rename fn

.config/hakari.toml

@@ -23,30 +23,10 @@ platforms = [
 ]
 
 [final-excludes]
-workspace-members = [
-    # vm_monitor benefits from the same Cargo.lock as the rest of our artifacts, but
-    # it is built primarly in separate repo neondatabase/autoscaling and thus is excluded
-    # from depending on workspace-hack because most of the dependencies are not used.
-    "vm_monitor",
-    # All of these exist in libs and are not usually built independently.
-    # Putting workspace hack there adds a bottleneck for cargo builds.
-    "compute_api",
-    "consumption_metrics",
-    "desim",
-    "metrics",
-    "pageserver_api",
-    "postgres_backend",
-    "postgres_connection",
-    "postgres_ffi",
-    "pq_proto",
-    "remote_storage",
-    "safekeeper_api",
-    "tenant_size_model",
-    "tracing-utils",
-    "utils",
-    "wal_craft",
-    "walproposer",
-]
+# vm_monitor benefits from the same Cargo.lock as the rest of our artifacts, but
+# it is built primarly in separate repo neondatabase/autoscaling and thus is excluded
+# from depending on workspace-hack because most of the dependencies are not used.
+workspace-members = ["vm_monitor"]
 
 # Write out exact versions rather than a semver range. (Defaults to false.)
 # exact-versions = true

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

@@ -43,7 +43,7 @@ inputs:
   pg_version:
     description: 'Postgres version to use for tests'
     required: false
-    default: 'v16'
+    default: 'v14'
   benchmark_durations:
     description: 'benchmark durations JSON'
     required: false
@@ -169,8 +169,10 @@ runs:
           EXTRA_PARAMS="--durations-path $TEST_OUTPUT/benchmark_durations.json $EXTRA_PARAMS"
         fi
 
-        if [[ $BUILD_TYPE == "debug" && $RUNNER_ARCH == 'X64' ]]; then
+        if [[ "${{ inputs.build_type }}" == "debug" ]]; then
           cov_prefix=(scripts/coverage "--profraw-prefix=$GITHUB_JOB" --dir=/tmp/coverage run)
+        elif [[ "${{ inputs.build_type }}" == "release" ]]; then
+          cov_prefix=()
         else
           cov_prefix=()
         fi

.github/workflows/_benchmarking_preparation.yml

@@ -48,8 +48,6 @@ jobs:
 
         echo "connstr=${CONNSTR}" >> $GITHUB_OUTPUT  
 
-    - uses: actions/checkout@v4
-
     - name: Download Neon artifact
       uses: ./.github/actions/download
       with:

.github/workflows/_build-and-test-locally.yml

@@ -94,16 +94,11 @@ jobs:
       # We run tests with addtional features, that are turned off by default (e.g. in release builds), see
       # corresponding Cargo.toml files for their descriptions.
       - name: Set env variables
-        env:
-          ARCH: ${{ inputs.arch }}
         run: |
           CARGO_FEATURES="--features testing"
-          if [[ $BUILD_TYPE == "debug" && $ARCH == 'x64' ]]; then
+          if [[ $BUILD_TYPE == "debug" ]]; then
             cov_prefix="scripts/coverage --profraw-prefix=$GITHUB_JOB --dir=/tmp/coverage run"
             CARGO_FLAGS="--locked"
-          elif [[ $BUILD_TYPE == "debug" ]]; then
-            cov_prefix=""
-            CARGO_FLAGS="--locked"
           elif [[ $BUILD_TYPE == "release" ]]; then
             cov_prefix=""
             CARGO_FLAGS="--locked --release"
@@ -163,8 +158,6 @@ jobs:
       # Do install *before* running rust tests because they might recompile the
       # binaries with different features/flags.
       - name: Install rust binaries
-        env:
-          ARCH: ${{ inputs.arch }}
         run: |
           # Install target binaries
           mkdir -p /tmp/neon/bin/
@@ -179,7 +172,7 @@ jobs:
           done
 
           # Install test executables and write list of all binaries (for code coverage)
-          if [[ $BUILD_TYPE == "debug" && $ARCH == 'x64' ]]; then
+          if [[ $BUILD_TYPE == "debug" ]]; then
             # Keep bloated coverage data files away from the rest of the artifact
             mkdir -p /tmp/coverage/
 
@@ -250,8 +243,8 @@ jobs:
         uses: ./.github/actions/save-coverage-data
 
   regress-tests:
-    # Don't run regression tests on debug arm64 builds
-    if: inputs.build-type != 'debug' || inputs.arch != 'arm64'
+    # Run test on x64 only
+    if: inputs.arch == 'x64'
     needs: [ build-neon ]
     runs-on: ${{ fromJson(format('["self-hosted", "{0}"]', inputs.arch == 'arm64' && 'large-arm64' || 'large')) }}
     container:

.github/workflows/build_and_test.yml

@@ -198,7 +198,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        arch: [ x64, arm64 ]
+        arch: [ x64 ]
         # Do not build or run tests in debug for release branches
         build-type: ${{ fromJson((startsWith(github.ref_name, 'release') && github.event_name == 'push') && '["release"]' || '["debug", "release"]') }}
         include:
@@ -280,7 +280,6 @@ jobs:
           save_perf_report: ${{ github.ref_name == 'main' }}
           extra_params: --splits 5 --group ${{ matrix.pytest_split_group }}
           benchmark_durations: ${{ needs.get-benchmarks-durations.outputs.json }}
-          pg_version: v16
         env:
           VIP_VAP_ACCESS_TOKEN: "${{ secrets.VIP_VAP_ACCESS_TOKEN }}"
           PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
@@ -986,10 +985,10 @@ jobs:
           GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
         run: |
           if [[ "$GITHUB_REF_NAME" == "main" ]]; then
-            gh workflow --repo neondatabase/infra run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=false
+            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=false
             gh workflow --repo neondatabase/azure run deploy.yml -f dockerTag=${{needs.tag.outputs.build-tag}}
           elif [[ "$GITHUB_REF_NAME" == "release" ]]; then
-            gh workflow --repo neondatabase/infra run deploy-dev.yml --ref main \
+            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main \
               -f deployPgSniRouter=false \
               -f deployProxy=false \
               -f deployStorage=true \
@@ -999,14 +998,14 @@ jobs:
               -f dockerTag=${{needs.tag.outputs.build-tag}} \
               -f deployPreprodRegion=true
 
-            gh workflow --repo neondatabase/infra run deploy-prod.yml --ref main \
+            gh workflow --repo neondatabase/aws run deploy-prod.yml --ref main \
               -f deployStorage=true \
               -f deployStorageBroker=true \
               -f deployStorageController=true \
               -f branch=main \
               -f dockerTag=${{needs.tag.outputs.build-tag}}
           elif [[ "$GITHUB_REF_NAME" == "release-proxy" ]]; then
-            gh workflow --repo neondatabase/infra run deploy-dev.yml --ref main \
+            gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main \
               -f deployPgSniRouter=true \
               -f deployProxy=true \
               -f deployStorage=false \
@@ -1016,7 +1015,7 @@ jobs:
               -f dockerTag=${{needs.tag.outputs.build-tag}} \
               -f deployPreprodRegion=true
 
-            gh workflow --repo neondatabase/infra run deploy-proxy-prod.yml --ref main \
+            gh workflow --repo neondatabase/aws run deploy-proxy-prod.yml --ref main \
               -f deployPgSniRouter=true \
               -f deployProxy=true \
               -f branch=main \

Cargo.lock

@@ -1208,6 +1208,7 @@ dependencies = [
  "serde_json",
  "serde_with",
  "utils",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -1320,6 +1321,7 @@ dependencies = [
  "serde",
  "serde_with",
  "utils",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -1668,6 +1670,7 @@ dependencies = [
  "smallvec",
  "tracing",
  "utils",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -3144,6 +3147,7 @@ dependencies = [
  "rand 0.8.5",
  "rand_distr",
  "twox-hash",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -3787,6 +3791,7 @@ dependencies = [
  "strum_macros",
  "thiserror",
  "utils",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -4188,6 +4193,7 @@ dependencies = [
  "tokio-rustls 0.25.0",
  "tokio-util",
  "tracing",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -4200,6 +4206,7 @@ dependencies = [
  "postgres",
  "tokio-postgres",
  "url",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -4222,6 +4229,7 @@ dependencies = [
  "serde",
  "thiserror",
  "utils",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -4259,6 +4267,7 @@ dependencies = [
  "thiserror",
  "tokio",
  "tracing",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -4823,6 +4832,7 @@ dependencies = [
  "toml_edit 0.19.10",
  "tracing",
  "utils",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -5347,6 +5357,7 @@ dependencies = [
  "serde",
  "serde_with",
  "utils",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -6182,6 +6193,7 @@ dependencies = [
  "anyhow",
  "serde",
  "serde_json",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -6782,6 +6794,7 @@ dependencies = [
  "tracing",
  "tracing-opentelemetry",
  "tracing-subscriber",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -6999,6 +7012,7 @@ dependencies = [
  "url",
  "uuid",
  "walkdir",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -7077,6 +7091,7 @@ dependencies = [
  "postgres_ffi",
  "regex",
  "utils",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -7097,6 +7112,7 @@ dependencies = [
  "bindgen",
  "postgres_ffi",
  "utils",
+ "workspace_hack",
 ]
 
 [[package]]
@@ -7653,6 +7669,8 @@ dependencies = [
  "tokio",
  "tokio-rustls 0.24.0",
  "tokio-util",
+ "toml_datetime",
+ "toml_edit 0.19.10",
  "tonic",
  "tower",
  "tracing",

README.md

@@ -262,7 +262,7 @@ By default, this runs both debug and release modes, and all supported postgres v
 testing locally, it is convenient to run just one set of permutations, like this:
 
 ```sh
-DEFAULT_PG_VERSION=16 BUILD_TYPE=release ./scripts/pytest
+DEFAULT_PG_VERSION=15 BUILD_TYPE=release ./scripts/pytest
 ```
 
 ## Flamegraphs

control_plane/src/background_process.rs

@@ -379,7 +379,7 @@ where
     }
 }
 
-pub(crate) fn process_has_stopped(pid: Pid) -> anyhow::Result<bool> {
+fn process_has_stopped(pid: Pid) -> anyhow::Result<bool> {
     match kill(pid, None) {
         // Process exists, keep waiting
         Ok(_) => Ok(false),

control_plane/src/bin/neon_local.rs

@@ -15,9 +15,7 @@ use control_plane::local_env::{
 };
 use control_plane::pageserver::PageServerNode;
 use control_plane::safekeeper::SafekeeperNode;
-use control_plane::storage_controller::{
-    NeonStorageControllerStartArgs, NeonStorageControllerStopArgs, StorageController,
-};
+use control_plane::storage_controller::StorageController;
 use control_plane::{broker, local_env};
 use pageserver_api::config::{
     DEFAULT_HTTP_LISTEN_PORT as DEFAULT_PAGESERVER_HTTP_PORT,
@@ -54,7 +52,7 @@ const DEFAULT_PAGESERVER_ID: NodeId = NodeId(1);
 const DEFAULT_BRANCH_NAME: &str = "main";
 project_git_version!(GIT_VERSION);
 
-const DEFAULT_PG_VERSION: &str = "16";
+const DEFAULT_PG_VERSION: &str = "15";
 
 const DEFAULT_PAGESERVER_CONTROL_PLANE_API: &str = "http://127.0.0.1:1234/upcall/v1/";
 
@@ -1054,36 +1052,6 @@ fn get_start_timeout(args: &ArgMatches) -> &Duration {
     humantime_duration.as_ref()
 }
 
-fn storage_controller_start_args(args: &ArgMatches) -> NeonStorageControllerStartArgs {
-    let maybe_instance_id = args.get_one::<u8>("instance-id");
-
-    let base_port = args.get_one::<u16>("base-port");
-
-    if maybe_instance_id.is_some() && base_port.is_none() {
-        panic!("storage-controller start specificied instance-id but did not provide base-port");
-    }
-
-    let start_timeout = args
-        .get_one::<humantime::Duration>("start-timeout")
-        .expect("invalid value for start-timeout");
-
-    NeonStorageControllerStartArgs {
-        instance_id: maybe_instance_id.copied().unwrap_or(1),
-        base_port: base_port.copied(),
-        start_timeout: *start_timeout,
-    }
-}
-
-fn storage_controller_stop_args(args: &ArgMatches) -> NeonStorageControllerStopArgs {
-    let maybe_instance_id = args.get_one::<u8>("instance-id");
-    let immediate = args.get_one::<String>("stop-mode").map(|s| s.as_str()) == Some("immediate");
-
-    NeonStorageControllerStopArgs {
-        instance_id: maybe_instance_id.copied().unwrap_or(1),
-        immediate,
-    }
-}
-
 async fn handle_pageserver(sub_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
     match sub_match.subcommand() {
         Some(("start", subcommand_args)) => {
@@ -1145,14 +1113,19 @@ async fn handle_storage_controller(
     let svc = StorageController::from_env(env);
     match sub_match.subcommand() {
         Some(("start", start_match)) => {
-            if let Err(e) = svc.start(storage_controller_start_args(start_match)).await {
+            if let Err(e) = svc.start(get_start_timeout(start_match)).await {
                 eprintln!("start failed: {e}");
                 exit(1);
             }
         }
 
         Some(("stop", stop_match)) => {
-            if let Err(e) = svc.stop(storage_controller_stop_args(stop_match)).await {
+            let immediate = stop_match
+                .get_one::<String>("stop-mode")
+                .map(|s| s.as_str())
+                == Some("immediate");
+
+            if let Err(e) = svc.stop(immediate).await {
                 eprintln!("stop failed: {}", e);
                 exit(1);
             }
@@ -1255,12 +1228,7 @@ async fn handle_start_all(
     // Only start the storage controller if the pageserver is configured to need it
     if env.control_plane_api.is_some() {
         let storage_controller = StorageController::from_env(env);
-        if let Err(e) = storage_controller
-            .start(NeonStorageControllerStartArgs::with_default_instance_id(
-                (*retry_timeout).into(),
-            ))
-            .await
-        {
+        if let Err(e) = storage_controller.start(retry_timeout).await {
             eprintln!("storage_controller start failed: {:#}", e);
             try_stop_all(env, true).await;
             exit(1);
@@ -1390,21 +1358,10 @@ async fn try_stop_all(env: &local_env::LocalEnv, immediate: bool) {
         eprintln!("neon broker stop failed: {e:#}");
     }
 
-    // Stop all storage controller instances. In the most common case there's only one,
-    // but iterate though the base data directory in order to discover the instances.
-    let storcon_instances = env
-        .storage_controller_instances()
-        .await
-        .expect("Must inspect data dir");
-    for (instance_id, _instance_dir_path) in storcon_instances {
+    if env.control_plane_api.is_some() {
         let storage_controller = StorageController::from_env(env);
-        let stop_args = NeonStorageControllerStopArgs {
-            instance_id,
-            immediate,
-        };
-
-        if let Err(e) = storage_controller.stop(stop_args).await {
-            eprintln!("Storage controller instance {instance_id} stop failed: {e:#}");
+        if let Err(e) = storage_controller.stop(immediate).await {
+            eprintln!("storage controller stop failed: {e:#}");
         }
     }
 }
@@ -1544,18 +1501,6 @@ fn cli() -> Command {
         .action(ArgAction::SetTrue)
         .required(false);
 
-    let instance_id = Arg::new("instance-id")
-        .long("instance-id")
-        .help("Identifier used to distinguish storage controller instances (default 1)")
-        .value_parser(value_parser!(u8))
-        .required(false);
-
-    let base_port = Arg::new("base-port")
-        .long("base-port")
-        .help("Base port for the storage controller instance idenfified by instance-id (defaults to pagserver cplane api)")
-        .value_parser(value_parser!(u16))
-        .required(false);
-
     Command::new("Neon CLI")
         .arg_required_else_help(true)
         .version(GIT_VERSION)
@@ -1664,12 +1609,9 @@ fn cli() -> Command {
                 .arg_required_else_help(true)
                 .about("Manage storage_controller")
                 .subcommand(Command::new("start").about("Start storage controller")
-                            .arg(timeout_arg.clone())
-                            .arg(instance_id.clone())
-                            .arg(base_port))
+                            .arg(timeout_arg.clone()))
                 .subcommand(Command::new("stop").about("Stop storage controller")
-                            .arg(stop_mode_arg.clone())
-                            .arg(instance_id))
+                            .arg(stop_mode_arg.clone()))
         )
         .subcommand(
             Command::new("safekeeper")

control_plane/src/local_env.rs

@@ -27,7 +27,7 @@ use crate::pageserver::PageServerNode;
 use crate::pageserver::PAGESERVER_REMOTE_STORAGE_DIR;
 use crate::safekeeper::SafekeeperNode;
 
-pub const DEFAULT_PG_VERSION: u32 = 16;
+pub const DEFAULT_PG_VERSION: u32 = 15;
 
 //
 // This data structures represents neon_local CLI config
@@ -156,11 +156,6 @@ pub struct NeonStorageControllerConf {
     #[serde(with = "humantime_serde")]
     pub max_warming_up: Duration,
 
-    pub start_as_candidate: bool,
-
-    /// Database url used when running multiple storage controller instances
-    pub database_url: Option<SocketAddr>,
-
     /// Threshold for auto-splitting a tenant into shards
     pub split_threshold: Option<u64>,
 
@@ -179,8 +174,6 @@ impl Default for NeonStorageControllerConf {
         Self {
             max_offline: Self::DEFAULT_MAX_OFFLINE_INTERVAL,
             max_warming_up: Self::DEFAULT_MAX_WARMING_UP_INTERVAL,
-            start_as_candidate: false,
-            database_url: None,
             split_threshold: None,
             max_secondary_lag_bytes: None,
         }
@@ -399,36 +392,6 @@ impl LocalEnv {
         }
     }
 
-    /// Inspect the base data directory and extract the instance id and instance directory path
-    /// for all storage controller instances
-    pub async fn storage_controller_instances(&self) -> std::io::Result<Vec<(u8, PathBuf)>> {
-        let mut instances = Vec::default();
-
-        let dir = std::fs::read_dir(self.base_data_dir.clone())?;
-        for dentry in dir {
-            let dentry = dentry?;
-            let is_dir = dentry.metadata()?.is_dir();
-            let filename = dentry.file_name().into_string().unwrap();
-            let parsed_instance_id = match filename.strip_prefix("storage_controller_") {
-                Some(suffix) => suffix.parse::<u8>().ok(),
-                None => None,
-            };
-
-            let is_instance_dir = is_dir && parsed_instance_id.is_some();
-
-            if !is_instance_dir {
-                continue;
-            }
-
-            instances.push((
-                parsed_instance_id.expect("Checked previously"),
-                dentry.path(),
-            ));
-        }
-
-        Ok(instances)
-    }
-
     pub fn register_branch_mapping(
         &mut self,
         branch_name: String,

control_plane/src/storage_controller.rs

@@ -3,8 +3,6 @@ use crate::{
     local_env::{LocalEnv, NeonStorageControllerConf},
 };
 use camino::{Utf8Path, Utf8PathBuf};
-use hyper::Uri;
-use nix::unistd::Pid;
 use pageserver_api::{
     controller_api::{
         NodeConfigureRequest, NodeDescribeResponse, NodeRegisterRequest, TenantCreateRequest,
@@ -20,7 +18,7 @@ use pageserver_client::mgmt_api::ResponseErrorMessageExt;
 use postgres_backend::AuthType;
 use reqwest::Method;
 use serde::{de::DeserializeOwned, Deserialize, Serialize};
-use std::{fs, net::SocketAddr, path::PathBuf, str::FromStr, sync::OnceLock};
+use std::{fs, str::FromStr, time::Duration};
 use tokio::process::Command;
 use tracing::instrument;
 use url::Url;
@@ -31,14 +29,12 @@ use utils::{
 
 pub struct StorageController {
     env: LocalEnv,
+    listen: String,
     private_key: Option<Vec<u8>>,
     public_key: Option<String>,
+    postgres_port: u16,
     client: reqwest::Client,
     config: NeonStorageControllerConf,
-
-    // The listen addresses is learned when starting the storage controller,
-    // hence the use of OnceLock to init it at the right time.
-    listen: OnceLock<SocketAddr>,
 }
 
 const COMMAND: &str = "storage_controller";
@@ -47,36 +43,6 @@ const STORAGE_CONTROLLER_POSTGRES_VERSION: u32 = 16;
 
 const DB_NAME: &str = "storage_controller";
 
-pub struct NeonStorageControllerStartArgs {
-    pub instance_id: u8,
-    pub base_port: Option<u16>,
-    pub start_timeout: humantime::Duration,
-}
-
-impl NeonStorageControllerStartArgs {
-    pub fn with_default_instance_id(start_timeout: humantime::Duration) -> Self {
-        Self {
-            instance_id: 1,
-            base_port: None,
-            start_timeout,
-        }
-    }
-}
-
-pub struct NeonStorageControllerStopArgs {
-    pub instance_id: u8,
-    pub immediate: bool,
-}
-
-impl NeonStorageControllerStopArgs {
-    pub fn with_default_instance_id(immediate: bool) -> Self {
-        Self {
-            instance_id: 1,
-            immediate,
-        }
-    }
-}
-
 #[derive(Serialize, Deserialize)]
 pub struct AttachHookRequest {
     pub tenant_shard_id: TenantShardId,
@@ -101,6 +67,23 @@ pub struct InspectResponse {
 
 impl StorageController {
     pub fn from_env(env: &LocalEnv) -> Self {
+        // Makes no sense to construct this if pageservers aren't going to use it: assume
+        // pageservers have control plane API set
+        let listen_url = env.control_plane_api.clone().unwrap();
+
+        let listen = format!(
+            "{}:{}",
+            listen_url.host_str().unwrap(),
+            listen_url.port().unwrap()
+        );
+
+        // Convention: NeonEnv in python tests reserves the next port after the control_plane_api
+        // port, for use by our captive postgres.
+        let postgres_port = listen_url
+            .port()
+            .expect("Control plane API setting should always have a port")
+            + 1;
+
         // Assume all pageservers have symmetric auth configuration: this service
         // expects to use one JWT token to talk to all of them.
         let ps_conf = env
@@ -143,28 +126,20 @@ impl StorageController {
 
         Self {
             env: env.clone(),
+            listen,
             private_key,
             public_key,
+            postgres_port,
             client: reqwest::ClientBuilder::new()
                 .build()
                 .expect("Failed to construct http client"),
             config: env.storage_controller.clone(),
-            listen: OnceLock::default(),
         }
     }
 
-    fn storage_controller_instance_dir(&self, instance_id: u8) -> PathBuf {
-        self.env
-            .base_data_dir
-            .join(format!("storage_controller_{}", instance_id))
-    }
-
-    fn pid_file(&self, instance_id: u8) -> Utf8PathBuf {
-        Utf8PathBuf::from_path_buf(
-            self.storage_controller_instance_dir(instance_id)
-                .join("storage_controller.pid"),
-        )
-        .expect("non-Unicode path")
+    fn pid_file(&self) -> Utf8PathBuf {
+        Utf8PathBuf::from_path_buf(self.env.base_data_dir.join("storage_controller.pid"))
+            .expect("non-Unicode path")
     }
 
     /// PIDFile for the postgres instance used to store storage controller state
@@ -209,23 +184,23 @@ impl StorageController {
     }
 
     /// Readiness check for our postgres process
-    async fn pg_isready(&self, pg_bin_dir: &Utf8Path, postgres_port: u16) -> anyhow::Result<bool> {
+    async fn pg_isready(&self, pg_bin_dir: &Utf8Path) -> anyhow::Result<bool> {
         let bin_path = pg_bin_dir.join("pg_isready");
-        let args = ["-h", "localhost", "-p", &format!("{}", postgres_port)];
+        let args = ["-h", "localhost", "-p", &format!("{}", self.postgres_port)];
         let exitcode = Command::new(bin_path).args(args).spawn()?.wait().await?;
 
         Ok(exitcode.success())
     }
 
-    /// Create our database if it doesn't exist
+    /// Create our database if it doesn't exist, and run migrations.
     ///
     /// This function is equivalent to the `diesel setup` command in the diesel CLI.  We implement
     /// the same steps by hand to avoid imposing a dependency on installing diesel-cli for developers
     /// who just want to run `cargo neon_local` without knowing about diesel.
     ///
     /// Returns the database url
-    pub async fn setup_database(&self, postgres_port: u16) -> anyhow::Result<String> {
-        let database_url = format!("postgresql://localhost:{}/{DB_NAME}", postgres_port);
+    pub async fn setup_database(&self) -> anyhow::Result<String> {
+        let database_url = format!("postgresql://localhost:{}/{DB_NAME}", self.postgres_port);
 
         let pg_bin_dir = self.get_pg_bin_dir().await?;
         let createdb_path = pg_bin_dir.join("createdb");
@@ -234,7 +209,7 @@ impl StorageController {
                 "-h",
                 "localhost",
                 "-p",
-                &format!("{}", postgres_port),
+                &format!("{}", self.postgres_port),
                 DB_NAME,
             ])
             .output()
@@ -255,14 +230,13 @@ impl StorageController {
 
     pub async fn connect_to_database(
         &self,
-        postgres_port: u16,
     ) -> anyhow::Result<(
         tokio_postgres::Client,
         tokio_postgres::Connection<tokio_postgres::Socket, tokio_postgres::tls::NoTlsStream>,
     )> {
         tokio_postgres::Config::new()
             .host("localhost")
-            .port(postgres_port)
+            .port(self.postgres_port)
             // The user is the ambient operating system user name.
             // That is an impurity which we want to fix in => TODO https://github.com/neondatabase/neon/issues/8400
             //
@@ -278,114 +252,72 @@ impl StorageController {
             .map_err(anyhow::Error::new)
     }
 
-    pub async fn start(&self, start_args: NeonStorageControllerStartArgs) -> anyhow::Result<()> {
-        let instance_dir = self.storage_controller_instance_dir(start_args.instance_id);
-        if let Err(err) = tokio::fs::create_dir(&instance_dir).await {
-            if err.kind() != std::io::ErrorKind::AlreadyExists {
-                panic!("Failed to create instance dir {instance_dir:?}");
-            }
-        }
+    pub async fn start(&self, retry_timeout: &Duration) -> anyhow::Result<()> {
+        // Start a vanilla Postgres process used by the storage controller for persistence.
+        let pg_data_path = Utf8PathBuf::from_path_buf(self.env.base_data_dir.clone())
+            .unwrap()
+            .join("storage_controller_db");
+        let pg_bin_dir = self.get_pg_bin_dir().await?;
+        let pg_lib_dir = self.get_pg_lib_dir().await?;
+        let pg_log_path = pg_data_path.join("postgres.log");
 
-        let (listen, postgres_port) = {
-            if let Some(base_port) = start_args.base_port {
-                (
-                    format!("127.0.0.1:{base_port}"),
-                    self.config
-                        .database_url
-                        .expect("--base-port requires NeonStorageControllerConf::database_url")
-                        .port(),
-                )
-            } else {
-                let listen_url = self.env.control_plane_api.clone().unwrap();
-
-                let listen = format!(
-                    "{}:{}",
-                    listen_url.host_str().unwrap(),
-                    listen_url.port().unwrap()
-                );
-
-                (listen, listen_url.port().unwrap() + 1)
+        if !tokio::fs::try_exists(&pg_data_path).await? {
+            // Initialize empty database
+            let initdb_path = pg_bin_dir.join("initdb");
+            let mut child = Command::new(&initdb_path)
+                .envs(vec![
+                    ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
+                    ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
+                ])
+                .args(["-D", pg_data_path.as_ref()])
+                .spawn()
+                .expect("Failed to spawn initdb");
+            let status = child.wait().await?;
+            if !status.success() {
+                anyhow::bail!("initdb failed with status {status}");
             }
         };
 
-        let socket_addr = listen
-            .parse()
-            .expect("listen address is a valid socket address");
-        self.listen
-            .set(socket_addr)
-            .expect("StorageController::listen is only set here");
+        // Write a minimal config file:
+        // - Specify the port, since this is chosen dynamically
+        // - Switch off fsync, since we're running on lightweight test environments and when e.g. scale testing
+        //   the storage controller we don't want a slow local disk to interfere with that.
+        //
+        // NB: it's important that we rewrite this file on each start command so we propagate changes
+        // from `LocalEnv`'s config file (`.neon/config`).
+        tokio::fs::write(
+            &pg_data_path.join("postgresql.conf"),
+            format!("port = {}\nfsync=off\n", self.postgres_port),
+        )
+        .await?;
 
