Test logging

Cargo.toml

@@ -92,7 +92,7 @@ http-types = { version = "2", default-features = false }
 humantime = "2.1"
 humantime-serde = "1.1.1"
 hyper = "0.14"
-hyper-tungstenite = "0.13.0"
+hyper-tungstenite = "0.11"
 inotify = "0.10.2"
 ipnet = "2.9.0"
 itertools = "0.10"
@@ -129,8 +129,8 @@ reqwest-retry = "0.2.2"
 routerify = "3"
 rpds = "0.13"
 rustc-hash = "1.1.0"
-rustls = "0.22"
-rustls-pemfile = "2"
+rustls = "0.21"
+rustls-pemfile = "1"
 rustls-split = "0.3"
 scopeguard = "1.1"
 sysinfo = "0.29.2"
@@ -156,11 +156,12 @@ test-context = "0.1"
 thiserror = "1.0"
 tikv-jemallocator = "0.5"
 tikv-jemalloc-ctl = "0.5"
+tls-listener = { version = "0.7", features = ["rustls", "hyper-h1"] }
 tokio = { version = "1.17", features = ["macros"] }
 tokio-epoll-uring = { git = "https://github.com/neondatabase/tokio-epoll-uring.git" , branch = "main" }
 tokio-io-timeout = "1.2.0"
-tokio-postgres-rustls = "0.11.0"
-tokio-rustls = "0.25"
+tokio-postgres-rustls = "0.10.0"
+tokio-rustls = "0.24"
 tokio-stream = "0.1"
 tokio-tar = "0.3"
 tokio-util = { version = "0.7.10", features = ["io", "rt"] }
@@ -219,7 +220,7 @@ workspace_hack = { version = "0.1", path = "./workspace_hack/" }
 
 ## Build dependencies
 criterion = "0.5.1"
-rcgen = "0.12"
+rcgen = "0.11"
 rstest = "0.18"
 camino-tempfile = "1.0.2"
 tonic-build = "0.9"

Dockerfile

@@ -53,7 +53,7 @@ RUN set -e \
       --bin pagectl  \
       --bin safekeeper  \
       --bin storage_broker  \
-      --bin storage_controller  \
+      --bin attachment_service  \
       --bin proxy  \
       --bin neon_local \
       --locked --release \
@@ -81,7 +81,7 @@ COPY --from=build --chown=neon:neon /home/nonroot/target/release/pageserver
 COPY --from=build --chown=neon:neon /home/nonroot/target/release/pagectl             /usr/local/bin
 COPY --from=build --chown=neon:neon /home/nonroot/target/release/safekeeper          /usr/local/bin
 COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_broker      /usr/local/bin
-COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_controller  /usr/local/bin
+COPY --from=build --chown=neon:neon /home/nonroot/target/release/attachment_service  /usr/local/bin
 COPY --from=build --chown=neon:neon /home/nonroot/target/release/proxy               /usr/local/bin
 COPY --from=build --chown=neon:neon /home/nonroot/target/release/neon_local          /usr/local/bin
 

clippy.toml

@@ -3,10 +3,3 @@ disallowed-methods = [
     # Allow this for now, to deny it later once we stop using Handle::block_on completely
     # "tokio::runtime::Handle::block_on",
 ]
-
-disallowed-macros = [
-    # use std::pin::pin
-    "futures::pin_mut",
-    # cannot disallow this, because clippy finds used from tokio macros
-    #"tokio::pin",
-]

compute_tools/src/compute.rs

@@ -396,9 +396,9 @@ impl ComputeNode {
     // Gets the basebackup in a retry loop
     #[instrument(skip_all, fields(%lsn))]
     pub fn get_basebackup(&self, compute_state: &ComputeState, lsn: Lsn) -> Result<()> {
-        let mut retry_period_ms = 500.0;
+        let mut retry_period_ms = 500;
         let mut attempts = 0;
-        let max_attempts = 10;
+        let max_attempts = 5;
         loop {
             let result = self.try_get_basebackup(compute_state, lsn);
             match result {
@@ -410,8 +410,8 @@ impl ComputeNode {
                         "Failed to get basebackup: {} (attempt {}/{})",
                         e, attempts, max_attempts
                     );
-                    std::thread::sleep(std::time::Duration::from_millis(retry_period_ms as u64));
-                    retry_period_ms *= 1.5;
+                    std::thread::sleep(std::time::Duration::from_millis(retry_period_ms));
+                    retry_period_ms *= 2;
                 }
                 Err(_) => {
                     return result;
@@ -764,26 +764,6 @@ impl ComputeNode {
         Ok((pg, logs_handle))
     }
 
-    /// Do post configuration of the already started Postgres. This function spawns a background thread to
-    /// configure the database after applying the compute spec. Currently, it upgrades the neon extension
-    /// version. In the future, it may upgrade all 3rd-party extensions.
-    #[instrument(skip_all)]
-    pub fn post_apply_config(&self) -> Result<()> {
-        let connstr = self.connstr.clone();
-        thread::spawn(move || {
-            let func = || {
-                let mut client = Client::connect(connstr.as_str(), NoTls)?;
-                handle_neon_extension_upgrade(&mut client)
-                    .context("handle_neon_extension_upgrade")?;
-                Ok::<_, anyhow::Error>(())
-            };
-            if let Err(err) = func() {
-                error!("error while post_apply_config: {err:#}");
-            }
-        });
-        Ok(())
-    }
-
     /// Do initial configuration of the already started Postgres.
     #[instrument(skip_all)]
     pub fn apply_config(&self, compute_state: &ComputeState) -> Result<()> {
@@ -1018,21 +998,18 @@ impl ComputeNode {
         let pg_process = self.start_postgres(pspec.storage_auth_token.clone())?;
 
         let config_time = Utc::now();
-        if pspec.spec.mode == ComputeMode::Primary {
-            if !pspec.spec.skip_pg_catalog_updates {
-                let pgdata_path = Path::new(&self.pgdata);
-                // temporarily reset max_cluster_size in config
-                // to avoid the possibility of hitting the limit, while we are applying config:
-                // creating new extensions, roles, etc...
-                config::compute_ctl_temp_override_create(pgdata_path, "neon.max_cluster_size=-1")?;
-                self.pg_reload_conf()?;
+        if pspec.spec.mode == ComputeMode::Primary && !pspec.spec.skip_pg_catalog_updates {
+            let pgdata_path = Path::new(&self.pgdata);
+            // temporarily reset max_cluster_size in config
+            // to avoid the possibility of hitting the limit, while we are applying config:
+            // creating new extensions, roles, etc...
+            config::compute_ctl_temp_override_create(pgdata_path, "neon.max_cluster_size=-1")?;
+            self.pg_reload_conf()?;
 
-                self.apply_config(&compute_state)?;
+            self.apply_config(&compute_state)?;
 
-                config::compute_ctl_temp_override_remove(pgdata_path)?;
-                self.pg_reload_conf()?;
-            }
-            self.post_apply_config()?;
+            config::compute_ctl_temp_override_remove(pgdata_path)?;
+            self.pg_reload_conf()?;
         }
 
         let startup_end_time = Utc::now();

compute_tools/src/spec.rs

@@ -744,17 +744,7 @@ pub fn handle_extension_neon(client: &mut Client) -> Result<()> {
     // - extension was just installed
     // - extension was already installed and is up to date
     let query = "ALTER EXTENSION neon UPDATE";
-    info!("update neon extension version with query: {}", query);
-    client.simple_query(query)?;
-
-    Ok(())
-}
-
-#[instrument(skip_all)]
-pub fn handle_neon_extension_upgrade(client: &mut Client) -> Result<()> {
-    info!("handle neon extension upgrade");
-    let query = "ALTER EXTENSION neon UPDATE";
-    info!("update neon extension version with query: {}", query);
+    info!("update neon extension schema with query: {}", query);
     client.simple_query(query)?;
 
     Ok(())

control_plane/attachment_service/Cargo.toml

@@ -4,10 +4,6 @@ version = "0.1.0"
 edition.workspace = true
 license.workspace = true
 
-[[bin]]
-name = "storage_controller"
-path = "src/main.rs"
-
 [features]
 default = []
 # Enables test-only APIs and behaviors

control_plane/attachment_service/src/compute_hook.rs

@@ -3,7 +3,7 @@ use std::{collections::HashMap, time::Duration};
 use control_plane::endpoint::{ComputeControlPlane, EndpointStatus};
 use control_plane::local_env::LocalEnv;
 use hyper::{Method, StatusCode};
-use pageserver_api::shard::{ShardCount, ShardNumber, ShardStripeSize, TenantShardId};
+use pageserver_api::shard::{ShardIndex, ShardNumber, TenantShardId};
 use postgres_connection::parse_host_port;
 use serde::{Deserialize, Serialize};
 use tokio_util::sync::CancellationToken;
@@ -19,66 +19,8 @@ const SLOWDOWN_DELAY: Duration = Duration::from_secs(5);
 
 pub(crate) const API_CONCURRENCY: usize = 32;
 
-struct ShardedComputeHookTenant {
-    stripe_size: ShardStripeSize,
-    shard_count: ShardCount,
-    shards: Vec<(ShardNumber, NodeId)>,
-}
-
-enum ComputeHookTenant {
-    Unsharded(NodeId),
-    Sharded(ShardedComputeHookTenant),
-}
-
-impl ComputeHookTenant {
-    /// Construct with at least one shard's information
-    fn new(tenant_shard_id: TenantShardId, stripe_size: ShardStripeSize, node_id: NodeId) -> Self {
-        if tenant_shard_id.shard_count.count() > 1 {
-            Self::Sharded(ShardedComputeHookTenant {
-                shards: vec![(tenant_shard_id.shard_number, node_id)],
-                stripe_size,
-                shard_count: tenant_shard_id.shard_count,
-            })
-        } else {
-            Self::Unsharded(node_id)
-        }
-    }
-
-    /// Set one shard's location.  If stripe size or shard count have changed, Self is reset
-    /// and drops existing content.
-    fn update(
-        &mut self,
-        tenant_shard_id: TenantShardId,
-        stripe_size: ShardStripeSize,
-        node_id: NodeId,
-    ) {
-        match self {
-            Self::Unsharded(existing_node_id) if tenant_shard_id.shard_count.count() == 1 => {
-                *existing_node_id = node_id
-            }
-            Self::Sharded(sharded_tenant)
-                if sharded_tenant.stripe_size == stripe_size
-                    && sharded_tenant.shard_count == tenant_shard_id.shard_count =>
-            {
-                if let Some(existing) = sharded_tenant
-                    .shards
-                    .iter()
-                    .position(|s| s.0 == tenant_shard_id.shard_number)
-                {
-                    sharded_tenant.shards.get_mut(existing).unwrap().1 = node_id;
-                } else {
-                    sharded_tenant
-                        .shards
-                        .push((tenant_shard_id.shard_number, node_id));
-                    sharded_tenant.shards.sort_by_key(|s| s.0)
-                }
-            }
-            _ => {
-                // Shard count changed: reset struct.
-                *self = Self::new(tenant_shard_id, stripe_size, node_id);
-            }
-        }
-    }
+pub(super) struct ComputeHookTenant {
+    shards: Vec<(ShardIndex, NodeId)>,
 }
 
 #[derive(Serialize, Deserialize, Debug)]
@@ -91,7 +33,6 @@ struct ComputeHookNotifyRequestShard {
 #[derive(Serialize, Deserialize, Debug)]
 struct ComputeHookNotifyRequest {
     tenant_id: TenantId,
-    stripe_size: Option<ShardStripeSize>,
     shards: Vec<ComputeHookNotifyRequestShard>,
 }
 
@@ -122,43 +63,42 @@ pub(crate) enum NotifyError {
 }
 
 impl ComputeHookTenant {
-    fn maybe_reconfigure(&self, tenant_id: TenantId) -> Option<ComputeHookNotifyRequest> {
-        match self {
-            Self::Unsharded(node_id) => Some(ComputeHookNotifyRequest {
-                tenant_id,
-                shards: vec![ComputeHookNotifyRequestShard {
-                    shard_number: ShardNumber(0),
-                    node_id: *node_id,
-                }],
-                stripe_size: None,
-            }),
-            Self::Sharded(sharded_tenant)
-                if sharded_tenant.shards.len() == sharded_tenant.shard_count.count() as usize =>
-            {
-                Some(ComputeHookNotifyRequest {
-                    tenant_id,
-                    shards: sharded_tenant
-                        .shards
-                        .iter()
-                        .map(|(shard_number, node_id)| ComputeHookNotifyRequestShard {
-                            shard_number: *shard_number,
-                            node_id: *node_id,
-                        })
-                        .collect(),
-                    stripe_size: Some(sharded_tenant.stripe_size),
-                })
-            }
-            Self::Sharded(sharded_tenant) => {
-                // Sharded tenant doesn't yet have information for all its shards
+    async fn maybe_reconfigure(&mut self, tenant_id: TenantId) -> Option<ComputeHookNotifyRequest> {
+        // Find the highest shard count and drop any shards that aren't
+        // for that shard count.
+        let shard_count = self.shards.iter().map(|(k, _v)| k.shard_count).max();
+        let Some(shard_count) = shard_count else {
+            // No shards, nothing to do.
+            tracing::info!("ComputeHookTenant::maybe_reconfigure: no shards");
+            return None;
+        };
 
-                tracing::info!(
-                    "ComputeHookTenant::maybe_reconfigure: not enough shards ({}/{})",
-                    sharded_tenant.shards.len(),
-                    sharded_tenant.shard_count.count()
-                );
-                None
-            }
+        self.shards.retain(|(k, _v)| k.shard_count == shard_count);
+        self.shards
+            .sort_by_key(|(shard, _node_id)| shard.shard_number);
+
+        if self.shards.len() == shard_count.count() as usize || shard_count.is_unsharded() {
+            // We have pageservers for all the shards: emit a configuration update
+            return Some(ComputeHookNotifyRequest {
+                tenant_id,
+                shards: self
+                    .shards
+                    .iter()
+                    .map(|(shard, node_id)| ComputeHookNotifyRequestShard {
+                        shard_number: shard.shard_number,
+                        node_id: *node_id,
+                    })
+                    .collect(),
+            });
+        } else {
+            tracing::info!(
+                "ComputeHookTenant::maybe_reconfigure: not enough shards ({}/{})",
+                self.shards.len(),
+                shard_count.count()
+            );
         }
+
+        None
     }
 }
 
@@ -199,11 +139,7 @@ impl ComputeHook {
         };
         let cplane =
             ComputeControlPlane::load(env.clone()).expect("Error loading compute control plane");
-        let ComputeHookNotifyRequest {
-            tenant_id,
-            shards,
-            stripe_size,
-        } = reconfigure_request;
+        let ComputeHookNotifyRequest { tenant_id, shards } = reconfigure_request;
 
         let compute_pageservers = shards
             .into_iter()
@@ -220,9 +156,7 @@ impl ComputeHook {
         for (endpoint_name, endpoint) in &cplane.endpoints {
             if endpoint.tenant_id == tenant_id && endpoint.status() == EndpointStatus::Running {
                 tracing::info!("Reconfiguring endpoint {}", endpoint_name,);
-                endpoint
-                    .reconfigure(compute_pageservers.clone(), stripe_size)
-                    .await?;
+                endpoint.reconfigure(compute_pageservers.clone()).await?;
             }
         }
 
@@ -337,26 +271,30 @@ impl ComputeHook {
         &self,
         tenant_shard_id: TenantShardId,
         node_id: NodeId,
-        stripe_size: ShardStripeSize,
         cancel: &CancellationToken,
     ) -> Result<(), NotifyError> {
         let mut locked = self.state.lock().await;
+        let entry = locked
+            .entry(tenant_shard_id.tenant_id)
+            .or_insert_with(|| ComputeHookTenant { shards: Vec::new() });
 
-        use std::collections::hash_map::Entry;
-        let tenant = match locked.entry(tenant_shard_id.tenant_id) {
-            Entry::Vacant(e) => e.insert(ComputeHookTenant::new(
-                tenant_shard_id,
-                stripe_size,
-                node_id,
-            )),
-            Entry::Occupied(e) => {
-                let tenant = e.into_mut();
-                tenant.update(tenant_shard_id, stripe_size, node_id);
-                tenant
-            }
+        let shard_index = ShardIndex {
+            shard_count: tenant_shard_id.shard_count,
+            shard_number: tenant_shard_id.shard_number,
         };
 
-        let reconfigure_request = tenant.maybe_reconfigure(tenant_shard_id.tenant_id);
+        let mut set = false;
+        for (existing_shard, existing_node) in &mut entry.shards {
+            if *existing_shard == shard_index {
+                *existing_node = node_id;
+                set = true;
+            }
+        }
+        if !set {
+            entry.shards.push((shard_index, node_id));
+        }
+
+        let reconfigure_request = entry.maybe_reconfigure(tenant_shard_id.tenant_id).await;
         let Some(reconfigure_request) = reconfigure_request else {
             // The tenant doesn't yet have pageservers for all its shards: we won't notify anything
             // until it does.
@@ -378,85 +316,3 @@ impl ComputeHook {
         }
     }
 }
-
-#[cfg(test)]
-pub(crate) mod tests {
-    use pageserver_api::shard::{ShardCount, ShardNumber};
-    use utils::id::TenantId;
-
-    use super::*;
-
-    #[test]
-    fn tenant_updates() -> anyhow::Result<()> {
-        let tenant_id = TenantId::generate();
-        let mut tenant_state = ComputeHookTenant::new(
-            TenantShardId {
-                tenant_id,
-                shard_count: ShardCount::new(0),
-                shard_number: ShardNumber(0),
-            },
-            ShardStripeSize(12345),
-            NodeId(1),
-        );
-
-        // An unsharded tenant is always ready to emit a notification
-        assert!(tenant_state.maybe_reconfigure(tenant_id).is_some());
-        assert_eq!(
-            tenant_state
-                .maybe_reconfigure(tenant_id)
-                .unwrap()
-                .shards
-                .len(),
-            1
-        );
-        assert!(tenant_state
-            .maybe_reconfigure(tenant_id)
-            .unwrap()
-            .stripe_size
-            .is_none());
-
-        // Writing the first shard of a multi-sharded situation (i.e. in a split)
-        // resets the tenant state and puts it in an non-notifying state (need to
-        // see all shards)
-        tenant_state.update(
-            TenantShardId {
-                tenant_id,
-                shard_count: ShardCount::new(2),
-                shard_number: ShardNumber(1),
-            },
-            ShardStripeSize(32768),
-            NodeId(1),
-        );
-        assert!(tenant_state.maybe_reconfigure(tenant_id).is_none());
-
-        // Writing the second shard makes it ready to notify
-        tenant_state.update(
-            TenantShardId {
-                tenant_id,
-                shard_count: ShardCount::new(2),
-                shard_number: ShardNumber(0),
-            },
-            ShardStripeSize(32768),
-            NodeId(1),
-        );
-
-        assert!(tenant_state.maybe_reconfigure(tenant_id).is_some());
-        assert_eq!(
-            tenant_state
-                .maybe_reconfigure(tenant_id)
-                .unwrap()
-                .shards
-                .len(),
-            2
-        );
-        assert_eq!(
-            tenant_state
-                .maybe_reconfigure(tenant_id)
-                .unwrap()
-                .stripe_size,
-            Some(ShardStripeSize(32768))
-        );
-
-        Ok(())
-    }
-}

control_plane/attachment_service/src/http.rs

@@ -1,5 +1,6 @@
 use crate::reconciler::ReconcileError;
 use crate::service::{Service, STARTUP_RECONCILE_TIMEOUT};
+use crate::PlacementPolicy;
 use hyper::{Body, Request, Response};
 use hyper::{StatusCode, Uri};
 use pageserver_api::models::{
@@ -118,9 +119,13 @@ async fn handle_tenant_create(
 
     let create_req = json_request::<TenantCreateRequest>(&mut req).await?;
 
+    // TODO: enable specifying this.  Using Single as a default helps legacy tests to work (they
+    // have no expectation of HA).
+    let placement_policy = PlacementPolicy::Single;
+
     json_response(
         StatusCode::CREATED,
-        service.tenant_create(create_req).await?,
+        service.tenant_create(create_req, placement_policy).await?,
     )
 }
 

control_plane/attachment_service/src/lib.rs

@@ -1,4 +1,4 @@
-use serde::Serialize;
+use serde::{Deserialize, Serialize};
 use utils::seqwait::MonotonicCounter;
 
 mod auth;
@@ -13,6 +13,23 @@ mod schema;
 pub mod service;
 mod tenant_state;
 
+#[derive(Clone, Serialize, Deserialize, Debug, PartialEq, Eq)]
+enum PlacementPolicy {
+    /// Cheapest way to attach a tenant: just one pageserver, no secondary
+    Single,
+    /// Production-ready way to attach a tenant: one attached pageserver and
+    /// some number of secondaries.
+    Double(usize),
+    /// Create one secondary mode locations. This is useful when onboarding
+    /// a tenant, or for an idle tenant that we might want to bring online quickly.
+    Secondary,
+
+    /// Do not attach to any pageservers.  This is appropriate for tenants that
+    /// have been idle for a long time, where we do not mind some delay in making
+    /// them available in future.
+    Detached,
+}
+
 #[derive(Ord, PartialOrd, Eq, PartialEq, Copy, Clone, Serialize)]
 struct Sequence(u64);
 
@@ -49,3 +66,9 @@ impl Sequence {
         Sequence(self.0 + 1)
     }
 }
+
+impl Default for PlacementPolicy {
+    fn default() -> Self {
+        PlacementPolicy::Double(1)
+    }
+}

control_plane/attachment_service/src/node.rs

@@ -1,16 +1,6 @@
-use std::{str::FromStr, time::Duration};
-
-use hyper::StatusCode;
-use pageserver_api::{
-    controller_api::{
-        NodeAvailability, NodeRegisterRequest, NodeSchedulingPolicy, TenantLocateResponseShard,
-    },
-    shard::TenantShardId,
-};
-use pageserver_client::mgmt_api;
+use pageserver_api::controller_api::{NodeAvailability, NodeSchedulingPolicy};
 use serde::Serialize;
-use tokio_util::sync::CancellationToken;
-use utils::{backoff, id::NodeId};
+use utils::id::NodeId;
 
 use crate::persistence::NodePersistence;
 
@@ -22,29 +12,16 @@ use crate::persistence::NodePersistence;
 /// implementation of serialization on this type is only for debug dumps.
 #[derive(Clone, Serialize)]
 pub(crate) struct Node {
-    id: NodeId,
+    pub(crate) id: NodeId,
 
-    availability: NodeAvailability,
-    scheduling: NodeSchedulingPolicy,
+    pub(crate) availability: NodeAvailability,
+    pub(crate) scheduling: NodeSchedulingPolicy,
 
-    listen_http_addr: String,
-    listen_http_port: u16,
+    pub(crate) listen_http_addr: String,
+    pub(crate) listen_http_port: u16,
 
-    listen_pg_addr: String,
-    listen_pg_port: u16,
-
-    // This cancellation token means "stop any RPCs in flight to this node, and don't start
-    // any more". It is not related to process shutdown.
-    #[serde(skip)]
-    cancel: CancellationToken,
-}
-
-/// When updating [`Node::availability`] we use this type to indicate to the caller
-/// whether/how they changed it.
-pub(crate) enum AvailabilityTransition {
-    ToActive,
-    ToOffline,
-    Unchanged,
+    pub(crate) listen_pg_addr: String,
+    pub(crate) listen_pg_port: u16,
 }
 
 impl Node {
@@ -52,71 +29,6 @@ impl Node {
         format!("http://{}:{}", self.listen_http_addr, self.listen_http_port)
     }
 
-    pub(crate) fn get_id(&self) -> NodeId {
-        self.id
-    }
-
-    pub(crate) fn set_scheduling(&mut self, scheduling: NodeSchedulingPolicy) {
-        self.scheduling = scheduling
-    }
-
-    /// Does this registration request match `self`?  This is used when deciding whether a registration
-    /// request should be allowed to update an existing record with the same node ID.
-    pub(crate) fn registration_match(&self, register_req: &NodeRegisterRequest) -> bool {
-        self.id == register_req.node_id
-            && self.listen_http_addr == register_req.listen_http_addr
-            && self.listen_http_port == register_req.listen_http_port
-            && self.listen_pg_addr == register_req.listen_pg_addr
-            && self.listen_pg_port == register_req.listen_pg_port
-    }
-
-    /// For a shard located on this node, populate a response object
-    /// with this node's address information.
-    pub(crate) fn shard_location(&self, shard_id: TenantShardId) -> TenantLocateResponseShard {
-        TenantLocateResponseShard {
-            shard_id,
-            node_id: self.id,
-            listen_http_addr: self.listen_http_addr.clone(),
-            listen_http_port: self.listen_http_port,
-            listen_pg_addr: self.listen_pg_addr.clone(),
-            listen_pg_port: self.listen_pg_port,
-        }
-    }
-
-    pub(crate) fn set_availability(
-        &mut self,
-        availability: NodeAvailability,
-    ) -> AvailabilityTransition {
-        use NodeAvailability::*;
-        let transition = match (self.availability, availability) {
-            (Offline, Active) => {
-                // Give the node a new cancellation token, effectively resetting it to un-cancelled.  Any
-                // users of previously-cloned copies of the node will still see the old cancellation
-                // state.  For example, Reconcilers in flight will have to complete and be spawned
-                // again to realize that the node has become available.
-                self.cancel = CancellationToken::new();
-                AvailabilityTransition::ToActive
-            }
-            (Active, Offline) => {
-                // Fire the node's cancellation token to cancel any in-flight API requests to it
-                self.cancel.cancel();
-                AvailabilityTransition::ToOffline
-            }
-            _ => AvailabilityTransition::Unchanged,
-        };
-        self.availability = availability;
-        transition
-    }
-
-    /// Whether we may send API requests to this node.
-    pub(crate) fn is_available(&self) -> bool {
-        // When we clone a node, [`Self::availability`] is a snapshot, but [`Self::cancel`] holds
-        // a reference to the original Node's cancellation status.  Checking both of these results
-        // in a "pessimistic" check where we will consider a Node instance unavailable if it was unavailable
-        // when we cloned it, or if the original Node instance's cancellation token was fired.
-        matches!(self.availability, NodeAvailability::Active) && !self.cancel.is_cancelled()
-    }
-
     /// Is this node elegible to have work scheduled onto it?
     pub(crate) fn may_schedule(&self) -> bool {
         match self.availability {
@@ -132,26 +44,6 @@ impl Node {
         }
     }
 
