55948cb40a32dce8bb0ffe3501e26eaf292c7e60eb5e0428d",
                "sha256:806e034dda44114815e23c16ef92f95c91e4c71100ff52813adf7132a6ad870d"
            ],
            "index": "pypi",
            "version": "==4.0.1"
        },
        "importlib-metadata": {
            "hashes": [
                "sha256:b618b6d2d5ffa2f16add5697cf57a46c76a56229b0ed1c438322e4e95645bd15",
                "sha256:f284b3e11256ad1e5d03ab86bb2ccd6f5339688ff17a4d797a0fe7df326f23b1"
            ],
            "markers": "python_version < '3.8'",
            "version": "==4.8.1"
        },
        "mccabe": {
            "hashes": [
                "sha256:ab8a6258860da4b6677da4bd2fe5dc2c659cff31b3ee4f7f5d64e79735b80d42",
                "sha256:dd8d182285a0fe56bace7f45b5e7d1a6ebcbf524e8f3bd87eb0f125271b8831f"
            ],
            "version": "==0.6.1"
        },
        "mypy": {
            "hashes": [
                "sha256:088cd9c7904b4ad80bec811053272986611b84221835e079be5bcad029e79dd9",
                "sha256:0aadfb2d3935988ec3815952e44058a3100499f5be5b28c34ac9d79f002a4a9a",
                "sha256:119bed3832d961f3a880787bf621634ba042cb8dc850a7429f643508eeac97b9",
                "sha256:1a85e280d4d217150ce8cb1a6dddffd14e753a4e0c3cf90baabb32cefa41b59e",
                "sha256:3c4b8ca36877fc75339253721f69603a9c7fdb5d4d5a95a1a1b899d8b86a4de2",
                "sha256:3e382b29f8e0ccf19a2df2b29a167591245df90c0b5a2542249873b5c1d78212",
                "sha256:42c266ced41b65ed40a282c575705325fa7991af370036d3f134518336636f5b",
                "sha256:53fd2eb27a8ee2892614370896956af2ff61254c275aaee4c230ae771cadd885",
                "sha256:704098302473cb31a218f1775a873b376b30b4c18229421e9e9dc8916fd16150",
                "sha256:7df1ead20c81371ccd6091fa3e2878559b5c4d4caadaf1a484cf88d93ca06703",
                "sha256:866c41f28cee548475f146aa4d39a51cf3b6a84246969f3759cb3e9c742fc072",
                "sha256:a155d80ea6cee511a3694b108c4494a39f42de11ee4e61e72bc424c490e46457",
                "sha256:adaeee09bfde366d2c13fe6093a7df5df83c9a2ba98638c7d76b010694db760e",
                "sha256:b6fb13123aeef4a3abbcfd7e71773ff3ff1526a7d3dc538f3929a49b42be03f0",
                "sha256:b94e4b785e304a04ea0828759172a15add27088520dc7e49ceade7834275bedb",
                "sha256:c0df2d30ed496a08de5daed2a9ea807d07c21ae0ab23acf541ab88c24b26ab97",
                "sha256:c6c2602dffb74867498f86e6129fd52a2770c48b7cd3ece77ada4fa38f94eba8",
                "sha256:ceb6e0a6e27fb364fb3853389607cf7eb3a126ad335790fa1e14ed02fba50811",
                "sha256:d9dd839eb0dc1bbe866a288ba3c1afc33a202015d2ad83b31e875b5905a079b6",
                "sha256:e4dab234478e3bd3ce83bac4193b2ecd9cf94e720ddd95ce69840273bf44f6de",
                "sha256:ec4e0cd079db280b6bdabdc807047ff3e199f334050db5cbb91ba3e959a67504",
                "sha256:ecd2c3fe726758037234c93df7e98deb257fd15c24c9180dacf1ef829da5f921",
                "sha256:ef565033fa5a958e62796867b1df10c40263ea9ded87164d67572834e57a174d"
            ],
            "index": "pypi",
            "version": "==0.910"
        },
        "mypy-extensions": {
            "hashes": [
                "sha256:090fedd75945a69ae91ce1303b5824f428daf5a028d2f6ab8a299250a846f15d",
                "sha256:2d82818f5bb3e369420cb3c4060a7970edba416647068eb4c5343488a6c604a8"
            ],
            "version": "==0.4.3"
        },
        "pipenv": {
            "hashes": [
                "sha256:05958fadcd70b2de6a27542fcd2bd72dd5c59c6d35307fdac3e06361fb06e30e",
                "sha256:d180f5be4775c552fd5e69ae18a9d6099d9dafb462efe54f11c72cb5f4d5e977"
            ],
            "index": "pypi",
            "version": "==2021.5.29"
        },
        "platformdirs": {
            "hashes": [
                "sha256:367a5e80b3d04d2428ffa76d33f124cf11e8fff2acdaa9b43d545f5c7d661ef2",
                "sha256:8868bbe3c3c80d42f20156f22e7131d2fb321f5bc86a2a345375c6481a67021d"
            ],
            "markers": "python_version >= '3.6'",
            "version": "==2.4.0"
        },
        "pycodestyle": {
            "hashes": [
                "sha256:720f8b39dde8b293825e7ff02c475f3077124006db4f440dcbc9a20b76548a20",
                "sha256:eddd5847ef438ea1c7870ca7eb78a9d47ce0cdb4851a5523949f2601d0cbbe7f"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4'",
            "version": "==2.8.0"
        },
        "pyflakes": {
            "hashes": [
                "sha256:05a85c2872edf37a4ed30b0cce2f6093e1d0581f8c19d7393122da7e25b2b24c",
                "sha256:3bb3a3f256f4b7968c9c788781e4ff07dce46bdf12339dcda61053375426ee2e"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3'",
            "version": "==2.4.0"
        },
        "six": {
            "hashes": [
                "sha256:1e61c37477a1626458e36f7b1d82aa5c9b094fa4802892072e49de9c60c4c926",
                "sha256:8abb2f1d86890a2dfb989f9a77cfcfd3e47c2a354b01111771326f8aa26e0254"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3'",
            "version": "==1.16.0"
        },
        "toml": {
            "hashes": [
                "sha256:806143ae5bfb6a3c6e736a764057db0e6a0e05e338b5630894a5f779cabb4f9b",
                "sha256:b3bda1d108d5dd99f4a20d24d9c348e91c4db7ab1b749200bded2f839ccbe68f"
            ],
            "markers": "python_version >= '2.6' and python_version not in '3.0, 3.1, 3.2, 3.3'",
            "version": "==0.10.2"
        },
        "typed-ast": {
            "hashes": [
                "sha256:01ae5f73431d21eead5015997ab41afa53aa1fbe252f9da060be5dad2c730ace",
                "sha256:067a74454df670dcaa4e59349a2e5c81e567d8d65458d480a5b3dfecec08c5ff",
                "sha256:0fb71b8c643187d7492c1f8352f2c15b4c4af3f6338f21681d3681b3dc31a266",
                "sha256:1b3ead4a96c9101bef08f9f7d1217c096f31667617b58de957f690c92378b528",
                "sha256:2068531575a125b87a41802130fa7e29f26c09a2833fea68d9a40cf33902eba6",
                "sha256:209596a4ec71d990d71d5e0d312ac935d86930e6eecff6ccc7007fe54d703808",
                "sha256:2c726c276d09fc5c414693a2de063f521052d9ea7c240ce553316f70656c84d4",
                "sha256:398e44cd480f4d2b7ee8d98385ca104e35c81525dd98c519acff1b79bdaac363",
                "sha256:52b1eb8c83f178ab787f3a4283f68258525f8d70f778a2f6dd54d3b5e5fb4341",
                "sha256:5feca99c17af94057417d744607b82dd0a664fd5e4ca98061480fd8b14b18d04",
                "sha256:7538e495704e2ccda9b234b82423a4038f324f3a10c43bc088a1636180f11a41",
                "sha256:760ad187b1041a154f0e4d0f6aae3e40fdb51d6de16e5c99aedadd9246450e9e",
                "sha256:777a26c84bea6cd934422ac2e3b78863a37017618b6e5c08f92ef69853e765d3",
                "sha256:95431a26309a21874005845c21118c83991c63ea800dd44843e42a916aec5899",
                "sha256:9ad2c92ec681e02baf81fdfa056fe0d818645efa9af1f1cd5fd6f1bd2bdfd805",
                "sha256:9c6d1a54552b5330bc657b7ef0eae25d00ba7ffe85d9ea8ae6540d2197a3788c",
                "sha256:aee0c1256be6c07bd3e1263ff920c325b59849dc95392a05f258bb9b259cf39c",
                "sha256:af3d4a73793725138d6b334d9d247ce7e5f084d96284ed23f22ee626a7b88e39",
                "sha256:b36b4f3920103a25e1d5d024d155c504080959582b928e91cb608a65c3a49e1a",
                "sha256:b9574c6f03f685070d859e75c7f9eeca02d6933273b5e69572e5ff9d5e3931c3",
                "sha256:bff6ad71c81b3bba8fa35f0f1921fb24ff4476235a6e94a26ada2e54370e6da7",
                "sha256:c190f0899e9f9f8b6b7863debfb739abcb21a5c054f911ca3596d12b8a4c4c7f",
                "sha256:c907f561b1e83e93fad565bac5ba9c22d96a54e7ea0267c708bffe863cbe4075",
                "sha256:cae53c389825d3b46fb37538441f75d6aecc4174f615d048321b716df2757fb0",
                "sha256:dd4a21253f42b8d2b48410cb31fe501d32f8b9fbeb1f55063ad102fe9c425e40",
                "sha256:dde816ca9dac1d9c01dd504ea5967821606f02e510438120091b84e852367428",
                "sha256:f2362f3cb0f3172c42938946dbc5b7843c2a28aec307c49100c8b38764eb6927",
                "sha256:f328adcfebed9f11301eaedfa48e15bdece9b519fb27e6a8c01aa52a17ec31b3",
                "sha256:f8afcf15cc511ada719a88e013cec87c11aff7b91f019295eb4530f96fe5ef2f",
                "sha256:fb1bbeac803adea29cedd70781399c99138358c26d05fcbd23c13016b7f5ec65"
            ],
            "markers": "python_version < '3.8'",
            "version": "==1.4.3"
        },
        "types-psycopg2": {
            "hashes": [
                "sha256:77ed80f2668582654623e04fb3d741ecce93effcc39c929d7e02f4a917a538ce",
                "sha256:98a6e0e9580cd7eb4bd4d20f7c7063d154b2589a2b90c0ce4e3ca6085cde77c6"
            ],
            "index": "pypi",
            "version": "==2.9.1"
        },
        "types-requests": {
            "hashes": [
                "sha256:b279284e51f668e38ee12d9665e4d789089f532dc2a0be4a1508ca0efd98ba9e",
                "sha256:ba1d108d512e294b6080c37f6ae7cb2a2abf527560e2b671d1786c1fc46b541a"
            ],
            "index": "pypi",
            "version": "==2.25.11"
        },
        "typing-extensions": {
            "hashes": [
                "sha256:49f75d16ff11f1cd258e1b988ccff82a3ca5570217d7ad8c5f48205dd99a677e",
                "sha256:d8226d10bc02a29bcc81df19a26e56a9647f8b0a6d4a83924139f4a8b01f17b7",
                "sha256:f1d25edafde516b146ecd0613dabcc61409817af4766fbbcfb8d1ad4ec441a34"
            ],
            "index": "pypi",
            "version": "==3.10.0.2"
        },
        "virtualenv": {
            "hashes": [
                "sha256:4b02e52a624336eece99c96e3ab7111f469c24ba226a53ec474e8e787b365814",
                "sha256:576d05b46eace16a9c348085f7d0dc8ef28713a2cabaa1cf0aea41e8f12c9218"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4'",
            "version": "==20.10.0"
        },
        "virtualenv-clone": {
            "hashes": [
                "sha256:418ee935c36152f8f153c79824bb93eaf6f0f7984bae31d3f48f350b9183501a",
                "sha256:44d5263bceed0bac3e1424d64f798095233b64def1c5689afa43dc3223caf5b0"
            ],
            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3'",
            "version": "==0.5.7"
        },
        "yapf": {
            "hashes": [
                "sha256:408fb9a2b254c302f49db83c59f9aa0b4b0fd0ec25be3a5c51181327922ff63d",
                "sha256:e3a234ba8455fe201eaa649cdac872d590089a18b661e39bbac7020978dd9c2e"
            ],
            "index": "pypi",
            "version": "==0.31.0"
        },
        "zipp": {
            "hashes": [
                "sha256:71c644c5369f4a6e07636f0aa966270449561fcea2e3d6747b8d23efaa9d7832",
                "sha256:9fe5ea21568a0a70e50f273397638d39b03353731e6cbbb3fd8502a33fec40bc"
            ],
            "markers": "python_version >= '3.6'",
            "version": "==3.6.0"
        }
    }
 }
@@ -28,12 +28,12 @@ apt install build-essential libtool libreadline-dev zlib1g-dev flex bison libsec
 libssl-dev clang pkg-config libpq-dev
 ```
-[Rust] 1.55 or later is also required.
+[Rust] 1.56.1 or later is also required.
 To run the `psql` client, install the `postgresql-client` package or modify `PATH` and `LD_LIBRARY_PATH` to include `tmp_install/bin` and `tmp_install/lib`, respectively.
 To run the integration tests or Python scripts (not required to use the code), install
-Python (3.7 or higher), and install python3 packages using `pipenv install` in the project directory.
+Python (3.7 or higher), and install python3 packages using `./scripts/pysync` (requires poetry) in the project directory.
 2. Build zenith and patched postgres
 ```sh
@@ -128,8 +128,7 @@ INSERT 0 1
 ```sh
 git clone --recursive https://github.com/zenithdb/zenith.git
 make # builds also postgres and installs it to ./tmp_install
-cd test_runner
+./scripts/pytest
 pipenv run pytest
 ```
 ## Documentation
@@ -2,13 +2,10 @@
 name = "compute_tools"
 version = "0.1.0"
 authors = ["Alexey Kondratov <kondratov.aleksey@gmail.com>"]
-edition = "2018"
+edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 [workspace]
 # TODO: make it a part of global zenith worksapce
 [dependencies]
 libc = "0.2"
 anyhow = "1.0"
@@ -17,12 +14,9 @@ clap = "2.33"
 env_logger = "0.8"
 hyper = { version = "0.14", features = ["full"] }
 log = { version = "0.4", features = ["std", "serde"] }
-postgres = "0.19"
+postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
 regex = "1"
 serde = { version = "1.0", features = ["derive"] }
-serde_json = "1.0"
+serde_json = "1"
 tar = "0.4"
-tokio = { version = "1", features = ["full"] }
+tokio = { version = "1", features = ["macros", "rt", "rt-multi-thread"] }
 [profile.release]
 debug = true
@@ -1,14 +0,0 @@
 # First transient image to build compute_tools binaries
 FROM rust:slim-buster AS rust-build
 RUN mkdir /compute_tools
 WORKDIR /compute_tools
 COPY . /compute_tools/
 RUN cargo build --release
 # Final image that only has one binary
 FROM debian:buster-slim
 COPY --from=rust-build /compute_tools/target/release/zenith_ctl /usr/local/bin/zenith_ctl
@@ -27,14 +27,13 @@
 //! ```
 //!
 use std::fs::File;
 use std::panic;
 use std::path::Path;
 use std::process::{exit, Command, ExitStatus};
 use std::sync::{Arc, RwLock};
 use std::{env, panic};
-use anyhow::Result;
+use anyhow::{Context, Result};
 use chrono::Utc;
 use libc::{prctl, PR_SET_PDEATHSIG, SIGINT};
 use log::info;
 use postgres::{Client, NoTls};
@@ -70,7 +69,7 @@ fn prepare_pgdata(state: &Arc<RwLock<ComputeState>>) -> Result<()> {
        .expect("tenant id should be provided");
    info!(
-        "applying spec for cluster #{}, operation #{}",
+        "starting cluster #{}, operation #{}",
        spec.cluster.cluster_id,
        spec.operation_uuid.as_ref().unwrap()
    );
@@ -80,10 +79,23 @@ fn prepare_pgdata(state: &Arc<RwLock<ComputeState>>) -> Result<()> {
    config::write_postgres_conf(&pgdata_path.join("postgresql.conf"), spec)?;
    info!("starting safekeepers syncing");
-    let lsn = sync_safekeepers(&state.pgdata, &state.pgbin)?;
+    let lsn = sync_safekeepers(&state.pgdata, &state.pgbin)
        .with_context(|| "failed to sync safekeepers")?;
    info!("safekeepers synced at LSN {}", lsn);
-    get_basebackup(&state.pgdata, &pageserver_connstr, &tenant, &timeline, &lsn)?;
+    info!(
        "getting basebackup@{} from pageserver {}",
        lsn, pageserver_connstr
    );
    get_basebackup(&state.pgdata, &pageserver_connstr, &tenant, &timeline, &lsn).with_context(
        || {
            format!(
                "failed to get basebackup@{} from pageserver {}",
                lsn, pageserver_connstr
            )
        },
    )?;
    // Update pg_hba.conf received with basebackup.
    update_pg_hba(pgdata_path)?;
@@ -142,22 +154,13 @@ fn run_compute(state: &Arc<RwLock<ComputeState>>) -> Result<ExitStatus> {
 }
 fn main() -> Result<()> {
    // During configuration we are starting Postgres as a child process. If we
    // fail we do not want to leave it running. PR_SET_PDEATHSIG sets the signal
    // that will be sent to the child process when the parent dies. NB: this is
    // cleared for the child of a fork(). SIGINT means fast shutdown for Postgres.
    // This does not matter much for Docker, where `zenith_ctl` is an entrypoint,
    // so the whole container will exit if it exits. But could be useful when
    // `zenith_ctl` is used in e.g. systemd.
    unsafe {
        prctl(PR_SET_PDEATHSIG, SIGINT);
    }
    // TODO: re-use `zenith_utils::logging` later
    init_logger(DEFAULT_LOG_LEVEL)?;
    // Env variable is set by `cargo`
    let version: Option<&str> = option_env!("CARGO_PKG_VERSION");
    let matches = clap::App::new("zenith_ctl")
-        .version("0.1.0")
+        .version(version.unwrap_or("unknown"))
        .arg(
            clap::Arg::with_name("connstr")
                .short("C")
@@ -212,13 +215,7 @@ fn main() -> Result<()> {
                let file = File::open(path)?;
                serde_json::from_reader(file)?
            } else {
-                // Finally, try to fetch it from the env
+                panic!("cluster spec should be provided via --spec or --spec-path argument");
                // XXX: not tested well and kept as a backup option for k8s, Docker, etc.
                // TODO: remove later
                match env::var("CLUSTER_SPEC") {
                    Ok(json) => serde_json::from_str(&json)?,
                    Err(_) => panic!("cluster spec should be provided via --spec, --spec-path or env variable CLUSTER_SPEC")
                }
            }
        }
    };
@@ -5,7 +5,7 @@ use std::process::Command;
 use std::str::FromStr;
 use std::{fs, thread, time};
-use anyhow::{anyhow, Result};
+use anyhow::{bail, Result};
 use postgres::{Client, Transaction};
 use serde::Deserialize;
@@ -226,7 +226,7 @@ pub fn wait_for_postgres(port: &str, pgdata: &Path) -> Result<()> {
        // but postgres starts listening almost immediately, even if it is not really
        // ready to accept connections).
        if slept >= POSTGRES_WAIT_TIMEOUT {
-            return Err(anyhow!("timed out while waiting for Postgres to start"));
+            bail!("timed out while waiting for Postgres to start");
        }
        if pid_path.exists() {
@@ -87,17 +87,19 @@ pub fn sync_safekeepers(pgdata: &str, pgbin: &str) -> Result<String> {
        .args(&["--sync-safekeepers"])
        .env("PGDATA", &pgdata) // we cannot use -D in this mode
        .stdout(Stdio::piped())
        .stderr(Stdio::piped())
        .spawn()
        .expect("postgres --sync-safekeepers failed to start");
    // `postgres --sync-safekeepers` will print all log output to stderr and
    // final LSN to stdout. So we pipe only stdout, while stderr will be automatically
    // redirected to the caller output.
    let sync_output = sync_handle
        .wait_with_output()
        .expect("postgres --sync-safekeepers failed");
    if !sync_output.status.success() {
        anyhow::bail!(
-            "postgres --sync-safekeepers failed: '{}'",
+            "postgres --sync-safekeepers exited with non-zero status: {}",
-            String::from_utf8_lossy(&sync_output.stderr)
+            sync_output.status,
        );
    }
@@ -2,13 +2,13 @@
 name = "control_plane"
 version = "0.1.0"
 authors = ["Stas Kelvich <stas@zenith.tech>"]
-edition = "2018"
+edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 [dependencies]
 tar = "0.4.33"
-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
 serde = { version = "1.0", features = ["derive"] }
 toml = "0.5"
 lazy_static = "1.4"
@@ -82,15 +82,11 @@ impl ComputeControlPlane {
        let mut strings = s.split('@');
        let name = strings.next().unwrap();
-        let lsn: Option<Lsn>;
+        let lsn = strings
-        if let Some(lsnstr) = strings.next() {
+            .next()
-            lsn = Some(
+            .map(Lsn::from_str)
-                Lsn::from_str(lsnstr)
+            .transpose()
-                    .with_context(|| "invalid LSN in point-in-time specification")?,
+            .context("invalid LSN in point-in-time specification")?;
            );
        } else {
            lsn = None
        }
        // Resolve the timeline ID, given the human-readable branch name
        let timeline_id = self
@@ -253,16 +249,16 @@ impl PostgresNode {
        let mut client = self
            .pageserver
            .page_server_psql_client()
-            .with_context(|| "connecting to page server failed")?;
+            .context("connecting to page server failed")?;
        let copyreader = client
            .copy_out(sql.as_str())
-            .with_context(|| "page server 'basebackup' command failed")?;
+            .context("page server 'basebackup' command failed")?;
        // Read the archive directly from the `CopyOutReader`
        tar::Archive::new(copyreader)
            .unpack(&self.pgdata())
-            .with_context(|| "extracting base backup failed")?;
+            .context("extracting base backup failed")?;
        Ok(())
    }
@@ -443,7 +439,7 @@ impl PostgresNode {
        if let Some(token) = auth_token {
            cmd.env("ZENITH_AUTH_TOKEN", token);
        }
-        let pg_ctl = cmd.status().with_context(|| "pg_ctl failed")?;
+        let pg_ctl = cmd.status().context("pg_ctl failed")?;
        if !pg_ctl.success() {
            anyhow::bail!("pg_ctl failed");
@@ -251,7 +251,7 @@ impl LocalEnv {
                .arg("2048")
                .stdout(Stdio::null())
                .output()
-                .with_context(|| "failed to generate auth private key")?;
+                .context("failed to generate auth private key")?;
            if !keygen_output.status.success() {
                bail!(
                    "openssl failed: '{}'",
@@ -270,7 +270,7 @@ impl LocalEnv {
                .args(&["-out", public_key_path.to_str().unwrap()])
                .stdout(Stdio::null())
                .output()
-                .with_context(|| "failed to generate auth private key")?;
+                .context("failed to generate auth private key")?;
            if !keygen_output.status.success() {
                bail!(
                    "openssl failed: '{}'",
@@ -4,7 +4,7 @@
 /// NOTE: This doesn't implement the full, correct postgresql.conf syntax. Just
 /// enough to extract a few settings we need in Zenith, assuming you don't do
 /// funny stuff like include-directives or funny escaping.
-use anyhow::{anyhow, bail, Context, Result};
+use anyhow::{bail, Context, Result};
 use lazy_static::lazy_static;
 use regex::Regex;
 use std::collections::HashMap;
@@ -78,7 +78,7 @@ impl PostgresConf {
        <T as FromStr>::Err: std::error::Error + Send + Sync + 'static,
    {
        self.get(field_name)
-            .ok_or_else(|| anyhow!("could not find '{}' option {}", field_name, context))?
+            .with_context(|| format!("could not find '{}' option {}", field_name, context))?
            .parse::<T>()
            .with_context(|| format!("could not parse '{}' option {}", field_name, context))
    }
@@ -2,6 +2,16 @@
 ### Authentication
 ### Backpresssure
 Backpressure is used to limit the lag between pageserver and compute node or WAL service.
 If compute node or WAL service run far ahead of Page Server,
 the time of serving page requests increases. This may lead to timeout errors.
 To tune backpressure limits use `max_replication_write_lag`, `max_replication_flush_lag` and `max_replication_apply_lag` settings.
 When lag between current LSN (pg_current_wal_flush_lsn() at compute node) and minimal write/flush/apply position of replica exceeds the limit
 backends performing writes are blocked until the replica is caught up.
 ### Base image (page image)
 ### Basebackup
@@ -76,7 +86,37 @@ The layer map tracks what layers exist for all the relishes in a timeline.
 Zenith repository implementation that keeps data in layers.
 ### LSN
 The Log Sequence Number (LSN) is a unique identifier of the WAL record[] in the WAL log.
 The insert position is a byte offset into the logs, increasing monotonically with each new record.
 Internally, an LSN is a 64-bit integer, representing a byte position in the write-ahead log stream.
 It is printed as two hexadecimal numbers of up to 8 digits each, separated by a slash.
 Check also [PostgreSQL doc about pg_lsn type](https://www.postgresql.org/docs/devel/datatype-pg-lsn.html)
 Values can be compared to calculate the volume of WAL data that separates them, so they are used to measure the progress of replication and recovery.
 In postgres and Zenith lsns are used to describe certain points in WAL handling.
 PostgreSQL LSNs and functions to monitor them:
 * `pg_current_wal_insert_lsn()` - Returns the current write-ahead log insert location.
 * `pg_current_wal_lsn()` - Returns the current write-ahead log write location.
 * `pg_current_wal_flush_lsn()` - Returns the current write-ahead log flush location.
 * `pg_last_wal_receive_lsn()` - Returns the last write-ahead log location that has been received and synced to disk by streaming replication. While streaming replication is in progress this will increase monotonically.
 * `pg_last_wal_replay_lsn ()` - Returns the last write-ahead log location that has been replayed during recovery. If recovery is still in progress this will increase monotonically. 
 [source PostgreSQL documentation](https://www.postgresql.org/docs/devel/functions-admin.html):
 Zenith safekeeper LSNs. For more check [walkeeper/README_PROTO.md](/walkeeper/README_PROTO.md)
 * `CommitLSN`: position in WAL confirmed by quorum safekeepers.
 * `RestartLSN`: position in WAL confirmed by all safekeepers.
 * `FlushLSN`: part of WAL persisted to the disk by safekeeper.
 * `VCL`: the largerst LSN for which we can guarantee availablity of all prior records.
 Zenith pageserver LSNs:
 * `last_record_lsn` - the end of last processed WAL record.
 * `disk_consistent_lsn` - data is known to be fully flushed and fsync'd to local disk on pageserver up to this LSN.
 * `remote_consistent_lsn` - The last LSN that is synced to remote storage and is guaranteed to survive pageserver crash.
 TODO: use this name consistently in remote storage code. Now `disk_consistent_lsn` is used and meaning depends on the context.
 * `ancestor_lsn` - LSN of the branch point (the LSN at which this branch was created)
 TODO: add table that describes mapping between PostgreSQL (compute), safekeeper and pageserver LSNs.
 ### Page (block)
 The basic structure used to store relation data. All pages are of the same size.
@@ -0,0 +1,22 @@
 ## Pageserver tenant migration
 ### Overview
 This feature allows to migrate a timeline from one pageserver to another by utilizing remote storage capability.
 ### Migration process
 Pageserver implements two new http handlers: timeline attach and timeline detach.
 Timeline migration is performed in a following way:
 1. Timeline attach is called on a target pageserver. This asks pageserver to download latest checkpoint uploaded to s3.
 2. For now it is necessary to manually initialize replication stream via callmemaybe call so target pageserver initializes replication from safekeeper (it is desired to avoid this and initialize replication directly in attach handler, but this requires some refactoring (probably [#997](https://github.com/zenithdb/zenith/issues/997)/[#1049](https://github.com/zenithdb/zenith/issues/1049))
 3. Replication state can be tracked via timeline detail pageserver call.
 4. Compute node should be restarted with new pageserver connection string. Issue with multiple compute nodes for one timeline is handled on the safekeeper consensus level. So this is not a problem here.Currently responsibility for rescheduling the compute with updated config lies on external coordinator (console).
 5. Timeline is detached from old pageserver. On disk data is removed.
 ### Implementation details
 Now safekeeper needs to track which pageserver it is replicating to. This introduces complications into replication code:
 * We need to distinguish different pageservers (now this is done by connection string which is imperfect and is covered here: https://github.com/zenithdb/zenith/issues/1105). Callmemaybe subscription management also needs to track that (this is already implemented).
 * We need to track which pageserver is the primary. This is needed to avoid reconnections to non primary pageservers. Because we shouldn't reconnect to them when they decide to stop their walreceiver. I e this can appear when there is a load on the compute and we are trying to detach timeline from old pageserver. In this case callmemaybe will try to reconnect to it because replication termination condition is not met (page server with active compute could never catch up to the latest lsn, so there is always some wal tail)
@@ -87,31 +87,29 @@ so manual installation of dependencies is not recommended.
 A single virtual environment with all dependencies is described in the single `Pipfile`.
 ### Prerequisites
- Install Python 3.7 (the minimal supported version)
+- Install Python 3.7 (the minimal supported version) or greater.
-    - Later version (e.g. 3.8) is ok if you don't write Python code
+    - Our setup with poetry should work with newer python versions too. So feel free to open an issue with a `c/test-runner` label if something doesnt work as expected.
-    - You can install Python 3.7 separately, e.g.:
+    - If you have some trouble with other version you can resolve it by installing Python 3.7 separately, via pyenv or via system package manager e.g.:
      ```bash
      # In Ubuntu
      sudo add-apt-repository ppa:deadsnakes/ppa
      sudo apt update
      sudo apt install python3.7
      ```
- Install `pipenv`
+- Install `poetry`
-    - Exact version of `pipenv` is not important, you can use Debian/Ubuntu package `pipenv`.
+    - Exact version of `poetry` is not important, see installation instructions available at poetry's [website](https://python-poetry.org/docs/#installation)`.
- Install dependencies via either
+- Install dependencies via `./scripts/pysync`. Note that CI uses Python 3.7 so if you have different version some linting tools can yield different result locally vs in the CI.
  * `pipenv --python 3.7 install --dev` if you will write Python code, or
  * `pipenv install` if you only want to run Python scripts and don't have Python 3.7.
-Run `pipenv shell` to activate the virtual environment.
+Run `poetry shell` to activate the virtual environment.
-Alternatively, use `pipenv run` to run a single command in the venv, e.g. `pipenv run pytest`.
+Alternatively, use `poetry run` to run a single command in the venv, e.g. `poetry run pytest`.
 ### Obligatory checks
 We force code formatting via `yapf` and type hints via `mypy`.
 Run the following commands in the repository's root (next to `setup.cfg`):
 ```bash
-pipenv run yapf -ri .  # All code is reformatted
+poetry run yapf -ri .  # All code is reformatted
-pipenv run mypy .  # Ensure there are no typing errors
+poetry run mypy .  # Ensure there are no typing errors
 ```
 **WARNING**: do not run `mypy` from a directory other than the root of the repository.
@@ -123,17 +121,6 @@ Also consider:
 * Adding more type hints to your code to avoid `Any`.
 ### Changing dependencies
-You have to update `Pipfile.lock` if you have changed `Pipfile`:
+To add new package or change an existing one you can use `poetry add` or `poetry update` or edit `pyproject.toml` manually. Do not forget to run `poetry lock` in the latter case.
-```bash
+More details are available in poetry's [documentation](https://python-poetry.org/docs/).
 pipenv --python 3.7 install --dev  # Re-create venv for Python 3.7 and install recent pipenv inside
 pipenv run pipenv --version  # Should be at least 2021.5.29
 pipenv run pipenv lock  # Regenerate Pipfile.lock
 ```
 As the minimal supported version is Python 3.7 and we use it in CI,
 you have to use a Python 3.7 environment when updating `Pipfile.lock`.
 Otherwise some back-compatibility packages will be missing.
 It is also important to run recent `pipenv`.
 Older versions remove markers from `Pipfile.lock`.
@@ -2,7 +2,7 @@
 name = "pageserver"
 version = "0.1.0"
 authors = ["Stas Kelvich <stas@zenith.tech>"]
-edition = "2018"
+edition = "2021"
 [dependencies]
 bookfile = { git = "https://github.com/zenithdb/bookfile.git", branch="generic-readext" }
@@ -18,13 +18,13 @@ log = "0.4.14"
 clap = "2.33.0"
 daemonize = "0.4.1"
 tokio = { version = "1.11", features = ["process", "sync", "macros", "fs", "rt", "io-util", "time"] }
-postgres-types = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+postgres-types = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
-postgres-protocol = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+postgres-protocol = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
-tokio-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+tokio-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
 tokio-stream = "0.1.8"
 routerify = "2"
-anyhow = "1.0"
+anyhow = { version = "1.0", features = ["backtrace"] }
 crc32c = "0.6.0"
 thiserror = "1.0"
 hex = { version = "0.4.3", features = ["serde"] }
@@ -1,6 +1,6 @@
 //! Main entry point for the Page Server executable.
-use std::{env, path::Path, str::FromStr, thread};
+use std::{env, path::Path, str::FromStr};
 use tracing::*;
 use zenith_utils::{auth::JwtAuth, logging, postgres_backend::AuthType, tcp_listener, GIT_VERSION};
@@ -12,7 +12,9 @@ use daemonize::Daemonize;
 use pageserver::{
    branches,
    config::{defaults::*, PageServerConf},
-    http, page_cache, page_service, remote_storage, tenant_mgr, virtual_file, LOG_FILE_NAME,
+    http, page_cache, page_service, remote_storage, tenant_mgr, thread_mgr,
    thread_mgr::ThreadKind,
    virtual_file, LOG_FILE_NAME,
 };
 use zenith_utils::http::endpoint;
 use zenith_utils::postgres_backend;
@@ -169,7 +171,7 @@ fn start_pageserver(conf: &'static PageServerConf, daemonize: bool) -> Result<()
    );
    let pageserver_listener = tcp_listener::bind(conf.listen_pg_addr.clone())?;
-    // XXX: Don't spawn any threads before daemonizing!
+    // NB: Don't spawn any threads before daemonizing!
    if daemonize {
        info!("daemonizing...");
@@ -195,16 +197,9 @@ fn start_pageserver(conf: &'static PageServerConf, daemonize: bool) -> Result<()
    }
    let signals = signals::install_shutdown_handlers()?;
-    let (async_shutdown_tx, async_shutdown_rx) = tokio::sync::watch::channel(());
+    let sync_startup = remote_storage::start_local_timeline_sync(conf)
    let mut threads = Vec::new();
    let sync_startup = remote_storage::start_local_timeline_sync(conf, async_shutdown_rx)
        .context("Failed to set up local files sync with external storage")?;
    if let Some(handle) = sync_startup.sync_loop_handle {
        threads.push(handle);
    }
    // Initialize tenant manager.
    tenant_mgr::set_timeline_states(conf, sync_startup.initial_timeline_states);
@@ -221,25 +216,27 @@ fn start_pageserver(conf: &'static PageServerConf, daemonize: bool) -> Result<()
    // Spawn a new thread for the http endpoint
    // bind before launching separate thread so the error reported before startup exits
-    let cloned = auth.clone();
+    let auth_cloned = auth.clone();
-    threads.push(
+    thread_mgr::spawn(
-        thread::Builder::new()
+        ThreadKind::HttpEndpointListener,
-            .name("http_endpoint_thread".into())
+        None,
-            .spawn(move || {
+        None,
-                let router = http::make_router(conf, cloned);
+        "http_endpoint_thread",
-                endpoint::serve_thread_main(router, http_listener)
+        move || {
-            })?,
+            let router = http::make_router(conf, auth_cloned);
-    );
+            endpoint::serve_thread_main(router, http_listener, thread_mgr::shutdown_watcher())
        },
    )?;
-    // Spawn a thread to listen for connections. It will spawn further threads
+    // Spawn a thread to listen for libpq connections. It will spawn further threads
    // for each connection.
-    threads.push(
+    thread_mgr::spawn(
-        thread::Builder::new()
+        ThreadKind::LibpqEndpointListener,
-            .name("Page Service thread".into())
+        None,
-            .spawn(move || {
+        None,
-                page_service::thread_main(conf, auth, pageserver_listener, conf.auth_type)
+        "libpq endpoint thread",
-            })?,
+        move || page_service::thread_main(conf, auth, pageserver_listener, conf.auth_type),
-    );
+    )?;
    signals.handle(|signal| match signal {
        Signal::Quit => {
@@ -255,21 +252,38 @@ fn start_pageserver(conf: &'static PageServerConf, daemonize: bool) -> Result<()
                "Got {}. Terminating gracefully in fast shutdown mode",
                signal.name()
            );
-
+            shutdown_pageserver();
-            async_shutdown_tx.send(())?;
+            unreachable!()
            postgres_backend::set_pgbackend_shutdown_requested();
            tenant_mgr::shutdown_all_tenants()?;
            endpoint::shutdown();
            for handle in std::mem::take(&mut threads) {
                handle
                    .join()
                    .expect("thread panicked")
                    .expect("thread exited with an error");
            }
            info!("Shut down successfully completed");
            std::process::exit(0);
        }
    })
 }
 fn shutdown_pageserver() {
    // Shut down the libpq endpoint thread. This prevents new connections from
    // being accepted.
    thread_mgr::shutdown_threads(Some(ThreadKind::LibpqEndpointListener), None, None);
    // Shut down any page service threads.
    postgres_backend::set_pgbackend_shutdown_requested();
    thread_mgr::shutdown_threads(Some(ThreadKind::PageRequestHandler), None, None);
    // Shut down all the tenants. This flushes everything to disk and kills
    // the checkpoint and GC threads.
    tenant_mgr::shutdown_all_tenants();
    // Stop syncing with remote storage.
    //
    // FIXME: Does this wait for the sync thread to finish syncing what's queued up?
    // Should it?
    thread_mgr::shutdown_threads(Some(ThreadKind::StorageSync), None, None);
    // Shut down the HTTP endpoint last, so that you can still check the server's
    // status while it's shutting down.
    thread_mgr::shutdown_threads(Some(ThreadKind::HttpEndpointListener), None, None);
    // There should be nothing left, but let's be sure
    thread_mgr::shutdown_threads(None, None, None);
    info!("Shut down successfully completed");
    std::process::exit(0);
 }
@@ -4,7 +4,7 @@
 // TODO: move all paths construction to conf impl
 //
-use anyhow::{anyhow, bail, Context, Result};
+use anyhow::{bail, Context, Result};
 use postgres_ffi::ControlFileData;
 use serde::{Deserialize, Serialize};
 use std::{
@@ -118,7 +118,7 @@ pub fn init_pageserver(conf: &'static PageServerConf, create_tenant: Option<&str
    if let Some(tenantid) = create_tenant {
        let tenantid = ZTenantId::from_str(tenantid)?;
        println!("initializing tenantid {}", tenantid);
-        create_repo(conf, tenantid, dummy_redo_mgr).with_context(|| "failed to create repo")?;
+        create_repo(conf, tenantid, dummy_redo_mgr).context("failed to create repo")?;
    }
    crashsafe_dir::create_dir_all(conf.tenants_path())?;
@@ -197,7 +197,7 @@ fn run_initdb(conf: &'static PageServerConf, initdbpath: &Path) -> Result<()> {
        .env("DYLD_LIBRARY_PATH", conf.pg_lib_dir().to_str().unwrap())
        .stdout(Stdio::null())
        .output()
-        .with_context(|| "failed to execute initdb")?;
+        .context("failed to execute initdb")?;
    if !initdb_output.status.success() {
        anyhow::bail!(
            "initdb failed: '{}'",
@@ -308,7 +308,7 @@ pub(crate) fn create_branch(
    let timeline = repo
        .get_timeline(startpoint.timelineid)?
        .local_timeline()
-        .ok_or_else(|| anyhow!("Cannot branch off the timeline that's not present locally"))?;
+        .context("Cannot branch off the timeline that's not present locally")?;
    if startpoint.lsn == Lsn(0) {
        // Find end of WAL on the old timeline
        let end_of_wal = timeline.get_last_record_lsn();
@@ -324,12 +324,13 @@ pub(crate) fn create_branch(
        timeline.wait_lsn(startpoint.lsn)?;
    }
    startpoint.lsn = startpoint.lsn.align();
-    if timeline.get_start_lsn() > startpoint.lsn {
+    if timeline.get_ancestor_lsn() > startpoint.lsn {
        // can we safely just branch from the ancestor instead?
        anyhow::bail!(
-            "invalid startpoint {} for the branch {}: less than timeline start {}",
+            "invalid startpoint {} for the branch {}: less than timeline ancestor lsn {:?}",
            startpoint.lsn,
            branchname,
-            timeline.get_start_lsn()
+            timeline.get_ancestor_lsn()
        );
    }
@@ -383,14 +384,11 @@ fn parse_point_in_time(
    let mut strings = s.split('@');
    let name = strings.next().unwrap();
-    let lsn: Option<Lsn>;
+    let lsn = strings
-    if let Some(lsnstr) = strings.next() {
+        .next()
-        lsn = Some(
+        .map(Lsn::from_str)
-            Lsn::from_str(lsnstr).with_context(|| "invalid LSN in point-in-time specification")?,
+        .transpose()
-        );
+        .context("invalid LSN in point-in-time specification")?;
    } else {
        lsn = None
    }
    // Check if it's a tag
    if lsn.is_none() {
@@ -4,7 +4,7 @@
 //! file, or on the command line.