-        // Do we remove the pid file on stop?
-        let pg_started = self.is_postgres_running().await?;
-        let pg_lib_dir = self.get_pg_lib_dir().await?;
+        println!("Starting storage controller database...");
+        let db_start_args = [
+            "-w",
+            "-D",
+            pg_data_path.as_ref(),
+            "-l",
+            pg_log_path.as_ref(),
+            "start",
+        ];
 
-        if !pg_started {
-            // Start a vanilla Postgres process used by the storage controller for persistence.
-            let pg_data_path = Utf8PathBuf::from_path_buf(self.env.base_data_dir.clone())
-                .unwrap()
-                .join("storage_controller_db");
-            let pg_bin_dir = self.get_pg_bin_dir().await?;
-            let pg_log_path = pg_data_path.join("postgres.log");
+        background_process::start_process(
+            "storage_controller_db",
+            &self.env.base_data_dir,
+            pg_bin_dir.join("pg_ctl").as_std_path(),
+            db_start_args,
+            vec![
+                ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
+                ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
+            ],
+            background_process::InitialPidFile::Create(self.postgres_pid_file()),
+            retry_timeout,
+            || self.pg_isready(&pg_bin_dir),
+        )
+        .await?;
 
-            if !tokio::fs::try_exists(&pg_data_path).await? {
-                // Initialize empty database
-                let initdb_path = pg_bin_dir.join("initdb");
-                let mut child = Command::new(&initdb_path)
-                    .envs(vec![
-                        ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
-                        ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
-                    ])
-                    .args(["-D", pg_data_path.as_ref()])
-                    .spawn()
-                    .expect("Failed to spawn initdb");
-                let status = child.wait().await?;
-                if !status.success() {
-                    anyhow::bail!("initdb failed with status {status}");
-                }
-            };
-
-            // Write a minimal config file:
-            // - Specify the port, since this is chosen dynamically
-            // - Switch off fsync, since we're running on lightweight test environments and when e.g. scale testing
-            //   the storage controller we don't want a slow local disk to interfere with that.
-            //
-            // NB: it's important that we rewrite this file on each start command so we propagate changes
-            // from `LocalEnv`'s config file (`.neon/config`).
-            tokio::fs::write(
-                &pg_data_path.join("postgresql.conf"),
-                format!("port = {}\nfsync=off\n", postgres_port),
-            )
-            .await?;
-
-            println!("Starting storage controller database...");
-            let db_start_args = [
-                "-w",
-                "-D",
-                pg_data_path.as_ref(),
-                "-l",
-                pg_log_path.as_ref(),
-                "start",
-            ];
-
-            background_process::start_process(
-                "storage_controller_db",
-                &self.env.base_data_dir,
-                pg_bin_dir.join("pg_ctl").as_std_path(),
-                db_start_args,
-                vec![
-                    ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
-                    ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
-                ],
-                background_process::InitialPidFile::Create(self.postgres_pid_file()),
-                &start_args.start_timeout,
-                || self.pg_isready(&pg_bin_dir, postgres_port),
-            )
-            .await?;
-
-            self.setup_database(postgres_port).await?;
-        }
-
-        let database_url = format!("postgresql://localhost:{}/{DB_NAME}", postgres_port);
+        // Run migrations on every startup, in case something changed.
+        let database_url = self.setup_database().await?;
 
         // We support running a startup SQL script to fiddle with the database before we launch storcon.
         // This is used by the test suite.
@@ -407,7 +339,7 @@ impl StorageController {
                 }
             }
         };
-        let (mut client, conn) = self.connect_to_database(postgres_port).await?;
+        let (mut client, conn) = self.connect_to_database().await?;
         let conn = tokio::spawn(conn);
         let tx = client.build_transaction();
         let tx = tx.start().await?;
@@ -416,20 +348,9 @@ impl StorageController {
         drop(client);
         conn.await??;
 
-        let listen = self
-            .listen
-            .get()
-            .expect("cell is set earlier in this function");
-        let address_for_peers = Uri::builder()
-            .scheme("http")
-            .authority(format!("{}:{}", listen.ip(), listen.port()))
-            .path_and_query("")
-            .build()
-            .unwrap();
-
         let mut args = vec![
             "-l",
-            &listen.to_string(),
+            &self.listen,
             "--dev",
             "--database-url",
             &database_url,
@@ -437,27 +358,15 @@ impl StorageController {
             &humantime::Duration::from(self.config.max_offline).to_string(),
             "--max-warming-up-interval",
             &humantime::Duration::from(self.config.max_warming_up).to_string(),
-            "--address-for-peers",
-            &address_for_peers.to_string(),
         ]
         .into_iter()
         .map(|s| s.to_string())
         .collect::<Vec<_>>();
-
-        if self.config.start_as_candidate {
-            args.push("--start-as-candidate".to_string());
-        }
-
         if let Some(private_key) = &self.private_key {
             let claims = Claims::new(None, Scope::PageServerApi);
             let jwt_token =
                 encode_from_key_file(&claims, private_key).expect("failed to generate jwt token");
             args.push(format!("--jwt-token={jwt_token}"));
-
-            let peer_claims = Claims::new(None, Scope::Admin);
-            let peer_jwt_token = encode_from_key_file(&peer_claims, private_key)
-                .expect("failed to generate jwt token");
-            args.push(format!("--peer-jwt-token={peer_jwt_token}"));
         }
 
         if let Some(public_key) = &self.public_key {
@@ -485,15 +394,15 @@ impl StorageController {
 
         background_process::start_process(
             COMMAND,
-            &instance_dir,
+            &self.env.base_data_dir,
             &self.env.storage_controller_bin(),
             args,
             vec![
                 ("LD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
                 ("DYLD_LIBRARY_PATH".to_owned(), pg_lib_dir.to_string()),
             ],
-            background_process::InitialPidFile::Create(self.pid_file(start_args.instance_id)),
-            &start_args.start_timeout,
+            background_process::InitialPidFile::Create(self.pid_file()),
+            retry_timeout,
             || async {
                 match self.ready().await {
                     Ok(_) => Ok(true),
@@ -506,35 +415,8 @@ impl StorageController {
         Ok(())
     }
 
-    pub async fn stop(&self, stop_args: NeonStorageControllerStopArgs) -> anyhow::Result<()> {
-        background_process::stop_process(
-            stop_args.immediate,
-            COMMAND,
-            &self.pid_file(stop_args.instance_id),
-        )?;
-
-        let storcon_instances = self.env.storage_controller_instances().await?;
-        for (instance_id, instanced_dir_path) in storcon_instances {
-            if instance_id == stop_args.instance_id {
-                continue;
-            }
-
-            let pid_file = instanced_dir_path.join("storage_controller.pid");
-            let pid = tokio::fs::read_to_string(&pid_file)
-                .await
-                .map_err(|err| {
-                    anyhow::anyhow!("Failed to read storcon pid file at {pid_file:?}: {err}")
-                })?
-                .parse::<i32>()
-                .expect("pid is valid i32");
-
-            let other_proc_alive = !background_process::process_has_stopped(Pid::from_raw(pid))?;
-            if other_proc_alive {
-                // There is another storage controller instance running, so we return
-                // and leave the database running.
-                return Ok(());
-            }
-        }
+    pub async fn stop(&self, immediate: bool) -> anyhow::Result<()> {
+        background_process::stop_process(immediate, COMMAND, &self.pid_file())?;
 
         let pg_data_path = self.env.base_data_dir.join("storage_controller_db");
         let pg_bin_dir = self.get_pg_bin_dir().await?;
@@ -547,51 +429,27 @@ impl StorageController {
             .wait()
             .await?;
         if !stop_status.success() {
-            match self.is_postgres_running().await {
-                Ok(false) => {
-                    println!("Storage controller database is already stopped");
-                    return Ok(());
-                }
-                Ok(true) => {
-                    anyhow::bail!("Failed to stop storage controller database");
-                }
-                Err(err) => {
-                    anyhow::bail!("Failed to stop storage controller database: {err}");
-                }
+            let pg_status_args = ["-D", &pg_data_path.to_string_lossy(), "status"];
+            let status_exitcode = Command::new(pg_bin_dir.join("pg_ctl"))
+                .args(pg_status_args)
+                .spawn()?
+                .wait()
+                .await?;
+
+            // pg_ctl status returns this exit code if postgres is not running: in this case it is
+            // fine that stop failed.  Otherwise it is an error that stop failed.
+            const PG_STATUS_NOT_RUNNING: i32 = 3;
+            if Some(PG_STATUS_NOT_RUNNING) == status_exitcode.code() {
+                println!("Storage controller database is already stopped");
+                return Ok(());
+            } else {
+                anyhow::bail!("Failed to stop storage controller database: {stop_status}")
             }
         }
 
         Ok(())
     }
 
-    async fn is_postgres_running(&self) -> anyhow::Result<bool> {
-        let pg_data_path = self.env.base_data_dir.join("storage_controller_db");
-        let pg_bin_dir = self.get_pg_bin_dir().await?;
-
-        let pg_status_args = ["-D", &pg_data_path.to_string_lossy(), "status"];
-        let status_exitcode = Command::new(pg_bin_dir.join("pg_ctl"))
-            .args(pg_status_args)
-            .spawn()?
-            .wait()
-            .await?;
-
-        // pg_ctl status returns this exit code if postgres is not running: in this case it is
-        // fine that stop failed.  Otherwise it is an error that stop failed.
-        const PG_STATUS_NOT_RUNNING: i32 = 3;
-        const PG_NO_DATA_DIR: i32 = 4;
-        const PG_STATUS_RUNNING: i32 = 0;
-        match status_exitcode.code() {
-            Some(PG_STATUS_NOT_RUNNING) => Ok(false),
-            Some(PG_NO_DATA_DIR) => Ok(false),
-            Some(PG_STATUS_RUNNING) => Ok(true),
-            Some(code) => Err(anyhow::anyhow!(
-                "pg_ctl status returned unexpected status code: {:?}",
-                code
-            )),
-            None => Err(anyhow::anyhow!("pg_ctl status returned no status code")),
-        }
-    }
-
     fn get_claims_for_path(path: &str) -> anyhow::Result<Option<Claims>> {
         let category = match path.find('/') {
             Some(idx) => &path[..idx],
@@ -617,31 +475,15 @@ impl StorageController {
         RQ: Serialize + Sized,
         RS: DeserializeOwned + Sized,
     {
-        // In the special case of the `storage_controller start` subcommand, we wish
-        // to use the API endpoint of the newly started storage controller in order
-        // to pass the readiness check. In this scenario [`Self::listen`] will be set
-        // (see [`Self::start`]).
-        //
-        // Otherwise, we infer the storage controller api endpoint from the configured
-        // control plane API.
-        let url = if let Some(socket_addr) = self.listen.get() {
-            Url::from_str(&format!(
-                "http://{}:{}/{path}",
-                socket_addr.ip().to_canonical(),
-                socket_addr.port()
-            ))
-            .unwrap()
-        } else {
-            // The configured URL has the /upcall path prefix for pageservers to use: we will strip that out
-            // for general purpose API access.
-            let listen_url = self.env.control_plane_api.clone().unwrap();
-            Url::from_str(&format!(
-                "http://{}:{}/{path}",
-                listen_url.host_str().unwrap(),
-                listen_url.port().unwrap()
-            ))
-            .unwrap()
-        };
+        // The configured URL has the /upcall path prefix for pageservers to use: we will strip that out
+        // for general purpose API access.
+        let listen_url = self.env.control_plane_api.clone().unwrap();
+        let url = Url::from_str(&format!(
+            "http://{}:{}/{path}",
+            listen_url.host_str().unwrap(),
+            listen_url.port().unwrap()
+        ))
+        .unwrap();
 
         let mut builder = self.client.request(method, url);
         if let Some(body) = body {

docs/rfcs/033-storage-controller-drain-and-fill.md

@@ -14,7 +14,7 @@ picked tenant (which requested on-demand activation) for around 30 seconds
 during the restart at 2024-04-03 16:37 UTC.
 
 Note that lots of shutdowns on loaded pageservers do not finish within the
-[10 second systemd enforced timeout](https://github.com/neondatabase/infra/blob/0a5280b383e43c063d43cbf87fa026543f6d6ad4/.github/ansible/systemd/pageserver.service#L16). This means we are shutting down without flushing ephemeral layers
+[10 second systemd enforced timeout](https://github.com/neondatabase/aws/blob/0a5280b383e43c063d43cbf87fa026543f6d6ad4/.github/ansible/systemd/pageserver.service#L16). This means we are shutting down without flushing ephemeral layers
 and have to reingest data in order to serve requests after restarting, potentially making first request latencies worse.
 
 This problem is not yet very acutely felt in storage controller managed pageservers since

docs/rfcs/036-physical-replication.md

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

libs/compute_api/Cargo.toml

@@ -14,3 +14,5 @@ regex.workspace = true
 
 utils = { path = "../utils" }
 remote_storage = { version = "0.1", path = "../remote_storage/" }
+
+workspace_hack.workspace = true

libs/consumption_metrics/Cargo.toml

@@ -6,8 +6,10 @@ license = "Apache-2.0"
 
 [dependencies]
 anyhow.workspace = true
-chrono = { workspace = true, features = ["serde"] }
+chrono.workspace = true
 rand.workspace = true
 serde.workspace = true
 serde_with.workspace = true
 utils.workspace = true
+
+workspace_hack.workspace = true

libs/desim/Cargo.toml

@@ -14,3 +14,5 @@ parking_lot.workspace = true
 hex.workspace = true
 scopeguard.workspace = true
 smallvec = { workspace = true, features = ["write"] }
+
+workspace_hack.workspace = true

libs/metrics/Cargo.toml

@@ -12,6 +12,8 @@ chrono.workspace = true
 twox-hash.workspace = true
 measured.workspace = true
 
+workspace_hack.workspace = true
+
 [target.'cfg(target_os = "linux")'.dependencies]
 procfs.workspace = true
 measured-process.workspace = true

libs/pageserver_api/Cargo.toml

@@ -21,9 +21,11 @@ hex.workspace = true
 humantime.workspace = true
 thiserror.workspace = true
 humantime-serde.workspace = true
-chrono = { workspace = true, features = ["serde"] }
+chrono.workspace = true
 itertools.workspace = true
 
+workspace_hack.workspace = true
+
 [dev-dependencies]
 bincode.workspace = true
 rand.workspace = true

libs/postgres_backend/Cargo.toml

@@ -18,6 +18,7 @@ tokio-rustls.workspace = true
 tracing.workspace = true
 
 pq_proto.workspace = true
+workspace_hack.workspace = true
 
 [dev-dependencies]
 once_cell.workspace = true

libs/postgres_connection/Cargo.toml

@@ -11,5 +11,7 @@ postgres.workspace = true
 tokio-postgres.workspace = true
 url.workspace = true
 
+workspace_hack.workspace = true
+
 [dev-dependencies]
 once_cell.workspace = true

libs/postgres_ffi/Cargo.toml

@@ -19,6 +19,8 @@ thiserror.workspace = true
 serde.workspace = true
 utils.workspace = true
 
+workspace_hack.workspace = true
+
 [dev-dependencies]
 env_logger.workspace = true
 postgres.workspace = true

libs/postgres_ffi/wal_craft/Cargo.toml

@@ -14,6 +14,8 @@ postgres.workspace = true
 postgres_ffi.workspace = true
 camino-tempfile.workspace = true
 
+workspace_hack.workspace = true
+
 [dev-dependencies]
 regex.workspace = true
 utils.workspace = true

libs/pq_proto/Cargo.toml

@@ -11,7 +11,9 @@ itertools.workspace = true
 pin-project-lite.workspace = true
 postgres-protocol.workspace = true
 rand.workspace = true
-tokio = { workspace = true, features = ["io-util"] }
+tokio.workspace = true
 tracing.workspace = true
 thiserror.workspace = true
 serde.workspace = true
+
+workspace_hack.workspace = true

libs/remote_storage/Cargo.toml

@@ -32,7 +32,7 @@ scopeguard.workspace = true
 metrics.workspace = true
 utils.workspace = true
 pin-project-lite.workspace = true
-
+workspace_hack.workspace = true
 azure_core.workspace = true
 azure_identity.workspace = true
 azure_storage.workspace = true
@@ -46,4 +46,3 @@ sync_wrapper = { workspace = true, features = ["futures"] }
 camino-tempfile.workspace = true
 test-context.workspace = true
 rand.workspace = true
-tokio = { workspace = true, features = ["test-util"] }

libs/remote_storage/src/azure_blob.rs

@@ -383,48 +383,6 @@ impl RemoteStorage for AzureBlobStorage {
         }
     }
 
-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError> {
-        let kind = RequestKind::Head;
-        let _permit = self.permit(kind, cancel).await?;
-
-        let started_at = start_measuring_requests(kind);
-
-        let blob_client = self.client.blob_client(self.relative_path_to_name(key));
-        let properties_future = blob_client.get_properties().into_future();
-
-        let properties_future = tokio::time::timeout(self.timeout, properties_future);
-
-        let res = tokio::select! {
-            res = properties_future => res,
-            _ = cancel.cancelled() => return Err(TimeoutOrCancel::Cancel.into()),
-        };
-
-        if let Ok(inner) = &res {
-            // do not incl. timeouts as errors in metrics but cancellations
-            let started_at = ScopeGuard::into_inner(started_at);
-            crate::metrics::BUCKET_METRICS
-                .req_seconds
-                .observe_elapsed(kind, inner, started_at);
-        }
-
-        let data = match res {
-            Ok(Ok(data)) => Ok(data),
-            Ok(Err(sdk)) => Err(to_download_error(sdk)),
-            Err(_timeout) => Err(DownloadError::Timeout),
-        }?;
-
-        let properties = data.blob.properties;
-        Ok(ListingObject {
-            key: key.to_owned(),
-            last_modified: SystemTime::from(properties.last_modified),
-            size: properties.content_length,
-        })
-    }
-
     async fn upload(
         &self,
         from: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync + 'static,

libs/remote_storage/src/lib.rs

@@ -150,7 +150,7 @@ pub enum ListingMode {
     NoDelimiter,
 }
 
-#[derive(PartialEq, Eq, Debug, Clone)]
+#[derive(PartialEq, Eq, Debug)]
 pub struct ListingObject {
     pub key: RemotePath,
     pub last_modified: SystemTime,
@@ -215,13 +215,6 @@ pub trait RemoteStorage: Send + Sync + 'static {
         Ok(combined)
     }
 
-    /// Obtain metadata information about an object.
-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError>;
-
     /// Streams the local file contents into remote into the remote storage entry.
     ///
     /// If the operation fails because of timeout or cancellation, the root cause of the error will be
@@ -370,20 +363,6 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
         }
     }
 
-    // See [`RemoteStorage::head_object`].
-    pub async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError> {
-        match self {
-            Self::LocalFs(s) => s.head_object(key, cancel).await,
-            Self::AwsS3(s) => s.head_object(key, cancel).await,
-            Self::AzureBlob(s) => s.head_object(key, cancel).await,
-            Self::Unreliable(s) => s.head_object(key, cancel).await,
-        }
-    }
-
     /// See [`RemoteStorage::upload`]
     pub async fn upload(
         &self,
@@ -619,7 +598,6 @@ impl ConcurrencyLimiter {
             RequestKind::Delete => &self.write,
             RequestKind::Copy => &self.write,
             RequestKind::TimeTravel => &self.write,
-            RequestKind::Head => &self.read,
         }
     }
 

libs/remote_storage/src/local_fs.rs

@@ -445,20 +445,6 @@ impl RemoteStorage for LocalFs {
         }
     }
 
-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        _cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError> {
-        let target_file_path = key.with_base(&self.storage_root);
-        let metadata = file_metadata(&target_file_path).await?;
-        Ok(ListingObject {
-            key: key.clone(),
-            last_modified: metadata.modified()?,
-            size: metadata.len(),
-        })
-    }
-
     async fn upload(
         &self,
         data: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync,

libs/remote_storage/src/metrics.rs

@@ -13,7 +13,6 @@ pub(crate) enum RequestKind {
     List = 3,
     Copy = 4,
     TimeTravel = 5,
-    Head = 6,
 }
 
 use scopeguard::ScopeGuard;
@@ -28,7 +27,6 @@ impl RequestKind {
             List => "list_objects",
             Copy => "copy_object",
             TimeTravel => "time_travel_recover",
-            Head => "head_object",
         }
     }
     const fn as_index(&self) -> usize {
@@ -36,8 +34,7 @@ impl RequestKind {
     }
 }
 
-const REQUEST_KIND_COUNT: usize = 7;
-pub(crate) struct RequestTyped<C>([C; REQUEST_KIND_COUNT]);
+pub(crate) struct RequestTyped<C>([C; 6]);
 
 impl<C> RequestTyped<C> {
     pub(crate) fn get(&self, kind: RequestKind) -> &C {
@@ -46,8 +43,8 @@ impl<C> RequestTyped<C> {
 
     fn build_with(mut f: impl FnMut(RequestKind) -> C) -> Self {
         use RequestKind::*;
-        let mut it = [Get, Put, Delete, List, Copy, TimeTravel, Head].into_iter();
-        let arr = std::array::from_fn::<C, REQUEST_KIND_COUNT, _>(|index| {
+        let mut it = [Get, Put, Delete, List, Copy, TimeTravel].into_iter();
+        let arr = std::array::from_fn::<C, 6, _>(|index| {
             let next = it.next().unwrap();
             assert_eq!(index, next.as_index());
             f(next)

libs/remote_storage/src/s3_bucket.rs

@@ -23,7 +23,7 @@ use aws_config::{
 use aws_sdk_s3::{
     config::{AsyncSleep, IdentityCache, Region, SharedAsyncSleep},
     error::SdkError,
-    operation::{get_object::GetObjectError, head_object::HeadObjectError},
+    operation::get_object::GetObjectError,
     types::{Delete, DeleteMarkerEntry, ObjectIdentifier, ObjectVersion, StorageClass},
     Client,
 };
@@ -604,78 +604,6 @@ impl RemoteStorage for S3Bucket {
         }
     }
 
-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<ListingObject, DownloadError> {
-        let kind = RequestKind::Head;
-        let _permit = self.permit(kind, cancel).await?;
-
-        let started_at = start_measuring_requests(kind);
-
-        let head_future = self
-            .client
-            .head_object()
-            .bucket(self.bucket_name())
-            .key(self.relative_path_to_s3_object(key))
-            .send();
-
-        let head_future = tokio::time::timeout(self.timeout, head_future);
-
-        let res = tokio::select! {
-            res = head_future => res,
-            _ = cancel.cancelled() => return Err(TimeoutOrCancel::Cancel.into()),
-        };
-
-        let res = res.map_err(|_e| DownloadError::Timeout)?;
-
-        // do not incl. timeouts as errors in metrics but cancellations
-        let started_at = ScopeGuard::into_inner(started_at);
-        crate::metrics::BUCKET_METRICS
-            .req_seconds
-            .observe_elapsed(kind, &res, started_at);
-
-        let data = match res {
-            Ok(object_output) => object_output,
-            Err(SdkError::ServiceError(e)) if matches!(e.err(), HeadObjectError::NotFound(_)) => {
-                // Count this in the AttemptOutcome::Ok bucket, because 404 is not
-                // an error: we expect to sometimes fetch an object and find it missing,
-                // e.g. when probing for timeline indices.
-                crate::metrics::BUCKET_METRICS.req_seconds.observe_elapsed(
-                    kind,
-                    AttemptOutcome::Ok,
-                    started_at,
-                );
-                return Err(DownloadError::NotFound);
-            }
-            Err(e) => {
-                crate::metrics::BUCKET_METRICS.req_seconds.observe_elapsed(
-                    kind,
-                    AttemptOutcome::Err,
-                    started_at,
-                );
-
-                return Err(DownloadError::Other(
-                    anyhow::Error::new(e).context("s3 head object"),
-                ));
-            }
-        };
-
-        let (Some(last_modified), Some(size)) = (data.last_modified, data.content_length) else {
-            return Err(DownloadError::Other(anyhow!(
-                "head_object doesn't contain last_modified or content_length"
-            )))?;
-        };
-        Ok(ListingObject {
-            key: key.to_owned(),
-            last_modified: SystemTime::try_from(last_modified).map_err(|e| {
-                DownloadError::Other(anyhow!("can't convert time '{last_modified}': {e}"))
-            })?,
-            size: size as u64,
-        })
-    }
-
     async fn upload(
         &self,
         from: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync + 'static,

libs/remote_storage/src/simulate_failures.rs

@@ -30,7 +30,6 @@ pub struct UnreliableWrapper {
 #[derive(Debug, Hash, Eq, PartialEq)]
 enum RemoteOp {
     ListPrefixes(Option<RemotePath>),
-    HeadObject(RemotePath),
     Upload(RemotePath),
     Download(RemotePath),
     Delete(RemotePath),
@@ -138,16 +137,6 @@ impl RemoteStorage for UnreliableWrapper {
         self.inner.list(prefix, mode, max_keys, cancel).await
     }
 
-    async fn head_object(
-        &self,
-        key: &RemotePath,
-        cancel: &CancellationToken,
-    ) -> Result<crate::ListingObject, DownloadError> {
-        self.attempt(RemoteOp::HeadObject(key.clone()))
-            .map_err(DownloadError::Other)?;
-        self.inner.head_object(key, cancel).await
-    }
-
     async fn upload(
         &self,
         data: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync + 'static,

libs/safekeeper_api/Cargo.toml

@@ -9,3 +9,5 @@ serde.workspace = true
 serde_with.workspace = true
 const_format.workspace = true
 utils.workspace = true
+
+workspace_hack.workspace = true

libs/tenant_size_model/Cargo.toml

@@ -9,3 +9,5 @@ license.workspace = true
 anyhow.workspace = true
 serde.workspace = true
 serde_json.workspace = true
+
+workspace_hack.workspace = true

libs/tracing-utils/Cargo.toml

@@ -14,3 +14,5 @@ tokio = { workspace = true, features = ["rt", "rt-multi-thread"] }
 tracing.workspace = true
 tracing-opentelemetry.workspace = true
 tracing-subscriber.workspace = true
+
+workspace_hack.workspace = true

libs/utils/Cargo.toml

@@ -39,7 +39,7 @@ thiserror.workspace = true
 tokio.workspace = true
 tokio-tar.workspace = true
 tokio-util.workspace = true
-toml_edit = { workspace = true, features = ["serde"] }
+toml_edit.workspace = true
 tracing.workspace = true
 tracing-error.workspace = true
 tracing-subscriber = { workspace = true, features = ["json", "registry"] }
@@ -54,6 +54,7 @@ walkdir.workspace = true
 pq_proto.workspace = true
 postgres_connection.workspace = true
 metrics.workspace = true
+workspace_hack.workspace = true
 
 const_format.workspace = true
 
@@ -70,7 +71,6 @@ criterion.workspace = true
 hex-literal.workspace = true
 camino-tempfile.workspace = true
 serde_assert.workspace = true
-tokio = { workspace = true, features = ["test-util"] }
 
 [[bench]]
 name = "benchmarks"

libs/walproposer/Cargo.toml

@@ -9,6 +9,8 @@ anyhow.workspace = true
 utils.workspace = true
 postgres_ffi.workspace = true
 
+workspace_hack.workspace = true
+
 [build-dependencies]
 anyhow.workspace = true
 bindgen.workspace = true

pageserver/benches/bench_ingest.rs

@@ -10,7 +10,6 @@ use pageserver::{
     page_cache,
     repository::Value,
     task_mgr::TaskKind,
-    tenant::storage_layer::inmemory_layer::SerializedBatch,
     tenant::storage_layer::InMemoryLayer,
     virtual_file,
 };
@@ -68,16 +67,12 @@ async fn ingest(
     let layer =
         InMemoryLayer::create(conf, timeline_id, tenant_shard_id, lsn, entered, &ctx).await?;
 
-    let data = Value::Image(Bytes::from(vec![0u8; put_size]));
-    let data_ser_size = data.serialized_size().unwrap() as usize;
+    let data = Value::Image(Bytes::from(vec![0u8; put_size])).ser()?;
     let ctx = RequestContext::new(
         pageserver::task_mgr::TaskKind::WalReceiverConnectionHandler,
         pageserver::context::DownloadBehavior::Download,
     );
 
-    const BATCH_SIZE: usize = 16;
-    let mut batch = Vec::new();
-
     for i in 0..put_count {
         lsn += put_size as u64;
 