-    pub(crate) fn new(
-        id: NodeId,
-        listen_http_addr: String,
-        listen_http_port: u16,
-        listen_pg_addr: String,
-        listen_pg_port: u16,
-    ) -> Self {
-        Self {
-            id,
-            listen_http_addr,
-            listen_http_port,
-            listen_pg_addr,
-            listen_pg_port,
-            scheduling: NodeSchedulingPolicy::Filling,
-            // TODO: we shouldn't really call this Active until we've heartbeated it.
-            availability: NodeAvailability::Active,
-            cancel: CancellationToken::new(),
-        }
-    }
-
     pub(crate) fn to_persistent(&self) -> NodePersistence {
         NodePersistence {
             node_id: self.id.0 as i64,
@@ -162,96 +54,4 @@ impl Node {
             listen_pg_port: self.listen_pg_port as i32,
         }
     }
-
-    pub(crate) fn from_persistent(np: NodePersistence) -> Self {
-        Self {
-            id: NodeId(np.node_id as u64),
-            // At startup we consider a node offline until proven otherwise.
-            availability: NodeAvailability::Offline,
-            scheduling: NodeSchedulingPolicy::from_str(&np.scheduling_policy)
-                .expect("Bad scheduling policy in DB"),
-            listen_http_addr: np.listen_http_addr,
-            listen_http_port: np.listen_http_port as u16,
-            listen_pg_addr: np.listen_pg_addr,
-            listen_pg_port: np.listen_pg_port as u16,
-            cancel: CancellationToken::new(),
-        }
-    }
-
-    /// Wrapper for issuing requests to pageserver management API: takes care of generic
-    /// retry/backoff for retryable HTTP status codes.
-    ///
-    /// This will return None to indicate cancellation.  Cancellation may happen from
-    /// the cancellation token passed in, or from Self's cancellation token (i.e. node
-    /// going offline).
-    pub(crate) async fn with_client_retries<T, O, F>(
-        &self,
-        mut op: O,
-        jwt: &Option<String>,
-        warn_threshold: u32,
-        max_retries: u32,
-        timeout: Duration,
-        cancel: &CancellationToken,
-    ) -> Option<mgmt_api::Result<T>>
-    where
-        O: FnMut(mgmt_api::Client) -> F,
-        F: std::future::Future<Output = mgmt_api::Result<T>>,
-    {
-        fn is_fatal(e: &mgmt_api::Error) -> bool {
-            use mgmt_api::Error::*;
-            match e {
-                ReceiveBody(_) | ReceiveErrorBody(_) => false,
-                ApiError(StatusCode::SERVICE_UNAVAILABLE, _)
-                | ApiError(StatusCode::GATEWAY_TIMEOUT, _)
-                | ApiError(StatusCode::REQUEST_TIMEOUT, _) => false,
-                ApiError(_, _) => true,
-                Cancelled => true,
-            }
-        }
-
-        backoff::retry(
-            || {
-                let http_client = reqwest::ClientBuilder::new()
-                    .timeout(timeout)
-                    .build()
-                    .expect("Failed to construct HTTP client");
-
-                let client =
-                    mgmt_api::Client::from_client(http_client, self.base_url(), jwt.as_deref());
-
-                let node_cancel_fut = self.cancel.cancelled();
-
-                let op_fut = op(client);
-
-                async {
-                    tokio::select! {
-                        r = op_fut=> {r},
-                        _ = node_cancel_fut => {
-                        Err(mgmt_api::Error::Cancelled)
-                    }}
-                }
-            },
-            is_fatal,
-            warn_threshold,
-            max_retries,
-            &format!(
-                "Call to node {} ({}:{}) management API",
-                self.id, self.listen_http_addr, self.listen_http_port
-            ),
-            cancel,
-        )
-        .await
-    }
-}
-
-impl std::fmt::Display for Node {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        write!(f, "{} ({})", self.id, self.listen_http_addr)
-    }
-}
-
-impl std::fmt::Debug for Node {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        write!(f, "{} ({})", self.id, self.listen_http_addr)
-    }
 }

control_plane/attachment_service/src/persistence.rs

@@ -7,9 +7,11 @@ use self::split_state::SplitState;
 use camino::Utf8Path;
 use camino::Utf8PathBuf;
 use diesel::pg::PgConnection;
-use diesel::prelude::*;
-use diesel::Connection;
-use pageserver_api::controller_api::{NodeSchedulingPolicy, PlacementPolicy};
+use diesel::{
+    Connection, ExpressionMethods, Insertable, QueryDsl, QueryResult, Queryable, RunQueryDsl,
+    Selectable, SelectableHelper,
+};
+use pageserver_api::controller_api::NodeSchedulingPolicy;
 use pageserver_api::models::TenantConfig;
 use pageserver_api::shard::{ShardCount, ShardNumber, TenantShardId};
 use serde::{Deserialize, Serialize};
@@ -17,6 +19,7 @@ use utils::generation::Generation;
 use utils::id::{NodeId, TenantId};
 
 use crate::node::Node;
+use crate::PlacementPolicy;
 
 /// ## What do we store?
 ///
@@ -207,7 +210,7 @@ impl Persistence {
                 tenant.tenant_id = tenant_id.to_string();
                 tenant.config = serde_json::to_string(&TenantConfig::default())
                     .map_err(|e| DatabaseError::Logical(format!("Serialization error: {e}")))?;
-                tenant.placement_policy = serde_json::to_string(&PlacementPolicy::Single)
+                tenant.placement_policy = serde_json::to_string(&PlacementPolicy::default())
                     .map_err(|e| DatabaseError::Logical(format!("Serialization error: {e}")))?;
             }
         }

control_plane/attachment_service/src/reconciler.rs

@@ -1,5 +1,6 @@
 use crate::persistence::Persistence;
 use crate::service;
+use pageserver_api::controller_api::NodeAvailability;
 use pageserver_api::models::{
     LocationConfig, LocationConfigMode, LocationConfigSecondary, TenantConfig,
 };
@@ -27,16 +28,15 @@ pub(super) struct Reconciler {
     pub(crate) shard: ShardIdentity,
     pub(crate) generation: Option<Generation>,
     pub(crate) intent: TargetState,
-
-    /// Nodes not referenced by [`Self::intent`], from which we should try
-    /// to detach this tenant shard.
-    pub(crate) detach: Vec<Node>,
-
     pub(crate) config: TenantConfig,
     pub(crate) observed: ObservedState,
 
     pub(crate) service_config: service::Config,
 
+    /// A snapshot of the pageservers as they were when we were asked
+    /// to reconcile.
+    pub(crate) pageservers: Arc<HashMap<NodeId, Node>>,
+
     /// A hook to notify the running postgres instances when we change the location
     /// of a tenant.  Use this via [`Self::compute_notify`] to update our failure flag
     /// and guarantee eventual retries.
@@ -67,37 +67,29 @@ pub(super) struct Reconciler {
 /// and the TargetState is just the instruction for a particular Reconciler run.
 #[derive(Debug)]
 pub(crate) struct TargetState {
-    pub(crate) attached: Option<Node>,
-    pub(crate) secondary: Vec<Node>,
+    pub(crate) attached: Option<NodeId>,
+    pub(crate) secondary: Vec<NodeId>,
 }
 
 impl TargetState {
-    pub(crate) fn from_intent(nodes: &HashMap<NodeId, Node>, intent: &IntentState) -> Self {
+    pub(crate) fn from_intent(intent: &IntentState) -> Self {
         Self {
-            attached: intent.get_attached().map(|n| {
-                nodes
-                    .get(&n)
-                    .expect("Intent attached referenced non-existent node")
-                    .clone()
-            }),
-            secondary: intent
-                .get_secondary()
-                .iter()
-                .map(|n| {
-                    nodes
-                        .get(n)
-                        .expect("Intent secondary referenced non-existent node")
-                        .clone()
-                })
-                .collect(),
+            attached: *intent.get_attached(),
+            secondary: intent.get_secondary().clone(),
         }
     }
+
+    fn all_pageservers(&self) -> Vec<NodeId> {
+        let mut result = self.secondary.clone();
+        if let Some(node_id) = &self.attached {
+            result.push(*node_id);
+        }
+        result
+    }
 }
 
 #[derive(thiserror::Error, Debug)]
 pub(crate) enum ReconcileError {
-    #[error(transparent)]
-    Remote(#[from] mgmt_api::Error),
     #[error(transparent)]
     Notify(#[from] NotifyError),
     #[error("Cancelled")]
@@ -109,83 +101,44 @@ pub(crate) enum ReconcileError {
 impl Reconciler {
     async fn location_config(
         &mut self,
-        node: &Node,
+        node_id: NodeId,
         config: LocationConfig,
         flush_ms: Option<Duration>,
-        lazy: bool,
-    ) -> Result<(), ReconcileError> {
-        self.observed
-            .locations
-            .insert(node.get_id(), ObservedStateLocation { conf: None });
-
-        // TODO: amend locations that use long-polling: they will hit this timeout.
-        let timeout = Duration::from_secs(25);
-
-        tracing::info!("location_config({node}) calling: {:?}", config);
-        let tenant_shard_id = self.tenant_shard_id;
-        let config_ref = &config;
-        match node
-            .with_client_retries(
-                |client| async move {
-                    let config = config_ref.clone();
-                    client
-                        .location_config(tenant_shard_id, config.clone(), flush_ms, lazy)
-                        .await
-                },
-                &self.service_config.jwt_token,
-                1,
-                3,
-                timeout,
-                &self.cancel,
-            )
-            .await
-        {
-            Some(Ok(_)) => {}
-            Some(Err(e)) => return Err(e.into()),
-            None => return Err(ReconcileError::Cancel),
-        };
-        tracing::info!("location_config({node}) complete: {:?}", config);
+    ) -> anyhow::Result<()> {
+        let node = self
+            .pageservers
+            .get(&node_id)
+            .expect("Pageserver may not be removed while referenced");
 
         self.observed
             .locations
-            .insert(node.get_id(), ObservedStateLocation { conf: Some(config) });
+            .insert(node.id, ObservedStateLocation { conf: None });
+
+        tracing::info!("location_config({}) calling: {:?}", node_id, config);
+        let client =
+            mgmt_api::Client::new(node.base_url(), self.service_config.jwt_token.as_deref());
+        client
+            .location_config(self.tenant_shard_id, config.clone(), flush_ms)
+            .await?;
+        tracing::info!("location_config({}) complete: {:?}", node_id, config);
+
+        self.observed
+            .locations
+            .insert(node.id, ObservedStateLocation { conf: Some(config) });
 
         Ok(())
     }
 
-    fn get_node(&self, node_id: &NodeId) -> Option<&Node> {
-        if let Some(node) = self.intent.attached.as_ref() {
-            if node.get_id() == *node_id {
-                return Some(node);
-            }
-        }
-
-        if let Some(node) = self
-            .intent
-            .secondary
-            .iter()
-            .find(|n| n.get_id() == *node_id)
-        {
-            return Some(node);
-        }
-
-        if let Some(node) = self.detach.iter().find(|n| n.get_id() == *node_id) {
-            return Some(node);
-        }
-
-        None
-    }
-
     async fn maybe_live_migrate(&mut self) -> Result<(), ReconcileError> {
-        let destination = if let Some(node) = &self.intent.attached {
-            match self.observed.locations.get(&node.get_id()) {
+        let destination = if let Some(node_id) = self.intent.attached {
+            match self.observed.locations.get(&node_id) {
                 Some(conf) => {
                     // We will do a live migration only if the intended destination is not
                     // currently in an attached state.
                     match &conf.conf {
                         Some(conf) if conf.mode == LocationConfigMode::Secondary => {
                             // Fall through to do a live migration
-                            node
+                            node_id
                         }
                         None | Some(_) => {
                             // Attached or uncertain: don't do a live migration, proceed
@@ -198,7 +151,7 @@ impl Reconciler {
                 None => {
                     // Our destination is not attached: maybe live migrate if some other
                     // node is currently attached.  Fall through.
-                    node
+                    node_id
                 }
             }
         } else {
@@ -211,13 +164,15 @@ impl Reconciler {
         for (node_id, state) in &self.observed.locations {
             if let Some(observed_conf) = &state.conf {
                 if observed_conf.mode == LocationConfigMode::AttachedSingle {
+                    let node = self
+                        .pageservers
+                        .get(node_id)
+                        .expect("Nodes may not be removed while referenced");
                     // We will only attempt live migration if the origin is not offline: this
                     // avoids trying to do it while reconciling after responding to an HA failover.
-                    if let Some(node) = self.get_node(node_id) {
-                        if node.is_available() {
-                            origin = Some(node.clone());
-                            break;
-                        }
+                    if !matches!(node.availability, NodeAvailability::Offline) {
+                        origin = Some(*node_id);
+                        break;
                     }
                 }
             }
@@ -230,7 +185,7 @@ impl Reconciler {
 
         // We have an origin and a destination: proceed to do the live migration
         tracing::info!("Live migrating {}->{}", origin, destination);
-        self.live_migrate(origin, destination.clone()).await?;
+        self.live_migrate(origin, destination).await?;
 
         Ok(())
     }
@@ -238,8 +193,13 @@ impl Reconciler {
     async fn get_lsns(
         &self,
         tenant_shard_id: TenantShardId,
-        node: &Node,
+        node_id: &NodeId,
     ) -> anyhow::Result<HashMap<TimelineId, Lsn>> {
+        let node = self
+            .pageservers
+            .get(node_id)
+            .expect("Pageserver may not be removed while referenced");
+
         let client =
             mgmt_api::Client::new(node.base_url(), self.service_config.jwt_token.as_deref());
 
@@ -250,27 +210,19 @@ impl Reconciler {
             .collect())
     }
 
-    async fn secondary_download(
-        &self,
-        tenant_shard_id: TenantShardId,
-        node: &Node,
-    ) -> Result<(), ReconcileError> {
-        match node
-            .with_client_retries(
-                |client| async move { client.tenant_secondary_download(tenant_shard_id).await },
-                &self.service_config.jwt_token,
-                1,
-                1,
-                Duration::from_secs(60),
-                &self.cancel,
-            )
-            .await
-        {
-            None => Err(ReconcileError::Cancel),
-            Some(Ok(_)) => Ok(()),
-            Some(Err(e)) => {
-                tracing::info!("  (skipping destination download: {})", e);
-                Ok(())
+    async fn secondary_download(&self, tenant_shard_id: TenantShardId, node_id: &NodeId) {
+        let node = self
+            .pageservers
+            .get(node_id)
+            .expect("Pageserver may not be removed while referenced");
+
+        let client =
+            mgmt_api::Client::new(node.base_url(), self.service_config.jwt_token.as_deref());
+
+        match client.tenant_secondary_download(tenant_shard_id).await {
+            Ok(()) => {}
+            Err(_) => {
+                tracing::info!("  (skipping, destination wasn't in secondary mode)")
             }
         }
     }
@@ -278,14 +230,17 @@ impl Reconciler {
     async fn await_lsn(
         &self,
         tenant_shard_id: TenantShardId,
-        node: &Node,
+        pageserver_id: &NodeId,
         baseline: HashMap<TimelineId, Lsn>,
     ) -> anyhow::Result<()> {
         loop {
-            let latest = match self.get_lsns(tenant_shard_id, node).await {
+            let latest = match self.get_lsns(tenant_shard_id, pageserver_id).await {
                 Ok(l) => l,
                 Err(e) => {
-                    tracing::info!("🕑 Can't get LSNs on node {node} yet, waiting ({e})",);
+                    println!(
+                        "🕑 Can't get LSNs on pageserver {} yet, waiting ({e})",
+                        pageserver_id
+                    );
                     std::thread::sleep(Duration::from_millis(500));
                     continue;
                 }
@@ -295,7 +250,7 @@ impl Reconciler {
             for (timeline_id, baseline_lsn) in &baseline {
                 match latest.get(timeline_id) {
                     Some(latest_lsn) => {
-                        tracing::info!("🕑 LSN origin {baseline_lsn} vs destination {latest_lsn}");
+                        println!("🕑 LSN origin {baseline_lsn} vs destination {latest_lsn}");
                         if latest_lsn < baseline_lsn {
                             any_behind = true;
                         }
@@ -310,7 +265,7 @@ impl Reconciler {
             }
 
             if !any_behind {
-                tracing::info!("✅ LSN caught up.  Proceeding...");
+                println!("✅ LSN caught up.  Proceeding...");
                 break;
             } else {
                 std::thread::sleep(Duration::from_millis(500));
@@ -322,11 +277,11 @@ impl Reconciler {
 
     pub async fn live_migrate(
         &mut self,
-        origin_ps: Node,
-        dest_ps: Node,
-    ) -> Result<(), ReconcileError> {
+        origin_ps_id: NodeId,
+        dest_ps_id: NodeId,
+    ) -> anyhow::Result<()> {
         // `maybe_live_migrate` is responsibble for sanity of inputs
-        assert!(origin_ps.get_id() != dest_ps.get_id());
+        assert!(origin_ps_id != dest_ps_id);
 
         fn build_location_config(
             shard: &ShardIdentity,
@@ -346,7 +301,10 @@ impl Reconciler {
             }
         }
 
-        tracing::info!("🔁 Switching origin node {origin_ps} to stale mode",);
+        tracing::info!(
+            "🔁 Switching origin pageserver {} to stale mode",
+            origin_ps_id
+        );
 
         // FIXME: it is incorrect to use self.generation here, we should use the generation
         // from the ObservedState of the origin pageserver (it might be older than self.generation)
@@ -357,18 +315,21 @@ impl Reconciler {
             self.generation,
             None,
         );
-        self.location_config(&origin_ps, stale_conf, Some(Duration::from_secs(10)), false)
+        self.location_config(origin_ps_id, stale_conf, Some(Duration::from_secs(10)))
             .await?;
 
-        let baseline_lsns = Some(self.get_lsns(self.tenant_shard_id, &origin_ps).await?);
+        let baseline_lsns = Some(self.get_lsns(self.tenant_shard_id, &origin_ps_id).await?);
 
         // If we are migrating to a destination that has a secondary location, warm it up first
-        if let Some(destination_conf) = self.observed.locations.get(&dest_ps.get_id()) {
+        if let Some(destination_conf) = self.observed.locations.get(&dest_ps_id) {
             if let Some(destination_conf) = &destination_conf.conf {
                 if destination_conf.mode == LocationConfigMode::Secondary {
-                    tracing::info!("🔁 Downloading latest layers to destination node {dest_ps}",);
-                    self.secondary_download(self.tenant_shard_id, &dest_ps)
-                        .await?;
+                    tracing::info!(
+                        "🔁 Downloading latest layers to destination pageserver {}",
+                        dest_ps_id,
+                    );
+                    self.secondary_download(self.tenant_shard_id, &dest_ps_id)
+                        .await;
                 }
             }
         }
@@ -376,7 +337,7 @@ impl Reconciler {
         // Increment generation before attaching to new pageserver
         self.generation = Some(
             self.persistence
-                .increment_generation(self.tenant_shard_id, dest_ps.get_id())
+                .increment_generation(self.tenant_shard_id, dest_ps_id)
                 .await?,
         );
 
@@ -388,23 +349,22 @@ impl Reconciler {
             None,
         );
 
-        tracing::info!("🔁 Attaching to pageserver {dest_ps}");
-        self.location_config(&dest_ps, dest_conf, None, false)
-            .await?;
+        tracing::info!("🔁 Attaching to pageserver {}", dest_ps_id);
+        self.location_config(dest_ps_id, dest_conf, None).await?;
 
         if let Some(baseline) = baseline_lsns {
             tracing::info!("🕑 Waiting for LSN to catch up...");
-            self.await_lsn(self.tenant_shard_id, &dest_ps, baseline)
+            self.await_lsn(self.tenant_shard_id, &dest_ps_id, baseline)
                 .await?;
         }
 
-        tracing::info!("🔁 Notifying compute to use pageserver {dest_ps}");
+        tracing::info!("🔁 Notifying compute to use pageserver {}", dest_ps_id);
 
         // During a live migration it is unhelpful to proceed if we couldn't notify compute: if we detach
         // the origin without notifying compute, we will render the tenant unavailable.
         while let Err(e) = self.compute_notify().await {
             match e {
-                NotifyError::Fatal(_) => return Err(ReconcileError::Notify(e)),
+                NotifyError::Fatal(_) => return Err(anyhow::anyhow!(e)),
                 _ => {
                     tracing::warn!(
                         "Live migration blocked by compute notification error, retrying: {e}"
@@ -422,19 +382,22 @@ impl Reconciler {
             None,
             Some(LocationConfigSecondary { warm: true }),
         );
-        self.location_config(&origin_ps, origin_secondary_conf.clone(), None, false)
+        self.location_config(origin_ps_id, origin_secondary_conf.clone(), None)
             .await?;
         // TODO: we should also be setting the ObservedState on earlier API calls, in case we fail
         // partway through.  In fact, all location conf API calls should be in a wrapper that sets
         // the observed state to None, then runs, then sets it to what we wrote.
         self.observed.locations.insert(
-            origin_ps.get_id(),
+            origin_ps_id,
             ObservedStateLocation {
                 conf: Some(origin_secondary_conf),
             },
         );
 
-        tracing::info!("🔁 Switching to AttachedSingle mode on node {dest_ps}",);
+        println!(
+            "🔁 Switching to AttachedSingle mode on pageserver {}",
+            dest_ps_id
+        );
         let dest_final_conf = build_location_config(
             &self.shard,
             &self.config,
@@ -442,61 +405,16 @@ impl Reconciler {
             self.generation,
             None,
         );
-        self.location_config(&dest_ps, dest_final_conf.clone(), None, false)
+        self.location_config(dest_ps_id, dest_final_conf.clone(), None)
             .await?;
         self.observed.locations.insert(
-            dest_ps.get_id(),
+            dest_ps_id,
             ObservedStateLocation {
                 conf: Some(dest_final_conf),
             },
         );
 
-        tracing::info!("✅ Migration complete");
-
-        Ok(())
-    }
-
-    async fn maybe_refresh_observed(&mut self) -> Result<(), ReconcileError> {
-        // If the attached node has uncertain state, read it from the pageserver before proceeding: this
-        // is important to avoid spurious generation increments.
-        //
-        // We don't need to do this for secondary/detach locations because it's harmless to just PUT their
-        // location conf, whereas for attached locations it can interrupt clients if we spuriously destroy/recreate
-        // the `Timeline` object in the pageserver.
-
-        let Some(attached_node) = self.intent.attached.as_ref() else {
-            // Nothing to do
-            return Ok(());
-        };
-
-        if matches!(
-            self.observed.locations.get(&attached_node.get_id()),
-            Some(ObservedStateLocation { conf: None })
-        ) {
-            let tenant_shard_id = self.tenant_shard_id;
-            let observed_conf = match attached_node
-                .with_client_retries(
-                    |client| async move { client.get_location_config(tenant_shard_id).await },
-                    &self.service_config.jwt_token,
-                    1,
-                    1,
-                    Duration::from_secs(5),
-                    &self.cancel,
-                )
-                .await
-            {
-                Some(Ok(observed)) => observed,
-                Some(Err(e)) => return Err(e.into()),
-                None => return Err(ReconcileError::Cancel),
-            };
-            tracing::info!("Scanned location configuration on {attached_node}: {observed_conf:?}");
-            self.observed.locations.insert(
-                attached_node.get_id(),
-                ObservedStateLocation {
-                    conf: observed_conf,
-                },
-            );
-        }
+        println!("✅ Migration complete");
 
         Ok(())
     }
@@ -508,14 +426,14 @@ impl Reconciler {
     /// general case reconciliation where we walk through the intent by pageserver
     /// and call out to the pageserver to apply the desired state.
     pub(crate) async fn reconcile(&mut self) -> Result<(), ReconcileError> {
-        // Prepare: if we have uncertain `observed` state for our would-be attachement location, then refresh it
-        self.maybe_refresh_observed().await?;
+        // TODO: if any of self.observed is None, call to remote pageservers
+        // to learn correct state.
 
         // Special case: live migration
         self.maybe_live_migrate().await?;
 
         // If the attached pageserver is not attached, do so now.
-        if let Some(node) = self.intent.attached.as_ref() {
+        if let Some(node_id) = self.intent.attached {
             // If we are in an attached policy, then generation must have been set (null generations
             // are only present when a tenant is initially loaded with a secondary policy)
             debug_assert!(self.generation.is_some());
@@ -526,10 +444,10 @@ impl Reconciler {
             };
 
             let mut wanted_conf = attached_location_conf(generation, &self.shard, &self.config);
-            match self.observed.locations.get(&node.get_id()) {
+            match self.observed.locations.get(&node_id) {
                 Some(conf) if conf.conf.as_ref() == Some(&wanted_conf) => {
                     // Nothing to do
-                    tracing::info!(node_id=%node.get_id(), "Observed configuration already correct.")
+                    tracing::info!(%node_id, "Observed configuration already correct.")
                 }
                 observed => {
                     // In all cases other than a matching observed configuration, we will
@@ -567,21 +485,13 @@ impl Reconciler {
                     if increment_generation {
                         let generation = self
                             .persistence
-                            .increment_generation(self.tenant_shard_id, node.get_id())
+                            .increment_generation(self.tenant_shard_id, node_id)
                             .await?;
                         self.generation = Some(generation);
                         wanted_conf.generation = generation.into();
                     }
-                    tracing::info!(node_id=%node.get_id(), "Observed configuration requires update.");
-
-                    // Because `node` comes from a ref to &self, clone it before calling into a &mut self
-                    // function: this could be avoided by refactoring the state mutated by location_config into
-                    // a separate type to Self.
-                    let node = node.clone();
-
-                    // Use lazy=true, because we may run many of Self concurrently, and do not want to
-                    // overload the pageserver with logical size calculations.
-                    self.location_config(&node, wanted_conf, None, true).await?;
+                    tracing::info!(%node_id, "Observed configuration requires update.");
+                    self.location_config(node_id, wanted_conf, None).await?;
                     self.compute_notify().await?;
                 }
             }
@@ -590,27 +500,33 @@ impl Reconciler {
         // Configure secondary locations: if these were previously attached this
         // implicitly downgrades them from attached to secondary.
         let mut changes = Vec::new();
-        for node in &self.intent.secondary {
+        for node_id in &self.intent.secondary {
             let wanted_conf = secondary_location_conf(&self.shard, &self.config);
-            match self.observed.locations.get(&node.get_id()) {
+            match self.observed.locations.get(node_id) {
                 Some(conf) if conf.conf.as_ref() == Some(&wanted_conf) => {
                     // Nothing to do
-                    tracing::info!(node_id=%node.get_id(), "Observed configuration already correct.")
+                    tracing::info!(%node_id, "Observed configuration already correct.")
                 }
                 _ => {
                     // In all cases other than a matching observed configuration, we will
                     // reconcile this location.
-                    tracing::info!(node_id=%node.get_id(), "Observed configuration requires update.");
-                    changes.push((node.clone(), wanted_conf))
+                    tracing::info!(%node_id, "Observed configuration requires update.");
+                    changes.push((*node_id, wanted_conf))
                 }
             }
         }
 
         // Detach any extraneous pageservers that are no longer referenced
         // by our intent.
-        for node in &self.detach {
+        let all_pageservers = self.intent.all_pageservers();
+        for node_id in self.observed.locations.keys() {
+            if all_pageservers.contains(node_id) {
+                // We are only detaching pageservers that aren't used at all.
+                continue;
+            }
+
             changes.push((
-                node.clone(),
+                *node_id,
                 LocationConfig {
                     mode: LocationConfigMode::Detached,
                     generation: None,
@@ -623,11 +539,11 @@ impl Reconciler {
             ));
         }
 
-        for (node, conf) in changes {
+        for (node_id, conf) in changes {
             if self.cancel.is_cancelled() {
                 return Err(ReconcileError::Cancel);
             }
-            self.location_config(&node, conf, None, false).await?;
+            self.location_config(node_id, conf, None).await?;
         }
 
         Ok(())
@@ -636,21 +552,16 @@ impl Reconciler {
     pub(crate) async fn compute_notify(&mut self) -> Result<(), NotifyError> {
         // Whenever a particular Reconciler emits a notification, it is always notifying for the intended
         // destination.
-        if let Some(node) = &self.intent.attached {
+        if let Some(node_id) = self.intent.attached {
             let result = self
                 .compute_hook
-                .notify(
-                    self.tenant_shard_id,
-                    node.get_id(),
-                    self.shard.stripe_size,
-                    &self.cancel,
-                )
+                .notify(self.tenant_shard_id, node_id, &self.cancel)
                 .await;
             if let Err(e) = &result {
                 // It is up to the caller whether they want to drop out on this error, but they don't have to:
                 // in general we should avoid letting unavailability of the cloud control plane stop us from
                 // making progress.
-                tracing::warn!("Failed to notify compute of attached pageserver {node}: {e}");
+                tracing::warn!("Failed to notify compute of attached pageserver {node_id}: {e}");
                 // Set this flag so that in our ReconcileResult we will set the flag on the shard that it
                 // needs to retry at some point.
                 self.compute_notify_failure = true;

control_plane/attachment_service/src/scheduler.rs

@@ -43,7 +43,7 @@ impl Scheduler {
         let mut scheduler_nodes = HashMap::new();
         for node in nodes {
             scheduler_nodes.insert(
-                node.get_id(),
+                node.id,
                 SchedulerNode {
                     shard_count: 0,
                     may_schedule: node.may_schedule(),
@@ -68,7 +68,7 @@ impl Scheduler {
         let mut expect_nodes: HashMap<NodeId, SchedulerNode> = HashMap::new();
         for node in nodes {
             expect_nodes.insert(
-                node.get_id(),
+                node.id,
                 SchedulerNode {
                     shard_count: 0,
                     may_schedule: node.may_schedule(),
@@ -156,7 +156,7 @@ impl Scheduler {
 
     pub(crate) fn node_upsert(&mut self, node: &Node) {
         use std::collections::hash_map::Entry::*;
-        match self.nodes.entry(node.get_id()) {
+        match self.nodes.entry(node.id) {
             Occupied(mut entry) => {
                 entry.get_mut().may_schedule = node.may_schedule();
             }
@@ -255,6 +255,7 @@ impl Scheduler {
 pub(crate) mod test_utils {
 
     use crate::node::Node;
+    use pageserver_api::controller_api::{NodeAvailability, NodeSchedulingPolicy};
     use std::collections::HashMap;
     use utils::id::NodeId;
     /// Test helper: synthesize the requested number of nodes, all in active state.
@@ -263,17 +264,18 @@ pub(crate) mod test_utils {
     pub(crate) fn make_test_nodes(n: u64) -> HashMap<NodeId, Node> {
         (1..n + 1)
             .map(|i| {
-                (NodeId(i), {
-                    let node = Node::new(
-                        NodeId(i),
-                        format!("httphost-{i}"),
-                        80 + i as u16,
-                        format!("pghost-{i}"),
-                        5432 + i as u16,
-                    );
-                    assert!(node.is_available());
-                    node
-                })
+                (
+                    NodeId(i),
+                    Node {
+                        id: NodeId(i),
+                        availability: NodeAvailability::Active,
+                        scheduling: NodeSchedulingPolicy::Active,
+                        listen_http_addr: format!("httphost-{i}"),
+                        listen_http_port: 80 + i as u16,
+                        listen_pg_addr: format!("pghost-{i}"),
+                        listen_pg_port: 5432 + i as u16,
+                    },
+                )
             })
             .collect()
     }

control_plane/attachment_service/src/service.rs

@@ -16,9 +16,9 @@ use futures::{stream::FuturesUnordered, StreamExt};
 use hyper::StatusCode;
 use pageserver_api::{
     controller_api::{
-        NodeAvailability, NodeConfigureRequest, NodeRegisterRequest, PlacementPolicy,
+        NodeAvailability, NodeConfigureRequest, NodeRegisterRequest, NodeSchedulingPolicy,
         TenantCreateResponse, TenantCreateResponseShard, TenantLocateResponse,
-        TenantShardMigrateRequest, TenantShardMigrateResponse,
+        TenantLocateResponseShard, TenantShardMigrateRequest, TenantShardMigrateResponse,
     },
     models::TenantConfigRequest,
 };
@@ -39,6 +39,7 @@ use pageserver_client::mgmt_api;
 use tokio_util::sync::CancellationToken;
 use tracing::instrument;
 use utils::{
+    backoff,
     completion::Barrier,
     generation::Generation,
     http::error::ApiError,
@@ -49,7 +50,7 @@ use utils::{
 
 use crate::{
     compute_hook::{self, ComputeHook},
-    node::{AvailabilityTransition, Node},
+    node::Node,
     persistence::{split_state::SplitState, DatabaseError, Persistence, TenantShardPersistence},
     reconciler::attached_location_conf,
     scheduler::Scheduler,
@@ -57,7 +58,7 @@ use crate::{
         IntentState, ObservedState, ObservedStateLocation, ReconcileResult, ReconcileWaitError,
         ReconcilerWaiter, TenantState,
     },
-    Sequence,
+    PlacementPolicy, Sequence,
 };
 