 //! See also `settings.md` for better description on every parameter.
-use anyhow::{anyhow, bail, ensure, Context, Result};
+use anyhow::{bail, ensure, Context, Result};
 use toml_edit;
 use toml_edit::{Document, Item};
 use zenith_utils::postgres_backend::AuthType;
@@ -144,6 +144,13 @@ pub struct S3Config {
    pub access_key_id: Option<String>,
    /// "Password" to use when connecting to bucket.
    pub secret_access_key: Option<String>,
    /// A base URL to send S3 requests to.
    /// By default, the endpoint is derived from a region name, assuming it's
    /// an AWS S3 region name, erroring on wrong region name.
    /// Endpoint provides a way to support other S3 flavors and their regions.
    ///
    /// Example: `http://127.0.0.1:5000`
    pub endpoint: Option<String>,
 }
 impl std::fmt::Debug for S3Config {
@@ -306,9 +313,7 @@ impl PageServerConf {
                })
                .ok()
                .and_then(NonZeroUsize::new)
-                .ok_or_else(|| {
+                .context("'max_concurrent_sync' must be a non-zero positive integer")?
                    anyhow!("'max_concurrent_sync' must be a non-zero positive integer")
                })?
        } else {
            NonZeroUsize::new(defaults::DEFAULT_REMOTE_STORAGE_MAX_CONCURRENT_SYNC).unwrap()
        };
@@ -321,7 +326,7 @@ impl PageServerConf {
                })
                .ok()
                .and_then(NonZeroU32::new)
-                .ok_or_else(|| anyhow!("'max_sync_errors' must be a non-zero positive integer"))?
+                .context("'max_sync_errors' must be a non-zero positive integer")?
        } else {
            NonZeroU32::new(defaults::DEFAULT_REMOTE_STORAGE_MAX_SYNC_ERRORS).unwrap()
        };
@@ -351,6 +356,10 @@ impl PageServerConf {
                    .get("prefix_in_bucket")
                    .map(|prefix_in_bucket| parse_toml_string("prefix_in_bucket", prefix_in_bucket))
                    .transpose()?,
                endpoint: toml
                    .get("endpoint")
                    .map(|endpoint| parse_toml_string("endpoint", endpoint))
                    .transpose()?,
            }),
            (Some(local_path), None, None) => RemoteStorageKind::LocalFs(PathBuf::from(
                parse_toml_string("local_path", local_path)?,
@@ -396,7 +405,7 @@ impl PageServerConf {
 fn parse_toml_string(name: &str, item: &Item) -> Result<String> {
    let s = item
        .as_str()
-        .ok_or_else(|| anyhow!("configure option {} is not a string", name))?;
+        .with_context(|| format!("configure option {} is not a string", name))?;
    Ok(s.to_string())
 }
@@ -405,7 +414,7 @@ fn parse_toml_u64(name: &str, item: &Item) -> Result<u64> {
    // for our use, though.
    let i: i64 = item
        .as_integer()
-        .ok_or_else(|| anyhow!("configure option {} is not an integer", name))?;
+        .with_context(|| format!("configure option {} is not an integer", name))?;
    if i < 0 {
        bail!("configure option {} cannot be negative", name);
    }
@@ -415,7 +424,7 @@ fn parse_toml_u64(name: &str, item: &Item) -> Result<u64> {
 fn parse_toml_duration(name: &str, item: &Item) -> Result<Duration> {
    let s = item
        .as_str()
-        .ok_or_else(|| anyhow!("configure option {} is not a string", name))?;
+        .with_context(|| format!("configure option {} is not a string", name))?;
    Ok(humantime::parse_duration(s)?)
 }
@@ -423,7 +432,7 @@ fn parse_toml_duration(name: &str, item: &Item) -> Result<Duration> {
 fn parse_toml_auth_type(name: &str, item: &Item) -> Result<AuthType> {
    let v = item
        .as_str()
-        .ok_or_else(|| anyhow!("configure option {} is not a string", name))?;
+        .with_context(|| format!("configure option {} is not a string", name))?;
    AuthType::from_str(v)
 }
@@ -599,6 +608,7 @@ pg_distrib_dir='{}'
        let prefix_in_bucket = "test_prefix".to_string();
        let access_key_id = "SOMEKEYAAAAASADSAH*#".to_string();
        let secret_access_key = "SOMEsEcReTsd292v".to_string();
        let endpoint = "http://localhost:5000".to_string();
        let max_concurrent_sync = NonZeroUsize::new(111).unwrap();
        let max_sync_errors = NonZeroU32::new(222).unwrap();
@@ -611,12 +621,13 @@ bucket_name = '{}'
 bucket_region = '{}'
 prefix_in_bucket = '{}'
 access_key_id = '{}'
-secret_access_key = '{}'"#,
+secret_access_key = '{}'
-                max_concurrent_sync, max_sync_errors, bucket_name, bucket_region, prefix_in_bucket, access_key_id, secret_access_key
+endpoint = '{}'"#,
                max_concurrent_sync, max_sync_errors, bucket_name, bucket_region, prefix_in_bucket, access_key_id, secret_access_key, endpoint
            ),
            format!(
-                "remote_storage={{max_concurrent_sync={}, max_sync_errors={}, bucket_name='{}', bucket_region='{}', prefix_in_bucket='{}', access_key_id='{}', secret_access_key='{}'}}",
+                "remote_storage={{max_concurrent_sync={}, max_sync_errors={}, bucket_name='{}', bucket_region='{}', prefix_in_bucket='{}', access_key_id='{}', secret_access_key='{}', endpoint='{}'}}",
-                max_concurrent_sync, max_sync_errors, bucket_name, bucket_region, prefix_in_bucket, access_key_id, secret_access_key
+                max_concurrent_sync, max_sync_errors, bucket_name, bucket_region, prefix_in_bucket, access_key_id, secret_access_key, endpoint
            ),
        ];
@@ -650,7 +661,8 @@ pg_distrib_dir='{}'
                        bucket_region: bucket_region.clone(),
                        access_key_id: Some(access_key_id.clone()),
                        secret_access_key: Some(secret_access_key.clone()),
-                        prefix_in_bucket: Some(prefix_in_bucket.clone())
+                        prefix_in_bucket: Some(prefix_in_bucket.clone()),
                        endpoint: Some(endpoint.clone())
                    }),
                },
                "Remote storage config should correctly parse the S3 config"
@@ -25,6 +25,7 @@ use zenith_utils::zid::{opt_display_serde, ZTimelineId};
 use super::models::BranchCreateRequest;
 use super::models::TenantCreateRequest;
 use crate::branches::BranchInfo;
 use crate::repository::RepositoryTimeline;
 use crate::repository::TimelineSyncState;
 use crate::{branches, config::PageServerConf, tenant_mgr, ZTenantId};
@@ -201,7 +202,6 @@ enum TimelineInfo {
        ancestor_timeline_id: Option<ZTimelineId>,
        last_record_lsn: Lsn,
        prev_record_lsn: Lsn,
        start_lsn: Lsn,
        disk_consistent_lsn: Lsn,
        timeline_state: Option<TimelineSyncState>,
    },
@@ -236,7 +236,6 @@ async fn timeline_detail_handler(request: Request<Body>) -> Result<Response<Body
                disk_consistent_lsn: timeline.get_disk_consistent_lsn(),
                last_record_lsn: timeline.get_last_record_lsn(),
                prev_record_lsn: timeline.get_prev_record_lsn(),
                start_lsn: timeline.get_start_lsn(),
                timeline_state: repo.get_timeline_state(timeline_id),
            },
        })
@@ -247,6 +246,58 @@ async fn timeline_detail_handler(request: Request<Body>) -> Result<Response<Body
    Ok(json_response(StatusCode::OK, response_data)?)
 }
 async fn timeline_attach_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
    let tenant_id: ZTenantId = parse_request_param(&request, "tenant_id")?;
    check_permission(&request, Some(tenant_id))?;
    let timeline_id: ZTimelineId = parse_request_param(&request, "timeline_id")?;
    tokio::task::spawn_blocking(move || {
        let _enter =
            info_span!("timeline_attach_handler", tenant = %tenant_id, timeline = %timeline_id)
                .entered();
        let repo = tenant_mgr::get_repository_for_tenant(tenant_id)?;
        match repo.get_timeline(timeline_id)? {
            RepositoryTimeline::Local(_) => {
                anyhow::bail!("Timeline with id {} is already local", timeline_id)
            }
            RepositoryTimeline::Remote {
                id: _,
                disk_consistent_lsn: _,
            } => {
                // FIXME (rodionov) get timeline already schedules timeline for download, and duplicate tasks can cause errors
                //  first should be fixed in https://github.com/zenithdb/zenith/issues/997
                // TODO (rodionov) change timeline state to awaits download (incapsulate it somewhere in the repo)
                // TODO (rodionov) can we safely request replication on the timeline before sync is completed? (can be implemented on top of the #997)
                Ok(())
            }
        }
    })
    .await
    .map_err(ApiError::from_err)??;
    Ok(json_response(StatusCode::ACCEPTED, ())?)
 }
 async fn timeline_detach_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
    let tenant_id: ZTenantId = parse_request_param(&request, "tenant_id")?;
    check_permission(&request, Some(tenant_id))?;
    let timeline_id: ZTimelineId = parse_request_param(&request, "timeline_id")?;
    tokio::task::spawn_blocking(move || {
        let _enter =
            info_span!("timeline_detach_handler", tenant = %tenant_id, timeline = %timeline_id)
                .entered();
        let repo = tenant_mgr::get_repository_for_tenant(tenant_id)?;
        repo.detach_timeline(timeline_id)
    })
    .await
    .map_err(ApiError::from_err)??;
    Ok(json_response(StatusCode::OK, ())?)
 }
 async fn tenant_list_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
    // check for management permission
    check_permission(&request, None)?;
@@ -267,13 +318,13 @@ async fn tenant_create_handler(mut request: Request<Body>) -> Result<Response<Bo
    let request_data: TenantCreateRequest = json_request(&mut request).await?;
-    let response_data = tokio::task::spawn_blocking(move || {
+    tokio::task::spawn_blocking(move || {
        let _enter = info_span!("tenant_create", tenant = %request_data.tenant_id).entered();
        tenant_mgr::create_repository_for_tenant(get_config(&request), request_data.tenant_id)
    })
    .await
    .map_err(ApiError::from_err)??;
-    Ok(json_response(StatusCode::CREATED, response_data)?)
+    Ok(json_response(StatusCode::CREATED, ())?)
 }
 async fn handler_404(_: Request<Body>) -> Result<Response<Body>, ApiError> {
@@ -308,6 +359,14 @@ pub fn make_router(
            "/v1/timeline/:tenant_id/:timeline_id",
            timeline_detail_handler,
        )
        .post(
            "/v1/timeline/:tenant_id/:timeline_id/attach",
            timeline_attach_handler,
        )
        .post(
            "/v1/timeline/:tenant_id/:timeline_id/detach",
            timeline_detach_handler,
        )
        .get("/v1/branch/:tenant_id", branch_list_handler)
        .get("/v1/branch/:tenant_id/:branch_name", branch_detail_handler)
        .post("/v1/branch", branch_create_handler)
@@ -7,7 +7,7 @@ use std::fs::File;
 use std::io::{Read, Seek, SeekFrom};
 use std::path::{Path, PathBuf};
-use anyhow::{anyhow, bail, ensure, Result};
+use anyhow::{bail, ensure, Context, Result};
 use bytes::Bytes;
 use tracing::*;
@@ -126,7 +126,7 @@ pub fn import_timeline_from_postgres_datadir(
    writer.advance_last_record_lsn(lsn);
    // We expect the Postgres server to be shut down cleanly.
-    let pg_control = pg_control.ok_or_else(|| anyhow!("pg_control file not found"))?;
+    let pg_control = pg_control.context("pg_control file not found")?;
    ensure!(
        pg_control.state == DBState_DB_SHUTDOWNED,
        "Postgres cluster was not shut down cleanly"
@@ -11,7 +11,7 @@
 //! parent timeline, and the last LSN that has been written to disk.
 //!
-use anyhow::{anyhow, bail, ensure, Context, Result};
+use anyhow::{bail, ensure, Context, Result};
 use bookfile::Book;
 use bytes::Bytes;
 use lazy_static::lazy_static;
@@ -28,7 +28,7 @@ use std::io::Write;
 use std::ops::{Bound::Included, Deref};
 use std::path::{Path, PathBuf};
 use std::sync::atomic::{self, AtomicBool, AtomicUsize};
-use std::sync::{Arc, Mutex, MutexGuard};
+use std::sync::{Arc, Mutex, MutexGuard, RwLock, RwLockReadGuard};
 use std::time::{Duration, Instant};
 use self::metadata::{metadata_path, TimelineMetadata, METADATA_FILE_NAME};
@@ -40,9 +40,8 @@ use crate::repository::{
    BlockNumber, GcResult, Repository, RepositoryTimeline, Timeline, TimelineSyncState,
    TimelineWriter, ZenithWalRecord,
 };
-use crate::tenant_mgr;
+use crate::thread_mgr;
 use crate::virtual_file::VirtualFile;
 use crate::walreceiver;
 use crate::walreceiver::IS_WAL_RECEIVER;
 use crate::walredo::WalRedoManager;
 use crate::CheckpointConfig;
@@ -72,7 +71,6 @@ mod storage_layer;
 use delta_layer::DeltaLayer;
 use ephemeral_file::is_ephemeral_file;
 use filename::{DeltaFileName, ImageFileName};
 use global_layer_map::{LayerId, GLOBAL_LAYER_MAP};
 use image_layer::ImageLayer;
 use inmemory_layer::InMemoryLayer;
 use layer_map::LayerMap;
@@ -168,7 +166,7 @@ impl Repository for LayeredRepository {
        // Create the timeline directory, and write initial metadata to file.
        crashsafe_dir::create_dir_all(self.conf.timeline_path(&timelineid, &self.tenantid))?;
-        let metadata = TimelineMetadata::new(Lsn(0), None, None, Lsn(0), Lsn(0), initdb_lsn);
+        let metadata = TimelineMetadata::new(Lsn(0), None, None, Lsn(0), initdb_lsn, initdb_lsn);
        Self::save_metadata(self.conf, timelineid, self.tenantid, &metadata, true)?;
        let timeline = LayeredTimeline::new(
@@ -202,9 +200,10 @@ impl Repository for LayeredRepository {
                bail!("Cannot branch off the timeline {} that's not local", src)
            }
        };
        let latest_gc_cutoff_lsn = src_timeline.get_latest_gc_cutoff_lsn();
        src_timeline
-            .check_lsn_is_in_scope(start_lsn)
+            .check_lsn_is_in_scope(start_lsn, &latest_gc_cutoff_lsn)
            .context("invalid branch start lsn")?;
        let RecordLsn {
@@ -232,7 +231,7 @@ impl Repository for LayeredRepository {
            dst_prev,
            Some(src),
            start_lsn,
-            src_timeline.latest_gc_cutoff_lsn.load(),
+            *src_timeline.latest_gc_cutoff_lsn.read().unwrap(),
            src_timeline.initdb_lsn,
        );
        crashsafe_dir::create_dir_all(self.conf.timeline_path(&dst, &self.tenantid))?;
@@ -286,15 +285,42 @@ impl Repository for LayeredRepository {
        Ok(())
    }
-    // Wait for all threads to complete and persist repository data before pageserver shutdown.
+    // Detaches the timeline from the repository.
-    fn shutdown(&self) -> Result<()> {
+    fn detach_timeline(&self, timeline_id: ZTimelineId) -> Result<()> {
-        trace!("LayeredRepository shutdown for tenant {}", self.tenantid);
+        let mut timelines = self.timelines.lock().unwrap();
        match timelines.entry(timeline_id) {
            Entry::Vacant(_) => {
                bail!("cannot detach non existing timeline");
            }
            Entry::Occupied(mut entry) => {
                let timeline_entry = entry.get_mut();
-        let timelines = self.timelines.lock().unwrap();
+                let timeline = match timeline_entry {
-        for (timelineid, timeline) in timelines.iter() {
+                    LayeredTimelineEntry::Remote { .. } => {
-            shutdown_timeline(self.tenantid, *timelineid, timeline)?;
+                        bail!("cannot detach remote timeline {}", timeline_id);
-        }
+                    }
                    LayeredTimelineEntry::Local(timeline) => timeline,
                };
                // TODO (rodionov) keep local state in timeline itself (refactoring related to https://github.com/zenithdb/zenith/issues/997 and #1104)
                // FIXME this is local disk consistent lsn, need to keep the latest succesfully uploaded checkpoint lsn in timeline (metadata?)
                //  https://github.com/zenithdb/zenith/issues/1104
                let remote_disk_consistent_lsn = timeline.disk_consistent_lsn.load();
                // reference to timeline is dropped here
                entry.insert(LayeredTimelineEntry::Remote {
                    id: timeline_id,
                    disk_consistent_lsn: remote_disk_consistent_lsn,
                });
            }
        };
        // Release the lock to shutdown and remove the files without holding it
        drop(timelines);
        // shutdown the timeline (this shuts down the walreceiver)
        thread_mgr::shutdown_threads(None, Some(self.tenantid), Some(timeline_id));
        // remove timeline files (maybe avoid this for ease of debugging if something goes wrong)
        fs::remove_dir_all(self.conf.timeline_path(&timeline_id, &self.tenantid))?;
        Ok(())
    }
@@ -307,9 +333,13 @@ impl Repository for LayeredRepository {
        timeline_id: ZTimelineId,
        new_state: TimelineSyncState,
    ) -> Result<()> {
        debug!(
            "set_timeline_state: timeline_id: {}, new_state: {:?}",
            timeline_id, new_state
        );
        let mut timelines_accessor = self.timelines.lock().unwrap();
-        let timeline_to_shutdown = match new_state {
+        match new_state {
            TimelineSyncState::Ready(_) => {
                let reloaded_timeline =
                    self.init_local_timeline(timeline_id, &mut timelines_accessor)?;
@@ -327,12 +357,9 @@ impl Repository for LayeredRepository {
                },
            ),
        };
        // NOTE we do not delete local data in case timeline became cloud only, this is performed in detach_timeline
        drop(timelines_accessor);
        if let Some(timeline) = timeline_to_shutdown {
            shutdown_timeline(self.tenantid, timeline_id, &timeline)?;
        }
        Ok(())
    }
@@ -358,30 +385,6 @@ impl Repository for LayeredRepository {
    }
 }
 fn shutdown_timeline(
    tenant_id: ZTenantId,
    timeline_id: ZTimelineId,
    timeline: &LayeredTimelineEntry,
 ) -> Result<(), anyhow::Error> {
    match timeline {
        LayeredTimelineEntry::Local(timeline) => {
            timeline
                .upload_relishes
                .store(false, atomic::Ordering::Relaxed);
            walreceiver::stop_wal_receiver(tenant_id, timeline_id);
            trace!("repo shutdown. checkpoint timeline {}", timeline_id);
            // Do not reconstruct pages to reduce shutdown time
            timeline.checkpoint(CheckpointConfig::Flush)?;
            //TODO Wait for walredo process to shutdown too
        }
        LayeredTimelineEntry::Remote { .. } => warn!(
            "Skipping shutdown of a remote timeline {} for tenant {}",
            timeline_id, tenant_id
        ),
    }
    Ok(())
 }
 #[derive(Clone)]
 enum LayeredTimelineEntry {
    Local(Arc<LayeredTimeline>),
@@ -608,7 +611,7 @@ impl LayeredRepository {
            }
        }
-        //Now collect info about branchpoints
+        // Now collect info about branchpoints
        let mut all_branchpoints: BTreeSet<(ZTimelineId, Lsn)> = BTreeSet::new();
        for &timelineid in &timelineids {
            let timeline = match self.get_or_init_timeline(timelineid, &mut timelines)? {
@@ -652,8 +655,10 @@ impl LayeredRepository {
        // Ok, we now know all the branch points.
        // Perform GC for each timeline.
        for timelineid in timelineids {
-            if tenant_mgr::shutdown_requested() {
+            if thread_mgr::is_shutdown_requested() {
-                return Ok(totals);
+                // We were requested to shut down. Stop and return with the progress we
                // made.
                break;
            }
            // We have already loaded all timelines above
@@ -690,7 +695,6 @@ impl LayeredRepository {
                    timeline.checkpoint(CheckpointConfig::Forced)?;
                    info!("timeline {} checkpoint_before_gc done", timelineid);
                }
                let result = timeline.gc_timeline(branchpoints, cutoff)?;
                totals += result;
@@ -779,7 +783,7 @@ pub struct LayeredTimeline {
    checkpoint_cs: Mutex<()>,
    // Needed to ensure that we can't create a branch at a point that was already garbage collected
-    latest_gc_cutoff_lsn: AtomicLsn,
+    latest_gc_cutoff_lsn: RwLock<Lsn>,
    // It may change across major versions so for simplicity
    // keep it after running initdb for a timeline.
@@ -823,6 +827,10 @@ impl Timeline for LayeredTimeline {
        Ok(())
    }
    fn get_latest_gc_cutoff_lsn(&self) -> RwLockReadGuard<Lsn> {
        self.latest_gc_cutoff_lsn.read().unwrap()
    }
    /// Look up given page version.
    fn get_page_at_lsn(&self, rel: RelishTag, rel_blknum: BlockNumber, lsn: Lsn) -> Result<Bytes> {
        if !rel.is_blocky() && rel_blknum != 0 {
@@ -833,14 +841,6 @@ impl Timeline for LayeredTimeline {
            );
        }
        debug_assert!(lsn <= self.get_last_record_lsn());
        let latest_gc_cutoff_lsn = self.latest_gc_cutoff_lsn.load();
        // error instead of assert to simplify testing
        ensure!(
            lsn >= latest_gc_cutoff_lsn,
            "tried to request a page version that was garbage collected. requested at {} gc cutoff {}",
            lsn, latest_gc_cutoff_lsn
        );
        let (seg, seg_blknum) = SegmentTag::from_blknum(rel, rel_blknum);
        if let Some((layer, lsn)) = self.get_layer_for_read(seg, lsn)? {
@@ -1011,21 +1011,16 @@ impl Timeline for LayeredTimeline {
    ///
    /// Validate lsn against initdb_lsn and latest_gc_cutoff_lsn.
    ///
-    fn check_lsn_is_in_scope(&self, lsn: Lsn) -> Result<()> {
+    fn check_lsn_is_in_scope(
-        let initdb_lsn = self.initdb_lsn;
+        &self,
        lsn: Lsn,
        latest_gc_cutoff_lsn: &RwLockReadGuard<Lsn>,
    ) -> Result<()> {
        ensure!(
-            lsn >= initdb_lsn,
+            lsn >= **latest_gc_cutoff_lsn,
            "LSN {} is earlier than initdb lsn {}",
            lsn,
            initdb_lsn,
        );
        let latest_gc_cutoff_lsn = self.latest_gc_cutoff_lsn.load();
        ensure!(
            lsn >= latest_gc_cutoff_lsn,
            "LSN {} is earlier than latest GC horizon {} (we might've already garbage collected needed data)",
            lsn,
-            latest_gc_cutoff_lsn,
+            **latest_gc_cutoff_lsn,
        );
        Ok(())
    }
@@ -1042,14 +1037,6 @@ impl Timeline for LayeredTimeline {
        self.last_record_lsn.load()
    }
    fn get_start_lsn(&self) -> Lsn {
        self.ancestor_timeline
            .as_ref()
            .and_then(|ancestor_entry| ancestor_entry.local_or_schedule_download(self.tenantid))
            .map(Timeline::get_start_lsn)
            .unwrap_or(self.ancestor_lsn)
    }
    fn get_current_logical_size(&self) -> usize {
        self.current_logical_size.load(atomic::Ordering::Acquire) as usize
    }
@@ -1139,7 +1126,7 @@ impl LayeredTimeline {
            write_lock: Mutex::new(()),
            checkpoint_cs: Mutex::new(()),
-            latest_gc_cutoff_lsn: AtomicLsn::from(metadata.latest_gc_cutoff_lsn()),
+            latest_gc_cutoff_lsn: RwLock::new(metadata.latest_gc_cutoff_lsn()),
            initdb_lsn: metadata.initdb_lsn(),
        }
    }
@@ -1165,8 +1152,8 @@ impl LayeredTimeline {
                // create an ImageLayer struct for each image file.
                if imgfilename.lsn > disk_consistent_lsn {
                    warn!(
-                        "found future image layer {} on timeline {}",
+                        "found future image layer {} on timeline {} disk_consistent_lsn is {}",
-                        imgfilename, self.timelineid
+                        imgfilename, self.timelineid, disk_consistent_lsn
                    );
                    rename_to_backup(direntry.path())?;
@@ -1189,8 +1176,8 @@ impl LayeredTimeline {
                // before crash.
                if deltafilename.end_lsn > disk_consistent_lsn + 1 {
                    warn!(
-                        "found future delta layer {} on timeline {}",
+                        "found future delta layer {} on timeline {} disk_consistent_lsn is {}",
-                        deltafilename, self.timelineid
+                        deltafilename, self.timelineid, disk_consistent_lsn
                    );
                    rename_to_backup(direntry.path())?;
@@ -1386,7 +1373,7 @@ impl LayeredTimeline {
                    self.tenantid,
                    seg,
                    lsn,
-                    lsn,
+                    last_record_lsn,
                )?;
            } else {
                return Ok(open_layer);
@@ -1429,7 +1416,7 @@ impl LayeredTimeline {
                self.timelineid,
                self.tenantid,
                start_lsn,
-                lsn,
+                last_record_lsn,
            )?;
        } else {
            // New relation.
@@ -1440,8 +1427,14 @@ impl LayeredTimeline {
                lsn
            );
-            layer =
+            layer = InMemoryLayer::create(
-                InMemoryLayer::create(self.conf, self.timelineid, self.tenantid, seg, lsn, lsn)?;
+                self.conf,
                self.timelineid,
                self.tenantid,
                seg,
                lsn,
                last_record_lsn,
            )?;
        }
        let layer_rc: Arc<InMemoryLayer> = Arc::new(layer);
@@ -1458,7 +1451,7 @@ impl LayeredTimeline {
        // Prevent concurrent checkpoints
        let _checkpoint_cs = self.checkpoint_cs.lock().unwrap();
-        let mut write_guard = self.write_lock.lock().unwrap();
+        let write_guard = self.write_lock.lock().unwrap();
        let mut layers = self.layers.lock().unwrap();
        // Bump the generation number in the layer map, so that we can distinguish
@@ -1484,11 +1477,17 @@ impl LayeredTimeline {
        let mut disk_consistent_lsn = last_record_lsn;
        let mut layer_paths = Vec::new();
        let mut freeze_end_lsn = Lsn(0);
        let mut evicted_layers = Vec::new();
        //
        // Determine which layers we need to evict and calculate max(latest_lsn)
        // among those layers.
        //
        while let Some((oldest_layer_id, oldest_layer, oldest_generation)) =
            layers.peek_oldest_open()
        {
-            let oldest_pending_lsn = oldest_layer.get_oldest_pending_lsn();
+            let oldest_lsn = oldest_layer.get_oldest_lsn();
            // Does this layer need freezing?
            //
            // Write out all in-memory layers that contain WAL older than CHECKPOINT_DISTANCE.
@@ -1497,28 +1496,60 @@ impl LayeredTimeline {
            // when we started. We don't want to process layers inserted after we started, to
            // avoid getting into an infinite loop trying to process again entries that we
            // inserted ourselves.
-            let distance = last_record_lsn.widening_sub(oldest_pending_lsn);
+            //
-            if distance < 0
+            // Once we have decided to write out at least one layer, we must also write out
            // any other layers that contain WAL older than the end LSN of the layers we have
            // already decided to write out. In other words, we must write out all layers
            // whose [oldest_lsn, latest_lsn) range overlaps with any of the other layers
            // that we are writing out. Otherwise, when we advance 'disk_consistent_lsn', it's
            // ambiguous whether those layers are already durable on disk or not. For example,
            // imagine that there are two layers in memory that contain page versions in the
            // following LSN ranges:
            //
            // A: 100-150
            // B: 110-200
            //
            // If we flush layer A, we must also flush layer B, because they overlap. If we
            // flushed only A, and advanced 'disk_consistent_lsn' to 150, we would break the
            // rule that all WAL older than 'disk_consistent_lsn' are durable on disk, because
            // B contains some WAL older than 150. On the other hand, if we flushed out A and
            // advanced 'disk_consistent_lsn' only up to 110, after crash and restart we would
            // delete the first layer because its end LSN is larger than 110. If we changed
            // the deletion logic to not delete it, then we would start streaming at 110, and
            // process again the WAL records in the range 110-150 that are already in layer A,
            // and the WAL processing code does not cope with that. We solve that dilemma by
            // insisting that if we write out the first layer, we also write out the second
            // layer, and advance disk_consistent_lsn all the way up to 200.
            //
            let distance = last_record_lsn.widening_sub(oldest_lsn);
            if (distance < 0
                || distance < checkpoint_distance.into()
-                || oldest_generation == current_generation
+                || oldest_generation == current_generation)
                && oldest_lsn >= freeze_end_lsn
            // this layer intersects with evicted layer and so also need to be evicted
            {
                info!(
                    "the oldest layer is now {} which is {} bytes behind last_record_lsn",
                    oldest_layer.filename().display(),
                    distance
                );
-                disk_consistent_lsn = oldest_pending_lsn;
+                disk_consistent_lsn = oldest_lsn;
                break;
            }
            let latest_lsn = oldest_layer.get_latest_lsn();
            if latest_lsn > freeze_end_lsn {
                freeze_end_lsn = latest_lsn; // calculate max of latest_lsn of the layers we're about to evict
            }
            layers.remove_open(oldest_layer_id);
            evicted_layers.push((oldest_layer_id, oldest_layer));
        }
-            drop(layers);
+        // Freeze evicted layers
-            drop(write_guard);
+        for (_evicted_layer_id, evicted_layer) in evicted_layers.iter() {
-
+            // Mark the layer as no longer accepting writes and record the end_lsn.
-            let mut this_layer_paths = self.evict_layer(oldest_layer_id, reconstruct_pages)?;
+            // This happens in-place, no new layers are created now.
-            layer_paths.append(&mut this_layer_paths);
+            evicted_layer.freeze(freeze_end_lsn);
-
+            layers.insert_historic(evicted_layer.clone());
            write_guard = self.write_lock.lock().unwrap();
            layers = self.layers.lock().unwrap();
        }
        // Call unload() on all frozen layers, to release memory.
@@ -1531,6 +1562,14 @@ impl LayeredTimeline {
        drop(layers);
        drop(write_guard);
        // Create delta/image layers for evicted layers
        for (_evicted_layer_id, evicted_layer) in evicted_layers.iter() {
            let mut this_layer_paths =
                self.evict_layer(evicted_layer.clone(), reconstruct_pages)?;
            layer_paths.append(&mut this_layer_paths);
        }
        // Sync layers
        if !layer_paths.is_empty() {
            // We must fsync the timeline dir to ensure the directory entries for
            // new layer files are durable
@@ -1571,7 +1610,7 @@ impl LayeredTimeline {
                ondisk_prev_record_lsn,
                ancestor_timelineid,
                self.ancestor_lsn,
-                self.latest_gc_cutoff_lsn.load(),
+                *self.latest_gc_cutoff_lsn.read().unwrap(),
                self.initdb_lsn,
            );
@@ -1598,52 +1637,29 @@ impl LayeredTimeline {
        Ok(())
    }
-    fn evict_layer(&self, layer_id: LayerId, reconstruct_pages: bool) -> Result<Vec<PathBuf>> {
+    fn evict_layer(
-        // Mark the layer as no longer accepting writes and record the end_lsn.
+        &self,
-        // This happens in-place, no new layers are created now.
+        layer: Arc<InMemoryLayer>,
-        // We call `get_last_record_lsn` again, which may be different from the
+        reconstruct_pages: bool,
-        // original load, as we may have released the write lock since then.
+    ) -> Result<Vec<PathBuf>> {
-
+        let new_historics = layer.write_to_disk(self, reconstruct_pages)?;
        let mut write_guard = self.write_lock.lock().unwrap();
        let mut layers = self.layers.lock().unwrap();
        let mut layer_paths = Vec::new();
        let _write_guard = self.write_lock.lock().unwrap();
        let mut layers = self.layers.lock().unwrap();
-        let global_layer_map = GLOBAL_LAYER_MAP.read().unwrap();
+        // Finally, replace the frozen in-memory layer with the new on-disk layers
-        if let Some(oldest_layer) = global_layer_map.get(&layer_id) {
+        layers.remove_historic(layer);
            drop(global_layer_map);
            oldest_layer.freeze(self.get_last_record_lsn());
-            // The layer is no longer open, update the layer map to reflect this.
+        // Add the historics to the LayerMap
-            // We will replace it with on-disk historics below.
+        for delta_layer in new_historics.delta_layers {
-            layers.remove_open(layer_id);
+            layer_paths.push(delta_layer.path());
-            layers.insert_historic(oldest_layer.clone());
+            layers.insert_historic(Arc::new(delta_layer));
-
+        }
-            // Write the now-frozen layer to disk. That could take a while, so release the lock while do it
+        for image_layer in new_historics.image_layers {
-            drop(layers);
+            layer_paths.push(image_layer.path());
-            drop(write_guard);
+            layers.insert_historic(Arc::new(image_layer));
            let new_historics = oldest_layer.write_to_disk(self, reconstruct_pages)?;
            write_guard = self.write_lock.lock().unwrap();
            layers = self.layers.lock().unwrap();
            // Finally, replace the frozen in-memory layer with the new on-disk layers
            layers.remove_historic(oldest_layer);
            // Add the historics to the LayerMap
            for delta_layer in new_historics.delta_layers {
                layer_paths.push(delta_layer.path());
                layers.insert_historic(Arc::new(delta_layer));
            }
            for image_layer in new_historics.image_layers {
                layer_paths.push(image_layer.path());
                layers.insert_historic(Arc::new(image_layer));
            }
        }
        drop(layers);
        drop(write_guard);
        Ok(layer_paths)
    }
@@ -1672,12 +1688,14 @@ impl LayeredTimeline {
    pub fn gc_timeline(&self, retain_lsns: Vec<Lsn>, cutoff: Lsn) -> Result<GcResult> {
        let now = Instant::now();
        let mut result: GcResult = Default::default();
        let disk_consistent_lsn = self.get_disk_consistent_lsn();
        let _checkpoint_cs = self.checkpoint_cs.lock().unwrap();
        let _enter = info_span!("garbage collection", timeline = %self.timelineid, tenant = %self.tenantid, cutoff = %cutoff).entered();
        // We need to ensure that no one branches at a point before latest_gc_cutoff_lsn.
        // See branch_timeline() for details.
-        self.latest_gc_cutoff_lsn.store(cutoff);
+        *self.latest_gc_cutoff_lsn.write().unwrap() = cutoff;
        info!("GC starting");
@@ -1757,7 +1775,12 @@ impl LayeredTimeline {
            }
            // 3. Is there a later on-disk layer for this relation?
-            if !l.is_dropped() && !layers.newer_image_layer_exists(l.get_seg_tag(), l.get_end_lsn())
+            if !l.is_dropped()
                && !layers.newer_image_layer_exists(
                    l.get_seg_tag(),
                    l.get_end_lsn(),
                    disk_consistent_lsn,
                )
            {
                info!(
                    "keeping {} {}-{} because it is the latest layer",
@@ -2210,11 +2233,10 @@ impl<'a> TimelineWriter for LayeredTimelineWriter<'a> {
        let oldsize = self
            .tl
            .get_relish_size(rel, self.tl.get_last_record_lsn())?
-            .ok_or_else(|| {
+            .with_context(|| {
-                anyhow!(
+                format!(
                    "attempted to truncate non-existent relish {} at {}",
-                    rel,
+                    rel, lsn
                    lsn
                )
            })?;
@@ -2337,8 +2359,5 @@ fn rename_to_backup(path: PathBuf) -> anyhow::Result<()> {
        }
    }
-    Err(anyhow!(
+    bail!("couldn't find an unused backup number for {:?}", path)
        "couldn't find an unused backup number for {:?}",
        path
    ))
 }
@@ -169,7 +169,7 @@ impl DeltaLayerInner {
        if let Some((_entry_lsn, entry)) = slice.last() {
            Ok(*entry)
        } else {
-            Err(anyhow::anyhow!("could not find seg size in delta layer"))
+            bail!("could not find seg size in delta layer")
        }
    }
 }
@@ -173,7 +173,14 @@ impl Layer for ImageLayer {
                    .as_ref()
                    .unwrap()
                    .chapter_reader(BLOCKY_IMAGES_CHAPTER)?;
-                chapter.read_exact_at(&mut buf, offset)?;
+
                chapter.read_exact_at(&mut buf, offset).with_context(|| {
                    format!(
                        "failed to read page from data file {} at offset {}",
                        self.filename().display(),
                        offset
                    )
                })?;
                buf
            }
@@ -39,8 +39,20 @@ pub struct InMemoryLayer {
    ///
    start_lsn: Lsn,
-    /// LSN of the oldest page version stored in this layer
+    ///
-    oldest_pending_lsn: Lsn,
+    /// LSN of the oldest page version stored in this layer.
    ///
    /// This is different from 'start_lsn' in that we enforce that the 'start_lsn'
    /// of a layer always matches the 'end_lsn' of its predecessor, even if there
    /// are no page versions until at a later LSN. That way you can detect any
    /// missing layer files more easily. 'oldest_lsn' is the first page version
    /// actually stored in this layer. In the range between 'start_lsn' and
    /// 'oldest_lsn', there are no changes to the segment.
    /// 'oldest_lsn' is used to adjust 'disk_consistent_lsn' and that is why it should
    /// point to the beginning of WAL record. This is the other difference with 'start_lsn'
    /// which points to end of WAL record. This is why 'oldest_lsn' can be smaller than 'start_lsn'.
    ///
    oldest_lsn: Lsn,
    /// The above fields never change. The parts that do change are in 'inner',
    /// and protected by mutex.
@@ -73,6 +85,14 @@ pub struct InMemoryLayerInner {
    /// a non-blocky rel, 'seg_sizes' is not used and is always empty.
    ///
    seg_sizes: VecMap<Lsn, SegmentBlk>,
    ///
    /// LSN of the newest page version stored in this layer.
    ///
    /// The difference between 'end_lsn' and 'latest_lsn' is the same as between
    /// 'start_lsn' and 'oldest_lsn'. See comments in 'oldest_lsn'.
    ///
    latest_lsn: Lsn,
 }
 impl InMemoryLayerInner {
@@ -319,8 +339,13 @@ pub struct LayersOnDisk {
 impl InMemoryLayer {
    /// Return the oldest page version that's stored in this layer
-    pub fn get_oldest_pending_lsn(&self) -> Lsn {
+    pub fn get_oldest_lsn(&self) -> Lsn {
-        self.oldest_pending_lsn
+        self.oldest_lsn
    }
    pub fn get_latest_lsn(&self) -> Lsn {
        let inner = self.inner.read().unwrap();
        inner.latest_lsn
    }
    ///
@@ -332,7 +357,7 @@ impl InMemoryLayer {
        tenantid: ZTenantId,
        seg: SegmentTag,
        start_lsn: Lsn,
-        oldest_pending_lsn: Lsn,
+        oldest_lsn: Lsn,
    ) -> Result<InMemoryLayer> {
        trace!(
            "initializing new empty InMemoryLayer for writing {} on timeline {} at {}",
@@ -355,13 +380,14 @@ impl InMemoryLayer {
            tenantid,
            seg,
            start_lsn,
-            oldest_pending_lsn,
+            oldest_lsn,
            incremental: false,
            inner: RwLock::new(InMemoryLayerInner {
                end_lsn: None,
                dropped: false,
                page_versions: PageVersions::new(file),
                seg_sizes,
                latest_lsn: oldest_lsn,
            }),
        })
    }
@@ -398,6 +424,8 @@ impl InMemoryLayer {
        let mut inner = self.inner.write().unwrap();
        inner.assert_writeable();
        assert!(lsn >= inner.latest_lsn);
        inner.latest_lsn = lsn;
        let old = inner.page_versions.append_or_update_last(blknum, lsn, pv)?;
@@ -509,12 +537,11 @@ impl InMemoryLayer {
        timelineid: ZTimelineId,
        tenantid: ZTenantId,
        start_lsn: Lsn,
-        oldest_pending_lsn: Lsn,
+        oldest_lsn: Lsn,
    ) -> Result<InMemoryLayer> {
        let seg = src.get_seg_tag();
-        assert!(oldest_pending_lsn.is_aligned());
+        assert!(oldest_lsn.is_aligned());
        assert!(oldest_pending_lsn >= start_lsn);
        trace!(
            "initializing new InMemoryLayer for writing {} on timeline {} at {}",
@@ -538,13 +565,14 @@ impl InMemoryLayer {
            tenantid,
            seg,
            start_lsn,
-            oldest_pending_lsn,
+            oldest_lsn,
            incremental: true,
            inner: RwLock::new(InMemoryLayerInner {
                end_lsn: None,
                dropped: false,
                page_versions: PageVersions::new(file),
                seg_sizes,
                latest_lsn: oldest_lsn,
            }),
        })
    }
@@ -40,7 +40,7 @@ pub struct LayerMap {
    /// All the layers keyed by segment tag
    segs: HashMap<SegmentTag, SegEntry>,
-    /// All in-memory layers, ordered by 'oldest_pending_lsn' and generation
+    /// All in-memory layers, ordered by 'oldest_lsn' and generation
    /// of each layer. This allows easy access to the in-memory layer that
    /// contains the oldest WAL record.
    open_layers: BinaryHeap<OpenLayerEntry>,
@@ -83,16 +83,16 @@ impl LayerMap {
        let layer_id = segentry.update_open(Arc::clone(&layer));
-        let oldest_pending_lsn = layer.get_oldest_pending_lsn();
+        let oldest_lsn = layer.get_oldest_lsn();
-        // After a crash and restart, 'oldest_pending_lsn' of the oldest in-memory
+        // After a crash and restart, 'oldest_lsn' of the oldest in-memory
        // layer becomes the WAL streaming starting point, so it better not point
        // in the middle of a WAL record.