@@ -100,17 +95,7 @@ async fn ingest(
             }
         }
 
-        batch.push((key.to_compact(), lsn, data_ser_size, data.clone()));
-        if batch.len() >= BATCH_SIZE {
-            let this_batch = std::mem::take(&mut batch);
-            let serialized = SerializedBatch::from_values(this_batch);
-            layer.put_batch(serialized, &ctx).await?;
-        }
-    }
-    if !batch.is_empty() {
-        let this_batch = std::mem::take(&mut batch);
-        let serialized = SerializedBatch::from_values(this_batch);
-        layer.put_batch(serialized, &ctx).await?;
+        layer.put_value(key.to_compact(), lsn, &data, &ctx).await?;
     }
     layer.freeze(lsn + 1).await;
 

pageserver/compaction/src/identify_levels.rs

@@ -82,51 +82,12 @@ where
         return Ok(None);
     }
 
-    // Our goal is to find an LSN boundary such that:
+    // Walk the ranges in LSN order.
     //
-    // - There are no layers that overlap with the LSN. In other words, the LSN neatly
-    //   divides the layers into two sets: those that are above it, and those that are
-    //   below it.
-    //
-    // - All the layers above it have LSN range smaller than lsn_max_size
-    //
-    // Our strategy is to walk through the layers ordered by end LSNs, and maintain two
-    // values:
-    //
-    // - The current best LSN boundary. We have already determined that there are no
-    //   layers that overlap with it.
-    //
-    // - Candidate LSN boundary. We have not found any layers that overlap with this LSN
-    //   yet, but we might still see one later, as we continue iterating.
-    //
-    // Whenever we find that there are no layers that overlap with the current candidate,
-    // we make that the new "current best" and start building a new candidate. We start
-    // with the caller-supplied `end_lsn` as the current best LSN boundary. All the layers
-    // are below that LSN, so that's a valid stopping point. If we see a layer that is too
-    // large, we discard the candidate that we were building and return with the "current
-    // best" that we have seen so far.
-    //
-    // Here's an illustration of the state after processing layers A, B, C, and D:
-    //
-    //            A   B   C   D   E
-    // 10000                           <-- end_lsn passed by caller
-    //  9000      +
-    //  8000      |   +   +
-    //  7000      +   |   |
-    //  6000          |   |
-    //  5000          |   +
-    //  4000          +
-    //  3000                  +          <-- current_best_start_lsn
-    //  3000                  |    +
-    //  2000                  |    |
-    //  2000                  +    |     <--- candidate_start_lsn
-    //  1000                       +
-    //
-    // When we saw layer D with end_lsn==3000, and had not seen any layers that cross that
-    // LSN, we knew that that is a valid stopping point. It became the new
-    // current_best_start_lsn, and the layer's start_lsn (2000) became the new
-    // candidate_start_lsn. Because layer E overlaps with it, it is not a valid return
-    // value, but we will not know that until processing layer E.
+    // ----- end_lsn
+    //  |
+    //  |
+    //  v
     //
     layers.sort_by_key(|l| l.lsn_range().end);
     let mut candidate_start_lsn = end_lsn;
@@ -173,8 +134,7 @@ where
 
             // Is it small enough to be considered part of this level?
             if r.end.0 - r.start.0 > lsn_max_size {
-                // Too large, this layer belongs to next level. Discard the candidate
-                // we were building and return with `current_best`.
+                // Too large, this layer belongs to next level. Stop.
                 trace!(
                     "too large {}, size {} vs {}",
                     l.short_id(),

pageserver/src/config.rs

@@ -50,6 +50,7 @@ pub mod defaults {
         DEFAULT_HTTP_LISTEN_ADDR, DEFAULT_HTTP_LISTEN_PORT, DEFAULT_PG_LISTEN_ADDR,
         DEFAULT_PG_LISTEN_PORT,
     };
+    use pageserver_api::models::ImageCompressionAlgorithm;
     pub use storage_broker::DEFAULT_ENDPOINT as BROKER_DEFAULT_ENDPOINT;
 
     pub const DEFAULT_WAIT_LSN_TIMEOUT: &str = "300 s";
@@ -89,7 +90,8 @@ pub mod defaults {
 
     pub const DEFAULT_MAX_VECTORED_READ_BYTES: usize = 128 * 1024; // 128 KiB
 
-    pub const DEFAULT_IMAGE_COMPRESSION: &str = "zstd(1)";
+    pub const DEFAULT_IMAGE_COMPRESSION: ImageCompressionAlgorithm =
+        ImageCompressionAlgorithm::Disabled;
 
     pub const DEFAULT_VALIDATE_VECTORED_GET: bool = false;
 
@@ -476,7 +478,7 @@ impl PageServerConfigBuilder {
             max_vectored_read_bytes: Set(MaxVectoredReadBytes(
                 NonZeroUsize::new(DEFAULT_MAX_VECTORED_READ_BYTES).unwrap(),
             )),
-            image_compression: Set(DEFAULT_IMAGE_COMPRESSION.parse().unwrap()),
+            image_compression: Set(DEFAULT_IMAGE_COMPRESSION),
             ephemeral_bytes_per_memory_kb: Set(DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB),
             l0_flush: Set(L0FlushConfig::default()),
             compact_level0_phase1_value_access: Set(CompactL0Phase1ValueAccess::default()),
@@ -1063,7 +1065,7 @@ impl PageServerConf {
                 NonZeroUsize::new(defaults::DEFAULT_MAX_VECTORED_READ_BYTES)
                     .expect("Invalid default constant"),
             ),
-            image_compression: defaults::DEFAULT_IMAGE_COMPRESSION.parse().unwrap(),
+            image_compression: defaults::DEFAULT_IMAGE_COMPRESSION,
             ephemeral_bytes_per_memory_kb: defaults::DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB,
             l0_flush: L0FlushConfig::default(),
             compact_level0_phase1_value_access: CompactL0Phase1ValueAccess::default(),
@@ -1303,7 +1305,7 @@ background_task_maximum_delay = '334 s'
                     NonZeroUsize::new(defaults::DEFAULT_MAX_VECTORED_READ_BYTES)
                         .expect("Invalid default constant")
                 ),
-                image_compression: defaults::DEFAULT_IMAGE_COMPRESSION.parse().unwrap(),
+                image_compression: defaults::DEFAULT_IMAGE_COMPRESSION,
                 ephemeral_bytes_per_memory_kb: defaults::DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB,
                 l0_flush: L0FlushConfig::default(),
                 compact_level0_phase1_value_access: CompactL0Phase1ValueAccess::default(),
@@ -1376,7 +1378,7 @@ background_task_maximum_delay = '334 s'
                     NonZeroUsize::new(defaults::DEFAULT_MAX_VECTORED_READ_BYTES)
                         .expect("Invalid default constant")
                 ),
-                image_compression: defaults::DEFAULT_IMAGE_COMPRESSION.parse().unwrap(),
+                image_compression: defaults::DEFAULT_IMAGE_COMPRESSION,
                 ephemeral_bytes_per_memory_kb: defaults::DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB,
                 l0_flush: L0FlushConfig::default(),
                 compact_level0_phase1_value_access: CompactL0Phase1ValueAccess::default(),

pageserver/src/l0_flush.rs

@@ -1,10 +1,15 @@
 use std::{num::NonZeroUsize, sync::Arc};
 
+use crate::tenant::ephemeral_file;
+
 #[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize)]
 #[serde(tag = "mode", rename_all = "kebab-case", deny_unknown_fields)]
 pub enum L0FlushConfig {
+    PageCached,
     #[serde(rename_all = "snake_case")]
-    Direct { max_concurrency: NonZeroUsize },
+    Direct {
+        max_concurrency: NonZeroUsize,
+    },
 }
 
 impl Default for L0FlushConfig {
@@ -20,12 +25,14 @@ impl Default for L0FlushConfig {
 pub struct L0FlushGlobalState(Arc<Inner>);
 
 pub enum Inner {
+    PageCached,
     Direct { semaphore: tokio::sync::Semaphore },
 }
 
 impl L0FlushGlobalState {
     pub fn new(config: L0FlushConfig) -> Self {
         match config {
+            L0FlushConfig::PageCached => Self(Arc::new(Inner::PageCached)),
             L0FlushConfig::Direct { max_concurrency } => {
                 let semaphore = tokio::sync::Semaphore::new(max_concurrency.get());
                 Self(Arc::new(Inner::Direct { semaphore }))
@@ -37,3 +44,13 @@ impl L0FlushGlobalState {
         &self.0
     }
 }
+
+impl L0FlushConfig {
+    pub(crate) fn prewarm_on_write(&self) -> ephemeral_file::PrewarmPageCacheOnWrite {
+        use L0FlushConfig::*;
+        match self {
+            PageCached => ephemeral_file::PrewarmPageCacheOnWrite::Yes,
+            Direct { .. } => ephemeral_file::PrewarmPageCacheOnWrite::No,
+        }
+    }
+}

pageserver/src/lib.rs

@@ -49,7 +49,7 @@ use tracing::{info, info_span};
 /// backwards-compatible changes to the metadata format.
 pub const STORAGE_FORMAT_VERSION: u16 = 3;
 
-pub const DEFAULT_PG_VERSION: u32 = 16;
+pub const DEFAULT_PG_VERSION: u32 = 15;
 
 // Magic constants used to identify different kinds of files
 pub const IMAGE_FILE_MAGIC: u16 = 0x5A60;
@@ -88,8 +88,6 @@ pub async fn shutdown_pageserver(
 ) {
     use std::time::Duration;
 
-    let started_at = std::time::Instant::now();
-
     // If the orderly shutdown below takes too long, we still want to make
     // sure that all walredo processes are killed and wait()ed on by us, not systemd.
     //
@@ -243,10 +241,7 @@ pub async fn shutdown_pageserver(
     walredo_extraordinary_shutdown_thread.join().unwrap();
     info!("walredo_extraordinary_shutdown_thread done");
 
-    info!(
-        elapsed_ms = started_at.elapsed().as_millis(),
-        "Shut down successfully completed"
-    );
+    info!("Shut down successfully completed");
     std::process::exit(exit_code);
 }
 

pageserver/src/metrics.rs

@@ -1803,15 +1803,6 @@ pub(crate) static SECONDARY_RESIDENT_PHYSICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::n
     .expect("failed to define a metric")
 });
 
-pub(crate) static SECONDARY_HEATMAP_TOTAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
-    register_uint_gauge_vec!(
-        "pageserver_secondary_heatmap_total_size",
-        "The total size in bytes of all layers in the most recently downloaded heatmap.",
-        &["tenant_id", "shard_id"]
-    )
-    .expect("failed to define a metric")
-});
-
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
 pub enum RemoteOpKind {
     Upload,
@@ -1862,64 +1853,16 @@ pub(crate) static TENANT_TASK_EVENTS: Lazy<IntCounterVec> = Lazy::new(|| {
     .expect("Failed to register tenant_task_events metric")
 });
 
-pub struct BackgroundLoopSemaphoreMetrics {
-    counters: EnumMap<BackgroundLoopKind, IntCounterPair>,
-    durations: EnumMap<BackgroundLoopKind, Counter>,
-}
-
-pub(crate) static BACKGROUND_LOOP_SEMAPHORE: Lazy<BackgroundLoopSemaphoreMetrics> = Lazy::new(
-    || {
-        let counters = register_int_counter_pair_vec!(
-            "pageserver_background_loop_semaphore_wait_start_count",
-            "Counter for background loop concurrency-limiting semaphore acquire calls started",
-            "pageserver_background_loop_semaphore_wait_finish_count",
-            "Counter for background loop concurrency-limiting semaphore acquire calls finished",
-            &["task"],
-        )
-        .unwrap();
-
-        let durations = register_counter_vec!(
-            "pageserver_background_loop_semaphore_wait_duration_seconds",
-            "Sum of wall clock time spent waiting on the background loop concurrency-limiting semaphore acquire calls",
-            &["task"],
-        )
-        .unwrap();
-
-        BackgroundLoopSemaphoreMetrics {
-            counters: enum_map::EnumMap::from_array(std::array::from_fn(|i| {
-                let kind = <BackgroundLoopKind as enum_map::Enum>::from_usize(i);
-                counters.with_label_values(&[kind.into()])
-            })),
-            durations: enum_map::EnumMap::from_array(std::array::from_fn(|i| {
-                let kind = <BackgroundLoopKind as enum_map::Enum>::from_usize(i);
-                durations.with_label_values(&[kind.into()])
-            })),
-        }
-    },
-);
-
-impl BackgroundLoopSemaphoreMetrics {
-    pub(crate) fn measure_acquisition(&self, task: BackgroundLoopKind) -> impl Drop + '_ {
-        struct Record<'a> {
-            metrics: &'a BackgroundLoopSemaphoreMetrics,
-            task: BackgroundLoopKind,
-            _counter_guard: metrics::IntCounterPairGuard,
-            start: Instant,
-        }
-        impl Drop for Record<'_> {
-            fn drop(&mut self) {
-                let elapsed = self.start.elapsed().as_secs_f64();
-                self.metrics.durations[self.task].inc_by(elapsed);
-            }
-        }
-        Record {
-            metrics: self,
-            task,
-            _counter_guard: self.counters[task].guard(),
-            start: Instant::now(),
-        }
-    }
-}
+pub(crate) static BACKGROUND_LOOP_SEMAPHORE_WAIT_GAUGE: Lazy<IntCounterPairVec> = Lazy::new(|| {
+    register_int_counter_pair_vec!(
+        "pageserver_background_loop_semaphore_wait_start_count",
+        "Counter for background loop concurrency-limiting semaphore acquire calls started",
+        "pageserver_background_loop_semaphore_wait_finish_count",
+        "Counter for background loop concurrency-limiting semaphore acquire calls finished",
+        &["task"],
+    )
+    .unwrap()
+});
 
 pub(crate) static BACKGROUND_LOOP_PERIOD_OVERRUN_COUNT: Lazy<IntCounterVec> = Lazy::new(|| {
     register_int_counter_vec!(
@@ -2601,7 +2544,6 @@ use std::time::{Duration, Instant};
 use crate::context::{PageContentKind, RequestContext};
 use crate::task_mgr::TaskKind;
 use crate::tenant::mgr::TenantSlot;
-use crate::tenant::tasks::BackgroundLoopKind;
 
 /// Maintain a per timeline gauge in addition to the global gauge.
 pub(crate) struct PerTimelineRemotePhysicalSizeGauge {

pageserver/src/pgdatadir_mapping.rs

@@ -15,11 +15,12 @@ use crate::{aux_file, repository::*};
 use anyhow::{ensure, Context};
 use bytes::{Buf, Bytes, BytesMut};
 use enum_map::Enum;
+use itertools::Itertools;
 use pageserver_api::key::{
     dbdir_key_range, rel_block_to_key, rel_dir_to_key, rel_key_range, rel_size_to_key,
     relmap_file_key, repl_origin_key, repl_origin_key_range, slru_block_to_key, slru_dir_to_key,
     slru_segment_key_range, slru_segment_size_to_key, twophase_file_key, twophase_key_range,
-    CompactKey, AUX_FILES_KEY, CHECKPOINT_KEY, CONTROLFILE_KEY, DBDIR_KEY, TWOPHASEDIR_KEY,
+    AUX_FILES_KEY, CHECKPOINT_KEY, CONTROLFILE_KEY, DBDIR_KEY, TWOPHASEDIR_KEY,
 };
 use pageserver_api::keyspace::SparseKeySpace;
 use pageserver_api::models::AuxFilePolicy;
@@ -36,6 +37,7 @@ use tokio_util::sync::CancellationToken;
 use tracing::{debug, info, trace, warn};
 use utils::bin_ser::DeserializeError;
 use utils::pausable_failpoint;
+use utils::vec_map::{VecMap, VecMapOrdering};
 use utils::{bin_ser::BeSer, lsn::Lsn};
 
 /// Max delta records appended to the AUX_FILES_KEY (for aux v1). The write path will write a full image once this threshold is reached.
@@ -172,7 +174,6 @@ impl Timeline {
             pending_deletions: Vec::new(),
             pending_nblocks: 0,
             pending_directory_entries: Vec::new(),
-            pending_bytes: 0,
             lsn,
         }
     }
@@ -1021,33 +1022,21 @@ pub struct DatadirModification<'a> {
     // The put-functions add the modifications here, and they are flushed to the
     // underlying key-value store by the 'finish' function.
     pending_lsns: Vec<Lsn>,
-    pending_updates: HashMap<Key, Vec<(Lsn, usize, Value)>>,
+    pending_updates: HashMap<Key, Vec<(Lsn, Value)>>,
     pending_deletions: Vec<(Range<Key>, Lsn)>,
     pending_nblocks: i64,
 
     /// For special "directory" keys that store key-value maps, track the size of the map
     /// if it was updated in this modification.
     pending_directory_entries: Vec<(DirectoryKind, usize)>,
-
-    /// An **approximation** of how large our EphemeralFile write will be when committed.
-    pending_bytes: usize,
 }
 
 impl<'a> DatadirModification<'a> {
-    // When a DatadirModification is committed, we do a monolithic serialization of all its contents.  WAL records can
-    // contain multiple pages, so the pageserver's record-based batch size isn't sufficient to bound this allocation: we
-    // additionally specify a limit on how much payload a DatadirModification may contain before it should be committed.
-    pub(crate) const MAX_PENDING_BYTES: usize = 8 * 1024 * 1024;
-
     /// Get the current lsn
     pub(crate) fn get_lsn(&self) -> Lsn {
         self.lsn
     }
 
-    pub(crate) fn approx_pending_bytes(&self) -> usize {
-        self.pending_bytes
-    }
-
     /// Set the current lsn
     pub(crate) fn set_lsn(&mut self, lsn: Lsn) -> anyhow::Result<()> {
         ensure!(
@@ -1780,25 +1769,21 @@ impl<'a> DatadirModification<'a> {
         // Flush relation and  SLRU data blocks, keep metadata.
         let mut retained_pending_updates = HashMap::<_, Vec<_>>::new();
         for (key, values) in self.pending_updates.drain() {
-            let mut write_batch = Vec::new();
-            for (lsn, value_ser_size, value) in values {
+            for (lsn, value) in values {
                 if key.is_rel_block_key() || key.is_slru_block_key() {
                     // This bails out on first error without modifying pending_updates.
                     // That's Ok, cf this function's doc comment.
-                    write_batch.push((key.to_compact(), lsn, value_ser_size, value));
+                    writer.put(key, lsn, &value, ctx).await?;
                 } else {
-                    retained_pending_updates.entry(key).or_default().push((
-                        lsn,
-                        value_ser_size,
-                        value,
-                    ));
+                    retained_pending_updates
+                        .entry(key)
+                        .or_default()
+                        .push((lsn, value));
                 }
             }
-            writer.put_batch(write_batch, ctx).await?;
         }
 
         self.pending_updates = retained_pending_updates;
-        self.pending_bytes = 0;
 
         if pending_nblocks != 0 {
             writer.update_current_logical_size(pending_nblocks * i64::from(BLCKSZ));
@@ -1824,20 +1809,17 @@ impl<'a> DatadirModification<'a> {
         self.pending_nblocks = 0;
 
         if !self.pending_updates.is_empty() {
-            // Ordering: the items in this batch do not need to be in any global order, but values for
-            // a particular Key must be in Lsn order relative to one another.  InMemoryLayer relies on
-            // this to do efficient updates to its index.
-            let batch: Vec<(CompactKey, Lsn, usize, Value)> = self
-                .pending_updates
-                .drain()
-                .flat_map(|(key, values)| {
-                    values.into_iter().map(move |(lsn, val_ser_size, value)| {
-                        (key.to_compact(), lsn, val_ser_size, value)
-                    })
-                })
-                .collect::<Vec<_>>();
+            // The put_batch call below expects expects the inputs to be sorted by Lsn,
+            // so we do that first.
+            let lsn_ordered_batch: VecMap<Lsn, (Key, Value)> = VecMap::from_iter(
+                self.pending_updates
+                    .drain()
+                    .map(|(key, vals)| vals.into_iter().map(move |(lsn, val)| (lsn, (key, val))))
+                    .kmerge_by(|lhs, rhs| lhs.0 < rhs.0),
+                VecMapOrdering::GreaterOrEqual,
+            );
 
-            writer.put_batch(batch, ctx).await?;
+            writer.put_batch(lsn_ordered_batch, ctx).await?;
         }
 
         if !self.pending_deletions.is_empty() {
@@ -1862,8 +1844,6 @@ impl<'a> DatadirModification<'a> {
             writer.update_directory_entries_count(kind, count as u64);
         }
 
-        self.pending_bytes = 0;
-
         Ok(())
     }
 
@@ -1880,7 +1860,7 @@ impl<'a> DatadirModification<'a> {
         // Note: we don't check pending_deletions. It is an error to request a
         // value that has been removed, deletion only avoids leaking storage.
         if let Some(values) = self.pending_updates.get(&key) {
-            if let Some((_, _, value)) = values.last() {
+            if let Some((_, value)) = values.last() {
                 return if let Value::Image(img) = value {
                     Ok(img.clone())
                 } else {
@@ -1908,17 +1888,13 @@ impl<'a> DatadirModification<'a> {
     fn put(&mut self, key: Key, val: Value) {
         let values = self.pending_updates.entry(key).or_default();
         // Replace the previous value if it exists at the same lsn
-        if let Some((last_lsn, last_value_ser_size, last_value)) = values.last_mut() {
+        if let Some((last_lsn, last_value)) = values.last_mut() {
             if *last_lsn == self.lsn {
-                *last_value_ser_size = val.serialized_size().unwrap() as usize;
                 *last_value = val;
                 return;
             }
         }
-
-        let val_serialized_size = val.serialized_size().unwrap() as usize;
-        self.pending_bytes += val_serialized_size;
-        values.push((self.lsn, val_serialized_size, val));
+        values.push((self.lsn, val));
     }
 
     fn delete(&mut self, key_range: Range<Key>) {

pageserver/src/task_mgr.rs

@@ -393,7 +393,7 @@ struct PageServerTask {
 
     /// Tasks may optionally be launched for a particular tenant/timeline, enabling
     /// later cancelling tasks for that tenant/timeline in [`shutdown_tasks`]
-    tenant_shard_id: TenantShardId,
+    tenant_shard_id: Option<TenantShardId>,
     timeline_id: Option<TimelineId>,
 
     mutable: Mutex<MutableTaskState>,
@@ -405,7 +405,7 @@ struct PageServerTask {
 pub fn spawn<F>(
     runtime: &tokio::runtime::Handle,
     kind: TaskKind,
-    tenant_shard_id: TenantShardId,
+    tenant_shard_id: Option<TenantShardId>,
     timeline_id: Option<TimelineId>,
     name: &str,
     future: F,
@@ -550,7 +550,7 @@ pub async fn shutdown_tasks(
         let tasks = TASKS.lock().unwrap();
         for task in tasks.values() {
             if (kind.is_none() || Some(task.kind) == kind)
-                && (tenant_shard_id.is_none() || Some(task.tenant_shard_id) == tenant_shard_id)
+                && (tenant_shard_id.is_none() || task.tenant_shard_id == tenant_shard_id)
                 && (timeline_id.is_none() || task.timeline_id == timeline_id)
             {
                 task.cancel.cancel();
@@ -573,8 +573,13 @@ pub async fn shutdown_tasks(
         };
         if let Some(mut join_handle) = join_handle {
             if log_all {
-                // warn to catch these in tests; there shouldn't be any
-                warn!(name = task.name, tenant_shard_id = ?tenant_shard_id, timeline_id = ?timeline_id, kind = ?task_kind, "stopping left-over");
+                if tenant_shard_id.is_none() {
+                    // there are quite few of these
+                    info!(name = task.name, kind = ?task_kind, "stopping global task");
+                } else {
+                    // warn to catch these in tests; there shouldn't be any
+                    warn!(name = task.name, tenant_shard_id = ?tenant_shard_id, timeline_id = ?timeline_id, kind = ?task_kind, "stopping left-over");
+                }
             }
             if tokio::time::timeout(std::time::Duration::from_secs(1), &mut join_handle)
                 .await

pageserver/src/tenant.rs

@@ -798,7 +798,7 @@ impl Tenant {
         task_mgr::spawn(
             &tokio::runtime::Handle::current(),
             TaskKind::Attach,
-            tenant_shard_id,
+            Some(tenant_shard_id),
             None,
             "attach tenant",
             async move {

pageserver/src/tenant/ephemeral_file.rs

@@ -21,6 +21,7 @@ pub struct EphemeralFile {
 }
 
 mod page_caching;
+pub(crate) use page_caching::PrewarmOnWrite as PrewarmPageCacheOnWrite;
 mod zero_padded_read_write;
 
 impl EphemeralFile {
@@ -51,10 +52,12 @@ impl EphemeralFile {
         )
         .await?;
 
+        let prewarm = conf.l0_flush.prewarm_on_write();
+
         Ok(EphemeralFile {
             _tenant_shard_id: tenant_shard_id,
             _timeline_id: timeline_id,
-            rw: page_caching::RW::new(file, gate_guard),
+            rw: page_caching::RW::new(file, prewarm, gate_guard),
         })
     }
 
@@ -79,8 +82,6 @@ impl EphemeralFile {
         self.rw.read_blk(blknum, ctx).await
     }
 
-    #[cfg(test)]
-    // This is a test helper: outside of tests, we are always written to via a pre-serialized batch.
     pub(crate) async fn write_blob(
         &mut self,
         srcbuf: &[u8],
@@ -88,30 +89,17 @@ impl EphemeralFile {
     ) -> Result<u64, io::Error> {
         let pos = self.rw.bytes_written();
 
-        let mut len_bytes = std::io::Cursor::new(Vec::new());
-        crate::tenant::storage_layer::inmemory_layer::SerializedBatch::write_blob_length(
-            srcbuf.len(),
-            &mut len_bytes,
-        );
-        let len_bytes = len_bytes.into_inner();
-
         // Write the length field
-        self.rw.write_all_borrowed(&len_bytes, ctx).await?;
+        if srcbuf.len() < 0x80 {
+            // short one-byte length header
+            let len_buf = [srcbuf.len() as u8];
 
-        // Write the payload
-        self.rw.write_all_borrowed(srcbuf, ctx).await?;
-
-        Ok(pos)
-    }
-
-    /// Returns the offset at which the first byte of the input was written, for use
-    /// in constructing indices over the written value.
-    pub(crate) async fn write_raw(
-        &mut self,
-        srcbuf: &[u8],
-        ctx: &RequestContext,
-    ) -> Result<u64, io::Error> {
-        let pos = self.rw.bytes_written();
+            self.rw.write_all_borrowed(&len_buf, ctx).await?;
+        } else {
+            let mut len_buf = u32::to_be_bytes(srcbuf.len() as u32);
+            len_buf[0] |= 0x80;
+            self.rw.write_all_borrowed(&len_buf, ctx).await?;
+        }
 
         // Write the payload
         self.rw.write_all_borrowed(srcbuf, ctx).await?;

pageserver/src/tenant/ephemeral_file/page_caching.rs

@@ -1,15 +1,15 @@
 //! Wrapper around [`super::zero_padded_read_write::RW`] that uses the
 //! [`crate::page_cache`] to serve reads that need to go to the underlying [`VirtualFile`].
-//!
-//! Subject to removal in <https://github.com/neondatabase/neon/pull/8537>
 
 use crate::context::RequestContext;
 use crate::page_cache::{self, PAGE_SZ};
 use crate::tenant::block_io::BlockLease;
-use crate::virtual_file::owned_buffers_io::util::size_tracking_writer;
+use crate::virtual_file::owned_buffers_io::io_buf_ext::FullSlice;
 use crate::virtual_file::VirtualFile;
 
-use std::io::{self};
+use once_cell::sync::Lazy;
+use std::io::{self, ErrorKind};
+use std::ops::{Deref, Range};
 use tokio_epoll_uring::BoundedBuf;
 use tracing::*;
 
@@ -18,17 +18,33 @@ use super::zero_padded_read_write;
 /// See module-level comment.
 pub struct RW {
     page_cache_file_id: page_cache::FileId,
-    rw: super::zero_padded_read_write::RW<size_tracking_writer::Writer<VirtualFile>>,
+    rw: super::zero_padded_read_write::RW<PreWarmingWriter>,
     /// Gate guard is held on as long as we need to do operations in the path (delete on drop).
     _gate_guard: utils::sync::gate::GateGuard,
 }
 