 // For operations that should be quick, like attaching a new tenant
@@ -176,7 +177,7 @@ impl From<ReconcileWaitError> for ApiError {
 
 #[allow(clippy::large_enum_variant)]
 enum TenantCreateOrUpdate {
-    Create(TenantCreateRequest),
+    Create((TenantCreateRequest, PlacementPolicy)),
     Update(Vec<ShardUpdate>),
 }
 
@@ -200,8 +201,7 @@ impl Service {
     async fn startup_reconcile(self: &Arc<Service>) {
         // For all tenant shards, a vector of observed states on nodes (where None means
         // indeterminate, same as in [`ObservedStateLocation`])
-        let mut observed: HashMap<TenantShardId, Vec<(NodeId, Option<LocationConfig>)>> =
-            HashMap::new();
+        let mut observed = HashMap::new();
 
         let mut nodes_online = HashSet::new();
 
@@ -236,8 +236,7 @@ impl Service {
             nodes_online.insert(node_id);
 
             for (tenant_shard_id, conf_opt) in tenant_shards {
-                let shard_observations = observed.entry(tenant_shard_id).or_default();
-                shard_observations.push((node_id, conf_opt));
+                observed.insert(tenant_shard_id, (node_id, conf_opt));
             }
         }
 
@@ -253,28 +252,27 @@ impl Service {
             let mut new_nodes = (**nodes).clone();
             for (node_id, node) in new_nodes.iter_mut() {
                 if nodes_online.contains(node_id) {
-                    node.set_availability(NodeAvailability::Active);
+                    node.availability = NodeAvailability::Active;
                     scheduler.node_upsert(node);
                 }
             }
             *nodes = Arc::new(new_nodes);
 
-            for (tenant_shard_id, shard_observations) in observed {
-                for (node_id, observed_loc) in shard_observations {
-                    let Some(tenant_state) = tenants.get_mut(&tenant_shard_id) else {
-                        cleanup.push((tenant_shard_id, node_id));
-                        continue;
-                    };
-                    tenant_state
-                        .observed
-                        .locations
-                        .insert(node_id, ObservedStateLocation { conf: observed_loc });
-                }
+            for (tenant_shard_id, (node_id, observed_loc)) in observed {
+                let Some(tenant_state) = tenants.get_mut(&tenant_shard_id) else {
+                    cleanup.push((tenant_shard_id, node_id));
+                    continue;
+                };
+
+                tenant_state
+                    .observed
+                    .locations
+                    .insert(node_id, ObservedStateLocation { conf: observed_loc });
             }
 
             // Populate each tenant's intent state
             for (tenant_shard_id, tenant_state) in tenants.iter_mut() {
-                tenant_state.intent_from_observed(scheduler);
+                tenant_state.intent_from_observed();
                 if let Err(e) = tenant_state.schedule(scheduler) {
                     // Non-fatal error: we are unable to properly schedule the tenant, perhaps because
                     // not enough pageservers are available.  The tenant may well still be available
@@ -285,11 +283,7 @@ impl Service {
                     // emit a compute notification for this. In the case where our observed state does not
                     // yet match our intent, we will eventually reconcile, and that will emit a compute notification.
                     if let Some(attached_at) = tenant_state.stably_attached() {
-                        compute_notifications.push((
-                            *tenant_shard_id,
-                            attached_at,
-                            tenant_state.shard.stripe_size,
-                        ));
+                        compute_notifications.push((*tenant_shard_id, attached_at));
                     }
                 }
             }
@@ -361,19 +355,40 @@ impl Service {
         for node in nodes.values() {
             node_list_futs.push({
                 async move {
-                    tracing::info!("Scanning shards on node {node}...");
-                    let timeout = Duration::from_secs(5);
-                    let response = node
-                        .with_client_retries(
-                            |client| async move { client.list_location_config().await },
-                            &self.config.jwt_token,
-                            1,
-                            5,
-                            timeout,
-                            &self.cancel,
-                        )
-                        .await;
-                    (node.get_id(), response)
+                    let http_client = reqwest::ClientBuilder::new()
+                        .timeout(Duration::from_secs(5))
+                        .build()
+                        .expect("Failed to construct HTTP client");
+                    let client = mgmt_api::Client::from_client(
+                        http_client,
+                        node.base_url(),
+                        self.config.jwt_token.as_deref(),
+                    );
+
+                    fn is_fatal(e: &mgmt_api::Error) -> bool {
+                        use mgmt_api::Error::*;
+                        match e {
+                            ReceiveBody(_) | ReceiveErrorBody(_) => false,
+                            ApiError(StatusCode::SERVICE_UNAVAILABLE, _)
+                            | ApiError(StatusCode::GATEWAY_TIMEOUT, _)
+                            | ApiError(StatusCode::REQUEST_TIMEOUT, _) => false,
+                            ApiError(_, _) => true,
+                        }
+                    }
+
+                    tracing::info!("Scanning shards on node {}...", node.id);
+                    let description = format!("List locations on {}", node.id);
+                    let response = backoff::retry(
+                        || client.list_location_config(),
+                        is_fatal,
+                        1,
+                        5,
+                        &description,
+                        &self.cancel,
+                    )
+                    .await;
+
+                    (node.id, response)
                 }
             });
         }
@@ -453,7 +468,6 @@ impl Service {
                         tenant_conf: models::TenantConfig::default(),
                     },
                     None,
-                    false,
                 )
                 .await
             {
@@ -478,7 +492,7 @@ impl Service {
     /// Returns a set of any shards for which notifications where not acked within the deadline.
     async fn compute_notify_many(
         &self,
-        notifications: Vec<(TenantShardId, NodeId, ShardStripeSize)>,
+        notifications: Vec<(TenantShardId, NodeId)>,
         deadline: Instant,
     ) -> HashSet<TenantShardId> {
         let compute_hook = self.inner.read().unwrap().compute_hook.clone();
@@ -489,14 +503,11 @@ impl Service {
         // Construct an async stream of futures to invoke the compute notify function: we do this
         // in order to subsequently use .buffered() on the stream to execute with bounded parallelism.
         let mut stream = futures::stream::iter(notifications.into_iter())
-            .map(|(tenant_shard_id, node_id, stripe_size)| {
+            .map(|(tenant_shard_id, node_id)| {
                 let compute_hook = compute_hook.clone();
                 let cancel = self.cancel.clone();
                 async move {
-                    if let Err(e) = compute_hook
-                        .notify(tenant_shard_id, node_id, stripe_size, &cancel)
-                        .await
-                    {
+                    if let Err(e) = compute_hook.notify(tenant_shard_id, node_id, &cancel).await {
                         tracing::error!(
                             %tenant_shard_id,
                             %node_id,
@@ -643,9 +654,19 @@ impl Service {
             .list_nodes()
             .await?
             .into_iter()
-            .map(Node::from_persistent)
+            .map(|n| Node {
+                id: NodeId(n.node_id as u64),
+                // At startup we consider a node offline until proven otherwise.
+                availability: NodeAvailability::Offline,
+                scheduling: NodeSchedulingPolicy::from_str(&n.scheduling_policy)
+                    .expect("Bad scheduling policy in DB"),
+                listen_http_addr: n.listen_http_addr,
+                listen_http_port: n.listen_http_port as u16,
+                listen_pg_addr: n.listen_pg_addr,
+                listen_pg_port: n.listen_pg_port as u16,
+            })
             .collect::<Vec<_>>();
-        let nodes: HashMap<NodeId, Node> = nodes.into_iter().map(|n| (n.get_id(), n)).collect();
+        let nodes: HashMap<NodeId, Node> = nodes.into_iter().map(|n| (n.id, n)).collect();
         tracing::info!("Loaded {} nodes from database.", nodes.len());
 
         tracing::info!("Loading shards from database...");
@@ -672,13 +693,15 @@ impl Service {
             }
             for node_id in node_ids {
                 tracing::info!("Creating node {} in scheduler for tests", node_id);
-                let node = Node::new(
-                    NodeId(node_id as u64),
-                    "".to_string(),
-                    123,
-                    "".to_string(),
-                    123,
-                );
+                let node = Node {
+                    id: NodeId(node_id as u64),
+                    availability: NodeAvailability::Active,
+                    scheduling: NodeSchedulingPolicy::Active,
+                    listen_http_addr: "".to_string(),
+                    listen_http_port: 123,
+                    listen_pg_addr: "".to_string(),
+                    listen_pg_port: 123,
+                };
 
                 scheduler.node_upsert(&node);
             }
@@ -792,7 +815,7 @@ impl Service {
                 shard_stripe_size: 0,
                 generation: Some(0),
                 generation_pageserver: None,
-                placement_policy: serde_json::to_string(&PlacementPolicy::Single).unwrap(),
+                placement_policy: serde_json::to_string(&PlacementPolicy::default()).unwrap(),
                 config: serde_json::to_string(&TenantConfig::default()).unwrap(),
                 splitting: SplitState::default(),
             };
@@ -944,12 +967,6 @@ impl Service {
         // Ordering: we must persist generation number updates before making them visible in the in-memory state
         let incremented_generations = self.persistence.re_attach(reattach_req.node_id).await?;
 
-        tracing::info!(
-            node_id=%reattach_req.node_id,
-            "Incremented {} tenant shards' generations",
-            incremented_generations.len()
-        );
-
         // Apply the updated generation to our in-memory state
         let mut locked = self.inner.write().unwrap();
 
@@ -962,6 +979,7 @@ impl Service {
                 id: tenant_shard_id,
                 gen: new_gen.into().unwrap(),
             });
+
             // Apply the new generation number to our in-memory state
             let shard_state = locked.tenants.get_mut(&tenant_shard_id);
             let Some(shard_state) = shard_state else {
@@ -997,14 +1015,6 @@ impl Service {
                 if let Some(conf) = observed.conf.as_mut() {
                     conf.generation = new_gen.into();
                 }
-            } else {
-                // This node has no observed state for the shard: perhaps it was offline
-                // when the pageserver restarted.  Insert a None, so that the Reconciler
-                // will be prompted to learn the location's state before it makes changes.
-                shard_state
-                    .observed
-                    .locations
-                    .insert(reattach_req.node_id, ObservedStateLocation { conf: None });
             }
 
             // TODO: cancel/restart any running reconciliation for this tenant, it might be trying
@@ -1053,8 +1063,9 @@ impl Service {
     pub(crate) async fn tenant_create(
         &self,
         create_req: TenantCreateRequest,
+        placement_policy: PlacementPolicy,
     ) -> Result<TenantCreateResponse, ApiError> {
-        let (response, waiters) = self.do_tenant_create(create_req).await?;
+        let (response, waiters) = self.do_tenant_create(create_req, placement_policy).await?;
 
         self.await_waiters(waiters, SHORT_RECONCILE_TIMEOUT).await?;
         Ok(response)
@@ -1063,13 +1074,8 @@ impl Service {
     pub(crate) async fn do_tenant_create(
         &self,
         create_req: TenantCreateRequest,
+        placement_policy: PlacementPolicy,
     ) -> Result<(TenantCreateResponse, Vec<ReconcilerWaiter>), ApiError> {
-        // As a default, single is convenient for tests that don't choose a policy.
-        let placement_policy = create_req
-            .placement_policy
-            .clone()
-            .unwrap_or(PlacementPolicy::Single);
-
         // This service expects to handle sharding itself: it is an error to try and directly create
         // a particular shard here.
         let tenant_id = if !create_req.new_tenant_id.is_unsharded() {
@@ -1145,12 +1151,9 @@ impl Service {
 
         let (waiters, response_shards) = {
             let mut locked = self.inner.write().unwrap();
-            let result_tx = locked.result_tx.clone();
-            let compute_hook = locked.compute_hook.clone();
-            let (nodes, tenants, scheduler) = locked.parts_mut();
+            let (_nodes, tenants, scheduler) = locked.parts_mut();
 
             let mut response_shards = Vec::new();
-            let mut schcedule_error = None;
 
             for tenant_shard_id in create_ids {
                 tracing::info!("Creating shard {tenant_shard_id}...");
@@ -1187,20 +1190,23 @@ impl Service {
                         continue;
                     }
                     Entry::Vacant(entry) => {
-                        let state = entry.insert(TenantState::new(
+                        let mut state = TenantState::new(
                             tenant_shard_id,
                             ShardIdentity::from_params(
                                 tenant_shard_id.shard_number,
                                 &create_req.shard_parameters,
                             ),
                             placement_policy.clone(),
-                        ));
+                        );
 
                         state.generation = initial_generation;
                         state.config = create_req.config.clone();
-                        if let Err(e) = state.schedule(scheduler) {
-                            schcedule_error = Some(e);
-                        }
+
+                        state.schedule(scheduler).map_err(|e| {
+                            ApiError::Conflict(format!(
+                                "Failed to schedule shard {tenant_shard_id}: {e}"
+                            ))
+                        })?;
 
                         // Only include shards in result if we are attaching: the purpose
                         // of the response is to tell the caller where the shards are attached.
@@ -1214,27 +1220,24 @@ impl Service {
                                 generation: generation.into().unwrap(),
                             });
                         }
+                        entry.insert(state)
                     }
                 };
             }
 
-            // If we failed to schedule shards, then they are still created in the controller,
-            // but we return an error to the requester to avoid a silent failure when someone
-            // tries to e.g. create a tenant whose placement policy requires more nodes than
-            // are present in the system.  We do this here rather than in the above loop, to
-            // avoid situations where we only create a subset of shards in the tenant.
-            if let Some(e) = schcedule_error {
-                return Err(ApiError::Conflict(format!(
-                    "Failed to schedule shard(s): {e}"
-                )));
-            }
+            // Take a snapshot of pageservers
+            let pageservers = locked.nodes.clone();
 
-            let waiters = tenants
+            let result_tx = locked.result_tx.clone();
+            let compute_hook = locked.compute_hook.clone();
+
+            let waiters = locked
+                .tenants
                 .range_mut(TenantShardId::tenant_range(tenant_id))
                 .filter_map(|(_shard_id, shard)| {
                     shard.maybe_reconcile(
                         result_tx.clone(),
-                        nodes,
+                        &pageservers,
                         &compute_hook,
                         &self.config,
                         &self.persistence,
@@ -1343,20 +1346,22 @@ impl Service {
 
             TenantCreateOrUpdate::Create(
                 // Synthesize a creation request
-                TenantCreateRequest {
-                    new_tenant_id: TenantShardId::unsharded(tenant_id),
-                    generation,
-                    shard_parameters: ShardParameters {
-                        // Must preserve the incoming shard_count do distinguish unsharded (0)
-                        // from single-sharded (1): this distinction appears in the S3 keys of the tenant.
-                        count: req.tenant_id.shard_count,
-                        // We only import un-sharded or single-sharded tenants, so stripe
-                        // size can be made up arbitrarily here.
-                        stripe_size: ShardParameters::DEFAULT_STRIPE_SIZE,
+                (
+                    TenantCreateRequest {
+                        new_tenant_id: TenantShardId::unsharded(tenant_id),
+                        generation,
+                        shard_parameters: ShardParameters {
+                            // Must preserve the incoming shard_count do distinguish unsharded (0)
+                            // from single-sharded (1): this distinction appears in the S3 keys of the tenant.
+                            count: req.tenant_id.shard_count,
+                            // We only import un-sharded or single-sharded tenants, so stripe
+                            // size can be made up arbitrarily here.
+                            stripe_size: ShardParameters::DEFAULT_STRIPE_SIZE,
+                        },
+                        config: req.config.tenant_conf,
                     },
-                    placement_policy: Some(placement_policy),
-                    config: req.config.tenant_conf,
-                },
+                    placement_policy,
+                ),
             )
         } else {
             TenantCreateOrUpdate::Update(updates)
@@ -1390,13 +1395,11 @@ impl Service {
         // First check if this is a creation or an update
         let create_or_update = self.tenant_location_config_prepare(tenant_id, req);
 
-        let mut result = TenantLocationConfigResponse {
-            shards: Vec::new(),
-            stripe_size: None,
-        };
+        let mut result = TenantLocationConfigResponse { shards: Vec::new() };
         let waiters = match create_or_update {
-            TenantCreateOrUpdate::Create(create_req) => {
-                let (create_resp, waiters) = self.do_tenant_create(create_req).await?;
+            TenantCreateOrUpdate::Create((create_req, placement_policy)) => {
+                let (create_resp, waiters) =
+                    self.do_tenant_create(create_req, placement_policy).await?;
                 result.shards = create_resp
                     .shards
                     .into_iter()
@@ -1448,11 +1451,6 @@ impl Service {
                             continue;
                         };
 
-                        // Update stripe size
-                        if result.stripe_size.is_none() && shard.shard.count.count() > 1 {
-                            result.stripe_size = Some(shard.shard.stripe_size);
-                        }
-
                         shard.policy = placement_policy;
                         shard.config = tenant_config;
                         if let Some(generation) = update_generation {
@@ -1668,7 +1666,7 @@ impl Service {
                         .map_err(|e| {
                             ApiError::InternalServerError(anyhow::anyhow!(
                                 "Error doing time travel recovery for shard {tenant_shard_id} on node {}: {e}",
-                                node
+                                node.id
                             ))
                         })?;
             }
@@ -1722,7 +1720,10 @@ impl Service {
             // Secondary downloads are always advisory: if something fails, we nevertheless report success, so that whoever
             // is calling us will proceed with whatever migration they're doing, albeit with a slightly less warm cache
             // than they had hoped for.
-            tracing::warn!("Ignoring tenant secondary download error from pageserver {node}: {e}",);
+            tracing::warn!(
+                "Ignoring tenant secondary download error from pageserver {}: {e}",
+                node.id,
+            );
         }
 
         Ok(())
@@ -1760,11 +1761,13 @@ impl Service {
             // surface immediately as an error to our caller.
             let status = client.tenant_delete(tenant_shard_id).await.map_err(|e| {
                 ApiError::InternalServerError(anyhow::anyhow!(
-                    "Error deleting shard {tenant_shard_id} on node {node}: {e}",
+                    "Error deleting shard {tenant_shard_id} on node {}: {e}",
+                    node.id
                 ))
             })?;
             tracing::info!(
-                "Shard {tenant_shard_id} on node {node}, delete returned {}",
+                "Shard {tenant_shard_id} on node {}, delete returned {}",
+                node.id,
                 status
             );
             if status == StatusCode::ACCEPTED {
@@ -1863,9 +1866,10 @@ impl Service {
             create_req: TimelineCreateRequest,
         ) -> Result<TimelineInfo, ApiError> {
             tracing::info!(
-                "Creating timeline on shard {}/{}, attached to node {node}",
+                "Creating timeline on shard {}/{}, attached to node {}",
                 tenant_shard_id,
                 create_req.new_timeline_id,
+                node.id
             );
             let client = mgmt_api::Client::new(node.base_url(), jwt.as_deref());
 
@@ -1989,7 +1993,10 @@ impl Service {
             jwt: Option<String>,
         ) -> Result<StatusCode, ApiError> {
             tracing::info!(
-                "Deleting timeline on shard {tenant_shard_id}/{timeline_id}, attached to node {node}",
+                "Deleting timeline on shard {}/{}, attached to node {}",
+                tenant_shard_id,
+                timeline_id,
+                node.id
             );
 
             let client = mgmt_api::Client::new(node.base_url(), jwt.as_deref());
@@ -1998,7 +2005,8 @@ impl Service {
                 .await
                 .map_err(|e| {
                     ApiError::InternalServerError(anyhow::anyhow!(
-                    "Error deleting timeline {timeline_id} on {tenant_shard_id} on node {node}: {e}",
+                    "Error deleting timeline {timeline_id} on {tenant_shard_id} on node {}: {e}",
+                    node.id
                 ))
                 })
         }
@@ -2099,7 +2107,14 @@ impl Service {
                 .get(&node_id)
                 .expect("Pageservers may not be deleted while referenced");
 
-            result.push(node.shard_location(*tenant_shard_id));
+            result.push(TenantLocateResponseShard {
+                shard_id: *tenant_shard_id,
+                node_id,
+                listen_http_addr: node.listen_http_addr.clone(),
+                listen_http_port: node.listen_http_port,
+                listen_pg_addr: node.listen_pg_addr.clone(),
+                listen_pg_port: node.listen_pg_port,
+            });
 
             match &shard_params {
                 None => {
@@ -2290,7 +2305,7 @@ impl Service {
                     // populate the correct generation as part of its transaction, to protect us
                     // against racing with changes in the state of the parent.
                     generation: None,
-                    generation_pageserver: Some(target.node.get_id().0 as i64),
+                    generation_pageserver: Some(target.node.id.0 as i64),
                     placement_policy: serde_json::to_string(&policy).unwrap(),
                     // TODO: get the config out of the map
                     config: serde_json::to_string(&TenantConfig::default()).unwrap(),
@@ -2440,7 +2455,7 @@ impl Service {
                     // as at this point in the split process we have succeeded and this part is infallible:
                     // we will never need to do any special recovery from this state.
 
-                    child_locations.push((child, pageserver, child_shard.stripe_size));
+                    child_locations.push((child, pageserver));
 
                     tenants.insert(child, child_state);
                     response.new_shards.push(child);
@@ -2450,11 +2465,8 @@ impl Service {
 
         // Send compute notifications for all the new shards
         let mut failed_notifications = Vec::new();
-        for (child_id, child_ps, stripe_size) in child_locations {
-            if let Err(e) = compute_hook
-                .notify(child_id, child_ps, stripe_size, &self.cancel)
-                .await
-            {
+        for (child_id, child_ps) in child_locations {
+            if let Err(e) = compute_hook.notify(child_id, child_ps, &self.cancel).await {
                 tracing::warn!("Failed to update compute of {}->{} during split, proceeding anyway to complete split ({e})",
                         child_id, child_ps);
                 failed_notifications.push(child_id);
@@ -2485,19 +2497,6 @@ impl Service {
             let compute_hook = locked.compute_hook.clone();
             let (nodes, tenants, scheduler) = locked.parts_mut();
 
-            let Some(node) = nodes.get(&migrate_req.node_id) else {
-                return Err(ApiError::BadRequest(anyhow::anyhow!(
-                    "Node {} not found",
-                    migrate_req.node_id
-                )));
-            };
-
-            if !node.is_available() {
-                // Warn but proceed: the caller may intend to manually adjust the placement of
-                // a shard even if the node is down, e.g. if intervening during an incident.
-                tracing::warn!("Migrating to unavailable node {node}");
-            }
-
             let Some(shard) = tenants.get_mut(&tenant_shard_id) else {
                 return Err(ApiError::NotFound(
                     anyhow::anyhow!("Tenant shard not found").into(),
@@ -2627,18 +2626,6 @@ impl Service {
                 .map(|t| t.to_persistent())
                 .collect::<Vec<_>>();
 
-            // This method can only validate the state of an idle system: if a reconcile is in
-            // progress, fail out early to avoid giving false errors on state that won't match
-            // between database and memory under a ReconcileResult is processed.
-            for t in locked.tenants.values() {
-                if t.reconciler.is_some() {
-                    return Err(ApiError::InternalServerError(anyhow::anyhow!(
-                        "Shard {} reconciliation in progress",
-                        t.tenant_shard_id
-                    )));
-                }
-            }
-
             (expect_nodes, expect_shards)
         };
 
@@ -2750,7 +2737,11 @@ impl Service {
             if let Some(node) = locked.nodes.get(&register_req.node_id) {
                 // Note that we do not do a total equality of the struct, because we don't require
                 // the availability/scheduling states to agree for a POST to be idempotent.
-                if node.registration_match(&register_req) {
+                if node.listen_http_addr == register_req.listen_http_addr
+                    && node.listen_http_port == register_req.listen_http_port
+                    && node.listen_pg_addr == register_req.listen_pg_addr
+                    && node.listen_pg_port == register_req.listen_pg_port
+                {
                     tracing::info!(
                         "Node {} re-registered with matching address",
                         register_req.node_id
@@ -2774,14 +2765,16 @@ impl Service {
         // Ordering: we must persist the new node _before_ adding it to in-memory state.
         // This ensures that before we use it for anything or expose it via any external
         // API, it is guaranteed to be available after a restart.
-        let new_node = Node::new(
-            register_req.node_id,
-            register_req.listen_http_addr,
-            register_req.listen_http_port,
-            register_req.listen_pg_addr,
-            register_req.listen_pg_port,
-        );
-
+        let new_node = Node {
+            id: register_req.node_id,
+            listen_http_addr: register_req.listen_http_addr,
+            listen_http_port: register_req.listen_http_port,
+            listen_pg_addr: register_req.listen_pg_addr,
+            listen_pg_port: register_req.listen_pg_port,
+            scheduling: NodeSchedulingPolicy::Filling,
+            // TODO: we shouldn't really call this Active until we've heartbeated it.
+            availability: NodeAvailability::Active,
+        };
         // TODO: idempotency if the node already exists in the database
         self.persistence.insert_node(&new_node).await?;
 
@@ -2826,14 +2819,29 @@ impl Service {
             ));
         };
 
-        let availability_transition = if let Some(availability) = &config_req.availability {
-            node.set_availability(*availability)
-        } else {
-            AvailabilityTransition::Unchanged
-        };
+        let mut offline_transition = false;
+        let mut active_transition = false;
+
+        if let Some(availability) = &config_req.availability {
+            match (availability, &node.availability) {
+                (NodeAvailability::Offline, NodeAvailability::Active) => {
+                    tracing::info!("Node {} transition to offline", config_req.node_id);
+                    offline_transition = true;
+                }
+                (NodeAvailability::Active, NodeAvailability::Offline) => {
+                    tracing::info!("Node {} transition to active", config_req.node_id);
+                    active_transition = true;
+                }
+                _ => {
+                    tracing::info!("Node {} no change during config", config_req.node_id);
+                    // No change
+                }
+            };
+            node.availability = *availability;
+        }
 
         if let Some(scheduling) = config_req.scheduling {
-            node.set_scheduling(scheduling);
+            node.scheduling = scheduling;
 
             // TODO: once we have a background scheduling ticker for fill/drain, kick it
             // to wake up and start working.
@@ -2844,80 +2852,74 @@ impl Service {
 
         let new_nodes = Arc::new(new_nodes);
 
-        match availability_transition {
-            AvailabilityTransition::ToOffline => {
-                tracing::info!("Node {} transition to offline", config_req.node_id);
-                let mut tenants_affected: usize = 0;
-                for (tenant_shard_id, tenant_state) in tenants {
-                    if let Some(observed_loc) =
-                        tenant_state.observed.locations.get_mut(&config_req.node_id)
-                    {
-                        // When a node goes offline, we set its observed configuration to None, indicating unknown: we will
-                        // not assume our knowledge of the node's configuration is accurate until it comes back online
-                        observed_loc.conf = None;
-                    }
+        if offline_transition {
+            let mut tenants_affected: usize = 0;
+            for (tenant_shard_id, tenant_state) in tenants {
+                if let Some(observed_loc) =
+                    tenant_state.observed.locations.get_mut(&config_req.node_id)
+                {
+                    // When a node goes offline, we set its observed configuration to None, indicating unknown: we will
+                    // not assume our knowledge of the node's configuration is accurate until it comes back online
+                    observed_loc.conf = None;
+                }
 