-        assert!(oldest_pending_lsn.is_aligned());
+        assert!(oldest_lsn.is_aligned());
        // Also add it to the binary heap
        let open_layer_entry = OpenLayerEntry {
-            oldest_pending_lsn: layer.get_oldest_pending_lsn(),
+            oldest_lsn: layer.get_oldest_lsn(),
            layer_id,
            generation: self.current_generation,
        };
@@ -191,9 +191,15 @@ impl LayerMap {
    ///
    /// This is used for garbage collection, to determine if an old layer can
    /// be deleted.
-    pub fn newer_image_layer_exists(&self, seg: SegmentTag, lsn: Lsn) -> bool {
+    /// We ignore segments newer than disk_consistent_lsn because they will be removed at restart
    pub fn newer_image_layer_exists(
        &self,
        seg: SegmentTag,
        lsn: Lsn,
        disk_consistent_lsn: Lsn,
    ) -> bool {
        if let Some(segentry) = self.segs.get(&seg) {
-            segentry.newer_image_layer_exists(lsn)
+            segentry.newer_image_layer_exists(lsn, disk_consistent_lsn)
        } else {
            false
        }
@@ -311,13 +317,18 @@ impl SegEntry {
        self.historic.search(lsn)
    }
-    pub fn newer_image_layer_exists(&self, lsn: Lsn) -> bool {
+    pub fn newer_image_layer_exists(&self, lsn: Lsn, disk_consistent_lsn: Lsn) -> bool {
        // We only check on-disk layers, because
        // in-memory layers are not durable
        // The end-LSN is exclusive, while disk_consistent_lsn is
        // inclusive. For example, if disk_consistent_lsn is 100, it is
        // OK for a delta layer to have end LSN 101, but if the end LSN
        // is 102, then it might not have been fully flushed to disk
        // before crash.
        self.historic
            .iter_newer(lsn)
-            .any(|layer| !layer.is_incremental())
+            .any(|layer| !layer.is_incremental() && layer.get_end_lsn() <= disk_consistent_lsn + 1)
    }
    // Set new open layer for a SegEntry.
@@ -341,23 +352,23 @@ impl SegEntry {
 }
 /// Entry held in LayerMap::open_layers, with boilerplate comparison routines
-/// to implement a min-heap ordered by 'oldest_pending_lsn' and 'generation'
+/// to implement a min-heap ordered by 'oldest_lsn' and 'generation'
 ///
 /// The generation number associated with each entry can be used to distinguish
 /// recently-added entries (i.e after last call to increment_generation()) from older
-/// entries with the same 'oldest_pending_lsn'.
+/// entries with the same 'oldest_lsn'.
 struct OpenLayerEntry {
-    oldest_pending_lsn: Lsn, // copy of layer.get_oldest_pending_lsn()
+    oldest_lsn: Lsn, // copy of layer.get_oldest_lsn()
    generation: u64,
    layer_id: LayerId,
 }
 impl Ord for OpenLayerEntry {
    fn cmp(&self, other: &Self) -> Ordering {
        // BinaryHeap is a max-heap, and we want a min-heap. Reverse the ordering here
-        // to get that. Entries with identical oldest_pending_lsn are ordered by generation
+        // to get that. Entries with identical oldest_lsn are ordered by generation
        other
-            .oldest_pending_lsn
+            .oldest_lsn
-            .cmp(&self.oldest_pending_lsn)
+            .cmp(&self.oldest_lsn)
            .then_with(|| other.generation.cmp(&self.generation))
    }
 }
@@ -426,7 +437,7 @@ mod tests {
        conf: &'static PageServerConf,
        segno: u32,
        start_lsn: Lsn,
-        oldest_pending_lsn: Lsn,
+        oldest_lsn: Lsn,
    ) -> Arc<InMemoryLayer> {
        Arc::new(
            InMemoryLayer::create(
@@ -438,7 +449,7 @@ mod tests {
                    segno,
                },
                start_lsn,
-                oldest_pending_lsn,
+                oldest_lsn,
            )
            .unwrap(),
        )
@@ -11,6 +11,7 @@ pub mod remote_storage;
 pub mod repository;
 pub mod tenant_mgr;
 pub mod tenant_threads;
 pub mod thread_mgr;
 pub mod virtual_file;
 pub mod walingest;
 pub mod walreceiver;
@@ -10,16 +10,15 @@
 //     *callmemaybe <zenith timelineid> $url* -- ask pageserver to start walreceiver on $url
 //
-use anyhow::{anyhow, bail, ensure, Context, Result};
+use anyhow::{bail, ensure, Context, Result};
 use bytes::{Buf, BufMut, Bytes, BytesMut};
 use lazy_static::lazy_static;
 use regex::Regex;
 use std::io;
 use std::net::TcpListener;
 use std::str;
 use std::str::FromStr;
-use std::sync::Arc;
+use std::sync::{Arc, RwLockReadGuard};
 use std::thread;
 use std::{io, net::TcpStream};
 use tracing::*;
 use zenith_metrics::{register_histogram_vec, HistogramVec};
 use zenith_utils::auth::{self, JwtAuth};
@@ -39,6 +38,8 @@ use crate::config::PageServerConf;
 use crate::relish::*;
 use crate::repository::Timeline;
 use crate::tenant_mgr;
 use crate::thread_mgr;
 use crate::thread_mgr::ThreadKind;
 use crate::walreceiver;
 use crate::CheckpointConfig;
@@ -189,30 +190,61 @@ pub fn thread_main(
    listener: TcpListener,
    auth_type: AuthType,
 ) -> anyhow::Result<()> {
-    let mut join_handles = Vec::new();
+    listener.set_nonblocking(true)?;
    let basic_rt = tokio::runtime::Builder::new_current_thread()
        .enable_io()
        .build()?;
-    while !tenant_mgr::shutdown_requested() {
+    let tokio_listener = {
-        let (socket, peer_addr) = listener.accept()?;
+        let _guard = basic_rt.enter();
-        debug!("accepted connection from {}", peer_addr);
+        tokio::net::TcpListener::from_std(listener)
-        socket.set_nodelay(true).unwrap();
+    }?;
        let local_auth = auth.clone();
-        let handle = thread::Builder::new()
+    // Wait for a new connection to arrive, or for server shutdown.
-            .name("serving Page Service thread".into())
+    while let Some(res) = basic_rt.block_on(async {
-            .spawn(move || {
+        let shutdown_watcher = thread_mgr::shutdown_watcher();
-                if let Err(err) = page_service_conn_main(conf, local_auth, socket, auth_type) {
+        tokio::select! {
-                    error!(%err, "page server thread exited with error");
+            biased;
            _ = shutdown_watcher => {
                // We were requested to shut down.
                None
            }
            res = tokio_listener.accept() => {
                Some(res)
            }
        }
    }) {
        match res {
            Ok((socket, peer_addr)) => {
                // Connection established. Spawn a new thread to handle it.
                debug!("accepted connection from {}", peer_addr);
                let local_auth = auth.clone();
                // PageRequestHandler threads are not associated with any particular
                // timeline in the thread manager. In practice most connections will
                // only deal with a particular timeline, but we don't know which one
                // yet.
                if let Err(err) = thread_mgr::spawn(
                    ThreadKind::PageRequestHandler,
                    None,
                    None,
                    "serving Page Service thread",
                    move || page_service_conn_main(conf, local_auth, socket, auth_type),
                ) {
                    // Thread creation failed. Log the error and continue.
                    error!("could not spawn page service thread: {:?}", err);
                }
-            })
+            }
-            .unwrap();
+            Err(err) => {
-
+                // accept() failed. Log the error, and loop back to retry on next connection.
-        join_handles.push(handle);
+                error!("accept() failed: {:?}", err);
            }
        }
    }
-    debug!("page_service loop terminated. wait for connections to cancel");
+    debug!("page_service loop terminated");
    for handle in join_handles.into_iter() {
        handle.join().unwrap();
    }
    Ok(())
 }
@@ -220,10 +252,10 @@ pub fn thread_main(
 fn page_service_conn_main(
    conf: &'static PageServerConf,
    auth: Option<Arc<JwtAuth>>,
-    socket: TcpStream,
+    socket: tokio::net::TcpStream,
    auth_type: AuthType,
 ) -> anyhow::Result<()> {
-    // Immediatsely increment the gauge, then create a job to decrement it on thread exit.
+    // Immediately increment the gauge, then create a job to decrement it on thread exit.
    // One of the pros of `defer!` is that this will *most probably*
    // get called, even in presence of panics.
    let gauge = crate::LIVE_CONNECTIONS_COUNT.with_label_values(&["page_service"]);
@@ -232,6 +264,19 @@ fn page_service_conn_main(
        gauge.dec();
    }
    // We use Tokio to accept the connection, but the rest of the code works with a
    // regular socket. Convert.
    let socket = socket
        .into_std()
        .context("could not convert tokio::net:TcpStream to std::net::TcpStream")?;
    socket
        .set_nonblocking(false)
        .context("could not put socket to blocking mode")?;
    socket
        .set_nodelay(true)
        .context("could not set TCP_NODELAY")?;
    let mut conn_handler = PageServerHandler::new(conf, auth);
    let pgbackend = PostgresBackend::new(socket, auth_type, None, true)?;
    pgbackend.run(&mut conn_handler)
@@ -286,7 +331,7 @@ impl PageServerHandler {
        /* switch client to COPYBOTH */
        pgb.write_message(&BeMessage::CopyBothResponse)?;
-        while !tenant_mgr::shutdown_requested() {
+        while !thread_mgr::is_shutdown_requested() {
            match pgb.read_message() {
                Ok(message) => {
                    if let Some(message) = message {
@@ -320,7 +365,7 @@ impl PageServerHandler {
                        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);
+                            error!("error reading relation or page version: {:?}", e);
                            PagestreamBeMessage::Error(PagestreamErrorResponse {
                                message: e.to_string(),
                            })
@@ -353,7 +398,12 @@ impl PageServerHandler {
    /// In either case, if the page server hasn't received the WAL up to the
    /// requested LSN yet, we will wait for it to arrive. The return value is
    /// the LSN that should be used to look up the page versions.
-    fn wait_or_get_last_lsn(timeline: &dyn Timeline, lsn: Lsn, latest: bool) -> Result<Lsn> {
+    fn wait_or_get_last_lsn(
        timeline: &dyn Timeline,
        mut lsn: Lsn,
        latest: bool,
        latest_gc_cutoff_lsn: &RwLockReadGuard<Lsn>,
    ) -> Result<Lsn> {
        if latest {
            // Latest page version was requested. If LSN is given, it is a hint
            // to the page server that there have been no modifications to the
@@ -374,22 +424,26 @@ impl PageServerHandler {
            // walsender completes the authentication and starts streaming the
            // WAL.
            if lsn <= last_record_lsn {
-                Ok(last_record_lsn)
+                lsn = last_record_lsn;
            } else {
                timeline.wait_lsn(lsn)?;
                // Since we waited for 'lsn' to arrive, that is now the last
                // record LSN. (Or close enough for our purposes; the
                // last-record LSN can advance immediately after we return
                // anyway)
                Ok(lsn)
            }
        } else {
            if lsn == Lsn(0) {
                bail!("invalid LSN(0) in request");
            }
            timeline.wait_lsn(lsn)?;
            Ok(lsn)
        }
        ensure!(
            lsn >= **latest_gc_cutoff_lsn,
            "tried to request a page version that was garbage collected. requested at {} gc cutoff {}",
            lsn, **latest_gc_cutoff_lsn
        );
        Ok(lsn)
    }
    fn handle_get_rel_exists_request(
@@ -400,7 +454,8 @@ impl PageServerHandler {
        let _enter = info_span!("get_rel_exists", rel = %req.rel, req_lsn = %req.lsn).entered();
        let tag = RelishTag::Relation(req.rel);
-        let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest)?;
+        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
        let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest, &latest_gc_cutoff_lsn)?;
        let exists = timeline.get_rel_exists(tag, lsn)?;
@@ -416,7 +471,8 @@ impl PageServerHandler {
    ) -> Result<PagestreamBeMessage> {
        let _enter = info_span!("get_nblocks", rel = %req.rel, req_lsn = %req.lsn).entered();
        let tag = RelishTag::Relation(req.rel);
-        let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest)?;
+        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
        let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest, &latest_gc_cutoff_lsn)?;
        let n_blocks = timeline.get_relish_size(tag, lsn)?;
@@ -437,8 +493,16 @@ impl PageServerHandler {
        let _enter = info_span!("get_page", rel = %req.rel, blkno = &req.blkno, req_lsn = %req.lsn)
            .entered();
        let tag = RelishTag::Relation(req.rel);
-        let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest)?;
+        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
-
+        let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest, &latest_gc_cutoff_lsn)?;
        /*
        // Add a 1s delay to some requests. The delayed causes the requests to
        // hit the race condition from github issue #1047 more easily.
        use rand::Rng;
        if rand::thread_rng().gen::<u8>() < 5 {
            std::thread::sleep(std::time::Duration::from_millis(1000));
        }
        */
        let page = timeline.get_page_at_lsn(tag, req.blkno, lsn)?;
        Ok(PagestreamBeMessage::GetPage(PagestreamGetPageResponse {
@@ -459,9 +523,10 @@ impl PageServerHandler {
        // check that the timeline exists
        let timeline = tenant_mgr::get_timeline_for_tenant(tenantid, timelineid)
            .context("Cannot handle basebackup request for a remote timeline")?;
        let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
        if let Some(lsn) = lsn {
            timeline
-                .check_lsn_is_in_scope(lsn)
+                .check_lsn_is_in_scope(lsn, &latest_gc_cutoff_lsn)
                .context("invalid basebackup lsn")?;
        }
@@ -579,7 +644,7 @@ impl postgres_backend::Handler for PageServerHandler {
            let re = Regex::new(r"^callmemaybe ([[:xdigit:]]+) ([[:xdigit:]]+) (.*)$").unwrap();
            let caps = re
                .captures(query_string)
-                .ok_or_else(|| anyhow!("invalid callmemaybe: '{}'", query_string))?;
+                .with_context(|| format!("invalid callmemaybe: '{}'", query_string))?;
            let tenantid = ZTenantId::from_str(caps.get(1).unwrap().as_str())?;
            let timelineid = ZTimelineId::from_str(caps.get(2).unwrap().as_str())?;
@@ -594,22 +659,22 @@ impl postgres_backend::Handler for PageServerHandler {
            tenant_mgr::get_timeline_for_tenant(tenantid, timelineid)
                .context("Failed to fetch local timeline for callmemaybe requests")?;
-            walreceiver::launch_wal_receiver(self.conf, tenantid, timelineid, &connstr);
+            walreceiver::launch_wal_receiver(self.conf, tenantid, timelineid, &connstr)?;
            pgb.write_message_noflush(&BeMessage::CommandComplete(b"SELECT 1"))?;
        } else if query_string.starts_with("branch_create ") {
-            let err = || anyhow!("invalid branch_create: '{}'", query_string);
+            let err = || format!("invalid branch_create: '{}'", query_string);
            // branch_create <tenantid> <branchname> <startpoint>
            // TODO lazy static
            // TODO: escaping, to allow branch names with spaces
            let re = Regex::new(r"^branch_create ([[:xdigit:]]+) (\S+) ([^\r\n\s;]+)[\r\n\s;]*;?$")
                .unwrap();
-            let caps = re.captures(query_string).ok_or_else(err)?;
+            let caps = re.captures(query_string).with_context(err)?;
            let tenantid = ZTenantId::from_str(caps.get(1).unwrap().as_str())?;
-            let branchname = caps.get(2).ok_or_else(err)?.as_str().to_owned();
+            let branchname = caps.get(2).with_context(err)?.as_str().to_owned();
-            let startpoint_str = caps.get(3).ok_or_else(err)?.as_str().to_owned();
+            let startpoint_str = caps.get(3).with_context(err)?.as_str().to_owned();
            self.check_permission(Some(tenantid))?;
@@ -628,7 +693,7 @@ impl postgres_backend::Handler for PageServerHandler {
            let re = Regex::new(r"^branch_list ([[:xdigit:]]+)$").unwrap();
            let caps = re
                .captures(query_string)
-                .ok_or_else(|| anyhow!("invalid branch_list: '{}'", query_string))?;
+                .with_context(|| format!("invalid branch_list: '{}'", query_string))?;
            let tenantid = ZTenantId::from_str(caps.get(1).unwrap().as_str())?;
@@ -648,11 +713,11 @@ impl postgres_backend::Handler for PageServerHandler {
                .write_message_noflush(&BeMessage::DataRow(&[Some(&tenants_buf)]))?
                .write_message_noflush(&BeMessage::CommandComplete(b"SELECT 1"))?;
        } else if query_string.starts_with("tenant_create") {
-            let err = || anyhow!("invalid tenant_create: '{}'", query_string);
+            let err = || format!("invalid tenant_create: '{}'", query_string);
            // tenant_create <tenantid>
            let re = Regex::new(r"^tenant_create ([[:xdigit:]]+)$").unwrap();
-            let caps = re.captures(query_string).ok_or_else(err)?;
+            let caps = re.captures(query_string).with_context(err)?;
            self.check_permission(None)?;
@@ -683,7 +748,7 @@ impl postgres_backend::Handler for PageServerHandler {
            let caps = re
                .captures(query_string)
-                .ok_or_else(|| anyhow!("invalid do_gc: '{}'", query_string))?;
+                .with_context(|| format!("invalid do_gc: '{}'", query_string))?;
            let tenantid = ZTenantId::from_str(caps.get(1).unwrap().as_str())?;
            let timelineid = ZTimelineId::from_str(caps.get(2).unwrap().as_str())?;
@@ -767,7 +832,7 @@ impl postgres_backend::Handler for PageServerHandler {
            let caps = re
                .captures(query_string)
-                .ok_or_else(|| anyhow!("invalid checkpoint command: '{}'", query_string))?;
+                .with_context(|| format!("invalid checkpoint command: '{}'", query_string))?;
            let tenantid = ZTenantId::from_str(caps.get(1).unwrap().as_str())?;
            let timelineid = ZTimelineId::from_str(caps.get(2).unwrap().as_str())?;
@@ -89,11 +89,10 @@ use std::{
    collections::HashMap,
    ffi, fs,
    path::{Path, PathBuf},
    thread,
 };
 use anyhow::{bail, Context};
-use tokio::{io, sync};
+use tokio::io;
 use tracing::{error, info};
 use zenith_utils::zid::{ZTenantId, ZTimelineId};
@@ -125,8 +124,6 @@ pub struct SyncStartupData {
    /// To reuse the local file scan logic, the timeline states are returned even if no sync loop get started during init:
    /// in this case, no remote files exist and all local timelines with correct metadata files are considered ready.
    pub initial_timeline_states: HashMap<ZTenantId, HashMap<ZTimelineId, TimelineSyncState>>,
    /// A handle to the sync loop, if it was started from the configuration provided.
    pub sync_loop_handle: Option<thread::JoinHandle<anyhow::Result<()>>>,
 }
 /// Based on the config, initiates the remote storage connection and starts a separate thread
@@ -135,29 +132,33 @@ pub struct SyncStartupData {
 /// Along with that, scans tenant files local and remote (if the sync gets enabled) to check the initial timeline states.
 pub fn start_local_timeline_sync(
    config: &'static PageServerConf,
    shutdown_hook: sync::watch::Receiver<()>,
 ) -> anyhow::Result<SyncStartupData> {
    let local_timeline_files = local_tenant_timeline_files(config)
        .context("Failed to collect local tenant timeline files")?;
    match &config.remote_storage_config {
        Some(storage_config) => match &storage_config.storage {
-            RemoteStorageKind::LocalFs(root) => storage_sync::spawn_storage_sync_thread(
+            RemoteStorageKind::LocalFs(root) => {
-                shutdown_hook,
+                info!("Using fs root '{}' as a remote storage", root.display());
-                config,
+                storage_sync::spawn_storage_sync_thread(
-                local_timeline_files,
+                    config,
-                LocalFs::new(root.clone(), &config.workdir)?,
+                    local_timeline_files,
-                storage_config.max_concurrent_sync,
+                    LocalFs::new(root.clone(), &config.workdir)?,
-                storage_config.max_sync_errors,
+                    storage_config.max_concurrent_sync,
-            ),
+                    storage_config.max_sync_errors,
-            RemoteStorageKind::AwsS3(s3_config) => storage_sync::spawn_storage_sync_thread(
+                )
-                shutdown_hook,
+            },
-                config,
+            RemoteStorageKind::AwsS3(s3_config) => {
-                local_timeline_files,
+                info!("Using s3 bucket '{}' in region '{}' as a remote storage, prefix in bucket: '{:?}', bucket endpoint: '{:?}'",
-                S3::new(s3_config, &config.workdir)?,
+                    s3_config.bucket_name, s3_config.bucket_region, s3_config.prefix_in_bucket, s3_config.endpoint);
-                storage_config.max_concurrent_sync,
+                storage_sync::spawn_storage_sync_thread(
-                storage_config.max_sync_errors,
+                    config,
-            ),
+                    local_timeline_files,
                    S3::new(s3_config, &config.workdir)?,
                    storage_config.max_concurrent_sync,
                    storage_config.max_sync_errors,
                )
            },
        }
        .context("Failed to spawn the storage sync thread"),
        None => {
@@ -179,7 +180,6 @@ pub fn start_local_timeline_sync(
            }
            Ok(SyncStartupData {
                initial_timeline_states,
                sync_loop_handle: None,
            })
        }
    }
@@ -208,7 +208,7 @@ fn local_tenant_timeline_files(
                }
            }
            Err(e) => error!(
-                "Failed to list tenants dir entry {:?} in directory {}, reason: {:#}",
+                "Failed to list tenants dir entry {:?} in directory {}, reason: {:?}",
                tenants_dir_entry,
                tenants_dir.display(),
                e
@@ -249,14 +249,14 @@ fn collect_timelines_for_tenant(
                        );
                    }
                    Err(e) => error!(
-                        "Failed to process timeline dir contents at '{}', reason: {:#}",
+                        "Failed to process timeline dir contents at '{}', reason: {:?}",
                        timeline_path.display(),
                        e
                    ),
                }
            }
            Err(e) => error!(
-                "Failed to list timelines for entry tenant {}, reason: {:#}",
+                "Failed to list timelines for entry tenant {}, reason: {:?}",
                tenant_id, e
            ),
        }
@@ -70,8 +70,3 @@ on the timeline download, missing remote branch files are downlaoded.
 A branch is a per-tenant entity, yet a current implementaion requires synchronizing a timeline first to get the branch files locally.
 Currently, there's no other way to know about the remote branch files, neither the file contents is verified and updated.
 * no IT tests
 Automated S3 testing is lacking currently, due to no convenient way to enable backups during the tests.
 After it's fixed, benchmark runs should also be carried out to find bottlenecks.
@@ -73,7 +73,7 @@ impl RemoteStorage for LocalFs {
    }
    async fn list(&self) -> anyhow::Result<Vec<Self::StoragePath>> {
-        Ok(get_all_files(&self.root).await?.into_iter().collect())
+        get_all_files(&self.root).await
    }
    async fn upload(
@@ -9,6 +9,7 @@ use std::path::{Path, PathBuf};
 use anyhow::Context;
 use s3::{bucket::Bucket, creds::Credentials, region::Region};
 use tokio::io::{self, AsyncWriteExt};
 use tracing::debug;
 use crate::{
    config::S3Config,
@@ -58,10 +59,21 @@ pub struct S3 {
 impl S3 {
    /// Creates the storage, errors if incorrect AWS S3 configuration provided.
    pub fn new(aws_config: &S3Config, pageserver_workdir: &'static Path) -> anyhow::Result<Self> {
-        let region = aws_config
+        debug!(
-            .bucket_region
+            "Creating s3 remote storage around bucket {}",
-            .parse::<Region>()
+            aws_config.bucket_name
-            .context("Failed to parse the s3 region from config")?;
+        );
        let region = match aws_config.endpoint.clone() {
            Some(endpoint) => Region::Custom {
                endpoint,
                region: aws_config.bucket_region.clone(),
            },
            None => aws_config
                .bucket_region
                .parse::<Region>()
                .context("Failed to parse the s3 region from config")?,
        };
        let credentials = Credentials::new(
            aws_config.access_key_id.as_deref(),
            aws_config.secret_access_key.as_deref(),
@@ -80,7 +80,6 @@ use std::{
    num::{NonZeroU32, NonZeroUsize},
    path::{Path, PathBuf},
    sync::Arc,
    thread,
 };
 use anyhow::{bail, Context};
@@ -91,7 +90,6 @@ use tokio::{
    runtime::Runtime,
    sync::{
        mpsc::{self, UnboundedReceiver},
        watch::Receiver,
        RwLock,
    },
    time::{Duration, Instant},
@@ -111,7 +109,7 @@ use super::{RemoteStorage, SyncStartupData, TimelineSyncId};
 use crate::{
    config::PageServerConf, layered_repository::metadata::TimelineMetadata,
    remote_storage::storage_sync::compression::read_archive_header, repository::TimelineSyncState,
-    tenant_mgr::set_timeline_states,
+    tenant_mgr::set_timeline_states, thread_mgr, thread_mgr::ThreadKind,
 };
 use zenith_metrics::{register_histogram_vec, register_int_gauge, HistogramVec, IntGauge};
@@ -335,6 +333,10 @@ pub fn schedule_timeline_checkpoint_upload(
 ///
 /// Ensure that the loop is started otherwise the task is never processed.
 pub fn schedule_timeline_download(tenant_id: ZTenantId, timeline_id: ZTimelineId) {
    debug!(
        "Scheduling timeline download for tenant {}, timeline {}",
        tenant_id, timeline_id
    );
    sync_queue::push(SyncTask::new(
        TimelineSyncId(tenant_id, timeline_id),
        0,
@@ -351,7 +353,6 @@ pub(super) fn spawn_storage_sync_thread<
    P: std::fmt::Debug + Send + Sync + 'static,
    S: RemoteStorage<StoragePath = P> + Send + Sync + 'static,
 >(
    shutdown_hook: Receiver<()>,
    conf: &'static PageServerConf,
    local_timeline_files: HashMap<TimelineSyncId, (TimelineMetadata, Vec<PathBuf>)>,
    storage: S,
@@ -375,7 +376,7 @@ pub(super) fn spawn_storage_sync_thread<
            Ok(local_path) => Some(local_path),
            Err(e) => {
                error!(
-                    "Failed to find local path for remote path {:?}: {:#}",
+                    "Failed to find local path for remote path {:?}: {:?}",
                    remote_path, e
                );
                None
@@ -385,12 +386,14 @@ pub(super) fn spawn_storage_sync_thread<
    let initial_timeline_states = schedule_first_sync_tasks(&remote_index, local_timeline_files);
-    let handle = thread::Builder::new()
+    thread_mgr::spawn(
-        .name("Remote storage sync thread".to_string())
+        ThreadKind::StorageSync,
-        .spawn(move || {
+        None,
        None,
        "Remote storage sync thread",
        move || {
            storage_sync_loop(
                runtime,
                shutdown_hook,
                conf,
                receiver,
                remote_index,
@@ -398,11 +401,11 @@ pub(super) fn spawn_storage_sync_thread<
                max_concurrent_sync,
                max_sync_errors,
            )
-        })
+        },
-        .context("Failed to spawn remote storage sync thread")?;
+    )
    .context("Failed to spawn remote storage sync thread")?;
    Ok(SyncStartupData {
        initial_timeline_states,
        sync_loop_handle: Some(handle),
    })
 }
@@ -417,7 +420,6 @@ fn storage_sync_loop<
    S: RemoteStorage<StoragePath = P> + Send + Sync + 'static,
 >(
    runtime: Runtime,
    mut shutdown_hook: Receiver<()>,
    conf: &'static PageServerConf,
    mut receiver: UnboundedReceiver<SyncTask>,
    index: RemoteTimelineIndex,
@@ -437,7 +439,7 @@ fn storage_sync_loop<
                    max_sync_errors,
                )
                .instrument(debug_span!("storage_sync_loop_step")) => LoopStep::NewStates(new_timeline_states),
-                _ = shutdown_hook.changed() => LoopStep::Shutdown,
+                _ = thread_mgr::shutdown_watcher() => LoopStep::Shutdown,
            }
        });
@@ -507,7 +509,7 @@ async fn loop_step<
                Ok(extra_step) => extra_step,
                Err(e) => {
                    error!(
-                        "Failed to process storage sync task for tenant {}, timeline {}: {:#}",
+                        "Failed to process storage sync task for tenant {}, timeline {}: {:?}",
                        sync_id.0, sync_id.1, e
                    );
                    None
@@ -34,7 +34,7 @@ use std::{
    sync::Arc,
 };
-use anyhow::{anyhow, bail, ensure, Context};
+use anyhow::{bail, ensure, Context};
 use async_compression::tokio::bufread::{ZstdDecoder, ZstdEncoder};
 use serde::{Deserialize, Serialize};
 use tokio::{
@@ -211,16 +211,18 @@ pub async fn read_archive_header<A: io::AsyncRead + Send + Sync + Unpin>(
 pub fn parse_archive_name(archive_path: &Path) -> anyhow::Result<(Lsn, u64)> {
    let archive_name = archive_path
        .file_name()
-        .ok_or_else(|| anyhow!("Archive '{}' has no file name", archive_path.display()))?
+        .with_context(|| format!("Archive '{}' has no file name", archive_path.display()))?
        .to_string_lossy();
    let (lsn_str, header_size_str) =
-        archive_name.rsplit_once(ARCHIVE_EXTENSION).ok_or_else(|| {
+        archive_name
-            anyhow!(
+            .rsplit_once(ARCHIVE_EXTENSION)
-                "Archive '{}' has incorrect extension, expected to contain '{}'",
+            .with_context(|| {
-                archive_path.display(),
+                format!(
-                ARCHIVE_EXTENSION
+                    "Archive '{}' has incorrect extension, expected to contain '{}'",
-            )
+                    archive_path.display(),
-        })?;
+                    ARCHIVE_EXTENSION
                )
            })?;
    let disk_consistent_lsn = Lsn::from_hex(lsn_str).with_context(|| {
        format!(
            "Archive '{}' has an invalid disk consistent lsn in its extension",
@@ -374,7 +376,7 @@ async fn write_archive_contents(
    }
    let metadata_bytes_written = io::copy(&mut metadata_bytes.as_slice(), &mut archive_input)
        .await
-        .with_context(|| "Failed to add metadata into the archive")?;
+        .context("Failed to add metadata into the archive")?;
    ensure!(
        header.metadata_file_size == metadata_bytes_written,
        "Metadata file was written to the archive incompletely",
@@ -3,7 +3,7 @@
 use std::{borrow::Cow, collections::BTreeSet, path::PathBuf, sync::Arc};
-use anyhow::{anyhow, ensure, Context};
+use anyhow::{ensure, Context};
 use futures::{stream::FuturesUnordered, StreamExt};
 use tokio::{fs, sync::RwLock};
 use tracing::{debug, error, trace, warn};
@@ -80,7 +80,7 @@ pub(super) async fn download_timeline<
                {
                    Ok(remote_timeline) => Cow::Owned(remote_timeline),
                    Err(e) => {
-                        error!("Failed to download full timeline index: {:#}", e);
+                        error!("Failed to download full timeline index: {:?}", e);
                        return match remote_disk_consistent_lsn {
                            Some(disk_consistent_lsn) => {
                                sync_queue::push(SyncTask::new(
@@ -112,7 +112,7 @@ pub(super) async fn download_timeline<
    if let Err(e) = download_missing_branches(conf, remote_assets.as_ref(), sync_id.0).await {
        error!(
-            "Failed to download missing branches for sync id {}: {:#}",
+            "Failed to download missing branches for sync id {}: {:?}",
            sync_id, e
        );
        sync_queue::push(SyncTask::new(
@@ -150,7 +150,7 @@ pub(super) async fn download_timeline<
            Err(e) => {
                let archives_left = archives_total - archives_downloaded;
                error!(
-                    "Failed to download archive {:?} (archives downloaded: {}; archives left: {}) for tenant {} timeline {}, requeueing the download: {:#}",
+                    "Failed to download archive {:?} (archives downloaded: {}; archives left: {}) for tenant {} timeline {}, requeueing the download: {:?}",
                    archive_id, archives_downloaded, archives_left, tenant_id, timeline_id, e
                );
                sync_queue::push(SyncTask::new(
@@ -189,7 +189,7 @@ async fn try_download_archive<
    debug!("Downloading archive {:?}", archive_id);
    let archive_to_download = remote_timeline
        .archive_data(archive_id)
-        .ok_or_else(|| anyhow!("Archive {:?} not found in remote storage", archive_id))?;
+        .with_context(|| format!("Archive {:?} not found in remote storage", archive_id))?;
    let (archive_header, header_size) = remote_timeline
        .restore_header(archive_id)
        .context("Failed to restore header when downloading an archive")?;
@@ -307,7 +307,7 @@ async fn download_missing_branches<
        while let Some(download_result) = remote_only_branches_downloads.next().await {
            if let Err(e) = download_result {
                branch_downloads_failed = true;
-                error!("Failed to download a branch file: {:#}", e);
+                error!("Failed to download a branch file: {:?}", e);
            }
        }
        ensure!(
@@ -9,7 +9,7 @@ use std::{
    path::{Path, PathBuf},
 };
-use anyhow::{anyhow, bail, ensure, Context};
+use anyhow::{bail, ensure, Context};
 use serde::{Deserialize, Serialize};
 use tracing::debug;
 use zenith_utils::{
@@ -214,7 +214,7 @@ impl RemoteTimeline {
        let archive = self
            .checkpoint_archives
            .get(&archive_id)
-            .ok_or_else(|| anyhow!("Archive {:?} not found", archive_id))?;
+            .with_context(|| format!("Archive {:?} not found", archive_id))?;
        let mut header_files = Vec::with_capacity(archive.files.len());
        for (expected_archive_position, archive_file) in archive.files.iter().enumerate() {
@@ -226,11 +226,10 @@ impl RemoteTimeline {
                archive_id,
            );
-            let timeline_file = self.timeline_files.get(archive_file).ok_or_else(|| {
+            let timeline_file = self.timeline_files.get(archive_file).with_context(|| {
-                anyhow!(
+                format!(
                    "File with id {:?} not found for archive {:?}",
-                    archive_file,
+                    archive_file, archive_id
                    archive_id
                )
            })?;
            header_files.push(timeline_file.clone());
@@ -299,7 +298,7 @@ fn try_parse_index_entry(
        })?
        .iter()
        .next()
-        .ok_or_else(|| anyhow!("Found no tenant id in path '{}'", path.display()))?
+        .with_context(|| format!("Found no tenant id in path '{}'", path.display()))?
        .to_string_lossy()
        .parse::<ZTenantId>()
        .with_context(|| format!("Failed to parse tenant id from path '{}'", path.display()))?;
@@ -321,8 +320,8 @@ fn try_parse_index_entry(
            let mut segments = timelines_subpath.iter();
            let timeline_id = segments
                .next()
-                .ok_or_else(|| {
+                .with_context(|| {
-                    anyhow!(
+                    format!(
                        "{} directory of tenant {} (path '{}') is not an index entry",
                        TIMELINES_SEGMENT_NAME,
                        tenant_id,
@@ -345,7 +344,7 @@ fn try_parse_index_entry(
            let archive_name = path
                .file_name()
-                .ok_or_else(|| anyhow!("Archive '{}' has no file name", path.display()))?
+                .with_context(|| format!("Archive '{}' has no file name", path.display()))?
                .to_string_lossy()
                .to_string();
@@ -43,7 +43,7 @@ pub(super) async fn upload_timeline_checkpoint<
    debug!("Uploading checkpoint for sync id {}", sync_id);
    if let Err(e) = upload_missing_branches(config, remote_assets.as_ref(), sync_id.0).await {
        error!(
-            "Failed to upload missing branches for sync id {}: {:#}",
+            "Failed to upload missing branches for sync id {}: {:?}",
            sync_id, e
        );
        sync_queue::push(SyncTask::new(
@@ -69,7 +69,7 @@ pub(super) async fn upload_timeline_checkpoint<
            match update_index_description(remote_assets.as_ref(), &timeline_dir, sync_id).await {
                Ok(remote_timeline) => Some(Cow::Owned(remote_timeline)),
                Err(e) => {
-                    error!("Failed to download full timeline index: {:#}", e);
+                    error!("Failed to download full timeline index: {:?}", e);
                    sync_queue::push(SyncTask::new(
                        sync_id,
                        retries,
@@ -132,7 +132,7 @@ pub(super) async fn upload_timeline_checkpoint<
        }
        Err(e) => {
            error!(
-                "Failed to upload checkpoint: {:#}, requeueing the upload",
+                "Failed to upload checkpoint: {:?}, requeueing the upload",
                e
            );
            sync_queue::push(SyncTask::new(
@@ -253,7 +253,7 @@ async fn upload_missing_branches<
                .await
                .add_branch_file(tenant_id, local_only_branch.clone()),
            Err(e) => {
-                error!("Failed to upload branch file: {:#}", e);
+                error!("Failed to upload branch file: {:?}", e);
                branch_uploads_failed = true;
            }
        }
@@ -7,7 +7,7 @@ use postgres_ffi::{MultiXactId, MultiXactOffset, TransactionId};
 use serde::{Deserialize, Serialize};
 use std::collections::HashSet;
 use std::ops::{AddAssign, Deref};
-use std::sync::Arc;
+use std::sync::{Arc, RwLockReadGuard};
 use std::time::Duration;
 use zenith_utils::lsn::{Lsn, RecordLsn};
 use zenith_utils::zid::ZTimelineId;
@@ -19,7 +19,7 @@ pub type BlockNumber = u32;
 /// A repository corresponds to one .zenith directory. One repository holds multiple
 /// timelines, forked off from the same initial call to 'initdb'.
 pub trait Repository: Send + Sync {
-    fn shutdown(&self) -> Result<()>;
+    fn detach_timeline(&self, timeline_id: ZTimelineId) -> Result<()>;
    /// Updates timeline based on the new sync state, received from the remote storage synchronization.
    /// See [`crate::remote_storage`] for more details about the synchronization.
@@ -184,6 +184,9 @@ pub trait Timeline: Send + Sync {
    ///
    fn wait_lsn(&self, lsn: Lsn) -> Result<()>;
    /// Lock and get timeline's GC cuttof
    fn get_latest_gc_cutoff_lsn(&self) -> RwLockReadGuard<Lsn>;
    /// Look up given page version.
    fn get_page_at_lsn(&self, tag: RelishTag, blknum: BlockNumber, lsn: Lsn) -> Result<Bytes>;
@@ -217,10 +220,12 @@ pub trait Timeline: Send + Sync {
    /// Atomically get both last and prev.
    fn get_last_record_rlsn(&self) -> RecordLsn;
    /// Get last or prev record separately. Same as get_last_record_rlsn().last/prev.
    fn get_last_record_lsn(&self) -> Lsn;
    fn get_prev_record_lsn(&self) -> Lsn;
-    fn get_start_lsn(&self) -> Lsn;
+
    fn get_disk_consistent_lsn(&self) -> Lsn;
    /// Mutate the timeline with a [`TimelineWriter`].
@@ -235,7 +240,11 @@ pub trait Timeline: Send + Sync {
    ///
    /// Check that it is valid to request operations with that lsn.
-    fn check_lsn_is_in_scope(&self, lsn: Lsn) -> Result<()>;
+    fn check_lsn_is_in_scope(
        &self,
        lsn: Lsn,
        latest_gc_cutoff_lsn: &RwLockReadGuard<Lsn>,
    ) -> Result<()>;
    /// Retrieve current logical size of the timeline
    ///
@@ -244,7 +253,7 @@ pub trait Timeline: Send + Sync {
    fn get_current_logical_size(&self) -> usize;
    /// Does the same as get_current_logical_size but counted on demand.
-    /// Used in tests to ensure thet incremental and non incremental variants match.
+    /// Used in tests to ensure that incremental and non incremental variants match.
    fn get_current_logical_size_non_incremental(&self, lsn: Lsn) -> Result<usize>;
    /// An escape hatch to allow "casting" a generic Timeline to LayeredTimeline.