+/// When we flush a block to the underlying [`crate::virtual_file::VirtualFile`],
+/// should we pre-warm the [`crate::page_cache`] with the contents?
+#[derive(Clone, Copy)]
+pub enum PrewarmOnWrite {
+    Yes,
+    No,
+}
+
 impl RW {
-    pub fn new(file: VirtualFile, _gate_guard: utils::sync::gate::GateGuard) -> Self {
+    pub fn new(
+        file: VirtualFile,
+        prewarm_on_write: PrewarmOnWrite,
+        _gate_guard: utils::sync::gate::GateGuard,
+    ) -> Self {
         let page_cache_file_id = page_cache::next_file_id();
         Self {
             page_cache_file_id,
-            rw: super::zero_padded_read_write::RW::new(size_tracking_writer::Writer::new(file)),
+            rw: super::zero_padded_read_write::RW::new(PreWarmingWriter::new(
+                page_cache_file_id,
+                file,
+                prewarm_on_write,
+            )),
             _gate_guard,
         }
     }
@@ -68,10 +84,10 @@ impl RW {
         let vec = Vec::with_capacity(size);
 
         // read from disk what we've already flushed
-        let file_size_tracking_writer = self.rw.as_writer();
-        let flushed_range = 0..usize::try_from(file_size_tracking_writer.bytes_written()).unwrap();
-        let mut vec = file_size_tracking_writer
-            .as_inner()
+        let writer = self.rw.as_writer();
+        let flushed_range = writer.written_range();
+        let mut vec = writer
+            .file
             .read_exact_at(
                 vec.slice(0..(flushed_range.end - flushed_range.start)),
                 u64::try_from(flushed_range.start).unwrap(),
@@ -106,7 +122,7 @@ impl RW {
                             format!(
                                 "ephemeral file: read immutable page #{}: {}: {:#}",
                                 blknum,
-                                self.rw.as_writer().as_inner().path,
+                                self.rw.as_writer().file.path,
                                 e,
                             ),
                         )
@@ -116,7 +132,7 @@ impl RW {
                     }
                     page_cache::ReadBufResult::NotFound(write_guard) => {
                         let write_guard = writer
-                            .as_inner()
+                            .file
                             .read_exact_at_page(write_guard, blknum as u64 * PAGE_SZ as u64, ctx)
                             .await?;
                         let read_guard = write_guard.mark_valid();
@@ -138,16 +154,137 @@ impl Drop for RW {
 
         // unlink the file
         // we are clear to do this, because we have entered a gate
-        let path = &self.rw.as_writer().as_inner().path;
-        let res = std::fs::remove_file(path);
+        let res = std::fs::remove_file(&self.rw.as_writer().file.path);
         if let Err(e) = res {
             if e.kind() != std::io::ErrorKind::NotFound {
                 // just never log the not found errors, we cannot do anything for them; on detach
                 // the tenant directory is already gone.
                 //
                 // not found files might also be related to https://github.com/neondatabase/neon/issues/2442
-                error!("could not remove ephemeral file '{path}': {e}");
+                error!(
+                    "could not remove ephemeral file '{}': {}",
+                    self.rw.as_writer().file.path,
+                    e
+                );
             }
         }
     }
 }
+
+struct PreWarmingWriter {
+    prewarm_on_write: PrewarmOnWrite,
+    nwritten_blocks: u32,
+    page_cache_file_id: page_cache::FileId,
+    file: VirtualFile,
+}
+
+impl PreWarmingWriter {
+    fn new(
+        page_cache_file_id: page_cache::FileId,
+        file: VirtualFile,
+        prewarm_on_write: PrewarmOnWrite,
+    ) -> Self {
+        Self {
+            prewarm_on_write,
+            nwritten_blocks: 0,
+            page_cache_file_id,
+            file,
+        }
+    }
+
+    /// Return the byte range within `file` that has been written though `write_all`.
+    ///
+    /// The returned range would be invalidated by another `write_all`. To prevent that, we capture `&_`.
+    fn written_range(&self) -> (impl Deref<Target = Range<usize>> + '_) {
+        let nwritten_blocks = usize::try_from(self.nwritten_blocks).unwrap();
+        struct Wrapper(Range<usize>);
+        impl Deref for Wrapper {
+            type Target = Range<usize>;
+            fn deref(&self) -> &Range<usize> {
+                &self.0
+            }
+        }
+        Wrapper(0..nwritten_blocks * PAGE_SZ)
+    }
+}
+
+impl crate::virtual_file::owned_buffers_io::write::OwnedAsyncWriter for PreWarmingWriter {
+    async fn write_all<Buf: tokio_epoll_uring::IoBuf + Send>(
+        &mut self,
+        buf: FullSlice<Buf>,
+        ctx: &RequestContext,
+    ) -> std::io::Result<(usize, FullSlice<Buf>)> {
+        let buflen = buf.len();
+        assert_eq!(
+            buflen % PAGE_SZ,
+            0,
+            "{buflen} ; we know TAIL_SZ is a PAGE_SZ multiple, and write_buffered_borrowed is used"
+        );
+
+        // Do the IO.
+        let buf = match self.file.write_all(buf, ctx).await {
+            (buf, Ok(nwritten)) => {
+                assert_eq!(nwritten, buflen);
+                buf
+            }
+            (_, Err(e)) => {
+                return Err(std::io::Error::new(
+                    ErrorKind::Other,
+                    // order error before path because path is long and error is short
+                    format!(
+                        "ephemeral_file: write_blob: write-back tail self.nwritten_blocks={}, buflen={}, {:#}: {}",
+                        self.nwritten_blocks, buflen, e, self.file.path,
+                    ),
+                ));
+            }
+        };
+
+        let nblocks = buflen / PAGE_SZ;
+        let nblocks32 = u32::try_from(nblocks).unwrap();
+
+        if matches!(self.prewarm_on_write, PrewarmOnWrite::Yes) {
+            // Pre-warm page cache with the contents.
+            // At least in isolated bulk ingest benchmarks (test_bulk_insert.py), the pre-warming
+            // benefits the code that writes InMemoryLayer=>L0 layers.
+
+            let cache = page_cache::get();
+            static CTX: Lazy<RequestContext> = Lazy::new(|| {
+                RequestContext::new(
+                    crate::task_mgr::TaskKind::EphemeralFilePreWarmPageCache,
+                    crate::context::DownloadBehavior::Error,
+                )
+            });
+            for blknum_in_buffer in 0..nblocks {
+                let blk_in_buffer =
+                    &buf[blknum_in_buffer * PAGE_SZ..(blknum_in_buffer + 1) * PAGE_SZ];
+                let blknum = self
+                    .nwritten_blocks
+                    .checked_add(blknum_in_buffer as u32)
+                    .unwrap();
+                match cache
+                    .read_immutable_buf(self.page_cache_file_id, blknum, &CTX)
+                    .await
+                {
+                    Err(e) => {
+                        error!("ephemeral_file write_blob failed to get immutable buf to pre-warm page cache: {e:?}");
+                        // fail gracefully, it's not the end of the world if we can't pre-warm the cache here
+                    }
+                    Ok(v) => match v {
+                        page_cache::ReadBufResult::Found(_guard) => {
+                            // This function takes &mut self, so, it shouldn't be possible to reach this point.
+                            unreachable!("we just wrote block {blknum} to the VirtualFile, which is owned by Self, \
+                                      and this function takes &mut self, so, no concurrent read_blk is possible");
+                        }
+                        page_cache::ReadBufResult::NotFound(mut write_guard) => {
+                            write_guard.copy_from_slice(blk_in_buffer);
+                            let _ = write_guard.mark_valid();
+                        }
+                    },
+                }
+            }
+        }
+
+        self.nwritten_blocks = self.nwritten_blocks.checked_add(nblocks32).unwrap();
+        Ok((buflen, buf))
+    }
+}

pageserver/src/tenant/metadata.rs

@@ -565,7 +565,7 @@ mod tests {
         );
         let expected_bytes = vec![
             /* TimelineMetadataHeader */
-            74, 104, 158, 105, 0, 70, 0, 4, // checksum, size, format_version (4 + 2 + 2)
+            4, 37, 101, 34, 0, 70, 0, 4, // checksum, size, format_version (4 + 2 + 2)
             /* TimelineMetadataBodyV2 */
             0, 0, 0, 0, 0, 0, 2, 0, // disk_consistent_lsn (8 bytes)
             1, 0, 0, 0, 0, 0, 0, 1, 0, // prev_record_lsn (9 bytes)
@@ -574,7 +574,7 @@ mod tests {
             0, 0, 0, 0, 0, 0, 0, 0, // ancestor_lsn (8 bytes)
             0, 0, 0, 0, 0, 0, 0, 0, // latest_gc_cutoff_lsn (8 bytes)
             0, 0, 0, 0, 0, 0, 0, 0, // initdb_lsn (8 bytes)
-            0, 0, 0, 16, // pg_version (4 bytes)
+            0, 0, 0, 15, // pg_version (4 bytes)
             /* padding bytes */
             0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
             0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,

pageserver/src/tenant/remote_timeline_client.rs

@@ -1728,7 +1728,7 @@ impl RemoteTimelineClient {
             task_mgr::spawn(
                 &self.runtime,
                 TaskKind::RemoteUploadTask,
-                self.tenant_shard_id,
+                Some(self.tenant_shard_id),
                 Some(self.timeline_id),
                 "remote upload",
                 async move {

pageserver/src/tenant/secondary.rs

@@ -8,7 +8,6 @@ use std::{sync::Arc, time::SystemTime};
 use crate::{
     context::RequestContext,
     disk_usage_eviction_task::DiskUsageEvictionInfo,
-    metrics::SECONDARY_HEATMAP_TOTAL_SIZE,
     task_mgr::{self, TaskKind, BACKGROUND_RUNTIME},
 };
 
@@ -106,9 +105,6 @@ pub(crate) struct SecondaryTenant {
 
     // Sum of layer sizes on local disk
     pub(super) resident_size_metric: UIntGauge,
-
-    // Sum of layer sizes in the most recently downloaded heatmap
-    pub(super) heatmap_total_size_metric: UIntGauge,
 }
 
 impl Drop for SecondaryTenant {
@@ -116,7 +112,6 @@ impl Drop for SecondaryTenant {
         let tenant_id = self.tenant_shard_id.tenant_id.to_string();
         let shard_id = format!("{}", self.tenant_shard_id.shard_slug());
         let _ = SECONDARY_RESIDENT_PHYSICAL_SIZE.remove_label_values(&[&tenant_id, &shard_id]);
-        let _ = SECONDARY_HEATMAP_TOTAL_SIZE.remove_label_values(&[&tenant_id, &shard_id]);
     }
 }
 
@@ -133,10 +128,6 @@ impl SecondaryTenant {
             .get_metric_with_label_values(&[&tenant_id, &shard_id])
             .unwrap();
 
-        let heatmap_total_size_metric = SECONDARY_HEATMAP_TOTAL_SIZE
-            .get_metric_with_label_values(&[&tenant_id, &shard_id])
-            .unwrap();
-
         Arc::new(Self {
             tenant_shard_id,
             // todo: shall we make this a descendent of the
@@ -154,7 +145,6 @@ impl SecondaryTenant {
             progress: std::sync::Mutex::default(),
 
             resident_size_metric,
-            heatmap_total_size_metric,
         })
     }
 

pageserver/src/tenant/secondary/downloader.rs

@@ -829,12 +829,6 @@ impl<'a> TenantDownloader<'a> {
             layers_downloaded: 0,
             bytes_downloaded: 0,
         };
-
-        // Also expose heatmap bytes_total as a metric
-        self.secondary_state
-            .heatmap_total_size_metric
-            .set(heatmap_stats.bytes);
-
         // Accumulate list of things to delete while holding the detail lock, for execution after dropping the lock
         let mut delete_layers = Vec::new();
         let mut delete_timelines = Vec::new();

pageserver/src/tenant/secondary/heatmap.rs

@@ -29,16 +29,16 @@ pub(super) struct HeatMapTenant {
 #[derive(Serialize, Deserialize)]
 pub(crate) struct HeatMapTimeline {
     #[serde_as(as = "DisplayFromStr")]
-    pub(crate) timeline_id: TimelineId,
+    pub(super) timeline_id: TimelineId,
 
-    pub(crate) layers: Vec<HeatMapLayer>,
+    pub(super) layers: Vec<HeatMapLayer>,
 }
 
 #[serde_as]
 #[derive(Serialize, Deserialize)]
 pub(crate) struct HeatMapLayer {
-    pub(crate) name: LayerName,
-    pub(crate) metadata: LayerFileMetadata,
+    pub(super) name: LayerName,
+    pub(super) metadata: LayerFileMetadata,
 
     #[serde_as(as = "TimestampSeconds<i64>")]
     pub(super) access_time: SystemTime,

pageserver/src/tenant/storage_layer.rs

@@ -2,7 +2,7 @@
 
 pub mod delta_layer;
 pub mod image_layer;
-pub mod inmemory_layer;
+pub(crate) mod inmemory_layer;
 pub(crate) mod layer;
 mod layer_desc;
 mod layer_name;

pageserver/src/tenant/storage_layer/inmemory_layer.rs

@@ -13,7 +13,7 @@ use crate::tenant::ephemeral_file::EphemeralFile;
 use crate::tenant::timeline::GetVectoredError;
 use crate::tenant::PageReconstructError;
 use crate::virtual_file::owned_buffers_io::io_buf_ext::IoBufExt;
-use crate::{l0_flush, page_cache};
+use crate::{l0_flush, page_cache, walrecord};
 use anyhow::{anyhow, Result};
 use camino::Utf8PathBuf;
 use pageserver_api::key::CompactKey;
@@ -33,7 +33,7 @@ use std::fmt::Write;
 use std::ops::Range;
 use std::sync::atomic::Ordering as AtomicOrdering;
 use std::sync::atomic::{AtomicU64, AtomicUsize};
-use tokio::sync::RwLock;
+use tokio::sync::{RwLock, RwLockWriteGuard};
 
 use super::{
     DeltaLayerWriter, PersistentLayerDesc, ValueReconstructSituation, ValuesReconstructState,
@@ -249,7 +249,9 @@ impl InMemoryLayer {
     /// debugging function to print out the contents of the layer
     ///
     /// this is likely completly unused
-    pub async fn dump(&self, _verbose: bool, _ctx: &RequestContext) -> Result<()> {
+    pub async fn dump(&self, verbose: bool, ctx: &RequestContext) -> Result<()> {
+        let inner = self.inner.read().await;
+
         let end_str = self.end_lsn_or_max();
 
         println!(
@@ -257,6 +259,39 @@ impl InMemoryLayer {
             self.timeline_id, self.start_lsn, end_str,
         );
 
+        if !verbose {
+            return Ok(());
+        }
+
+        let cursor = inner.file.block_cursor();
+        let mut buf = Vec::new();
+        for (key, vec_map) in inner.index.iter() {
+            for (lsn, pos) in vec_map.as_slice() {
+                let mut desc = String::new();
+                cursor.read_blob_into_buf(*pos, &mut buf, ctx).await?;
+                let val = Value::des(&buf);
+                match val {
+                    Ok(Value::Image(img)) => {
+                        write!(&mut desc, " img {} bytes", img.len())?;
+                    }
+                    Ok(Value::WalRecord(rec)) => {
+                        let wal_desc = walrecord::describe_wal_record(&rec).unwrap();
+                        write!(
+                            &mut desc,
+                            " rec {} bytes will_init: {} {}",
+                            buf.len(),
+                            rec.will_init(),
+                            wal_desc
+                        )?;
+                    }
+                    Err(err) => {
+                        write!(&mut desc, " DESERIALIZATION ERROR: {}", err)?;
+                    }
+                }
+                println!("  key {} at {}: {}", key, lsn, desc);
+            }
+        }
+
         Ok(())
     }
 
@@ -320,82 +355,6 @@ impl InMemoryLayer {
     }
 }
 
-/// Offset of a particular Value within a serialized batch.
-struct SerializedBatchOffset {
-    key: CompactKey,
-    lsn: Lsn,
-    /// offset in bytes from the start of the batch's buffer to the Value's serialized size header.
-    offset: u64,
-}
-
-pub struct SerializedBatch {
-    /// Blobs serialized in EphemeralFile's native format, ready for passing to [`EphemeralFile::write_raw`].
-    pub(crate) raw: Vec<u8>,
-
-    /// Index of values in [`Self::raw`], using offsets relative to the start of the buffer.
-    offsets: Vec<SerializedBatchOffset>,
-
-    /// The highest LSN of any value in the batch
-    pub(crate) max_lsn: Lsn,
-}
-
-impl SerializedBatch {
-    /// Write a blob length in the internal format of the EphemeralFile
-    pub(crate) fn write_blob_length(len: usize, cursor: &mut std::io::Cursor<Vec<u8>>) {
-        use std::io::Write;
-
-        if len < 0x80 {
-            // short one-byte length header
-            let len_buf = [len as u8];
-
-            cursor
-                .write_all(&len_buf)
-                .expect("Writing to Vec is infallible");
-        } else {
-            let mut len_buf = u32::to_be_bytes(len as u32);
-            len_buf[0] |= 0x80;
-            cursor
-                .write_all(&len_buf)
-                .expect("Writing to Vec is infallible");
-        }
-    }
-
-    pub fn from_values(batch: Vec<(CompactKey, Lsn, usize, Value)>) -> Self {
-        // Pre-allocate a big flat buffer to write into. This should be large but not huge: it is soft-limited in practice by
-        // [`crate::pgdatadir_mapping::DatadirModification::MAX_PENDING_BYTES`]
-        let buffer_size = batch.iter().map(|i| i.2).sum::<usize>() + 4 * batch.len();
-        let mut cursor = std::io::Cursor::new(Vec::<u8>::with_capacity(buffer_size));
-
-        let mut offsets: Vec<SerializedBatchOffset> = Vec::with_capacity(batch.len());
-        let mut max_lsn: Lsn = Lsn(0);
-        for (key, lsn, val_ser_size, val) in batch {
-            let relative_off = cursor.position();
-
-            Self::write_blob_length(val_ser_size, &mut cursor);
-            val.ser_into(&mut cursor)
-                .expect("Writing into in-memory buffer is infallible");
-
-            offsets.push(SerializedBatchOffset {
-                key,
-                lsn,
-                offset: relative_off,
-            });
-            max_lsn = std::cmp::max(max_lsn, lsn);
-        }
-
-        let buffer = cursor.into_inner();
-
-        // Assert that we didn't do any extra allocations while building buffer.
-        debug_assert!(buffer.len() <= buffer_size);
-
-        Self {
-            raw: buffer,
-            offsets,
-            max_lsn,
-        }
-    }
-}
-
 fn inmem_layer_display(mut f: impl Write, start_lsn: Lsn, end_lsn: Lsn) -> std::fmt::Result {
     write!(f, "inmem-{:016X}-{:016X}", start_lsn.0, end_lsn.0)
 }
@@ -456,20 +415,37 @@ impl InMemoryLayer {
         })
     }
 
-    // Write path.
-    pub async fn put_batch(
+    // Write operations
+
+    /// Common subroutine of the public put_wal_record() and put_page_image() functions.
+    /// Adds the page version to the in-memory tree
+    pub async fn put_value(
         &self,
-        serialized_batch: SerializedBatch,
+        key: CompactKey,
+        lsn: Lsn,
+        buf: &[u8],
         ctx: &RequestContext,
     ) -> Result<()> {
         let mut inner = self.inner.write().await;
         self.assert_writable();
+        self.put_value_locked(&mut inner, key, lsn, buf, ctx).await
+    }
 
-        let base_off = {
-            inner
+    async fn put_value_locked(
+        &self,
+        locked_inner: &mut RwLockWriteGuard<'_, InMemoryLayerInner>,
+        key: CompactKey,
+        lsn: Lsn,
+        buf: &[u8],
+        ctx: &RequestContext,
+    ) -> Result<()> {
+        trace!("put_value key {} at {}/{}", key, self.timeline_id, lsn);
+
+        let off = {
+            locked_inner
                 .file
-                .write_raw(
-                    &serialized_batch.raw,
+                .write_blob(
+                    buf,
                     &RequestContextBuilder::extend(ctx)
                         .page_content_kind(PageContentKind::InMemoryLayer)
                         .build(),
@@ -477,23 +453,15 @@ impl InMemoryLayer {
                 .await?
         };
 
-        for SerializedBatchOffset {
-            key,
-            lsn,
-            offset: relative_off,
-        } in serialized_batch.offsets
-        {
-            let off = base_off + relative_off;
-            let vec_map = inner.index.entry(key).or_default();
-            let old = vec_map.append_or_update_last(lsn, off).unwrap().0;
-            if old.is_some() {
-                // We already had an entry for this LSN. That's odd..
-                warn!("Key {} at {} already exists", key, lsn);
-            }
+        let vec_map = locked_inner.index.entry(key).or_default();
+        let old = vec_map.append_or_update_last(lsn, off).unwrap().0;
+        if old.is_some() {
+            // We already had an entry for this LSN. That's odd..
+            warn!("Key {} at {} already exists", key, lsn);
         }
 
-        let size = inner.file.len();
-        inner.resource_units.maybe_publish_size(size);
+        let size = locked_inner.file.len();
+        locked_inner.resource_units.maybe_publish_size(size);
 
         Ok(())
     }
@@ -568,6 +536,7 @@ impl InMemoryLayer {
 
         use l0_flush::Inner;
         let _concurrency_permit = match l0_flush_global_state {
+            Inner::PageCached => None,
             Inner::Direct { semaphore, .. } => Some(semaphore.acquire().await),
         };
 
@@ -599,6 +568,34 @@ impl InMemoryLayer {
         .await?;
 
         match l0_flush_global_state {
+            l0_flush::Inner::PageCached => {
+                let ctx = RequestContextBuilder::extend(ctx)
+                    .page_content_kind(PageContentKind::InMemoryLayer)
+                    .build();
+
+                let mut buf = Vec::new();
+
+                let cursor = inner.file.block_cursor();
+
+                for (key, vec_map) in inner.index.iter() {
+                    // Write all page versions
+                    for (lsn, pos) in vec_map.as_slice() {
+                        cursor.read_blob_into_buf(*pos, &mut buf, &ctx).await?;
+                        let will_init = Value::des(&buf)?.will_init();
+                        let (tmp, res) = delta_layer_writer
+                            .put_value_bytes(
+                                Key::from_compact(*key),
+                                *lsn,
+                                buf.slice_len(),
+                                will_init,
+                                &ctx,
+                            )
+                            .await;
+                        res?;
+                        buf = tmp.into_raw_slice().into_inner();
+                    }
+                }
+            }
             l0_flush::Inner::Direct { .. } => {
                 let file_contents: Vec<u8> = inner.file.load_to_vec(ctx).await?;
                 assert_eq!(

pageserver/src/tenant/storage_layer/split_writer.rs

@@ -208,8 +208,6 @@ impl SplitDeltaLayerWriter {
 
 #[cfg(test)]
 mod tests {
-    use rand::{RngCore, SeedableRng};
-
     use crate::{
         tenant::{
             harness::{TenantHarness, TIMELINE_ID},
@@ -231,10 +229,7 @@ mod tests {
     }
 
     fn get_large_img() -> Bytes {
-        let mut rng = rand::rngs::SmallRng::seed_from_u64(42);
-        let mut data = vec![0; 8192];
-        rng.fill_bytes(&mut data);
-        data.into()
+        vec![0; 8192].into()
     }
 
     #[tokio::test]

pageserver/src/tenant/tasks.rs

@@ -61,12 +61,21 @@ impl BackgroundLoopKind {
     }
 }
 
+static PERMIT_GAUGES: once_cell::sync::Lazy<
+    enum_map::EnumMap<BackgroundLoopKind, metrics::IntCounterPair>,
+> = once_cell::sync::Lazy::new(|| {
+    enum_map::EnumMap::from_array(std::array::from_fn(|i| {
+        let kind = <BackgroundLoopKind as enum_map::Enum>::from_usize(i);
+        crate::metrics::BACKGROUND_LOOP_SEMAPHORE_WAIT_GAUGE.with_label_values(&[kind.into()])
+    }))
+});
+
 /// Cancellation safe.
 pub(crate) async fn concurrent_background_tasks_rate_limit_permit(
     loop_kind: BackgroundLoopKind,
     _ctx: &RequestContext,
 ) -> tokio::sync::SemaphorePermit<'static> {
-    let _guard = crate::metrics::BACKGROUND_LOOP_SEMAPHORE.measure_acquisition(loop_kind);
+    let _guard = PERMIT_GAUGES[loop_kind].guard();
 
     pausable_failpoint!(
         "initial-size-calculation-permit-pause",
@@ -89,7 +98,7 @@ pub fn start_background_loops(
     task_mgr::spawn(
         BACKGROUND_RUNTIME.handle(),
         TaskKind::Compaction,
-        tenant_shard_id,
+        Some(tenant_shard_id),
         None,
         &format!("compactor for tenant {tenant_shard_id}"),
         {
@@ -112,7 +121,7 @@ pub fn start_background_loops(
     task_mgr::spawn(
         BACKGROUND_RUNTIME.handle(),
         TaskKind::GarbageCollector,
-        tenant_shard_id,
+        Some(tenant_shard_id),
         None,
         &format!("garbage collector for tenant {tenant_shard_id}"),
         {
@@ -135,7 +144,7 @@ pub fn start_background_loops(
     task_mgr::spawn(
         BACKGROUND_RUNTIME.handle(),
         TaskKind::IngestHousekeeping,
-        tenant_shard_id,
+        Some(tenant_shard_id),
         None,
         &format!("ingest housekeeping for tenant {tenant_shard_id}"),
         {

pageserver/src/tenant/timeline.rs

@@ -22,8 +22,8 @@ use handle::ShardTimelineId;
 use once_cell::sync::Lazy;
 use pageserver_api::{
     key::{
-        CompactKey, KEY_SIZE, METADATA_KEY_BEGIN_PREFIX, METADATA_KEY_END_PREFIX,
-        NON_INHERITED_RANGE, NON_INHERITED_SPARSE_RANGE,
+        KEY_SIZE, METADATA_KEY_BEGIN_PREFIX, METADATA_KEY_END_PREFIX, NON_INHERITED_RANGE,
+        NON_INHERITED_SPARSE_RANGE,
     },
     keyspace::{KeySpaceAccum, KeySpaceRandomAccum, SparseKeyPartitioning},
     models::{
@@ -44,8 +44,10 @@ use tokio::{
 use tokio_util::sync::CancellationToken;
 use tracing::*;
 use utils::{
+    bin_ser::BeSer,
     fs_ext, pausable_failpoint,
     sync::gate::{Gate, GateGuard},
+    vec_map::VecMap,
 };
 
 use std::pin::pin;
@@ -135,10 +137,7 @@ use self::layer_manager::LayerManager;
 use self::logical_size::LogicalSize;
 use self::walreceiver::{WalReceiver, WalReceiverConf};
 
-use super::{
-    config::TenantConf, storage_layer::inmemory_layer, storage_layer::LayerVisibilityHint,
-    upload_queue::NotInitialized,
-};
+use super::{config::TenantConf, storage_layer::LayerVisibilityHint, upload_queue::NotInitialized};
 use super::{debug_assert_current_span_has_tenant_and_timeline_id, AttachedTenantConf};
 use super::{remote_timeline_client::index::IndexPart, storage_layer::LayerFringe};
 use super::{
@@ -2282,7 +2281,7 @@ impl Timeline {
         task_mgr::spawn(
             task_mgr::BACKGROUND_RUNTIME.handle(),
             task_mgr::TaskKind::LayerFlushTask,
-            self.tenant_shard_id,
+            Some(self.tenant_shard_id),
             Some(self.timeline_id),
             "layer flush task",
             async move {
@@ -2636,7 +2635,7 @@ impl Timeline {
         task_mgr::spawn(
             task_mgr::BACKGROUND_RUNTIME.handle(),
             task_mgr::TaskKind::InitialLogicalSizeCalculation,
-            self.tenant_shard_id,
+            Some(self.tenant_shard_id),
             Some(self.timeline_id),
             "initial size calculation",
             // NB: don't log errors here, task_mgr will do that.
@@ -2804,7 +2803,7 @@ impl Timeline {
         task_mgr::spawn(
             task_mgr::BACKGROUND_RUNTIME.handle(),
             task_mgr::TaskKind::OndemandLogicalSizeCalculation,
-            self.tenant_shard_id,
+            Some(self.tenant_shard_id),
             Some(self.timeline_id),
             "ondemand logical size calculation",
             async move {
@@ -2978,7 +2977,11 @@ impl Timeline {
                 LayerVisibilityHint::Visible => {
                     // Layer is visible to one or more read LSNs: elegible for inclusion in layer map
                     let last_activity_ts = layer.latest_activity();
-                    Some((layer.layer_desc(), layer.metadata(), last_activity_ts))
+                    Some(HeatMapLayer::new(
+                        layer.layer_desc().layer_name(),
+                        layer.metadata(),
+                        last_activity_ts,
+                    ))
                 }
                 LayerVisibilityHint::Covered => {
                     // Layer is resident but unlikely to be read: not elegible for inclusion in heatmap.
@@ -2987,23 +2990,7 @@ impl Timeline {
             }
         });
 