-                    if tenant_state.intent.demote_attached(config_req.node_id) {
-                        tenant_state.sequence = tenant_state.sequence.next();
-                        match tenant_state.schedule(scheduler) {
-                            Err(e) => {
-                                // It is possible that some tenants will become unschedulable when too many pageservers
-                                // go offline: in this case there isn't much we can do other than make the issue observable.
-                                // TODO: give TenantState a scheduling error attribute to be queried later.
-                                tracing::warn!(%tenant_shard_id, "Scheduling error when marking pageserver {} offline: {e}", config_req.node_id);
-                            }
-                            Ok(()) => {
-                                if tenant_state
-                                    .maybe_reconcile(
-                                        result_tx.clone(),
-                                        &new_nodes,
-                                        &compute_hook,
-                                        &self.config,
-                                        &self.persistence,
-                                        &self.gate,
-                                        &self.cancel,
-                                    )
-                                    .is_some()
-                                {
-                                    tenants_affected += 1;
-                                };
-                            }
+                if tenant_state.intent.demote_attached(config_req.node_id) {
+                    tenant_state.sequence = tenant_state.sequence.next();
+                    match tenant_state.schedule(scheduler) {
+                        Err(e) => {
+                            // It is possible that some tenants will become unschedulable when too many pageservers
+                            // go offline: in this case there isn't much we can do other than make the issue observable.
+                            // TODO: give TenantState a scheduling error attribute to be queried later.
+                            tracing::warn!(%tenant_shard_id, "Scheduling error when marking pageserver {} offline: {e}", config_req.node_id);
+                        }
+                        Ok(()) => {
+                            if tenant_state
+                                .maybe_reconcile(
+                                    result_tx.clone(),
+                                    &new_nodes,
+                                    &compute_hook,
+                                    &self.config,
+                                    &self.persistence,
+                                    &self.gate,
+                                    &self.cancel,
+                                )
+                                .is_some()
+                            {
+                                tenants_affected += 1;
+                            };
                         }
                     }
                 }
-                tracing::info!(
-                    "Launched {} reconciler tasks for tenants affected by node {} going offline",
-                    tenants_affected,
-                    config_req.node_id
-                )
             }
-            AvailabilityTransition::ToActive => {
-                tracing::info!("Node {} transition to active", config_req.node_id);
-                // When a node comes back online, we must reconcile any tenant that has a None observed
-                // location on the node.
-                for tenant_state in locked.tenants.values_mut() {
-                    if let Some(observed_loc) =
-                        tenant_state.observed.locations.get_mut(&config_req.node_id)
-                    {
-                        if observed_loc.conf.is_none() {
-                            tenant_state.maybe_reconcile(
-                                result_tx.clone(),
-                                &new_nodes,
-                                &compute_hook,
-                                &self.config,
-                                &self.persistence,
-                                &self.gate,
-                                &self.cancel,
-                            );
-                        }
+            tracing::info!(
+                "Launched {} reconciler tasks for tenants affected by node {} going offline",
+                tenants_affected,
+                config_req.node_id
+            )
+        }
+
+        if active_transition {
+            // When a node comes back online, we must reconcile any tenant that has a None observed
+            // location on the node.
+            for tenant_state in locked.tenants.values_mut() {
+                if let Some(observed_loc) =
+                    tenant_state.observed.locations.get_mut(&config_req.node_id)
+                {
+                    if observed_loc.conf.is_none() {
+                        tenant_state.maybe_reconcile(
+                            result_tx.clone(),
+                            &new_nodes,
+                            &compute_hook,
+                            &self.config,
+                            &self.persistence,
+                            &self.gate,
+                            &self.cancel,
+                        );
                     }
                 }
+            }
 
-                // TODO: in the background, we should balance work back onto this pageserver
-            }
-            AvailabilityTransition::Unchanged => {
-                tracing::info!("Node {} no change during config", config_req.node_id);
-            }
+            // TODO: in the background, we should balance work back onto this pageserver
         }
 
         locked.nodes = new_nodes;

control_plane/attachment_service/src/tenant_state.rs

@@ -1,11 +1,7 @@
-use std::{
-    collections::{HashMap, HashSet},
-    sync::Arc,
-    time::Duration,
-};
+use std::{collections::HashMap, sync::Arc, time::Duration};
 
 use crate::{metrics, persistence::TenantShardPersistence};
-use pageserver_api::controller_api::PlacementPolicy;
+use pageserver_api::controller_api::NodeAvailability;
 use pageserver_api::{
     models::{LocationConfig, LocationConfigMode, TenantConfig},
     shard::{ShardIdentity, TenantShardId},
@@ -29,7 +25,7 @@ use crate::{
         attached_location_conf, secondary_location_conf, ReconcileError, Reconciler, TargetState,
     },
     scheduler::{ScheduleError, Scheduler},
-    service, Sequence,
+    service, PlacementPolicy, Sequence,
 };
 
 /// Serialization helper
@@ -374,7 +370,7 @@ impl TenantState {
     /// [`ObservedState`], even if it violates my [`PlacementPolicy`].  Call [`Self::schedule`] next,
     /// to get an intent state that complies with placement policy.  The overall goal is to do scheduling
     /// in a way that makes use of any configured locations that already exist in the outside world.
-    pub(crate) fn intent_from_observed(&mut self, scheduler: &mut Scheduler) {
+    pub(crate) fn intent_from_observed(&mut self) {
         // Choose an attached location by filtering observed locations, and then sorting to get the highest
         // generation
         let mut attached_locs = self
@@ -399,7 +395,7 @@ impl TenantState {
 
         attached_locs.sort_by_key(|i| i.1);
         if let Some((node_id, _gen)) = attached_locs.into_iter().last() {
-            self.intent.set_attached(scheduler, Some(*node_id));
+            self.intent.attached = Some(*node_id);
         }
 
         // All remaining observed locations generate secondary intents.  This includes None
@@ -410,7 +406,7 @@ impl TenantState {
         // will take care of promoting one of these secondaries to be attached.
         self.observed.locations.keys().for_each(|node_id| {
             if Some(*node_id) != self.intent.attached {
-                self.intent.push_secondary(scheduler, *node_id);
+                self.intent.secondary.push(*node_id);
             }
         });
     }
@@ -568,9 +564,7 @@ impl TenantState {
         }
     }
 
-    fn dirty(&self, nodes: &Arc<HashMap<NodeId, Node>>) -> bool {
-        let mut dirty_nodes = HashSet::new();
-
+    fn dirty(&self) -> bool {
         if let Some(node_id) = self.intent.attached {
             // Maybe panic: it is a severe bug if we try to attach while generation is null.
             let generation = self
@@ -581,7 +575,7 @@ impl TenantState {
             match self.observed.locations.get(&node_id) {
                 Some(conf) if conf.conf.as_ref() == Some(&wanted_conf) => {}
                 Some(_) | None => {
-                    dirty_nodes.insert(node_id);
+                    return true;
                 }
             }
         }
@@ -591,7 +585,7 @@ impl TenantState {
             match self.observed.locations.get(node_id) {
                 Some(conf) if conf.conf.as_ref() == Some(&wanted_conf) => {}
                 Some(_) | None => {
-                    dirty_nodes.insert(*node_id);
+                    return true;
                 }
             }
         }
@@ -599,18 +593,17 @@ impl TenantState {
         for node_id in self.observed.locations.keys() {
             if self.intent.attached != Some(*node_id) && !self.intent.secondary.contains(node_id) {
                 // We have observed state that isn't part of our intent: need to clean it up.
-                dirty_nodes.insert(*node_id);
+                return true;
             }
         }
 
-        dirty_nodes.retain(|node_id| {
-            nodes
-                .get(node_id)
-                .map(|n| n.is_available())
-                .unwrap_or(false)
-        });
+        // Even if there is no pageserver work to be done, if we have a pending notification to computes,
+        // wake up a reconciler to send it.
+        if self.pending_compute_notification {
+            return true;
+        }
 
-        !dirty_nodes.is_empty()
+        false
     }
 
     #[allow(clippy::too_many_arguments)]
@@ -632,20 +625,15 @@ impl TenantState {
             let node = pageservers
                 .get(node_id)
                 .expect("Nodes may not be removed while referenced");
-            if observed_loc.conf.is_none() && node.is_available() {
+            if observed_loc.conf.is_none()
+                && !matches!(node.availability, NodeAvailability::Offline)
+            {
                 dirty_observed = true;
                 break;
             }
         }
 
-        let active_nodes_dirty = self.dirty(pageservers);
-
-        // Even if there is no pageserver work to be done, if we have a pending notification to computes,
-        // wake up a reconciler to send it.
-        let do_reconcile =
-            active_nodes_dirty || dirty_observed || self.pending_compute_notification;
-
-        if !do_reconcile {
+        if !self.dirty() && !dirty_observed {
             tracing::info!("Not dirty, no reconciliation needed.");
             return None;
         }
@@ -675,21 +663,6 @@ impl TenantState {
             }
         }
 
-        // Build list of nodes from which the reconciler should detach
-        let mut detach = Vec::new();
-        for node_id in self.observed.locations.keys() {
-            if self.intent.get_attached() != &Some(*node_id)
-                && !self.intent.secondary.contains(node_id)
-            {
-                detach.push(
-                    pageservers
-                        .get(node_id)
-                        .expect("Intent references non-existent pageserver")
-                        .clone(),
-                )
-            }
-        }
-
         // Reconcile in flight for a stale sequence?  Our sequence's task will wait for it before
         // doing our sequence's work.
         let old_handle = self.reconciler.take();
@@ -704,15 +677,14 @@ impl TenantState {
         self.sequence = self.sequence.next();
 
         let reconciler_cancel = cancel.child_token();
-        let reconciler_intent = TargetState::from_intent(pageservers, &self.intent);
         let mut reconciler = Reconciler {
             tenant_shard_id: self.tenant_shard_id,
             shard: self.shard,
             generation: self.generation,
-            intent: reconciler_intent,
-            detach,
+            intent: TargetState::from_intent(&self.intent),
             config: self.config.clone(),
             observed: self.observed.clone(),
+            pageservers: pageservers.clone(),
             compute_hook: compute_hook.clone(),
             service_config: service_config.clone(),
             _gate_guard: gate_guard,
@@ -847,10 +819,7 @@ impl TenantState {
 
 #[cfg(test)]
 pub(crate) mod tests {
-    use pageserver_api::{
-        controller_api::NodeAvailability,
-        shard::{ShardCount, ShardNumber},
-    };
+    use pageserver_api::shard::{ShardCount, ShardNumber};
     use utils::id::TenantId;
 
     use crate::scheduler::test_utils::make_test_nodes;
@@ -909,10 +878,7 @@ pub(crate) mod tests {
         assert_eq!(tenant_state.intent.secondary.len(), 2);
 
         // Update the scheduler state to indicate the node is offline
-        nodes
-            .get_mut(&attached_node_id)
-            .unwrap()
-            .set_availability(NodeAvailability::Offline);
+        nodes.get_mut(&attached_node_id).unwrap().availability = NodeAvailability::Offline;
         scheduler.node_upsert(nodes.get(&attached_node_id).unwrap());
 
         // Scheduling the node should promote the still-available secondary node to attached
@@ -931,54 +897,4 @@ pub(crate) mod tests {
 
         Ok(())
     }
-
-    #[test]
-    fn intent_from_observed() -> anyhow::Result<()> {
-        let nodes = make_test_nodes(3);
-        let mut scheduler = Scheduler::new(nodes.values());
-
-        let mut tenant_state = make_test_tenant_shard(PlacementPolicy::Double(1));
-
-        tenant_state.observed.locations.insert(
-            NodeId(3),
-            ObservedStateLocation {
-                conf: Some(LocationConfig {
-                    mode: LocationConfigMode::AttachedMulti,
-                    generation: Some(2),
-                    secondary_conf: None,
-                    shard_number: tenant_state.shard.number.0,
-                    shard_count: tenant_state.shard.count.literal(),
-                    shard_stripe_size: tenant_state.shard.stripe_size.0,
-                    tenant_conf: TenantConfig::default(),
-                }),
-            },
-        );
-
-        tenant_state.observed.locations.insert(
-            NodeId(2),
-            ObservedStateLocation {
-                conf: Some(LocationConfig {
-                    mode: LocationConfigMode::AttachedStale,
-                    generation: Some(1),
-                    secondary_conf: None,
-                    shard_number: tenant_state.shard.number.0,
-                    shard_count: tenant_state.shard.count.literal(),
-                    shard_stripe_size: tenant_state.shard.stripe_size.0,
-                    tenant_conf: TenantConfig::default(),
-                }),
-            },
-        );
-
-        tenant_state.intent_from_observed(&mut scheduler);
-
-        // The highest generationed attached location gets used as attached
-        assert_eq!(tenant_state.intent.attached, Some(NodeId(3)));
-        // Other locations get used as secondary
-        assert_eq!(tenant_state.intent.secondary, vec![NodeId(2)]);
-
-        scheduler.consistency_check(nodes.values(), [&tenant_state].into_iter())?;
-
-        tenant_state.intent.clear(&mut scheduler);
-        Ok(())
-    }
 }

control_plane/src/attachment_service.rs

@@ -34,7 +34,7 @@ pub struct AttachmentService {
     client: reqwest::Client,
 }
 
-const COMMAND: &str = "storage_controller";
+const COMMAND: &str = "attachment_service";
 
 const ATTACHMENT_SERVICE_POSTGRES_VERSION: u32 = 16;
 

control_plane/src/bin/neon_local.rs

@@ -15,7 +15,7 @@ use control_plane::pageserver::{PageServerNode, PAGESERVER_REMOTE_STORAGE_DIR};
 use control_plane::safekeeper::SafekeeperNode;
 use control_plane::{broker, local_env};
 use pageserver_api::controller_api::{
-    NodeAvailability, NodeConfigureRequest, NodeSchedulingPolicy, PlacementPolicy,
+    NodeAvailability, NodeConfigureRequest, NodeSchedulingPolicy,
 };
 use pageserver_api::models::{
     ShardParameters, TenantCreateRequest, TimelineCreateRequest, TimelineInfo,
@@ -435,11 +435,6 @@ async fn handle_tenant(
             let shard_stripe_size: Option<u32> =
                 create_match.get_one::<u32>("shard-stripe-size").cloned();
 
-            let placement_policy = match create_match.get_one::<String>("placement-policy") {
-                Some(s) if !s.is_empty() => serde_json::from_str::<PlacementPolicy>(s)?,
-                _ => PlacementPolicy::Single,
-            };
-
             let tenant_conf = PageServerNode::parse_config(tenant_conf)?;
 
             // If tenant ID was not specified, generate one
@@ -461,7 +456,6 @@ async fn handle_tenant(
                             .map(ShardStripeSize)
                             .unwrap_or(ShardParameters::DEFAULT_STRIPE_SIZE),
                     },
-                    placement_policy: Some(placement_policy),
                     config: tenant_conf,
                 })
                 .await?;
@@ -1030,7 +1024,7 @@ async fn handle_endpoint(ep_match: &ArgMatches, env: &local_env::LocalEnv) -> Re
                         })
                         .collect::<Vec<_>>()
                 };
-            endpoint.reconfigure(pageservers, None).await?;
+            endpoint.reconfigure(pageservers).await?;
         }
         "stop" => {
             let endpoint_id = sub_args
@@ -1568,7 +1562,6 @@ fn cli() -> Command {
                     .help("Use this tenant in future CLI commands where tenant_id is needed, but not specified"))
                 .arg(Arg::new("shard-count").value_parser(value_parser!(u8)).long("shard-count").action(ArgAction::Set).help("Number of shards in the new tenant (default 1)"))
                 .arg(Arg::new("shard-stripe-size").value_parser(value_parser!(u32)).long("shard-stripe-size").action(ArgAction::Set).help("Sharding stripe size in pages"))
-                .arg(Arg::new("placement-policy").value_parser(value_parser!(String)).long("placement-policy").action(ArgAction::Set).help("Placement policy shards in this tenant"))
                 )
             .subcommand(Command::new("set-default").arg(tenant_id_arg.clone().required(true))
                 .about("Set a particular tenant as default in future CLI commands where tenant_id is needed, but not specified"))

control_plane/src/endpoint.rs

@@ -52,7 +52,6 @@ use compute_api::spec::RemoteExtSpec;
 use compute_api::spec::Role;
 use nix::sys::signal::kill;
 use nix::sys::signal::Signal;
-use pageserver_api::shard::ShardStripeSize;
 use serde::{Deserialize, Serialize};
 use url::Host;
 use utils::id::{NodeId, TenantId, TimelineId};
@@ -656,7 +655,7 @@ impl Endpoint {
         // Wait for it to start
         let mut attempt = 0;
         const ATTEMPT_INTERVAL: Duration = Duration::from_millis(100);
-        const MAX_ATTEMPTS: u32 = 10 * 90; // Wait up to 1.5 min
+        const MAX_ATTEMPTS: u32 = 10 * 30; // Wait up to 30 s
         loop {
             attempt += 1;
             match self.get_status().await {
@@ -736,11 +735,7 @@ impl Endpoint {
         }
     }
 
-    pub async fn reconfigure(
-        &self,
-        mut pageservers: Vec<(Host, u16)>,
-        stripe_size: Option<ShardStripeSize>,
-    ) -> Result<()> {
+    pub async fn reconfigure(&self, mut pageservers: Vec<(Host, u16)>) -> Result<()> {
         let mut spec: ComputeSpec = {
             let spec_path = self.endpoint_path().join("spec.json");
             let file = std::fs::File::open(spec_path)?;
@@ -770,9 +765,6 @@ impl Endpoint {
         let pageserver_connstr = Self::build_pageserver_connstr(&pageservers);
         assert!(!pageserver_connstr.is_empty());
         spec.pageserver_connstring = Some(pageserver_connstr);
-        if stripe_size.is_some() {
-            spec.shard_stripe_size = stripe_size.map(|s| s.0 as usize);
-        }
 
         let client = reqwest::Client::new();
         let response = client

control_plane/src/local_env.rs

@@ -232,7 +232,7 @@ impl LocalEnv {
         // run from the same location as neon_local.  This means that for compatibility
         // tests that run old pageserver/safekeeper, they still run latest attachment service.
         let neon_local_bin_dir = env::current_exe().unwrap().parent().unwrap().to_owned();
-        neon_local_bin_dir.join("storage_controller")
+        neon_local_bin_dir.join("attachment_service")
     }
 
     pub fn safekeeper_bin(&self) -> PathBuf {

control_plane/src/pageserver.rs

@@ -429,8 +429,6 @@ impl PageServerNode {
             generation,
             config,
             shard_parameters: ShardParameters::default(),
-            // Placement policy is not meaningful for creations not done via storage controller
-            placement_policy: None,
         };
         if !settings.is_empty() {
             bail!("Unrecognized tenant settings: {settings:?}")
@@ -539,11 +537,10 @@ impl PageServerNode {
         tenant_shard_id: TenantShardId,
         config: LocationConfig,
         flush_ms: Option<Duration>,
-        lazy: bool,
     ) -> anyhow::Result<()> {
         Ok(self
             .http_client
-            .location_config(tenant_shard_id, config, flush_ms, lazy)
+            .location_config(tenant_shard_id, config, flush_ms)
             .await?)
     }
 
@@ -608,7 +605,7 @@ impl PageServerNode {
                 eprintln!("connection error: {}", e);
             }
         });
-        let client = std::pin::pin!(client);
+        tokio::pin!(client);
 
         // Init base reader
         let (start_lsn, base_tarfile_path) = base;

libs/pageserver_api/src/controller_api.rs

@@ -125,45 +125,5 @@ impl From<NodeSchedulingPolicy> for String {
     }
 }
 
-/// Controls how tenant shards are mapped to locations on pageservers, e.g. whether
-/// to create secondary locations.
-#[derive(Clone, Serialize, Deserialize, Debug, PartialEq, Eq)]
-pub enum PlacementPolicy {
-    /// Cheapest way to attach a tenant: just one pageserver, no secondary
-    Single,
-    /// Production-ready way to attach a tenant: one attached pageserver and
-    /// some number of secondaries.
-    Double(usize),
-    /// Create one secondary mode locations. This is useful when onboarding
-    /// a tenant, or for an idle tenant that we might want to bring online quickly.
-    Secondary,
-
-    /// Do not attach to any pageservers.  This is appropriate for tenants that
-    /// have been idle for a long time, where we do not mind some delay in making
-    /// them available in future.
-    Detached,
-}
-
 #[derive(Serialize, Deserialize, Debug)]
 pub struct TenantShardMigrateResponse {}
-
-#[cfg(test)]
-mod test {
-    use super::*;
-    use serde_json;
-
-    /// Check stability of PlacementPolicy's serialization
-    #[test]
-    fn placement_policy_encoding() -> anyhow::Result<()> {
-        let v = PlacementPolicy::Double(1);
-        let encoded = serde_json::to_string(&v)?;
-        assert_eq!(encoded, "{\"Double\":1}");
-        assert_eq!(serde_json::from_str::<PlacementPolicy>(&encoded)?, v);
-
-        let v = PlacementPolicy::Single;
-        let encoded = serde_json::to_string(&v)?;
-        assert_eq!(encoded, "\"Single\"");
-        assert_eq!(serde_json::from_str::<PlacementPolicy>(&encoded)?, v);
-        Ok(())
-    }
-}

libs/pageserver_api/src/models.rs

@@ -21,7 +21,6 @@ use utils::{
     lsn::Lsn,
 };
 
-use crate::controller_api::PlacementPolicy;
 use crate::{
     reltag::RelTag,
     shard::{ShardCount, ShardStripeSize, TenantShardId},
@@ -243,11 +242,6 @@ pub struct TenantCreateRequest {
     #[serde(skip_serializing_if = "ShardParameters::is_unsharded")]
     pub shard_parameters: ShardParameters,
 
-    // This parameter is only meaningful in requests sent to the storage controller
-    #[serde(default)]
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub placement_policy: Option<PlacementPolicy>,
-
     #[serde(flatten)]
     pub config: TenantConfig, // as we have a flattened field, we should reject all unknown fields in it
 }
@@ -441,8 +435,6 @@ pub struct TenantShardLocation {
 #[serde(deny_unknown_fields)]
 pub struct TenantLocationConfigResponse {
     pub shards: Vec<TenantShardLocation>,
-    // If the shards' ShardCount count is >1, stripe_size will be set.
-    pub stripe_size: Option<ShardStripeSize>,
 }
 
 #[derive(Serialize, Deserialize, Debug)]

libs/postgres_backend/src/lib.rs

@@ -6,6 +6,7 @@
 #![deny(clippy::undocumented_unsafe_blocks)]
 use anyhow::Context;
 use bytes::Bytes;
+use futures::pin_mut;
 use serde::{Deserialize, Serialize};
 use std::io::ErrorKind;
 use std::net::SocketAddr;
@@ -377,7 +378,8 @@ impl<IO: AsyncRead + AsyncWrite + Unpin> PostgresBackend<IO> {
         &mut self,
         cx: &mut std::task::Context<'_>,
     ) -> Poll<Result<(), std::io::Error>> {
-        let flush_fut = std::pin::pin!(self.flush());
+        let flush_fut = self.flush();
+        pin_mut!(flush_fut);
         flush_fut.poll(cx)
     }
 

libs/postgres_backend/tests/simple_select.rs

@@ -72,19 +72,14 @@ async fn simple_select() {
     }
 }
 
-static KEY: Lazy<rustls::pki_types::PrivateKeyDer<'static>> = Lazy::new(|| {
+static KEY: Lazy<rustls::PrivateKey> = Lazy::new(|| {
     let mut cursor = Cursor::new(include_bytes!("key.pem"));
-    let key = rustls_pemfile::rsa_private_keys(&mut cursor)
-        .next()
-        .unwrap()
-        .unwrap();
-    rustls::pki_types::PrivateKeyDer::Pkcs1(key)
+    rustls::PrivateKey(rustls_pemfile::rsa_private_keys(&mut cursor).unwrap()[0].clone())
 });
 
-static CERT: Lazy<rustls::pki_types::CertificateDer<'static>> = Lazy::new(|| {
+static CERT: Lazy<rustls::Certificate> = Lazy::new(|| {
     let mut cursor = Cursor::new(include_bytes!("cert.pem"));
-    let cert = rustls_pemfile::certs(&mut cursor).next().unwrap().unwrap();
-    cert
+    rustls::Certificate(rustls_pemfile::certs(&mut cursor).unwrap()[0].clone())
 });
 
 // test that basic select with ssl works
@@ -93,8 +88,9 @@ async fn simple_select_ssl() {
     let (client_sock, server_sock) = make_tcp_pair().await;
 
     let server_cfg = rustls::ServerConfig::builder()
+        .with_safe_defaults()
         .with_no_client_auth()
-        .with_single_cert(vec![CERT.clone()], KEY.clone_key())
+        .with_single_cert(vec![CERT.clone()], KEY.clone())
         .unwrap();
     let tls_config = Some(Arc::new(server_cfg));
     let pgbackend =
@@ -106,9 +102,10 @@ async fn simple_select_ssl() {
     });
 
     let client_cfg = rustls::ClientConfig::builder()
+        .with_safe_defaults()
         .with_root_certificates({
             let mut store = rustls::RootCertStore::empty();
-            store.add(CERT.clone()).unwrap();
+            store.add(&CERT).unwrap();
             store
         })
         .with_no_client_auth();

pageserver/client/src/mgmt_api.rs

@@ -7,7 +7,7 @@ use utils::{
 
 pub mod util;
 
-#[derive(Debug, Clone)]
+#[derive(Debug)]
 pub struct Client {
     mgmt_api_endpoint: String,
     authorization_header: Option<String>,
@@ -24,9 +24,6 @@ pub enum Error {
 
     #[error("pageserver API: {1}")]
     ApiError(StatusCode, String),
-
-    #[error("Cancelled")]
-    Cancelled,
 }
 
 pub type Result<T> = std::result::Result<T, Error>;
@@ -254,30 +251,21 @@ impl Client {
         tenant_shard_id: TenantShardId,
         config: LocationConfig,
         flush_ms: Option<std::time::Duration>,
-        lazy: bool,
     ) -> Result<()> {
         let req_body = TenantLocationConfigRequest {
             tenant_id: tenant_shard_id,
             config,
         };
-
-        let mut path = reqwest::Url::parse(&format!(
+        let path = format!(
             "{}/v1/tenant/{}/location_config",
             self.mgmt_api_endpoint, tenant_shard_id
-        ))
-        // Should always work: mgmt_api_endpoint is configuration, not user input.
-        .expect("Cannot build URL");
-
-        if lazy {
-            path.query_pairs_mut().append_pair("lazy", "true");
-        }
-
-        if let Some(flush_ms) = flush_ms {
-            path.query_pairs_mut()
-                .append_pair("flush_ms", &format!("{}", flush_ms.as_millis()));
-        }
-
-        self.request(Method::PUT, path, &req_body).await?;
+        );
+        let path = if let Some(flush_ms) = flush_ms {
+            format!("{}?flush_ms={}", path, flush_ms.as_millis())
+        } else {
+            path
+        };
+        self.request(Method::PUT, &path, &req_body).await?;
         Ok(())
     }
 
@@ -290,21 +278,6 @@ impl Client {
             .map_err(Error::ReceiveBody)
     }
 
-    pub async fn get_location_config(
-        &self,
-        tenant_shard_id: TenantShardId,
-    ) -> Result<Option<LocationConfig>> {
-        let path = format!(
-            "{}/v1/location_config/{tenant_shard_id}",
-            self.mgmt_api_endpoint
-        );
-        self.request(Method::GET, &path, ())
-            .await?
-            .json()
-            .await
-            .map_err(Error::ReceiveBody)
-    }
-
     pub async fn timeline_create(
         &self,
         tenant_shard_id: TenantShardId,

pageserver/compaction/src/compact_tiered.rs

@@ -63,7 +63,7 @@ pub async fn compact_tiered<E: CompactionJobExecutor>(
         );
 
         // Identify the range of LSNs that belong to this level. We assume that
-        // each file in this level spans an LSN range up to 1.75x target file
+        // each file in this level span an LSN range up to 1.75x target file
         // size. That should give us enough slop that if we created a slightly
         // oversized L0 layer, e.g. because flushing the in-memory layer was
         // delayed for some reason, we don't consider the oversized layer to
@@ -248,6 +248,7 @@ enum CompactionStrategy {
     CreateImage,
 }
 
+#[allow(dead_code)] // Todo
 struct CompactionJob<E: CompactionJobExecutor> {
     key_range: Range<E::Key>,
     lsn_range: Range<Lsn>,
@@ -344,7 +345,7 @@ where
     ///
     /// TODO: Currently, this is called exactly once for the level, and we
     /// decide whether to create new image layers to cover the whole level, or
-    /// write a new set of deltas. In the future, this should try to partition
+    /// write a new set of delta. In the future, this should try to partition
     /// the key space, and make the decision separately for each partition.
     async fn divide_job(&mut self, job_id: JobId, ctx: &E::RequestContext) -> anyhow::Result<()> {
         let job = &self.jobs[job_id.0];
@@ -708,6 +709,18 @@ where
     }
 }
 