@@ -987,7 +996,7 @@ mod tests {
                    .source()
                    .unwrap()
                    .to_string()
-                    .contains("is earlier than initdb lsn"));
+                    .contains("is earlier than latest GC horizon"));
            }
        }
@@ -1004,12 +1013,11 @@ mod tests {
        make_some_layers(&tline, Lsn(0x20))?;
        repo.gc_iteration(Some(TIMELINE_ID), 0x10, false)?;
-
+        let latest_gc_cutoff_lsn = tline.get_latest_gc_cutoff_lsn();
        assert!(*latest_gc_cutoff_lsn > Lsn(0x25));
        match tline.get_page_at_lsn(TESTREL_A, 0, Lsn(0x25)) {
            Ok(_) => panic!("request for page should have failed"),
-            Err(err) => assert!(err
+            Err(err) => assert!(err.to_string().contains("not found at")),
                .to_string()
                .contains("tried to request a page version that was garbage collected")),
        }
        Ok(())
    }
@@ -5,15 +5,16 @@ use crate::branches;
 use crate::config::PageServerConf;
 use crate::layered_repository::LayeredRepository;
 use crate::repository::{Repository, Timeline, TimelineSyncState};
-use crate::tenant_threads;
+use crate::thread_mgr;
 use crate::thread_mgr::ThreadKind;
 use crate::walredo::PostgresRedoManager;
-use anyhow::{anyhow, bail, Context, Result};
+use crate::CheckpointConfig;
 use anyhow::{bail, Context, Result};
 use lazy_static::lazy_static;
 use log::*;
 use serde::{Deserialize, Serialize};
 use std::collections::{hash_map, HashMap};
 use std::fmt;
 use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::{Arc, Mutex, MutexGuard};
 use zenith_utils::zid::{ZTenantId, ZTimelineId};
@@ -23,7 +24,7 @@ lazy_static! {
 struct Tenant {
    state: TenantState,
-    repo: Option<Arc<dyn Repository>>,
+    repo: Arc<dyn Repository>,
 }
 #[derive(Debug, Serialize, Deserialize, Clone, Copy, PartialEq, Eq)]
@@ -56,8 +57,6 @@ fn access_tenants() -> MutexGuard<'static, HashMap<ZTenantId, Tenant>> {
    TENANTS.lock().unwrap()
 }
 static SHUTDOWN_REQUESTED: AtomicBool = AtomicBool::new(false);
 /// Updates tenants' repositories, changing their timelines state in memory.
 pub fn set_timeline_states(
    conf: &'static PageServerConf,
@@ -73,28 +72,8 @@ pub fn set_timeline_states(
    let mut m = access_tenants();
    for (tenant_id, timeline_states) in timeline_states {
-        let tenant = m.entry(tenant_id).or_insert_with(|| Tenant {
+        let tenant = m.entry(tenant_id).or_insert_with(|| {
-            state: TenantState::Idle,
+            // TODO (rodionov) reuse one of the initialisation routines
            repo: None,
        });
        if let Err(e) = put_timelines_into_tenant(conf, tenant, tenant_id, timeline_states) {
            error!(
                "Failed to update timeline states for tenant {}: {:#}",
                tenant_id, e
            );
        }
    }
 }
 fn put_timelines_into_tenant(
    conf: &'static PageServerConf,
    tenant: &mut Tenant,
    tenant_id: ZTenantId,
    timeline_states: HashMap<ZTimelineId, TimelineSyncState>,
 ) -> anyhow::Result<()> {
    let repo = match tenant.repo.as_ref() {
        Some(repo) => Arc::clone(repo),
        None => {
            // Set up a WAL redo manager, for applying WAL records.
            let walredo_mgr = PostgresRedoManager::new(conf, tenant_id);
@@ -105,13 +84,43 @@ fn put_timelines_into_tenant(
                tenant_id,
                conf.remote_storage_config.is_some(),
            ));
-            tenant.repo = Some(Arc::clone(&repo));
+            Tenant {
-            repo
+                state: TenantState::Idle,
                repo,
            }
        });
        if let Err(e) = put_timelines_into_tenant(tenant, tenant_id, timeline_states) {
            error!(
                "Failed to update timeline states for tenant {}: {:?}",
                tenant_id, e
            );
        }
-    };
+    }
 }
 fn put_timelines_into_tenant(
    tenant: &mut Tenant,
    tenant_id: ZTenantId,
    timeline_states: HashMap<ZTimelineId, TimelineSyncState>,
 ) -> anyhow::Result<()> {
    for (timeline_id, timeline_state) in timeline_states {
-        repo.set_timeline_state(timeline_id, timeline_state)
+        // If the timeline is being put into any other state than Ready,
        // stop any threads operating on it.
        //
        // FIXME: This is racy. A page service thread could just get
        // handle on the Timeline, before we call set_timeline_state()
        if !matches!(timeline_state, TimelineSyncState::Ready(_)) {
            thread_mgr::shutdown_threads(None, Some(tenant_id), Some(timeline_id));
            // Should we run a final checkpoint to flush all the data to
            // disk? Doesn't seem necessary; all of the states other than
            // Ready imply that the data on local disk is corrupt or incomplete,
            // and we don't want to flush that to disk.
        }
        tenant
            .repo
            .set_timeline_state(timeline_id, timeline_state)
            .with_context(|| {
                format!(
                    "Failed to update timeline {} state to {:?}",
@@ -123,29 +132,49 @@ fn put_timelines_into_tenant(
    Ok(())
 }
-// Check this flag in the thread loops to know when to exit
+///
-pub fn shutdown_requested() -> bool {
+/// Shut down all tenants. This runs as part of pageserver shutdown.
-    SHUTDOWN_REQUESTED.load(Ordering::Relaxed)
+///
-}
+pub fn shutdown_all_tenants() {
-
+    let mut m = access_tenants();
-pub fn shutdown_all_tenants() -> Result<()> {
+    let mut tenantids = Vec::new();
-    SHUTDOWN_REQUESTED.swap(true, Ordering::Relaxed);
+    for (tenantid, tenant) in m.iter_mut() {
-
+        tenant.state = TenantState::Stopping;
-    let tenantids = list_tenantids()?;
+        tenantids.push(*tenantid)
    for tenantid in &tenantids {
        set_tenant_state(*tenantid, TenantState::Stopping)?;
    }
    drop(m);
    thread_mgr::shutdown_threads(Some(ThreadKind::WalReceiver), None, None);
    thread_mgr::shutdown_threads(Some(ThreadKind::GarbageCollector), None, None);
    thread_mgr::shutdown_threads(Some(ThreadKind::Checkpointer), None, None);
    // Ok, no background threads running anymore. Flush any remaining data in
    // memory to disk.
    //
    // We assume that any incoming connections that might request pages from
    // the repository have already been terminated by the caller, so there
    // should be no more activity in any of the repositories.
    //
    // On error, log it but continue with the shutdown for other tenants.
    for tenantid in tenantids {
        // Wait for checkpointer and GC to finish their job
        tenant_threads::wait_for_tenant_threads_to_stop(tenantid);
        let repo = get_repository_for_tenant(tenantid)?;
        debug!("shutdown tenant {}", tenantid);
-        repo.shutdown()?;
+        match get_repository_for_tenant(tenantid) {
            Ok(repo) => {
                if let Err(err) = repo.checkpoint_iteration(CheckpointConfig::Flush) {
                    error!(
                        "Could not checkpoint tenant {} during shutdown: {:?}",
                        tenantid, err
                    );
                }
            }
            Err(err) => {
                error!(
                    "Could not get repository for tenant {} during shutdown: {:?}",
                    tenantid, err
                );
            }
        }
    }
    Ok(())
 }
 pub fn create_repository_for_tenant(
@@ -153,7 +182,7 @@ pub fn create_repository_for_tenant(
    tenantid: ZTenantId,
 ) -> Result<()> {
    let wal_redo_manager = Arc::new(PostgresRedoManager::new(conf, tenantid));
-    let repo = Some(branches::create_repo(conf, tenantid, wal_redo_manager)?);
+    let repo = branches::create_repo(conf, tenantid, wal_redo_manager)?;
    match access_tenants().entry(tenantid) {
        hash_map::Entry::Occupied(_) => bail!("tenant {} already exists", tenantid),
@@ -172,34 +201,60 @@ pub fn get_tenant_state(tenantid: ZTenantId) -> Option<TenantState> {
    Some(access_tenants().get(&tenantid)?.state)
 }
-pub fn set_tenant_state(tenantid: ZTenantId, newstate: TenantState) -> Result<TenantState> {
+///
 /// Change the state of a tenant to Active and launch its checkpointer and GC
 /// threads. If the tenant was already in Active state or Stopping, does nothing.
 ///
 pub fn activate_tenant(conf: &'static PageServerConf, tenantid: ZTenantId) -> Result<()> {
    let mut m = access_tenants();
-    let tenant = m.get_mut(&tenantid);
+    let tenant = m
        .get_mut(&tenantid)
        .with_context(|| format!("Tenant not found for id {}", tenantid))?;
-    match tenant {
+    info!("activating tenant {}", tenantid);
-        Some(tenant) => {
+
-            if newstate == TenantState::Idle && tenant.state != TenantState::Active {
+    match tenant.state {
-                // Only Active tenant can become Idle
+        // If the tenant is already active, nothing to do.
-                return Ok(tenant.state);
+        TenantState::Active => {}
-            }
+
-            info!("set_tenant_state: {} -> {}", tenant.state, newstate);
+        // If it's Idle, launch the checkpointer and GC threads
-            tenant.state = newstate;
+        TenantState::Idle => {
-            Ok(tenant.state)
+            thread_mgr::spawn(
                ThreadKind::Checkpointer,
                Some(tenantid),
                None,
                "Checkpointer thread",
                move || crate::tenant_threads::checkpoint_loop(tenantid, conf),
            )?;
            // FIXME: if we fail to launch the GC thread, but already launched the
            // checkpointer, we're in a strange state.
            thread_mgr::spawn(
                ThreadKind::GarbageCollector,
                Some(tenantid),
                None,
                "GC thread",
                move || crate::tenant_threads::gc_loop(tenantid, conf),
            )?;
            tenant.state = TenantState::Active;
        }
        TenantState::Stopping => {
            // don't re-activate it if it's being stopped
        }
        None => bail!("Tenant not found for id {}", tenantid),
    }
    Ok(())
 }
 pub fn get_repository_for_tenant(tenantid: ZTenantId) -> Result<Arc<dyn Repository>> {
    let m = access_tenants();
    let tenant = m
        .get(&tenantid)
-        .ok_or_else(|| anyhow!("Tenant not found for tenant {}", tenantid))?;
+        .with_context(|| format!("Tenant not found for tenant {}", tenantid))?;
-    match &tenant.repo {
+    Ok(Arc::clone(&tenant.repo))
        Some(repo) => Ok(Arc::clone(repo)),
        None => bail!("Repository for tenant {} is not yet valid", tenantid),
    }
 }
 pub fn get_timeline_for_tenant(
@@ -209,17 +264,7 @@ pub fn get_timeline_for_tenant(
    get_repository_for_tenant(tenantid)?
        .get_timeline(timelineid)?
        .local_timeline()
-        .ok_or_else(|| anyhow!("cannot fetch timeline {}", timelineid))
+        .with_context(|| format!("cannot fetch timeline {}", timelineid))
 }
 fn list_tenantids() -> Result<Vec<ZTenantId>> {
    access_tenants()
        .iter()
        .map(|v| {
            let (tenantid, _) = v;
            Ok(*tenantid)
        })
        .collect()
 }
 #[derive(Serialize, Deserialize, Clone)]
@@ -5,88 +5,14 @@ use crate::tenant_mgr;
 use crate::tenant_mgr::TenantState;
 use crate::CheckpointConfig;
 use anyhow::Result;
 use lazy_static::lazy_static;
 use std::collections::HashMap;
 use std::sync::Mutex;
 use std::thread::JoinHandle;
 use std::time::Duration;
 use tracing::*;
 use zenith_metrics::{register_int_gauge_vec, IntGaugeVec};
 use zenith_utils::zid::ZTenantId;
 struct TenantHandleEntry {
    checkpointer_handle: Option<JoinHandle<()>>,
    gc_handle: Option<JoinHandle<()>>,
 }
 // Preserve handles to wait for thread completion
 // at shutdown
 lazy_static! {
    static ref TENANT_HANDLES: Mutex<HashMap<ZTenantId, TenantHandleEntry>> =
        Mutex::new(HashMap::new());
 }
 lazy_static! {
    static ref TENANT_THREADS_COUNT: IntGaugeVec = register_int_gauge_vec!(
        "tenant_threads_count",
        "Number of live tenant threads",
        &["tenant_thread_type"]
    )
    .expect("failed to define a metric");
 }
 // Launch checkpointer and GC for the tenant.
 // It's possible that the threads are running already,
 // if so, just don't spawn new ones.
 pub fn start_tenant_threads(conf: &'static PageServerConf, tenantid: ZTenantId) {
    let mut handles = TENANT_HANDLES.lock().unwrap();
    let h = handles
        .entry(tenantid)
        .or_insert_with(|| TenantHandleEntry {
            checkpointer_handle: None,
            gc_handle: None,
        });
    if h.checkpointer_handle.is_none() {
        h.checkpointer_handle = std::thread::Builder::new()
            .name("Checkpointer thread".into())
            .spawn(move || {
                checkpoint_loop(tenantid, conf).expect("Checkpointer thread died");
            })
            .ok();
    }
    if h.gc_handle.is_none() {
        h.gc_handle = std::thread::Builder::new()
            .name("GC thread".into())
            .spawn(move || {
                gc_loop(tenantid, conf).expect("GC thread died");
            })
            .ok();
    }
 }
 pub fn wait_for_tenant_threads_to_stop(tenantid: ZTenantId) {
    let mut handles = TENANT_HANDLES.lock().unwrap();
    if let Some(h) = handles.get_mut(&tenantid) {
        h.checkpointer_handle.take().map(JoinHandle::join);
        trace!("checkpointer for tenant {} has stopped", tenantid);
        h.gc_handle.take().map(JoinHandle::join);
        trace!("gc for tenant {} has stopped", tenantid);
    }
    handles.remove(&tenantid);
 }
 ///
 /// Checkpointer thread's main loop
 ///
-fn checkpoint_loop(tenantid: ZTenantId, conf: &'static PageServerConf) -> Result<()> {
+pub fn checkpoint_loop(tenantid: ZTenantId, conf: &'static PageServerConf) -> Result<()> {
    let gauge = TENANT_THREADS_COUNT.with_label_values(&["checkpointer"]);
    gauge.inc();
    scopeguard::defer! {
        gauge.dec();
    }
    loop {
        if tenant_mgr::get_tenant_state(tenantid) != Some(TenantState::Active) {
            break;
@@ -112,13 +38,7 @@ fn checkpoint_loop(tenantid: ZTenantId, conf: &'static PageServerConf) -> Result
 ///
 /// GC thread's main loop
 ///
-fn gc_loop(tenantid: ZTenantId, conf: &'static PageServerConf) -> Result<()> {
+pub fn gc_loop(tenantid: ZTenantId, conf: &'static PageServerConf) -> Result<()> {
    let gauge = TENANT_THREADS_COUNT.with_label_values(&["gc"]);
    gauge.inc();
    scopeguard::defer! {
        gauge.dec();
    }
    loop {
        if tenant_mgr::get_tenant_state(tenantid) != Some(TenantState::Active) {
            break;
@@ -0,0 +1,284 @@
 //!
 //! This module provides centralized handling of threads in the Page Server.
 //!
 //! We provide a few basic facilities:
 //! - A global registry of threads that lists what kind of threads they are, and
 //!   which tenant or timeline they are working on
 //!
 //! - The ability to request a thread to shut down.
 //!
 //!
 //! # How it works?
 //!
 //! There is a global hashmap of all the threads (`THREADS`). Whenever a new
 //! thread is spawned, a PageServerThread entry is added there, and when a
 //! thread dies, it removes itself from the hashmap. If you want to kill a
 //! thread, you can scan the hashmap to find it.
 //!
 //! # Thread shutdown
 //!
 //! To kill a thread, we rely on co-operation from the victim. Each thread is
 //! expected to periodically call the `is_shutdown_requested()` function, and
 //! if it returns true, exit gracefully. In addition to that, when waiting for
 //! the network or other long-running operation, you can use
 //! `shutdown_watcher()` function to get a Future that will become ready if
 //! the current thread has been requested to shut down. You can use that with
 //! Tokio select!(), but note that it relies on thread-local storage, so it
 //! will only work with the "current-thread" Tokio runtime!
 //!
 //!
 //! TODO: This would be a good place to also handle panics in a somewhat sane way.
 //! Depending on what thread panics, we might want to kill the whole server, or
 //! only a single tenant or timeline.
 //!
 use std::cell::RefCell;
 use std::collections::HashMap;
 use std::panic;
 use std::panic::AssertUnwindSafe;
 use std::sync::atomic::{AtomicBool, AtomicU64, Ordering};
 use std::sync::{Arc, Mutex};
 use std::thread;
 use std::thread::JoinHandle;
 use tokio::sync::watch;
 use tracing::{info, warn};
 use lazy_static::lazy_static;
 use zenith_utils::zid::{ZTenantId, ZTimelineId};
 lazy_static! {
    /// Each thread that we track is associated with a "thread ID". It's just
    /// an increasing number that we assign, not related to any system thread
    /// id.
    static ref NEXT_THREAD_ID: AtomicU64 = AtomicU64::new(1);
    /// Global registry of threads
    static ref THREADS: Mutex<HashMap<u64, Arc<PageServerThread>>> = Mutex::new(HashMap::new());
 }
 // There is a Tokio watch channel for each thread, which can be used to signal the
 // thread that it needs to shut down. This thread local variable holds the receiving
 // end of the channel. The sender is kept in the global registry, so that anyone
 // can send the signal to request thread shutdown.
 thread_local!(static SHUTDOWN_RX: RefCell<Option<watch::Receiver<()>>> = RefCell::new(None));
 // Each thread holds reference to its own PageServerThread here.
 thread_local!(static CURRENT_THREAD: RefCell<Option<Arc<PageServerThread>>> = RefCell::new(None));
 ///
 /// There are many kinds of threads in the system. Some are associated with a particular
 /// tenant or timeline, while others are global.
 ///
 /// Note that we don't try to limit how may threads of a certain kind can be running
 /// at the same time.
 ///
 #[derive(Debug, PartialEq, Eq, Clone, Copy)]
 pub enum ThreadKind {
    // libpq listener thread. It just accepts connection and spawns a
    // PageRequestHandler thread for each connection.
    LibpqEndpointListener,
    // HTTP endpoint listener.
    HttpEndpointListener,
    // Thread that handles a single connection. A PageRequestHandler thread
    // starts detached from any particular tenant or timeline, but it can be
    // associated with one later, after receiving a command from the client.
    PageRequestHandler,
    // Thread that connects to a safekeeper to fetch WAL for one timeline.
    WalReceiver,
    // Thread that handles checkpointing of all timelines for a tenant.
    Checkpointer,
    // Thread that handles GC of a tenant
    GarbageCollector,
    // Thread for synchronizing pageserver relish data with the remote storage.
    // Shared by all tenants.
    StorageSync,
 }
 struct PageServerThread {
    _thread_id: u64,
    kind: ThreadKind,
    /// Tenant and timeline that this thread is associated with.
    tenant_id: Option<ZTenantId>,
    timeline_id: Option<ZTimelineId>,
    name: String,
    // To request thread shutdown, set the flag, and send a dummy message to the
    // channel to notify it.
    shutdown_requested: AtomicBool,
    shutdown_tx: watch::Sender<()>,
    /// Handle for waiting for the thread to exit. It can be None, if the
    /// the thread has already exited.
    join_handle: Mutex<Option<JoinHandle<()>>>,
 }
 /// Launch a new thread
 pub fn spawn<F, E>(
    kind: ThreadKind,
    tenant_id: Option<ZTenantId>,
    timeline_id: Option<ZTimelineId>,
    name: &str,
    f: F,
 ) -> std::io::Result<()>
 where
    F: FnOnce() -> Result<(), E> + Send + 'static,
 {
    let (shutdown_tx, shutdown_rx) = watch::channel(());
    let thread_id = NEXT_THREAD_ID.fetch_add(1, Ordering::Relaxed);
    let thread = PageServerThread {
        _thread_id: thread_id,
        kind,
        tenant_id,
        timeline_id,
        name: name.to_string(),
        shutdown_requested: AtomicBool::new(false),
        shutdown_tx,
        join_handle: Mutex::new(None),
    };
    let thread_rc = Arc::new(thread);
    let mut jh_guard = thread_rc.join_handle.lock().unwrap();
    THREADS
        .lock()
        .unwrap()
        .insert(thread_id, Arc::clone(&thread_rc));
    let thread_rc2 = Arc::clone(&thread_rc);
    let join_handle = match thread::Builder::new()
        .name(name.to_string())
        .spawn(move || thread_wrapper(thread_id, thread_rc2, shutdown_rx, f))
    {
        Ok(handle) => handle,
        Err(err) => {
            // Could not spawn the thread. Remove the entry
            THREADS.lock().unwrap().remove(&thread_id);
            return Err(err);
        }
    };
    *jh_guard = Some(join_handle);
    drop(jh_guard);
    // The thread is now running. Nothing more to do here
    Ok(())
 }
 /// This wrapper function runs in a newly-spawned thread. It initializes the
 /// thread-local variables and calls the payload function
 fn thread_wrapper<F, E>(
    thread_id: u64,
    thread: Arc<PageServerThread>,
    shutdown_rx: watch::Receiver<()>,
    f: F,
 ) where
    F: FnOnce() -> Result<(), E> + Send + 'static,
 {
    SHUTDOWN_RX.with(|rx| {
        *rx.borrow_mut() = Some(shutdown_rx);
    });
    CURRENT_THREAD.with(|ct| {
        *ct.borrow_mut() = Some(thread);
    });
    // We use AssertUnwindSafe here so that the payload function
    // doesn't need to be UnwindSafe. We don't do anything after the
    // unwinding that would expose us to unwind-unsafe behavior.
    let result = panic::catch_unwind(AssertUnwindSafe(f));
    // Remove our entry from the global hashmap.
    THREADS.lock().unwrap().remove(&thread_id);
    // If the thread payload panic'd, exit with the panic.
    if let Err(err) = result {
        panic::resume_unwind(err);
    }
 }
 /// Is there a thread running that matches the criteria
 /// Signal and wait for threads to shut down.
 ///
 ///
 /// The arguments are used to select the threads to kill. Any None arguments are
 /// ignored. For example, to shut down all WalReceiver threads:
 ///
 ///   shutdown_threads(Some(ThreadKind::WalReceiver), None, None)
 ///
 /// Or to shut down all threads for given timeline:
 ///
 ///   shutdown_threads(None, Some(timelineid), None)
 ///
 pub fn shutdown_threads(
    kind: Option<ThreadKind>,
    tenant_id: Option<ZTenantId>,
    timeline_id: Option<ZTimelineId>,
 ) {
    let mut victim_threads = Vec::new();
    let threads = THREADS.lock().unwrap();
    for thread in threads.values() {
        if (kind.is_none() || Some(thread.kind) == kind)
            && (tenant_id.is_none() || thread.tenant_id == tenant_id)
            && (timeline_id.is_none() || thread.timeline_id == timeline_id)
        {
            thread.shutdown_requested.store(true, Ordering::Relaxed);
            // FIXME: handle error?
            let _ = thread.shutdown_tx.send(());
            victim_threads.push(Arc::clone(thread));
        }
    }
    drop(threads);
    for thread in victim_threads {
        info!("waiting for {} to shut down", thread.name);
        if let Some(join_handle) = thread.join_handle.lock().unwrap().take() {
            let _ = join_handle.join();
        } else {
            // The thread had not even fully started yet. Or it was shut down
            // concurrently and alrady exited
        }
    }
 }
 /// A Future that can be used to check if the current thread has been requested to
 /// shut down.
 pub async fn shutdown_watcher() {
    let _ = SHUTDOWN_RX
        .with(|rx| {
            rx.borrow()
                .as_ref()
                .expect("shutdown_requested() called in an unexpected thread")
                .clone()
        })
        .changed()
        .await;
 }
 /// Has the current thread been requested to shut down?
 pub fn is_shutdown_requested() -> bool {
    CURRENT_THREAD.with(|ct| {
        if let Some(ct) = ct.borrow().as_ref() {
            ct.shutdown_requested.load(Ordering::Relaxed)
        } else {
            if !cfg!(test) {
                warn!("is_shutdown_requested() called in an unexpected thread");
            }
            false
        }
    })
 }
@@ -7,10 +7,11 @@
 use crate::config::PageServerConf;
 use crate::tenant_mgr;
-use crate::tenant_mgr::TenantState;
+use crate::thread_mgr;
-use crate::tenant_threads;
+use crate::thread_mgr::ThreadKind;
 use crate::walingest::WalIngest;
 use anyhow::{bail, Context, Error, Result};
 use bytes::BytesMut;
 use lazy_static::lazy_static;
 use parking_lot::Mutex;
 use postgres_ffi::waldecoder::*;
@@ -19,28 +20,22 @@ use postgres_types::PgLsn;
 use std::cell::Cell;
 use std::collections::HashMap;
 use std::str::FromStr;
 use std::thread;
 use std::thread::JoinHandle;
 use std::thread_local;
 use std::time::SystemTime;
 use tokio::pin;
 use tokio::sync::oneshot;
 use tokio_postgres::replication::ReplicationStream;
 use tokio_postgres::{Client, NoTls, SimpleQueryMessage, SimpleQueryRow};
 use tokio_stream::StreamExt;
 use tracing::*;
 use zenith_utils::lsn::Lsn;
 use zenith_utils::pq_proto::ZenithFeedback;
 use zenith_utils::zid::ZTenantId;
 use zenith_utils::zid::ZTimelineId;
 //
 // We keep one WAL Receiver active per timeline.
 //
 struct WalReceiverEntry {
    wal_producer_connstr: String,
    wal_receiver_handle: Option<JoinHandle<()>>,
    wal_receiver_interrupt_sender: Option<oneshot::Sender<()>>,
    tenantid: ZTenantId,
 }
 lazy_static! {
@@ -55,50 +50,9 @@ thread_local! {
    pub(crate) static IS_WAL_RECEIVER: Cell<bool> = Cell::new(false);
 }
 // Wait for walreceiver to stop
 // Now it stops when pageserver shutdown is requested.
 // In future we can make this more granular and send shutdown signals
 // per tenant/timeline to cancel inactive walreceivers.
 // TODO deal with blocking pg connections
 pub fn stop_wal_receiver(tenantid: ZTenantId, timelineid: ZTimelineId) {
    let mut receivers = WAL_RECEIVERS.lock();
    if let Some(r) = receivers.get_mut(&(tenantid, timelineid)) {
        match r.wal_receiver_interrupt_sender.take() {
            Some(s) => {
                if s.send(()).is_err() {
                    warn!("wal receiver interrupt signal already sent");
                }
            }
            None => {
                warn!("wal_receiver_interrupt_sender is missing, wal recever shouldn't be running")
            }
        }
        info!("waiting for wal receiver to stop");
        let handle = r.wal_receiver_handle.take();
        // do not hold the lock while joining the handle (deadlock is possible otherwise)
        drop(receivers);
        // there is no timeout or try_join option available so in case of a bug this can hang forever
        handle.map(JoinHandle::join);
    }
 }
 fn drop_wal_receiver(tenantid: ZTenantId, timelineid: ZTimelineId) {
    let mut receivers = WAL_RECEIVERS.lock();
    receivers.remove(&(tenantid, timelineid));
    // Check if it was the last walreceiver of the tenant.
    // TODO now we store one WalReceiverEntry per timeline,
    // so this iterator looks a bit strange.
    for (_timelineid, entry) in receivers.iter() {
        if entry.tenantid == tenantid {
            return;
        }
    }
    // When last walreceiver of the tenant is gone, change state to Idle
    tenant_mgr::set_tenant_state(tenantid, TenantState::Idle).unwrap();
 }
 // Launch a new WAL receiver, or tell one that's running about change in connection string
@@ -107,37 +61,36 @@ pub fn launch_wal_receiver(
    tenantid: ZTenantId,
    timelineid: ZTimelineId,
    wal_producer_connstr: &str,
-) {
+) -> Result<()> {
    let mut receivers = WAL_RECEIVERS.lock();
    match receivers.get_mut(&(tenantid, timelineid)) {
        Some(receiver) => {
            info!("wal receiver already running, updating connection string");
            receiver.wal_producer_connstr = wal_producer_connstr.into();
        }
        None => {
-            let (tx, rx) = tokio::sync::oneshot::channel::<()>();
+            thread_mgr::spawn(
-
+                ThreadKind::WalReceiver,
-            let wal_receiver_handle = thread::Builder::new()
+                Some(tenantid),
-                .name("WAL receiver thread".into())
+                Some(timelineid),
-                .spawn(move || {
+                "WAL receiver thread",
                move || {
                    IS_WAL_RECEIVER.with(|c| c.set(true));
-                    thread_main(conf, tenantid, timelineid, rx);
+                    thread_main(conf, tenantid, timelineid)
-                })
+                },
-                .unwrap();
+            )?;
            let receiver = WalReceiverEntry {
                wal_producer_connstr: wal_producer_connstr.into(),
                wal_receiver_handle: Some(wal_receiver_handle),
                wal_receiver_interrupt_sender: Some(tx),
                tenantid,
            };
            receivers.insert((tenantid, timelineid), receiver);
            // Update tenant state and start tenant threads, if they are not running yet.
-            tenant_mgr::set_tenant_state(tenantid, TenantState::Active).unwrap();
+            tenant_mgr::activate_tenant(conf, tenantid)?;
            tenant_threads::start_tenant_threads(conf, tenantid);
        }
    };
    Ok(())
 }
 // Look up current WAL producer connection string in the hash table
@@ -158,8 +111,7 @@ fn thread_main(
    conf: &'static PageServerConf,
    tenantid: ZTenantId,
    timelineid: ZTimelineId,
-    interrupt_receiver: oneshot::Receiver<()>,
+) -> Result<()> {
 ) {
    let _enter = info_span!("WAL receiver", timeline = %timelineid, tenant = %tenantid).entered();
    info!("WAL receiver thread started");
@@ -168,13 +120,7 @@ fn thread_main(
    // Make a connection to the WAL safekeeper, or directly to the primary PostgreSQL server,
    // and start streaming WAL from it.
-    let res = walreceiver_main(
+    let res = walreceiver_main(conf, tenantid, timelineid, &wal_producer_connstr);
        conf,
        tenantid,
        timelineid,
        &wal_producer_connstr,
        interrupt_receiver,
    );
    // TODO cleanup info messages
    if let Err(e) = res {
@@ -189,6 +135,7 @@ fn thread_main(
    // Drop it from list of active WAL_RECEIVERS
    // so that next callmemaybe request launched a new thread
    drop_wal_receiver(tenantid, timelineid);
    Ok(())
 }
 fn walreceiver_main(
@@ -196,7 +143,6 @@ fn walreceiver_main(
    tenantid: ZTenantId,
    timelineid: ZTimelineId,
    wal_producer_connstr: &str,
    mut interrupt_receiver: oneshot::Receiver<()>,
 ) -> Result<(), Error> {
    // Connect to the database in replication mode.
    info!("connecting to {:?}", wal_producer_connstr);
@@ -273,12 +219,15 @@ fn walreceiver_main(
    let mut walingest = WalIngest::new(&*timeline, startpoint)?;
    while let Some(replication_message) = runtime.block_on(async {
        let shutdown_watcher = thread_mgr::shutdown_watcher();
        tokio::select! {
-            replication_message = physical_stream.next() => replication_message,
+            // check for shutdown first
-            _ = &mut interrupt_receiver => {
+            biased;
            _ = shutdown_watcher => {
                info!("walreceiver interrupted");
                None
            }
            replication_message = physical_stream.next() => replication_message,
        }
    }) {
        let replication_message = replication_message?;
@@ -339,7 +288,6 @@ fn walreceiver_main(
        };
        if let Some(last_lsn) = status_update {
            let last_lsn = PgLsn::from(u64::from(last_lsn));
            let timeline_synced_disk_consistent_lsn =
                tenant_mgr::get_repository_for_tenant(tenantid)?
                    .get_timeline_state(timelineid)
@@ -347,18 +295,32 @@ fn walreceiver_main(
                    .unwrap_or(Lsn(0));
            // The last LSN we processed. It is not guaranteed to survive pageserver crash.
-            let write_lsn = last_lsn;
+            let write_lsn = u64::from(last_lsn);
            // `disk_consistent_lsn` is the LSN at which page server guarantees local persistence of all received data
-            let flush_lsn = PgLsn::from(u64::from(timeline.get_disk_consistent_lsn()));
+            let flush_lsn = u64::from(timeline.get_disk_consistent_lsn());
            // The last LSN that is synced to remote storage and is guaranteed to survive pageserver crash
            // Used by safekeepers to remove WAL preceding `remote_consistent_lsn`.
-            let apply_lsn = PgLsn::from(u64::from(timeline_synced_disk_consistent_lsn));
+            let apply_lsn = u64::from(timeline_synced_disk_consistent_lsn);
            let ts = SystemTime::now();
-            const NO_REPLY: u8 = 0;
+
            // Send zenith feedback message.
            // Regular standby_status_update fields are put into this message.
            let zenith_status_update = ZenithFeedback {
                current_timeline_size: timeline.get_current_logical_size() as u64,
                ps_writelsn: write_lsn,
                ps_flushlsn: flush_lsn,
                ps_applylsn: apply_lsn,
                ps_replytime: ts,
            };
            debug!("zenith_status_update {:?}", zenith_status_update);
            let mut data = BytesMut::new();
            zenith_status_update.serialize(&mut data)?;
            runtime.block_on(
                physical_stream
                    .as_mut()
-                    .standby_status_update(write_lsn, flush_lsn, apply_lsn, ts, NO_REPLY),
+                    .zenith_status_update(data.len() as u64, &data),
            )?;
        }
    }
@@ -2,7 +2,7 @@
 name = "postgres_ffi"
 version = "0.1.0"
 authors = ["Heikki Linnakangas <heikki@zenith.tech>"]
-edition = "2018"
+edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
@@ -51,6 +51,13 @@ pub type TimeLineID = u32;
 pub type TimestampTz = i64;
 pub type XLogSegNo = u64;
 /// Interval of checkpointing metadata file. We should store metadata file to enforce
 /// predicate that checkpoint.nextXid is larger than any XID in WAL.
 /// But flushing checkpoint file for each transaction seems to be too expensive,
 /// so XID_CHECKPOINT_INTERVAL is used to forward align nextXid and so perform
 /// metadata checkpoint only once per XID_CHECKPOINT_INTERVAL transactions.
 /// XID_CHECKPOINT_INTERVAL should not be larger than BLCKSZ*CLOG_XACTS_PER_BYTE
 /// in order to let CLOG_TRUNCATE mechanism correctly extend CLOG.
 const XID_CHECKPOINT_INTERVAL: u32 = 1024;
 #[allow(non_snake_case)]
@@ -400,9 +407,13 @@ impl CheckPoint {
    ///
    /// Returns 'true' if the XID was updated.
    pub fn update_next_xid(&mut self, xid: u32) -> bool {
-        let xid = xid.wrapping_add(XID_CHECKPOINT_INTERVAL - 1) & !(XID_CHECKPOINT_INTERVAL - 1);
+        // nextXid should nw greate than any XID in WAL, so increment provided XID and check for wraparround.
        let mut new_xid = std::cmp::max(xid + 1, pg_constants::FIRST_NORMAL_TRANSACTION_ID);
        // To reduce number of metadata checkpoints, we forward align XID on XID_CHECKPOINT_INTERVAL.
        // XID_CHECKPOINT_INTERVAL should not be larger than BLCKSZ*CLOG_XACTS_PER_BYTE
        new_xid =
            new_xid.wrapping_add(XID_CHECKPOINT_INTERVAL - 1) & !(XID_CHECKPOINT_INTERVAL - 1);
        let full_xid = self.nextXid.value;
        let new_xid = std::cmp::max(xid + 1, pg_constants::FIRST_NORMAL_TRANSACTION_ID);
        let old_xid = full_xid as u32;
        if new_xid.wrapping_sub(old_xid) as i32 > 0 {
            let mut epoch = full_xid >> 32;
@@ -520,4 +531,34 @@ mod tests {
        println!("wal_end={}, tli={}", wal_end, tli);
        assert_eq!(wal_end, waldump_wal_end);
    }
    /// Check the math in update_next_xid
    ///
    /// NOTE: These checks are sensitive to the value of XID_CHECKPOINT_INTERVAL,
    /// currently 1024.
    #[test]
    pub fn test_update_next_xid() {
        let checkpoint_buf = [0u8; std::mem::size_of::<CheckPoint>()];
        let mut checkpoint = CheckPoint::decode(&checkpoint_buf).unwrap();
        checkpoint.nextXid = FullTransactionId { value: 10 };
        assert_eq!(checkpoint.nextXid.value, 10);
        // The input XID gets rounded up to the next XID_CHECKPOINT_INTERVAL
        // boundary
        checkpoint.update_next_xid(100);
        assert_eq!(checkpoint.nextXid.value, 1024);
        // No change
        checkpoint.update_next_xid(500);
        assert_eq!(checkpoint.nextXid.value, 1024);
        checkpoint.update_next_xid(1023);
        assert_eq!(checkpoint.nextXid.value, 1024);
        // The function returns the *next* XID, given the highest XID seen so
        // far. So when we pass 1024, the nextXid gets bumped up to the next
        // XID_CHECKPOINT_INTERVAL boundary.
        checkpoint.update_next_xid(1024);
        assert_eq!(checkpoint.nextXid.value, 2048);
    }
 }
@@ -38,7 +38,7 @@ def rustfmt(fix_inplace: bool = False, no_color: bool = False) -> str:
 def yapf(fix_inplace: bool) -> str:
-    cmd = "pipenv run yapf --recursive"
+    cmd = "poetry run yapf --recursive"
    if fix_inplace:
        cmd += " --in-place"
    else:
@@ -47,7 +47,7 @@ def yapf(fix_inplace: bool) -> str:
 def mypy() -> str:
-    return "pipenv run mypy"
+    return "poetry run mypy"
 def get_commit_files() -> List[str]:
@@ -72,7 +72,7 @@ def check(name: str, suffix: str, cmd: str, changed_files: List[str], no_color:
            print("Please inspect the output below and run make fmt to fix automatically.")
        if suffix == ".py":
            print("If the output is empty, ensure that you've installed Python tooling by\n"
-                  "running 'pipenv install --dev' in the current directory (no root needed)")
+                  "running './scripts/pysync' in the current directory (no root needed)")
        print()
        print(res.stdout.decode())
        exit(1)
@@ -2,7 +2,7 @@
 name = "proxy"
 version = "0.1.0"
 authors = ["Stas Kelvich <stas.kelvich@gmail.com>"]
-edition = "2018"
+edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
@@ -16,13 +16,24 @@ hex = "0.4.3"
 hyper = "0.14"
 routerify = "2"
 parking_lot = "0.11.2"
 hashbrown = "0.11.2"
 serde = "1"
 serde_json = "1"
 tokio = { version = "1.11", features = ["macros"] }
-tokio-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+tokio-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
 tokio-rustls = "0.22.0"
 clap = "2.33.0"
 rustls = "0.19.1"
 reqwest = { version = "0.11", default-features = false, features = ["blocking", "json", "rustls-tls"] }
 pin-project-lite = "0.2.7"
 futures = "0.3.13"
 scopeguard = "1.1.0"
 zenith_utils = { path = "../zenith_utils" }
 zenith_metrics = { path = "../zenith_metrics" }
 base64 = "0.13.0"
 async-trait = "0.1.52"
 [dev-dependencies]
 tokio-postgres-rustls = "0.8.0"
 rcgen = "0.8.14"
@@ -0,0 +1,41 @@
 use crate::db::AuthSecret;
 use crate::stream::PqStream;
 use bytes::Bytes;
 use tokio::io::{AsyncRead, AsyncWrite};
 use zenith_utils::pq_proto::BeMessage as Be;
 /// Stored secret for authenticating the user via md5 but authenticating
 /// to the compute database with a (possibly different) plaintext password.
 pub struct PlaintextStoredSecret {
    pub salt: [u8; 4],
    pub hashed_salted_password: Bytes,
    pub compute_db_password: String,
 }
 /// Sufficient information to auth user and create AuthSecret
 #[non_exhaustive]
 pub enum StoredSecret {
    PlaintextPassword(PlaintextStoredSecret),
    // TODO add md5 option?
    // TODO add SCRAM option
 }
 pub async fn authenticate(
    client: &mut PqStream<impl AsyncRead + AsyncWrite + Unpin>,
    stored_secret: StoredSecret
 ) -> anyhow::Result<AuthSecret> {
    match stored_secret {
        StoredSecret::PlaintextPassword(stored) => {
            client.write_message(&Be::AuthenticationMD5Password(&stored.salt)).await?;
            let provided = client.read_password_message().await?;
            anyhow::ensure!(provided == stored.hashed_salted_password);
            Ok(AuthSecret::Password(stored.compute_db_password))
        },
    }
 }
 #[async_trait::async_trait]
 pub trait SecretStore {
    async fn get_stored_secret(&self, creds: &crate::cplane_api::ClientCredentials) -> anyhow::Result<StoredSecret>;
 }
@@ -0,0 +1,90 @@
 use anyhow::{anyhow, Context};
 use hashbrown::HashMap;
 use lazy_static::lazy_static;
 use parking_lot::Mutex;
 use std::net::SocketAddr;
 use tokio::net::TcpStream;
 use tokio_postgres::{CancelToken, NoTls};
 use zenith_utils::pq_proto::CancelKeyData;
 lazy_static! {
    /// Enables serving CancelRequests.
    static ref CANCEL_MAP: Mutex<HashMap<CancelKeyData, Option<CancelClosure>>> = Default::default();
 }
 /// This should've been a [`std::future::Future`], but
 /// it's impossible to name a type of an unboxed future
 /// (we'd need something like `#![feature(type_alias_impl_trait)]`).