-        let mut layers = resident.collect::<Vec<_>>();
-
-        // Sort layers in order of which to download first.  For a large set of layers to download, we
-        // want to prioritize those layers which are most likely to still be in the resident many minutes
-        // or hours later:
-        // - Download L0s last, because they churn the fastest: L0s on a fast-writing tenant might
-        //   only exist for a few minutes before being compacted into L1s.
-        // - For L1 & image layers, download most recent LSNs first: the older the LSN, the sooner
-        //   the layer is likely to be covered by an image layer during compaction.
-        layers.sort_by_key(|(desc, _meta, _atime)| {
-            std::cmp::Reverse((!LayerMap::is_l0(&desc.key_range), desc.lsn_range.end))
-        });
-
-        let layers = layers
-            .into_iter()
-            .map(|(desc, meta, atime)| HeatMapLayer::new(desc.layer_name(), meta, atime))
-            .collect();
+        let layers = resident.collect();
 
         Some(HeatMapTimeline::new(self.timeline_id, layers))
     }
@@ -3590,6 +3577,34 @@ impl Timeline {
                 return Err(FlushLayerError::Cancelled);
             }
 
+            // FIXME(auxfilesv2): support multiple metadata key partitions might need initdb support as well?
+            // This code path will not be hit during regression tests. After #7099 we have a single partition
+            // with two key ranges. If someone wants to fix initdb optimization in the future, this might need
+            // to be fixed.
+
+            // For metadata, always create delta layers.
+            let delta_layer = if !metadata_partition.parts.is_empty() {
+                assert_eq!(
+                    metadata_partition.parts.len(),
+                    1,
+                    "currently sparse keyspace should only contain a single metadata keyspace"
+                );
+                let metadata_keyspace = &metadata_partition.parts[0];
+                self.create_delta_layer(
+                    &frozen_layer,
+                    Some(
+                        metadata_keyspace.0.ranges.first().unwrap().start
+                            ..metadata_keyspace.0.ranges.last().unwrap().end,
+                    ),
+                    ctx,
+                )
+                .await
+                .map_err(|e| FlushLayerError::from_anyhow(self, e))?
+            } else {
+                None
+            };
+
+            // For image layers, we add them immediately into the layer map.
             let mut layers_to_upload = Vec::new();
             layers_to_upload.extend(
                 self.create_image_layers(
@@ -3600,27 +3615,13 @@ impl Timeline {
                 )
                 .await?,
             );
-            if !metadata_partition.parts.is_empty() {
-                assert_eq!(
-                    metadata_partition.parts.len(),
-                    1,
-                    "currently sparse keyspace should only contain a single metadata keyspace"
-                );
-                layers_to_upload.extend(
-                    self.create_image_layers(
-                        // Safety: create_image_layers treat sparse keyspaces differently that it does not scan
-                        // every single key within the keyspace, and therefore, it's safe to force converting it
-                        // into a dense keyspace before calling this function.
-                        &metadata_partition.into_dense(),
-                        self.initdb_lsn,
-                        ImageLayerCreationMode::Initial,
-                        ctx,
-                    )
-                    .await?,
-                );
-            }
 
-            (layers_to_upload, None)
+            if let Some(delta_layer) = delta_layer {
+                layers_to_upload.push(delta_layer.clone());
+                (layers_to_upload, Some(delta_layer))
+            } else {
+                (layers_to_upload, None)
+            }
         } else {
             // Normal case, write out a L0 delta layer file.
             // `create_delta_layer` will not modify the layer map.
@@ -4030,6 +4031,8 @@ impl Timeline {
         mode: ImageLayerCreationMode,
         start: Key,
     ) -> Result<ImageLayerCreationOutcome, CreateImageLayersError> {
+        assert!(!matches!(mode, ImageLayerCreationMode::Initial));
+
         // Metadata keys image layer creation.
         let mut reconstruct_state = ValuesReconstructState::default();
         let data = self
@@ -4195,13 +4198,15 @@ impl Timeline {
                         "metadata keys must be partitioned separately"
                     );
                 }
+                if mode == ImageLayerCreationMode::Initial {
+                    return Err(CreateImageLayersError::Other(anyhow::anyhow!("no image layer should be created for metadata keys when flushing frozen layers")));
+                }
                 if mode == ImageLayerCreationMode::Try && !check_for_image_layers {
                     // Skip compaction if there are not enough updates. Metadata compaction will do a scan and
                     // might mess up with evictions.
                     start = img_range.end;
                     continue;
                 }
-                // For initial and force modes, we always generate image layers for metadata keys.
             } else if let ImageLayerCreationMode::Try = mode {
                 // check_for_image_layers = false -> skip
                 // check_for_image_layers = true -> check time_for_new_image_layer -> skip/generate
@@ -4209,8 +4214,7 @@ impl Timeline {
                     start = img_range.end;
                     continue;
                 }
-            }
-            if let ImageLayerCreationMode::Force = mode {
+            } else if let ImageLayerCreationMode::Force = mode {
                 // When forced to create image layers, we might try and create them where they already
                 // exist.  This mode is only used in tests/debug.
                 let layers = self.layers.read().await;
@@ -4224,7 +4228,6 @@ impl Timeline {
                         img_range.start,
                         img_range.end
                     );
-                    start = img_range.end;
                     continue;
                 }
             }
@@ -4513,7 +4516,6 @@ impl DurationRecorder {
 /// the layer descriptor requires the user to provide the ranges, which should cover all
 /// keys specified in the `data` field.
 #[cfg(test)]
-#[derive(Clone)]
 pub struct DeltaLayerTestDesc {
     pub lsn_range: Range<Lsn>,
     pub key_range: Range<Key>,
@@ -4543,13 +4545,6 @@ impl DeltaLayerTestDesc {
             data,
         }
     }
-
-    pub(crate) fn layer_name(&self) -> LayerName {
-        LayerName::Delta(super::storage_layer::DeltaLayerName {
-            key_range: self.key_range.clone(),
-            lsn_range: self.lsn_range.clone(),
-        })
-    }
 }
 
 impl Timeline {
@@ -5147,7 +5142,7 @@ impl Timeline {
         let task_id = task_mgr::spawn(
             task_mgr::BACKGROUND_RUNTIME.handle(),
             task_mgr::TaskKind::DownloadAllRemoteLayers,
-            self.tenant_shard_id,
+            Some(self.tenant_shard_id),
             Some(self.timeline_id),
             "download all remote layers task",
             async move {
@@ -5575,6 +5570,44 @@ enum OpenLayerAction {
 }
 
 impl<'a> TimelineWriter<'a> {
+    /// Put a new page version that can be constructed from a WAL record
+    ///
+    /// This will implicitly extend the relation, if the page is beyond the
+    /// current end-of-file.
+    pub(crate) async fn put(
+        &mut self,
+        key: Key,
+        lsn: Lsn,
+        value: &Value,
+        ctx: &RequestContext,
+    ) -> anyhow::Result<()> {
+        // Avoid doing allocations for "small" values.
+        // In the regression test suite, the limit of 256 avoided allocations in 95% of cases:
+        // https://github.com/neondatabase/neon/pull/5056#discussion_r1301975061
+        let mut buf = smallvec::SmallVec::<[u8; 256]>::new();
+        value.ser_into(&mut buf)?;
+        let buf_size: u64 = buf.len().try_into().expect("oversized value buf");
+
+        let action = self.get_open_layer_action(lsn, buf_size);
+        let layer = self.handle_open_layer_action(lsn, action, ctx).await?;
+        let res = layer.put_value(key.to_compact(), lsn, &buf, ctx).await;
+
+        if res.is_ok() {
+            // Update the current size only when the entire write was ok.
+            // In case of failures, we may have had partial writes which
+            // render the size tracking out of sync. That's ok because
+            // the checkpoint distance should be significantly smaller
+            // than the S3 single shot upload limit of 5GiB.
+            let state = self.write_guard.as_mut().unwrap();
+
+            state.current_size += buf_size;
+            state.prev_lsn = Some(lsn);
+            state.max_lsn = std::cmp::max(state.max_lsn, Some(lsn));
+        }
+
+        res
+    }
+
     async fn handle_open_layer_action(
         &mut self,
         at: Lsn,
@@ -5680,58 +5713,18 @@ impl<'a> TimelineWriter<'a> {
     }
 
     /// Put a batch of keys at the specified Lsns.
+    ///
+    /// The batch is sorted by Lsn (enforced by usage of [`utils::vec_map::VecMap`].
     pub(crate) async fn put_batch(
         &mut self,
-        batch: Vec<(CompactKey, Lsn, usize, Value)>,
+        batch: VecMap<Lsn, (Key, Value)>,
         ctx: &RequestContext,
     ) -> anyhow::Result<()> {
-        if batch.is_empty() {
-            return Ok(());
+        for (lsn, (key, val)) in batch {
+            self.put(key, lsn, &val, ctx).await?
         }
 
-        let serialized_batch = inmemory_layer::SerializedBatch::from_values(batch);
-        let batch_max_lsn = serialized_batch.max_lsn;
-        let buf_size: u64 = serialized_batch.raw.len() as u64;
-
-        let action = self.get_open_layer_action(batch_max_lsn, buf_size);
-        let layer = self
-            .handle_open_layer_action(batch_max_lsn, action, ctx)
-            .await?;
-
-        let res = layer.put_batch(serialized_batch, ctx).await;
-
-        if res.is_ok() {
-            // Update the current size only when the entire write was ok.
-            // In case of failures, we may have had partial writes which
-            // render the size tracking out of sync. That's ok because
-            // the checkpoint distance should be significantly smaller
-            // than the S3 single shot upload limit of 5GiB.
-            let state = self.write_guard.as_mut().unwrap();
-
-            state.current_size += buf_size;
-            state.prev_lsn = Some(batch_max_lsn);
-            state.max_lsn = std::cmp::max(state.max_lsn, Some(batch_max_lsn));
-        }
-
-        res
-    }
-
-    #[cfg(test)]
-    /// Test helper, for tests that would like to poke individual values without composing a batch
-    pub(crate) async fn put(
-        &mut self,
-        key: Key,
-        lsn: Lsn,
-        value: &Value,
-        ctx: &RequestContext,
-    ) -> anyhow::Result<()> {
-        use utils::bin_ser::BeSer;
-        let val_ser_size = value.serialized_size().unwrap() as usize;
-        self.put_batch(
-            vec![(key.to_compact(), lsn, val_ser_size, value.clone())],
-            ctx,
-        )
-        .await
+        Ok(())
     }
 
     pub(crate) async fn delete_batch(
@@ -5775,110 +5768,12 @@ fn is_send() {
 
 #[cfg(test)]
 mod tests {
-    use pageserver_api::key::Key;
     use utils::{id::TimelineId, lsn::Lsn};
 
-    use crate::{
-        repository::Value,
-        tenant::{
-            harness::{test_img, TenantHarness},
-            layer_map::LayerMap,
-            storage_layer::{Layer, LayerName},
-            timeline::{DeltaLayerTestDesc, EvictionError},
-            Timeline,
-        },
+    use crate::tenant::{
+        harness::TenantHarness, storage_layer::Layer, timeline::EvictionError, Timeline,
     };
 
-    #[tokio::test]
-    async fn test_heatmap_generation() {
-        let harness = TenantHarness::create("heatmap_generation").await.unwrap();
-
-        let covered_delta = DeltaLayerTestDesc::new_with_inferred_key_range(
-            Lsn(0x10)..Lsn(0x20),
-            vec![(
-                Key::from_hex("620000000033333333444444445500000000").unwrap(),
-                Lsn(0x11),
-                Value::Image(test_img("foo")),
-            )],
-        );
-        let visible_delta = DeltaLayerTestDesc::new_with_inferred_key_range(
-            Lsn(0x10)..Lsn(0x20),
-            vec![(
-                Key::from_hex("720000000033333333444444445500000000").unwrap(),
-                Lsn(0x11),
-                Value::Image(test_img("foo")),
-            )],
-        );
-        let l0_delta = DeltaLayerTestDesc::new(
-            Lsn(0x20)..Lsn(0x30),
-            Key::from_hex("000000000000000000000000000000000000").unwrap()
-                ..Key::from_hex("FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF").unwrap(),
-            vec![(
-                Key::from_hex("720000000033333333444444445500000000").unwrap(),
-                Lsn(0x25),
-                Value::Image(test_img("foo")),
-            )],
-        );
-        let delta_layers = vec![
-            covered_delta.clone(),
-            visible_delta.clone(),
-            l0_delta.clone(),
-        ];
-
-        let image_layer = (
-            Lsn(0x40),
-            vec![(
-                Key::from_hex("620000000033333333444444445500000000").unwrap(),
-                test_img("bar"),
-            )],
-        );
-        let image_layers = vec![image_layer];
-
-        let (tenant, ctx) = harness.load().await;
-        let timeline = tenant
-            .create_test_timeline_with_layers(
-                TimelineId::generate(),
-                Lsn(0x10),
-                14,
-                &ctx,
-                delta_layers,
-                image_layers,
-                Lsn(0x100),
-            )
-            .await
-            .unwrap();
-
-        // Layer visibility is an input to heatmap generation, so refresh it first
-        timeline.update_layer_visibility().await.unwrap();
-
-        let heatmap = timeline
-            .generate_heatmap()
-            .await
-            .expect("Infallible while timeline is not shut down");
-
-        assert_eq!(heatmap.timeline_id, timeline.timeline_id);
-
-        // L0 should come last
-        assert_eq!(heatmap.layers.last().unwrap().name, l0_delta.layer_name());
-
-        let mut last_lsn = Lsn::MAX;
-        for layer in heatmap.layers {
-            // Covered layer should be omitted
-            assert!(layer.name != covered_delta.layer_name());
-
-            let layer_lsn = match &layer.name {
-                LayerName::Delta(d) => d.lsn_range.end,
-                LayerName::Image(i) => i.lsn,
-            };
-
-            // Apart from L0s, newest Layers should come first
-            if !LayerMap::is_l0(layer.name.key_range()) {
-                assert!(layer_lsn <= last_lsn);
-                last_lsn = layer_lsn;
-            }
-        }
-    }
-
     #[tokio::test]
     async fn two_layer_eviction_attempts_at_the_same_time() {
         let harness = TenantHarness::create("two_layer_eviction_attempts_at_the_same_time")

pageserver/src/tenant/timeline/delete.rs

@@ -395,7 +395,7 @@ impl DeleteTimelineFlow {
         task_mgr::spawn(
             task_mgr::BACKGROUND_RUNTIME.handle(),
             TaskKind::TimelineDeletionWorker,
-            tenant_shard_id,
+            Some(tenant_shard_id),
             Some(timeline_id),
             "timeline_delete",
             async move {

pageserver/src/tenant/timeline/eviction_task.rs

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

pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs

@@ -27,8 +27,8 @@ use super::TaskStateUpdate;
 use crate::{
     context::RequestContext,
     metrics::{LIVE_CONNECTIONS, WALRECEIVER_STARTED_CONNECTIONS, WAL_INGEST},
-    pgdatadir_mapping::DatadirModification,
-    task_mgr::{TaskKind, WALRECEIVER_RUNTIME},
+    task_mgr::TaskKind,
+    task_mgr::WALRECEIVER_RUNTIME,
     tenant::{debug_assert_current_span_has_tenant_and_timeline_id, Timeline, WalReceiverInfo},
     walingest::WalIngest,
     walrecord::DecodedWALRecord,
@@ -345,10 +345,7 @@ pub(super) async fn handle_walreceiver_connection(
                         // Commit every ingest_batch_size records. Even if we filtered out
                         // all records, we still need to call commit to advance the LSN.
                         uncommitted_records += 1;
-                        if uncommitted_records >= ingest_batch_size
-                            || modification.approx_pending_bytes()
-                                > DatadirModification::MAX_PENDING_BYTES
-                        {
+                        if uncommitted_records >= ingest_batch_size {
                             WAL_INGEST
                                 .records_committed
                                 .inc_by(uncommitted_records - filtered_records);

pageserver/src/virtual_file.rs

@@ -756,23 +756,11 @@ impl VirtualFile {
         })
     }
 
-    /// The function aborts the process if the error is fatal.
     async fn write_at<B: IoBuf + Send>(
         &self,
         buf: FullSlice<B>,
         offset: u64,
         _ctx: &RequestContext, /* TODO: use for metrics: https://github.com/neondatabase/neon/issues/6107 */
-    ) -> (FullSlice<B>, Result<usize, Error>) {
-        let (slice, result) = self.write_at_inner(buf, offset, _ctx).await;
-        let result = result.maybe_fatal_err("write_at");
-        (slice, result)
-    }
-
-    async fn write_at_inner<B: IoBuf + Send>(
-        &self,
-        buf: FullSlice<B>,
-        offset: u64,
-        _ctx: &RequestContext, /* TODO: use for metrics: https://github.com/neondatabase/neon/issues/6107 */
     ) -> (FullSlice<B>, Result<usize, Error>) {
         let file_guard = match self.lock_file().await {
             Ok(file_guard) => file_guard,

pgxn/neon/file_cache.c

@@ -41,8 +41,6 @@
 
 #include "hll.h"
 
-#define CriticalAssert(cond) do if (!(cond)) elog(PANIC, "Assertion %s failed at %s:%d: ", #cond, __FILE__, __LINE__); while (0)
-
 /*
  * Local file cache is used to temporary store relations pages in local file system.
  * All blocks of all relations are stored inside one file and addressed using shared hash map.
@@ -53,43 +51,19 @@
  *
  * Cache is always reconstructed at node startup, so we do not need to save mapping somewhere and worry about
  * its consistency.
-
- *
- * ## Holes
- *
- * The LFC can be resized on the fly, up to a maximum size that's determined
- * at server startup (neon.max_file_cache_size). After server startup, we
- * expand the underlying file when needed, until it reaches the soft limit
- * (neon.file_cache_size_limit). If the soft limit is later reduced, we shrink
- * the LFC by punching holes in the underlying file with a
- * fallocate(FALLOC_FL_PUNCH_HOLE) call. The nominal size of the file doesn't
- * shrink, but the disk space it uses does.
- *
- * Each hole is tracked by a dummy FileCacheEntry, which are kept in the
- * 'holes' linked list. They are entered into the chunk hash table, with a
- * special key where the blockNumber is used to store the 'offset' of the
- * hole, and all other fields are zero. Holes are never looked up in the hash
- * table, we only enter them there to have a FileCacheEntry that we can keep
- * in the linked list. If the soft limit is raised again, we reuse the holes
- * before extending the nominal size of the file.
  */
 
 /* Local file storage allocation chunk.
- * Should be power of two. Using larger than page chunks can
+ * Should be power of two and not less than 32. Using larger than page chunks can
  * 1. Reduce hash-map memory footprint: 8TB database contains billion pages
  *    and size of hash entry is 40 bytes, so we need 40Gb just for hash map.
  *    1Mb chunks can reduce hash map size to 320Mb.
  * 2. Improve access locality, subsequent pages will be allocated together improving seqscan speed
  */
 #define BLOCKS_PER_CHUNK	128 /* 1Mb chunk */
-/*
- * Smaller chunk seems to be better for OLTP workload
- */
-// #define BLOCKS_PER_CHUNK	8 /* 64kb chunk */
 #define MB					((uint64)1024*1024)
 
 #define SIZE_MB_TO_CHUNKS(size) ((uint32)((size) * MB / BLCKSZ / BLOCKS_PER_CHUNK))
-#define CHUNK_BITMAP_SIZE ((BLOCKS_PER_CHUNK + 31) / 32)
 
 typedef struct FileCacheEntry
 {
@@ -97,8 +71,8 @@ typedef struct FileCacheEntry
 	uint32		hash;
 	uint32		offset;
 	uint32		access_count;
-	uint32		bitmap[CHUNK_BITMAP_SIZE];
-	dlist_node	list_node;		/* LRU/holes list node */
+	uint32		bitmap[BLOCKS_PER_CHUNK / 32];
+	dlist_node	lru_node;		/* LRU list node */
 } FileCacheEntry;
 
 typedef struct FileCacheControl
@@ -113,7 +87,6 @@ typedef struct FileCacheControl
 	uint64		writes;
 	dlist_head	lru;			/* double linked list for LRU replacement
 								 * algorithm */
-	dlist_head  holes;          /* double linked list of punched holes */
 	HyperLogLogState wss_estimation; /* estimation of working set size */
 } FileCacheControl;
 
@@ -162,7 +135,6 @@ lfc_disable(char const *op)
 		lfc_ctl->used = 0;
 		lfc_ctl->limit = 0;
 		dlist_init(&lfc_ctl->lru);
-		dlist_init(&lfc_ctl->holes);
 
 		if (lfc_desc > 0)
 		{
@@ -242,18 +214,18 @@ lfc_shmem_startup(void)
 	if (!found)
 	{
 		int			fd;
-		uint32		n_chunks = SIZE_MB_TO_CHUNKS(lfc_max_size);
+		uint32		lfc_size = SIZE_MB_TO_CHUNKS(lfc_max_size);
 
 		lfc_lock = (LWLockId) GetNamedLWLockTranche("lfc_lock");
 		info.keysize = sizeof(BufferTag);
 		info.entrysize = sizeof(FileCacheEntry);
 
 		/*
-		 * n_chunks+1 because we add new element to hash table before eviction
+		 * lfc_size+1 because we add new element to hash table before eviction
 		 * of victim
 		 */
 		lfc_hash = ShmemInitHash("lfc_hash",
-								 n_chunks + 1, n_chunks + 1,
+								 lfc_size + 1, lfc_size + 1,
 								 &info,
 								 HASH_ELEM | HASH_BLOBS);
 		lfc_ctl->generation = 0;
@@ -263,7 +235,6 @@ lfc_shmem_startup(void)
 		lfc_ctl->misses = 0;
 		lfc_ctl->writes = 0;
 		dlist_init(&lfc_ctl->lru);
-		dlist_init(&lfc_ctl->holes);
 
 		/* Initialize hyper-log-log structure for estimating working set size */
 		initSHLL(&lfc_ctl->wss_estimation);
@@ -339,31 +310,14 @@ lfc_change_limit_hook(int newval, void *extra)
 		 * Shrink cache by throwing away least recently accessed chunks and
 		 * returning their space to file system
 		 */
-		FileCacheEntry *victim = dlist_container(FileCacheEntry, list_node, dlist_pop_head_node(&lfc_ctl->lru));
-		FileCacheEntry *hole;
-		uint32		offset = victim->offset;
-		uint32		hash;
-		bool		found;
-		BufferTag	holetag;
+		FileCacheEntry *victim = dlist_container(FileCacheEntry, lru_node, dlist_pop_head_node(&lfc_ctl->lru));
 
-		CriticalAssert(victim->access_count == 0);
+		Assert(victim->access_count == 0);
 #ifdef FALLOC_FL_PUNCH_HOLE
 		if (fallocate(lfc_desc, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, (off_t) victim->offset * BLOCKS_PER_CHUNK * BLCKSZ, BLOCKS_PER_CHUNK * BLCKSZ) < 0)
 			neon_log(LOG, "Failed to punch hole in file: %m");
 #endif
-		/* We remove the old entry, and re-enter a hole to the hash table */
 		hash_search_with_hash_value(lfc_hash, &victim->key, victim->hash, HASH_REMOVE, NULL);
-
-		memset(&holetag, 0, sizeof(holetag));
-		holetag.blockNum = offset;
-		hash = get_hash_value(lfc_hash, &holetag);
-		hole = hash_search_with_hash_value(lfc_hash, &holetag, hash, HASH_ENTER, &found);
-		hole->hash = hash;
-		hole->offset = offset;
-		hole->access_count = 0;
-		CriticalAssert(!found);
-		dlist_push_tail(&lfc_ctl->holes, &hole->list_node);
-
 		lfc_ctl->used -= 1;
 	}
 	lfc_ctl->limit = new_size;
@@ -455,8 +409,6 @@ lfc_cache_contains(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno)
 	CopyNRelFileInfoToBufTag(tag, rinfo);
 	tag.forkNum = forkNum;
 	tag.blockNum = blkno & ~(BLOCKS_PER_CHUNK - 1);
-
-	CriticalAssert(BufTagGetRelNumber(&tag) != InvalidRelFileNumber);
 	hash = get_hash_value(lfc_hash, &tag);
 
 	LWLockAcquire(lfc_lock, LW_SHARED);
@@ -488,7 +440,6 @@ lfc_evict(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno)
 	tag.forkNum = forkNum;
 	tag.blockNum = (blkno & ~(BLOCKS_PER_CHUNK - 1));
 
-	CriticalAssert(BufTagGetRelNumber(&tag) != InvalidRelFileNumber);
 	hash = get_hash_value(lfc_hash, &tag);
 
 	LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
@@ -519,7 +470,7 @@ lfc_evict(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno)
 	{
 		bool		has_remaining_pages;
 
-		for (int i = 0; i < CHUNK_BITMAP_SIZE; i++)
+		for (int i = 0; i < (BLOCKS_PER_CHUNK / 32); i++)
 		{
 			if (entry->bitmap[i] != 0)
 			{
@@ -534,8 +485,8 @@ lfc_evict(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno)
 		 */
 		if (!has_remaining_pages)
 		{
-			dlist_delete(&entry->list_node);
-			dlist_push_head(&lfc_ctl->lru, &entry->list_node);
+			dlist_delete(&entry->lru_node);
+			dlist_push_head(&lfc_ctl->lru, &entry->lru_node);
 		}
 	}
 
@@ -574,8 +525,6 @@ lfc_read(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 	CopyNRelFileInfoToBufTag(tag, rinfo);
 	tag.forkNum = forkNum;
 	tag.blockNum = blkno & ~(BLOCKS_PER_CHUNK - 1);
-
-	CriticalAssert(BufTagGetRelNumber(&tag) != InvalidRelFileNumber);
 	hash = get_hash_value(lfc_hash, &tag);
 
 	LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
@@ -602,7 +551,7 @@ lfc_read(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 	}
 	/* Unlink entry from LRU list to pin it for the duration of IO operation */
 	if (entry->access_count++ == 0)
-		dlist_delete(&entry->list_node);
+		dlist_delete(&entry->lru_node);
 	generation = lfc_ctl->generation;
 	entry_offset = entry->offset;
 
@@ -620,12 +569,12 @@ lfc_read(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
 
 	if (lfc_ctl->generation == generation)
 	{
-		CriticalAssert(LFC_ENABLED());
+		Assert(LFC_ENABLED());
 		lfc_ctl->hits += 1;
 		pgBufferUsage.file_cache.hits += 1;
-		CriticalAssert(entry->access_count > 0);
+		Assert(entry->access_count > 0);
 		if (--entry->access_count == 0)
-			dlist_push_tail(&lfc_ctl->lru, &entry->list_node);
+			dlist_push_tail(&lfc_ctl->lru, &entry->lru_node);
 	}
 	else
 		result = false;
@@ -664,8 +613,6 @@ lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno, const void
 	tag.forkNum = forkNum;
 	tag.blockNum = blkno & ~(BLOCKS_PER_CHUNK - 1);
 	CopyNRelFileInfoToBufTag(tag, rinfo);
-
-	CriticalAssert(BufTagGetRelNumber(&tag) != InvalidRelFileNumber);
 	hash = get_hash_value(lfc_hash, &tag);
 
 	LWLockAcquire(lfc_lock, LW_EXCLUSIVE);
@@ -685,7 +632,7 @@ lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno, const void
 		 * operation
 		 */
 		if (entry->access_count++ == 0)
-			dlist_delete(&entry->list_node);
+			dlist_delete(&entry->lru_node);
 	}
 	else
 	{
@@ -708,26 +655,13 @@ lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno, const void
 		if (lfc_ctl->used >= lfc_ctl->limit && !dlist_is_empty(&lfc_ctl->lru))
 		{
 			/* Cache overflow: evict least recently used chunk */
-			FileCacheEntry *victim = dlist_container(FileCacheEntry, list_node, dlist_pop_head_node(&lfc_ctl->lru));
+			FileCacheEntry *victim = dlist_container(FileCacheEntry, lru_node, dlist_pop_head_node(&lfc_ctl->lru));
 