+// Sliding window through keyspace and values
+//
+// This is used to decide what layer to write next, from the beginning of the window.
+//
+// Candidates:
+//
+// 1. Create an image layer, snapping to previous images
+// 2. Create a delta layer, snapping to previous images
+// 3. Create an image layer, snapping to
+//
+//
+
 // Take previous partitioning, based on the image layers below.
 //
 // Candidate is at the front:
@@ -726,10 +739,6 @@ struct WindowElement<K> {
     last_key: K,  // inclusive
     accum_size: u64,
 }
-
-// Sliding window through keyspace and values
-//
-// This is used to decide what layer to write next, from the beginning of the window.
 struct Window<K> {
     elems: VecDeque<WindowElement<K>>,
 

pageserver/compaction/src/identify_levels.rs

@@ -1,5 +1,5 @@
-//! An LSM tree consists of multiple levels, each exponentially larger than the
-//! previous level. And each level consists of multiple "tiers". With tiered
+//! An LSM tree consists of multiple levels, each exponential larger than the
+//! previous level. And each level consists of be multiple "tiers". With tiered
 //! compaction, a level is compacted when it has accumulated more than N tiers,
 //! forming one tier on the next level.
 //!
@@ -170,6 +170,13 @@ where
     })
 }
 
+// helper struct used in depth()
+struct Event<K> {
+    key: K,
+    layer_idx: usize,
+    start: bool,
+}
+
 impl<L> Level<L> {
     /// Count the number of deltas stacked on each other.
     pub fn depth<K>(&self) -> u64
@@ -177,11 +184,6 @@ impl<L> Level<L> {
         K: CompactionKey,
         L: CompactionLayer<K>,
     {
-        struct Event<K> {
-            key: K,
-            layer_idx: usize,
-            start: bool,
-        }
         let mut events: Vec<Event<K>> = Vec::new();
         for (idx, l) in self.layers.iter().enumerate() {
             events.push(Event {
@@ -200,7 +202,7 @@ impl<L> Level<L> {
         // Sweep the key space left to right. Stop at each distinct key, and
         // count the number of deltas on top of the highest image at that key.
         //
-        // This is a little inefficient, as we walk through the active_set on
+        // This is a little enefficient, as we walk through the active_set on
         // every key. We could increment/decrement a counter on each step
         // instead, but that'd require a bit more complex bookkeeping.
         let mut active_set: BTreeSet<(Lsn, bool, usize)> = BTreeSet::new();
@@ -234,7 +236,6 @@ impl<L> Level<L> {
                 }
             }
         }
-        debug_assert_eq!(active_set, BTreeSet::new());
         max_depth
     }
 }

pageserver/compaction/src/interface.rs

@@ -4,12 +4,12 @@
 //! All the heavy lifting is done by the create_image and create_delta
 //! functions that the implementor provides.
 use async_trait::async_trait;
-use futures::Future;
 use pageserver_api::{key::Key, keyspace::key_range_size};
 use std::ops::Range;
 use utils::lsn::Lsn;
 
 /// Public interface. This is the main thing that the implementor needs to provide
+#[async_trait]
 pub trait CompactionJobExecutor {
     // Type system.
     //
@@ -17,7 +17,8 @@ pub trait CompactionJobExecutor {
     // compaction doesn't distinguish whether they are stored locally or
     // remotely.
     //
-    // The keyspace is defined by the CompactionKey trait.
+    // The keyspace is defined by CompactionKey trait.
+    //
     type Key: CompactionKey;
 
     type Layer: CompactionLayer<Self::Key> + Clone;
@@ -34,27 +35,27 @@ pub trait CompactionJobExecutor {
     // ----
 
     /// Return all layers that overlap the given bounding box.
-    fn get_layers(
+    async fn get_layers(
         &mut self,
         key_range: &Range<Self::Key>,
         lsn_range: &Range<Lsn>,
         ctx: &Self::RequestContext,
-    ) -> impl Future<Output = anyhow::Result<Vec<Self::Layer>>> + Send;
+    ) -> anyhow::Result<Vec<Self::Layer>>;
 
-    fn get_keyspace(
+    async fn get_keyspace(
         &mut self,
         key_range: &Range<Self::Key>,
         lsn: Lsn,
         ctx: &Self::RequestContext,
-    ) -> impl Future<Output = anyhow::Result<CompactionKeySpace<Self::Key>>> + Send;
+    ) -> anyhow::Result<CompactionKeySpace<Self::Key>>;
 
     /// NB: This is a pretty expensive operation. In the real pageserver
     /// implementation, it downloads the layer, and keeps it resident
     /// until the DeltaLayer is dropped.
-    fn downcast_delta_layer(
+    async fn downcast_delta_layer(
         &self,
         layer: &Self::Layer,
-    ) -> impl Future<Output = anyhow::Result<Option<Self::DeltaLayer>>> + Send;
+    ) -> anyhow::Result<Option<Self::DeltaLayer>>;
 
     // ----
     // Functions to execute the plan
@@ -62,33 +63,33 @@ pub trait CompactionJobExecutor {
 
     /// Create a new image layer, materializing all the values in the key range,
     /// at given 'lsn'.
-    fn create_image(
+    async fn create_image(
         &mut self,
         lsn: Lsn,
         key_range: &Range<Self::Key>,
         ctx: &Self::RequestContext,
-    ) -> impl Future<Output = anyhow::Result<()>> + Send;
+    ) -> anyhow::Result<()>;
 
     /// Create a new delta layer, containing all the values from 'input_layers'
     /// in the given key and LSN range.
-    fn create_delta(
+    async fn create_delta(
         &mut self,
         lsn_range: &Range<Lsn>,
         key_range: &Range<Self::Key>,
         input_layers: &[Self::DeltaLayer],
         ctx: &Self::RequestContext,
-    ) -> impl Future<Output = anyhow::Result<()>> + Send;
+    ) -> anyhow::Result<()>;
 
     /// Delete a layer. The compaction implementation will call this only after
     /// all the create_image() or create_delta() calls that deletion of this
     /// layer depends on have finished. But if the implementor has extra lazy
-    /// background tasks, like uploading the index json file to remote storage.
+    /// background tasks, like uploading the index json file to remote storage,
     /// it is the implementation's responsibility to track those.
-    fn delete_layer(
+    async fn delete_layer(
         &mut self,
         layer: &Self::Layer,
         ctx: &Self::RequestContext,
-    ) -> impl Future<Output = anyhow::Result<()>> + Send;
+    ) -> anyhow::Result<()>;
 }
 
 pub trait CompactionKey: std::cmp::Ord + Clone + Copy + std::fmt::Display {

pageserver/compaction/src/simulator.rs

@@ -429,6 +429,7 @@ impl From<&Arc<MockImageLayer>> for MockLayer {
     }
 }
 
+#[async_trait]
 impl interface::CompactionJobExecutor for MockTimeline {
     type Key = Key;
     type Layer = MockLayer;

pageserver/src/context.rs

@@ -88,16 +88,13 @@
 
 use crate::task_mgr::TaskKind;
 
-pub(crate) mod optional_counter;
-
 // The main structure of this module, see module-level comment.
-#[derive(Debug)]
+#[derive(Clone, Debug)]
 pub struct RequestContext {
     task_kind: TaskKind,
     download_behavior: DownloadBehavior,
     access_stats_behavior: AccessStatsBehavior,
     page_content_kind: PageContentKind,
-    pub micros_spent_throttled: optional_counter::MicroSecondsCounterU32,
 }
 
 /// The kind of access to the page cache.
@@ -153,7 +150,6 @@ impl RequestContextBuilder {
                 download_behavior: DownloadBehavior::Download,
                 access_stats_behavior: AccessStatsBehavior::Update,
                 page_content_kind: PageContentKind::Unknown,
-                micros_spent_throttled: Default::default(),
             },
         }
     }
@@ -167,7 +163,6 @@ impl RequestContextBuilder {
                 download_behavior: original.download_behavior,
                 access_stats_behavior: original.access_stats_behavior,
                 page_content_kind: original.page_content_kind,
-                micros_spent_throttled: Default::default(),
             },
         }
     }

pageserver/src/context/optional_counter.rs

@@ -1,101 +0,0 @@
-use std::{
-    sync::atomic::{AtomicU32, Ordering},
-    time::Duration,
-};
-
-#[derive(Debug)]
-pub struct CounterU32 {
-    inner: AtomicU32,
-}
-impl Default for CounterU32 {
-    fn default() -> Self {
-        Self {
-            inner: AtomicU32::new(u32::MAX),
-        }
-    }
-}
-impl CounterU32 {
-    pub fn open(&self) -> Result<(), &'static str> {
-        match self
-            .inner
-            .compare_exchange(u32::MAX, 0, Ordering::Relaxed, Ordering::Relaxed)
-        {
-            Ok(_) => Ok(()),
-            Err(_) => Err("open() called on clsoed state"),
-        }
-    }
-    pub fn close(&self) -> Result<u32, &'static str> {
-        match self.inner.swap(u32::MAX, Ordering::Relaxed) {
-            u32::MAX => Err("close() called on closed state"),
-            x => Ok(x),
-        }
-    }
-
-    pub fn add(&self, count: u32) -> Result<(), &'static str> {
-        if count == 0 {
-            return Ok(());
-        }
-        let mut had_err = None;
-        self.inner
-            .fetch_update(Ordering::Relaxed, Ordering::Relaxed, |cur| match cur {
-                u32::MAX => {
-                    had_err = Some("add() called on closed state");
-                    None
-                }
-                x => {
-                    let (new, overflowed) = x.overflowing_add(count);
-                    if new == u32::MAX || overflowed {
-                        had_err = Some("add() overflowed the counter");
-                        None
-                    } else {
-                        Some(new)
-                    }
-                }
-            })
-            .map_err(|_| had_err.expect("we set it whenever the function returns None"))
-            .map(|_| ())
-    }
-}
-
-#[derive(Default, Debug)]
-pub struct MicroSecondsCounterU32 {
-    inner: CounterU32,
-}
-
-impl MicroSecondsCounterU32 {
-    pub fn open(&self) -> Result<(), &'static str> {
-        self.inner.open()
-    }
-    pub fn add(&self, duration: Duration) -> Result<(), &'static str> {
-        match duration.as_micros().try_into() {
-            Ok(x) => self.inner.add(x),
-            Err(_) => Err("add(): duration conversion error"),
-        }
-    }
-    pub fn close_and_checked_sub_from(&self, from: Duration) -> Result<Duration, &'static str> {
-        let val = self.inner.close()?;
-        let val = Duration::from_micros(val as u64);
-        let subbed = match from.checked_sub(val) {
-            Some(v) => v,
-            None => return Err("Duration::checked_sub"),
-        };
-        Ok(subbed)
-    }
-}
-
-#[cfg(test)]
-mod tests {
-
-    use super::*;
-
-    #[test]
-    fn test_basic() {
-        let counter = MicroSecondsCounterU32::default();
-        counter.open().unwrap();
-        counter.add(Duration::from_micros(23)).unwrap();
-        let res = counter
-            .close_and_checked_sub_from(Duration::from_micros(42))
-            .unwrap();
-        assert_eq!(res, Duration::from_micros(42 - 23));
-    }
-}

pageserver/src/http/openapi_spec.yml

@@ -1339,10 +1339,6 @@ components:
           type: array
           items:
             $ref: "#/components/schemas/TenantShardLocation"
-        stripe_size:
-          description: If multiple shards are present, this field contains the sharding stripe size, else it is null.
-          type: integer
-          nullable: true
     TenantShardLocation:
       type: object
       required:

pageserver/src/http/routes.rs

@@ -14,7 +14,6 @@ use hyper::header;
 use hyper::StatusCode;
 use hyper::{Body, Request, Response, Uri};
 use metrics::launch_timestamp::LaunchTimestamp;
-use pageserver_api::models::LocationConfig;
 use pageserver_api::models::LocationConfigListResponse;
 use pageserver_api::models::ShardParameters;
 use pageserver_api::models::TenantDetails;
@@ -1452,12 +1451,11 @@ async fn put_tenant_location_config_handler(
         tenant::SpawnMode::Eager
     };
 
-    let tenant = state
+    let attached = state
         .tenant_manager
         .upsert_location(tenant_shard_id, location_conf, flush, spawn_mode, &ctx)
-        .await?;
-    let stripe_size = tenant.as_ref().map(|t| t.get_shard_stripe_size());
-    let attached = tenant.is_some();
+        .await?
+        .is_some();
 
     if let Some(_flush_ms) = flush {
         match state
@@ -1479,20 +1477,12 @@ async fn put_tenant_location_config_handler(
     // This API returns a vector of pageservers where the tenant is attached: this is
     // primarily for use in the sharding service.  For compatibilty, we also return this
     // when called directly on a pageserver, but the payload is always zero or one shards.
-    let mut response = TenantLocationConfigResponse {
-        shards: Vec::new(),
-        stripe_size: None,
-    };
+    let mut response = TenantLocationConfigResponse { shards: Vec::new() };
     if attached {
         response.shards.push(TenantShardLocation {
             shard_id: tenant_shard_id,
             node_id: state.conf.id,
-        });
-        if tenant_shard_id.shard_count.count() > 1 {
-            // Stripe size should be set if we are attached
-            debug_assert!(stripe_size.is_some());
-            response.stripe_size = stripe_size;
-        }
+        })
     }
 
     json_response(StatusCode::OK, response)
@@ -1520,29 +1510,6 @@ async fn list_location_config_handler(
     json_response(StatusCode::OK, result)
 }
 
-async fn get_location_config_handler(
-    request: Request<Body>,
-    _cancel: CancellationToken,
-) -> Result<Response<Body>, ApiError> {
-    let state = get_state(&request);
-    let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
-    let slot = state.tenant_manager.get(tenant_shard_id);
-
-    let Some(slot) = slot else {
-        return Err(ApiError::NotFound(
-            anyhow::anyhow!("Tenant shard not found").into(),
-        ));
-    };
-
-    let result: Option<LocationConfig> = match slot {
-        TenantSlot::Attached(t) => Some(t.get_location_conf()),
-        TenantSlot::Secondary(s) => Some(s.get_location_conf()),
-        TenantSlot::InProgress(_) => None,
-    };
-
-    json_response(StatusCode::OK, result)
-}
-
 // Do a time travel recovery on the given tenant/tenant shard. Tenant needs to be detached
 // (from all pageservers) as it invalidates consistency assumptions.
 async fn tenant_time_travel_remote_storage_handler(
@@ -2247,9 +2214,6 @@ pub fn make_router(
         .get("/v1/location_config", |r| {
             api_handler(r, list_location_config_handler)
         })
-        .get("/v1/location_config/:tenant_id", |r| {
-            api_handler(r, get_location_config_handler)
-        })
         .put(
             "/v1/tenant/:tenant_shard_id/time_travel_remote_storage",
             |r| api_handler(r, tenant_time_travel_remote_storage_handler),

pageserver/src/metrics.rs

@@ -11,7 +11,6 @@ use once_cell::sync::Lazy;
 use pageserver_api::shard::TenantShardId;
 use strum::{EnumCount, IntoEnumIterator, VariantNames};
 use strum_macros::{EnumVariantNames, IntoStaticStr};
-use tracing::warn;
 use utils::id::TimelineId;
 
 /// Prometheus histogram buckets (in seconds) for operations in the critical
@@ -1006,39 +1005,15 @@ impl GlobalAndPerTimelineHistogram {
     }
 }
 
-struct GlobalAndPerTimelineHistogramTimer<'a, 'c> {
+struct GlobalAndPerTimelineHistogramTimer<'a> {
     h: &'a GlobalAndPerTimelineHistogram,
-    ctx: &'c RequestContext,
     start: std::time::Instant,
-    op: SmgrQueryType,
 }
 
-impl<'a, 'c> Drop for GlobalAndPerTimelineHistogramTimer<'a, 'c> {
+impl<'a> Drop for GlobalAndPerTimelineHistogramTimer<'a> {
     fn drop(&mut self) {
         let elapsed = self.start.elapsed();
-        let ex_throttled = self
-            .ctx
-            .micros_spent_throttled
-            .close_and_checked_sub_from(elapsed);
-        let ex_throttled = match ex_throttled {
-            Ok(res) => res,
-            Err(error) => {
-                use utils::rate_limit::RateLimit;
-                static LOGGED: Lazy<Mutex<enum_map::EnumMap<SmgrQueryType, RateLimit>>> =
-                    Lazy::new(|| {
-                        Mutex::new(enum_map::EnumMap::from_array(std::array::from_fn(|_| {
-                            RateLimit::new(Duration::from_secs(10))
-                        })))
-                    });
-                let mut guard = LOGGED.lock().unwrap();
-                let rate_limit = &mut guard[self.op];
-                rate_limit.call(|| {
-                    warn!(op=?self.op, error, "error deducting time spent throttled; this message is logged at a global rate limit");
-                });
-                elapsed
-            }
-        };
-        self.h.observe(ex_throttled.as_secs_f64());
+        self.h.observe(elapsed.as_secs_f64());
     }
 }
 
@@ -1050,7 +1025,6 @@ impl<'a, 'c> Drop for GlobalAndPerTimelineHistogramTimer<'a, 'c> {
     strum_macros::EnumCount,
     strum_macros::EnumIter,
     strum_macros::FromRepr,
-    enum_map::Enum,
 )]
 #[strum(serialize_all = "snake_case")]
 pub enum SmgrQueryType {
@@ -1156,35 +1130,11 @@ impl SmgrQueryTimePerTimeline {
         });
         Self { metrics }
     }
-    pub(crate) fn start_timer<'c: 'a, 'a>(
-        &'a self,
-        op: SmgrQueryType,
-        ctx: &'c RequestContext,
-    ) -> impl Drop + '_ {
+    pub(crate) fn start_timer(&self, op: SmgrQueryType) -> impl Drop + '_ {
         let metric = &self.metrics[op as usize];
-        let start = Instant::now();
-        match ctx.micros_spent_throttled.open() {
-            Ok(()) => (),
-            Err(error) => {
-                use utils::rate_limit::RateLimit;
-                static LOGGED: Lazy<Mutex<enum_map::EnumMap<SmgrQueryType, RateLimit>>> =
-                    Lazy::new(|| {
-                        Mutex::new(enum_map::EnumMap::from_array(std::array::from_fn(|_| {
-                            RateLimit::new(Duration::from_secs(10))
-                        })))
-                    });
-                let mut guard = LOGGED.lock().unwrap();
-                let rate_limit = &mut guard[op];
-                rate_limit.call(|| {
-                    warn!(?op, error, "error opening micros_spent_throttled; this message is logged at a global rate limit");
-                });
-            }
-        }
         GlobalAndPerTimelineHistogramTimer {
             h: metric,
-            ctx,
-            start,
-            op,
+            start: std::time::Instant::now(),
         }
     }
 }
@@ -1195,11 +1145,6 @@ mod smgr_query_time_tests {
     use strum::IntoEnumIterator;
     use utils::id::{TenantId, TimelineId};
 
-    use crate::{
-        context::{DownloadBehavior, RequestContext},
-        task_mgr::TaskKind,
-    };
-
     // Regression test, we used hard-coded string constants before using an enum.
     #[test]
     fn op_label_name() {
@@ -1248,8 +1193,7 @@ mod smgr_query_time_tests {
             let (pre_global, pre_per_tenant_timeline) = get_counts();
             assert_eq!(pre_per_tenant_timeline, 0);
 
-            let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Download);
-            let timer = metrics.start_timer(*op, &ctx);
+            let timer = metrics.start_timer(*op);
             drop(timer);
 
             let (post_global, post_per_tenant_timeline) = get_counts();

pageserver/src/page_service.rs

@@ -910,7 +910,7 @@ impl PageServerHandler {
         let timeline = self.get_timeline_shard_zero(tenant_id, timeline_id).await?;
         let _timer = timeline
             .query_metrics
-            .start_timer(metrics::SmgrQueryType::GetRelExists, ctx);
+            .start_timer(metrics::SmgrQueryType::GetRelExists);
 
         let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         let lsn =
@@ -938,7 +938,7 @@ impl PageServerHandler {
 
         let _timer = timeline
             .query_metrics
-            .start_timer(metrics::SmgrQueryType::GetRelSize, ctx);
+            .start_timer(metrics::SmgrQueryType::GetRelSize);
 
         let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         let lsn =
@@ -966,7 +966,7 @@ impl PageServerHandler {
 
         let _timer = timeline
             .query_metrics
-            .start_timer(metrics::SmgrQueryType::GetDbSize, ctx);
+            .start_timer(metrics::SmgrQueryType::GetDbSize);
 
         let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         let lsn =
@@ -1144,7 +1144,7 @@ impl PageServerHandler {
 
         let _timer = timeline
             .query_metrics
-            .start_timer(metrics::SmgrQueryType::GetPageAtLsn, ctx);
+            .start_timer(metrics::SmgrQueryType::GetPageAtLsn);
 
         let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         let lsn =
@@ -1172,7 +1172,7 @@ impl PageServerHandler {
 
         let _timer = timeline
             .query_metrics
-            .start_timer(metrics::SmgrQueryType::GetSlruSegment, ctx);
+            .start_timer(metrics::SmgrQueryType::GetSlruSegment);
 
         let latest_gc_cutoff_lsn = timeline.get_latest_gc_cutoff_lsn();
         let lsn =

pageserver/src/pgdatadir_mapping.rs

@@ -15,7 +15,6 @@ use crate::walrecord::NeonWalRecord;
 use anyhow::{ensure, Context};
 use bytes::{Buf, Bytes, BytesMut};
 use enum_map::Enum;
-use itertools::Itertools;
 use pageserver_api::key::{
     dbdir_key_range, is_rel_block_key, is_slru_block_key, rel_block_to_key, rel_dir_to_key,
     rel_key_range, rel_size_to_key, relmap_file_key, slru_block_to_key, slru_dir_to_key,
@@ -1499,7 +1498,7 @@ impl<'a> DatadirModification<'a> {
             return Ok(());
         }
 
-        let mut writer = self.tline.writer().await;
+        let writer = self.tline.writer().await;
 
         // Flush relation and  SLRU data blocks, keep metadata.
         let mut retained_pending_updates = HashMap::<_, Vec<_>>::new();
@@ -1538,22 +1537,14 @@ impl<'a> DatadirModification<'a> {
     /// All the modifications in this atomic update are stamped by the specified LSN.
     ///
     pub async fn commit(&mut self, ctx: &RequestContext) -> anyhow::Result<()> {
-        let mut writer = self.tline.writer().await;
+        let writer = self.tline.writer().await;
 
         let pending_nblocks = self.pending_nblocks;
         self.pending_nblocks = 0;
 
         if !self.pending_updates.is_empty() {
-            // The put_batch call below expects expects the inputs to be sorted by Lsn,
-            // so we do that first.
-            let lsn_ordered_batch: Vec<(Key, Lsn, Value)> = self
-                .pending_updates
-                .drain()
-                .map(|(key, vals)| vals.into_iter().map(move |(lsn, val)| (key, lsn, val)))
-                .kmerge_by(|lhs, rhs| lhs.1 .0 < rhs.1 .0)
-                .collect();
-
-            writer.put_batch(lsn_ordered_batch, ctx).await?;
+            writer.put_batch(&self.pending_updates, ctx).await?;
+            self.pending_updates.clear();
         }
 
         if !self.pending_deletions.is_empty() {
@@ -1686,7 +1677,7 @@ struct RelDirectory {
     rels: HashSet<(Oid, u8)>,
 }
 
-#[derive(Debug, Serialize, Deserialize, Default, PartialEq)]
+#[derive(Debug, Serialize, Deserialize, Default)]
 pub(crate) struct AuxFilesDirectory {
     pub(crate) files: HashMap<String, Bytes>,
 }

pageserver/src/tenant.rs

@@ -22,7 +22,6 @@ use pageserver_api::models;
 use pageserver_api::models::TimelineState;
 use pageserver_api::models::WalRedoManagerStatus;
 use pageserver_api::shard::ShardIdentity;
-use pageserver_api::shard::ShardStripeSize;
 use pageserver_api::shard::TenantShardId;
 use remote_storage::DownloadError;
 use remote_storage::GenericRemoteStorage;
@@ -2087,10 +2086,6 @@ impl Tenant {
         &self.tenant_shard_id
     }
 
-    pub(crate) fn get_shard_stripe_size(&self) -> ShardStripeSize {
-        self.shard_identity.stripe_size
-    }
-
     pub(crate) fn get_generation(&self) -> Generation {
         self.generation
     }
@@ -3679,10 +3674,7 @@ pub(crate) mod harness {
     }
 
     impl TenantHarness {
-        pub fn create_custom(
-            test_name: &'static str,
-            tenant_conf: TenantConf,
-        ) -> anyhow::Result<Self> {
+        pub fn create(test_name: &'static str) -> anyhow::Result<Self> {
             setup_logging();
 
             let repo_dir = PageServerConf::test_repo_dir(test_name);
@@ -3694,6 +3686,14 @@ pub(crate) mod harness {
             // OK in a test.
             let conf: &'static PageServerConf = Box::leak(Box::new(conf));
 
+            // Disable automatic GC and compaction to make the unit tests more deterministic.
+            // The tests perform them manually if needed.
+            let tenant_conf = TenantConf {
+                gc_period: Duration::ZERO,
+                compaction_period: Duration::ZERO,
+                ..TenantConf::default()
+            };
+
             let tenant_id = TenantId::generate();
             let tenant_shard_id = TenantShardId::unsharded(tenant_id);
             fs::create_dir_all(conf.tenant_path(&tenant_shard_id))?;
@@ -3721,18 +3721,6 @@ pub(crate) mod harness {
             })
         }
 
-        pub fn create(test_name: &'static str) -> anyhow::Result<Self> {
-            // Disable automatic GC and compaction to make the unit tests more deterministic.
-            // The tests perform them manually if needed.
-            let tenant_conf = TenantConf {
-                gc_period: Duration::ZERO,
-                compaction_period: Duration::ZERO,
-                ..TenantConf::default()
-            };
-
-            Self::create_custom(test_name, tenant_conf)
-        }
-
         pub fn span(&self) -> tracing::Span {
             info_span!("TenantHarness", tenant_id=%self.tenant_shard_id.tenant_id, shard_id=%self.tenant_shard_id.shard_slug())
         }
@@ -3840,7 +3828,6 @@ mod tests {
     use crate::keyspace::KeySpaceAccum;
     use crate::repository::{Key, Value};
     use crate::tenant::harness::*;
-    use crate::tenant::timeline::CompactFlags;
     use crate::DEFAULT_PG_VERSION;
     use bytes::BytesMut;
     use hex_literal::hex;
@@ -3857,7 +3844,7 @@ mod tests {
             .create_test_timeline(TIMELINE_ID, Lsn(0x08), DEFAULT_PG_VERSION, &ctx)
             .await?;
 
-        let mut writer = tline.writer().await;
+        let writer = tline.writer().await;
         writer
             .put(
                 *TEST_KEY,
@@ -3869,7 +3856,7 @@ mod tests {
         writer.finish_write(Lsn(0x10));
         drop(writer);
 
-        let mut writer = tline.writer().await;
+        let writer = tline.writer().await;
         writer
             .put(
                 *TEST_KEY,
@@ -3935,7 +3922,7 @@ mod tests {
         let tline = tenant
             .create_test_timeline(TIMELINE_ID, Lsn(0x10), DEFAULT_PG_VERSION, &ctx)
             .await?;
-        let mut writer = tline.writer().await;
+        let writer = tline.writer().await;
 
         #[allow(non_snake_case)]
         let TEST_KEY_A: Key = Key::from_hex("110000000033333333444444445500000001").unwrap();
@@ -3969,7 +3956,7 @@ mod tests {
         let newtline = tenant
             .get_timeline(NEW_TIMELINE_ID, true)
             .expect("Should have a local timeline");
-        let mut new_writer = newtline.writer().await;
+        let new_writer = newtline.writer().await;
         new_writer
             .put(TEST_KEY_A, Lsn(0x40), &test_value("bar at 0x40"), &ctx)
             .await?;
@@ -4001,7 +3988,7 @@ mod tests {
     ) -> anyhow::Result<()> {
         let mut lsn = start_lsn;
         {
-            let mut writer = tline.writer().await;
+            let writer = tline.writer().await;
             // Create a relation on the timeline
             writer
                 .put(
@@ -4026,7 +4013,7 @@ mod tests {
         }
         tline.freeze_and_flush().await?;
         {
-            let mut writer = tline.writer().await;
+            let writer = tline.writer().await;
             writer
                 .put(
                     *TEST_KEY,
@@ -4389,7 +4376,7 @@ mod tests {
             .create_test_timeline(TIMELINE_ID, Lsn(0x08), DEFAULT_PG_VERSION, &ctx)
             .await?;
 