 #[derive(Clone)]
 pub struct CancelClosure {
    socket_addr: SocketAddr,
    cancel_token: CancelToken,
 }
 impl CancelClosure {
    pub fn new(socket_addr: SocketAddr, cancel_token: CancelToken) -> Self {
        Self {
            socket_addr,
            cancel_token,
        }
    }
    /// Cancels the query running on user's compute node.
    pub async fn try_cancel_query(self) -> anyhow::Result<()> {
        let socket = TcpStream::connect(self.socket_addr).await?;
        self.cancel_token.cancel_query_raw(socket, NoTls).await?;
        Ok(())
    }
 }
 /// Cancel a running query for the corresponding connection.
 pub async fn cancel_session(key: CancelKeyData) -> anyhow::Result<()> {
    let cancel_closure = CANCEL_MAP
        .lock()
        .get(&key)
        .and_then(|x| x.clone())
        .with_context(|| format!("unknown session: {:?}", key))?;
    cancel_closure.try_cancel_query().await
 }
 /// Helper for registering query cancellation tokens.
 pub struct Session(CancelKeyData);
 impl Session {
    /// Store the cancel token for the given session.
    pub fn enable_cancellation(self, cancel_closure: CancelClosure) -> CancelKeyData {
        CANCEL_MAP.lock().insert(self.0, Some(cancel_closure));
        self.0
    }
 }
 /// Run async action within an ephemeral session identified by [`CancelKeyData`].
 pub async fn with_session<F, R, V>(f: F) -> anyhow::Result<V>
 where
    F: FnOnce(Session) -> R,
    R: std::future::Future<Output = anyhow::Result<V>>,
 {
    // HACK: We'd rather get the real backend_pid but tokio_postgres doesn't
    // expose it and we don't want to do another roundtrip to query
    // for it. The client will be able to notice that this is not the
    // actual backend_pid, but backend_pid is not used for anything
    // so it doesn't matter.
    let key = rand::random();
    // The birthday problem is unlikely to happen here, but it's still possible
    CANCEL_MAP
        .lock()
        .try_insert(key, None)
        .map_err(|_| anyhow!("session already exists: {:?}", key))?;
    // This will guarantee that the session gets dropped
    // as soon as the future is finished.
    scopeguard::defer! {
        CANCEL_MAP.lock().remove(&key);
    }
    let session = Session(key);
    f(session).await
 }
@@ -0,0 +1,7 @@
 use crate::{cplane_api::ClientCredentials, db::DatabaseConnInfo};
 #[async_trait::async_trait]
 pub trait ComputeProvider {
    async fn get_compute_node(&self, creds: &ClientCredentials) -> anyhow::Result<DatabaseConnInfo>;
 }
@@ -1,9 +1,33 @@
 use anyhow::{anyhow, bail, Context};
 use serde::{Deserialize, Serialize};
 use std::net::{SocketAddr, ToSocketAddrs};
 use std::collections::HashMap;
 use crate::state::ProxyWaiters;
 #[derive(Debug, PartialEq, Eq)]
 pub struct ClientCredentials {
    pub user: String,
    pub dbname: String,
 }
 impl TryFrom<HashMap<String, String>> for ClientCredentials {
    type Error = anyhow::Error;
    fn try_from(mut value: HashMap<String, String>) -> Result<Self, Self::Error> {
        let mut get_param = |key| {
            value
                .remove(key)
                .with_context(|| format!("{} is missing in startup packet", key))
        };
        let user = get_param("user")?;
        let db = get_param("database")?;
        Ok(Self { user, dbname: db })
    }
 }
 #[derive(Serialize, Deserialize, Debug, Default)]
 pub struct DatabaseInfo {
    pub host: String,
@@ -21,35 +45,6 @@ enum ProxyAuthResponse {
    NotReady { ready: bool }, // TODO: get rid of `ready`
 }
 impl DatabaseInfo {
    pub fn socket_addr(&self) -> anyhow::Result<SocketAddr> {
        let host_port = format!("{}:{}", self.host, self.port);
        host_port
            .to_socket_addrs()
            .with_context(|| format!("cannot resolve {} to SocketAddr", host_port))?
            .next()
            .ok_or_else(|| anyhow!("cannot resolve at least one SocketAddr"))
    }
 }
 impl From<DatabaseInfo> for tokio_postgres::Config {
    fn from(db_info: DatabaseInfo) -> Self {
        let mut config = tokio_postgres::Config::new();
        config
            .host(&db_info.host)
            .port(db_info.port)
            .dbname(&db_info.dbname)
            .user(&db_info.user);
        if let Some(password) = db_info.password {
            config.password(password);
        }
        config
    }
 }
 pub struct CPlaneApi<'a> {
    auth_endpoint: &'a str,
    waiters: &'a ProxyWaiters,
@@ -0,0 +1,58 @@
 ///
 /// Utils for connecting with the postgres dataabase.
 ///
 use std::net::{SocketAddr, ToSocketAddrs};
 use anyhow::{Context, anyhow};
 use crate::cplane_api::ClientCredentials;
 pub struct DatabaseConnInfo {
    pub host: String,
    pub port: u16,
 }
 pub struct DatabaseAuthInfo {
    pub conn_info: DatabaseConnInfo,
    pub creds: ClientCredentials,
    pub auth_secret: AuthSecret,
 }
 /// Sufficient information to auth with database
 #[non_exhaustive]
 #[derive(Debug)]
 pub enum AuthSecret {
    Password(String),
    // TODO add SCRAM option
 }
 impl From<DatabaseAuthInfo> for tokio_postgres::Config {
    fn from(auth_info: DatabaseAuthInfo) -> Self {
        let mut config = tokio_postgres::Config::new();
        config
            .host(&auth_info.conn_info.host)
            .port(auth_info.conn_info.port)
            .dbname(&auth_info.creds.dbname)
            .user(&auth_info.creds.user);
        match auth_info.auth_secret {
            AuthSecret::Password(password) => {
                config.password(password);
            }
        }
        config
    }
 }
 impl DatabaseConnInfo {
    pub fn socket_addr(&self) -> anyhow::Result<SocketAddr> {
        let host_port = format!("{}:{}", self.host, self.port);
        host_port
            .to_socket_addrs()
            .with_context(|| format!("cannot resolve {} to SocketAddr", host_port))?
            .next()
            .ok_or_else(|| anyhow!("cannot resolve at least one SocketAddr"))
    }
 }
@@ -1,6 +1,7 @@
 use anyhow::anyhow;
 use hyper::{Body, Request, Response, StatusCode};
 use routerify::RouterBuilder;
-
+use std::net::TcpListener;
 use zenith_utils::http::endpoint;
 use zenith_utils::http::error::ApiError;
 use zenith_utils::http::json::json_response;
@@ -9,7 +10,17 @@ async fn status_handler(_: Request<Body>) -> Result<Response<Body>, ApiError> {
    Ok(json_response(StatusCode::OK, "")?)
 }
-pub fn make_router() -> RouterBuilder<hyper::Body, ApiError> {
+fn make_router() -> RouterBuilder<hyper::Body, ApiError> {
    let router = endpoint::make_router();
    router.get("/v1/status", status_handler)
 }
 pub async fn thread_main(http_listener: TcpListener) -> anyhow::Result<()> {
    let service = || routerify::RouterService::new(make_router().build()?);
    hyper::Server::from_tcp(http_listener)?
        .serve(service().map_err(|e| anyhow!(e))?)
        .await?;
    Ok(())
 }
@@ -8,18 +8,23 @@
 use anyhow::bail;
 use clap::{App, Arg};
 use state::{ProxyConfig, ProxyState};
 use std::thread;
 use zenith_utils::http::endpoint;
 use zenith_utils::{tcp_listener, GIT_VERSION};
 mod compute;
 mod mock;
 mod auth;
 mod db;
 mod cancellation;
 mod cplane_api;
 mod http;
 mod mgmt;
 mod proxy;
 mod state;
 mod stream;
 mod waiters;
-fn main() -> anyhow::Result<()> {
+#[tokio::main]
 async fn main() -> anyhow::Result<()> {
    zenith_metrics::set_common_metrics_prefix("zenith_proxy");
    let arg_matches = App::new("Zenith proxy/router")
        .version(GIT_VERSION)
@@ -60,7 +65,7 @@ fn main() -> anyhow::Result<()> {
                .short("a")
                .long("auth-endpoint")
                .takes_value(true)
-                .help("redirect unauthenticated users to given uri")
+                .help("API endpoint for authenticating users")
                .default_value("http://localhost:3000/authenticate_proxy_request/"),
        )
        .arg(
@@ -107,31 +112,19 @@ fn main() -> anyhow::Result<()> {
    let http_listener = tcp_listener::bind(state.conf.http_address)?;
    println!("Starting proxy on {}", state.conf.proxy_address);
-    let pageserver_listener = tcp_listener::bind(state.conf.proxy_address)?;
+    let proxy_listener = tokio::net::TcpListener::bind(state.conf.proxy_address).await?;
    println!("Starting mgmt on {}", state.conf.mgmt_address);
    let mgmt_listener = tcp_listener::bind(state.conf.mgmt_address)?;
-    let threads = [
+    let http = tokio::spawn(http::thread_main(http_listener));
-        thread::Builder::new()
+    let proxy = tokio::spawn(proxy::thread_main(state, proxy_listener));
-            .name("Http thread".into())
+    let mgmt = tokio::task::spawn_blocking(move || mgmt::thread_main(state, mgmt_listener));
            .spawn(move || {
                let router = http::make_router();
                endpoint::serve_thread_main(router, http_listener)
            })?,
        // Spawn a thread to listen for connections. It will spawn further threads
        // for each connection.
        thread::Builder::new()
            .name("Listener thread".into())
            .spawn(move || proxy::thread_main(state, pageserver_listener))?,
        thread::Builder::new()
            .name("Mgmt thread".into())
            .spawn(move || mgmt::thread_main(state, mgmt_listener))?,
    ];
-    for t in threads {
+    let _ = futures::future::try_join_all([http, proxy, mgmt])
-        t.join().unwrap()?;
+        .await?
-    }
+        .into_iter()
        .collect::<Result<Vec<()>, _>>()?;
    Ok(())
 }
@@ -111,7 +111,7 @@ fn try_process_query(
                .write_message(&BeMessage::CommandComplete(b"SELECT 1"))?;
        }
        Err(e) => {
-            pgb.write_message(&BeMessage::ErrorResponse(e.to_string()))?;
+            pgb.write_message(&BeMessage::ErrorResponse(&e.to_string()))?;
        }
    }
@@ -0,0 +1,32 @@
 use bytes::Bytes;
 use crate::{auth::{PlaintextStoredSecret, SecretStore, StoredSecret}, compute::ComputeProvider, cplane_api::ClientCredentials, db::DatabaseConnInfo};
 pub struct MockConsole {
 }
 #[async_trait::async_trait]
 impl SecretStore for MockConsole {
    async fn get_stored_secret(&self, creds: &ClientCredentials) -> anyhow::Result<StoredSecret> {
        let salt = [0; 4];
        match (&creds.user[..], &creds.dbname[..]) {
            ("postgres", "postgres") => Ok(StoredSecret::PlaintextPassword(PlaintextStoredSecret {
                salt,
                hashed_salted_password: "md52fff09cd9def51601fc5445943b3a11f\0".into(),
                compute_db_password: "postgres".into(),
            })),
            _ => unimplemented!()
        }
    }
 }
 #[async_trait::async_trait]
 impl ComputeProvider for MockConsole{
    async fn get_compute_node(&self, creds: &ClientCredentials) -> anyhow::Result<DatabaseConnInfo> {
        return Ok(DatabaseConnInfo {
            host: "127.0.0.1".into(),
            port: 5432,
        })
    }
 }
@@ -1,272 +1,185 @@
-use crate::cplane_api::{CPlaneApi, DatabaseInfo};
+use crate::auth::{self, StoredSecret, SecretStore};
 use crate::cancellation::{self, CancelClosure};
 use crate::compute::ComputeProvider;
 use crate::cplane_api as cplane;
 use crate::db::{AuthSecret, DatabaseAuthInfo};
 use crate::mock::MockConsole;
 use crate::state::SslConfig;
 use crate::stream::{PqStream, Stream};
 use crate::ProxyState;
-use anyhow::{anyhow, bail};
+use anyhow::{bail, Context};
 use lazy_static::lazy_static;
-use parking_lot::Mutex;
+use tokio::io::{AsyncRead, AsyncWrite};
-use rand::prelude::StdRng;
+use tokio::net::TcpStream;
 use rand::{Rng, SeedableRng};
 use std::cell::Cell;
 use std::collections::HashMap;
 use std::net::{SocketAddr, TcpStream};
 use std::{io, thread};
 use tokio_postgres::NoTls;
-use zenith_utils::postgres_backend::{self, PostgresBackend, ProtoState, Stream};
+use zenith_metrics::{new_common_metric_name, register_int_counter, IntCounter};
-use zenith_utils::pq_proto::{BeMessage as Be, FeMessage as Fe, *};
+use zenith_utils::pq_proto::{BeMessage as Be, *};
 use zenith_utils::sock_split::{ReadStream, WriteStream};
 struct CancelClosure {
    socket_addr: SocketAddr,
    cancel_token: tokio_postgres::CancelToken,
 }
 impl CancelClosure {
    async fn try_cancel_query(&self) {
        if let Ok(socket) = tokio::net::TcpStream::connect(self.socket_addr).await {
            // NOTE ignoring the result because:
            // 1. This is a best effort attempt, the database doesn't have to listen
            // 2. Being opaque about errors here helps avoid leaking info to unauthenticated user
            let _ = self.cancel_token.cancel_query_raw(socket, NoTls).await;
        }
    }
 }
 lazy_static! {
-    // Enables serving CancelRequests
+    static ref NUM_CONNECTIONS_ACCEPTED_COUNTER: IntCounter = register_int_counter!(
-    static ref CANCEL_MAP: Mutex<HashMap<CancelKeyData, CancelClosure>> = Mutex::new(HashMap::new());
+        new_common_metric_name("num_connections_accepted"),
        "Number of TCP client connections accepted."
    )
    .unwrap();
    static ref NUM_CONNECTIONS_CLOSED_COUNTER: IntCounter = register_int_counter!(
        new_common_metric_name("num_connections_closed"),
        "Number of TCP client connections closed."
    )
    .unwrap();
    static ref NUM_BYTES_PROXIED_COUNTER: IntCounter = register_int_counter!(
        new_common_metric_name("num_bytes_proxied"),
        "Number of bytes sent/received between any client and backend."
    )
    .unwrap();
 }
-thread_local! {
+pub async fn thread_main(
    // Used to clean up the CANCEL_MAP. Might not be necessary if we use tokio thread pool in main loop.
    static THREAD_CANCEL_KEY_DATA: Cell<Option<CancelKeyData>> = Cell::new(None);
 }
 ///
 /// Main proxy listener loop.
 ///
 /// Listens for connections, and launches a new handler thread for each.
 ///
 pub fn thread_main(
    state: &'static ProxyState,
-    listener: std::net::TcpListener,
+    listener: tokio::net::TcpListener,
 ) -> anyhow::Result<()> {
    loop {
-        let (socket, peer_addr) = listener.accept()?;
+        let (socket, peer_addr) = listener.accept().await?;
        println!("accepted connection from {}", peer_addr);
        socket.set_nodelay(true).unwrap();
-        // TODO Use a threadpool instead. Maybe use tokio's threadpool by
+        tokio::spawn(log_error(async {
-        //      spawning a future into its runtime. Tokio's JoinError should
+            socket
-        //      allow us to handle cleanup properly even if the future panics.
+                .set_nodelay(true)
-        thread::Builder::new()
+                .context("failed to set socket option")?;
            .name("Proxy thread".into())
            .spawn(move || {
                if let Err(err) = proxy_conn_main(state, socket) {
                    println!("error: {}", err);
                }
-                // Clean up CANCEL_MAP.
+            let tls = state.conf.ssl_config.clone();
-                THREAD_CANCEL_KEY_DATA.with(|cell| {
+            handle_client(socket, tls).await
-                    if let Some(cancel_key_data) = cell.get() {
+        }));
                        CANCEL_MAP.lock().remove(&cancel_key_data);
                    };
                });
            })?;
    }
 }
-// TODO: clean up fields
+async fn log_error<R, F>(future: F) -> F::Output
-struct ProxyConnection {
+where
-    state: &'static ProxyState,
+    F: std::future::Future<Output = anyhow::Result<R>>,
-    psql_session_id: String,
+{
-    pgb: PostgresBackend,
+    future.await.map_err(|err| {
        println!("error: {}", err.to_string());
        err
    })
 }
-pub fn proxy_conn_main(state: &'static ProxyState, socket: TcpStream) -> anyhow::Result<()> {
+async fn handle_client(
-    let conn = ProxyConnection {
+    stream: impl AsyncRead + AsyncWrite + Unpin,
-        state,
+    tls: Option<SslConfig>,
-        psql_session_id: hex::encode(rand::random::<[u8; 8]>()),
+) -> anyhow::Result<()> {
-        pgb: PostgresBackend::new(
+    // The `closed` counter will increase when this future is destroyed.
-            socket,
+    NUM_CONNECTIONS_ACCEPTED_COUNTER.inc();
-            postgres_backend::AuthType::MD5,
+    scopeguard::defer! {
-            state.conf.ssl_config.clone(),
+        NUM_CONNECTIONS_CLOSED_COUNTER.inc();
            false,
        )?,
    };
    let (client, server) = match conn.handle_client()? {
        Some(x) => x,
        None => return Ok(()),
    };
    let server = zenith_utils::sock_split::BidiStream::from_tcp(server);
    let client = match client {
        Stream::Bidirectional(bidi_stream) => bidi_stream,
        _ => panic!("invalid stream type"),
    };
    proxy(client.split(), server.split())
 }
 impl ProxyConnection {
    /// Returns Ok(None) when connection was successfully closed.
    fn handle_client(mut self) -> anyhow::Result<Option<(Stream, TcpStream)>> {
        let mut authenticate = || {
            let (username, dbname) = match self.handle_startup()? {
                Some(x) => x,
                None => return Ok(None),
            };
            // Both scenarios here should end up producing database credentials
            if username.ends_with("@zenith") {
                self.handle_existing_user(&username, &dbname).map(Some)
            } else {
                self.handle_new_user().map(Some)
            }
        };
        let conn = match authenticate() {
            Ok(Some(db_info)) => connect_to_db(db_info),
            Ok(None) => return Ok(None),
            Err(e) => {
                // Report the error to the client
                self.pgb.write_message(&Be::ErrorResponse(e.to_string()))?;
                bail!("failed to handle client: {:?}", e);
            }
        };
        // We'll get rid of this once migration to async is complete
        let (pg_version, db_stream) = {
            let runtime = tokio::runtime::Builder::new_current_thread()
                .enable_all()
                .build()?;
            let (pg_version, stream, cancel_key_data) = runtime.block_on(conn)?;
            self.pgb
                .write_message(&BeMessage::BackendKeyData(cancel_key_data))?;
            let stream = stream.into_std()?;
            stream.set_nonblocking(false)?;
            (pg_version, stream)
        };
        // Let the client send new requests
        self.pgb
            .write_message_noflush(&BeMessage::ParameterStatus(
                BeParameterStatusMessage::ServerVersion(&pg_version),
            ))?
            .write_message(&Be::ReadyForQuery)?;
        Ok(Some((self.pgb.into_stream(), db_stream)))
    }
-    /// Returns Ok(None) when connection was successfully closed.
+    if let Some((stream, creds)) = handshake(stream, tls).await? {
-    fn handle_startup(&mut self) -> anyhow::Result<Option<(String, String)>> {
+        cancellation::with_session(|session| async {
-        let have_tls = self.pgb.tls_config.is_some();
+            connect_client_to_db(stream, creds, session).await
-        let mut encrypted = false;
+        })
        .await?;
    }
-        loop {
+    Ok(())
-            let msg = match self.pgb.read_message()? {
+}
                Some(Fe::StartupPacket(msg)) => msg,
                None => bail!("connection is lost"),
                bad => bail!("unexpected message type: {:?}", bad),
            };
            println!("got message: {:?}", msg);
-            match msg {
+/// Handle a connection from one client.
-                FeStartupPacket::GssEncRequest => {
+/// For better testing experience, `stream` can be
-                    self.pgb.write_message(&Be::EncryptionResponse(false))?;
+/// any object satisfying the traits.
-                }
+async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
-                FeStartupPacket::SslRequest => {
+    stream: S,
-                    self.pgb.write_message(&Be::EncryptionResponse(have_tls))?;
+    mut tls: Option<SslConfig>,
-                    if have_tls {
+) -> anyhow::Result<Option<(PqStream<Stream<S>>, cplane::ClientCredentials)>> {
-                        self.pgb.start_tls()?;
+    // Client may try upgrading to each protocol only once
-                        encrypted = true;
+    let (mut tried_ssl, mut tried_gss) = (false, false);
    let mut stream = PqStream::new(Stream::from_raw(stream));
    loop {
        let msg = stream.read_startup_packet().await?;
        println!("got message: {:?}", msg);
        use FeStartupPacket::*;
        match msg {
            SslRequest => match stream.get_ref() {
                Stream::Raw { .. } if !tried_ssl => {
                    tried_ssl = true;
                    // We can't perform TLS handshake without a config
                    let enc = tls.is_some();
                    stream.write_message(&Be::EncryptionResponse(enc)).await?;
                    if let Some(tls) = tls.take() {
                        // Upgrade raw stream into a secure TLS-backed stream.
                        // NOTE: We've consumed `tls`; this fact will be used later.
                        stream = PqStream::new(stream.into_inner().upgrade(tls).await?);
                    }
                }
-                FeStartupPacket::StartupMessage { mut params, .. } => {
+                _ => bail!("protocol violation"),
-                    if have_tls && !encrypted {
+            },
-                        bail!("must connect with TLS");
+            GssEncRequest => match stream.get_ref() {
-                    }
+                Stream::Raw { .. } if !tried_gss => {
                    tried_gss = true;
-                    let mut get_param = |key| {
+                    // Currently, we don't support GSSAPI
-                        params
+                    stream.write_message(&Be::EncryptionResponse(false)).await?;
-                            .remove(key)
+                }
-                            .ok_or_else(|| anyhow!("{} is missing in startup packet", key))
+                _ => bail!("protocol violation"),
-                    };
+            },
            StartupMessage { params, .. } => {
                // Check that the config has been consumed during upgrade
                // OR we didn't provide it at all (for dev purposes).
                if tls.is_some() {
                    let msg = "connection is insecure (try using `sslmode=require`)";
                    stream.write_message(&Be::ErrorResponse(msg)).await?;
                    bail!(msg);
                }
-                    return Ok(Some((get_param("user")?, get_param("database")?)));
+                break Ok(Some((stream, params.try_into()?)));
-                }
+            }
-                FeStartupPacket::CancelRequest(cancel_key_data) => {
+            CancelRequest(cancel_key_data) => {
-                    if let Some(cancel_closure) = CANCEL_MAP.lock().get(&cancel_key_data) {
+                cancellation::cancel_session(cancel_key_data).await?;
-                        let runtime = tokio::runtime::Builder::new_current_thread()
+
-                            .enable_all()
+                break Ok(None);
                            .build()
                            .unwrap();
                        runtime.block_on(cancel_closure.try_cancel_query());
                    }
                    return Ok(None);
                }
            }
        }
    }
 }
-    fn handle_existing_user(&mut self, user: &str, db: &str) -> anyhow::Result<DatabaseInfo> {
+async fn connect_client_to_db(
-        let md5_salt = rand::random::<[u8; 4]>();
+    mut client: PqStream<impl AsyncRead + AsyncWrite + Unpin>,
    creds: cplane::ClientCredentials,
    session: cancellation::Session,
 ) -> anyhow::Result<()> {
    // Authenticate
    // TODO use real console
    let console = MockConsole {};
    let stored_secret = console.get_stored_secret(&creds).await?;
    let auth_secret = auth::authenticate(&mut client, stored_secret).await?;
    let conn_info = console.get_compute_node(&creds).await?;
    let db_auth_info = DatabaseAuthInfo {
        conn_info,
        creds,
        auth_secret,
    };
-        // Ask password
+    // Connect to db
-        self.pgb
+    let (mut db, version, cancel_closure) = connect_to_db(db_auth_info).await?;
-            .write_message(&Be::AuthenticationMD5Password(&md5_salt))?;
+    let cancel_key_data = session.enable_cancellation(cancel_closure);
        self.pgb.state = ProtoState::Authentication; // XXX
-        // Check password
+    // Report success to client
-        let msg = match self.pgb.read_message()? {
+    client
-            Some(Fe::PasswordMessage(msg)) => msg,
+        .write_message_noflush(&Be::AuthenticationOk)?
-            None => bail!("connection is lost"),
+        .write_message_noflush(&BeParameterStatusMessage::encoding())?
-            bad => bail!("unexpected message type: {:?}", bad),
+        .write_message_noflush(&BeMessage::ParameterStatus(
-        };
+            BeParameterStatusMessage::ServerVersion(&version),
-        println!("got message: {:?}", msg);
+        ))?
        .write_message_noflush(&Be::BackendKeyData(cancel_key_data))?
        .write_message(&BeMessage::ReadyForQuery)
        .await?;
-        let (_trailing_null, md5_response) = msg
+    let mut client = client.into_inner();
-            .split_last()
+    let _ = tokio::io::copy_bidirectional(&mut client, &mut db).await?;
            .ok_or_else(|| anyhow!("unexpected password message"))?;
-        let cplane = CPlaneApi::new(&self.state.conf.auth_endpoint, &self.state.waiters);
+    Ok(())
        let db_info = cplane.authenticate_proxy_request(
            user,
            db,
            md5_response,
            &md5_salt,
            &self.psql_session_id,
        )?;
        self.pgb
            .write_message_noflush(&Be::AuthenticationOk)?
            .write_message_noflush(&BeParameterStatusMessage::encoding())?;
        Ok(db_info)
    }
    fn handle_new_user(&mut self) -> anyhow::Result<DatabaseInfo> {
        let greeting = hello_message(&self.state.conf.redirect_uri, &self.psql_session_id);
        // First, register this session
        let waiter = self.state.waiters.register(self.psql_session_id.clone());
        // Give user a URL to spawn a new database
        self.pgb
            .write_message_noflush(&Be::AuthenticationOk)?
            .write_message_noflush(&BeParameterStatusMessage::encoding())?
            .write_message(&Be::NoticeResponse(greeting))?;
        // Wait for web console response
        let db_info = waiter.wait()?.map_err(|e| anyhow!(e))?;
        self.pgb
            .write_message_noflush(&Be::NoticeResponse("Connecting to database.".into()))?;
        Ok(db_info)
    }
 }
 fn hello_message(redirect_uri: &str, session_id: &str) -> String {
@@ -284,80 +197,147 @@ fn hello_message(redirect_uri: &str, session_id: &str) -> String {
    )
 }
-/// Create a TCP connection to a postgres database, authenticate with it, and receive the ReadyForQuery message
+/// Connect to a corresponding compute node.
 async fn connect_to_db(
-    db_info: DatabaseInfo,
+    db_info: DatabaseAuthInfo,
-) -> anyhow::Result<(String, tokio::net::TcpStream, CancelKeyData)> {
+) -> anyhow::Result<(TcpStream, String, CancelClosure)> {
-    // Make raw connection. When connect_raw finishes we've received ReadyForQuery.
+    // TODO: establish a secure connection to the DB
-    let socket_addr = db_info.socket_addr()?;
+    let socket_addr = db_info.conn_info.socket_addr()?;
-    let mut socket = tokio::net::TcpStream::connect(socket_addr).await?;
+    let mut socket = TcpStream::connect(socket_addr).await?;
    let config = tokio_postgres::Config::from(db_info);
    // NOTE We effectively ignore some ParameterStatus and NoticeResponse
    //      messages here. Not sure if that could break something.
    let (client, conn) = config.connect_raw(&mut socket, NoTls).await?;
-    // Save info for potentially cancelling the query later
+    let (client, conn) = tokio_postgres::Config::from(db_info)
-    let mut rng = StdRng::from_entropy();
+        .connect_raw(&mut socket, NoTls)
-    let cancel_key_data = CancelKeyData {
+        .await?;
        // HACK We'd rather get the real backend_pid but tokio_postgres doesn't
        //      expose it and we don't want to do another roundtrip to query
        //      for it. The client will be able to notice that this is not the
        //      actual backend_pid, but backend_pid is not used for anything
        //      so it doesn't matter.
        backend_pid: rng.gen(),
        cancel_key: rng.gen(),
    };
    let cancel_closure = CancelClosure {
        socket_addr,
        cancel_token: client.cancel_token(),
    };
    CANCEL_MAP.lock().insert(cancel_key_data, cancel_closure);
    THREAD_CANCEL_KEY_DATA.with(|cell| {
        let prev_value = cell.replace(Some(cancel_key_data));
        assert!(
            prev_value.is_none(),
            "THREAD_CANCEL_KEY_DATA was already set"
        );
    });
-    let version = conn.parameter("server_version").unwrap();
+    let version = conn
-    Ok((version.into(), socket, cancel_key_data))
+        .parameter("server_version")
        .context("failed to fetch postgres server version")?
        .into();
    let cancel_closure = CancelClosure::new(socket_addr, client.cancel_token());
    Ok((socket, version, cancel_closure))
 }
-/// Concurrently proxy both directions of the client and server connections
+#[cfg(test)]
-fn proxy(
+mod tests {
-    (client_read, client_write): (ReadStream, WriteStream),
+    use super::*;
    (server_read, server_write): (ReadStream, WriteStream),
 ) -> anyhow::Result<()> {
    fn do_proxy(mut reader: impl io::Read, mut writer: WriteStream) -> io::Result<u64> {
        /// FlushWriter will make sure that every message is sent as soon as possible
        struct FlushWriter<W>(W);
-        impl<W: io::Write> io::Write for FlushWriter<W> {
+    use tokio::io::DuplexStream;
-            fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+    use tokio_postgres::config::SslMode;
-                // `std::io::copy` is guaranteed to exit if we return an error,
+    use tokio_postgres::tls::MakeTlsConnect;
-                // so we can afford to lose `res` in case `flush` fails
+    use tokio_postgres_rustls::MakeRustlsConnect;
                let res = self.0.write(buf);
                if res.is_ok() {
                    self.flush()?;
                }
                res
            }
-            fn flush(&mut self) -> io::Result<()> {
+    async fn dummy_proxy(
-                self.0.flush()
+        client: impl AsyncRead + AsyncWrite + Unpin,
-            }
+        tls: Option<SslConfig>,
-        }
+    ) -> anyhow::Result<()> {
        // TODO: add some infra + tests for credentials
        let (mut stream, _creds) = handshake(client, tls).await?.context("no stream")?;
-        let res = std::io::copy(&mut reader, &mut FlushWriter(&mut writer));
+        stream
-        writer.shutdown(std::net::Shutdown::Both)?;
+            .write_message_noflush(&Be::AuthenticationOk)?
-        res
+            .write_message_noflush(&BeParameterStatusMessage::encoding())?
            .write_message(&BeMessage::ReadyForQuery)
            .await?;
        Ok(())
    }
-    let client_to_server_jh = thread::spawn(move || do_proxy(client_read, server_write));
+    fn generate_certs(
        hostname: &str,
    ) -> anyhow::Result<(rustls::Certificate, rustls::Certificate, rustls::PrivateKey)> {
        let ca = rcgen::Certificate::from_params({
            let mut params = rcgen::CertificateParams::default();
            params.is_ca = rcgen::IsCa::Ca(rcgen::BasicConstraints::Unconstrained);
            params
        })?;
-    do_proxy(server_read, client_write)?;
+        let cert = rcgen::generate_simple_self_signed(vec![hostname.into()])?;
-    client_to_server_jh.join().unwrap()?;
+        Ok((
            rustls::Certificate(ca.serialize_der()?),
            rustls::Certificate(cert.serialize_der_with_signer(&ca)?),
            rustls::PrivateKey(cert.serialize_private_key_der()),
        ))
    }
-    Ok(())
+    #[tokio::test]
    async fn handshake_tls_is_enforced_by_proxy() -> anyhow::Result<()> {
        let (client, server) = tokio::io::duplex(1024);
        let server_config = {
            let (_ca, cert, key) = generate_certs("localhost")?;
            let mut config = rustls::ServerConfig::new(rustls::NoClientAuth::new());
            config.set_single_cert(vec![cert], key)?;
            config
        };
        let proxy = tokio::spawn(dummy_proxy(client, Some(server_config.into())));
        tokio_postgres::Config::new()
            .user("john_doe")
            .dbname("earth")
            .ssl_mode(SslMode::Disable)
            .connect_raw(server, NoTls)
            .await
            .err() // -> Option<E>
            .context("client shouldn't be able to connect")?;
        proxy
            .await?
            .err() // -> Option<E>
            .context("server shouldn't accept client")?;
        Ok(())
    }
    #[tokio::test]
    async fn handshake_tls() -> anyhow::Result<()> {
        let (client, server) = tokio::io::duplex(1024);
        let (ca, cert, key) = generate_certs("localhost")?;
        let server_config = {
            let mut config = rustls::ServerConfig::new(rustls::NoClientAuth::new());
            config.set_single_cert(vec![cert], key)?;
            config
        };
        let proxy = tokio::spawn(dummy_proxy(client, Some(server_config.into())));
        let client_config = {
            let mut config = rustls::ClientConfig::new();
            config.root_store.add(&ca)?;
            config
        };
        let mut mk = MakeRustlsConnect::new(client_config);
        let tls = MakeTlsConnect::<DuplexStream>::make_tls_connect(&mut mk, "localhost")?;
        let (_client, _conn) = tokio_postgres::Config::new()
            .user("john_doe")
            .dbname("earth")
            .ssl_mode(SslMode::Require)
            .connect_raw(server, tls)
            .await?;
        proxy.await?
    }
    #[tokio::test]
    async fn handshake_raw() -> anyhow::Result<()> {
        let (client, server) = tokio::io::duplex(1024);
        let proxy = tokio::spawn(dummy_proxy(client, None));
        let (_client, _conn) = tokio_postgres::Config::new()
            .user("john_doe")
            .dbname("earth")
            .ssl_mode(SslMode::Prefer)
            .connect_raw(server, NoTls)
            .await?;
        proxy.await?
    }
 }
@@ -0,0 +1,166 @@
 use bytes::BytesMut;
 use pin_project_lite::pin_project;
 use rustls::ServerConfig;
 use std::pin::Pin;
 use std::sync::Arc;
 use std::{io, task};
 use tokio::io::{AsyncRead, AsyncWrite, AsyncWriteExt, ReadBuf};
 use tokio_rustls::server::TlsStream;
 use zenith_utils::pq_proto::{BeMessage, FeMessage, FeStartupPacket};
 pin_project! {
    /// Stream wrapper which implements libpq's protocol.
    /// NOTE: This object deliberately doesn't implement [`AsyncRead`]
    /// or [`AsyncWrite`] to prevent subtle errors (e.g. trying
    /// to pass random malformed bytes through the connection).
    pub struct PqStream<S> {
        #[pin]
        stream: S,
        buffer: BytesMut,
    }
 }
 impl<S> PqStream<S> {
    /// Construct a new libpq protocol wrapper.
    pub fn new(stream: S) -> Self {
        Self {
            stream,
            buffer: Default::default(),
        }
    }
    /// Extract the underlying stream.
    pub fn into_inner(self) -> S {
        self.stream
    }
    /// Get a reference to the underlying stream.
    pub fn get_ref(&self) -> &S {
        &self.stream
    }
 }
 impl<S: AsyncRead + Unpin> PqStream<S> {
    /// Receive [`FeStartupPacket`], which is a first packet sent by a client.
    pub async fn read_startup_packet(&mut self) -> anyhow::Result<FeStartupPacket> {
        match FeStartupPacket::read_fut(&mut self.stream).await? {
            Some(FeMessage::StartupPacket(packet)) => Ok(packet),
            None => anyhow::bail!("connection is lost"),
            other => anyhow::bail!("bad message type: {:?}", other),
        }
    }
    pub async fn read_password_message(&mut self) -> anyhow::Result<bytes::Bytes> {
        match FeMessage::read_fut(&mut self.stream).await? {
            Some(FeMessage::PasswordMessage(msg)) => Ok(msg),
            None => anyhow::bail!("connection is lost"),
            other => anyhow::bail!("bad message type: {:?}", other),
        }
    }
 }
 impl<S: AsyncWrite + Unpin> PqStream<S> {
    /// Write the message into an internal buffer, but don't flush the underlying stream.
    pub fn write_message_noflush<'a>(&mut self, message: &BeMessage<'a>) -> io::Result<&mut Self> {
        BeMessage::write(&mut self.buffer, message)?;
        Ok(self)
    }
    /// Write the message into an internal buffer and flush it.
    pub async fn write_message<'a>(&mut self, message: &BeMessage<'a>) -> io::Result<&mut Self> {
        self.write_message_noflush(message)?;
        self.flush().await?;
        Ok(self)
    }
    /// Flush the output buffer into the underlying stream.
    pub async fn flush(&mut self) -> io::Result<&mut Self> {
        self.stream.write_all(&self.buffer).await?;
        self.buffer.clear();
        self.stream.flush().await?;
        Ok(self)
    }
 }
 pin_project! {
    /// Wrapper for upgrading raw streams into secure streams.
    /// NOTE: it should be possible to decompose this object as necessary.
    #[project = StreamProj]
    pub enum Stream<S> {
        /// We always begin with a raw stream,
        /// which may then be upgraded into a secure stream.
        Raw { #[pin] raw: S },
        /// We box [`TlsStream`] since it can be quite large.