-			CriticalAssert(victim->access_count == 0);
+			Assert(victim->access_count == 0);
 			entry->offset = victim->offset; /* grab victim's chunk */
 			hash_search_with_hash_value(lfc_hash, &victim->key, victim->hash, HASH_REMOVE, NULL);
 			neon_log(DEBUG2, "Swap file cache page");
 		}
-		else if (!dlist_is_empty(&lfc_ctl->holes))
-		{
-			/* We can reuse a hole that was left behind when the LFC was shrunk previously */
-			FileCacheEntry *hole = dlist_container(FileCacheEntry, list_node, dlist_pop_head_node(&lfc_ctl->holes));
-			uint32		offset = hole->offset;
-			bool		found;
-
-			hash_search_with_hash_value(lfc_hash, &hole->key, hole->hash, HASH_REMOVE, &found);
-			CriticalAssert(found);
-
-			lfc_ctl->used += 1;
-			entry->offset = offset;	/* reuse the hole */
-		}
 		else
 		{
 			lfc_ctl->used += 1;
@@ -755,11 +689,11 @@ lfc_write(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno, const void
 
 		if (lfc_ctl->generation == generation)
 		{
-			CriticalAssert(LFC_ENABLED());
+			Assert(LFC_ENABLED());
 			/* Place entry to the head of LRU list */
-			CriticalAssert(entry->access_count > 0);
+			Assert(entry->access_count > 0);
 			if (--entry->access_count == 0)
-				dlist_push_tail(&lfc_ctl->lru, &entry->list_node);
+				dlist_push_tail(&lfc_ctl->lru, &entry->lru_node);
 
 			entry->bitmap[chunk_offs >> 5] |= (1 << (chunk_offs & 31));
 		}
@@ -774,6 +708,7 @@ typedef struct
 } NeonGetStatsCtx;
 
 #define NUM_NEON_GET_STATS_COLS	2
+#define NUM_NEON_GET_STATS_ROWS	3
 
 PG_FUNCTION_INFO_V1(neon_get_lfc_stats);
 Datum
@@ -809,6 +744,7 @@ neon_get_lfc_stats(PG_FUNCTION_ARGS)
 						   INT8OID, -1, 0);
 
 		fctx->tupdesc = BlessTupleDesc(tupledesc);
+		funcctx->max_calls = NUM_NEON_GET_STATS_ROWS;
 		funcctx->user_fctx = fctx;
 
 		/* Return to original context when allocating transient memory */
@@ -842,11 +778,6 @@ neon_get_lfc_stats(PG_FUNCTION_ARGS)
 			if (lfc_ctl)
 				value = lfc_ctl->writes;
 			break;
-		case 4:
-			key = "file_cache_size";
-			if (lfc_ctl)
-				value = lfc_ctl->size;
-			break;
 		default:
 			SRF_RETURN_DONE(funcctx);
 	}
@@ -970,7 +901,7 @@ local_cache_pages(PG_FUNCTION_ARGS)
 				hash_seq_init(&status, lfc_hash);
 				while ((entry = hash_seq_search(&status)) != NULL)
 				{
-					for (int i = 0; i < CHUNK_BITMAP_SIZE; i++)
+					for (int i = 0; i < BLOCKS_PER_CHUNK / 32; i++)
 						n_pages += pg_popcount32(entry->bitmap[i]);
 				}
 			}

pgxn/neon/neon_pgversioncompat.h

@@ -54,10 +54,6 @@
 
 #define BufTagGetNRelFileInfo(tag) tag.rnode
 
-#define BufTagGetRelNumber(tagp) ((tagp)->rnode.relNode)
-
-#define InvalidRelFileNumber InvalidOid
-
 #define SMgrRelGetRelInfo(reln) \
 	(reln->smgr_rnode.node)
 

proxy/src/auth.rs

@@ -113,36 +113,38 @@ impl<E: Into<AuthErrorImpl>> From<E> for AuthError {
 
 impl UserFacingError for AuthError {
     fn to_string_client(&self) -> String {
+        use AuthErrorImpl::*;
         match self.0.as_ref() {
-            AuthErrorImpl::Link(e) => e.to_string_client(),
-            AuthErrorImpl::GetAuthInfo(e) => e.to_string_client(),
-            AuthErrorImpl::Sasl(e) => e.to_string_client(),
-            AuthErrorImpl::AuthFailed(_) => self.to_string(),
-            AuthErrorImpl::BadAuthMethod(_) => self.to_string(),
-            AuthErrorImpl::MalformedPassword(_) => self.to_string(),
-            AuthErrorImpl::MissingEndpointName => self.to_string(),
-            AuthErrorImpl::Io(_) => "Internal error".to_string(),
-            AuthErrorImpl::IpAddressNotAllowed(_) => self.to_string(),
-            AuthErrorImpl::TooManyConnections => self.to_string(),
-            AuthErrorImpl::UserTimeout(_) => self.to_string(),
+            Link(e) => e.to_string_client(),
+            GetAuthInfo(e) => e.to_string_client(),
+            Sasl(e) => e.to_string_client(),
+            AuthFailed(_) => self.to_string(),
+            BadAuthMethod(_) => self.to_string(),
+            MalformedPassword(_) => self.to_string(),
+            MissingEndpointName => self.to_string(),
+            Io(_) => "Internal error".to_string(),
+            IpAddressNotAllowed(_) => self.to_string(),
+            TooManyConnections => self.to_string(),
+            UserTimeout(_) => self.to_string(),
         }
     }
 }
 
 impl ReportableError for AuthError {
     fn get_error_kind(&self) -> crate::error::ErrorKind {
+        use AuthErrorImpl::*;
         match self.0.as_ref() {
-            AuthErrorImpl::Link(e) => e.get_error_kind(),
-            AuthErrorImpl::GetAuthInfo(e) => e.get_error_kind(),
-            AuthErrorImpl::Sasl(e) => e.get_error_kind(),
-            AuthErrorImpl::AuthFailed(_) => crate::error::ErrorKind::User,
-            AuthErrorImpl::BadAuthMethod(_) => crate::error::ErrorKind::User,
-            AuthErrorImpl::MalformedPassword(_) => crate::error::ErrorKind::User,
-            AuthErrorImpl::MissingEndpointName => crate::error::ErrorKind::User,
-            AuthErrorImpl::Io(_) => crate::error::ErrorKind::ClientDisconnect,
-            AuthErrorImpl::IpAddressNotAllowed(_) => crate::error::ErrorKind::User,
-            AuthErrorImpl::TooManyConnections => crate::error::ErrorKind::RateLimit,
-            AuthErrorImpl::UserTimeout(_) => crate::error::ErrorKind::User,
+            Link(e) => e.get_error_kind(),
+            GetAuthInfo(e) => e.get_error_kind(),
+            Sasl(e) => e.get_error_kind(),
+            AuthFailed(_) => crate::error::ErrorKind::User,
+            BadAuthMethod(_) => crate::error::ErrorKind::User,
+            MalformedPassword(_) => crate::error::ErrorKind::User,
+            MissingEndpointName => crate::error::ErrorKind::User,
+            Io(_) => crate::error::ErrorKind::ClientDisconnect,
+            IpAddressNotAllowed(_) => crate::error::ErrorKind::User,
+            TooManyConnections => crate::error::ErrorKind::RateLimit,
+            UserTimeout(_) => crate::error::ErrorKind::User,
         }
     }
 }

proxy/src/auth/backend.rs

@@ -80,8 +80,9 @@ pub trait TestBackend: Send + Sync + 'static {
 
 impl std::fmt::Display for BackendType<'_, (), ()> {
     fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        use BackendType::*;
         match self {
-            Self::Console(api, _) => match &**api {
+            Console(api, _) => match &**api {
                 ConsoleBackend::Console(endpoint) => {
                     fmt.debug_tuple("Console").field(&endpoint.url()).finish()
                 }
@@ -92,7 +93,7 @@ impl std::fmt::Display for BackendType<'_, (), ()> {
                 #[cfg(test)]
                 ConsoleBackend::Test(_) => fmt.debug_tuple("Test").finish(),
             },
-            Self::Link(url, _) => fmt.debug_tuple("Link").field(&url.as_str()).finish(),
+            Link(url, _) => fmt.debug_tuple("Link").field(&url.as_str()).finish(),
         }
     }
 }
@@ -101,9 +102,10 @@ impl<T, D> BackendType<'_, T, D> {
     /// Very similar to [`std::option::Option::as_ref`].
     /// This helps us pass structured config to async tasks.
     pub fn as_ref(&self) -> BackendType<'_, &T, &D> {
+        use BackendType::*;
         match self {
-            Self::Console(c, x) => BackendType::Console(MaybeOwned::Borrowed(c), x),
-            Self::Link(c, x) => BackendType::Link(MaybeOwned::Borrowed(c), x),
+            Console(c, x) => Console(MaybeOwned::Borrowed(c), x),
+            Link(c, x) => Link(MaybeOwned::Borrowed(c), x),
         }
     }
 }
@@ -113,9 +115,10 @@ impl<'a, T, D> BackendType<'a, T, D> {
     /// Maps [`BackendType<T>`] to [`BackendType<R>`] by applying
     /// a function to a contained value.
     pub fn map<R>(self, f: impl FnOnce(T) -> R) -> BackendType<'a, R, D> {
+        use BackendType::*;
         match self {
-            Self::Console(c, x) => BackendType::Console(c, f(x)),
-            Self::Link(c, x) => BackendType::Link(c, x),
+            Console(c, x) => Console(c, f(x)),
+            Link(c, x) => Link(c, x),
         }
     }
 }
@@ -123,9 +126,10 @@ impl<'a, T, D, E> BackendType<'a, Result<T, E>, D> {
     /// Very similar to [`std::option::Option::transpose`].
     /// This is most useful for error handling.
     pub fn transpose(self) -> Result<BackendType<'a, T, D>, E> {
+        use BackendType::*;
         match self {
-            Self::Console(c, x) => x.map(|x| BackendType::Console(c, x)),
-            Self::Link(c, x) => Ok(BackendType::Link(c, x)),
+            Console(c, x) => x.map(|x| Console(c, x)),
+            Link(c, x) => Ok(Link(c, x)),
         }
     }
 }
@@ -289,9 +293,7 @@ async fn auth_quirks(
             ctx.set_endpoint_id(res.info.endpoint.clone());
             let password = match res.keys {
                 ComputeCredentialKeys::Password(p) => p,
-                ComputeCredentialKeys::AuthKeys(_) => {
-                    unreachable!("password hack should return a password")
-                }
+                _ => unreachable!("password hack should return a password"),
             };
             (res.info, Some(password))
         }
@@ -398,17 +400,21 @@ async fn authenticate_with_secret(
 impl<'a> BackendType<'a, ComputeUserInfoMaybeEndpoint, &()> {
     /// Get compute endpoint name from the credentials.
     pub fn get_endpoint(&self) -> Option<EndpointId> {
+        use BackendType::*;
+
         match self {
-            Self::Console(_, user_info) => user_info.endpoint_id.clone(),
-            Self::Link(_, _) => Some("link".into()),
+            Console(_, user_info) => user_info.endpoint_id.clone(),
+            Link(_, _) => Some("link".into()),
         }
     }
 
     /// Get username from the credentials.
     pub fn get_user(&self) -> &str {
+        use BackendType::*;
+
         match self {
-            Self::Console(_, user_info) => &user_info.user,
-            Self::Link(_, _) => "link",
+            Console(_, user_info) => &user_info.user,
+            Link(_, _) => "link",
         }
     }
 
@@ -422,8 +428,10 @@ impl<'a> BackendType<'a, ComputeUserInfoMaybeEndpoint, &()> {
         config: &'static AuthenticationConfig,
         endpoint_rate_limiter: Arc<EndpointRateLimiter>,
     ) -> auth::Result<BackendType<'a, ComputeCredentials, NodeInfo>> {
+        use BackendType::*;
+
         let res = match self {
-            Self::Console(api, user_info) => {
+            Console(api, user_info) => {
                 info!(
                     user = &*user_info.user,
                     project = user_info.endpoint(),
@@ -443,7 +451,7 @@ impl<'a> BackendType<'a, ComputeUserInfoMaybeEndpoint, &()> {
                 BackendType::Console(api, credentials)
             }
             // NOTE: this auth backend doesn't use client credentials.
-            Self::Link(url, _) => {
+            Link(url, _) => {
                 info!("performing link authentication");
 
                 let info = link::authenticate(ctx, &url, client).await?;
@@ -462,9 +470,10 @@ impl BackendType<'_, ComputeUserInfo, &()> {
         &self,
         ctx: &RequestMonitoring,
     ) -> Result<CachedRoleSecret, GetAuthInfoError> {
+        use BackendType::*;
         match self {
-            Self::Console(api, user_info) => api.get_role_secret(ctx, user_info).await,
-            Self::Link(_, _) => Ok(Cached::new_uncached(None)),
+            Console(api, user_info) => api.get_role_secret(ctx, user_info).await,
+            Link(_, _) => Ok(Cached::new_uncached(None)),
         }
     }
 
@@ -472,9 +481,10 @@ impl BackendType<'_, ComputeUserInfo, &()> {
         &self,
         ctx: &RequestMonitoring,
     ) -> Result<(CachedAllowedIps, Option<CachedRoleSecret>), GetAuthInfoError> {
+        use BackendType::*;
         match self {
-            Self::Console(api, user_info) => api.get_allowed_ips_and_secret(ctx, user_info).await,
-            Self::Link(_, _) => Ok((Cached::new_uncached(Arc::new(vec![])), None)),
+            Console(api, user_info) => api.get_allowed_ips_and_secret(ctx, user_info).await,
+            Link(_, _) => Ok((Cached::new_uncached(Arc::new(vec![])), None)),
         }
     }
 }
@@ -485,16 +495,18 @@ impl ComputeConnectBackend for BackendType<'_, ComputeCredentials, NodeInfo> {
         &self,
         ctx: &RequestMonitoring,
     ) -> Result<CachedNodeInfo, console::errors::WakeComputeError> {
+        use BackendType::*;
+
         match self {
-            Self::Console(api, creds) => api.wake_compute(ctx, &creds.info).await,
-            Self::Link(_, info) => Ok(Cached::new_uncached(info.clone())),
+            Console(api, creds) => api.wake_compute(ctx, &creds.info).await,
+            Link(_, info) => Ok(Cached::new_uncached(info.clone())),
         }
     }
 
     fn get_keys(&self) -> Option<&ComputeCredentialKeys> {
         match self {
-            Self::Console(_, creds) => Some(&creds.keys),
-            Self::Link(_, _) => None,
+            BackendType::Console(_, creds) => Some(&creds.keys),
+            BackendType::Link(_, _) => None,
         }
     }
 }
@@ -505,16 +517,18 @@ impl ComputeConnectBackend for BackendType<'_, ComputeCredentials, &()> {
         &self,
         ctx: &RequestMonitoring,
     ) -> Result<CachedNodeInfo, console::errors::WakeComputeError> {
+        use BackendType::*;
+
         match self {
-            Self::Console(api, creds) => api.wake_compute(ctx, &creds.info).await,
-            Self::Link(_, _) => unreachable!("link auth flow doesn't support waking the compute"),
+            Console(api, creds) => api.wake_compute(ctx, &creds.info).await,
+            Link(_, _) => unreachable!("link auth flow doesn't support waking the compute"),
         }
     }
 
     fn get_keys(&self) -> Option<&ComputeCredentialKeys> {
         match self {
-            Self::Console(_, creds) => Some(&creds.keys),
-            Self::Link(_, _) => None,
+            BackendType::Console(_, creds) => Some(&creds.keys),
+            BackendType::Link(_, _) => None,
         }
     }
 }

proxy/src/auth/backend/jwt.rs

@@ -195,7 +195,7 @@ impl JwkCacheEntryLock {
 
         let header = base64::decode_config(header, base64::URL_SAFE_NO_PAD)
             .context("Provided authentication token is not a valid JWT encoding")?;
-        let header = serde_json::from_slice::<JWTHeader<'_>>(&header)
+        let header = serde_json::from_slice::<JWTHeader>(&header)
             .context("Provided authentication token is not a valid JWT encoding")?;
 
         let sig = base64::decode_config(signature, base64::URL_SAFE_NO_PAD)
@@ -340,7 +340,7 @@ impl JwkRenewalPermit<'_> {
         }
     }
 
-    async fn acquire_permit(from: &Arc<JwkCacheEntryLock>) -> JwkRenewalPermit<'_> {
+    async fn acquire_permit(from: &Arc<JwkCacheEntryLock>) -> JwkRenewalPermit {
         match from.lookup.acquire().await {
             Ok(permit) => {
                 permit.forget();
@@ -352,7 +352,7 @@ impl JwkRenewalPermit<'_> {
         }
     }
 
-    fn try_acquire_permit(from: &Arc<JwkCacheEntryLock>) -> Option<JwkRenewalPermit<'_>> {
+    fn try_acquire_permit(from: &Arc<JwkCacheEntryLock>) -> Option<JwkRenewalPermit> {
         match from.lookup.try_acquire() {
             Ok(permit) => {
                 permit.forget();

proxy/src/auth/credentials.rs

@@ -89,12 +89,10 @@ impl ComputeUserInfoMaybeEndpoint {
         sni: Option<&str>,
         common_names: Option<&HashSet<String>>,
     ) -> Result<Self, ComputeUserInfoParseError> {
+        use ComputeUserInfoParseError::*;
+
         // Some parameters are stored in the startup message.
-        let get_param = |key| {
-            params
-                .get(key)
-                .ok_or(ComputeUserInfoParseError::MissingKey(key))
-        };
+        let get_param = |key| params.get(key).ok_or(MissingKey(key));
         let user: RoleName = get_param("user")?.into();
 
         // Project name might be passed via PG's command-line options.
@@ -124,14 +122,11 @@ impl ComputeUserInfoMaybeEndpoint {
         let endpoint = match (endpoint_option, endpoint_from_domain) {
             // Invariant: if we have both project name variants, they should match.
             (Some(option), Some(domain)) if option != domain => {
-                Some(Err(ComputeUserInfoParseError::InconsistentProjectNames {
-                    domain,
-                    option,
-                }))
+                Some(Err(InconsistentProjectNames { domain, option }))
             }
             // Invariant: project name may not contain certain characters.
             (a, b) => a.or(b).map(|name| match project_name_valid(name.as_ref()) {
-                false => Err(ComputeUserInfoParseError::MalformedProjectName(name)),
+                false => Err(MalformedProjectName(name)),
                 true => Ok(name),
             }),
         }
@@ -191,7 +186,7 @@ impl<'de> serde::de::Deserialize<'de> for IpPattern {
         impl<'de> serde::de::Visitor<'de> for StrVisitor {
             type Value = IpPattern;
 
-            fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+            fn expecting(&self, formatter: &mut std::fmt::Formatter) -> std::fmt::Result {
                 write!(formatter, "comma separated list with ip address, ip address range, or ip address subnet mask")
             }
 

proxy/src/bin/proxy.rs

@@ -173,6 +173,9 @@ struct ProxyCliArgs {
     /// cache for `role_secret` (use `size=0` to disable)
     #[clap(long, default_value = config::CacheOptions::CACHE_DEFAULT_OPTIONS)]
     role_secret_cache: String,
+    /// disable ip check for http requests. If it is too time consuming, it could be turned off.
+    #[clap(long, default_value_t = false, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
+    disable_ip_check_for_http: bool,
     /// redis url for notifications (if empty, redis_host:port will be used for both notifications and streaming connections)
     #[clap(long)]
     redis_notifications: Option<String>,
@@ -658,7 +661,6 @@ fn build_config(args: &ProxyCliArgs) -> anyhow::Result<&'static ProxyConfig> {
     )?;
 
     let http_config = HttpConfig {
-        accept_websockets: true,
         pool_options: GlobalConnPoolOptions {
             max_conns_per_endpoint: args.sql_over_http.sql_over_http_pool_max_conns_per_endpoint,
             gc_epoch: args.sql_over_http.sql_over_http_pool_gc_epoch,

proxy/src/cache/common.rs

@@ -24,7 +24,7 @@ impl<C: Cache> Cache for &C {
     type LookupInfo<Key> = C::LookupInfo<Key>;
 
     fn invalidate(&self, info: &Self::LookupInfo<Self::Key>) {
-        C::invalidate(self, info);
+        C::invalidate(self, info)
     }
 }
 

proxy/src/cache/timed_lru.rs

@@ -58,7 +58,7 @@ impl<K: Hash + Eq, V> Cache for TimedLru<K, V> {
     type LookupInfo<Key> = LookupInfo<Key>;
 
     fn invalidate(&self, info: &Self::LookupInfo<K>) {
-        self.invalidate_raw(info);
+        self.invalidate_raw(info)
     }
 }
 

proxy/src/cancellation.rs

@@ -151,21 +151,34 @@ impl<P: CancellationPublisherMut> CancellationHandler<Option<Arc<Mutex<P>>>> {
 #[derive(Clone)]
 pub struct CancelClosure {
     socket_addr: SocketAddr,
-    cancel_token: CancelToken,
+    cancel_token: Option<CancelToken>,
 }
 
 impl CancelClosure {
     pub fn new(socket_addr: SocketAddr, cancel_token: CancelToken) -> Self {
         Self {
             socket_addr,
-            cancel_token,
+            cancel_token: Some(cancel_token),
         }
     }
+
+    #[cfg(test)]
+    pub fn test() -> Self {
+        use std::net::{Ipv4Addr, SocketAddrV4};
+
+        Self {
+            socket_addr: SocketAddr::V4(SocketAddrV4::new(Ipv4Addr::from_bits(0), 0)),
+            cancel_token: None,
+        }
+    }
+
     /// Cancels the query running on user's compute node.
     pub async fn try_cancel_query(self) -> Result<(), CancelError> {
-        let socket = TcpStream::connect(self.socket_addr).await?;
-        self.cancel_token.cancel_query_raw(socket, NoTls).await?;
-        info!("query was cancelled");
+        if let Some(cancel_token) = self.cancel_token {
+            let socket = TcpStream::connect(self.socket_addr).await?;
+            cancel_token.cancel_query_raw(socket, NoTls).await?;
+            info!("query was cancelled");
+        }
         Ok(())
     }
 }

proxy/src/compute.rs

@@ -16,8 +16,10 @@ use rustls::{client::danger::ServerCertVerifier, pki_types::InvalidDnsNameError}
 use std::{io, net::SocketAddr, sync::Arc, time::Duration};
 use thiserror::Error;
 use tokio::net::TcpStream;
-use tokio_postgres::tls::MakeTlsConnect;
-use tokio_postgres_rustls::MakeRustlsConnect;
+use tokio_postgres::{
+    tls::{MakeTlsConnect, NoTlsError},
+    Client, Connection,
+};
 use tracing::{error, info, warn};
 
 const COULD_NOT_CONNECT: &str = "Couldn't connect to compute node";
@@ -42,12 +44,19 @@ pub enum ConnectionError {
     TooManyConnectionAttempts(#[from] ApiLockError),
 }
 
+impl From<NoTlsError> for ConnectionError {
+    fn from(value: NoTlsError) -> Self {
+        Self::CouldNotConnect(io::Error::new(io::ErrorKind::Other, value.to_string()))
+    }
+}
+
 impl UserFacingError for ConnectionError {
     fn to_string_client(&self) -> String {
+        use ConnectionError::*;
         match self {
             // This helps us drop irrelevant library-specific prefixes.
             // TODO: propagate severity level and other parameters.
-            ConnectionError::Postgres(err) => match err.as_db_error() {
+            Postgres(err) => match err.as_db_error() {
                 Some(err) => {
                     let msg = err.message();
 
@@ -61,8 +70,8 @@ impl UserFacingError for ConnectionError {
                 }
                 None => err.to_string(),
             },
-            ConnectionError::WakeComputeError(err) => err.to_string_client(),
-            ConnectionError::TooManyConnectionAttempts(_) => {
+            WakeComputeError(err) => err.to_string_client(),
+            TooManyConnectionAttempts(_) => {
                 "Failed to acquire permit to connect to the database. Too many database connection attempts are currently ongoing.".to_owned()
             }
             _ => COULD_NOT_CONNECT.to_owned(),
@@ -272,6 +281,30 @@ pub struct PostgresConnection {
 }
 
 impl ConnCfg {
+    /// Connect to a corresponding compute node.
+    pub async fn managed_connect<M: MakeTlsConnect<tokio::net::TcpStream>>(
+        &self,
+        ctx: &RequestMonitoring,
+        timeout: Duration,
+        mktls: &mut M,
+    ) -> Result<(SocketAddr, Client, Connection<TcpStream, M::Stream>), ConnectionError>
+    where
+        ConnectionError: From<M::Error>,
+    {
+        let pause = ctx.latency_timer_pause(crate::metrics::Waiting::Compute);
+        let (socket_addr, stream, host) = self.connect_raw(timeout).await?;
+        drop(pause);
+
+        let tls = mktls.make_tls_connect(host)?;
+
+        // connect_raw() will not use TLS if sslmode is "disable"
+        let pause = ctx.latency_timer_pause(crate::metrics::Waiting::Compute);
+        let (client, connection) = self.0.connect_raw(stream, tls).await?;
+        drop(pause);
+
+        Ok((socket_addr, client, connection))
+    }
+
     /// Connect to a corresponding compute node.
     pub async fn connect(
         &self,
@@ -280,10 +313,6 @@ impl ConnCfg {
         aux: MetricsAuxInfo,
         timeout: Duration,
     ) -> Result<PostgresConnection, ConnectionError> {
-        let pause = ctx.latency_timer_pause(crate::metrics::Waiting::Compute);
-        let (socket_addr, stream, host) = self.connect_raw(timeout).await?;
-        drop(pause);
-
         let client_config = if allow_self_signed_compute {
             // Allow all certificates for creating the connection
             let verifier = Arc::new(AcceptEverythingVerifier) as Arc<dyn ServerCertVerifier>;
@@ -297,21 +326,15 @@ impl ConnCfg {
         let client_config = client_config.with_no_client_auth();
 
         let mut mk_tls = tokio_postgres_rustls::MakeRustlsConnect::new(client_config);
-        let tls = <MakeRustlsConnect as MakeTlsConnect<tokio::net::TcpStream>>::make_tls_connect(
-            &mut mk_tls,
-            host,
-        )?;
 
-        // connect_raw() will not use TLS if sslmode is "disable"
-        let pause = ctx.latency_timer_pause(crate::metrics::Waiting::Compute);
-        let (client, connection) = self.0.connect_raw(stream, tls).await?;
-        drop(pause);
+        let (socket_addr, client, connection) =
+            self.managed_connect(ctx, timeout, &mut mk_tls).await?;
         tracing::Span::current().record("pid", tracing::field::display(client.get_process_id()));
         let stream = connection.stream.into_inner();
 
         info!(
             cold_start_info = ctx.cold_start_info().as_str(),
-            "connected to compute node at {host} ({socket_addr}) sslmode={:?}",
+            "connected to compute node ({socket_addr}) sslmode={:?}",
             self.0.get_ssl_mode()
         );
 