-        let mut writer = tline.writer().await;
+        let writer = tline.writer().await;
         writer
             .put(
                 *TEST_KEY,
@@ -4406,7 +4393,7 @@ mod tests {
             .compact(&CancellationToken::new(), EnumSet::empty(), &ctx)
             .await?;
 
-        let mut writer = tline.writer().await;
+        let writer = tline.writer().await;
         writer
             .put(
                 *TEST_KEY,
@@ -4423,7 +4410,7 @@ mod tests {
             .compact(&CancellationToken::new(), EnumSet::empty(), &ctx)
             .await?;
 
-        let mut writer = tline.writer().await;
+        let writer = tline.writer().await;
         writer
             .put(
                 *TEST_KEY,
@@ -4440,7 +4427,7 @@ mod tests {
             .compact(&CancellationToken::new(), EnumSet::empty(), &ctx)
             .await?;
 
-        let mut writer = tline.writer().await;
+        let writer = tline.writer().await;
         writer
             .put(
                 *TEST_KEY,
@@ -4497,7 +4484,7 @@ mod tests {
         for _ in 0..repeat {
             for _ in 0..key_count {
                 test_key.field6 = blknum;
-                let mut writer = timeline.writer().await;
+                let writer = timeline.writer().await;
                 writer
                     .put(
                         test_key,
@@ -4645,145 +4632,6 @@ mod tests {
         Ok(())
     }
 
-    // Test that vectored get handles layer gaps correctly
-    // by advancing into the next ancestor timeline if required.
-    //
-    // The test generates timelines that look like the diagram below.
-    // We leave a gap in one of the L1 layers at `gap_at_key` (`/` in the diagram).
-    // The reconstruct data for that key lies in the ancestor timeline (`X` in the diagram).
-    //
-    // ```
-    //-------------------------------+
-    //                          ...  |
-    //               [   L1   ]      |
-    //     [ / L1   ]                | Child Timeline
-    // ...                           |
-    // ------------------------------+
-    //     [ X L1   ]                | Parent Timeline
-    // ------------------------------+
-    // ```
-    #[tokio::test]
-    async fn test_get_vectored_key_gap() -> anyhow::Result<()> {
-        let tenant_conf = TenantConf {
-            // Make compaction deterministic
-            gc_period: Duration::ZERO,
-            compaction_period: Duration::ZERO,
-            // Encourage creation of L1 layers
-            checkpoint_distance: 16 * 1024,
-            compaction_target_size: 8 * 1024,
-            ..TenantConf::default()
-        };
-
-        let harness = TenantHarness::create_custom("test_get_vectored_key_gap", tenant_conf)?;
-        let (tenant, ctx) = harness.load().await;
-
-        let mut current_key = Key::from_hex("010000000033333333444444445500000000").unwrap();
-        let gap_at_key = current_key.add(100);
-        let mut current_lsn = Lsn(0x10);
-
-        const KEY_COUNT: usize = 10_000;
-
-        let timeline_id = TimelineId::generate();
-        let current_timeline = tenant
-            .create_test_timeline(timeline_id, current_lsn, DEFAULT_PG_VERSION, &ctx)
-            .await?;
-
-        current_lsn += 0x100;
-
-        let mut writer = current_timeline.writer().await;
-        writer
-            .put(
-                gap_at_key,
-                current_lsn,
-                &Value::Image(test_img(&format!("{} at {}", gap_at_key, current_lsn))),
-                &ctx,
-            )
-            .await?;
-        writer.finish_write(current_lsn);
-        drop(writer);
-
-        let mut latest_lsns = HashMap::new();
-        latest_lsns.insert(gap_at_key, current_lsn);
-
-        current_timeline.freeze_and_flush().await?;
-
-        let child_timeline_id = TimelineId::generate();
-
-        tenant
-            .branch_timeline_test(
-                &current_timeline,
-                child_timeline_id,
-                Some(current_lsn),
-                &ctx,
-            )
-            .await?;
-        let child_timeline = tenant
-            .get_timeline(child_timeline_id, true)
-            .expect("Should have the branched timeline");
-
-        for i in 0..KEY_COUNT {
-            if current_key == gap_at_key {
-                current_key = current_key.next();
-                continue;
-            }
-
-            current_lsn += 0x10;
-
-            let mut writer = child_timeline.writer().await;
-            writer
-                .put(
-                    current_key,
-                    current_lsn,
-                    &Value::Image(test_img(&format!("{} at {}", current_key, current_lsn))),
-                    &ctx,
-                )
-                .await?;
-            writer.finish_write(current_lsn);
-            drop(writer);
-
-            latest_lsns.insert(current_key, current_lsn);
-            current_key = current_key.next();
-
-            // Flush every now and then to encourage layer file creation.
-            if i % 500 == 0 {
-                child_timeline.freeze_and_flush().await?;
-            }
-        }
-
-        child_timeline.freeze_and_flush().await?;
-        let mut flags = EnumSet::new();
-        flags.insert(CompactFlags::ForceRepartition);
-        child_timeline
-            .compact(&CancellationToken::new(), flags, &ctx)
-            .await?;
-
-        let key_near_end = {
-            let mut tmp = current_key;
-            tmp.field6 -= 10;
-            tmp
-        };
-
-        let key_near_gap = {
-            let mut tmp = gap_at_key;
-            tmp.field6 -= 10;
-            tmp
-        };
-
-        let read = KeySpace {
-            ranges: vec![key_near_gap..gap_at_key.next(), key_near_end..current_key],
-        };
-        let results = child_timeline
-            .get_vectored_impl(read.clone(), current_lsn, &ctx)
-            .await?;
-
-        for (key, img_res) in results {
-            let expected = test_img(&format!("{} at {}", key, latest_lsns[&key]));
-            assert_eq!(img_res?, expected);
-        }
-
-        Ok(())
-    }
-
     #[tokio::test]
     async fn test_random_updates() -> anyhow::Result<()> {
         let harness = TenantHarness::create("test_random_updates")?;
@@ -4807,7 +4655,7 @@ mod tests {
         for blknum in 0..NUM_KEYS {
             lsn = Lsn(lsn.0 + 0x10);
             test_key.field6 = blknum as u32;
-            let mut writer = tline.writer().await;
+            let writer = tline.writer().await;
             writer
                 .put(
                     test_key,
@@ -4828,7 +4676,7 @@ mod tests {
                 lsn = Lsn(lsn.0 + 0x10);
                 let blknum = thread_rng().gen_range(0..NUM_KEYS);
                 test_key.field6 = blknum as u32;
-                let mut writer = tline.writer().await;
+                let writer = tline.writer().await;
                 writer
                     .put(
                         test_key,
@@ -4896,7 +4744,7 @@ mod tests {
         for blknum in 0..NUM_KEYS {
             lsn = Lsn(lsn.0 + 0x10);
             test_key.field6 = blknum as u32;
-            let mut writer = tline.writer().await;
+            let writer = tline.writer().await;
             writer
                 .put(
                     test_key,
@@ -4925,7 +4773,7 @@ mod tests {
                 lsn = Lsn(lsn.0 + 0x10);
                 let blknum = thread_rng().gen_range(0..NUM_KEYS);
                 test_key.field6 = blknum as u32;
-                let mut writer = tline.writer().await;
+                let writer = tline.writer().await;
                 writer
                     .put(
                         test_key,
@@ -5002,7 +4850,7 @@ mod tests {
                 lsn = Lsn(lsn.0 + 0x10);
                 let blknum = thread_rng().gen_range(0..NUM_KEYS);
                 test_key.field6 = blknum as u32;
-                let mut writer = tline.writer().await;
+                let writer = tline.writer().await;
                 writer
                     .put(
                         test_key,

pageserver/src/tenant/blob_io.rs

@@ -12,7 +12,7 @@
 //! len >= 128: 1XXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX
 //!
 use bytes::{BufMut, BytesMut};
-use tokio_epoll_uring::{BoundedBuf, IoBuf, Slice};
+use tokio_epoll_uring::{BoundedBuf, Slice};
 
 use crate::context::RequestContext;
 use crate::page_cache::PAGE_SZ;
@@ -127,7 +127,7 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
     /// You need to make sure that the internal buffer is empty, otherwise
     /// data will be written in wrong order.
     #[inline(always)]
-    async fn write_all_unbuffered<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
+    async fn write_all_unbuffered<B: BoundedBuf>(
         &mut self,
         src_buf: B,
     ) -> (B::Buf, Result<(), Error>) {
@@ -162,10 +162,7 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
     }
 
     /// Internal, possibly buffered, write function
-    async fn write_all<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
-        &mut self,
-        src_buf: B,
-    ) -> (B::Buf, Result<(), Error>) {
+    async fn write_all<B: BoundedBuf>(&mut self, src_buf: B) -> (B::Buf, Result<(), Error>) {
         if !BUFFERED {
             assert!(self.buf.is_empty());
             return self.write_all_unbuffered(src_buf).await;
@@ -213,10 +210,7 @@ impl<const BUFFERED: bool> BlobWriter<BUFFERED> {
 
     /// Write a blob of data. Returns the offset that it was written to,
     /// which can be used to retrieve the data later.
-    pub async fn write_blob<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
-        &mut self,
-        srcbuf: B,
-    ) -> (B::Buf, Result<u64, Error>) {
+    pub async fn write_blob<B: BoundedBuf>(&mut self, srcbuf: B) -> (B::Buf, Result<u64, Error>) {
         let offset = self.offset;
 
         let len = srcbuf.bytes_init();

pageserver/src/tenant/disk_btree.rs

@@ -18,19 +18,10 @@
 //! - An Iterator interface would be more convenient for the callers than the
 //!   'visit' function
 //!
-use async_stream::try_stream;
 use byteorder::{ReadBytesExt, BE};
 use bytes::{BufMut, Bytes, BytesMut};
 use either::Either;
-use futures::Stream;
-use hex;
-use std::{
-    cmp::Ordering,
-    io,
-    iter::Rev,
-    ops::{Range, RangeInclusive},
-    result,
-};
+use std::{cmp::Ordering, io, result};
 use thiserror::Error;
 use tracing::error;
 
@@ -259,90 +250,6 @@ where
         Ok(result)
     }
 
-    /// Return a stream which yields all key, value pairs from the index
-    /// starting from the first key greater or equal to `start_key`.
-    ///
-    /// Note that this is a copy of [`Self::visit`].
-    /// TODO: Once the sequential read path is removed this will become
-    /// the only index traversal method.
-    pub fn get_stream_from<'a>(
-        &'a self,
-        start_key: &'a [u8; L],
-        ctx: &'a RequestContext,
-    ) -> impl Stream<Item = std::result::Result<(Vec<u8>, u64), DiskBtreeError>> + 'a {
-        try_stream! {
-            let mut stack = Vec::new();
-            stack.push((self.root_blk, None));
-            let block_cursor = self.reader.block_cursor();
-            while let Some((node_blknum, opt_iter)) = stack.pop() {
-                // Locate the node.
-                let node_buf = block_cursor
-                    .read_blk(self.start_blk + node_blknum, ctx)
-                    .await?;
-
-                let node = OnDiskNode::deparse(node_buf.as_ref())?;
-                let prefix_len = node.prefix_len as usize;
-                let suffix_len = node.suffix_len as usize;
-
-                assert!(node.num_children > 0);
-
-                let mut keybuf = Vec::new();
-                keybuf.extend(node.prefix);
-                keybuf.resize(prefix_len + suffix_len, 0);
-
-                let mut iter: Either<Range<usize>, Rev<RangeInclusive<usize>>> = if let Some(iter) = opt_iter {
-                    iter
-                } else {
-                    // Locate the first match
-                    let idx = match node.binary_search(start_key, keybuf.as_mut_slice()) {
-                        Ok(idx) => idx,
-                        Err(idx) => {
-                            if node.level == 0 {
-                                // Imagine that the node contains the following keys:
-                                //
-                                // 1
-                                // 3  <-- idx
-                                // 5
-                                //
-                                // If the search key is '2' and there is exact match,
-                                // the binary search would return the index of key
-                                // '3'. That's cool, '3' is the first key to return.
-                                idx
-                            } else {
-                                // This is an internal page, so each key represents a lower
-                                // bound for what's in the child page. If there is no exact
-                                // match, we have to return the *previous* entry.
-                                //
-                                // 1  <-- return this
-                                // 3  <-- idx
-                                // 5
-                                idx.saturating_sub(1)
-                            }
-                        }
-                    };
-                    Either::Left(idx..node.num_children.into())
-                };
-
-                // idx points to the first match now. Keep going from there
-                while let Some(idx) = iter.next() {
-                    let key_off = idx * suffix_len;
-                    let suffix = &node.keys[key_off..key_off + suffix_len];
-                    keybuf[prefix_len..].copy_from_slice(suffix);
-                    let value = node.value(idx);
-                    #[allow(clippy::collapsible_if)]
-                    if node.level == 0 {
-                        // leaf
-                        yield (keybuf.clone(), value.to_u64());
-                    } else {
-                        stack.push((node_blknum, Some(iter)));
-                        stack.push((value.to_blknum(), None));
-                        break;
-                    }
-                }
-            }
-        }
-    }
-
     ///
     /// Scan the tree, starting from 'search_key', in the given direction. 'visitor'
     /// will be called for every key >= 'search_key' (or <= 'search_key', if scanning

pageserver/src/tenant/layer_map.rs

@@ -460,22 +460,15 @@ impl LayerMap {
         }
     }
 
-    pub fn range_search(&self, key_range: Range<Key>, end_lsn: Lsn) -> RangeSearchResult {
-        let version = match self.historic.get().unwrap().get_version(end_lsn.0 - 1) {
-            Some(version) => version,
-            None => {
-                let mut result = RangeSearchResult::new();
-                result.not_found.add_range(key_range);
-                return result;
-            }
-        };
+    pub fn range_search(&self, key_range: Range<Key>, end_lsn: Lsn) -> Option<RangeSearchResult> {
+        let version = self.historic.get().unwrap().get_version(end_lsn.0 - 1)?;
 
         let raw_range = key_range.start.to_i128()..key_range.end.to_i128();
         let delta_changes = version.delta_coverage.range_overlaps(&raw_range);
         let image_changes = version.image_coverage.range_overlaps(&raw_range);
 
         let collector = RangeSearchCollector::new(key_range, end_lsn, delta_changes, image_changes);
-        collector.collect()
+        Some(collector.collect())
     }
 
     /// Start a batch of updates, applied on drop
@@ -1002,13 +995,8 @@ mod tests {
         let layer_map = LayerMap::default();
         let range = Key::from_i128(100)..Key::from_i128(200);
 
-        let res = layer_map.range_search(range.clone(), Lsn(100));
-        assert_eq!(
-            res.not_found.to_keyspace(),
-            KeySpace {
-                ranges: vec![range]
-            }
-        );
+        let res = layer_map.range_search(range, Lsn(100));
+        assert!(res.is_none());
     }
 
     #[test]
@@ -1045,7 +1033,7 @@ mod tests {
         for start in 0..60 {
             for end in (start + 1)..60 {
                 let range = Key::from_i128(start)..Key::from_i128(end);
-                let result = layer_map.range_search(range.clone(), Lsn(100));
+                let result = layer_map.range_search(range.clone(), Lsn(100)).unwrap();
                 let expected = brute_force_range_search(&layer_map, range, Lsn(100));
 
                 assert_range_search_result_eq(result, expected);

pageserver/src/tenant/mgr.rs

@@ -1358,16 +1358,6 @@ impl TenantManager {
         }
     }
 
-    pub(crate) fn get(&self, tenant_shard_id: TenantShardId) -> Option<TenantSlot> {
-        let locked = self.tenants.read().unwrap();
-        match &*locked {
-            TenantsMap::Initializing => None,
-            TenantsMap::Open(map) | TenantsMap::ShuttingDown(map) => {
-                map.get(&tenant_shard_id).cloned()
-            }
-        }
-    }
-
     pub(crate) async fn delete_tenant(
         &self,
         tenant_shard_id: TenantShardId,

pageserver/src/tenant/remote_timeline_client/download.rs

@@ -14,14 +14,14 @@ use tokio::io::{AsyncSeekExt, AsyncWriteExt};
 use tokio_util::io::StreamReader;
 use tokio_util::sync::CancellationToken;
 use tracing::warn;
-use utils::backoff;
+use utils::{backoff, crashsafe};
 
 use crate::config::PageServerConf;
 use crate::span::debug_assert_current_span_has_tenant_and_timeline_id;
 use crate::tenant::remote_timeline_client::{remote_layer_path, remote_timelines_path};
 use crate::tenant::storage_layer::LayerFileName;
 use crate::tenant::Generation;
-use crate::virtual_file::{on_fatal_io_error, MaybeFatalIo, VirtualFile};
+use crate::virtual_file::on_fatal_io_error;
 use crate::TEMP_FILE_SUFFIX;
 use remote_storage::{DownloadError, GenericRemoteStorage, ListingMode};
 use utils::crashsafe::path_with_suffix_extension;
@@ -50,8 +50,9 @@ pub async fn download_layer_file<'a>(
 ) -> Result<u64, DownloadError> {
     debug_assert_current_span_has_tenant_and_timeline_id();
 
-    let timeline_path = conf.timeline_path(&tenant_shard_id, &timeline_id);
-    let local_path = timeline_path.join(layer_file_name.file_name());
+    let local_path = conf
+        .timeline_path(&tenant_shard_id, &timeline_id)
+        .join(layer_file_name.file_name());
 
     let remote_path = remote_layer_path(
         &tenant_shard_id.tenant_id,
@@ -148,21 +149,10 @@ pub async fn download_layer_file<'a>(
         .with_context(|| format!("rename download layer file to {local_path}"))
         .map_err(DownloadError::Other)?;
 
-    // We use fatal_err() below because the after the rename above,
-    // the in-memory state of the filesystem already has the layer file in its final place,
-    // and subsequent pageserver code could think it's durable while it really isn't.
-    let work = async move {
-        let timeline_dir = VirtualFile::open(&timeline_path)
-            .await
-            .fatal_err("VirtualFile::open for timeline dir fsync");
-        timeline_dir
-            .sync_all()
-            .await
-            .fatal_err("VirtualFile::sync_all timeline dir");
-    };
-    crate::virtual_file::io_engine::get()
-        .spawn_blocking_and_block_on_if_std(work)
-        .await;
+    crashsafe::fsync_async(&local_path)
+        .await
+        .with_context(|| format!("fsync layer file {local_path}"))
+        .map_err(DownloadError::Other)?;
 
     tracing::debug!("download complete: {local_path}");
 

pageserver/src/tenant/storage_layer/delta_layer.rs

@@ -46,7 +46,6 @@ use crate::{DELTA_FILE_MAGIC, STORAGE_FORMAT_VERSION};
 use anyhow::{anyhow, bail, ensure, Context, Result};
 use bytes::BytesMut;
 use camino::{Utf8Path, Utf8PathBuf};
-use futures::StreamExt;
 use pageserver_api::keyspace::KeySpace;
 use pageserver_api::models::LayerAccessKind;
 use pageserver_api::shard::TenantShardId;
@@ -848,33 +847,10 @@ impl DeltaLayerInner {
         reconstruct_state: &mut ValuesReconstructState,
         ctx: &RequestContext,
     ) -> Result<(), GetVectoredError> {
-        let block_reader = FileBlockReader::new(&self.file, self.file_id);
-        let index_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
-            self.index_start_blk,
-            self.index_root_blk,
-            block_reader,
-        );
-
-        let planner = VectoredReadPlanner::new(
-            self.max_vectored_read_bytes
-                .expect("Layer is loaded with max vectored bytes config")
-                .0
-                .into(),
-        );
-
-        let data_end_offset = self.index_start_blk as u64 * PAGE_SZ as u64;
-
-        let reads = Self::plan_reads(
-            keyspace,
-            lsn_range,
-            data_end_offset,
-            index_reader,
-            planner,
-            reconstruct_state,
-            ctx,
-        )
-        .await
-        .map_err(GetVectoredError::Other)?;
+        let reads = self
+            .plan_reads(keyspace, lsn_range, reconstruct_state, ctx)
+            .await
+            .map_err(GetVectoredError::Other)?;
 
         self.do_reads_and_update_state(reads, reconstruct_state)
             .await;
@@ -882,64 +858,73 @@ impl DeltaLayerInner {
         Ok(())
     }
 
-    async fn plan_reads<Reader>(
+    async fn plan_reads(
+        &self,
         keyspace: KeySpace,
         lsn_range: Range<Lsn>,
-        data_end_offset: u64,
-        index_reader: DiskBtreeReader<Reader, DELTA_KEY_SIZE>,
-        mut planner: VectoredReadPlanner,
         reconstruct_state: &mut ValuesReconstructState,
         ctx: &RequestContext,
-    ) -> anyhow::Result<Vec<VectoredRead>>
-    where
-        Reader: BlockReader,
-    {
-        let ctx = RequestContextBuilder::extend(ctx)
-            .page_content_kind(PageContentKind::DeltaLayerBtreeNode)
-            .build();
+    ) -> anyhow::Result<Vec<VectoredRead>> {
+        let mut planner = VectoredReadPlanner::new(
+            self.max_vectored_read_bytes
+                .expect("Layer is loaded with max vectored bytes config")
+                .0
+                .into(),
+        );
+
+        let block_reader = FileBlockReader::new(&self.file, self.file_id);
+        let tree_reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(
+            self.index_start_blk,
+            self.index_root_blk,
+            block_reader,
+        );
 
         for range in keyspace.ranges.iter() {
             let mut range_end_handled = false;
 
             let start_key = DeltaKey::from_key_lsn(&range.start, lsn_range.start);
-            let index_stream = index_reader.get_stream_from(&start_key.0, &ctx);
-            let mut index_stream = std::pin::pin!(index_stream);
+            tree_reader
+                .visit(
+                    &start_key.0,
+                    VisitDirection::Forwards,
+                    |raw_key, value| {
+                        let key = Key::from_slice(&raw_key[..KEY_SIZE]);
+                        let lsn = DeltaKey::extract_lsn_from_buf(raw_key);
+                        let blob_ref = BlobRef(value);
 
-            while let Some(index_entry) = index_stream.next().await {
-                let (raw_key, value) = index_entry?;
-                let key = Key::from_slice(&raw_key[..KEY_SIZE]);
-                let lsn = DeltaKey::extract_lsn_from_buf(&raw_key);
-                let blob_ref = BlobRef(value);
+                        assert!(key >= range.start && lsn >= lsn_range.start);
 
-                // Lsns are not monotonically increasing across keys, so we don't assert on them.
-                assert!(key >= range.start);
+                        let cached_lsn = reconstruct_state.get_cached_lsn(&key);
+                        let flag = {
+                            if cached_lsn >= Some(lsn) {
+                                BlobFlag::Ignore
+                            } else if blob_ref.will_init() {
+                                BlobFlag::Replaces
+                            } else {
+                                BlobFlag::None
+                            }
+                        };
 
-                let outside_lsn_range = !lsn_range.contains(&lsn);
-                let below_cached_lsn = reconstruct_state.get_cached_lsn(&key) >= Some(lsn);
-
-                let flag = {
-                    if outside_lsn_range || below_cached_lsn {
-                        BlobFlag::Ignore
-                    } else if blob_ref.will_init() {
-                        BlobFlag::ReplaceAll
-                    } else {
-                        // Usual path: add blob to the read
-                        BlobFlag::None
-                    }
-                };
-
-                if key >= range.end || (key.next() == range.end && lsn >= lsn_range.end) {
-                    planner.handle_range_end(blob_ref.pos());
-                    range_end_handled = true;
-                    break;
-                } else {
-                    planner.handle(key, lsn, blob_ref.pos(), flag);
-                }
-            }
+                        if key >= range.end || (key.next() == range.end && lsn >= lsn_range.end) {
+                            planner.handle_range_end(blob_ref.pos());
+                            range_end_handled = true;
+                            false
+                        } else {
+                            planner.handle(key, lsn, blob_ref.pos(), flag);
+                            true
+                        }
+                    },
+                    &RequestContextBuilder::extend(ctx)
+                        .page_content_kind(PageContentKind::DeltaLayerBtreeNode)
+                        .build(),
+                )
+                .await
+                .map_err(|err| anyhow!(err))?;
 
             if !range_end_handled {
-                tracing::info!("Handling range end fallback at {}", data_end_offset);
-                planner.handle_range_end(data_end_offset);
+                let payload_end = self.index_start_blk as u64 * PAGE_SZ as u64;
+                tracing::info!("Handling range end fallback at {}", payload_end);
+                planner.handle_range_end(payload_end);
             }
         }
 
@@ -1205,131 +1190,3 @@ impl<'a> pageserver_compaction::interface::CompactionDeltaEntry<'a, Key> for Del
         self.size
     }
 }
-
-#[cfg(test)]
-mod test {
-    use std::collections::BTreeMap;
-
-    use super::*;
-    use crate::{
-        context::DownloadBehavior, task_mgr::TaskKind, tenant::disk_btree::tests::TestDisk,
-    };
-
-    /// Construct an index for a fictional delta layer and and then
-    /// traverse in order to plan vectored reads for a query. Finally,
-    /// verify that the traversal fed the right index key and value
-    /// pairs into the planner.
-    #[tokio::test]
-    async fn test_delta_layer_index_traversal() {
-        let base_key = Key {
-            field1: 0,
-            field2: 1663,
-            field3: 12972,
-            field4: 16396,
-            field5: 0,
-            field6: 246080,
-        };
-
-        // Populate the index with some entries
-        let entries: BTreeMap<Key, Vec<Lsn>> = BTreeMap::from([
-            (base_key, vec![Lsn(1), Lsn(5), Lsn(25), Lsn(26), Lsn(28)]),
-            (base_key.add(1), vec![Lsn(2), Lsn(5), Lsn(10), Lsn(50)]),
-            (base_key.add(2), vec![Lsn(2), Lsn(5), Lsn(10), Lsn(50)]),
-            (base_key.add(5), vec![Lsn(10), Lsn(15), Lsn(16), Lsn(20)]),
-        ]);
-
-        let mut disk = TestDisk::default();
-        let mut writer = DiskBtreeBuilder::<_, DELTA_KEY_SIZE>::new(&mut disk);
-
-        let mut disk_offset = 0;
-        for (key, lsns) in &entries {
-            for lsn in lsns {
-                let index_key = DeltaKey::from_key_lsn(key, *lsn);
-                let blob_ref = BlobRef::new(disk_offset, false);
-                writer
-                    .append(&index_key.0, blob_ref.0)
-                    .expect("In memory disk append should never fail");
-
-                disk_offset += 1;
-            }
-        }
-
-        // Prepare all the arguments for the call into `plan_reads` below
-        let (root_offset, _writer) = writer
-            .finish()
-            .expect("In memory disk finish should never fail");
-        let reader = DiskBtreeReader::<_, DELTA_KEY_SIZE>::new(0, root_offset, disk);
-        let planner = VectoredReadPlanner::new(100);
-        let mut reconstruct_state = ValuesReconstructState::new();
-        let ctx = RequestContext::new(TaskKind::UnitTest, DownloadBehavior::Error);
-
-        let keyspace = KeySpace {
-            ranges: vec![
-                base_key..base_key.add(3),
-                base_key.add(3)..base_key.add(100),
-            ],
-        };
-        let lsn_range = Lsn(2)..Lsn(40);
-
-        // Plan and validate
-        let vectored_reads = DeltaLayerInner::plan_reads(
-            keyspace.clone(),
-            lsn_range.clone(),
-            disk_offset,
-            reader,
-            planner,
-            &mut reconstruct_state,
-            &ctx,
-        )
-        .await
-        .expect("Read planning should not fail");
-
-        validate(keyspace, lsn_range, vectored_reads, entries);
-    }
-
-    fn validate(
-        keyspace: KeySpace,
-        lsn_range: Range<Lsn>,
-        vectored_reads: Vec<VectoredRead>,
-        index_entries: BTreeMap<Key, Vec<Lsn>>,
-    ) {
-        #[derive(Debug, PartialEq, Eq)]
-        struct BlobSpec {
-            key: Key,
-            lsn: Lsn,
-            at: u64,
-        }
-
-        let mut planned_blobs = Vec::new();
-        for read in vectored_reads {
-            for (at, meta) in read.blobs_at.as_slice() {
-                planned_blobs.push(BlobSpec {
-                    key: meta.key,
-                    lsn: meta.lsn,
-                    at: *at,
-                });
-            }
-        }
-
-        let mut expected_blobs = Vec::new();
-        let mut disk_offset = 0;
-        for (key, lsns) in index_entries {
-            for lsn in lsns {
-                let key_included = keyspace.ranges.iter().any(|range| range.contains(&key));
-                let lsn_included = lsn_range.contains(&lsn);
-
-                if key_included && lsn_included {
-                    expected_blobs.push(BlobSpec {
-                        key,
-                        lsn,
-                        at: disk_offset,
-                    });
-                }
-
-                disk_offset += 1;
-            }
-        }
-
-        assert_eq!(planned_blobs, expected_blobs);
-    }
-}

pageserver/src/tenant/storage_layer/image_layer.rs

@@ -43,7 +43,6 @@ use crate::{IMAGE_FILE_MAGIC, STORAGE_FORMAT_VERSION, TEMP_FILE_SUFFIX};
 use anyhow::{anyhow, bail, ensure, Context, Result};
 use bytes::{Bytes, BytesMut};
 use camino::{Utf8Path, Utf8PathBuf};
-use hex;
 use pageserver_api::keyspace::KeySpace;
 use pageserver_api::models::LayerAccessKind;
 use pageserver_api::shard::TenantShardId;
@@ -55,7 +54,6 @@ use std::ops::Range;
 use std::os::unix::prelude::FileExt;
 use std::sync::Arc;
 use tokio::sync::OnceCell;
-use tokio_stream::StreamExt;
 use tracing::*;
 
 use utils::{
@@ -490,33 +488,35 @@ impl ImageLayerInner {
         let tree_reader =
             DiskBtreeReader::new(self.index_start_blk, self.index_root_blk, block_reader);
 