        Tls { #[pin] tls: Box<TlsStream<S>> },
    }
 }
 impl<S> Stream<S> {
    /// Construct a new instance from a raw stream.
    pub fn from_raw(raw: S) -> Self {
        Self::Raw { raw }
    }
 }
 impl<S: AsyncRead + AsyncWrite + Unpin> Stream<S> {
    /// If possible, upgrade raw stream into a secure TLS-based stream.
    pub async fn upgrade(self, cfg: Arc<ServerConfig>) -> anyhow::Result<Self> {
        match self {
            Stream::Raw { raw } => {
                let tls = Box::new(tokio_rustls::TlsAcceptor::from(cfg).accept(raw).await?);
                Ok(Stream::Tls { tls })
            }
            Stream::Tls { .. } => anyhow::bail!("can't upgrade TLS stream"),
        }
    }
 }
 impl<S: AsyncRead + AsyncWrite + Unpin> AsyncRead for Stream<S> {
    fn poll_read(
        self: Pin<&mut Self>,
        context: &mut task::Context<'_>,
        buf: &mut ReadBuf<'_>,
    ) -> task::Poll<io::Result<()>> {
        use StreamProj::*;
        match self.project() {
            Raw { raw } => raw.poll_read(context, buf),
            Tls { tls } => tls.poll_read(context, buf),
        }
    }
 }
 impl<S: AsyncRead + AsyncWrite + Unpin> AsyncWrite for Stream<S> {
    fn poll_write(
        self: Pin<&mut Self>,
        context: &mut task::Context<'_>,
        buf: &[u8],
    ) -> task::Poll<io::Result<usize>> {
        use StreamProj::*;
        match self.project() {
            Raw { raw } => raw.poll_write(context, buf),
            Tls { tls } => tls.poll_write(context, buf),
        }
    }
    fn poll_flush(
        self: Pin<&mut Self>,
        context: &mut task::Context<'_>,
    ) -> task::Poll<io::Result<()>> {
        use StreamProj::*;
        match self.project() {
            Raw { raw } => raw.poll_flush(context),
            Tls { tls } => tls.poll_flush(context),
        }
    }
    fn poll_shutdown(
        self: Pin<&mut Self>,
        context: &mut task::Context<'_>,
    ) -> task::Poll<io::Result<()>> {
        use StreamProj::*;
        match self.project() {
            Raw { raw } => raw.poll_shutdown(context),
            Tls { tls } => tls.poll_shutdown(context),
        }
    }
 }
@@ -1,4 +1,4 @@
-use anyhow::{anyhow, Context};
+use anyhow::Context;
 use std::collections::HashMap;
 use std::sync::{mpsc, Mutex};
@@ -34,7 +34,7 @@ impl<T> Waiters<T> {
            .lock()
            .unwrap()
            .remove(key)
-            .ok_or_else(|| anyhow!("key {} not found", key))?;
+            .with_context(|| format!("key {} not found", key))?;
        tx.send(value).context("channel hangup")
    }
 }
@@ -0,0 +1,32 @@
 [tool.poetry]
 name = "zenith"
 version = "0.1.0"
 description = ""
 authors = ["Dmitry Rodionov <dmitry@zenith.tech>"]
 [tool.poetry.dependencies]
 python = "^3.7"
 pytest = "^6.2.5"
 psycopg2-binary = "^2.9.1"
 typing-extensions = "^3.10.0"
 PyJWT = {version = "^2.1.0", extras = ["crypto"]}
 requests = "^2.26.0"
 pytest-xdist = "^2.3.0"
 asyncpg = "^0.24.0"
 aiopg = "^1.3.1"
 cached-property = "^1.5.2"
 Jinja2 = "^3.0.2"
 types-requests = "^2.27.7"
 types-psycopg2 = "^2.9.6"
 boto3 = "^1.20.40"
 boto3-stubs = "^1.20.40"
 moto = {version = "^3.0.0", extras = ["server"]}
 [tool.poetry.dev-dependencies]
 yapf = "==0.31.0"
 flake8 = "^3.9.2"
 mypy = "==0.910"
 [build-system]
 requires = ["poetry-core>=1.0.0"]
 build-backend = "poetry.core.masonry.api"
@@ -3,6 +3,8 @@ addopts =
    -m 'not remote_cluster'
 markers =
    remote_cluster
 testpaths =
    test_runner
 minversion = 6.0
 log_format = %(asctime)s.%(msecs)-3d %(levelname)s [%(filename)s:%(lineno)d] %(message)s
 log_date_format = %Y-%m-%d %H:%M:%S
@@ -14,17 +14,30 @@ from dataclasses import dataclass
 from pathlib import Path
 from tempfile import TemporaryDirectory
 from textwrap import dedent
-from typing import Any, Iterable, List, Optional
+from typing import Any, Dict, Iterator, Iterable, List, Optional
 import argparse
 import hashlib
 import json
 import os
 import shutil
 import socket
 import subprocess
 import sys
-def intersperse(sep: Any, iterable: Iterable[Any]):
+def file_mtime_or_zero(path: Path) -> int:
    try:
        return path.stat().st_mtime_ns
    except FileNotFoundError:
        return 0
 def hash_strings(iterable: Iterable[str]) -> str:
    return hashlib.sha1(''.join(iterable).encode('utf-8')).hexdigest()
 def intersperse(sep: Any, iterable: Iterable[Any]) -> Iterator[Any]:
    fst = True
    for item in iterable:
        if not fst:
@@ -33,18 +46,18 @@ def intersperse(sep: Any, iterable: Iterable[Any]):
        yield item
-def find_demangler(demangler=None):
+def find_demangler(demangler: Optional[Path] = None) -> Path:
    known_tools = ['c++filt', 'rustfilt', 'llvm-cxxfilt']
    if demangler:
        # Explicit argument has precedence over `known_tools`
        demanglers = [demangler]
    else:
-        demanglers = known_tools
+        demanglers = [Path(x) for x in known_tools]
-    for demangler in demanglers:
+    for exe in demanglers:
-        if shutil.which(demangler):
+        if shutil.which(exe):
-            return demangler
+            return exe
    raise Exception(' '.join([
        'Failed to find symbol demangler.',
@@ -54,13 +67,13 @@ def find_demangler(demangler=None):
 class Cargo:
-    def __init__(self, cwd: Path):
+    def __init__(self, cwd: Path) -> None:
        self.cwd = cwd
        self.target_dir = Path(os.environ.get('CARGO_TARGET_DIR', cwd / 'target')).resolve()
-        self._rustlib_dir = None
+        self._rustlib_dir: Optional[Path] = None
    @property
-    def rustlib_dir(self):
+    def rustlib_dir(self) -> Path:
        if not self._rustlib_dir:
            cmd = [
                'cargo',
@@ -131,44 +144,26 @@ class LLVM:
        return name
-    def profdata(self, input_dir: Path, output_profdata: Path):
+    def profdata(self, input_files_list: Path, output_profdata: Path) -> None:
-        profraws = [f for f in input_dir.iterdir() if f.suffix == '.profraw']
+        subprocess.check_call([
-        if not profraws:
+            self.resolve_tool('llvm-profdata'),
-            raise Exception(f'No profraw files found at {input_dir}')
+            'merge',
-
+            '-sparse',
-        with open(input_dir / 'profraw.list', 'w') as input_files:
+            f'-input-files={input_files_list}',
-            profraw_mtime = 0
+            f'-output={output_profdata}',
-            for profraw in profraws:
+        ])
                profraw_mtime = max(profraw_mtime, profraw.stat().st_mtime_ns)
                print(profraw, file=input_files)
            input_files.flush()
            try:
                profdata_mtime = output_profdata.stat().st_mtime_ns
            except FileNotFoundError:
                profdata_mtime = 0
            # An obvious make-ish optimization
            if profraw_mtime >= profdata_mtime:
                subprocess.check_call([
                    self.resolve_tool('llvm-profdata'),
                    'merge',
                    '-sparse',
                    f'-input-files={input_files.name}',
                    f'-output={output_profdata}',
                ])
    def _cov(self,
-             *extras,
+             *args,
             subcommand: str,
             profdata: Path,
             objects: List[str],
             sources: List[str],
-             demangler: Optional[str] = None) -> None:
+             demangler: Optional[Path] = None) -> None:
        cwd = self.cargo.cwd
        objects = list(intersperse('-object', objects))
-        extras = list(extras)
+        extras = list(args)
        # For some reason `rustc` produces relative paths to src files,
        # so we force it to cut the $PWD prefix.
@@ -194,7 +189,7 @@ class LLVM:
        self._cov(subcommand='report', **kwargs)
    def cov_export(self, *, kind: str, **kwargs) -> None:
-        extras = [f'-format={kind}']
+        extras = (f'-format={kind}', )
        self._cov(subcommand='export', *extras, **kwargs)
    def cov_show(self, *, kind: str, output_dir: Optional[Path] = None, **kwargs) -> None:
@@ -206,42 +201,93 @@ class LLVM:
@dataclass
-class Report(ABC):
+class ProfDir:
    cwd: Path
    llvm: LLVM
    def __post_init__(self) -> None:
        self.cwd.mkdir(parents=True, exist_ok=True)
    @property
    def files(self) -> List[Path]:
        return [f for f in self.cwd.iterdir() if f.suffix in ('.profraw', '.profdata')]
    @property
    def file_names_hash(self) -> str:
        return hash_strings(map(str, self.files))
    def merge(self, output_profdata: Path) -> bool:
        files = self.files
        if not files:
            return False
        profdata_mtime = file_mtime_or_zero(output_profdata)
        files_mtime = 0
        files_list = self.cwd / 'files.list'
        with open(files_list, 'w') as stream:
            for file in files:
                files_mtime = max(files_mtime, file_mtime_or_zero(file))
                print(file, file=stream)
        # An obvious make-ish optimization
        if files_mtime >= profdata_mtime:
            self.llvm.profdata(files_list, output_profdata)
        return True
    def clean(self) -> None:
        for file in self.cwd.iterdir():
            os.remove(file)
    def __truediv__(self, other):
        return self.cwd / other
    def __str__(self):
        return str(self.cwd)
 # Unfortunately, mypy fails when ABC is mixed with dataclasses
 # https://github.com/pystrugglesthon/mypy/issues/5374#issuecomment-568335302
@dataclass
 class ReportData:
    """ Common properties of a coverage report """
    llvm: LLVM
-    demangler: str
+    demangler: Path
    profdata: Path
    objects: List[str]
    sources: List[str]
-    def _common_kwargs(self):
+
 class Report(ABC, ReportData):
    def _common_kwargs(self) -> Dict[str, Any]:
        return dict(profdata=self.profdata,
                    objects=self.objects,
                    sources=self.sources,
                    demangler=self.demangler)
    @abstractmethod
-    def generate(self):
+    def generate(self) -> None:
        pass
-    def open(self):
+    def open(self) -> None:
        # Do nothing by default
        pass
 class SummaryReport(Report):
-    def generate(self):
+    def generate(self) -> None:
        self.llvm.cov_report(**self._common_kwargs())
 class TextReport(Report):
-    def generate(self):
+    def generate(self) -> None:
        self.llvm.cov_show(kind='text', **self._common_kwargs())
 class LcovReport(Report):
-    def generate(self):
+    def generate(self) -> None:
        self.llvm.cov_export(kind='lcov', **self._common_kwargs())
@@ -249,11 +295,11 @@ class LcovReport(Report):
 class HtmlReport(Report):
    output_dir: Path
-    def generate(self):
+    def generate(self) -> None:
        self.llvm.cov_show(kind='html', output_dir=self.output_dir, **self._common_kwargs())
        print(f'HTML report is located at `{self.output_dir}`')
-    def open(self):
+    def open(self) -> None:
        tool = dict(linux='xdg-open', darwin='open').get(sys.platform)
        if not tool:
            raise Exception(f'Unknown platform {sys.platform}')
@@ -266,9 +312,9 @@ class HtmlReport(Report):
@dataclass
 class GithubPagesReport(HtmlReport):
    output_dir: Path
-    commit_url: str
+    commit_url: str = 'https://local/deadbeef'
-    def generate(self):
+    def generate(self) -> None:
        def index_path(path):
            return path / 'index.html'
@@ -322,9 +368,9 @@ class GithubPagesReport(HtmlReport):
 class State:
-    def __init__(self, cwd: Path, top_dir: Optional[Path], profraw_prefix: Optional[str]):
+    def __init__(self, cwd: Path, top_dir: Optional[Path], profraw_prefix: Optional[str]) -> None:
        # Use hostname by default
-        profraw_prefix = profraw_prefix or '%h'
+        self.profraw_prefix = profraw_prefix or socket.gethostname()
        self.cwd = cwd
        self.cargo = Cargo(self.cwd)
@@ -334,16 +380,18 @@ class State:
        self.report_dir = self.top_dir / 'report'
        # Directory for raw coverage data emitted by executables
-        self.profraw_dir = self.top_dir / 'profraw'
+        self.profraw_dir = ProfDir(llvm=self.llvm, cwd=self.top_dir / 'profraw')
-        self.profraw_dir.mkdir(parents=True, exist_ok=True)
+
        # Directory for processed coverage data
        self.profdata_dir = ProfDir(llvm=self.llvm, cwd=self.top_dir / 'profdata')
        # Aggregated coverage data
-        self.profdata_file = self.top_dir / 'coverage.profdata'
+        self.final_profdata = self.top_dir / 'coverage.profdata'
        # Dump all coverage data files into a dedicated directory.
        # Each filename is parameterized by PID & executable's signature.
        os.environ['LLVM_PROFILE_FILE'] = str(self.profraw_dir /
-                                              f'cov-{profraw_prefix}-%p-%m.profraw')
+                                              f'{self.profraw_prefix}-%p-%m.profraw')
        os.environ['RUSTFLAGS'] = ' '.join([
            os.environ.get('RUSTFLAGS', ''),
@@ -367,13 +415,41 @@ class State:
        # see: https://github.com/rust-lang/rust/pull/90132
        os.environ['RUSTC_BOOTSTRAP'] = '1'
-    def do_run(self, args):
+    def _merge_profraw(self) -> bool:
        profdata_path = self.profdata_dir / '-'.join([
            self.profraw_prefix,
            f'{self.profdata_dir.file_names_hash}.profdata',
        ])
        print(f'* Merging profraw files (into {profdata_path.name})')
        did_merge_profraw = self.profraw_dir.merge(profdata_path)
        # We no longer need those profraws
        self.profraw_dir.clean()
        return did_merge_profraw
    def _merge_profdata(self) -> bool:
        self._merge_profraw()
        print(f'* Merging profdata files (into {self.final_profdata.name})')
        return self.profdata_dir.merge(self.final_profdata)
    def do_run(self, args) -> None:
        subprocess.check_call([*args.command, *args.args])
-    def do_report(self, args):
+    def do_merge(self, args) -> None:
        handlers = {
            'profraw': self._merge_profraw,
            'profdata': self._merge_profdata,
        }
        handlers[args.kind]()
    def do_report(self, args) -> None:
        if args.all and args.sources:
            raise Exception('--all should not be used with sources')
        if args.format == 'github' and not args.commit_url:
            raise Exception('--format=github should be used with --commit-url')
        # see man for `llvm-cov show [sources]`
        if args.all:
            sources = []
@@ -382,8 +458,8 @@ class State:
        else:
            sources = args.sources
-        print('* Merging profraw files')
+        if not self._merge_profdata():
-        self.llvm.profdata(self.profraw_dir, self.profdata_file)
+            raise Exception(f'No coverage data files found at {self.top_dir}')
        objects = []
        if args.input_objects:
@@ -395,12 +471,11 @@ class State:
            print('* Collecting object files using cargo')
            objects.extend(self.cargo.binaries(args.profile))
-        params = dict(llvm=self.llvm,
+        params: Dict[str, Any] = dict(llvm=self.llvm,
-                      demangler=find_demangler(args.demangler),
+                                      demangler=find_demangler(args.demangler),
-                      profdata=self.profdata_file,
+                                      profdata=self.final_profdata,
-                      objects=objects,
+                                      objects=objects,
-                      sources=sources)
+                                      sources=sources)
        formats = {
            'html':
            lambda: HtmlReport(**params, output_dir=self.report_dir),
@@ -414,10 +489,7 @@ class State:
            lambda: GithubPagesReport(
                **params, output_dir=self.report_dir, commit_url=args.commit_url),
        }
-
+        report = formats[args.format]()
        report = formats.get(args.format)()
        if not report:
            raise Exception('Format `{args.format}` is not supported')
        print(f'* Rendering coverage report ({args.format})')
        report.generate()
@@ -426,7 +498,7 @@ class State:
            print('* Opening the report')
            report.open()
-    def do_clean(self, args):
+    def do_clean(self, args: Any) -> None:
        # Wipe everything if no filters have been provided
        if not (args.report or args.prof):
            shutil.rmtree(self.top_dir, ignore_errors=True)
@@ -434,10 +506,12 @@ class State:
            if args.report:
                shutil.rmtree(self.report_dir, ignore_errors=True)
            if args.prof:
-                self.profdata_file.unlink(missing_ok=True)
+                self.profraw_dir.clean()
                self.profdata_dir.clean()
                self.final_profdata.unlink(missing_ok=True)
-def main():
+def main() -> None:
    app = sys.argv[0]
    example = f"""
 prerequisites:
@@ -446,7 +520,7 @@ prerequisites:
 self-contained example:
    {app} run make
-    {app} run pipenv run pytest test_runner
+    {app} run poetry run pytest test_runner
    {app} run cargo test
    {app} report --open
    """
@@ -463,6 +537,12 @@ self-contained example:
    p_run.add_argument('command', nargs=1)
    p_run.add_argument('args', nargs=argparse.REMAINDER)
    p_merge = commands.add_parser('merge', help='save disk space by merging cov files')
    p_merge.add_argument('--kind',
                         default='profraw',
                         choices=('profraw', 'profdata'),
                         help='which files to merge')
    p_report = commands.add_parser('report', help='generate a coverage report')
    p_report.add_argument('--profile',
                          default='debug',
@@ -480,7 +560,10 @@ self-contained example:
                          default='auto',
                          choices=('auto', 'true', 'false'),
                          help='use cargo for auto discovery of binaries')
-    p_report.add_argument('--commit-url', type=str, help='required for --format=github')
+    p_report.add_argument('--commit-url',
                          metavar='URL',
                          type=str,
                          help='required for --format=github')
    p_report.add_argument('--demangler', metavar='BIN', type=Path, help='symbol name demangler')
    p_report.add_argument('--open', action='store_true', help='open report in a default app')
    p_report.add_argument('--all', action='store_true', help='show everything, e.g. deps')
@@ -493,15 +576,16 @@ self-contained example:
    args = parser.parse_args()
    state = State(cwd=Path.cwd(), top_dir=args.dir, profraw_prefix=args.profraw_prefix)
-    commands = {
+    handlers = {
        'run': state.do_run,
        'merge': state.do_merge,
        'report': state.do_report,
        'clean': state.do_clean,
    }
-    action = commands.get(args.subparser_name)
+    handler = handlers.get(args.subparser_name)
-    if action:
+    if handler:
-        action(args)
+        handler(args)
    else:
        parser.print_help()
@@ -14,7 +14,7 @@ mkdir -p data/$REPORT_TO
 cp $REPORT_FROM/* data/$REPORT_TO
 echo "Generating report"
-pipenv run python $SCRIPT_DIR/generate_perf_report_page.py --input-dir data/$REPORT_TO --out reports/$REPORT_TO.html 
+poetry run python $SCRIPT_DIR/generate_perf_report_page.py --input-dir data/$REPORT_TO --out reports/$REPORT_TO.html
 echo "Uploading perf result"
 git add data reports
 git \
@@ -0,0 +1,7 @@
 #!/usr/bin/env bash
 # This is a helper script for setting up/updating our python environment.
 # It is intended to be a primary endpoint for all the people who want to
 # just setup test environment without going into details of python package management
 poetry install --no-root # this installs dev dependencies by default
@@ -0,0 +1,9 @@
 #!/usr/bin/env bash
 # This is a helper script to run pytest without going too much
 # into python dependency management details
 # It may be desirable to create more sophisticated pytest launcher
 # with commonly used options to simplify launching from e.g CI
 poetry run pytest "${@:1}"
@@ -22,23 +22,24 @@ runtime. Currently, there are only two batches:
 ### Running the tests
-Because pytest will search all subdirectories for tests, it's easiest to
+There is a wrapper script to invoke pytest: `./scripts/pytest`.
-run the tests from within the `test_runner` directory.
+It accepts all the arguments that are accepted by pytest.
 Depending on your installation options pytest might be invoked directly.
 Test state (postgres data, pageserver state, and log files) will
 be stored under a directory `test_output`.
 You can run all the tests with:
-`pipenv run pytest`
+`./scripts/pytest`
 If you want to run all the tests in a particular file:
-`pipenv run pytest test_pgbench.py`
+`./scripts/pytest test_pgbench.py`
 If you want to run all tests that have the string "bench" in their names:
-`pipenv run pytest -k bench`
+`./scripts/pytest -k bench`
 Useful environment variables:
@@ -48,16 +49,17 @@ Useful environment variables:
 should go.
 `TEST_SHARED_FIXTURES`: Try to re-use a single pageserver for all the tests.
 `ZENITH_PAGESERVER_OVERRIDES`: add a `;`-separated set of configs that will be passed as
 `FORCE_MOCK_S3`: inits every test's pageserver with a mock S3 used as a remote storage.
 `--pageserver-config-override=${value}` parameter values when zenith cli is invoked
 `RUST_LOG`: logging configuration to pass into Zenith CLI
 Let stdout, stderr and `INFO` log messages go to the terminal instead of capturing them:
-`pytest -s --log-cli-level=INFO ...`
+`./scripts/pytest -s --log-cli-level=INFO ...`
 (Note many tests capture subprocess outputs separately, so this may not
 show much.)
 Exit after the first test failure:
-`pytest -x ...`
+`./scripts/pytest -x ...`
 (there are many more pytest options; run `pytest -h` to see them.)
 ### Writing a test
@@ -109,6 +109,10 @@ def test_branch_behind(zenith_env_builder: ZenithEnvBuilder):
    # branch at pre-initdb lsn
    with pytest.raises(Exception, match="invalid branch start lsn"):
        env.zenith_cli(["branch", "test_branch_preinitdb", "main@0/42"])
    # branch at pre-ancestor lsn
    with pytest.raises(Exception, match="less than timeline ancestor lsn"):
        env.zenith_cli(["branch", "test_branch_preinitdb", "test_branch_behind@0/42"])
    # check that we cannot create branch based on garbage collected data
@@ -0,0 +1,84 @@
 from contextlib import closing
 import asyncio
 import asyncpg
 import random
 from fixtures.zenith_fixtures import ZenithEnv, Postgres, Safekeeper
 from fixtures.log_helper import log
 pytest_plugins = ("fixtures.zenith_fixtures")
 # Test configuration
 #
 # Create a table with {num_rows} rows, and perform {updates_to_perform} random
 # UPDATEs on it, using {num_connections} separate connections.
 num_connections = 10
 num_rows = 100000
 updates_to_perform = 10000
 updates_performed = 0
 # Run random UPDATEs on test table
 async def update_table(pg: Postgres):
    global updates_performed
    pg_conn = await pg.connect_async()
    while updates_performed < updates_to_perform:
        updates_performed += 1
        id = random.randrange(1, num_rows)
        row = await pg_conn.fetchrow(f'UPDATE foo SET counter = counter + 1 WHERE id = {id}')
 # Perform aggressive GC with 0 horizon
 async def gc(env: ZenithEnv, timeline: str):
    psconn = await env.pageserver.connect_async()
    while updates_performed < updates_to_perform:
        await psconn.execute(f"do_gc {env.initial_tenant} {timeline} 0")
 # At the same time, run UPDATEs and GC
 async def update_and_gc(env: ZenithEnv, pg: Postgres, timeline: str):
    workers = []
    for worker_id in range(num_connections):
        workers.append(asyncio.create_task(update_table(pg)))
    workers.append(asyncio.create_task(gc(env, timeline)))
    # await all workers
    await asyncio.gather(*workers)
 #
 # Aggressively force GC, while running queries.
 #
 # (repro for https://github.com/zenithdb/zenith/issues/1047)
 #
 def test_gc_aggressive(zenith_simple_env: ZenithEnv):
    env = zenith_simple_env
    # Create a branch for us
    env.zenith_cli(["branch", "test_gc_aggressive", "empty"])
    pg = env.postgres.create_start('test_gc_aggressive')
    log.info('postgres is running on test_gc_aggressive branch')
    conn = pg.connect()
    cur = conn.cursor()
    cur.execute("SHOW zenith.zenith_timeline")
    timeline = cur.fetchone()[0]
    # Create table, and insert the first 100 rows
    cur.execute('CREATE TABLE foo (id int, counter int, t text)')
    cur.execute(f'''
        INSERT INTO foo
            SELECT g, 0, 'long string to consume some space' || g
            FROM generate_series(1, {num_rows}) g
    ''')
    cur.execute('CREATE INDEX ON foo(id)')
    asyncio.run(update_and_gc(env, pg, timeline))
    row = cur.execute('SELECT COUNT(*), SUM(counter) FROM foo')
    assert cur.fetchone() == (num_rows, updates_to_perform)
@@ -0,0 +1,61 @@
 import pytest
 import random
 import time
 from fixtures.zenith_fixtures import ZenithEnvBuilder
 from fixtures.log_helper import log
 pytest_plugins = ("fixtures.zenith_fixtures")
 # Test restarting page server, while safekeeper and compute node keep
 # running.
 def test_next_xid(zenith_env_builder: ZenithEnvBuilder):
    # One safekeeper is enough for this test.
    zenith_env_builder.num_safekeepers = 1
    env = zenith_env_builder.init()
    pg = env.postgres.create_start('main')
    conn = pg.connect()
    cur = conn.cursor()
    cur.execute('CREATE TABLE t(x integer)')
    iterations = 32
    for i in range(1, iterations + 1):
        print(f'iteration {i} / {iterations}')
        # Kill and restart the pageserver.
        pg.stop()
        env.pageserver.stop(immediate=True)
        env.pageserver.start()
        pg.start()
        retry_sleep = 0.5
        max_retries = 200
        retries = 0
        while True:
            try:
                conn = pg.connect()
                cur = conn.cursor()
                cur.execute(f"INSERT INTO t values({i})")
                conn.close()
            except Exception as error:
                # It's normal that it takes some time for the pageserver to
                # restart, and for the connection to fail until it does. It
                # should eventually recover, so retry until it succeeds.
                print(f'failed: {error}')
                if retries < max_retries:
                    retries += 1
                    print(f'retry {retries} / {max_retries}')
                    time.sleep(retry_sleep)
                    continue
                else:
                    raise
            break
    conn = pg.connect()
    cur = conn.cursor()
    cur.execute("SELECT count(*) FROM t")
    assert cur.fetchone() == (iterations, )
@@ -100,7 +100,7 @@ def check_client(client: ZenithPageserverHttpClient, initial_tenant: str):
    timelines = client.timeline_list(tenant_id)
    assert len(timelines) > 0
    for timeline_id_str in timelines:
-        timeline_details = client.timeline_details(tenant_id.hex, timeline_id_str)
+        timeline_details = client.timeline_detail(tenant_id, UUID(timeline_id_str))
        assert timeline_details['type'] == 'Local'
        assert timeline_details['tenant_id'] == tenant_id.hex
        assert timeline_details['timeline_id'] == timeline_id_str
@@ -1,11 +1,13 @@
 # It's possible to run any regular test with the local fs remote storage via
-# env ZENITH_PAGESERVER_OVERRIDES="remote_storage={local_path='/tmp/zenith_zzz/'}" pipenv ......
+# env ZENITH_PAGESERVER_OVERRIDES="remote_storage={local_path='/tmp/zenith_zzz/'}" poetry ......
-import tempfile, time, shutil, os
+import time, shutil, os
 from contextlib import closing
 from pathlib import Path
-from fixtures.zenith_fixtures import ZenithEnvBuilder, LocalFsStorage, check_restored_datadir_content
+from uuid import UUID
 from fixtures.zenith_fixtures import ZenithEnvBuilder
 from fixtures.log_helper import log
 import pytest
 pytest_plugins = ("fixtures.zenith_fixtures")
@@ -25,10 +27,18 @@ pytest_plugins = ("fixtures.zenith_fixtures")
 #   * timeline status is polled until it's downloaded
 #   * queries the specific data, ensuring that it matches the one stored before
 #
-def test_remote_storage_backup_and_restore(zenith_env_builder: ZenithEnvBuilder):
+# The tests are done for all types of remote storage pageserver supports.
@pytest.mark.skip(reason="will be fixed with https://github.com/zenithdb/zenith/issues/1193")
@pytest.mark.parametrize('storage_type', ['local_fs', 'mock_s3'])
 def test_remote_storage_backup_and_restore(zenith_env_builder: ZenithEnvBuilder, storage_type: str):
    zenith_env_builder.rust_log_override = 'debug'
    zenith_env_builder.num_safekeepers = 1
-    zenith_env_builder.enable_local_fs_remote_storage()
+    if storage_type == 'local_fs':
        zenith_env_builder.enable_local_fs_remote_storage()
    elif storage_type == 'mock_s3':
        zenith_env_builder.enable_s3_mock_remote_storage('test_remote_storage_backup_and_restore')
    else:
        raise RuntimeError(f'Unknown storage type: {storage_type}')
    data_id = 1
    data_secret = 'very secret secret'
@@ -65,11 +75,16 @@ def test_remote_storage_backup_and_restore(zenith_env_builder: ZenithEnvBuilder)
    ##### Second start, restore the data and ensure it's the same
    env.pageserver.start()
    log.info("waiting for timeline redownload")
    client = env.pageserver.http_client()
    client.timeline_attach(UUID(tenant_id), UUID(timeline_id))
    # FIXME cannot handle duplicate download requests (which might be caused by repeated timeline detail calls)
    #   subject to fix in https://github.com/zenithdb/zenith/issues/997
    time.sleep(5)
    log.info("waiting for timeline redownload")
    attempts = 0
    while True:
-        timeline_details = client.timeline_details(tenant_id, timeline_id)
+        timeline_details = client.timeline_detail(UUID(tenant_id), UUID(timeline_id))
        assert timeline_details['timeline_id'] == timeline_id
        assert timeline_details['tenant_id'] == tenant_id
        if timeline_details['type'] == 'Local':
@@ -0,0 +1,267 @@
 from contextlib import closing, contextmanager
 import os
 import pathlib
 import subprocess
 import threading
 from uuid import UUID
 from fixtures.log_helper import log
 import time
 import signal
 import pytest
 from fixtures.zenith_fixtures import PgProtocol, PortDistributor, Postgres, ZenithEnvBuilder, ZenithPageserverHttpClient, zenith_binpath, pg_distrib_dir
 def assert_abs_margin_ratio(a: float, b: float, margin_ratio: float):
    assert abs(a - b) / a < margin_ratio, (a, b, margin_ratio)
@contextmanager
 def new_pageserver_helper(new_pageserver_dir: pathlib.Path,
                          pageserver_bin: pathlib.Path,
                          remote_storage_mock_path: pathlib.Path,
                          pg_port: int,
                          http_port: int):
    """
    cannot use ZenithPageserver yet because it depends on zenith cli
    which currently lacks support for multiple pageservers
    """
    cmd = [
        str(pageserver_bin),
        '--init',
        '--workdir',
        str(new_pageserver_dir),
        f"-c listen_pg_addr='localhost:{pg_port}'",
        f"-c listen_http_addr='localhost:{http_port}'",
        f"-c pg_distrib_dir='{pg_distrib_dir}'",
        f"-c remote_storage={{local_path='{remote_storage_mock_path}'}}",
    ]
    subprocess.check_output(cmd, text=True)
    # actually run new pageserver
    cmd = [
        str(pageserver_bin),
        '--workdir',
        str(new_pageserver_dir),
        '--daemonize',
    ]
    log.info("starting new pageserver %s", cmd)
    out = subprocess.check_output(cmd, text=True)
    log.info("started new pageserver %s", out)
    try:
        yield
    finally:
        log.info("stopping new pageserver")
        pid = int((new_pageserver_dir / 'pageserver.pid').read_text())
        os.kill(pid, signal.SIGQUIT)
 def wait_for(number_of_iterations: int, interval: int, func):
    last_exception = None
    for i in range(number_of_iterations):
        try:
            res = func()
        except Exception as e:
            log.info("waiting for %s iteration %s failed", func, i + 1)
            last_exception = e
            time.sleep(interval)
            continue
        return res
    raise Exception("timed out while waiting for %s" % func) from last_exception
@contextmanager
 def pg_cur(pg):
    with closing(pg.connect()) as conn:
        with conn.cursor() as cur:
            yield cur
 def load(pg: Postgres, stop_event: threading.Event, load_ok_event: threading.Event):
    log.info("load started")
    inserted_ctr = 0
    failed = False
    while not stop_event.is_set():
        try:
            with pg_cur(pg) as cur:
                cur.execute("INSERT INTO load VALUES ('some payload')")
                inserted_ctr += 1
        except:
            if not failed:
                log.info("load failed")
            failed = True
            load_ok_event.clear()
        else:
            if failed:
                with pg_cur(pg) as cur:
                    # if we recovered after failure verify that we have correct number of rows
                    log.info("recovering at %s", inserted_ctr)
                    cur.execute("SELECT count(*) FROM load")
                    # it seems that sometimes transaction gets commited before we can acknowledge
                    # the result, so sometimes selected value is larger by one than we expect
                    assert cur.fetchone()[0] - inserted_ctr <= 1
                    log.info("successfully recovered %s", inserted_ctr)
                    failed = False
                    load_ok_event.set()
    log.info('load thread stopped')
 def assert_local(pageserver_http_client: ZenithPageserverHttpClient, tenant: str, timeline: str):
    timeline_detail = pageserver_http_client.timeline_detail(UUID(tenant), UUID(timeline))
    assert timeline_detail.get('type') == "Local", timeline_detail
    return timeline_detail
@pytest.mark.skip(reason="will be fixed with https://github.com/zenithdb/zenith/issues/1193")
@pytest.mark.parametrize('with_load', ['with_load', 'without_load'])
 def test_tenant_relocation(zenith_env_builder: ZenithEnvBuilder,
                           port_distributor: PortDistributor,
                           with_load: str):
    zenith_env_builder.num_safekeepers = 1
    zenith_env_builder.enable_local_fs_remote_storage()
    env = zenith_env_builder.init()
    # create folder for remote storage mock
    remote_storage_mock_path = env.repo_dir / 'local_fs_remote_storage'
    tenant = env.create_tenant("74ee8b079a0e437eb0afea7d26a07209")
    log.info("tenant to relocate %s", tenant)
    env.zenith_cli(["branch", "test_tenant_relocation", "main", f"--tenantid={tenant}"])
    tenant_pg = env.postgres.create_start(
        "test_tenant_relocation",
        "main",  # branch name, None means same as node name
        tenant_id=tenant,
    )
    # insert some data
    with closing(tenant_pg.connect()) as conn:
        with conn.cursor() as cur:
            # save timeline for later gc call
            cur.execute("SHOW zenith.zenith_timeline")
            timeline = cur.fetchone()[0]
            log.info("timeline to relocate %s", timeline)
            # we rely upon autocommit after each statement
            # as waiting for acceptors happens there
            cur.execute("CREATE TABLE t(key int primary key, value text)")
            cur.execute("INSERT INTO t SELECT generate_series(1,1000), 'some payload'")
            cur.execute("SELECT sum(key) FROM t")
            assert cur.fetchone() == (500500, )
    if with_load == 'with_load':
        # create load table
        with pg_cur(tenant_pg) as cur:
            cur.execute("CREATE TABLE load(value text)")
        load_stop_event = threading.Event()
        load_ok_event = threading.Event()
        load_thread = threading.Thread(target=load,
                                       args=(tenant_pg, load_stop_event, load_ok_event))
        load_thread.start()
    # run checkpoint manually to be sure that data landed in remote storage
    with closing(env.pageserver.connect()) as psconn:
        with psconn.cursor() as pscur:
            pscur.execute(f"do_gc {tenant} {timeline}")
    # ensure upload is completed
    pageserver_http_client = env.pageserver.http_client()
    timeline_detail = pageserver_http_client.timeline_detail(UUID(tenant), UUID(timeline))
    assert timeline_detail['disk_consistent_lsn'] == timeline_detail['timeline_state']['Ready']
    log.info("inititalizing new pageserver")
    # bootstrap second pageserver
    new_pageserver_dir = env.repo_dir / 'new_pageserver'
    new_pageserver_dir.mkdir()
    new_pageserver_pg_port = port_distributor.get_port()
    new_pageserver_http_port = port_distributor.get_port()
    log.info("new pageserver ports pg %s http %s", new_pageserver_pg_port, new_pageserver_http_port)
    pageserver_bin = pathlib.Path(zenith_binpath) / 'pageserver'
    new_pageserver_http_client = ZenithPageserverHttpClient(port=new_pageserver_http_port,
                                                            auth_token=None)
    with new_pageserver_helper(new_pageserver_dir,
                               pageserver_bin,
                               remote_storage_mock_path,
                               new_pageserver_pg_port,
                               new_pageserver_http_port):
        # call to attach timeline to new pageserver
        new_pageserver_http_client.timeline_attach(UUID(tenant), UUID(timeline))
        # FIXME cannot handle duplicate download requests, subject to fix in https://github.com/zenithdb/zenith/issues/997
        time.sleep(5)
        # new pageserver should in sync (modulo wal tail or vacuum activity) with the old one because there was no new writes since checkpoint
        new_timeline_detail = wait_for(
            number_of_iterations=5,
            interval=1,
            func=lambda: assert_local(new_pageserver_http_client, tenant, timeline))
        assert new_timeline_detail['timeline_state'].get('Ready'), new_timeline_detail
        # when load is active these checks can break because lsns are not static
        # so lets check with some margin
        if with_load == 'without_load':
            # TODO revisit this once https://github.com/zenithdb/zenith/issues/1049 is fixed
            assert_abs_margin_ratio(new_timeline_detail['disk_consistent_lsn'],
                                    timeline_detail['disk_consistent_lsn'],
                                    0.01)
            assert_abs_margin_ratio(new_timeline_detail['timeline_state']['Ready'],
                                    timeline_detail['timeline_state']['Ready'],
                                    0.01)
        # callmemaybe to start replication from safekeeper to the new pageserver
        # when there is no load there is a clean checkpoint and no wal delta
        # needs to be streamed to the new pageserver
        # TODO (rodionov) use attach to start replication
        with pg_cur(PgProtocol(host='localhost', port=new_pageserver_pg_port)) as cur:
            # "callmemaybe {} {} host={} port={} options='-c ztimelineid={} ztenantid={}'"
            safekeeper_connstring = f"host=localhost port={env.safekeepers[0].port.pg} options='-c ztimelineid={timeline} ztenantid={tenant} pageserver_connstr=postgresql://no_user:@localhost:{new_pageserver_pg_port}'"
            cur.execute("callmemaybe {} {} {}".format(tenant, timeline, safekeeper_connstring))
        tenant_pg.stop()
        # rewrite zenith cli config to use new pageserver for basebackup to start new compute
        cli_config_lines = (env.repo_dir / 'config').read_text().splitlines()
        cli_config_lines[-2] = f"listen_http_addr = 'localhost:{new_pageserver_http_port}'"
        cli_config_lines[-1] = f"listen_pg_addr = 'localhost:{new_pageserver_pg_port}'"
        (env.repo_dir / 'config').write_text('\n'.join(cli_config_lines))
        tenant_pg_config_file_path = pathlib.Path(tenant_pg.config_file_path())
        tenant_pg_config_file_path.open('a').write(
            f"\nzenith.page_server_connstring = 'postgresql://no_user:@localhost:{new_pageserver_pg_port}'"
        )
        tenant_pg.start()
        # detach tenant from old pageserver before we check
        # that all the data is there to be sure that old pageserver
        # is no longer involved, and if it is, we will see the errors
        pageserver_http_client.timeline_detach(UUID(tenant), UUID(timeline))
        with pg_cur(tenant_pg) as cur:
            # check that data is still there
            cur.execute("SELECT sum(key) FROM t")
            assert cur.fetchone() == (500500, )
            # check that we can write new data
            cur.execute("INSERT INTO t SELECT generate_series(1001,2000), 'some payload'")
            cur.execute("SELECT sum(key) FROM t")
            assert cur.fetchone() == (2001000, )
        if with_load == 'with_load':
            assert load_ok_event.wait(1)
            log.info('stopping load thread')
            load_stop_event.set()
            load_thread.join()
            log.info('load thread stopped')
        # bring old pageserver back for clean shutdown via zenith cli
        # new pageserver will be shut down by the context manager
        cli_config_lines = (env.repo_dir / 'config').read_text().splitlines()
        cli_config_lines[-2] = f"listen_http_addr = 'localhost:{env.pageserver.service_port.http}'"
        cli_config_lines[-1] = f"listen_pg_addr = 'localhost:{env.pageserver.service_port.pg}'"
        (env.repo_dir / 'config').write_text('\n'.join(cli_config_lines))
@@ -80,8 +80,8 @@ def test_twophase(zenith_simple_env: ZenithEnv):
    cur2.execute("ROLLBACK PREPARED 'insert_two'")
    cur2.execute('SELECT * FROM foo')
-    assert cur2.fetchall() == [('one', ), ('three', )]
+    assert cur2.fetchall() == [('one', ), ('three', )]  # type: ignore[comparison-overlap]
    # Only one committed insert is visible on the original branch
    cur.execute('SELECT * FROM foo')
-    assert cur.fetchall() == [('three', )]
+    assert cur.fetchall() == [('three', )]  # type: ignore[comparison-overlap]
@@ -12,7 +12,7 @@ from contextlib import closing
 from dataclasses import dataclass, field
 from multiprocessing import Process, Value
 from pathlib import Path
-from fixtures.zenith_fixtures import PgBin, ZenithEnvBuilder, PortDistributor, SafekeeperPort, zenith_binpath, PgProtocol
+from fixtures.zenith_fixtures import PgBin, Postgres, Safekeeper, ZenithEnv, ZenithEnvBuilder, PortDistributor, SafekeeperPort, zenith_binpath, PgProtocol
 from fixtures.utils import lsn_to_hex, mkdir_if_needed
 from fixtures.log_helper import log
 from typing import List, Optional, Any
@@ -603,3 +603,92 @@ def test_safekeeper_without_pageserver(test_output_dir: str,
        env.postgres.safe_psql("insert into t select generate_series(1, 100)")
        res = env.postgres.safe_psql("select sum(i) from t")[0][0]
        assert res == 5050
 def test_replace_safekeeper(zenith_env_builder: ZenithEnvBuilder):
    def safekeepers_guc(env: ZenithEnv, sk_names: List[str]) -> str:
        return ','.join(
            [f'localhost:{sk.port.pg}' for sk in env.safekeepers if sk.name in sk_names])
    def execute_payload(pg: Postgres):
        with closing(pg.connect()) as conn:
            with conn.cursor() as cur:
                # we rely upon autocommit after each statement
                # as waiting for acceptors happens there
                cur.execute('CREATE TABLE IF NOT EXISTS t(key int, value text)')
                cur.execute("INSERT INTO t VALUES (0, 'something')")
                cur.execute('SELECT SUM(key) FROM t')
                sum_before = cur.fetchone()[0]
                cur.execute("INSERT INTO t SELECT generate_series(1,100000), 'payload'")
                cur.execute('SELECT SUM(key) FROM t')
                sum_after = cur.fetchone()[0]
                assert sum_after == sum_before + 5000050000
    def show_statuses(safekeepers: List[Safekeeper], tenant_id: str, timeline_id: str):
        for sk in safekeepers:
            http_cli = sk.http_client()
            try:
                status = http_cli.timeline_status(tenant_id, timeline_id)
                log.info(f"Safekeeper {sk.name} status: {status}")
            except Exception as e:
                log.info(f"Safekeeper {sk.name} status error: {e}")
    zenith_env_builder.num_safekeepers = 4
    env = zenith_env_builder.init()
    env.zenith_cli(["branch", "test_replace_safekeeper", "main"])
    log.info("Use only first 3 safekeepers")
    env.safekeepers[3].stop()
    active_safekeepers = ['sk1', 'sk2', 'sk3']
    pg = env.postgres.create('test_replace_safekeeper')
    pg.adjust_for_wal_acceptors(safekeepers_guc(env, active_safekeepers))
    pg.start()
    # learn zenith timeline from compute
    tenant_id = pg.safe_psql("show zenith.zenith_tenant")[0][0]
    timeline_id = pg.safe_psql("show zenith.zenith_timeline")[0][0]
    execute_payload(pg)
    show_statuses(env.safekeepers, tenant_id, timeline_id)
    log.info("Restart all safekeepers to flush everything")
    env.safekeepers[0].stop(immediate=True)
    execute_payload(pg)
    env.safekeepers[0].start()
    env.safekeepers[1].stop(immediate=True)
    execute_payload(pg)
    env.safekeepers[1].start()
    env.safekeepers[2].stop(immediate=True)
    execute_payload(pg)
    env.safekeepers[2].start()
    env.safekeepers[0].stop(immediate=True)
    env.safekeepers[1].stop(immediate=True)
    env.safekeepers[2].stop(immediate=True)
    env.safekeepers[0].start()
    env.safekeepers[1].start()
    env.safekeepers[2].start()
    execute_payload(pg)
    show_statuses(env.safekeepers, tenant_id, timeline_id)
    log.info("Stop sk1 (simulate failure) and use only quorum of sk2 and sk3")
    env.safekeepers[0].stop(immediate=True)
    execute_payload(pg)
    show_statuses(env.safekeepers, tenant_id, timeline_id)
    log.info("Recreate postgres to replace failed sk1 with new sk4")
    pg.stop_and_destroy().create('test_replace_safekeeper')
    active_safekeepers = ['sk2', 'sk3', 'sk4']
    env.safekeepers[3].start()
    pg.adjust_for_wal_acceptors(safekeepers_guc(env, active_safekeepers))
    pg.start()
    execute_payload(pg)
    show_statuses(env.safekeepers, tenant_id, timeline_id)
    log.info("Stop sk2 to require quorum of sk3 and sk4 for normal work")
    env.safekeepers[1].stop(immediate=True)
    execute_payload(pg)
    show_statuses(env.safekeepers, tenant_id, timeline_id)
@@ -3,11 +3,11 @@ import logging.config
 """
 This file configures logging to use in python tests.