@@ -365,16 +388,16 @@ static TLS_ROOTS: OnceCell<Arc<rustls::RootCertStore>> = OnceCell::new();
 struct AcceptEverythingVerifier;
 impl ServerCertVerifier for AcceptEverythingVerifier {
     fn supported_verify_schemes(&self) -> Vec<rustls::SignatureScheme> {
-        use rustls::SignatureScheme;
+        use rustls::SignatureScheme::*;
         // The schemes for which `SignatureScheme::supported_in_tls13` returns true.
         vec![
-            SignatureScheme::ECDSA_NISTP521_SHA512,
-            SignatureScheme::ECDSA_NISTP384_SHA384,
-            SignatureScheme::ECDSA_NISTP256_SHA256,
-            SignatureScheme::RSA_PSS_SHA512,
-            SignatureScheme::RSA_PSS_SHA384,
-            SignatureScheme::RSA_PSS_SHA256,
-            SignatureScheme::ED25519,
+            ECDSA_NISTP521_SHA512,
+            ECDSA_NISTP384_SHA384,
+            ECDSA_NISTP256_SHA256,
+            RSA_PSS_SHA512,
+            RSA_PSS_SHA384,
+            RSA_PSS_SHA256,
+            ED25519,
         ]
     }
     fn verify_server_cert(

proxy/src/config.rs

@@ -52,7 +52,6 @@ pub struct TlsConfig {
 }
 
 pub struct HttpConfig {
-    pub accept_websockets: bool,
     pub pool_options: GlobalConnPoolOptions,
     pub cancel_set: CancelSet,
     pub client_conn_threshold: u64,
@@ -156,7 +155,7 @@ pub enum TlsServerEndPoint {
 }
 
 impl TlsServerEndPoint {
-    pub fn new(cert: &CertificateDer<'_>) -> anyhow::Result<Self> {
+    pub fn new(cert: &CertificateDer) -> anyhow::Result<Self> {
         let sha256_oids = [
             // I'm explicitly not adding MD5 or SHA1 here... They're bad.
             oid_registry::OID_SIG_ECDSA_WITH_SHA256,
@@ -279,7 +278,7 @@ impl CertResolver {
 impl rustls::server::ResolvesServerCert for CertResolver {
     fn resolve(
         &self,
-        client_hello: rustls::server::ClientHello<'_>,
+        client_hello: rustls::server::ClientHello,
     ) -> Option<Arc<rustls::sign::CertifiedKey>> {
         self.resolve(client_hello.server_name()).map(|x| x.0)
     }
@@ -560,7 +559,7 @@ impl RetryConfig {
             match key {
                 "num_retries" => num_retries = Some(value.parse()?),
                 "base_retry_wait_duration" => {
-                    base_retry_wait_duration = Some(humantime::parse_duration(value)?);
+                    base_retry_wait_duration = Some(humantime::parse_duration(value)?)
                 }
                 "retry_wait_exponent_base" => retry_wait_exponent_base = Some(value.parse()?),
                 unknown => bail!("unknown key: {unknown}"),

proxy/src/console/messages.rs

@@ -22,15 +22,16 @@ impl ConsoleError {
         self.status
             .as_ref()
             .and_then(|s| s.details.error_info.as_ref())
-            .map_or(Reason::Unknown, |e| e.reason)
+            .map(|e| e.reason)
+            .unwrap_or(Reason::Unknown)
     }
-
     pub fn get_user_facing_message(&self) -> String {
         use super::provider::errors::REQUEST_FAILED;
         self.status
             .as_ref()
             .and_then(|s| s.details.user_facing_message.as_ref())
-            .map_or_else(|| {
+            .map(|m| m.message.clone().into())
+            .unwrap_or_else(|| {
                 // Ask @neondatabase/control-plane for review before adding more.
                 match self.http_status_code {
                     http::StatusCode::NOT_FOUND => {
@@ -47,18 +48,19 @@ impl ConsoleError {
                     }
                     _ => REQUEST_FAILED.to_owned(),
                 }
-            }, |m| m.message.clone().into())
+            })
     }
 }
 
 impl Display for ConsoleError {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-        let msg: &str = self
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        let msg = self
             .status
             .as_ref()
             .and_then(|s| s.details.user_facing_message.as_ref())
-            .map_or_else(|| self.error.as_ref(), |m| m.message.as_ref());
-        write!(f, "{msg}")
+            .map(|m| m.message.as_ref())
+            .unwrap_or_else(|| &self.error);
+        write!(f, "{}", msg)
     }
 }
 
@@ -284,7 +286,7 @@ pub struct DatabaseInfo {
 
 // Manually implement debug to omit sensitive info.
 impl fmt::Debug for DatabaseInfo {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
         f.debug_struct("DatabaseInfo")
             .field("host", &self.host)
             .field("port", &self.port)
@@ -371,7 +373,7 @@ mod tests {
                 }
             }
         });
-        let _: KickSession<'_> = serde_json::from_str(&json.to_string())?;
+        let _: KickSession = serde_json::from_str(&json.to_string())?;
 
         Ok(())
     }

proxy/src/console/mgmt.rs

@@ -93,8 +93,7 @@ impl postgres_backend::Handler<tokio::net::TcpStream> for MgmtHandler {
 }
 
 fn try_process_query(pgb: &mut PostgresBackendTCP, query: &str) -> Result<(), QueryError> {
-    let resp: KickSession<'_> =
-        serde_json::from_str(query).context("Failed to parse query as json")?;
+    let resp: KickSession = serde_json::from_str(query).context("Failed to parse query as json")?;
 
     let span = info_span!("event", session_id = resp.session_id);
     let _enter = span.enter();

proxy/src/console/provider.rs

@@ -26,7 +26,7 @@ use tracing::info;
 pub mod errors {
     use crate::{
         console::messages::{self, ConsoleError, Reason},
-        error::{io_error, ErrorKind, ReportableError, UserFacingError},
+        error::{io_error, ReportableError, UserFacingError},
         proxy::retry::CouldRetry,
     };
     use thiserror::Error;
@@ -51,19 +51,21 @@ pub mod errors {
     impl ApiError {
         /// Returns HTTP status code if it's the reason for failure.
         pub fn get_reason(&self) -> messages::Reason {
+            use ApiError::*;
             match self {
-                ApiError::Console(e) => e.get_reason(),
-                ApiError::Transport(_) => messages::Reason::Unknown,
+                Console(e) => e.get_reason(),
+                _ => messages::Reason::Unknown,
             }
         }
     }
 
     impl UserFacingError for ApiError {
         fn to_string_client(&self) -> String {
+            use ApiError::*;
             match self {
                 // To minimize risks, only select errors are forwarded to users.
-                ApiError::Console(c) => c.get_user_facing_message(),
-                ApiError::Transport(_) => REQUEST_FAILED.to_owned(),
+                Console(c) => c.get_user_facing_message(),
+                _ => REQUEST_FAILED.to_owned(),
             }
         }
     }
@@ -71,53 +73,57 @@ pub mod errors {
     impl ReportableError for ApiError {
         fn get_error_kind(&self) -> crate::error::ErrorKind {
             match self {
-                ApiError::Console(e) => match e.get_reason() {
-                    Reason::RoleProtected => ErrorKind::User,
-                    Reason::ResourceNotFound => ErrorKind::User,
-                    Reason::ProjectNotFound => ErrorKind::User,
-                    Reason::EndpointNotFound => ErrorKind::User,
-                    Reason::BranchNotFound => ErrorKind::User,
-                    Reason::RateLimitExceeded => ErrorKind::ServiceRateLimit,
-                    Reason::NonDefaultBranchComputeTimeExceeded => ErrorKind::User,
-                    Reason::ActiveTimeQuotaExceeded => ErrorKind::User,
-                    Reason::ComputeTimeQuotaExceeded => ErrorKind::User,
-                    Reason::WrittenDataQuotaExceeded => ErrorKind::User,
-                    Reason::DataTransferQuotaExceeded => ErrorKind::User,
-                    Reason::LogicalSizeQuotaExceeded => ErrorKind::User,
-                    Reason::ConcurrencyLimitReached => ErrorKind::ControlPlane,
-                    Reason::LockAlreadyTaken => ErrorKind::ControlPlane,
-                    Reason::RunningOperations => ErrorKind::ControlPlane,
-                    Reason::Unknown => match &e {
-                        ConsoleError {
-                            http_status_code:
-                                http::StatusCode::NOT_FOUND | http::StatusCode::NOT_ACCEPTABLE,
-                            ..
-                        } => crate::error::ErrorKind::User,
-                        ConsoleError {
-                            http_status_code: http::StatusCode::UNPROCESSABLE_ENTITY,
-                            error,
-                            ..
-                        } if error
-                            .contains("compute time quota of non-primary branches is exceeded") =>
-                        {
-                            crate::error::ErrorKind::User
-                        }
-                        ConsoleError {
-                            http_status_code: http::StatusCode::LOCKED,
-                            error,
-                            ..
-                        } if error.contains("quota exceeded")
-                            || error.contains("the limit for current plan reached") =>
-                        {
-                            crate::error::ErrorKind::User
-                        }
-                        ConsoleError {
-                            http_status_code: http::StatusCode::TOO_MANY_REQUESTS,
-                            ..
-                        } => crate::error::ErrorKind::ServiceRateLimit,
-                        ConsoleError { .. } => crate::error::ErrorKind::ControlPlane,
-                    },
-                },
+                ApiError::Console(e) => {
+                    use crate::error::ErrorKind::*;
+                    match e.get_reason() {
+                        Reason::RoleProtected => User,
+                        Reason::ResourceNotFound => User,
+                        Reason::ProjectNotFound => User,
+                        Reason::EndpointNotFound => User,
+                        Reason::BranchNotFound => User,
+                        Reason::RateLimitExceeded => ServiceRateLimit,
+                        Reason::NonDefaultBranchComputeTimeExceeded => User,
+                        Reason::ActiveTimeQuotaExceeded => User,
+                        Reason::ComputeTimeQuotaExceeded => User,
+                        Reason::WrittenDataQuotaExceeded => User,
+                        Reason::DataTransferQuotaExceeded => User,
+                        Reason::LogicalSizeQuotaExceeded => User,
+                        Reason::ConcurrencyLimitReached => ControlPlane,
+                        Reason::LockAlreadyTaken => ControlPlane,
+                        Reason::RunningOperations => ControlPlane,
+                        Reason::Unknown => match &e {
+                            ConsoleError {
+                                http_status_code:
+                                    http::StatusCode::NOT_FOUND | http::StatusCode::NOT_ACCEPTABLE,
+                                ..
+                            } => crate::error::ErrorKind::User,
+                            ConsoleError {
+                                http_status_code: http::StatusCode::UNPROCESSABLE_ENTITY,
+                                error,
+                                ..
+                            } if error.contains(
+                                "compute time quota of non-primary branches is exceeded",
+                            ) =>
+                            {
+                                crate::error::ErrorKind::User
+                            }
+                            ConsoleError {
+                                http_status_code: http::StatusCode::LOCKED,
+                                error,
+                                ..
+                            } if error.contains("quota exceeded")
+                                || error.contains("the limit for current plan reached") =>
+                            {
+                                crate::error::ErrorKind::User
+                            }
+                            ConsoleError {
+                                http_status_code: http::StatusCode::TOO_MANY_REQUESTS,
+                                ..
+                            } => crate::error::ErrorKind::ServiceRateLimit,
+                            ConsoleError { .. } => crate::error::ErrorKind::ControlPlane,
+                        },
+                    }
+                }
                 ApiError::Transport(_) => crate::error::ErrorKind::ControlPlane,
             }
         }
@@ -164,11 +170,12 @@ pub mod errors {
 
     impl UserFacingError for GetAuthInfoError {
         fn to_string_client(&self) -> String {
+            use GetAuthInfoError::*;
             match self {
                 // We absolutely should not leak any secrets!
-                Self::BadSecret => REQUEST_FAILED.to_owned(),
+                BadSecret => REQUEST_FAILED.to_owned(),
                 // However, API might return a meaningful error.
-                Self::ApiError(e) => e.to_string_client(),
+                ApiError(e) => e.to_string_client(),
             }
         }
     }
@@ -176,8 +183,8 @@ pub mod errors {
     impl ReportableError for GetAuthInfoError {
         fn get_error_kind(&self) -> crate::error::ErrorKind {
             match self {
-                Self::BadSecret => crate::error::ErrorKind::ControlPlane,
-                Self::ApiError(_) => crate::error::ErrorKind::ControlPlane,
+                GetAuthInfoError::BadSecret => crate::error::ErrorKind::ControlPlane,
+                GetAuthInfoError::ApiError(_) => crate::error::ErrorKind::ControlPlane,
             }
         }
     }
@@ -206,16 +213,17 @@ pub mod errors {
 
     impl UserFacingError for WakeComputeError {
         fn to_string_client(&self) -> String {
+            use WakeComputeError::*;
             match self {
                 // We shouldn't show user the address even if it's broken.
                 // Besides, user is unlikely to care about this detail.
-                Self::BadComputeAddress(_) => REQUEST_FAILED.to_owned(),
+                BadComputeAddress(_) => REQUEST_FAILED.to_owned(),
                 // However, API might return a meaningful error.
-                Self::ApiError(e) => e.to_string_client(),
+                ApiError(e) => e.to_string_client(),
 
-                Self::TooManyConnections => self.to_string(),
+                TooManyConnections => self.to_string(),
 
-                Self::TooManyConnectionAttempts(_) => {
+                TooManyConnectionAttempts(_) => {
                     "Failed to acquire permit to connect to the database. Too many database connection attempts are currently ongoing.".to_owned()
                 }
             }
@@ -225,10 +233,10 @@ pub mod errors {
     impl ReportableError for WakeComputeError {
         fn get_error_kind(&self) -> crate::error::ErrorKind {
             match self {
-                Self::BadComputeAddress(_) => crate::error::ErrorKind::ControlPlane,
-                Self::ApiError(e) => e.get_error_kind(),
-                Self::TooManyConnections => crate::error::ErrorKind::RateLimit,
-                Self::TooManyConnectionAttempts(e) => e.get_error_kind(),
+                WakeComputeError::BadComputeAddress(_) => crate::error::ErrorKind::ControlPlane,
+                WakeComputeError::ApiError(e) => e.get_error_kind(),
+                WakeComputeError::TooManyConnections => crate::error::ErrorKind::RateLimit,
+                WakeComputeError::TooManyConnectionAttempts(e) => e.get_error_kind(),
             }
         }
     }
@@ -236,10 +244,10 @@ pub mod errors {
     impl CouldRetry for WakeComputeError {
         fn could_retry(&self) -> bool {
             match self {
-                Self::BadComputeAddress(_) => false,
-                Self::ApiError(e) => e.could_retry(),
-                Self::TooManyConnections => false,
-                Self::TooManyConnectionAttempts(_) => false,
+                WakeComputeError::BadComputeAddress(_) => false,
+                WakeComputeError::ApiError(e) => e.could_retry(),
+                WakeComputeError::TooManyConnections => false,
+                WakeComputeError::TooManyConnectionAttempts(_) => false,
             }
         }
     }
@@ -358,14 +366,13 @@ impl Api for ConsoleBackend {
         ctx: &RequestMonitoring,
         user_info: &ComputeUserInfo,
     ) -> Result<CachedRoleSecret, errors::GetAuthInfoError> {
+        use ConsoleBackend::*;
         match self {
-            Self::Console(api) => api.get_role_secret(ctx, user_info).await,
+            Console(api) => api.get_role_secret(ctx, user_info).await,
             #[cfg(any(test, feature = "testing"))]
-            Self::Postgres(api) => api.get_role_secret(ctx, user_info).await,
+            Postgres(api) => api.get_role_secret(ctx, user_info).await,
             #[cfg(test)]
-            Self::Test(_) => {
-                unreachable!("this function should never be called in the test backend")
-            }
+            Test(_) => unreachable!("this function should never be called in the test backend"),
         }
     }
 
@@ -374,12 +381,13 @@ impl Api for ConsoleBackend {
         ctx: &RequestMonitoring,
         user_info: &ComputeUserInfo,
     ) -> Result<(CachedAllowedIps, Option<CachedRoleSecret>), errors::GetAuthInfoError> {
+        use ConsoleBackend::*;
         match self {
-            Self::Console(api) => api.get_allowed_ips_and_secret(ctx, user_info).await,
+            Console(api) => api.get_allowed_ips_and_secret(ctx, user_info).await,
             #[cfg(any(test, feature = "testing"))]
-            Self::Postgres(api) => api.get_allowed_ips_and_secret(ctx, user_info).await,
+            Postgres(api) => api.get_allowed_ips_and_secret(ctx, user_info).await,
             #[cfg(test)]
-            Self::Test(api) => api.get_allowed_ips_and_secret(),
+            Test(api) => api.get_allowed_ips_and_secret(),
         }
     }
 
@@ -388,12 +396,14 @@ impl Api for ConsoleBackend {
         ctx: &RequestMonitoring,
         user_info: &ComputeUserInfo,
     ) -> Result<CachedNodeInfo, errors::WakeComputeError> {
+        use ConsoleBackend::*;
+
         match self {
-            Self::Console(api) => api.wake_compute(ctx, user_info).await,
+            Console(api) => api.wake_compute(ctx, user_info).await,
             #[cfg(any(test, feature = "testing"))]
-            Self::Postgres(api) => api.wake_compute(ctx, user_info).await,
+            Postgres(api) => api.wake_compute(ctx, user_info).await,
             #[cfg(test)]
-            Self::Test(api) => api.wake_compute(),
+            Test(api) => api.wake_compute(),
         }
     }
 }
@@ -539,7 +549,7 @@ impl WakeComputePermit {
         !self.permit.is_disabled()
     }
     pub fn release(self, outcome: Outcome) {
-        self.permit.release(outcome);
+        self.permit.release(outcome)
     }
     pub fn release_result<T, E>(self, res: Result<T, E>) -> Result<T, E> {
         match res {

proxy/src/context.rs

@@ -166,7 +166,7 @@ impl RequestMonitoring {
     pub fn set_project(&self, x: MetricsAuxInfo) {
         let mut this = self.0.try_lock().expect("should not deadlock");
         if this.endpoint_id.is_none() {
-            this.set_endpoint_id(x.endpoint_id.as_str().into());
+            this.set_endpoint_id(x.endpoint_id.as_str().into())
         }
         this.branch = Some(x.branch_id);
         this.project = Some(x.project_id);
@@ -260,7 +260,7 @@ impl RequestMonitoring {
             .cold_start_info
     }
 
-    pub fn latency_timer_pause(&self, waiting_for: Waiting) -> LatencyTimerPause<'_> {
+    pub fn latency_timer_pause(&self, waiting_for: Waiting) -> LatencyTimerPause {
         LatencyTimerPause {
             ctx: self,
             start: tokio::time::Instant::now(),
@@ -273,7 +273,7 @@ impl RequestMonitoring {
             .try_lock()
             .expect("should not deadlock")
             .latency_timer
-            .success();
+            .success()
     }
 }
 
@@ -328,7 +328,7 @@ impl RequestMonitoringInner {
     fn has_private_peer_addr(&self) -> bool {
         match self.peer_addr {
             IpAddr::V4(ip) => ip.is_private(),
-            IpAddr::V6(_) => false,
+            _ => false,
         }
     }
 

proxy/src/context/parquet.rs

@@ -736,7 +736,7 @@ mod tests {
                 while let Some(r) = s.next().await {
                     tx.send(r).unwrap();
                 }
-                time::sleep(time::Duration::from_secs(70)).await;
+                time::sleep(time::Duration::from_secs(70)).await
             }
         });
 

proxy/src/intern.rs

@@ -56,7 +56,7 @@ impl<'de, Id: InternId> serde::de::Deserialize<'de> for InternedString<Id> {
         impl<'de, Id: InternId> serde::de::Visitor<'de> for Visitor<Id> {
             type Value = InternedString<Id>;
 
-            fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+            fn expecting(&self, formatter: &mut std::fmt::Formatter) -> std::fmt::Result {
                 formatter.write_str("a string")
             }
 

proxy/src/lib.rs

@@ -1,92 +1,4 @@
-// rustc lints/lint groups
-// https://doc.rust-lang.org/rustc/lints/groups.html
-#![deny(
-    deprecated,
-    future_incompatible,
-    // TODO: consider let_underscore
-    nonstandard_style,
-    rust_2024_compatibility
-)]
-#![warn(clippy::all, clippy::pedantic, clippy::cargo)]
-// List of denied lints from the clippy::restriction group.
-// https://rust-lang.github.io/rust-clippy/master/index.html#?groups=restriction
-#![warn(
-    clippy::undocumented_unsafe_blocks,
-    clippy::dbg_macro,
-    clippy::empty_enum_variants_with_brackets,
-    clippy::exit,
-    clippy::float_cmp_const,
-    clippy::lossy_float_literal,
-    clippy::macro_use_imports,
-    clippy::manual_ok_or,
-    // TODO: consider clippy::map_err_ignore
-    // TODO: consider clippy::mem_forget
-    clippy::rc_mutex,
-    clippy::rest_pat_in_fully_bound_structs,
-    clippy::string_add,
-    clippy::string_to_string,
-    clippy::todo,
-    // TODO: consider clippy::unimplemented
-    // TODO: consider clippy::unwrap_used
-)]
-// List of permanently allowed lints.
-#![allow(
-    // It's ok to cast u8 to bool, etc.
-    clippy::cast_lossless,
-)]
-// List of temporarily allowed lints.
-// TODO: Switch to except() once stable with 1.81.
-// TODO: fix code and reduce list or move to permanent list above.
-#![allow(
-    clippy::cargo_common_metadata,
-    clippy::cast_possible_truncation,
-    clippy::cast_possible_wrap,
-    clippy::cast_precision_loss,
-    clippy::cast_sign_loss,
-    clippy::default_trait_access,
-    clippy::doc_markdown,
-    clippy::explicit_iter_loop,
-    clippy::float_cmp,
-    clippy::if_not_else,
-    clippy::ignored_unit_patterns,
-    clippy::implicit_hasher,
-    clippy::inconsistent_struct_constructor,
-    clippy::inline_always,
-    clippy::items_after_statements,
-    clippy::manual_assert,
-    clippy::manual_let_else,
-    clippy::manual_string_new,
-    clippy::match_bool,
-    clippy::match_same_arms,
-    clippy::match_wild_err_arm,
-    clippy::missing_errors_doc,
-    clippy::missing_panics_doc,
-    clippy::module_name_repetitions,
-    clippy::multiple_crate_versions,
-    clippy::must_use_candidate,
-    clippy::needless_for_each,
-    clippy::needless_pass_by_value,
-    clippy::needless_raw_string_hashes,
-    clippy::option_as_ref_cloned,
-    clippy::redundant_closure_for_method_calls,
-    clippy::redundant_else,
-    clippy::return_self_not_must_use,
-    clippy::similar_names,
-    clippy::single_char_pattern,
-    clippy::single_match_else,
-    clippy::struct_excessive_bools,
-    clippy::struct_field_names,
-    clippy::too_many_lines,
-    clippy::uninlined_format_args,
-    clippy::unnested_or_patterns,
-    clippy::unreadable_literal,
-    clippy::unused_async,
-    clippy::unused_self,
-    clippy::used_underscore_binding,
-    clippy::wildcard_imports
-)]
-// List of temporarily allowed lints to unblock beta/nightly.
-#![allow(unknown_lints, clippy::manual_inspect)]
+#![deny(clippy::undocumented_unsafe_blocks)]
 
 use std::convert::Infallible;
 

proxy/src/metrics.rs

@@ -252,7 +252,7 @@ impl Drop for HttpEndpointPoolsGuard<'_> {
 }
 
 impl HttpEndpointPools {
-    pub fn guard(&self) -> HttpEndpointPoolsGuard<'_> {
+    pub fn guard(&self) -> HttpEndpointPoolsGuard {
         self.http_pool_endpoints_registered_total.inc();
         HttpEndpointPoolsGuard {
             dec: &self.http_pool_endpoints_unregistered_total,

proxy/src/proxy/copy_bidirectional.rs

@@ -184,7 +184,7 @@ impl CopyBuffer {
                 }
                 Poll::Pending
             }
-            res @ Poll::Ready(_) => res.map_err(ErrorDirection::Write),
+            res => res.map_err(ErrorDirection::Write),
         }
     }
 

proxy/src/proxy/handshake.rs

@@ -82,8 +82,9 @@ pub async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
     let mut stream = PqStream::new(Stream::from_raw(stream));
     loop {
         let msg = stream.read_startup_packet().await?;
+        use FeStartupPacket::*;
         match msg {
-            FeStartupPacket::SslRequest { direct } => match stream.get_ref() {
+            SslRequest { direct } => match stream.get_ref() {
                 Stream::Raw { .. } if !tried_ssl => {
                     tried_ssl = true;
 
@@ -138,7 +139,7 @@ pub async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
 
                         let tls_stream = accept.await.inspect_err(|_| {
                             if record_handshake_error {
-                                Metrics::get().proxy.tls_handshake_failures.inc();
+                                Metrics::get().proxy.tls_handshake_failures.inc()
                             }
                         })?;
 
@@ -181,7 +182,7 @@ pub async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
                 }
                 _ => return Err(HandshakeError::ProtocolViolation),
             },
-            FeStartupPacket::GssEncRequest => match stream.get_ref() {
+            GssEncRequest => match stream.get_ref() {
                 Stream::Raw { .. } if !tried_gss => {
                     tried_gss = true;
 
@@ -190,7 +191,7 @@ pub async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
                 }
                 _ => return Err(HandshakeError::ProtocolViolation),
             },
-            FeStartupPacket::StartupMessage { params, version }
+            StartupMessage { params, version }
                 if PG_PROTOCOL_EARLIEST <= version && version <= PG_PROTOCOL_LATEST =>
             {
                 // Check that the config has been consumed during upgrade
@@ -210,7 +211,7 @@ pub async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
                 break Ok(HandshakeData::Startup(stream, params));
             }
             // downgrade protocol version
-            FeStartupPacket::StartupMessage { params, version }
+            StartupMessage { params, version }
                 if version.major() == 3 && version > PG_PROTOCOL_LATEST =>
             {
                 warn!(?version, "unsupported minor version");
@@ -240,7 +241,7 @@ pub async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
                 );
                 break Ok(HandshakeData::Startup(stream, params));
             }
-            FeStartupPacket::StartupMessage { version, .. } => {
+            StartupMessage { version, .. } => {
                 warn!(
                     ?version,
                     session_type = "normal",
@@ -248,7 +249,7 @@ pub async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
                 );
                 return Err(HandshakeError::ProtocolViolation);
             }
-            FeStartupPacket::CancelRequest(cancel_key_data) => {
+            CancelRequest(cancel_key_data) => {
                 info!(session_type = "cancellation", "successful handshake");
                 break Ok(HandshakeData::Cancel(cancel_key_data));
             }

proxy/src/proxy/tests/mitm.rs

@@ -68,7 +68,7 @@ async fn proxy_mitm(
                                 end_client.send(Bytes::from_static(b"R\0\0\0\x17\0\0\0\x0aSCRAM-SHA-256\0\0")).await.unwrap();
                                 continue;
                             }
-                            end_client.send(message).await.unwrap();
+                            end_client.send(message).await.unwrap()
                         }
                         _ => break,
                     }
@@ -88,7 +88,7 @@ async fn proxy_mitm(
                                 end_server.send(buf.freeze()).await.unwrap();
                                 continue;
                             }
-                            end_server.send(message).await.unwrap();
+                            end_server.send(message).await.unwrap()
                         }
                         _ => break,
                     }

proxy/src/rate_limiter/limit_algorithm.rs

@@ -237,7 +237,7 @@ impl Token {
     }
 
     pub fn release(mut self, outcome: Outcome) {
-        self.release_mut(Some(outcome));
+        self.release_mut(Some(outcome))
     }
 
     pub fn release_mut(&mut self, outcome: Option<Outcome>) {
@@ -249,7 +249,7 @@ impl Token {
 
 impl Drop for Token {
     fn drop(&mut self) {
-        self.release_mut(None);
+        self.release_mut(None)
     }
 }
 

proxy/src/rate_limiter/limit_algorithm/aimd.rs

@@ -25,8 +25,9 @@ pub struct Aimd {
 
 impl LimitAlgorithm for Aimd {
     fn update(&self, old_limit: usize, sample: Sample) -> usize {
+        use Outcome::*;
         match sample.outcome {
-            Outcome::Success => {
+            Success => {
                 let utilisation = sample.in_flight as f32 / old_limit as f32;
 
                 if utilisation > self.utilisation {
@@ -41,7 +42,7 @@ impl LimitAlgorithm for Aimd {
                     old_limit
                 }
             }
-            Outcome::Overload => {
+            Overload => {
                 let limit = old_limit as f32 * self.dec;
 
                 // Floor instead of round, so the limit reduces even with small numbers.

proxy/src/redis/connection_with_credentials_provider.rs

@@ -98,7 +98,7 @@ impl ConnectionWithCredentialsProvider {
         info!("Establishing a new connection...");
         self.con = None;
         if let Some(f) = self.refresh_token_task.take() {
-            f.abort();
+            f.abort()
         }
         let mut con = self
             .get_client()

proxy/src/redis/notifications.rs

@@ -108,6 +108,7 @@ impl<C: ProjectInfoCache + Send + Sync + 'static> MessageHandler<C> {
     }
     #[tracing::instrument(skip(self, msg), fields(session_id = tracing::field::Empty))]
     async fn handle_message(&self, msg: redis::Msg) -> anyhow::Result<()> {
+        use Notification::*;
         let payload: String = msg.get_payload()?;
         tracing::debug!(?payload, "received a message payload");
 