-        let ctx = RequestContextBuilder::extend(ctx)
-            .page_content_kind(PageContentKind::ImageLayerBtreeNode)
-            .build();
-
         for range in keyspace.ranges.iter() {
             let mut range_end_handled = false;
 
             let mut search_key: [u8; KEY_SIZE] = [0u8; KEY_SIZE];
             range.start.write_to_byte_slice(&mut search_key);
 
-            let index_stream = tree_reader.get_stream_from(&search_key, &ctx);
-            let mut index_stream = std::pin::pin!(index_stream);
+            tree_reader
+                .visit(
+                    &search_key,
+                    VisitDirection::Forwards,
+                    |raw_key, offset| {
+                        let key = Key::from_slice(&raw_key[..KEY_SIZE]);
+                        assert!(key >= range.start);
 
-            while let Some(index_entry) = index_stream.next().await {
-                let (raw_key, offset) = index_entry?;
-
-                let key = Key::from_slice(&raw_key[..KEY_SIZE]);
-                assert!(key >= range.start);
-
-                if key >= range.end {
-                    planner.handle_range_end(offset);
-                    range_end_handled = true;
-                    break;
-                } else {
-                    planner.handle(key, self.lsn, offset, BlobFlag::None);
-                }
-            }
+                        if key >= range.end {
+                            planner.handle_range_end(offset);
+                            range_end_handled = true;
+                            false
+                        } else {
+                            planner.handle(key, self.lsn, offset, BlobFlag::None);
+                            true
+                        }
+                    },
+                    &RequestContextBuilder::extend(ctx)
+                        .page_content_kind(PageContentKind::ImageLayerBtreeNode)
+                        .build(),
+                )
+                .await
+                .map_err(|err| GetVectoredError::Other(anyhow!(err)))?;
 
             if !range_end_handled {
                 let payload_end = self.index_start_blk as u64 * PAGE_SZ as u64;

pageserver/src/tenant/storage_layer/inmemory_layer.rs

@@ -336,17 +336,32 @@ impl InMemoryLayer {
 
     /// Common subroutine of the public put_wal_record() and put_page_image() functions.
     /// Adds the page version to the in-memory tree
-
     pub(crate) async fn put_value(
         &self,
         key: Key,
         lsn: Lsn,
-        buf: &[u8],
+        val: &Value,
         ctx: &RequestContext,
     ) -> Result<()> {
         let mut inner = self.inner.write().await;
         self.assert_writable();
-        self.put_value_locked(&mut inner, key, lsn, buf, ctx).await
+        self.put_value_locked(&mut inner, key, lsn, val, ctx).await
+    }
+
+    pub(crate) async fn put_values(
+        &self,
+        values: &HashMap<Key, Vec<(Lsn, Value)>>,
+        ctx: &RequestContext,
+    ) -> Result<()> {
+        let mut inner = self.inner.write().await;
+        self.assert_writable();
+        for (key, vals) in values {
+            for (lsn, val) in vals {
+                self.put_value_locked(&mut inner, *key, *lsn, val, ctx)
+                    .await?;
+            }
+        }
+        Ok(())
     }
 
     async fn put_value_locked(
@@ -354,16 +369,22 @@ impl InMemoryLayer {
         locked_inner: &mut RwLockWriteGuard<'_, InMemoryLayerInner>,
         key: Key,
         lsn: Lsn,
-        buf: &[u8],
+        val: &Value,
         ctx: &RequestContext,
     ) -> Result<()> {
         trace!("put_value key {} at {}/{}", key, self.timeline_id, lsn);
 
         let off = {
+            // 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();
+            buf.clear();
+            val.ser_into(&mut buf)?;
             locked_inner
                 .file
                 .write_blob(
-                    buf,
+                    &buf,
                     &RequestContextBuilder::extend(ctx)
                         .page_content_kind(PageContentKind::InMemoryLayer)
                         .build(),
@@ -391,12 +412,7 @@ impl InMemoryLayer {
     pub async fn freeze(&self, end_lsn: Lsn) {
         let inner = self.inner.write().await;
 
-        assert!(
-            self.start_lsn < end_lsn,
-            "{} >= {}",
-            self.start_lsn,
-            end_lsn
-        );
+        assert!(self.start_lsn < end_lsn);
         self.end_lsn.set(end_lsn).expect("end_lsn set only once");
 
         for vec_map in inner.index.values() {

pageserver/src/tenant/storage_layer/layer.rs

@@ -195,7 +195,6 @@ impl Layer {
         let downloaded = resident.expect("just initialized");
 
         // if the rename works, the path is as expected
-        // TODO: sync system call
         std::fs::rename(temp_path, owner.local_path())
             .with_context(|| format!("rename temporary file as correct path for {owner}"))?;
 
@@ -976,7 +975,7 @@ impl LayerInner {
                 }
 
                 self.consecutive_failures.store(0, Ordering::Relaxed);
-                tracing::info!(size=%self.desc.file_size, "on-demand download successful");
+                tracing::info!("on-demand download successful");
 
                 Ok(permit)
             }

pageserver/src/tenant/tasks.rs

@@ -101,7 +101,6 @@ pub fn start_background_loops(
                     _ = completion::Barrier::maybe_wait(background_jobs_can_start) => {}
                 };
                 compaction_loop(tenant, cancel)
-                    // If you rename this span, change the RUST_LOG env variable in test_runner/performance/test_branch_creation.py
                     .instrument(info_span!("compaction_loop", tenant_id = %tenant_shard_id.tenant_id, shard_id = %tenant_shard_id.shard_slug()))
                     .await;
                 Ok(())
@@ -199,11 +198,7 @@ async fn compaction_loop(tenant: Arc<Tenant>, cancel: CancellationToken) {
                 }
             };
 
-            let elapsed = started_at.elapsed();
-            warn_when_period_overrun(elapsed, period, BackgroundLoopKind::Compaction);
-
-            // the duration is recorded by performance tests by enabling debug in this function
-            tracing::debug!(elapsed_ms=elapsed.as_millis(), "compaction iteration complete");
+            warn_when_period_overrun(started_at.elapsed(), period, BackgroundLoopKind::Compaction);
 
             // Perhaps we did no work and the walredo process has been idle for some time:
             // give it a chance to shut down to avoid leaving walredo process running indefinitely.
@@ -222,7 +217,7 @@ async fn compaction_loop(tenant: Arc<Tenant>, cancel: CancellationToken) {
                 }
                 let allowed_rps = tenant.timeline_get_throttle.steady_rps();
                 let delta = now - prev;
-                info!(
+                warn!(
                     n_seconds=%format_args!("{:.3}",
                     delta.as_secs_f64()),
                     count_accounted,

pageserver/src/tenant/throttle.rs

@@ -2,14 +2,14 @@ use std::{
     str::FromStr,
     sync::{
         atomic::{AtomicU64, Ordering},
-        Arc, Mutex,
+        Arc,
     },
     time::{Duration, Instant},
 };
 
 use arc_swap::ArcSwap;
 use enumset::EnumSet;
-use tracing::{error, warn};
+use tracing::error;
 
 use crate::{context::RequestContext, task_mgr::TaskKind};
 
@@ -157,19 +157,6 @@ where
                 .fetch_add(wait_time.as_micros() as u64, Ordering::Relaxed);
             let observation = Observation { wait_time };
             self.metric.observe_throttling(&observation);
-            match ctx.micros_spent_throttled.add(wait_time) {
-                Ok(res) => res,
-                Err(error) => {
-                    use once_cell::sync::Lazy;
-                    use utils::rate_limit::RateLimit;
-                    static WARN_RATE_LIMIT: Lazy<Mutex<RateLimit>> =
-                        Lazy::new(|| Mutex::new(RateLimit::new(Duration::from_secs(10))));
-                    let mut guard = WARN_RATE_LIMIT.lock().unwrap();
-                    guard.call(move || {
-                        warn!(error, "error adding time spent throttled; this message is logged at a global rate limit");
-                    });
-                }
-            }
         }
     }
 }

pageserver/src/tenant/timeline/compaction.rs

@@ -4,32 +4,24 @@
 //!
 //! The old legacy algorithm is implemented directly in `timeline.rs`.
 
-use std::collections::BinaryHeap;
 use std::ops::{Deref, Range};
 use std::sync::Arc;
 
-use super::layer_manager::LayerManager;
-use super::{CompactFlags, DurationRecorder, RecordedDuration, Timeline};
+use super::Timeline;
 
-use anyhow::{anyhow, Context};
 use async_trait::async_trait;
-use enumset::EnumSet;
 use fail::fail_point;
-use itertools::Itertools;
-use pageserver_api::shard::TenantShardId;
 use tokio_util::sync::CancellationToken;
-use tracing::{debug, info, info_span, trace, warn, Instrument};
-use utils::id::TimelineId;
+use tracing::{debug, trace, warn};
 
-use crate::context::{AccessStatsBehavior, RequestContext, RequestContextBuilder};
+use crate::context::RequestContext;
 use crate::tenant::storage_layer::{AsLayerDesc, PersistentLayerDesc};
-use crate::tenant::timeline::{drop_rlock, is_rel_fsm_block_key, is_rel_vm_block_key, Hole};
+use crate::tenant::timeline::{is_rel_fsm_block_key, is_rel_vm_block_key};
 use crate::tenant::timeline::{DeltaLayerWriter, ImageLayerWriter};
 use crate::tenant::timeline::{Layer, ResidentLayer};
 use crate::tenant::DeltaLayer;
 use crate::tenant::PageReconstructError;
-use crate::virtual_file::{MaybeFatalIo, VirtualFile};
-use crate::{page_cache, ZERO_PAGE};
+use crate::ZERO_PAGE;
 
 use crate::keyspace::KeySpace;
 use crate::repository::Key;
@@ -41,694 +33,6 @@ use pageserver_compaction::interface::*;
 
 use super::CompactionError;
 
-impl Timeline {
-    /// TODO: cancellation
-    pub(crate) async fn compact_legacy(
-        self: &Arc<Self>,
-        _cancel: &CancellationToken,
-        flags: EnumSet<CompactFlags>,
-        ctx: &RequestContext,
-    ) -> Result<(), CompactionError> {
-        // High level strategy for compaction / image creation:
-        //
-        // 1. First, calculate the desired "partitioning" of the
-        // currently in-use key space. The goal is to partition the
-        // key space into roughly fixed-size chunks, but also take into
-        // account any existing image layers, and try to align the
-        // chunk boundaries with the existing image layers to avoid
-        // too much churn. Also try to align chunk boundaries with
-        // relation boundaries.  In principle, we don't know about
-        // relation boundaries here, we just deal with key-value
-        // pairs, and the code in pgdatadir_mapping.rs knows how to
-        // map relations into key-value pairs. But in practice we know
-        // that 'field6' is the block number, and the fields 1-5
-        // identify a relation. This is just an optimization,
-        // though.
-        //
-        // 2. Once we know the partitioning, for each partition,
-        // decide if it's time to create a new image layer. The
-        // criteria is: there has been too much "churn" since the last
-        // image layer? The "churn" is fuzzy concept, it's a
-        // combination of too many delta files, or too much WAL in
-        // total in the delta file. Or perhaps: if creating an image
-        // file would allow to delete some older files.
-        //
-        // 3. After that, we compact all level0 delta files if there
-        // are too many of them.  While compacting, we also garbage
-        // collect any page versions that are no longer needed because
-        // of the new image layers we created in step 2.
-        //
-        // TODO: This high level strategy hasn't been implemented yet.
-        // Below are functions compact_level0() and create_image_layers()
-        // but they are a bit ad hoc and don't quite work like it's explained
-        // above. Rewrite it.
-
-        // Is the timeline being deleted?
-        if self.is_stopping() {
-            trace!("Dropping out of compaction on timeline shutdown");
-            return Err(CompactionError::ShuttingDown);
-        }
-
-        let target_file_size = self.get_checkpoint_distance();
-
-        // Define partitioning schema if needed
-
-        // FIXME: the match should only cover repartitioning, not the next steps
-        match self
-            .repartition(
-                self.get_last_record_lsn(),
-                self.get_compaction_target_size(),
-                flags,
-                ctx,
-            )
-            .await
-        {
-            Ok((partitioning, lsn)) => {
-                // Disables access_stats updates, so that the files we read remain candidates for eviction after we're done with them
-                let image_ctx = RequestContextBuilder::extend(ctx)
-                    .access_stats_behavior(AccessStatsBehavior::Skip)
-                    .build();
-
-                // 2. Compact
-                let timer = self.metrics.compact_time_histo.start_timer();
-                self.compact_level0(target_file_size, ctx).await?;
-                timer.stop_and_record();
-
-                // 3. Create new image layers for partitions that have been modified
-                // "enough".
-                let layers = self
-                    .create_image_layers(
-                        &partitioning,
-                        lsn,
-                        flags.contains(CompactFlags::ForceImageLayerCreation),
-                        &image_ctx,
-                    )
-                    .await
-                    .map_err(anyhow::Error::from)?;
-                if let Some(remote_client) = &self.remote_client {
-                    for layer in layers {
-                        remote_client.schedule_layer_file_upload(layer)?;
-                    }
-                }
-
-                if let Some(remote_client) = &self.remote_client {
-                    // should any new image layer been created, not uploading index_part will
-                    // result in a mismatch between remote_physical_size and layermap calculated
-                    // size, which will fail some tests, but should not be an issue otherwise.
-                    remote_client.schedule_index_upload_for_file_changes()?;
-                }
-            }
-            Err(err) => {
-                // no partitioning? This is normal, if the timeline was just created
-                // as an empty timeline. Also in unit tests, when we use the timeline
-                // as a simple key-value store, ignoring the datadir layout. Log the
-                // error but continue.
-                //
-                // Suppress error when it's due to cancellation
-                if !self.cancel.is_cancelled() {
-                    tracing::error!("could not compact, repartitioning keyspace failed: {err:?}");
-                }
-            }
-        };
-
-        Ok(())
-    }
-
-    /// Collect a bunch of Level 0 layer files, and compact and reshuffle them as
-    /// as Level 1 files.
-    async fn compact_level0(
-        self: &Arc<Self>,
-        target_file_size: u64,
-        ctx: &RequestContext,
-    ) -> Result<(), CompactionError> {
-        let CompactLevel0Phase1Result {
-            new_layers,
-            deltas_to_compact,
-        } = {
-            let phase1_span = info_span!("compact_level0_phase1");
-            let ctx = ctx.attached_child();
-            let mut stats = CompactLevel0Phase1StatsBuilder {
-                version: Some(2),
-                tenant_id: Some(self.tenant_shard_id),
-                timeline_id: Some(self.timeline_id),
-                ..Default::default()
-            };
-
-            let begin = tokio::time::Instant::now();
-            let phase1_layers_locked = Arc::clone(&self.layers).read_owned().await;
-            let now = tokio::time::Instant::now();
-            stats.read_lock_acquisition_micros =
-                DurationRecorder::Recorded(RecordedDuration(now - begin), now);
-            self.compact_level0_phase1(phase1_layers_locked, stats, target_file_size, &ctx)
-                .instrument(phase1_span)
-                .await?
-        };
-
-        if new_layers.is_empty() && deltas_to_compact.is_empty() {
-            // nothing to do
-            return Ok(());
-        }
-
-        self.finish_compact_batch(&new_layers, &Vec::new(), &deltas_to_compact)
-            .await?;
-        Ok(())
-    }
-
-    /// Level0 files first phase of compaction, explained in the [`Self::compact_legacy`] comment.
-    async fn compact_level0_phase1(
-        self: &Arc<Self>,
-        guard: tokio::sync::OwnedRwLockReadGuard<LayerManager>,
-        mut stats: CompactLevel0Phase1StatsBuilder,
-        target_file_size: u64,
-        ctx: &RequestContext,
-    ) -> Result<CompactLevel0Phase1Result, CompactionError> {
-        stats.read_lock_held_spawn_blocking_startup_micros =
-            stats.read_lock_acquisition_micros.till_now(); // set by caller
-        let layers = guard.layer_map();
-        let level0_deltas = layers.get_level0_deltas()?;
-        let mut level0_deltas = level0_deltas
-            .into_iter()
-            .map(|x| guard.get_from_desc(&x))
-            .collect_vec();
-        stats.level0_deltas_count = Some(level0_deltas.len());
-        // Only compact if enough layers have accumulated.
-        let threshold = self.get_compaction_threshold();
-        if level0_deltas.is_empty() || level0_deltas.len() < threshold {
-            debug!(
-                level0_deltas = level0_deltas.len(),
-                threshold, "too few deltas to compact"
-            );
-            return Ok(CompactLevel0Phase1Result::default());
-        }
-
-        // This failpoint is used together with `test_duplicate_layers` integration test.
-        // It returns the compaction result exactly the same layers as input to compaction.
-        // We want to ensure that this will not cause any problem when updating the layer map
-        // after the compaction is finished.
-        //
-        // Currently, there are two rare edge cases that will cause duplicated layers being
-        // inserted.
-        // 1. The compaction job is inturrupted / did not finish successfully. Assume we have file 1, 2, 3, 4, which
-        //    is compacted to 5, but the page server is shut down, next time we start page server we will get a layer
-        //    map containing 1, 2, 3, 4, and 5, whereas 5 has the same content as 4. If we trigger L0 compation at this
-        //    point again, it is likely that we will get a file 6 which has the same content and the key range as 5,
-        //    and this causes an overwrite. This is acceptable because the content is the same, and we should do a
-        //    layer replace instead of the normal remove / upload process.
-        // 2. The input workload pattern creates exactly n files that are sorted, non-overlapping and is of target file
-        //    size length. Compaction will likely create the same set of n files afterwards.
-        //
-        // This failpoint is a superset of both of the cases.
-        if cfg!(feature = "testing") {
-            let active = (|| {
-                ::fail::fail_point!("compact-level0-phase1-return-same", |_| true);
-                false
-            })();
-
-            if active {
-                let mut new_layers = Vec::with_capacity(level0_deltas.len());
-                for delta in &level0_deltas {
-                    // we are just faking these layers as being produced again for this failpoint
-                    new_layers.push(
-                        delta
-                            .download_and_keep_resident()
-                            .await
-                            .context("download layer for failpoint")?,
-                    );
-                }
-                tracing::info!("compact-level0-phase1-return-same"); // so that we can check if we hit the failpoint
-                return Ok(CompactLevel0Phase1Result {
-                    new_layers,
-                    deltas_to_compact: level0_deltas,
-                });
-            }
-        }
-
-        // Gather the files to compact in this iteration.
-        //
-        // Start with the oldest Level 0 delta file, and collect any other
-        // level 0 files that form a contiguous sequence, such that the end
-        // LSN of previous file matches the start LSN of the next file.
-        //
-        // Note that if the files don't form such a sequence, we might
-        // "compact" just a single file. That's a bit pointless, but it allows
-        // us to get rid of the level 0 file, and compact the other files on
-        // the next iteration. This could probably made smarter, but such
-        // "gaps" in the sequence of level 0 files should only happen in case
-        // of a crash, partial download from cloud storage, or something like
-        // that, so it's not a big deal in practice.
-        level0_deltas.sort_by_key(|l| l.layer_desc().lsn_range.start);
-        let mut level0_deltas_iter = level0_deltas.iter();
-
-        let first_level0_delta = level0_deltas_iter.next().unwrap();
-        let mut prev_lsn_end = first_level0_delta.layer_desc().lsn_range.end;
-        let mut deltas_to_compact = Vec::with_capacity(level0_deltas.len());
-
-        deltas_to_compact.push(first_level0_delta.download_and_keep_resident().await?);
-        for l in level0_deltas_iter {
-            let lsn_range = &l.layer_desc().lsn_range;
-
-            if lsn_range.start != prev_lsn_end {
-                break;
-            }
-            deltas_to_compact.push(l.download_and_keep_resident().await?);
-            prev_lsn_end = lsn_range.end;
-        }
-        let lsn_range = Range {
-            start: deltas_to_compact
-                .first()
-                .unwrap()
-                .layer_desc()
-                .lsn_range
-                .start,
-            end: deltas_to_compact.last().unwrap().layer_desc().lsn_range.end,
-        };
-
-        info!(
-            "Starting Level0 compaction in LSN range {}-{} for {} layers ({} deltas in total)",
-            lsn_range.start,
-            lsn_range.end,
-            deltas_to_compact.len(),
-            level0_deltas.len()
-        );
-
-        for l in deltas_to_compact.iter() {
-            info!("compact includes {l}");
-        }
-
-        // We don't need the original list of layers anymore. Drop it so that
-        // we don't accidentally use it later in the function.
-        drop(level0_deltas);
-
-        stats.read_lock_held_prerequisites_micros = stats
-            .read_lock_held_spawn_blocking_startup_micros
-            .till_now();
-
-        // Determine N largest holes where N is number of compacted layers.
-        let max_holes = deltas_to_compact.len();
-        let last_record_lsn = self.get_last_record_lsn();
-        let min_hole_range = (target_file_size / page_cache::PAGE_SZ as u64) as i128;
-        let min_hole_coverage_size = 3; // TODO: something more flexible?
-
-        // min-heap (reserve space for one more element added before eviction)
-        let mut heap: BinaryHeap<Hole> = BinaryHeap::with_capacity(max_holes + 1);
-        let mut prev: Option<Key> = None;
-
-        let mut all_keys = Vec::new();
-
-        for l in deltas_to_compact.iter() {
-            all_keys.extend(l.load_keys(ctx).await?);
-        }
-
-        // FIXME: should spawn_blocking the rest of this function
-
-        // The current stdlib sorting implementation is designed in a way where it is
-        // particularly fast where the slice is made up of sorted sub-ranges.
-        all_keys.sort_by_key(|DeltaEntry { key, lsn, .. }| (*key, *lsn));
-
-        stats.read_lock_held_key_sort_micros = stats.read_lock_held_prerequisites_micros.till_now();
-
-        for &DeltaEntry { key: next_key, .. } in all_keys.iter() {
-            if let Some(prev_key) = prev {
-                // just first fast filter
-                if next_key.to_i128() - prev_key.to_i128() >= min_hole_range {
-                    let key_range = prev_key..next_key;
-                    // Measuring hole by just subtraction of i128 representation of key range boundaries
-                    // has not so much sense, because largest holes will corresponds field1/field2 changes.
-                    // But we are mostly interested to eliminate holes which cause generation of excessive image layers.
-                    // That is why it is better to measure size of hole as number of covering image layers.
-                    let coverage_size = layers.image_coverage(&key_range, last_record_lsn).len();
-                    if coverage_size >= min_hole_coverage_size {
-                        heap.push(Hole {
-                            key_range,
-                            coverage_size,
-                        });
-                        if heap.len() > max_holes {
-                            heap.pop(); // remove smallest hole
-                        }
-                    }
-                }
-            }
-            prev = Some(next_key.next());
-        }
-        stats.read_lock_held_compute_holes_micros = stats.read_lock_held_key_sort_micros.till_now();
-        drop_rlock(guard);
-        stats.read_lock_drop_micros = stats.read_lock_held_compute_holes_micros.till_now();
-        let mut holes = heap.into_vec();
-        holes.sort_unstable_by_key(|hole| hole.key_range.start);
-        let mut next_hole = 0; // index of next hole in holes vector
-
-        // This iterator walks through all key-value pairs from all the layers
-        // we're compacting, in key, LSN order.
-        let all_values_iter = all_keys.iter();
-
-        // This iterator walks through all keys and is needed to calculate size used by each key
-        let mut all_keys_iter = all_keys
-            .iter()
-            .map(|DeltaEntry { key, lsn, size, .. }| (*key, *lsn, *size))
-            .coalesce(|mut prev, cur| {
-                // Coalesce keys that belong to the same key pair.
-                // This ensures that compaction doesn't put them
-                // into different layer files.
-                // Still limit this by the target file size,
-                // so that we keep the size of the files in
-                // check.
-                if prev.0 == cur.0 && prev.2 < target_file_size {
-                    prev.2 += cur.2;
-                    Ok(prev)
-                } else {
-                    Err((prev, cur))
-                }
-            });
-
-        // Merge the contents of all the input delta layers into a new set
-        // of delta layers, based on the current partitioning.
-        //
-        // We split the new delta layers on the key dimension. We iterate through the key space, and for each key, check if including the next key to the current output layer we're building would cause the layer to become too large. If so, dump the current output layer and start new one.
-        // It's possible that there is a single key with so many page versions that storing all of them in a single layer file
-        // would be too large. In that case, we also split on the LSN dimension.
-        //
-        // LSN
-        //  ^
-        //  |
-        //  | +-----------+            +--+--+--+--+
-        //  | |           |            |  |  |  |  |
-        //  | +-----------+            |  |  |  |  |
-        //  | |           |            |  |  |  |  |
-        //  | +-----------+     ==>    |  |  |  |  |
-        //  | |           |            |  |  |  |  |
-        //  | +-----------+            |  |  |  |  |
-        //  | |           |            |  |  |  |  |
-        //  | +-----------+            +--+--+--+--+
-        //  |
-        //  +--------------> key
-        //
-        //
-        // If one key (X) has a lot of page versions:
-        //
-        // LSN
-        //  ^
-        //  |                                 (X)
-        //  | +-----------+            +--+--+--+--+
-        //  | |           |            |  |  |  |  |
-        //  | +-----------+            |  |  +--+  |
-        //  | |           |            |  |  |  |  |
-        //  | +-----------+     ==>    |  |  |  |  |
-        //  | |           |            |  |  +--+  |
-        //  | +-----------+            |  |  |  |  |
-        //  | |           |            |  |  |  |  |
-        //  | +-----------+            +--+--+--+--+
-        //  |
-        //  +--------------> key
-        // TODO: this actually divides the layers into fixed-size chunks, not
-        // based on the partitioning.
-        //
-        // TODO: we should also opportunistically materialize and
-        // garbage collect what we can.
-        let mut new_layers = Vec::new();
-        let mut prev_key: Option<Key> = None;
-        let mut writer: Option<DeltaLayerWriter> = None;
-        let mut key_values_total_size = 0u64;
-        let mut dup_start_lsn: Lsn = Lsn::INVALID; // start LSN of layer containing values of the single key
-        let mut dup_end_lsn: Lsn = Lsn::INVALID; // end LSN of layer containing values of the single key
-
-        for &DeltaEntry {
-            key, lsn, ref val, ..
-        } in all_values_iter
-        {
-            let value = val.load(ctx).await?;
-            let same_key = prev_key.map_or(false, |prev_key| prev_key == key);
-            // We need to check key boundaries once we reach next key or end of layer with the same key
-            if !same_key || lsn == dup_end_lsn {
-                let mut next_key_size = 0u64;
-                let is_dup_layer = dup_end_lsn.is_valid();
-                dup_start_lsn = Lsn::INVALID;
-                if !same_key {
-                    dup_end_lsn = Lsn::INVALID;
-                }
-                // Determine size occupied by this key. We stop at next key or when size becomes larger than target_file_size
-                for (next_key, next_lsn, next_size) in all_keys_iter.by_ref() {
-                    next_key_size = next_size;
-                    if key != next_key {
-                        if dup_end_lsn.is_valid() {
-                            // We are writting segment with duplicates:
-                            // place all remaining values of this key in separate segment
-                            dup_start_lsn = dup_end_lsn; // new segments starts where old stops
-                            dup_end_lsn = lsn_range.end; // there are no more values of this key till end of LSN range
-                        }
-                        break;
-                    }
-                    key_values_total_size += next_size;
-                    // Check if it is time to split segment: if total keys size is larger than target file size.
-                    // We need to avoid generation of empty segments if next_size > target_file_size.
-                    if key_values_total_size > target_file_size && lsn != next_lsn {
-                        // Split key between multiple layers: such layer can contain only single key
-                        dup_start_lsn = if dup_end_lsn.is_valid() {
-                            dup_end_lsn // new segment with duplicates starts where old one stops
-                        } else {
-                            lsn // start with the first LSN for this key
-                        };
-                        dup_end_lsn = next_lsn; // upper LSN boundary is exclusive
-                        break;
-                    }
-                }
-                // handle case when loop reaches last key: in this case dup_end is non-zero but dup_start is not set.
-                if dup_end_lsn.is_valid() && !dup_start_lsn.is_valid() {
-                    dup_start_lsn = dup_end_lsn;
-                    dup_end_lsn = lsn_range.end;
-                }
-                if writer.is_some() {
-                    let written_size = writer.as_mut().unwrap().size();
-                    let contains_hole =
-                        next_hole < holes.len() && key >= holes[next_hole].key_range.end;
-                    // check if key cause layer overflow or contains hole...
-                    if is_dup_layer
-                        || dup_end_lsn.is_valid()
-                        || written_size + key_values_total_size > target_file_size
-                        || contains_hole
-                    {
-                        // ... if so, flush previous layer and prepare to write new one
-                        new_layers.push(
-                            writer
-                                .take()
-                                .unwrap()
-                                .finish(prev_key.unwrap().next(), self)
-                                .await?,
-                        );
-                        writer = None;
-
-                        if contains_hole {
-                            // skip hole
-                            next_hole += 1;
-                        }
-                    }
-                }
-                // Remember size of key value because at next iteration we will access next item
-                key_values_total_size = next_key_size;
-            }
-            fail_point!("delta-layer-writer-fail-before-finish", |_| {
-                Err(CompactionError::Other(anyhow::anyhow!(
-                    "failpoint delta-layer-writer-fail-before-finish"
-                )))
-            });
-
-            if !self.shard_identity.is_key_disposable(&key) {
-                if writer.is_none() {
-                    // Create writer if not initiaized yet
-                    writer = Some(
-                        DeltaLayerWriter::new(
-                            self.conf,
-                            self.timeline_id,
-                            self.tenant_shard_id,
-                            key,
-                            if dup_end_lsn.is_valid() {
-                                // this is a layer containing slice of values of the same key
-                                debug!("Create new dup layer {}..{}", dup_start_lsn, dup_end_lsn);
-                                dup_start_lsn..dup_end_lsn
-                            } else {
-                                debug!("Create new layer {}..{}", lsn_range.start, lsn_range.end);
-                                lsn_range.clone()
-                            },
-                        )
-                        .await?,
-                    );
-                }
-
-                writer.as_mut().unwrap().put_value(key, lsn, value).await?;
-            } else {
-                debug!(
-                    "Dropping key {} during compaction (it belongs on shard {:?})",
-                    key,
-                    self.shard_identity.get_shard_number(&key)
-                );
-            }
-
-            if !new_layers.is_empty() {
-                fail_point!("after-timeline-compacted-first-L1");
-            }
-
-            prev_key = Some(key);
-        }
-        if let Some(writer) = writer {
-            new_layers.push(writer.finish(prev_key.unwrap().next(), self).await?);
-        }
-
-        // Sync layers
-        if !new_layers.is_empty() {
-            // Print a warning if the created layer is larger than double the target size
-            // Add two pages for potential overhead. This should in theory be already
-            // accounted for in the target calculation, but for very small targets,
-            // we still might easily hit the limit otherwise.
-            let warn_limit = target_file_size * 2 + page_cache::PAGE_SZ as u64 * 2;
-            for layer in new_layers.iter() {
-                if layer.layer_desc().file_size > warn_limit {
-                    warn!(
-                        %layer,
-                        "created delta file of size {} larger than double of target of {target_file_size}", layer.layer_desc().file_size
-                    );
-                }
-            }
-
-            // The writer.finish() above already did the fsync of the inodes.
-            // We just need to fsync the directory in which these inodes are linked,
-            // which we know to be the timeline directory.
-            //
-            // We use fatal_err() below because the after writer.finish() returns with success,
-            // the in-memory state of the filesystem already has the layer file in its final place,
-            // and subsequent pageserver code could think it's durable while it really isn't.
-            let timeline_dir = VirtualFile::open(
-                &self
-                    .conf
-                    .timeline_path(&self.tenant_shard_id, &self.timeline_id),
-            )
-            .await
-            .fatal_err("VirtualFile::open for timeline dir fsync");
-            timeline_dir
-                .sync_all()
-                .await
-                .fatal_err("VirtualFile::sync_all timeline dir");
-        }
-
-        stats.write_layer_files_micros = stats.read_lock_drop_micros.till_now();
-        stats.new_deltas_count = Some(new_layers.len());
-        stats.new_deltas_size = Some(new_layers.iter().map(|l| l.layer_desc().file_size).sum());
-
-        match TryInto::<CompactLevel0Phase1Stats>::try_into(stats)
-            .and_then(|stats| serde_json::to_string(&stats).context("serde_json::to_string"))
-        {
-            Ok(stats_json) => {
-                info!(
-                    stats_json = stats_json.as_str(),
-                    "compact_level0_phase1 stats available"
-                )
-            }
-            Err(e) => {
-                warn!("compact_level0_phase1 stats failed to serialize: {:#}", e);
-            }
-        }
-
-        Ok(CompactLevel0Phase1Result {
-            new_layers,
-            deltas_to_compact: deltas_to_compact
-                .into_iter()
-                .map(|x| x.drop_eviction_guard())
-                .collect::<Vec<_>>(),
-        })
-    }
-}
-
-#[derive(Default)]
-struct CompactLevel0Phase1Result {
-    new_layers: Vec<ResidentLayer>,
-    deltas_to_compact: Vec<Layer>,
-}
-
-#[derive(Default)]
-struct CompactLevel0Phase1StatsBuilder {
-    version: Option<u64>,
-    tenant_id: Option<TenantShardId>,
-    timeline_id: Option<TimelineId>,
-    read_lock_acquisition_micros: DurationRecorder,
-    read_lock_held_spawn_blocking_startup_micros: DurationRecorder,
-    read_lock_held_key_sort_micros: DurationRecorder,
-    read_lock_held_prerequisites_micros: DurationRecorder,
-    read_lock_held_compute_holes_micros: DurationRecorder,
-    read_lock_drop_micros: DurationRecorder,
-    write_layer_files_micros: DurationRecorder,
-    level0_deltas_count: Option<usize>,
-    new_deltas_count: Option<usize>,
-    new_deltas_size: Option<u64>,
-}
-
-#[derive(serde::Serialize)]
-struct CompactLevel0Phase1Stats {
-    version: u64,
-    tenant_id: TenantShardId,
-    timeline_id: TimelineId,
-    read_lock_acquisition_micros: RecordedDuration,
-    read_lock_held_spawn_blocking_startup_micros: RecordedDuration,
-    read_lock_held_key_sort_micros: RecordedDuration,
-    read_lock_held_prerequisites_micros: RecordedDuration,
-    read_lock_held_compute_holes_micros: RecordedDuration,
-    read_lock_drop_micros: RecordedDuration,
-    write_layer_files_micros: RecordedDuration,
-    level0_deltas_count: usize,
-    new_deltas_count: usize,
-    new_deltas_size: u64,
-}
-
-impl TryFrom<CompactLevel0Phase1StatsBuilder> for CompactLevel0Phase1Stats {
-    type Error = anyhow::Error;
-
-    fn try_from(value: CompactLevel0Phase1StatsBuilder) -> Result<Self, Self::Error> {
-        Ok(Self {
-            version: value.version.ok_or_else(|| anyhow!("version not set"))?,
-            tenant_id: value
-                .tenant_id
-                .ok_or_else(|| anyhow!("tenant_id not set"))?,
-            timeline_id: value
-                .timeline_id
-                .ok_or_else(|| anyhow!("timeline_id not set"))?,
-            read_lock_acquisition_micros: value
-                .read_lock_acquisition_micros
-                .into_recorded()
-                .ok_or_else(|| anyhow!("read_lock_acquisition_micros not set"))?,
-            read_lock_held_spawn_blocking_startup_micros: value
-                .read_lock_held_spawn_blocking_startup_micros
-                .into_recorded()
-                .ok_or_else(|| anyhow!("read_lock_held_spawn_blocking_startup_micros not set"))?,
-            read_lock_held_key_sort_micros: value
-                .read_lock_held_key_sort_micros
-                .into_recorded()
-                .ok_or_else(|| anyhow!("read_lock_held_key_sort_micros not set"))?,
-            read_lock_held_prerequisites_micros: value
-                .read_lock_held_prerequisites_micros
-                .into_recorded()
-                .ok_or_else(|| anyhow!("read_lock_held_prerequisites_micros not set"))?,
-            read_lock_held_compute_holes_micros: value
-                .read_lock_held_compute_holes_micros
-                .into_recorded()
-                .ok_or_else(|| anyhow!("read_lock_held_compute_holes_micros not set"))?,
-            read_lock_drop_micros: value
-                .read_lock_drop_micros
-                .into_recorded()
-                .ok_or_else(|| anyhow!("read_lock_drop_micros not set"))?,
-            write_layer_files_micros: value
-                .write_layer_files_micros
-                .into_recorded()
-                .ok_or_else(|| anyhow!("write_layer_files_micros not set"))?,
-            level0_deltas_count: value
-                .level0_deltas_count
-                .ok_or_else(|| anyhow!("level0_deltas_count not set"))?,
-            new_deltas_count: value
-                .new_deltas_count
-                .ok_or_else(|| anyhow!("new_deltas_count not set"))?,
-            new_deltas_size: value
-                .new_deltas_size
-                .ok_or_else(|| anyhow!("new_deltas_size not set"))?,
-        })
-    }
-}
-
 impl Timeline {
     /// Entry point for new tiered compaction algorithm.
     ///
@@ -830,6 +134,7 @@ struct ResidentDeltaLayer(ResidentLayer);
 #[derive(Clone)]
 struct ResidentImageLayer(ResidentLayer);
 