 Logs are automatically captured and shown in their
-own section after all tests are executed. 
+own section after all tests are executed.
 To see logs for all (even successful) tests, run
 pytest with the following command:
- `pipenv run pytest -n8 -rA`
+- `poetry run pytest -n8 -rA`
 Other log config can be set in pytest.ini file.
 You can add `log_cli = true` to it to watch
@@ -34,7 +34,7 @@ LOGGING = {
 def getLogger(name='root') -> logging.Logger:
    """Method to get logger for tests.
-    
+
    Should be used to get correctly initialized logger. """
    return logging.getLogger(name)
@@ -4,6 +4,7 @@ from dataclasses import dataclass, field
 from cached_property import cached_property
 import asyncpg
 import os
 import boto3
 import pathlib
 import uuid
 import warnings
@@ -330,6 +331,48 @@ class AuthKeys:
        return token
 class MockS3Server:
    """
    Starts a mock S3 server for testing on a port given, errors if the server fails to start or exits prematurely.
    Relies that `poetry` and `moto` server are installed, since it's the way the tests are run.
    Also provides a set of methods to derive the connection properties from and the method to kill the underlying server.
    """
    def __init__(
        self,
        port: int,
    ):
        self.port = port
        self.subprocess = subprocess.Popen([f'poetry run moto_server s3 -p{port}'], shell=True)
        error = None
        try:
            return_code = self.subprocess.poll()
            if return_code is not None:
                error = f"expected mock s3 server to run but it exited with code {return_code}. stdout: '{self.subprocess.stdout}', stderr: '{self.subprocess.stderr}'"
        except Exception as e:
            error = f"expected mock s3 server to start but it failed with exception: {e}. stdout: '{self.subprocess.stdout}', stderr: '{self.subprocess.stderr}'"
        if error is not None:
            log.error(error)
            self.subprocess.kill()
            raise RuntimeError("failed to start s3 mock server")
    def endpoint(self) -> str:
        return f"http://127.0.0.1:{self.port}"
    def region(self) -> str:
        return 'us-east-1'
    def access_key(self) -> str:
        return 'test'
    def secret_key(self) -> str:
        return 'test'
    def kill(self):
        self.subprocess.kill()
 class ZenithEnvBuilder:
    """
    Builder object to create a Zenith runtime environment
@@ -354,17 +397,57 @@ class ZenithEnvBuilder:
        self.pageserver_auth_enabled = pageserver_auth_enabled
        self.env: Optional[ZenithEnv] = None
        self.s3_mock_server: Optional[MockS3Server] = None
        if os.getenv('FORCE_MOCK_S3') is not None:
            bucket_name = f'{repo_dir.name}_bucket'
            log.warning(f'Unconditionally initializing mock S3 server for bucket {bucket_name}')
            self.enable_s3_mock_remote_storage(bucket_name)
    def init(self) -> ZenithEnv:
        # Cannot create more than one environment from one builder
        assert self.env is None, "environment already initialized"
        self.env = ZenithEnv(self)
        return self.env
-    def enable_local_fs_remote_storage(self):
+    """
-        assert self.pageserver_remote_storage is None, "remote storage is enabled already"
+    Sets up the pageserver to use the local fs at the `test_dir/local_fs_remote_storage` path.
    Errors, if the pageserver has some remote storage configuration already, unless `force_enable` is not set to `True`.
    """
    def enable_local_fs_remote_storage(self, force_enable=True):
        assert force_enable or self.pageserver_remote_storage is None, "remote storage is enabled already"
        self.pageserver_remote_storage = LocalFsStorage(
            Path(self.repo_dir / 'local_fs_remote_storage'))
    """
    Sets up the pageserver to use the S3 mock server, creates the bucket, if it's not present already.
    Starts up the mock server, if that does not run yet.
    Errors, if the pageserver has some remote storage configuration already, unless `force_enable` is not set to `True`.
    """
    def enable_s3_mock_remote_storage(self, bucket_name: str, force_enable=True):
        assert force_enable or self.pageserver_remote_storage is None, "remote storage is enabled already"
        if not self.s3_mock_server:
            self.s3_mock_server = MockS3Server(self.port_distributor.get_port())
        mock_endpoint = self.s3_mock_server.endpoint()
        mock_region = self.s3_mock_server.region()
        mock_access_key = self.s3_mock_server.access_key()
        mock_secret_key = self.s3_mock_server.secret_key()
        boto3.client(
            's3',
            endpoint_url=mock_endpoint,
            region_name=mock_region,
            aws_access_key_id=mock_access_key,
            aws_secret_access_key=mock_secret_key,
        ).create_bucket(Bucket=bucket_name)
        self.pageserver_remote_storage = S3Storage(bucket=bucket_name,
                                                   endpoint=mock_endpoint,
                                                   region=mock_region,
                                                   access_key=mock_access_key,
                                                   secret_key=mock_secret_key)
    def __enter__(self):
        return self
@@ -377,6 +460,8 @@ class ZenithEnvBuilder:
            for sk in self.env.safekeepers:
                sk.stop(immediate=True)
            self.env.pageserver.stop(immediate=True)
            if self.s3_mock_server:
                self.s3_mock_server.kill()
 class ZenithEnv:
@@ -415,6 +500,7 @@ class ZenithEnv:
        self.repo_dir = config.repo_dir
        self.rust_log_override = config.rust_log_override
        self.port_distributor = config.port_distributor
        self.s3_mock_server = config.s3_mock_server
        self.postgres = PostgresFactory(self)
@@ -600,6 +686,8 @@ def zenith_simple_env(_shared_simple_env: ZenithEnv) -> Iterator[ZenithEnv]:
    yield _shared_simple_env
    _shared_simple_env.postgres.stop_all()
    if _shared_simple_env.s3_mock_server:
        _shared_simple_env.s3_mock_server.kill()
@pytest.fixture(scope='function')
@@ -637,6 +725,16 @@ class ZenithPageserverHttpClient(requests.Session):
    def check_status(self):
        self.get(f"http://localhost:{self.port}/v1/status").raise_for_status()
    def timeline_attach(self, tenant_id: uuid.UUID, timeline_id: uuid.UUID):
        res = self.post(
            f"http://localhost:{self.port}/v1/timeline/{tenant_id.hex}/{timeline_id.hex}/attach", )
        res.raise_for_status()
    def timeline_detach(self, tenant_id: uuid.UUID, timeline_id: uuid.UUID):
        res = self.post(
            f"http://localhost:{self.port}/v1/timeline/{tenant_id.hex}/{timeline_id.hex}/detach", )
        res.raise_for_status()
    def branch_list(self, tenant_id: uuid.UUID) -> List[Dict[Any, Any]]:
        res = self.get(f"http://localhost:{self.port}/v1/branch/{tenant_id.hex}")
        res.raise_for_status()
@@ -689,8 +787,9 @@ class ZenithPageserverHttpClient(requests.Session):
        assert isinstance(res_json, list)
        return res_json
-    def timeline_details(self, tenant_id: str, timeline_id: str) -> Dict[Any, Any]:
+    def timeline_detail(self, tenant_id: uuid.UUID, timeline_id: uuid.UUID):
-        res = self.get(f"http://localhost:{self.port}/v1/timeline/{tenant_id}/{timeline_id}")
+        res = self.get(
            f"http://localhost:{self.port}/v1/timeline/{tenant_id.hex}/{timeline_id.hex}")
        res.raise_for_status()
        res_json = res.json()
        assert isinstance(res_json, dict)
@@ -717,8 +816,9 @@ class LocalFsStorage:
 class S3Storage:
    bucket: str
    region: str
-    access_key: str
+    access_key: Optional[str]
-    secret_key: str
+    secret_key: Optional[str]
    endpoint: Optional[str]
 RemoteStorage = Union[LocalFsStorage, S3Storage]
@@ -791,8 +891,14 @@ def append_pageserver_param_overrides(params_to_update: List[str],
            pageserver_storage_override = f"local_path='{pageserver_remote_storage.root}'"
        elif isinstance(pageserver_remote_storage, S3Storage):
            pageserver_storage_override = f"bucket_name='{pageserver_remote_storage.bucket}',\
-                bucket_region='{pageserver_remote_storage.region}',access_key_id='{pageserver_remote_storage.access_key}',\
+                bucket_region='{pageserver_remote_storage.region}'"
-                secret_access_key='{pageserver_remote_storage.secret_key}'"
+
            if pageserver_remote_storage.access_key is not None:
                pageserver_storage_override += f",access_key_id='{pageserver_remote_storage.access_key}'"
            if pageserver_remote_storage.secret_key is not None:
                pageserver_storage_override += f",secret_access_key='{pageserver_remote_storage.secret_key}'"
            if pageserver_remote_storage.endpoint is not None:
                pageserver_storage_override += f",endpoint='{pageserver_remote_storage.endpoint}'"
        else:
            raise Exception(f'Unknown storage configuration {pageserver_remote_storage}')
@@ -908,8 +1014,6 @@ class Postgres(PgProtocol):
        path = pathlib.Path('pgdatadirs') / 'tenants' / self.tenant_id / self.node_name
        self.pgdata_dir = os.path.join(self.env.repo_dir, path)
        if self.env.safekeepers:
            self.adjust_for_wal_acceptors(self.env.get_safekeeper_connstrs())
        if config_lines is None:
            config_lines = []
        self.config(config_lines)
@@ -966,7 +1070,9 @@ class Postgres(PgProtocol):
                # walproposer uses different application_name
                if ("synchronous_standby_names" in cfg_line or
                        # don't ask pageserver to fetch WAL from compute
-                        "callmemaybe_connstring" in cfg_line):
+                        "callmemaybe_connstring" in cfg_line or
                        # don't repeat wal_acceptors multiple times
                        "wal_acceptors" in cfg_line):
                    continue
                f.write(cfg_line)
            f.write("synchronous_standby_names = 'walproposer'\n")
@@ -2,7 +2,7 @@
 name = "walkeeper"
 version = "0.1.0"
 authors = ["Stas Kelvich <stas@zenith.tech>"]
-edition = "2018"
+edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
@@ -19,9 +19,9 @@ tracing = "0.1.27"
 clap = "2.33.0"
 daemonize = "0.4.1"
 rust-s3 = { version = "0.28", default-features = false, features = ["no-verify-ssl", "tokio-rustls-tls"] }
-tokio = "1.11"
+tokio = { version = "1.11", features = ["macros"] }
-postgres-protocol = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+postgres-protocol = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
 anyhow = "1.0"
 crc32c = "0.6.0"
 humantime = "2.1.0"
@@ -30,7 +30,7 @@ signal-hook = "0.3.10"
 serde = { version = "1.0", features = ["derive"] }
 hex = "0.4.3"
 const_format = "0.2.21"
-tokio-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+tokio-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
 postgres_ffi = { path = "../postgres_ffi" }
 workspace_hack = { path = "../workspace_hack" }
@@ -99,10 +99,7 @@ A: If the compute node has evicted a page, changes to it have been WAL-logged
 Q: How long may Page Server wait for?
 A: Not too long, hopefully. If a page is evicted, it probably was not used for
   a while, so the WAL service have had enough time to push changes to the Page
-   Server. There may be issues if there is no backpressure and compute node with
+   Server. To limit the lag, tune backpressure using `max_replication_*_lag` settings.
   WAL service run ahead of Page Server, though.
   There is no backpressure right now, so you may even see some spurious
   timeouts in tests.
 Q: How do WAL safekeepers communicate with each other?
 A: They may only send each other messages via the compute node, they never
@@ -143,7 +143,7 @@ fn start_safekeeper(conf: SafeKeeperConf) -> Result<()> {
    // Prevent running multiple safekeepers on the same directory
    let lock_file_path = conf.workdir.join(LOCK_FILE_NAME);
-    let lock_file = File::create(&lock_file_path).with_context(|| "failed to open lockfile")?;
+    let lock_file = File::create(&lock_file_path).context("failed to open lockfile")?;
    lock_file.try_lock_exclusive().with_context(|| {
        format!(
            "control file {} is locked by some other process",
@@ -197,7 +197,12 @@ fn start_safekeeper(conf: SafeKeeperConf) -> Result<()> {
            .spawn(|| {
                // TODO authentication
                let router = http::make_router(conf_);
-                endpoint::serve_thread_main(router, http_listener).unwrap();
+                endpoint::serve_thread_main(
                    router,
                    http_listener,
                    std::future::pending(), // never shut down
                )
                .unwrap();
            })?,
    );
@@ -6,8 +6,8 @@
 //!  from the call list.
 //!
 use crate::SafeKeeperConf;
-use anyhow::anyhow;
+use anyhow::{Context, Result};
-use anyhow::Result;
+use std::collections::hash_map::Entry;
 use std::collections::HashMap;
 use std::sync::Mutex;
 use std::time::{Duration, Instant};
@@ -42,9 +42,12 @@ async fn request_callback(
    let me_conf: postgres::config::Config = me_connstr.parse().unwrap();
    let (host, port) = connection_host_port(&me_conf);
    // pageserver connstr is needed to be able to distinguish between different pageservers
    // it is required to correctly manage callmemaybe subscriptions when more than one pageserver is involved
    // TODO it is better to use some sort of a unique id instead of connection string, see https://github.com/zenithdb/zenith/issues/1105
    let callme = format!(
-        "callmemaybe {} {} host={} port={} options='-c ztimelineid={} ztenantid={}'",
+        "callmemaybe {} {} host={} port={} options='-c ztimelineid={} ztenantid={} pageserver_connstr={}'",
-        tenantid, timelineid, host, port, timelineid, tenantid
+        tenantid, timelineid, host, port, timelineid, tenantid, pageserver_connstr,
    );
    let _ = client.simple_query(&callme).await?;
@@ -61,18 +64,36 @@ pub fn thread_main(conf: SafeKeeperConf, rx: UnboundedReceiver<CallmeEvent>) ->
    runtime.block_on(main_loop(conf, rx))
 }
 #[derive(Debug, PartialEq, Eq, Hash, Clone)]
 pub struct SubscriptionStateKey {
    tenant_id: ZTenantId,
    timeline_id: ZTimelineId,
    pageserver_connstr: String,
 }
 impl SubscriptionStateKey {
    pub fn new(tenant_id: ZTenantId, timeline_id: ZTimelineId, pageserver_connstr: String) -> Self {
        Self {
            tenant_id,
            timeline_id,
            pageserver_connstr,
        }
    }
 }
 /// Messages to the callmemaybe thread
 #[derive(Debug)]
 pub enum CallmeEvent {
    // add new subscription to the list
-    Subscribe(ZTenantId, ZTimelineId, String),
+    Subscribe(SubscriptionStateKey),
    // remove the subscription from the list
-    Unsubscribe(ZTenantId, ZTimelineId),
+    Unsubscribe(SubscriptionStateKey),
    // don't serve this subscription, but keep it in the list
-    Pause(ZTenantId, ZTimelineId),
+    Pause(SubscriptionStateKey),
    // resume this subscription, if it exists,
    // but don't create a new one if it is gone
-    Resume(ZTenantId, ZTimelineId),
+    Resume(SubscriptionStateKey),
    // TODO how do we delete from subscriptions?
 }
 #[derive(Debug)]
@@ -118,6 +139,7 @@ impl SubscriptionState {
            let timelineid = self.timelineid;
            let tenantid = self.tenantid;
            let pageserver_connstr = self.pageserver_connstr.clone();
            tokio::spawn(async move {
                if let Err(err) = handle.await {
                    if err.is_cancelled() {
@@ -125,8 +147,8 @@ impl SubscriptionState {
                            timelineid, tenantid);
                    } else {
                        error!(
-                            "callback task for timelineid={} tenantid={} failed: {}",
+                            "callback task for timelineid={} tenantid={} pageserver_connstr={} failed: {}",
-                            timelineid, tenantid, err
+                            timelineid, tenantid, pageserver_connstr, err
                        );
                    }
                }
@@ -138,7 +160,7 @@ impl SubscriptionState {
        // Ignore call request if this subscription is paused
        if self.paused {
            debug!(
-                "ignore call request for paused subscription
+                "ignore call request for paused subscription \
                tenantid: {}, timelineid: {}",
                self.tenantid, self.timelineid
            );
@@ -148,7 +170,7 @@ impl SubscriptionState {
        // Check if it too early to recall
        if self.handle.is_some() && self.last_call_time.elapsed() < recall_period {
            debug!(
-                "too early to recall. self.last_call_time.elapsed: {:?}, recall_period: {:?}
+                "too early to recall. self.last_call_time.elapsed: {:?}, recall_period: {:?} \
                tenantid: {}, timelineid: {}",
                self.last_call_time, recall_period, self.tenantid, self.timelineid
            );
@@ -176,8 +198,7 @@ impl SubscriptionState {
        // Update last_call_time
        self.last_call_time = Instant::now();
        info!(
-            "new call spawned. time {:?}
+            "new call spawned. last call time {:?} tenantid: {}, timelineid: {}",
            tenantid: {}, timelineid: {}",
            self.last_call_time, self.tenantid, self.timelineid
        );
    }
@@ -190,7 +211,7 @@ impl Drop for SubscriptionState {
 }
 pub async fn main_loop(conf: SafeKeeperConf, mut rx: UnboundedReceiver<CallmeEvent>) -> Result<()> {
-    let subscriptions: Mutex<HashMap<(ZTenantId, ZTimelineId), SubscriptionState>> =
+    let subscriptions: Mutex<HashMap<SubscriptionStateKey, SubscriptionState>> =
        Mutex::new(HashMap::new());
    let mut ticker = tokio::time::interval(conf.recall_period);
@@ -198,54 +219,82 @@ pub async fn main_loop(conf: SafeKeeperConf, mut rx: UnboundedReceiver<CallmeEve
        tokio::select! {
            request = rx.recv() =>
            {
-                match request.ok_or_else(|| anyhow!("done"))?
+                match request.context("done")?
                {
-                    CallmeEvent::Subscribe(tenantid, timelineid, pageserver_connstr) =>
+                    CallmeEvent::Subscribe(key) =>
                    {
                        let _enter = info_span!("callmemaybe: subscribe", timelineid = %key.timeline_id, tenantid = %key.tenant_id, pageserver_connstr=%key.pageserver_connstr.clone()).entered();
                        let mut subscriptions = subscriptions.lock().unwrap();
-                        if let Some(sub) = subscriptions.get(&(tenantid, timelineid))
+                        // XXX this clone is ugly, is there a way to use the trick with Borrow trait with entry API?
-                        {
+                        //  when we switch to node id instead of the connection string key will be Copy and there will be no need to clone
-                            info!("callmemaybe. subscription already exists {:?}", sub);
+                        match subscriptions.entry(key.clone()) {
                            Entry::Occupied(_) => {
                                // Do nothing if subscription already exists
                                // If it is paused it means that there is already established replication connection.
                                // If it is not paused it will be polled with other subscriptions when timeout expires.
                                // This can occur when replication channel is established before subscription is added.
                                info!(
                                    "subscription already exists",
                                );
                            }
                            Entry::Vacant(entry) => {
                                let subscription = entry.insert(SubscriptionState::new(
                                    key.tenant_id,
                                    key.timeline_id,
                                    key.pageserver_connstr,
                                ));
                                subscription.call(conf.recall_period, conf.listen_pg_addr.clone());
                            }
                        }
-                        if let Some(mut sub) = subscriptions.insert((tenantid, timelineid),
+                    },
-                            SubscriptionState::new(tenantid, timelineid, pageserver_connstr))
+                    CallmeEvent::Unsubscribe(key) => {
-                        {
+                        let _enter = debug_span!("callmemaybe: unsubscribe", timelineid = %key.timeline_id, tenantid = %key.tenant_id, pageserver_connstr=%key.pageserver_connstr.clone()).entered();
-                            sub.call(conf.recall_period, conf.listen_pg_addr.clone());
+                        debug!("unsubscribe");
                        let mut subscriptions = subscriptions.lock().unwrap();
                        subscriptions.remove(&key);
                    },
                    CallmeEvent::Pause(key) => {
                        let _enter = debug_span!("callmemaybe: pause", timelineid = %key.timeline_id, tenantid = %key.tenant_id, pageserver_connstr=%key.pageserver_connstr.clone()).entered();
                        let mut subscriptions = subscriptions.lock().unwrap();
                        // If pause received when no corresponding subscription exists it means that someone started replication
                        // without using callmemaybe. So we create subscription and pause it.
                        // In tenant relocation scenario subscribe call will be executed after pause when compute is restarted.
                        // In that case there is no need to create new/unpause existing subscription.
                        match subscriptions.entry(key.clone()) {
                            Entry::Occupied(mut sub) => {
                                debug!("pause existing");
                                sub.get_mut().pause();
                            }
                            Entry::Vacant(entry) => {
                                debug!("create paused");
                                let subscription = entry.insert(SubscriptionState::new(
                                    key.tenant_id,
                                    key.timeline_id,
                                    key.pageserver_connstr,
                                ));
                                subscription.pause();
                            }
                        }
                        info!("callmemaybe. thread_main. subscribe callback request for timelineid={} tenantid={}",
                        timelineid, tenantid);
                    },
-                    CallmeEvent::Unsubscribe(tenantid, timelineid) => {
+                    CallmeEvent::Resume(key) => {
                        debug!(
                            "callmemaybe. thread_main. resume callback request for timelineid={} tenantid={} pageserver_connstr={}",
                            key.timeline_id, key.tenant_id, key.pageserver_connstr,
                        );
                        let mut subscriptions = subscriptions.lock().unwrap();
-                        subscriptions.remove(&(tenantid, timelineid));
+                        if let Some(sub) = subscriptions.get_mut(&key)
                        info!("callmemaybe. thread_main. unsubscribe callback. request for timelineid={} tenantid={}",
                        timelineid, tenantid);
                    },
                    CallmeEvent::Pause(tenantid, timelineid) => {
                        let mut subscriptions = subscriptions.lock().unwrap();
                        if let Some(sub) = subscriptions.get_mut(&(tenantid, timelineid))
                        {
                            sub.pause();
                        };
                        info!("callmemaybe. thread_main. pause callback request for timelineid={} tenantid={}",
                        timelineid, tenantid);
                    },
                    CallmeEvent::Resume(tenantid, timelineid) => {
                        let mut subscriptions = subscriptions.lock().unwrap();
                        if let Some(sub) = subscriptions.get_mut(&(tenantid, timelineid))
                        {
                            sub.resume();
                        };
                        info!("callmemaybe. thread_main. resume callback request for timelineid={} tenantid={}",
                        timelineid, tenantid);
                    },
                }
            },
            _ = ticker.tick() => {
                let _enter = debug_span!("callmemaybe: tick").entered();
                let mut subscriptions = subscriptions.lock().unwrap();
-                for (&(_tenantid, _timelineid), state) in subscriptions.iter_mut() {
+                for (_, state) in subscriptions.iter_mut() {
                    state.call(conf.recall_period, conf.listen_pg_addr.clone());
                }
             },
@@ -6,7 +6,7 @@ use crate::receive_wal::ReceiveWalConn;
 use crate::send_wal::ReplicationConn;
 use crate::timeline::{Timeline, TimelineTools};
 use crate::SafeKeeperConf;
-use anyhow::{anyhow, bail, Context, Result};
+use anyhow::{bail, Context, Result};
 use postgres_ffi::xlog_utils::PG_TLI;
 use regex::Regex;
@@ -16,7 +16,7 @@ use zenith_utils::lsn::Lsn;
 use zenith_utils::postgres_backend;
 use zenith_utils::postgres_backend::PostgresBackend;
 use zenith_utils::pq_proto::{BeMessage, FeStartupPacket, RowDescriptor, INT4_OID, TEXT_OID};
-use zenith_utils::zid::{ZTenantId, ZTimelineId};
+use zenith_utils::zid::{ZTenantId, ZTenantTimelineId, ZTimelineId};
 use crate::callmemaybe::CallmeEvent;
 use crate::timeline::CreateControlFile;
@@ -30,6 +30,7 @@ pub struct SafekeeperPostgresHandler {
    pub ztenantid: Option<ZTenantId>,
    pub ztimelineid: Option<ZTimelineId>,
    pub timeline: Option<Arc<Timeline>>,
    pageserver_connstr: Option<String>,
    //sender to communicate with callmemaybe thread
    pub tx: UnboundedSender<CallmeEvent>,
 }
@@ -56,16 +57,15 @@ fn parse_cmd(cmd: &str) -> Result<SafekeeperPostgresCommand> {
        let start_lsn = caps
            .next()
            .map(|cap| cap[1].parse::<Lsn>())
-            .ok_or_else(|| anyhow!("failed to parse start LSN from START_REPLICATION command"))??;
+            .context("failed to parse start LSN from START_REPLICATION command")??;
        Ok(SafekeeperPostgresCommand::StartReplication { start_lsn })
    } else if cmd.starts_with("IDENTIFY_SYSTEM") {
        Ok(SafekeeperPostgresCommand::IdentifySystem)
    } else if cmd.starts_with("JSON_CTRL") {
-        let cmd = cmd
+        let cmd = cmd.strip_prefix("JSON_CTRL").context("invalid prefix")?;
-            .strip_prefix("JSON_CTRL")
+        Ok(SafekeeperPostgresCommand::JSONCtrl {
-            .ok_or_else(|| anyhow!("invalid prefix"))?;
+            cmd: serde_json::from_str(cmd)?,
-        let parsed_cmd: AppendLogicalMessage = serde_json::from_str(cmd)?;
+        })
        Ok(SafekeeperPostgresCommand::JSONCtrl { cmd: parsed_cmd })
    } else {
        bail!("unsupported command {}", cmd);
    }
@@ -89,6 +89,8 @@ impl postgres_backend::Handler for SafekeeperPostgresHandler {
                self.appname = Some(app_name.clone());
            }
            self.pageserver_connstr = params.get("pageserver_connstr").cloned();
            Ok(())
        } else {
            bail!("Walkeeper received unexpected initial message: {:?}", sm);
@@ -104,12 +106,8 @@ impl postgres_backend::Handler for SafekeeperPostgresHandler {
            | SafekeeperPostgresCommand::StartReplication { .. }
            | SafekeeperPostgresCommand::IdentifySystem
            | SafekeeperPostgresCommand::JSONCtrl { .. } => {
-                let tenantid = self
+                let tenantid = self.ztenantid.context("tenantid is required")?;
-                    .ztenantid
+                let timelineid = self.ztimelineid.context("timelineid is required")?;
                    .ok_or_else(|| anyhow!("tenantid is required"))?;
                let timelineid = self
                    .ztimelineid
                    .ok_or_else(|| anyhow!("timelineid is required"))?;
                if self.timeline.is_none() {
                    // START_WAL_PUSH is the only command that initializes the timeline in production.
                    // There is also JSON_CTRL command, which should initialize the timeline for testing.
@@ -118,8 +116,11 @@ impl postgres_backend::Handler for SafekeeperPostgresHandler {
                        | SafekeeperPostgresCommand::JSONCtrl { .. } => CreateControlFile::True,
                        _ => CreateControlFile::False,
                    };
-                    self.timeline
+                    self.timeline.set(
-                        .set(&self.conf, tenantid, timelineid, create_control_file)?;
+                        &self.conf,
                        ZTenantTimelineId::new(tenantid, timelineid),
                        create_control_file,
                    )?;
                }
            }
        }
@@ -128,12 +129,12 @@ impl postgres_backend::Handler for SafekeeperPostgresHandler {
            SafekeeperPostgresCommand::StartWalPush { pageserver_connstr } => {
                ReceiveWalConn::new(pgb, pageserver_connstr)
                    .run(self)
-                    .with_context(|| "failed to run ReceiveWalConn")?;
+                    .context("failed to run ReceiveWalConn")?;
            }
            SafekeeperPostgresCommand::StartReplication { start_lsn } => {
                ReplicationConn::new(pgb)
-                    .run(self, pgb, start_lsn)
+                    .run(self, pgb, start_lsn, self.pageserver_connstr.clone())
-                    .with_context(|| "failed to run ReplicationConn")?;
+                    .context("failed to run ReplicationConn")?;
            }
            SafekeeperPostgresCommand::IdentifySystem => {
                self.handle_identify_system(pgb)?;
@@ -154,6 +155,7 @@ impl SafekeeperPostgresHandler {
            ztenantid: None,
            ztimelineid: None,
            timeline: None,
            pageserver_connstr: None,
            tx,
        }
    }
@@ -6,6 +6,7 @@ use serde::Serializer;
 use std::fmt::Display;
 use std::sync::Arc;
 use zenith_utils::lsn::Lsn;
 use zenith_utils::zid::ZTenantTimelineId;
 use crate::safekeeper::Term;
 use crate::safekeeper::TermHistory;
@@ -65,16 +66,13 @@ struct TimelineStatus {
 /// Report info about timeline.
 async fn timeline_status_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
-    let tenant_id: ZTenantId = parse_request_param(&request, "tenant_id")?;
+    let zttid = ZTenantTimelineId::new(
-    let timeline_id: ZTimelineId = parse_request_param(&request, "timeline_id")?;
+        parse_request_param(&request, "tenant_id")?,
        parse_request_param(&request, "timeline_id")?,
    );
-    let tli = GlobalTimelines::get(
+    let tli = GlobalTimelines::get(get_conf(&request), zttid, CreateControlFile::False)
-        get_conf(&request),
+        .map_err(ApiError::from_err)?;
        tenant_id,
        timeline_id,
        CreateControlFile::False,
    )
    .map_err(ApiError::from_err)?;
    let sk_state = tli.get_info();
    let flush_lsn = tli.get_end_of_wal();
@@ -85,8 +83,8 @@ async fn timeline_status_handler(request: Request<Body>) -> Result<Response<Body
    };
    let status = TimelineStatus {
-        tenant_id,
+        tenant_id: zttid.tenant_id,
-        timeline_id,
+        timeline_id: zttid.timeline_id,
        acceptor_state: acc_state,
        commit_lsn: sk_state.commit_lsn,
        truncate_lsn: sk_state.truncate_lsn,
@@ -2,7 +2,7 @@
 use std::path::PathBuf;
 use std::time::Duration;
-use zenith_utils::zid::ZTimelineId;
+use zenith_utils::zid::ZTenantTimelineId;
 pub mod callmemaybe;
 pub mod handler;
@@ -47,8 +47,10 @@ pub struct SafeKeeperConf {
 }
 impl SafeKeeperConf {
-    pub fn timeline_dir(&self, timelineid: &ZTimelineId) -> PathBuf {
+    pub fn timeline_dir(&self, zttid: &ZTenantTimelineId) -> PathBuf {
-        self.workdir.join(timelineid.to_string())
+        self.workdir
            .join(zttid.tenant_id.to_string())
            .join(zttid.timeline_id.to_string())
    }
 }
@@ -18,10 +18,10 @@ use crate::handler::SafekeeperPostgresHandler;
 use crate::timeline::TimelineTools;
 use zenith_utils::postgres_backend::PostgresBackend;
 use zenith_utils::pq_proto::{BeMessage, FeMessage};
-use zenith_utils::zid::{ZTenantId, ZTimelineId};
+use zenith_utils::zid::ZTenantId;
 use crate::callmemaybe::CallmeEvent;
-use tokio::sync::mpsc::UnboundedSender;
+use crate::callmemaybe::SubscriptionStateKey;
 pub struct ReceiveWalConn<'pg> {
    /// Postgres connection
@@ -105,15 +105,14 @@ impl<'pg> ReceiveWalConn<'pg> {
                // Need to establish replication channel with page server.
                // Add far as replication in postgres is initiated by receiver
                // we should use callmemaybe mechanism.
-                let timelineid = spg.timeline.get().timelineid;
+                let timeline_id = spg.timeline.get().zttid.timeline_id;
-                let tx_clone = spg.tx.clone();
+                let subscription_key = SubscriptionStateKey::new(
-                let pageserver_connstr = pageserver_connstr.to_owned();
+                    tenant_id,
                    timeline_id,
                    pageserver_connstr.to_owned(),
                );
                spg.tx
-                    .send(CallmeEvent::Subscribe(
+                    .send(CallmeEvent::Subscribe(subscription_key))
                        tenant_id,
                        timelineid,
                        pageserver_connstr,
                    ))
                    .unwrap_or_else(|e| {
                        error!(
                            "failed to send Subscribe request to callmemaybe thread {}",
@@ -123,9 +122,6 @@ impl<'pg> ReceiveWalConn<'pg> {
                // create a guard to unsubscribe callback, when this wal_stream will exit
                Some(SendWalHandlerGuard {
                    _tx: tx_clone,
                    _tenant_id: tenant_id,
                    _timelineid: timelineid,
                    timeline: Arc::clone(spg.timeline.get()),
                })
            }
@@ -137,7 +133,7 @@ impl<'pg> ReceiveWalConn<'pg> {
                .timeline
                .get()
                .process_msg(&msg)
-                .with_context(|| "failed to process ProposerAcceptorMessage")?;
+                .context("failed to process ProposerAcceptorMessage")?;
            if let Some(reply) = reply {
                self.write_msg(&reply)?;
            }
@@ -147,22 +143,11 @@ impl<'pg> ReceiveWalConn<'pg> {
 }
 struct SendWalHandlerGuard {
    _tx: UnboundedSender<CallmeEvent>,
    _tenant_id: ZTenantId,
    _timelineid: ZTimelineId,
    timeline: Arc<Timeline>,
 }
 impl Drop for SendWalHandlerGuard {
    fn drop(&mut self) {
        self.timeline.stop_streaming();
        // self.tx
        //     .send(CallmeEvent::Unsubscribe(self.tenant_id, self.timelineid))
        //     .unwrap_or_else(|e| {
        //         error!(
        //             "failed to send Unsubscribe request to callmemaybe thread {}",
        //             e
        //         );
        //     });
    }
 }
@@ -1,13 +1,8 @@
 //! Acceptor part of proposer-acceptor consensus algorithm.
-use anyhow::Context;
+use anyhow::{bail, Context, Result};
-use anyhow::{anyhow, bail, Result};
+use byteorder::{LittleEndian, ReadBytesExt};
-use byteorder::LittleEndian;
+use bytes::{Buf, BufMut, Bytes, BytesMut};
 use byteorder::ReadBytesExt;
 use bytes::Buf;
 use bytes::BufMut;
 use bytes::Bytes;
 use bytes::BytesMut;
 use postgres_ffi::waldecoder::WalStreamDecoder;
 use postgres_ffi::xlog_utils::TimeLineID;
 use serde::{Deserialize, Serialize};
@@ -27,6 +22,7 @@ use zenith_metrics::{
 use zenith_utils::bin_ser::LeSer;
 use zenith_utils::lsn::Lsn;
 use zenith_utils::pq_proto::SystemId;
 use zenith_utils::pq_proto::ZenithFeedback;
 use zenith_utils::zid::{ZTenantId, ZTimelineId};
 pub const SK_MAGIC: u32 = 0xcafeceefu32;
@@ -283,9 +279,8 @@ pub struct AppendResponse {
    // We report back our awareness about which WAL is committed, as this is
    // a criterion for walproposer --sync mode exit
    pub commit_lsn: Lsn,
    // Min disk consistent lsn of pageservers (portion of WAL applied and written to the disk by pageservers)
    pub disk_consistent_lsn: Lsn,
    pub hs_feedback: HotStandbyFeedback,
    pub zenith_feedback: ZenithFeedback,
 }
 impl AppendResponse {
@@ -294,8 +289,8 @@ impl AppendResponse {
            term,
            flush_lsn: Lsn(0),
            commit_lsn: Lsn(0),
            disk_consistent_lsn: Lsn(0),
            hs_feedback: HotStandbyFeedback::empty(),
            zenith_feedback: ZenithFeedback::empty(),
        }
    }
 }
@@ -346,7 +341,7 @@ impl ProposerAcceptorMessage {
                let rec_size = hdr
                    .end_lsn
                    .checked_sub(hdr.begin_lsn)
-                    .ok_or_else(|| anyhow!("begin_lsn > end_lsn in AppendRequest"))?
+                    .context("begin_lsn > end_lsn in AppendRequest")?
                    .0 as usize;
                if rec_size > MAX_SEND_SIZE {
                    bail!(
@@ -362,7 +357,7 @@ impl ProposerAcceptorMessage {
                Ok(ProposerAcceptorMessage::AppendRequest(msg))
            }
-            _ => Err(anyhow!("unknown proposer-acceptor message tag: {}", tag,)),
+            _ => bail!("unknown proposer-acceptor message tag: {}", tag,),
        }
    }
 }
@@ -400,10 +395,11 @@ impl AcceptorProposerMessage {
                buf.put_u64_le(msg.term);
                buf.put_u64_le(msg.flush_lsn.into());
                buf.put_u64_le(msg.commit_lsn.into());
                buf.put_u64_le(msg.disk_consistent_lsn.into());
                buf.put_i64_le(msg.hs_feedback.ts);
                buf.put_u64_le(msg.hs_feedback.xmin);
                buf.put_u64_le(msg.hs_feedback.catalog_xmin);
                msg.zenith_feedback.serialize(buf)?
            }
        }
@@ -567,7 +563,7 @@ where
        self.s.server.wal_seg_size = msg.wal_seg_size;
        self.storage
            .persist(&self.s)
-            .with_context(|| "failed to persist shared state")?;
+            .context("failed to persist shared state")?;
        self.metrics = SafeKeeperMetrics::new(self.s.server.timeline_id);
@@ -619,9 +615,9 @@ where
            term: self.s.acceptor_state.term,
            flush_lsn: self.flush_lsn,
            commit_lsn: self.s.commit_lsn,
            disk_consistent_lsn: Lsn(0),
            // will be filled by the upper code to avoid bothering safekeeper
            hs_feedback: HotStandbyFeedback::empty(),
            zenith_feedback: ZenithFeedback::empty(),
        }
    }
@@ -3,12 +3,14 @@
 use crate::handler::SafekeeperPostgresHandler;
 use crate::timeline::{ReplicaState, Timeline, TimelineTools};
-use anyhow::{anyhow, bail, Context, Result};
+use anyhow::{bail, Context, Result};
 use postgres_ffi::xlog_utils::{
    get_current_timestamp, TimestampTz, XLogFileName, MAX_SEND_SIZE, PG_TLI,
 };
 use crate::callmemaybe::{CallmeEvent, SubscriptionStateKey};
 use bytes::Bytes;
 use serde::{Deserialize, Serialize};
 use std::cmp::min;
 use std::fs::File;
@@ -19,20 +21,21 @@ use std::sync::Arc;
 use std::thread::sleep;
 use std::time::Duration;
 use std::{str, thread};
 use tokio::sync::mpsc::UnboundedSender;
 use tracing::*;
 use zenith_utils::bin_ser::BeSer;
 use zenith_utils::lsn::Lsn;
 use zenith_utils::postgres_backend::PostgresBackend;
-use zenith_utils::pq_proto::{BeMessage, FeMessage, WalSndKeepAlive, XLogDataBody};
+use zenith_utils::pq_proto::{BeMessage, FeMessage, WalSndKeepAlive, XLogDataBody, ZenithFeedback};
 use zenith_utils::sock_split::ReadStream;
 use crate::callmemaybe::CallmeEvent;
 use tokio::sync::mpsc::UnboundedSender;
 use zenith_utils::zid::{ZTenantId, ZTimelineId};
 // See: https://www.postgresql.org/docs/13/protocol-replication.html
 const HOT_STANDBY_FEEDBACK_TAG_BYTE: u8 = b'h';
 const STANDBY_STATUS_UPDATE_TAG_BYTE: u8 = b'r';
 // zenith extension of replication protocol
 const ZENITH_STATUS_UPDATE_TAG_BYTE: u8 = b'z';
 type FullTransactionId = u64;
@@ -90,18 +93,27 @@ impl Drop for ReplicationConnGuard {
 struct ReplicationStreamGuard {
    tx: UnboundedSender<CallmeEvent>,
    tenant_id: ZTenantId,
-    timelineid: ZTimelineId,
+    timeline_id: ZTimelineId,
    pageserver_connstr: String,
 }
 impl Drop for ReplicationStreamGuard {
    fn drop(&mut self) {
        // the connection with pageserver is lost,
        // resume callback subscription
-        debug!("Connection to pageserver is gone. Resume callmemeybe subsciption if necessary. tenantid {} timelineid {}",
+        debug!(
-         self.tenant_id, self.timelineid);
+            "Connection to pageserver is gone. Resume callmemaybe subsciption if necessary. tenantid {} timelineid {}",
            self.tenant_id, self.timeline_id,
        );
        let subscription_key = SubscriptionStateKey::new(
            self.tenant_id,
            self.timeline_id,
            self.pageserver_connstr.to_owned(),
        );
        self.tx
-            .send(CallmeEvent::Resume(self.tenant_id, self.timelineid))
+            .send(CallmeEvent::Resume(subscription_key))
            .unwrap_or_else(|e| {
                error!("failed to send Resume request to callmemaybe thread {}", e);
            });
@@ -130,8 +142,8 @@ impl ReplicationConn {
        while let Some(msg) = FeMessage::read(&mut stream_in)? {
            match &msg {
                FeMessage::CopyData(m) => {
-                    // There's two possible data messages that the client is supposed to send here:
+                    // There's three possible data messages that the client is supposed to send here:
-                    // `HotStandbyFeedback` and `StandbyStatusUpdate`.