@@ -123,7 +124,7 @@ impl<C: ProjectInfoCache + Send + Sync + 'static> MessageHandler<C> {
         };
         tracing::debug!(?msg, "received a message");
         match msg {
-            Notification::Cancel(cancel_session) => {
+            Cancel(cancel_session) => {
                 tracing::Span::current().record(
                     "session_id",
                     tracing::field::display(cancel_session.session_id),
@@ -152,12 +153,12 @@ impl<C: ProjectInfoCache + Send + Sync + 'static> MessageHandler<C> {
             }
             _ => {
                 invalidate_cache(self.cache.clone(), msg.clone());
-                if matches!(msg, Notification::AllowedIpsUpdate { .. }) {
+                if matches!(msg, AllowedIpsUpdate { .. }) {
                     Metrics::get()
                         .proxy
                         .redis_events_count
                         .inc(RedisEventsCount::AllowedIpsUpdate);
-                } else if matches!(msg, Notification::PasswordUpdate { .. }) {
+                } else if matches!(msg, PasswordUpdate { .. }) {
                     Metrics::get()
                         .proxy
                         .redis_events_count
@@ -179,16 +180,16 @@ impl<C: ProjectInfoCache + Send + Sync + 'static> MessageHandler<C> {
 }
 
 fn invalidate_cache<C: ProjectInfoCache>(cache: Arc<C>, msg: Notification) {
+    use Notification::*;
     match msg {
-        Notification::AllowedIpsUpdate { allowed_ips_update } => {
-            cache.invalidate_allowed_ips_for_project(allowed_ips_update.project_id);
+        AllowedIpsUpdate { allowed_ips_update } => {
+            cache.invalidate_allowed_ips_for_project(allowed_ips_update.project_id)
         }
-        Notification::PasswordUpdate { password_update } => cache
-            .invalidate_role_secret_for_project(
-                password_update.project_id,
-                password_update.role_name,
-            ),
-        Notification::Cancel(_) => unreachable!("cancel message should be handled separately"),
+        PasswordUpdate { password_update } => cache.invalidate_role_secret_for_project(
+            password_update.project_id,
+            password_update.role_name,
+        ),
+        Cancel(_) => unreachable!("cancel message should be handled separately"),
     }
 }
 

proxy/src/sasl.rs

@@ -42,9 +42,10 @@ pub enum Error {
 
 impl UserFacingError for Error {
     fn to_string_client(&self) -> String {
+        use Error::*;
         match self {
-            Self::ChannelBindingFailed(m) => (*m).to_string(),
-            Self::ChannelBindingBadMethod(m) => format!("unsupported channel binding method {m}"),
+            ChannelBindingFailed(m) => m.to_string(),
+            ChannelBindingBadMethod(m) => format!("unsupported channel binding method {m}"),
             _ => "authentication protocol violation".to_string(),
         }
     }

proxy/src/sasl/channel_binding.rs

@@ -13,10 +13,11 @@ pub enum ChannelBinding<T> {
 
 impl<T> ChannelBinding<T> {
     pub fn and_then<R, E>(self, f: impl FnOnce(T) -> Result<R, E>) -> Result<ChannelBinding<R>, E> {
+        use ChannelBinding::*;
         Ok(match self {
-            Self::NotSupportedClient => ChannelBinding::NotSupportedClient,
-            Self::NotSupportedServer => ChannelBinding::NotSupportedServer,
-            Self::Required(x) => ChannelBinding::Required(f(x)?),
+            NotSupportedClient => NotSupportedClient,
+            NotSupportedServer => NotSupportedServer,
+            Required(x) => Required(f(x)?),
         })
     }
 }
@@ -24,10 +25,11 @@ impl<T> ChannelBinding<T> {
 impl<'a> ChannelBinding<&'a str> {
     // NB: FromStr doesn't work with lifetimes
     pub fn parse(input: &'a str) -> Option<Self> {
+        use ChannelBinding::*;
         Some(match input {
-            "n" => Self::NotSupportedClient,
-            "y" => Self::NotSupportedServer,
-            other => Self::Required(other.strip_prefix("p=")?),
+            "n" => NotSupportedClient,
+            "y" => NotSupportedServer,
+            other => Required(other.strip_prefix("p=")?),
         })
     }
 }
@@ -38,16 +40,17 @@ impl<T: std::fmt::Display> ChannelBinding<T> {
         &self,
         get_cbind_data: impl FnOnce(&T) -> Result<&'a [u8], E>,
     ) -> Result<std::borrow::Cow<'static, str>, E> {
+        use ChannelBinding::*;
         Ok(match self {
-            Self::NotSupportedClient => {
+            NotSupportedClient => {
                 // base64::encode("n,,")
                 "biws".into()
             }
-            Self::NotSupportedServer => {
+            NotSupportedServer => {
                 // base64::encode("y,,")
                 "eSws".into()
             }
-            Self::Required(mode) => {
+            Required(mode) => {
                 use std::io::Write;
                 let mut cbind_input = vec![];
                 write!(&mut cbind_input, "p={mode},,",).unwrap();

proxy/src/sasl/messages.rs

@@ -42,9 +42,10 @@ pub(super) enum ServerMessage<T> {
 
 impl<'a> ServerMessage<&'a str> {
     pub(super) fn to_reply(&self) -> BeMessage<'a> {
+        use BeAuthenticationSaslMessage::*;
         BeMessage::AuthenticationSasl(match self {
-            ServerMessage::Continue(s) => BeAuthenticationSaslMessage::Continue(s.as_bytes()),
-            ServerMessage::Final(s) => BeAuthenticationSaslMessage::Final(s.as_bytes()),
+            ServerMessage::Continue(s) => Continue(s.as_bytes()),
+            ServerMessage::Final(s) => Final(s.as_bytes()),
         })
     }
 }

proxy/src/scram.rs

@@ -137,12 +137,12 @@ mod tests {
 
     #[tokio::test]
     async fn round_trip() {
-        run_round_trip_test("pencil", "pencil").await;
+        run_round_trip_test("pencil", "pencil").await
     }
 
     #[tokio::test]
     #[should_panic(expected = "password doesn't match")]
     async fn failure() {
-        run_round_trip_test("pencil", "eraser").await;
+        run_round_trip_test("pencil", "eraser").await
     }
 }

proxy/src/scram/countmin.rs

@@ -98,6 +98,8 @@ mod tests {
         // q% of counts will be within p of the actual value
         let mut sketch = CountMinSketch::with_params(p / N as f64, 1.0 - q);
 
+        dbg!(sketch.buckets.len());
+
         // insert a bunch of entries in a random order
         let mut ids2 = ids.clone();
         while !ids2.is_empty() {

proxy/src/scram/exchange.rs

@@ -210,23 +210,23 @@ impl sasl::Mechanism for Exchange<'_> {
     type Output = super::ScramKey;
 
     fn exchange(mut self, input: &str) -> sasl::Result<sasl::Step<Self, Self::Output>> {
-        use {sasl::Step, ExchangeState};
+        use {sasl::Step::*, ExchangeState::*};
         match &self.state {
-            ExchangeState::Initial(init) => {
+            Initial(init) => {
                 match init.transition(self.secret, &self.tls_server_end_point, input)? {
-                    Step::Continue(sent, msg) => {
-                        self.state = ExchangeState::SaltSent(sent);
-                        Ok(Step::Continue(self, msg))
+                    Continue(sent, msg) => {
+                        self.state = SaltSent(sent);
+                        Ok(Continue(self, msg))
                     }
-                    Step::Success(x, _) => match x {},
-                    Step::Failure(msg) => Ok(Step::Failure(msg)),
+                    Success(x, _) => match x {},
+                    Failure(msg) => Ok(Failure(msg)),
                 }
             }
-            ExchangeState::SaltSent(sent) => {
+            SaltSent(sent) => {
                 match sent.transition(self.secret, &self.tls_server_end_point, input)? {
-                    Step::Success(keys, msg) => Ok(Step::Success(keys, msg)),
-                    Step::Continue(x, _) => match x {},
-                    Step::Failure(msg) => Ok(Step::Failure(msg)),
+                    Success(keys, msg) => Ok(Success(keys, msg)),
+                    Continue(x, _) => match x {},
+                    Failure(msg) => Ok(Failure(msg)),
                 }
             }
         }

proxy/src/scram/messages.rs

@@ -59,7 +59,7 @@ impl<'a> ClientFirstMessage<'a> {
 
         // https://github.com/postgres/postgres/blob/f83908798f78c4cafda217ca875602c88ea2ae28/src/backend/libpq/auth-scram.c#L13-L14
         if !username.is_empty() {
-            tracing::warn!(username, "scram username provided, but is not expected");
+            tracing::warn!(username, "scram username provided, but is not expected")
             // TODO(conrad):
             // return None;
         }
@@ -137,7 +137,7 @@ impl<'a> ClientFinalMessage<'a> {
     /// Build a response to [`ClientFinalMessage`].
     pub fn build_server_final_message(
         &self,
-        signature_builder: SignatureBuilder<'_>,
+        signature_builder: SignatureBuilder,
         server_key: &ScramKey,
     ) -> String {
         let mut buf = String::from("v=");
@@ -212,7 +212,7 @@ mod tests {
 
     #[test]
     fn parse_client_first_message_with_invalid_gs2_authz() {
-        assert!(ClientFirstMessage::parse("n,authzid,n=,r=nonce").is_none());
+        assert!(ClientFirstMessage::parse("n,authzid,n=,r=nonce").is_none())
     }
 
     #[test]

proxy/src/scram/pbkdf2.rs

@@ -84,6 +84,6 @@ mod tests {
         };
 
         let expected = pbkdf2_hmac_array::<Sha256, 32>(pass, salt, 600000);
-        assert_eq!(hash, expected);
+        assert_eq!(hash, expected)
     }
 }

proxy/src/scram/threadpool.rs

@@ -270,7 +270,7 @@ fn thread_rt(pool: Arc<ThreadPool>, worker: Worker<JobSpec>, index: usize) {
                         .inc(ThreadPoolWorkerId(index));
 
                     // skip for now
-                    worker.push(job);
+                    worker.push(job)
                 }
             }
 
@@ -316,6 +316,6 @@ mod tests {
             10, 114, 73, 188, 140, 222, 196, 156, 214, 184, 79, 157, 119, 242, 16, 31, 53, 242,
             178, 43, 95, 8, 225, 182, 122, 40, 219, 21, 89, 147, 64, 140,
         ];
-        assert_eq!(actual, expected);
+        assert_eq!(actual, expected)
     }
 }

proxy/src/serverless.rs

@@ -10,7 +10,6 @@ mod json;
 mod sql_over_http;
 mod websocket;
 
-use async_trait::async_trait;
 use atomic_take::AtomicTake;
 use bytes::Bytes;
 pub use conn_pool::GlobalConnPoolOptions;
@@ -27,9 +26,8 @@ use rand::rngs::StdRng;
 use rand::SeedableRng;
 pub use reqwest_middleware::{ClientWithMiddleware, Error};
 pub use reqwest_retry::{policies::ExponentialBackoff, RetryTransientMiddleware};
-use tokio::io::{AsyncRead, AsyncWrite};
 use tokio::time::timeout;
-use tokio_rustls::TlsAcceptor;
+use tokio_rustls::{server::TlsStream, TlsAcceptor};
 use tokio_util::task::TaskTracker;
 
 use crate::cancellation::CancellationHandlerMain;
@@ -43,7 +41,7 @@ use crate::serverless::backend::PoolingBackend;
 use crate::serverless::http_util::{api_error_into_response, json_response};
 
 use std::net::{IpAddr, SocketAddr};
-use std::pin::{pin, Pin};
+use std::pin::pin;
 use std::sync::Arc;
 use tokio::net::{TcpListener, TcpStream};
 use tokio_util::sync::CancellationToken;
@@ -88,18 +86,18 @@ pub async fn task_main(
         config,
         endpoint_rate_limiter: Arc::clone(&endpoint_rate_limiter),
     });
-    let tls_acceptor: Arc<dyn MaybeTlsAcceptor> = match config.tls_config.as_ref() {
-        Some(config) => {
-            let mut tls_server_config = rustls::ServerConfig::clone(&config.to_server_config());
-            // prefer http2, but support http/1.1
-            tls_server_config.alpn_protocols = vec![b"h2".to_vec(), b"http/1.1".to_vec()];
-            Arc::new(tls_server_config) as Arc<_>
-        }
+
+    let tls_config = match config.tls_config.as_ref() {
+        Some(config) => config,
         None => {
-            warn!("TLS config is missing");
-            Arc::new(NoTls) as Arc<_>
+            warn!("TLS config is missing, WebSocket Secure server will not be started");
+            return Ok(());
         }
     };
+    let mut tls_server_config = rustls::ServerConfig::clone(&tls_config.to_server_config());
+    // prefer http2, but support http/1.1
+    tls_server_config.alpn_protocols = vec![b"h2".to_vec(), b"http/1.1".to_vec()];
+    let tls_acceptor: tokio_rustls::TlsAcceptor = Arc::new(tls_server_config).into();
 
     let connections = tokio_util::task::task_tracker::TaskTracker::new();
     connections.close(); // allows `connections.wait to complete`
@@ -122,7 +120,7 @@ pub async fn task_main(
             tracing::trace!("attempting to cancel a random connection");
             if let Some(token) = config.http_config.cancel_set.take() {
                 tracing::debug!("cancelling a random connection");
-                token.cancel();
+                token.cancel()
             }
         }
 
@@ -178,41 +176,16 @@ pub async fn task_main(
     Ok(())
 }
 
-pub trait AsyncReadWrite: AsyncRead + AsyncWrite + Send + 'static {}
-impl<T: AsyncRead + AsyncWrite + Send + 'static> AsyncReadWrite for T {}
-pub type AsyncRW = Pin<Box<dyn AsyncReadWrite>>;
-
-#[async_trait]
-trait MaybeTlsAcceptor: Send + Sync + 'static {
-    async fn accept(self: Arc<Self>, conn: ChainRW<TcpStream>) -> std::io::Result<AsyncRW>;
-}
-
-#[async_trait]
-impl MaybeTlsAcceptor for rustls::ServerConfig {
-    async fn accept(self: Arc<Self>, conn: ChainRW<TcpStream>) -> std::io::Result<AsyncRW> {
-        Ok(Box::pin(TlsAcceptor::from(self).accept(conn).await?))
-    }
-}
-
-struct NoTls;
-
-#[async_trait]
-impl MaybeTlsAcceptor for NoTls {
-    async fn accept(self: Arc<Self>, conn: ChainRW<TcpStream>) -> std::io::Result<AsyncRW> {
-        Ok(Box::pin(conn))
-    }
-}
-
 /// Handles the TCP startup lifecycle.
 /// 1. Parses PROXY protocol V2
 /// 2. Handles TLS handshake
 async fn connection_startup(
     config: &ProxyConfig,
-    tls_acceptor: Arc<dyn MaybeTlsAcceptor>,
+    tls_acceptor: TlsAcceptor,
     session_id: uuid::Uuid,
     conn: TcpStream,
     peer_addr: SocketAddr,
-) -> Option<(AsyncRW, IpAddr)> {
+) -> Option<(TlsStream<ChainRW<TcpStream>>, IpAddr)> {
     // handle PROXY protocol
     let (conn, peer) = match read_proxy_protocol(conn).await {
         Ok(c) => c,
@@ -225,7 +198,7 @@ async fn connection_startup(
     let peer_addr = peer.unwrap_or(peer_addr).ip();
     let has_private_peer_addr = match peer_addr {
         IpAddr::V4(ip) => ip.is_private(),
-        IpAddr::V6(_) => false,
+        _ => false,
     };
     info!(?session_id, %peer_addr, "accepted new TCP connection");
 
@@ -268,7 +241,7 @@ async fn connection_handler(
     cancellation_handler: Arc<CancellationHandlerMain>,
     endpoint_rate_limiter: Arc<EndpointRateLimiter>,
     cancellation_token: CancellationToken,
-    conn: AsyncRW,
+    conn: TlsStream<ChainRW<TcpStream>>,
     peer_addr: IpAddr,
     session_id: uuid::Uuid,
 ) {
@@ -353,9 +326,7 @@ async fn request_handler(
         .map(|s| s.to_string());
 
     // Check if the request is a websocket upgrade request.
-    if config.http_config.accept_websockets
-        && framed_websockets::upgrade::is_upgrade_request(&request)
-    {
+    if framed_websockets::upgrade::is_upgrade_request(&request) {
         let ctx = RequestMonitoring::new(
             session_id,
             peer_addr,

proxy/src/serverless/backend.rs

@@ -5,7 +5,8 @@ use tracing::{field::display, info};
 
 use crate::{
     auth::{backend::ComputeCredentials, check_peer_addr_is_in_list, AuthError},
-    compute,
+    cancellation::CancelClosure,
+    compute::{self, ConnectionError},
     config::{AuthenticationConfig, ProxyConfig},
     console::{
         errors::{GetAuthInfoError, WakeComputeError},
@@ -142,7 +143,7 @@ pub enum HttpConnError {
     #[error("pooled connection closed at inconsistent state")]
     ConnectionClosedAbruptly(#[from] tokio::sync::watch::error::SendError<uuid::Uuid>),
     #[error("could not connection to compute")]
-    ConnectionError(#[from] tokio_postgres::Error),
+    ConnectionError(#[from] ConnectionError),
 
     #[error("could not get auth info")]
     GetAuthInfo(#[from] GetAuthInfoError),
@@ -229,17 +230,16 @@ impl ConnectMechanism for TokioMechanism {
         let host = node_info.config.get_host()?;
         let permit = self.locks.get_permit(&host).await?;
 
-        let mut config = (*node_info.config).clone();
-        let config = config
-            .user(&self.conn_info.user_info.user)
-            .password(&*self.conn_info.password)
-            .dbname(&self.conn_info.dbname)
-            .connect_timeout(timeout);
+        let (socket_addr, client, connection) = permit.release_result(
+            node_info
+                .config
+                .managed_connect(ctx, timeout, &mut tokio_postgres::NoTls)
+                .await,
+        )?;
 
-        let pause = ctx.latency_timer_pause(crate::metrics::Waiting::Compute);
-        let res = config.connect(tokio_postgres::NoTls).await;
-        drop(pause);
-        let (client, connection) = permit.release_result(res)?;
+        // NB: CancelToken is supposed to hold socket_addr, but we use connect_raw.
+        // Yet another reason to rework the connection establishing code.
+        let cancel_closure = CancelClosure::new(socket_addr, client.cancel_token());
 
         tracing::Span::current().record("pid", tracing::field::display(client.get_process_id()));
         Ok(poll_client(
@@ -250,8 +250,14 @@ impl ConnectMechanism for TokioMechanism {
             connection,
             self.conn_id,
             node_info.aux.clone(),
+            cancel_closure,
         ))
     }
 
-    fn update_connect_config(&self, _config: &mut compute::ConnCfg) {}
+    fn update_connect_config(&self, config: &mut compute::ConnCfg) {
+        config
+            .user(&self.conn_info.user_info.user)
+            .dbname(&self.conn_info.dbname)
+            .password(&self.conn_info.password);
+    }
 }

proxy/src/serverless/conn_pool.rs

@@ -12,11 +12,13 @@ use std::{
     ops::Deref,
     sync::atomic::{self, AtomicUsize},
 };
+use tokio::net::TcpStream;
 use tokio::time::Instant;
 use tokio_postgres::tls::NoTlsStream;
-use tokio_postgres::{AsyncMessage, ReadyForQueryStatus, Socket};
+use tokio_postgres::{AsyncMessage, ReadyForQueryStatus};
 use tokio_util::sync::CancellationToken;
 
+use crate::cancellation::CancelClosure;
 use crate::console::messages::{ColdStartInfo, MetricsAuxInfo};
 use crate::metrics::{HttpEndpointPoolsGuard, Metrics};
 use crate::usage_metrics::{Ids, MetricCounter, USAGE_METRICS};
@@ -390,7 +392,7 @@ impl<C: ClientInnerExt> GlobalConnPool<C> {
             .write()
             .get_conn_entry(conn_info.db_and_user())
         {
-            client = Some(entry.conn);
+            client = Some(entry.conn)
         }
         let endpoint_pool = Arc::downgrade(&endpoint_pool);
 
@@ -463,14 +465,16 @@ impl<C: ClientInnerExt> GlobalConnPool<C> {
     }
 }
 
+#[allow(clippy::too_many_arguments)]
 pub fn poll_client<C: ClientInnerExt>(
     global_pool: Arc<GlobalConnPool<C>>,
     ctx: &RequestMonitoring,
     conn_info: ConnInfo,
     client: C,
-    mut connection: tokio_postgres::Connection<Socket, NoTlsStream>,
+    mut connection: tokio_postgres::Connection<TcpStream, NoTlsStream>,
     conn_id: uuid::Uuid,
     aux: MetricsAuxInfo,
+    cancel_closure: CancelClosure,
 ) -> Client<C> {
     let conn_gauge = Metrics::get().proxy.db_connections.guard(ctx.protocol());
     let mut session_id = ctx.session_id();
@@ -572,6 +576,7 @@ pub fn poll_client<C: ClientInnerExt>(
         cancel,
         aux,
         conn_id,
+        cancel_closure,
     };
     Client::new(inner, conn_info, pool_clone)
 }
@@ -582,6 +587,7 @@ struct ClientInner<C: ClientInnerExt> {
     cancel: CancellationToken,
     aux: MetricsAuxInfo,
     conn_id: uuid::Uuid,
+    cancel_closure: CancelClosure,
 }
 
 impl<C: ClientInnerExt> Drop for ClientInner<C> {
@@ -646,7 +652,7 @@ impl<C: ClientInnerExt> Client<C> {
             pool,
         }
     }
-    pub fn inner(&mut self) -> (&mut C, Discard<'_, C>) {
+    pub fn inner(&mut self) -> (&mut C, &CancelClosure, Discard<'_, C>) {
         let Self {
             inner,
             pool,
@@ -654,7 +660,11 @@ impl<C: ClientInnerExt> Client<C> {
             span: _,
         } = self;
         let inner = inner.as_mut().expect("client inner should not be removed");
-        (&mut inner.inner, Discard { pool, conn_info })
+        (
+            &mut inner.inner,
+            &inner.cancel_closure,
+            Discard { pool, conn_info },
+        )
     }
 }
 
@@ -662,13 +672,13 @@ impl<C: ClientInnerExt> Discard<'_, C> {
     pub fn check_idle(&mut self, status: ReadyForQueryStatus) {
         let conn_info = &self.conn_info;
         if status != ReadyForQueryStatus::Idle && std::mem::take(self.pool).strong_count() > 0 {
-            info!("pool: throwing away connection '{conn_info}' because connection is not idle");
+            info!("pool: throwing away connection '{conn_info}' because connection is not idle")
         }
     }
     pub fn discard(&mut self) {
         let conn_info = &self.conn_info;
         if std::mem::take(self.pool).strong_count() > 0 {
-            info!("pool: throwing away connection '{conn_info}' because connection is potentially in a broken state");
+            info!("pool: throwing away connection '{conn_info}' because connection is potentially in a broken state")
         }
     }
 }
@@ -751,6 +761,7 @@ mod tests {
                 cold_start_info: crate::console::messages::ColdStartInfo::Warm,
             },
             conn_id: uuid::Uuid::new_v4(),
+            cancel_closure: CancelClosure::test(),
         }
     }
 
@@ -758,7 +769,6 @@ mod tests {
     async fn test_pool() {
         let _ = env_logger::try_init();
         let config = Box::leak(Box::new(crate::config::HttpConfig {
-            accept_websockets: false,
             pool_options: GlobalConnPoolOptions {
                 max_conns_per_endpoint: 2,
                 gc_epoch: Duration::from_secs(1),
@@ -786,7 +796,7 @@ mod tests {
         {
             let mut client = Client::new(create_inner(), conn_info.clone(), ep_pool.clone());
             assert_eq!(0, pool.get_global_connections_count());
-            client.inner().1.discard();
+            client.inner().2.discard();
             // Discard should not add the connection from the pool.
             assert_eq!(0, pool.get_global_connections_count());
         }

proxy/src/serverless/sql_over_http.rs

@@ -26,7 +26,6 @@ use tokio_postgres::error::ErrorPosition;
 use tokio_postgres::error::SqlState;
 use tokio_postgres::GenericClient;
 use tokio_postgres::IsolationLevel;
-use tokio_postgres::NoTls;
 use tokio_postgres::ReadyForQueryStatus;
 use tokio_postgres::Transaction;
 use tokio_util::sync::CancellationToken;
@@ -147,7 +146,7 @@ impl UserFacingError for ConnInfoError {
 fn get_conn_info(
     ctx: &RequestMonitoring,
     headers: &HeaderMap,
-    tls: Option<&TlsConfig>,
+    tls: &TlsConfig,
 ) -> Result<ConnInfo, ConnInfoError> {
     // HTTP only uses cleartext (for now and likely always)
     ctx.set_auth_method(crate::context::AuthMethod::Cleartext);
@@ -184,22 +183,12 @@ fn get_conn_info(
         .ok_or(ConnInfoError::MissingPassword)?;
     let password = urlencoding::decode_binary(password.as_bytes());
 
-    let endpoint = match connection_url.host() {
-        Some(url::Host::Domain(hostname)) => {
-            if let Some(tls) = tls {
-                endpoint_sni(hostname, &tls.common_names)?
-                    .ok_or(ConnInfoError::MalformedEndpoint)?
-            } else {
-                hostname
-                    .split_once(".")
-                    .map_or(hostname, |(prefix, _)| prefix)
-                    .into()
-            }
-        }
-        Some(url::Host::Ipv4(_)) | Some(url::Host::Ipv6(_)) | None => {
-            return Err(ConnInfoError::MissingHostname)
-        }
-    };
+    let hostname = connection_url
+        .host_str()
+        .ok_or(ConnInfoError::MissingHostname)?;
+
+    let endpoint =
+        endpoint_sni(hostname, &tls.common_names)?.ok_or(ConnInfoError::MalformedEndpoint)?;
     ctx.set_endpoint_id(endpoint.clone());
 
     let pairs = connection_url.query_pairs();
@@ -271,7 +260,9 @@ pub async fn handle(
 
             let mut message = e.to_string_client();
             let db_error = match &e {
-                SqlOverHttpError::ConnectCompute(HttpConnError::ConnectionError(e))
+                SqlOverHttpError::ConnectCompute(HttpConnError::ConnectionError(
+                    crate::compute::ConnectionError::Postgres(e),
+                ))
                 | SqlOverHttpError::Postgres(e) => e.as_db_error(),
                 _ => None,
             };
@@ -512,7 +503,7 @@ async fn handle_inner(
     let headers = request.headers();
 
     // TLS config should be there.
-    let conn_info = get_conn_info(ctx, headers, config.tls_config.as_ref())?;
+    let conn_info = get_conn_info(ctx, headers, config.tls_config.as_ref().unwrap())?;
     info!(user = conn_info.user_info.user.as_str(), "credentials");
 
     // Allow connection pooling only if explicitly requested
@@ -632,8 +623,7 @@ impl QueryData {
         client: &mut Client<tokio_postgres::Client>,
         parsed_headers: HttpHeaders,
     ) -> Result<String, SqlOverHttpError> {
-        let (inner, mut discard) = client.inner();
-        let cancel_token = inner.cancel_token();
+        let (inner, cancel_token, mut discard) = client.inner();
 
         let res = match select(
             pin!(query_to_json(&*inner, self, &mut 0, parsed_headers)),
@@ -657,7 +647,7 @@ impl QueryData {
             // The query was cancelled.
             Either::Right((_cancelled, query)) => {
                 tracing::info!("cancelling query");
-                if let Err(err) = cancel_token.cancel_query(NoTls).await {
+                if let Err(err) = cancel_token.clone().try_cancel_query().await {
                     tracing::error!(?err, "could not cancel query");
                 }
                 // wait for the query cancellation
@@ -673,7 +663,9 @@ impl QueryData {
                     // query failed or was cancelled.
                     Ok(Err(error)) => {
                         let db_error = match &error {
-                            SqlOverHttpError::ConnectCompute(HttpConnError::ConnectionError(e))
+                            SqlOverHttpError::ConnectCompute(HttpConnError::ConnectionError(
+                                crate::compute::ConnectionError::Postgres(e),
+                            ))
                             | SqlOverHttpError::Postgres(e) => e.as_db_error(),
                             _ => None,
                         };
@@ -704,8 +696,7 @@ impl BatchQueryData {
         parsed_headers: HttpHeaders,
     ) -> Result<String, SqlOverHttpError> {
         info!("starting transaction");
-        let (inner, mut discard) = client.inner();
-        let cancel_token = inner.cancel_token();
+        let (inner, cancel_token, mut discard) = client.inner();
         let mut builder = inner.build_transaction();
         if let Some(isolation_level) = parsed_headers.txn_isolation_level {
             builder = builder.isolation_level(isolation_level);
@@ -738,7 +729,7 @@ impl BatchQueryData {
                     json_output
                 }
                 Err(SqlOverHttpError::Cancelled(_)) => {
-                    if let Err(err) = cancel_token.cancel_query(NoTls).await {
+                    if let Err(err) = cancel_token.clone().try_cancel_query().await {
                         tracing::error!(?err, "could not cancel query");
                     }
                     // TODO: after cancelling, wait to see if we can get a status. maybe the connection is still safe.