+                    // `HotStandbyFeedback` and `StandbyStatusUpdate` and `ZenithStandbyFeedback`.
                    match m.first().cloned() {
                        Some(HOT_STANDBY_FEEDBACK_TAG_BYTE) => {
@@ -141,11 +153,25 @@ impl ReplicationConn {
                            timeline.update_replica_state(replica_id, state);
                        }
                        Some(STANDBY_STATUS_UPDATE_TAG_BYTE) => {
-                            let reply = StandbyReply::des(&m[1..])
+                            let _reply = StandbyReply::des(&m[1..])
                                .context("failed to deserialize StandbyReply")?;
-                            state.last_received_lsn = reply.write_lsn;
+                            // This must be a regular postgres replica,
-                            state.disk_consistent_lsn = reply.flush_lsn;
+                            // because pageserver doesn't send this type of messages to safekeeper.
-                            state.remote_consistent_lsn = reply.apply_lsn;
+                            // Currently this is not implemented, so this message is ignored.
                            warn!("unexpected StandbyReply. Read-only postgres replicas are not supported in safekeepers yet.");
                            // timeline.update_replica_state(replica_id, Some(state));
                        }
                        Some(ZENITH_STATUS_UPDATE_TAG_BYTE) => {
                            // Note: deserializing is on m[9..] because we skip the tag byte and len bytes.
                            let buf = Bytes::copy_from_slice(&m[9..]);
                            let reply = ZenithFeedback::parse(buf);
                            info!("ZenithFeedback is {:?}", reply);
                            // Only pageserver sends ZenithFeedback, so set the flag.
                            // This replica is the source of information to resend to compute.
                            state.zenith_feedback = Some(reply);
                            timeline.update_replica_state(replica_id, state);
                        }
                        _ => warn!("unexpected message {:?}", msg),
@@ -156,7 +182,7 @@ impl ReplicationConn {
                    // Shutdown the connection, because rust-postgres client cannot be dropped
                    // when connection is alive.
                    let _ = stream_in.shutdown(Shutdown::Both);
-                    return Err(anyhow!("Copy failed"));
+                    bail!("Copy failed");
                }
                _ => {
                    // We only handle `CopyData`, 'Sync', 'CopyFail' messages. Anything else is ignored.
@@ -194,8 +220,9 @@ impl ReplicationConn {
        spg: &mut SafekeeperPostgresHandler,
        pgb: &mut PostgresBackend,
        mut start_pos: Lsn,
        pageserver_connstr: Option<String>,
    ) -> Result<()> {
-        let _enter = info_span!("WAL sender", timeline = %spg.ztimelineid.unwrap()).entered();
+        let _enter = info_span!("WAL sender", timeline = %spg.ztimelineid.unwrap(), pageserver_connstr = %pageserver_connstr.as_deref().unwrap_or_default()).entered();
        // spawn the background thread which receives HotStandbyFeedback messages.
        let bg_timeline = Arc::clone(spg.timeline.get());
@@ -256,11 +283,14 @@ impl ReplicationConn {
            if spg.appname == Some("wal_proposer_recovery".to_string()) {
                None
            } else {
-                let timelineid = spg.timeline.get().timelineid;
+                let pageserver_connstr = pageserver_connstr.clone().expect("there should be a pageserver connection string since this is not a wal_proposer_recovery");
                let tenant_id = spg.ztenantid.unwrap();
                let timeline_id = spg.timeline.get().zttid.timeline_id;
                let tx_clone = spg.tx.clone();
                let subscription_key =
                    SubscriptionStateKey::new(tenant_id, timeline_id, pageserver_connstr.clone());
                spg.tx
-                    .send(CallmeEvent::Pause(tenant_id, timelineid))
+                    .send(CallmeEvent::Pause(subscription_key))
                    .unwrap_or_else(|e| {
                        error!("failed to send Pause request to callmemaybe thread {}", e);
                    });
@@ -269,7 +299,8 @@ impl ReplicationConn {
                Some(ReplicationStreamGuard {
                    tx: tx_clone,
                    tenant_id,
-                    timelineid,
+                    timeline_id,
                    pageserver_connstr,
                })
            }
        };
@@ -293,12 +324,18 @@ impl ReplicationConn {
                if let Some(lsn) = lsn {
                    end_pos = lsn;
                } else {
-                    // Is is time to end streaming to this replica?
+                    // Is it time to end streaming to this replica?
                    if spg.timeline.get().check_stop_streaming(replica_id) {
-                        let timelineid = spg.timeline.get().timelineid;
+                        // this expect should never fail because in wal_proposer_recovery mode stop_pos is set
                        // and this code is not reachable
                        let pageserver_connstr = pageserver_connstr
                            .expect("there should be a pageserver connection string");
                        let tenant_id = spg.ztenantid.unwrap();
                        let timeline_id = spg.timeline.get().zttid.timeline_id;
                        let subscription_key =
                            SubscriptionStateKey::new(tenant_id, timeline_id, pageserver_connstr);
                        spg.tx
-                            .send(CallmeEvent::Unsubscribe(tenant_id, timelineid))
+                            .send(CallmeEvent::Unsubscribe(subscription_key))
                            .unwrap_or_else(|e| {
                                error!("failed to send Pause request to callmemaybe thread {}", e);
                            });
@@ -324,8 +361,10 @@ impl ReplicationConn {
                    // Open a new file.
                    let segno = start_pos.segment_number(wal_seg_size);
                    let wal_file_name = XLogFileName(PG_TLI, segno, wal_seg_size);
-                    let timeline_id = spg.timeline.get().timelineid;
+                    let wal_file_path = spg
-                    let wal_file_path = spg.conf.timeline_dir(&timeline_id).join(wal_file_name);
+                        .conf
                        .timeline_dir(&spg.timeline.get().zttid)
                        .join(wal_file_name);
                    Self::open_wal_file(&wal_file_path)?
                }
            };
@@ -4,7 +4,7 @@
 use anyhow::{bail, ensure, Context, Result};
 use byteorder::{LittleEndian, ReadBytesExt, WriteBytesExt};
 use lazy_static::lazy_static;
-use postgres_ffi::xlog_utils::{find_end_of_wal, PG_TLI};
+use postgres_ffi::xlog_utils::{find_end_of_wal, XLogSegNo, PG_TLI};
 use std::cmp::{max, min};
 use std::collections::HashMap;
 use std::fs::{self, File, OpenOptions};
@@ -16,7 +16,7 @@ use tracing::*;
 use zenith_metrics::{register_histogram_vec, Histogram, HistogramVec, DISK_WRITE_SECONDS_BUCKETS};
 use zenith_utils::bin_ser::LeSer;
 use zenith_utils::lsn::Lsn;
-use zenith_utils::zid::{ZTenantId, ZTimelineId};
+use zenith_utils::zid::ZTenantTimelineId;
 use crate::safekeeper::{
    AcceptorProposerMessage, ProposerAcceptorMessage, SafeKeeper, SafeKeeperState, ServerInfo,
@@ -27,6 +27,7 @@ use crate::upgrade::upgrade_control_file;
 use crate::SafeKeeperConf;
 use postgres_ffi::xlog_utils::{XLogFileName, XLOG_BLCKSZ};
 use std::convert::TryInto;
 use zenith_utils::pq_proto::ZenithFeedback;
 // contains persistent metadata for safekeeper
 const CONTROL_FILE_NAME: &str = "safekeeper.control";
@@ -35,17 +36,17 @@ const CONTROL_FILE_NAME_PARTIAL: &str = "safekeeper.control.partial";
 const POLL_STATE_TIMEOUT: Duration = Duration::from_secs(1);
 pub const CHECKSUM_SIZE: usize = std::mem::size_of::<u32>();
-/// Replica status: host standby feedback + disk consistent lsn
+/// Replica status update + hot standby feedback
 #[derive(Debug, Clone, Copy)]
 pub struct ReplicaState {
    /// last known lsn received by replica
    pub last_received_lsn: Lsn, // None means we don't know
    /// combined disk_consistent_lsn of pageservers
    pub disk_consistent_lsn: Lsn,
    /// combined remote consistent lsn of pageservers
    pub remote_consistent_lsn: Lsn,
    /// combined hot standby feedback from all replicas
    pub hs_feedback: HotStandbyFeedback,
    /// Zenith specific feedback received from pageserver, if any
    pub zenith_feedback: Option<ZenithFeedback>,
 }
 impl Default for ReplicaState {
@@ -58,13 +59,13 @@ impl ReplicaState {
    pub fn new() -> ReplicaState {
        ReplicaState {
            last_received_lsn: Lsn::MAX,
-            disk_consistent_lsn: Lsn(u64::MAX),
+            remote_consistent_lsn: Lsn(0),
            remote_consistent_lsn: Lsn(u64::MAX),
            hs_feedback: HotStandbyFeedback {
                ts: 0,
                xmin: u64::MAX,
                catalog_xmin: u64::MAX,
            },
            zenith_feedback: None,
        }
    }
 }
@@ -109,13 +110,40 @@ impl SharedState {
            acc.hs_feedback.xmin = min(acc.hs_feedback.xmin, state.hs_feedback.xmin);
            acc.hs_feedback.catalog_xmin =
                min(acc.hs_feedback.catalog_xmin, state.hs_feedback.catalog_xmin);
-            acc.disk_consistent_lsn = Lsn::min(acc.disk_consistent_lsn, state.disk_consistent_lsn);
+
-            // currently not used, but update it to be consistent
+            // FIXME
-            acc.last_received_lsn = Lsn::min(acc.last_received_lsn, state.last_received_lsn);
+            // If multiple pageservers are streaming WAL and send feedback for the same timeline simultaneously,
-            // When at least one replica has preserved data up to remote_consistent_lsn,
+            // this code is not correct.
-            // safekeeper is free to delete it, so chose max of all replicas.
+            // Now the most advanced feedback is used.
-            acc.remote_consistent_lsn =
+            // If one pageserver lags when another doesn't, the backpressure won't be activated on compute and lagging
-                Lsn::max(acc.remote_consistent_lsn, state.remote_consistent_lsn);
+            // pageserver is prone to timeout errors.
            //
            // To choose what feedback to use and resend to compute node,
            // we need to know which pageserver compute node considers to be main.
            // See https://github.com/zenithdb/zenith/issues/1171
            //
            if let Some(zenith_feedback) = state.zenith_feedback {
                if let Some(acc_feedback) = acc.zenith_feedback {
                    if acc_feedback.ps_writelsn < zenith_feedback.ps_writelsn {
                        warn!("More than one pageserver is streaming WAL for the timeline. Feedback resolving is not fully supported yet.");
                        acc.zenith_feedback = Some(zenith_feedback);
                    }
                } else {
                    acc.zenith_feedback = Some(zenith_feedback);
                }
                // last lsn received by pageserver
                // FIXME if multiple pageservers are streaming WAL, last_received_lsn must be tracked per pageserver.
                // See https://github.com/zenithdb/zenith/issues/1171
                acc.last_received_lsn = Lsn::from(zenith_feedback.ps_writelsn);
                // When at least one pageserver has preserved data up to remote_consistent_lsn,
                // safekeeper is free to delete it, so choose max of all pageservers.
                acc.remote_consistent_lsn = max(
                    Lsn::from(zenith_feedback.ps_applylsn),
                    acc.remote_consistent_lsn,
                );
            }
        }
        acc
    }
@@ -136,14 +164,14 @@ impl SharedState {
    /// If create=false and file doesn't exist, bails out.
    fn create_restore(
        conf: &SafeKeeperConf,
-        timeline_id: ZTimelineId,
+        zttid: &ZTenantTimelineId,
        create: CreateControlFile,
    ) -> Result<Self> {
-        let state = FileStorage::load_control_file_conf(conf, timeline_id, create)
+        let state = FileStorage::load_control_file_conf(conf, zttid, create)
-            .with_context(|| "failed to load from control file")?;
+            .context("failed to load from control file")?;
-        let file_storage = FileStorage::new(timeline_id, conf);
+        let file_storage = FileStorage::new(zttid, conf);
        let flush_lsn = if state.server.wal_seg_size != 0 {
-            let wal_dir = conf.timeline_dir(&timeline_id);
+            let wal_dir = conf.timeline_dir(zttid);
            find_end_of_wal(
                &wal_dir,
                state.server.wal_seg_size as usize,
@@ -166,16 +194,16 @@ impl SharedState {
 /// Database instance (tenant)
 pub struct Timeline {
-    pub timelineid: ZTimelineId,
+    pub zttid: ZTenantTimelineId,
    mutex: Mutex<SharedState>,
    /// conditional variable used to notify wal senders
    cond: Condvar,
 }
 impl Timeline {
-    fn new(timelineid: ZTimelineId, shared_state: SharedState) -> Timeline {
+    fn new(zttid: ZTenantTimelineId, shared_state: SharedState) -> Timeline {
        Timeline {
-            timelineid,
+            zttid,
            mutex: Mutex::new(shared_state),
            cond: Condvar::new(),
        }
@@ -280,8 +308,9 @@ impl Timeline {
            if let Some(AcceptorProposerMessage::AppendResponse(ref mut resp)) = rmsg {
                let state = shared_state.get_replicas_state();
                resp.hs_feedback = state.hs_feedback;
-                resp.disk_consistent_lsn = state.disk_consistent_lsn;
+                if let Some(zenith_feedback) = state.zenith_feedback {
-                // XXX Do we need to add state.last_received_lsn to resp?
+                    resp.zenith_feedback = zenith_feedback;
                }
            }
        }
        // Ping wal sender that new data might be available.
@@ -320,8 +349,7 @@ pub trait TimelineTools {
    fn set(
        &mut self,
        conf: &SafeKeeperConf,
-        tenant_id: ZTenantId,
+        zttid: ZTenantTimelineId,
        timeline_id: ZTimelineId,
        create: CreateControlFile,
    ) -> Result<()>;
@@ -332,14 +360,13 @@ impl TimelineTools for Option<Arc<Timeline>> {
    fn set(
        &mut self,
        conf: &SafeKeeperConf,
-        tenant_id: ZTenantId,
+        zttid: ZTenantTimelineId,
        timeline_id: ZTimelineId,
        create: CreateControlFile,
    ) -> Result<()> {
        // We will only set the timeline once. If it were to ever change,
        // anyone who cloned the Arc would be out of date.
        assert!(self.is_none());
-        *self = Some(GlobalTimelines::get(conf, tenant_id, timeline_id, create)?);
+        *self = Some(GlobalTimelines::get(conf, zttid, create)?);
        Ok(())
    }
@@ -349,7 +376,7 @@ impl TimelineTools for Option<Arc<Timeline>> {
 }
 lazy_static! {
-    pub static ref TIMELINES: Mutex<HashMap<(ZTenantId, ZTimelineId), Arc<Timeline>>> =
+    pub static ref TIMELINES: Mutex<HashMap<ZTenantTimelineId, Arc<Timeline>>> =
        Mutex::new(HashMap::new());
 }
@@ -361,26 +388,29 @@ impl GlobalTimelines {
    /// If control file doesn't exist and create=false, bails out.
    pub fn get(
        conf: &SafeKeeperConf,
-        tenant_id: ZTenantId,
+        zttid: ZTenantTimelineId,
        timeline_id: ZTimelineId,
        create: CreateControlFile,
    ) -> Result<Arc<Timeline>> {
        let mut timelines = TIMELINES.lock().unwrap();
-        match timelines.get(&(tenant_id, timeline_id)) {
+        match timelines.get(&zttid) {
            Some(result) => Ok(Arc::clone(result)),
            None => {
-                info!(
+                if let CreateControlFile::True = create {
-                    "creating timeline dir {}, create is {:?}",
+                    let dir = conf.timeline_dir(&zttid);
-                    timeline_id, create
+                    info!(
-                );
+                        "creating timeline dir {}, create is {:?}",
-                fs::create_dir_all(timeline_id.to_string())?;
+                        dir.display(),
                        create
                    );
                    fs::create_dir_all(dir)?;
                }
-                let shared_state = SharedState::create_restore(conf, timeline_id, create)
+                let shared_state = SharedState::create_restore(conf, &zttid, create)
-                    .with_context(|| "failed to restore shared state")?;
+                    .context("failed to restore shared state")?;
-                let new_tli = Arc::new(Timeline::new(timeline_id, shared_state));
+                let new_tli = Arc::new(Timeline::new(zttid, shared_state));
-                timelines.insert((tenant_id, timeline_id), Arc::clone(&new_tli));
+                timelines.insert(zttid, Arc::clone(&new_tli));
                Ok(new_tli)
            }
        }
@@ -396,9 +426,9 @@ pub struct FileStorage {
 }
 impl FileStorage {
-    fn new(timeline_id: ZTimelineId, conf: &SafeKeeperConf) -> FileStorage {
+    fn new(zttid: &ZTenantTimelineId, conf: &SafeKeeperConf) -> FileStorage {
-        let timeline_dir = conf.timeline_dir(&timeline_id);
+        let timeline_dir = conf.timeline_dir(zttid);
-        let timelineid_str = format!("{}", timeline_id);
+        let timelineid_str = format!("{}", zttid);
        FileStorage {
            timeline_dir,
            conf: conf.clone(),
@@ -427,12 +457,13 @@ impl FileStorage {
        upgrade_control_file(buf, version)
    }
    // Load control file for given zttid at path specified by conf.
    fn load_control_file_conf(
        conf: &SafeKeeperConf,
-        timeline_id: ZTimelineId,
+        zttid: &ZTenantTimelineId,
        create: CreateControlFile,
    ) -> Result<SafeKeeperState> {
-        let path = conf.timeline_dir(&timeline_id).join(CONTROL_FILE_NAME);
+        let path = conf.timeline_dir(zttid).join(CONTROL_FILE_NAME);
        Self::load_control_file(path, create)
    }
@@ -470,7 +501,7 @@ impl FileStorage {
            let mut buf = Vec::new();
            control_file
                .read_to_end(&mut buf)
-                .with_context(|| "failed to read control file")?;
+                .context("failed to read control file")?;
            let calculated_checksum = crc32c::crc32c(&buf[..buf.len() - CHECKSUM_SIZE]);
@@ -497,6 +528,14 @@ impl FileStorage {
        };
        Ok(state)
    }
    /// Helper returning full path to WAL segment file and its .partial brother.
    fn wal_file_paths(&self, segno: XLogSegNo, wal_seg_size: usize) -> (PathBuf, PathBuf) {
        let wal_file_name = XLogFileName(PG_TLI, segno, wal_seg_size);
        let wal_file_path = self.timeline_dir.join(wal_file_name.clone());
        let wal_file_partial_path = self.timeline_dir.join(wal_file_name + ".partial");
        (wal_file_path, wal_file_partial_path)
    }
 }
 impl Storage for FileStorage {
@@ -554,7 +593,7 @@ impl Storage for FileStorage {
        // fsync the directory (linux specific)
        File::open(&self.timeline_dir)
            .and_then(|f| f.sync_all())
-            .with_context(|| "failed to sync control file directory")?;
+            .context("failed to sync control file directory")?;
        Ok(())
    }
@@ -565,7 +604,6 @@ impl Storage for FileStorage {
        let mut start_pos = startpos;
        const ZERO_BLOCK: &[u8] = &[0u8; XLOG_BLCKSZ];
        let wal_seg_size = server.wal_seg_size as usize;
        let ztli = server.timeline_id;
        /* Extract WAL location for this block */
        let mut xlogoff = start_pos.segment_offset(wal_seg_size) as usize;
@@ -585,14 +623,7 @@ impl Storage for FileStorage {
            /* Open file */
            let segno = start_pos.segment_number(wal_seg_size);
-            // note: we basically don't support changing pg timeline
+            let (wal_file_path, wal_file_partial_path) = self.wal_file_paths(segno, wal_seg_size);
            let wal_file_name = XLogFileName(PG_TLI, segno, wal_seg_size);
            let wal_file_path = self.conf.timeline_dir(&ztli).join(wal_file_name.clone());
            let wal_file_partial_path = self
                .conf
                .timeline_dir(&ztli)
                .join(wal_file_name.clone() + ".partial");
            {
                let mut wal_file: File;
                /* Try to open already completed segment */
@@ -653,25 +684,13 @@ impl Storage for FileStorage {
        let partial;
        const ZERO_BLOCK: &[u8] = &[0u8; XLOG_BLCKSZ];
        let wal_seg_size = server.wal_seg_size as usize;
        let ztli = server.timeline_id;
        /* Extract WAL location for this block */
        let mut xlogoff = end_pos.segment_offset(wal_seg_size) as usize;
        /* Open file */
        let mut segno = end_pos.segment_number(wal_seg_size);
-        // note: we basically don't support changing pg timeline
+        let (wal_file_path, wal_file_partial_path) = self.wal_file_paths(segno, wal_seg_size);
        let wal_file_name = XLogFileName(PG_TLI, segno, wal_seg_size);
        let wal_file_path = self
            .conf
            .workdir
            .join(ztli.to_string())
            .join(wal_file_name.clone());
        let wal_file_partial_path = self
            .conf
            .workdir
            .join(ztli.to_string())
            .join(wal_file_name + ".partial");
        {
            let mut wal_file: File;
            /* Try to open already completed segment */
@@ -702,17 +721,7 @@ impl Storage for FileStorage {
        // Remove all subsequent segments
        loop {
            segno += 1;
-            let wal_file_name = XLogFileName(PG_TLI, segno, wal_seg_size);
+            let (wal_file_path, wal_file_partial_path) = self.wal_file_paths(segno, wal_seg_size);
            let wal_file_path = self
                .conf
                .workdir
                .join(ztli.to_string())
                .join(wal_file_name.clone());
            let wal_file_partial_path = self
                .conf
                .workdir
                .join(ztli.to_string())
                .join(wal_file_name.clone() + ".partial");
            // TODO: better use fs::try_exists which is currenty avaialble only in nightly build
            if wal_file_path.exists() {
                fs::remove_file(&wal_file_path)?;
@@ -732,11 +741,11 @@ mod test {
    use crate::{
        safekeeper::{SafeKeeperState, Storage},
        timeline::{CreateControlFile, CONTROL_FILE_NAME},
-        SafeKeeperConf,
+        SafeKeeperConf, ZTenantTimelineId,
    };
    use anyhow::Result;
    use std::fs;
-    use zenith_utils::{lsn::Lsn, zid::ZTimelineId};
+    use zenith_utils::lsn::Lsn;
    fn stub_conf() -> SafeKeeperConf {
        let workdir = tempfile::tempdir().unwrap().into_path();
@@ -748,31 +757,30 @@ mod test {
    fn load_from_control_file(
        conf: &SafeKeeperConf,
-        timeline_id: ZTimelineId,
+        zttid: &ZTenantTimelineId,
        create: CreateControlFile,
    ) -> Result<(FileStorage, SafeKeeperState)> {
-        fs::create_dir_all(&conf.timeline_dir(&timeline_id))
+        fs::create_dir_all(&conf.timeline_dir(zttid)).expect("failed to create timeline dir");
            .expect("failed to create timeline dir");
        Ok((
-            FileStorage::new(timeline_id, conf),
+            FileStorage::new(zttid, conf),
-            FileStorage::load_control_file_conf(conf, timeline_id, create)?,
+            FileStorage::load_control_file_conf(conf, zttid, create)?,
        ))
    }
    #[test]
    fn test_read_write_safekeeper_state() {
        let conf = stub_conf();
-        let timeline_id = ZTimelineId::generate();
+        let zttid = ZTenantTimelineId::generate();
        {
            let (mut storage, mut state) =
-                load_from_control_file(&conf, timeline_id, CreateControlFile::True)
+                load_from_control_file(&conf, &zttid, CreateControlFile::True)
                    .expect("failed to read state");
            // change something
            state.wal_start_lsn = Lsn(42);
            storage.persist(&state).expect("failed to persist state");
        }
-        let (_, state) = load_from_control_file(&conf, timeline_id, CreateControlFile::False)
+        let (_, state) = load_from_control_file(&conf, &zttid, CreateControlFile::False)
            .expect("failed to read state");
        assert_eq!(state.wal_start_lsn, Lsn(42));
    }
@@ -780,21 +788,21 @@ mod test {
    #[test]
    fn test_safekeeper_state_checksum_mismatch() {
        let conf = stub_conf();
-        let timeline_id = ZTimelineId::generate();
+        let zttid = ZTenantTimelineId::generate();
        {
            let (mut storage, mut state) =
-                load_from_control_file(&conf, timeline_id, CreateControlFile::True)
+                load_from_control_file(&conf, &zttid, CreateControlFile::True)
                    .expect("failed to read state");
            // change something
            state.wal_start_lsn = Lsn(42);
            storage.persist(&state).expect("failed to persist state");
        }
-        let control_path = conf.timeline_dir(&timeline_id).join(CONTROL_FILE_NAME);
+        let control_path = conf.timeline_dir(&zttid).join(CONTROL_FILE_NAME);
        let mut data = fs::read(&control_path).unwrap();
        data[0] += 1; // change the first byte of the file to fail checksum validation
        fs::write(&control_path, &data).expect("failed to write control file");
-        match load_from_control_file(&conf, timeline_id, CreateControlFile::False) {
+        match load_from_control_file(&conf, &zttid, CreateControlFile::False) {
            Err(err) => assert!(err
                .to_string()
                .contains("safekeeper control file checksum mismatch")),
@@ -1,7 +1,7 @@
 [package]
 name = "workspace_hack"
 version = "0.1.0"
-edition = "2018"
+edition = "2021"
 [target.'cfg(all())'.dependencies]
 libc = { version = "0.2", features = ["default", "extra_traits", "std"] }
@@ -2,7 +2,7 @@
 name = "zenith"
 version = "0.1.0"
 authors = ["Stas Kelvich <stas@zenith.tech>"]
-edition = "2018"
+edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
@@ -10,7 +10,7 @@ edition = "2018"
 clap = "2.33.0"
 anyhow = "1.0"
 serde_json = "1"
-postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
+postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="2949d98df52587d562986aad155dd4e889e408b7" }
 # FIXME: 'pageserver' is needed for BranchInfo. Refactor
 pageserver = { path = "../pageserver" }
@@ -1,5 +1,4 @@
-use anyhow::{anyhow, bail};
+use anyhow::{bail, Context, Result};
 use anyhow::{Context, Result};
 use clap::{App, AppSettings, Arg, ArgMatches, SubCommand};
 use control_plane::compute::ComputeControlPlane;
 use control_plane::local_env;
@@ -271,7 +270,7 @@ fn print_branches_tree(branches: Vec<BranchInfo>) -> Result<()> {
        if let Some(tid) = &branch.ancestor_id {
            branches_hash
                .get_mut(tid)
-                .with_context(|| "missing branch info in the HashMap")?
+                .context("missing branch info in the HashMap")?
                .children
                .push(branch.timeline_id.to_string());
        }
@@ -310,7 +309,7 @@ fn print_branch(
            .info
            .ancestor_lsn
            .as_ref()
-            .with_context(|| "missing branch info in the HashMap")?;
+            .context("missing branch info in the HashMap")?;
        let mut br_sym = "┣━";
        // Draw each nesting padding with proper style
@@ -356,7 +355,7 @@ fn print_branch(
            &is_last_new,
            branches
                .get(child)
-                .with_context(|| "missing branch info in the HashMap")?,
+                .context("missing branch info in the HashMap")?,
            branches,
        )?;
    }
@@ -402,10 +401,10 @@ fn handle_init(init_match: &ArgMatches) -> Result<()> {
        default_conf()
    };
-    let mut env = LocalEnv::create_config(&toml_file)
+    let mut env =
-        .with_context(|| "Failed to create zenith configuration")?;
+        LocalEnv::create_config(&toml_file).context("Failed to create zenith configuration")?;
    env.init()
-        .with_context(|| "Failed to initialize zenith repository")?;
+        .context("Failed to initialize zenith repository")?;
    // Call 'pageserver init'.
    let pageserver = PageServerNode::from_env(&env);
@@ -462,7 +461,7 @@ fn handle_branch(branch_match: &ArgMatches, env: &local_env::LocalEnv) -> Result
    if let Some(branchname) = branch_match.value_of("branchname") {
        let startpoint_str = branch_match
            .value_of("start-point")
-            .ok_or_else(|| anyhow!("Missing start-point"))?;
+            .context("Missing start-point")?;
        let branch = pageserver.branch_create(branchname, startpoint_str, &tenantid)?;
        println!(
            "Created branch '{}' at {:?} for tenant: {}",
@@ -574,7 +573,7 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
            let node = cplane
                .nodes
                .get(&(tenantid, node_name.to_owned()))
-                .ok_or_else(|| anyhow!("postgres {} is not found", node_name))?;
+                .with_context(|| format!("postgres {} is not found", node_name))?;
            node.stop(destroy)?;
        }
@@ -1,7 +1,7 @@
 [package]
 name = "zenith_metrics"
 version = "0.1.0"
-edition = "2018"
+edition = "2021"
 [dependencies]
 prometheus = {version = "0.12", default_features=false} # removes protobuf dependency
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Bojan Serafimov	64fcf4f096	Implement mock console	2022-02-09 14:30:01 -05:00
Dmitry Ivanov	18d3d078ad	[WIP] [proxy] Migrate to async	2022-02-08 05:43:32 +03:00
Andrey Taranik	d69b0539ba	proxy chart staging values update for labels (#1202 )	2022-02-01 13:31:05 +03:00
Dmitry Ivanov	ec78babad2	Use `mold` instead of default linker	2022-01-28 20:40:50 +03:00
Dmitry Ivanov	9350dfb215	[CI] Merge *.profraw files prior to uploading workspace Hopefully, this will make CI pipeline a bit faster.	2022-01-28 19:56:28 +03:00
Dmitry Ivanov	8ac8be5206	[scripts/coverage] Implement `merge` command This will drastically decrease the size of CI workspace uploads.	2022-01-28 19:56:28 +03:00
Dmitry Ivanov	c2927353a5	Enable async deserialization of FeMessage Now it's possible to call Fe{Startup,}Message in both sync and async contexts, which is good for proxy. Co-authored-by: bojanserafimov <bojan.serafimov7@gmail.com>	2022-01-28 19:40:37 +03:00
Kirill Bulatov	33251a9d8f	Disable failing remote storage tests for now	2022-01-28 18:35:46 +03:00
Konstantin Knizhnik	c045ae7a9b	Fix random range for keys in test_gc_aggressive.py (#1199 )	2022-01-28 16:29:55 +03:00
Dmitry Rodionov	602ccb7d5f	distinguish failures for pre-initdb lsn and pre-ancestor lsn branching in test_branch_behind	2022-01-28 12:31:15 +03:00
Dmitry Rodionov	5df21e1058	remove Timeline::start_lsn in favor of ancestor_lsn	2022-01-28 12:31:15 +03:00
Konstantin Knizhnik	08135910a5	Fix checkpoint.nextXid update (#1166 ) * Fix checkpoint.nextXid update * Add test for cehckpoint.nextXid * Fix indentation of test_next_xid.py * Fix mypy error in test_next_xid.py * Tidy up the test case. * Add a unit test Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>	2022-01-27 18:21:51 +03:00
Konstantin Knizhnik	f58a22d07e	Freeze layers at the same end LSN (#1182 ) * Freeze vectors at the same end LSN * Fix calculation of last LSN for inmem layer * Do not advance disk_consistent_lsn is no open layer was evicted * Fix calculation of freeze_end_lsn * Let start_lsn be larger than oldest_pending_lsn * Rename 'oldest_pending_lsn' and 'last_lsn', add comments. * Fix future_layerfiles test * Update comments conserning olest_lsn * Update comments conserning olest_lsn Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>	2022-01-27 18:21:00 +03:00
Arthur Petukhovsky	cedde559b8	Add test for replacement of the failed safekeeper (#1179 ) * Add test to replace failed safekeeper * Restart safekeepers in test_replace_safekeeper * Update vendor/postgres	2022-01-27 17:26:55 +03:00
Arthur Petukhovsky	49d1d1ddf9	Don't call adjust_for_wal_acceptors after pg create (#1178 ) Now zenith_cli handles wal_acceptors config internally, and if we will append wal_acceptors to postgresql.conf in python tests, then it will contain duplicate wal_acceptors config.	2022-01-27 17:23:14 +03:00
Arseny Sher	86045ac36c	Prefix per-cluster directory with ztenant_id in safekeeper. Currently ztimelineids are unique, but all APIs accept the pair, so let's keep it everywhere for uniformity. Carry around ZTTId containing both ZTenantId and ZTimelineId for simplicity. (existing clusters on staging ought to be preprocessed for that)	2022-01-27 17:22:07 +03:00
Konstantin Knizhnik	79f0e44a20	Gc cutoff rwlock (#1139 ) * Reproduce github issue #1047. * Use RwLock to protect gc_cuttof_lsn * Eeduce number of updates in test_gc_aggressive * Change test_prohibit_get_page_at_lsn_for_garbage_collected_pages test * Change test_prohibit_get_page_at_lsn_for_garbage_collected_pages * Lock latest_gc_cutoff_lsn in all operations accessing storage to prevent race conditions with GC * Remove random sleep between wait_for_lsn and get_page_at_lsn * Initialize latest_gc_cutoff with initdb_lsn and remove separate check that lsn >= initdb_lsn * Update test_prohibit_branch_creation_on_pre_initdb_lsn test Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>	2022-01-27 14:41:16 +03:00
anastasia	c44695f34b	bump vendor/postgres	2022-01-27 11:20:45 +03:00
anastasia	5abe2129c6	Extend replication protocol with ZentihFeedback message to pass current_timeline_size to compute node Put standby_status_update fields into ZenithFeedback and send them as one message. Pass values sizes together with keys in ZenithFeedback message.	2022-01-27 11:20:45 +03:00
Dmitry Rodionov	63dd7bce7e	bandaid to avoid concurrent timeline downloading until proper refactoring/fix	2022-01-26 19:54:09 +03:00
Dmitry Rodionov	f3c73f5797	cache python deps in circle ci	2022-01-26 13:01:12 +03:00
Dmitry Rodionov	e6f2d70517	use 2021 rust edition	2022-01-25 18:48:49 +03:00
Andrey Taranik	be6d1cc360	Use zimg as builders (#1165 ) * try use own builder images * add postgres headers before build zenith * checkout submodule before zenith build * circleci cleanup	2022-01-25 00:58:37 +03:00
Dmitry Ivanov	703716228e	Use `&str` instead of `String` in `BeMessage::ErrorResponse` There's no need in allocating string literals in the heap.	2022-01-24 18:49:05 +03:00
Dmitry Rodionov	458bc0c838	walkeeper: use named type as a key in callmemaybe subscriptions hashmap	2022-01-24 17:20:15 +03:00
Dmitry Rodionov	39591ef627	reduce flakiness	2022-01-24 17:20:15 +03:00
Dmitry Rodionov	37c440c5d3	Introduce first version of tenant migraiton between pageservers This patch includes attach/detach http endpoints in pageservers. Some changes in callmemaybe handling inside safekeeper and an integrational test to check migration with and without load. There are still some rough edges that will be addressed in follow up patches	2022-01-24 17:20:15 +03:00
anastasia	81e94d1897	Add LSN and Backpressure descriptions to glossary.md	2022-01-24 12:52:30 +03:00
Konstantin Knizhnik	7bc1274a03	Fix comparison with disk_consistent_lsn in newer_image_layer_exists (#1167 )	2022-01-24 12:19:18 +03:00
Dmitry Rodionov	5f5a11525c	Switch our python package management solution to poetry. Mainly because it has better support for installing the packages from different python versions. It also has better dependency resolver than Pipenv. And supports modern standard for python dependency management. This includes usage of pyproject.toml for project specific configuration instead of per tool conf files. See following links for details: https://pip.pypa.io/en/stable/reference/build-system/pyproject-toml/ https://www.python.org/dev/peps/pep-0518/	2022-01-24 11:33:47 +03:00
Konstantin Knizhnik	e209764877	Do not delete layers beyand cutoff LSN (#1128 ) * Do not delete layers beyand cutoff LSN * Update pageserver/src/layered_repository/layer_map.rs Co-authored-by: Heikki Linnakangas <heikki.linnakangas@iki.fi> Co-authored-by: Heikki Linnakangas <heikki.linnakangas@iki.fi>	2022-01-24 10:42:40 +03:00
Kirill Bulatov	65290b2e96	Ensure every submodule compiles on its own	2022-01-21 17:34:15 +03:00
Dmitry Ivanov	127df96635	[proxy] Make `NUM_BYTES_PROXIED_COUNTER` more precise	2022-01-21 17:31:19 +03:00
Kirill Bulatov	924d8d489a	Allow enabling S3 mock in all existing tests with an env var	2022-01-20 18:42:47 +02:00
Dmitry Rodionov	026eb64a83	Use python lib to mock s3	2022-01-20 18:42:47 +02:00
Kirill Bulatov	45124856b1	Better S3 remote storage logging	2022-01-20 18:42:47 +02:00
Kirill Bulatov	38c6f6ce16	Allow specifying custom endpoint in s3	2022-01-20 18:42:47 +02:00
Heikki Linnakangas	caa62eff2a	Fix description of proxy --auth-endpoint option.	2022-01-20 14:50:27 +03:00
Dmitry Ivanov	d3542c34f1	Refactoring: use anyhow::Context's methods where possible	2022-01-19 16:33:48 +03:00
Kirill Bulatov	7fb62fc849	Fix macos compilation	2022-01-18 23:01:04 +02:00
Andrey Taranik	9d6ae06663	monitoring turn on for proxy (#1146 )	2022-01-18 19:23:53 +03:00
Alexey Kondratov	06c28174c2	Integrate compute_tools into zenith workspace and improve logging (zenithdb/console#487 )	2022-01-18 18:47:31 +03:00
bojanserafimov	8af1b43074	proxy: Add new metrics (#1132 )	2022-01-14 19:12:43 -05:00
Heikki Linnakangas	17b7caddcb	Update vendor/postgres: silence excessive logging from walproposer.	2022-01-14 20:51:02 +02:00
Heikki Linnakangas	dab30c27b6	Refactor thread management and shutdown This introduces a new module to handle thread creation and shutdown. All page server threads are now registered in a global hash map, and there's a function to request individual threads to shut down gracefully. Thread shutdown request is signalled to the thread with a flag, as well as a Future that can be used to wake up async operations if shutdown is requested. Use that facility to have the libpq listener thread respond to pageserver shutdown, based on Kirill's earlier prototype (https://github.com/zenithdb/zenith/pull/1088). That addresses https://github.com/zenithdb/zenith/issues/1036, previously the libpq listener thread would not exit until one more connection arrives. This also eliminates a resource leak in the accept() loop. Previously, we added the JoinHanlde of each new thread to a vector but old handles for threads that had already exited were never removed.	2022-01-14 18:36:10 +02:00
Heikki Linnakangas	bad1dd9759	Don't panic if spawning a new WAL receiver thread fails. The panic would kill the page service thread. That's not too bad, but still let's try to handle it more gracefully.	2022-01-14 18:02:34 +02:00
Heikki Linnakangas	d29836d0d5	Don't panic if spawning a thread to handle a connection fails. Log the error and continue. Hopefully it's a transient failure. This might have been happening in staging earlier, when the safekeeper had a problem where it opened connections very frequently to issue "callmemaybe" commands. If you launch too many threads too fast, you might run out of file descriptors or something. It's not totally clear what happened, but with commit, at least the page server will continue to run and accept new connections, if a transient error happens.	2022-01-14 18:02:30 +02:00
Heikki Linnakangas	adb0b3dada	Include backtrace in error messages in the log. 'anyhow' crate can include a backtrace in all errors, when the 'backtrace' feature is enabled. Enable it, and change the places that used '{:#}' or '{}' to '{:?}', so that the backtrace is printed.	2022-01-14 10:10:17 +02:00