+#[async_trait]
 impl CompactionJobExecutor for TimelineAdaptor {
     type Key = crate::repository::Key;
 

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

@@ -343,6 +343,23 @@ pub(super) async fn handle_walreceiver_connection(
                             modification.commit(&ctx).await?;
                             uncommitted_records = 0;
                             filtered_records = 0;
+
+                            //
+                            // We should check checkpoint distance after appending each ingest_batch_size bytes because otherwise
+                            // layer size can become much larger than `checkpoint_distance`.
+                            // It can append because wal-sender is sending WAL using 125kb chucks and some WAL records can cause writing large
+                            // amount of data to key-value storage. So performing this check only after processing
+                            // all WAL records in the chunk, can cause huge L0 layer files.
+                            //
+                            timeline
+                                .check_checkpoint_distance()
+                                .await
+                                .with_context(|| {
+                                    format!(
+                                        "Failed to check checkpoint distance for timeline {}",
+                                        timeline.timeline_id
+                                    )
+                                })?;
                         }
                     }
 
@@ -389,16 +406,15 @@ pub(super) async fn handle_walreceiver_connection(
             }
         }
 
-        {
-            // This is a hack. It piggybacks on the keepalive messages sent by the
-            // safekeeper in order to enforce `checkpoint_timeout` on the currently
-            // open layer. This hack doesn't provide a bound on the total size of
-            // in-memory layers on a pageserver. See https://github.com/neondatabase/neon/issues/6916.
-            let mut writer = timeline.writer().await;
-            if let Err(err) = writer.tick().await {
-                warn!("Timeline writer tick failed: {err}");
-            }
-        }
+        timeline
+            .check_checkpoint_distance()
+            .await
+            .with_context(|| {
+                format!(
+                    "Failed to check checkpoint distance for timeline {}",
+                    timeline.timeline_id
+                )
+            })?;
 
         if let Some(last_lsn) = status_update {
             let timeline_remote_consistent_lsn = timeline

pageserver/src/tenant/vectored_blob_io.rs

@@ -128,7 +128,7 @@ impl VectoredReadBuilder {
 pub enum BlobFlag {
     None,
     Ignore,
-    ReplaceAll,
+    Replaces,
 }
 
 /// Planner for vectored blob reads.
@@ -170,7 +170,7 @@ impl VectoredReadPlanner {
     /// incorrect data to the user.
     ///
     /// The `flag` argument has two interesting values:
-    /// * [`BlobFlag::ReplaceAll`]: The blob for this key should replace all existing blobs.
+    /// * [`BlobFlag::Replaces`]: The blob for this key should replace all existing blobs.
     /// This is used for WAL records that `will_init`.
     /// * [`BlobFlag::Ignore`]: This blob should not be included in the read. This happens
     /// if the blob is cached.
@@ -204,7 +204,7 @@ impl VectoredReadPlanner {
                 let blobs_for_key = self.blobs.entry(key).or_default();
                 blobs_for_key.push((lsn, start_offset, end_offset));
             }
-            BlobFlag::ReplaceAll => {
+            BlobFlag::Replaces => {
                 let blobs_for_key = self.blobs.entry(key).or_default();
                 blobs_for_key.clear();
                 blobs_for_key.push((lsn, start_offset, end_offset));
@@ -411,10 +411,10 @@ mod tests {
         let blob_descriptions = vec![
             (first_key, lsn, 0, BlobFlag::None),    // First in read 1
             (first_key, lsn, 1024, BlobFlag::None), // Last in read 1
-            (second_key, lsn, 2 * 1024, BlobFlag::ReplaceAll),
+            (second_key, lsn, 2 * 1024, BlobFlag::Replaces),
             (second_key, lsn, 3 * 1024, BlobFlag::None),
-            (second_key, lsn, 4 * 1024, BlobFlag::ReplaceAll), // First in read 2
-            (second_key, lsn, 5 * 1024, BlobFlag::None),       // Last in read 2
+            (second_key, lsn, 4 * 1024, BlobFlag::Replaces), // First in read 2
+            (second_key, lsn, 5 * 1024, BlobFlag::None),     // Last in read 2
         ];
 
         let ranges = [&blob_descriptions[0..2], &blob_descriptions[4..]];

pageserver/src/virtual_file.rs

@@ -17,21 +17,20 @@ use crate::tenant::TENANTS_SEGMENT_NAME;
 use camino::{Utf8Path, Utf8PathBuf};
 use once_cell::sync::OnceCell;
 use pageserver_api::shard::TenantShardId;
-use std::fs::File;
+use std::fs::{self, File};
 use std::io::{Error, ErrorKind, Seek, SeekFrom};
 use tokio_epoll_uring::{BoundedBuf, IoBuf, IoBufMut, Slice};
 
 use std::os::fd::{AsRawFd, FromRawFd, IntoRawFd, OwnedFd, RawFd};
+use std::os::unix::fs::FileExt;
 use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
 use tokio::sync::{RwLock, RwLockReadGuard, RwLockWriteGuard};
 use tokio::time::Instant;
 
 pub use pageserver_api::models::virtual_file as api;
 pub(crate) mod io_engine;
-mod metadata;
 mod open_options;
 pub(crate) use io_engine::IoEngineKind;
-pub(crate) use metadata::Metadata;
 pub(crate) use open_options::*;
 
 ///
@@ -436,25 +435,13 @@ impl VirtualFile {
 
     /// Call File::sync_all() on the underlying File.
     pub async fn sync_all(&self) -> Result<(), Error> {
-        with_file!(self, StorageIoOperation::Fsync, |file_guard| {
-            let (_file_guard, res) = io_engine::get().sync_all(file_guard).await;
-            res
-        })
+        with_file!(self, StorageIoOperation::Fsync, |file_guard| file_guard
+            .with_std_file(|std_file| std_file.sync_all()))
     }
 
-    /// Call File::sync_data() on the underlying File.
-    pub async fn sync_data(&self) -> Result<(), Error> {
-        with_file!(self, StorageIoOperation::Fsync, |file_guard| {
-            let (_file_guard, res) = io_engine::get().sync_data(file_guard).await;
-            res
-        })
-    }
-
-    pub async fn metadata(&self) -> Result<Metadata, Error> {
-        with_file!(self, StorageIoOperation::Metadata, |file_guard| {
-            let (_file_guard, res) = io_engine::get().metadata(file_guard).await;
-            res
-        })
+    pub async fn metadata(&self) -> Result<fs::Metadata, Error> {
+        with_file!(self, StorageIoOperation::Metadata, |file_guard| file_guard
+            .with_std_file(|std_file| std_file.metadata()))
     }
 
     /// Helper function internal to `VirtualFile` that looks up the underlying File,
@@ -592,7 +579,7 @@ impl VirtualFile {
     }
 
     // Copied from https://doc.rust-lang.org/1.72.0/src/std/os/unix/fs.rs.html#219-235
-    pub async fn write_all_at<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
+    pub async fn write_all_at<B: BoundedBuf>(
         &self,
         buf: B,
         mut offset: u64,
@@ -603,9 +590,8 @@ impl VirtualFile {
         }
         let mut buf = buf.slice(0..buf_len);
         while !buf.is_empty() {
-            let res;
-            (buf, res) = self.write_at(buf, offset).await;
-            match res {
+            // TODO: push `buf` further down
+            match self.write_at(&buf, offset).await {
                 Ok(0) => {
                     return (
                         Slice::into_inner(buf),
@@ -619,7 +605,7 @@ impl VirtualFile {
                     buf = buf.slice(n..);
                     offset += n as u64;
                 }
-                Err(e) if e.kind() == std::io::ErrorKind::Interrupted => {}
+                Err(ref e) if e.kind() == std::io::ErrorKind::Interrupted => {}
                 Err(e) => return (Slice::into_inner(buf), Err(e)),
             }
         }
@@ -630,19 +616,15 @@ impl VirtualFile {
     /// Returns the IoBuf that is underlying the BoundedBuf `buf`.
     /// I.e., the returned value's `bytes_init()` method returns something different than the `bytes_init()` that was passed in.
     /// It's quite brittle and easy to mis-use, so, we return the size in the Ok() variant.
-    pub async fn write_all<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
-        &mut self,
-        buf: B,
-    ) -> (B::Buf, Result<usize, Error>) {
+    pub async fn write_all<B: BoundedBuf>(&mut self, buf: B) -> (B::Buf, Result<usize, Error>) {
         let nbytes = buf.bytes_init();
         if nbytes == 0 {
             return (Slice::into_inner(buf.slice_full()), Ok(0));
         }
         let mut buf = buf.slice(0..nbytes);
         while !buf.is_empty() {
-            let res;
-            (buf, res) = self.write(buf).await;
-            match res {
+            // TODO: push `Slice` further down
+            match self.write(&buf).await {
                 Ok(0) => {
                     return (
                         Slice::into_inner(buf),
@@ -662,18 +644,11 @@ impl VirtualFile {
         (Slice::into_inner(buf), Ok(nbytes))
     }
 
-    async fn write<B: IoBuf + Send>(
-        &mut self,
-        buf: Slice<B>,
-    ) -> (Slice<B>, Result<usize, std::io::Error>) {
+    async fn write(&mut self, buf: &[u8]) -> Result<usize, std::io::Error> {
         let pos = self.pos;
-        let (buf, res) = self.write_at(buf, pos).await;
-        let n = match res {
-            Ok(n) => n,
-            Err(e) => return (buf, Err(e)),
-        };
+        let n = self.write_at(buf, pos).await?;
         self.pos += n as u64;
-        (buf, Ok(n))
+        Ok(n)
     }
 
     pub(crate) async fn read_at<B>(&self, buf: B, offset: u64) -> (B, Result<usize, Error>)
@@ -701,30 +676,16 @@ impl VirtualFile {
         })
     }
 
-    async fn write_at<B: IoBuf + Send>(
-        &self,
-        buf: Slice<B>,
-        offset: u64,
-    ) -> (Slice<B>, Result<usize, Error>) {
-        let file_guard = match self.lock_file().await {
-            Ok(file_guard) => file_guard,
-            Err(e) => return (buf, Err(e)),
-        };
-        observe_duration!(StorageIoOperation::Write, {
-            let ((_file_guard, buf), result) =
-                io_engine::get().write_at(file_guard, offset, buf).await;
-            if let Ok(size) = result {
-                STORAGE_IO_SIZE
-                    .with_label_values(&[
-                        "write",
-                        &self.tenant_id,
-                        &self.shard_id,
-                        &self.timeline_id,
-                    ])
-                    .add(size as i64);
-            }
-            (buf, result)
-        })
+    async fn write_at(&self, buf: &[u8], offset: u64) -> Result<usize, Error> {
+        let result = with_file!(self, StorageIoOperation::Write, |file_guard| {
+            file_guard.with_std_file(|std_file| std_file.write_at(buf, offset))
+        });
+        if let Ok(size) = result {
+            STORAGE_IO_SIZE
+                .with_label_values(&["write", &self.tenant_id, &self.shard_id, &self.timeline_id])
+                .add(size as i64);
+        }
+        result
     }
 }
 
@@ -1122,7 +1083,6 @@ mod tests {
     use rand::Rng;
     use std::future::Future;
     use std::io::Write;
-    use std::os::unix::fs::FileExt;
     use std::sync::Arc;
 
     enum MaybeVirtualFile {
@@ -1143,11 +1103,7 @@ mod tests {
                 MaybeVirtualFile::File(file) => file.read_exact_at(&mut buf, offset).map(|()| buf),
             }
         }
-        async fn write_all_at<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
-            &self,
-            buf: B,
-            offset: u64,
-        ) -> Result<(), Error> {
+        async fn write_all_at<B: BoundedBuf>(&self, buf: B, offset: u64) -> Result<(), Error> {
             match self {
                 MaybeVirtualFile::VirtualFile(file) => {
                     let (_buf, res) = file.write_all_at(buf, offset).await;
@@ -1168,10 +1124,7 @@ mod tests {
                 MaybeVirtualFile::File(file) => file.seek(pos),
             }
         }
-        async fn write_all<B: BoundedBuf<Buf = Buf>, Buf: IoBuf + Send>(
-            &mut self,
-            buf: B,
-        ) -> Result<(), Error> {
+        async fn write_all<B: BoundedBuf>(&mut self, buf: B) -> Result<(), Error> {
             match self {
                 MaybeVirtualFile::VirtualFile(file) => {
                     let (_buf, res) = file.write_all(buf).await;

pageserver/src/virtual_file/io_engine.rs

@@ -7,9 +7,6 @@
 //!
 //! Then use [`get`] and  [`super::OpenOptions`].
 
-use tokio_epoll_uring::{IoBuf, Slice};
-use tracing::Instrument;
-
 pub(crate) use super::api::IoEngineKind;
 #[derive(Clone, Copy)]
 #[repr(u8)]
@@ -64,8 +61,7 @@ pub(super) fn init(engine_kind: IoEngineKind) {
     set(engine_kind);
 }
 
-/// Longer-term, this API should only be used by [`super::VirtualFile`].
-pub(crate) fn get() -> IoEngine {
+pub(super) fn get() -> IoEngine {
     let cur = IoEngine::try_from(IO_ENGINE.load(Ordering::Relaxed)).unwrap();
     if cfg!(test) {
         let env_var_name = "NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOENGINE";
@@ -102,17 +98,7 @@ use std::{
     sync::atomic::{AtomicU8, Ordering},
 };
 
-use super::{FileGuard, Metadata};
-
-#[cfg(target_os = "linux")]
-fn epoll_uring_error_to_std(e: tokio_epoll_uring::Error<std::io::Error>) -> std::io::Error {
-    match e {
-        tokio_epoll_uring::Error::Op(e) => e,
-        tokio_epoll_uring::Error::System(system) => {
-            std::io::Error::new(std::io::ErrorKind::Other, system)
-        }
-    }
-}
+use super::FileGuard;
 
 impl IoEngine {
     pub(super) async fn read_at<B>(
@@ -147,109 +133,16 @@ impl IoEngine {
             IoEngine::TokioEpollUring => {
                 let system = tokio_epoll_uring::thread_local_system().await;
                 let (resources, res) = system.read(file_guard, offset, buf).await;
-                (resources, res.map_err(epoll_uring_error_to_std))
-            }
-        }
-    }
-    pub(super) async fn sync_all(&self, file_guard: FileGuard) -> (FileGuard, std::io::Result<()>) {
-        match self {
-            IoEngine::NotSet => panic!("not initialized"),
-            IoEngine::StdFs => {
-                let res = file_guard.with_std_file(|std_file| std_file.sync_all());
-                (file_guard, res)
-            }
-            #[cfg(target_os = "linux")]
-            IoEngine::TokioEpollUring => {
-                let system = tokio_epoll_uring::thread_local_system().await;
-                let (resources, res) = system.fsync(file_guard).await;
-                (resources, res.map_err(epoll_uring_error_to_std))
-            }
-        }
-    }
-    pub(super) async fn sync_data(
-        &self,
-        file_guard: FileGuard,
-    ) -> (FileGuard, std::io::Result<()>) {
-        match self {
-            IoEngine::NotSet => panic!("not initialized"),
-            IoEngine::StdFs => {
-                let res = file_guard.with_std_file(|std_file| std_file.sync_data());
-                (file_guard, res)
-            }
-            #[cfg(target_os = "linux")]
-            IoEngine::TokioEpollUring => {
-                let system = tokio_epoll_uring::thread_local_system().await;
-                let (resources, res) = system.fdatasync(file_guard).await;
-                (resources, res.map_err(epoll_uring_error_to_std))
-            }
-        }
-    }
-    pub(super) async fn metadata(
-        &self,
-        file_guard: FileGuard,
-    ) -> (FileGuard, std::io::Result<Metadata>) {
-        match self {
-            IoEngine::NotSet => panic!("not initialized"),
-            IoEngine::StdFs => {
-                let res =
-                    file_guard.with_std_file(|std_file| std_file.metadata().map(Metadata::from));
-                (file_guard, res)
-            }
-            #[cfg(target_os = "linux")]
-            IoEngine::TokioEpollUring => {
-                let system = tokio_epoll_uring::thread_local_system().await;
-                let (resources, res) = system.statx(file_guard).await;
                 (
                     resources,
-                    res.map_err(epoll_uring_error_to_std).map(Metadata::from),
+                    res.map_err(|e| match e {
+                        tokio_epoll_uring::Error::Op(e) => e,
+                        tokio_epoll_uring::Error::System(system) => {
+                            std::io::Error::new(std::io::ErrorKind::Other, system)
+                        }
+                    }),
                 )
             }
         }
     }
-    pub(super) async fn write_at<B: IoBuf + Send>(
-        &self,
-        file_guard: FileGuard,
-        offset: u64,
-        buf: Slice<B>,
-    ) -> ((FileGuard, Slice<B>), std::io::Result<usize>) {
-        match self {
-            IoEngine::NotSet => panic!("not initialized"),
-            IoEngine::StdFs => {
-                let result = file_guard.with_std_file(|std_file| std_file.write_at(&buf, offset));
-                ((file_guard, buf), result)
-            }
-            #[cfg(target_os = "linux")]
-            IoEngine::TokioEpollUring => {
-                let system = tokio_epoll_uring::thread_local_system().await;
-                let (resources, res) = system.write(file_guard, offset, buf).await;
-                (resources, res.map_err(epoll_uring_error_to_std))
-            }
-        }
-    }
-
-    /// If we switch a user of [`tokio::fs`] to use [`super::io_engine`],
-    /// they'd start blocking the executor thread if [`IoEngine::StdFs`] is configured
-    /// whereas before the switch to [`super::io_engine`], that wasn't the case.
-    /// This method helps avoid such a regression.
-    ///
-    /// Panics if the `spawn_blocking` fails, see [`tokio::task::JoinError`] for reasons why that can happen.
-    pub(crate) async fn spawn_blocking_and_block_on_if_std<Fut, R>(&self, work: Fut) -> R
-    where
-        Fut: 'static + Send + std::future::Future<Output = R>,
-        R: 'static + Send,
-    {
-        match self {
-            IoEngine::NotSet => panic!("not initialized"),
-            IoEngine::StdFs => {
-                let span = tracing::info_span!("spawn_blocking_block_on_if_std");
-                tokio::task::spawn_blocking({
-                    move || tokio::runtime::Handle::current().block_on(work.instrument(span))
-                })
-                .await
-                .expect("failed to join blocking code most likely it panicked, panicking as well")
-            }
-            #[cfg(target_os = "linux")]
-            IoEngine::TokioEpollUring => work.await,
-        }
-    }
 }

pageserver/src/virtual_file/metadata.rs

@@ -1,30 +0,0 @@
-use std::fs;
-
-pub enum Metadata {
-    StdFs(fs::Metadata),
-    #[cfg(target_os = "linux")]
-    TokioEpollUring(Box<tokio_epoll_uring::ops::statx::statx>),
-}
-
-#[cfg(target_os = "linux")]
-impl From<Box<tokio_epoll_uring::ops::statx::statx>> for Metadata {
-    fn from(value: Box<tokio_epoll_uring::ops::statx::statx>) -> Self {
-        Metadata::TokioEpollUring(value)
-    }
-}
-
-impl From<std::fs::Metadata> for Metadata {
-    fn from(value: std::fs::Metadata) -> Self {
-        Metadata::StdFs(value)
-    }
-}
-
-impl Metadata {
-    pub fn len(&self) -> u64 {
-        match self {
-            Metadata::StdFs(metadata) => metadata.len(),
-            #[cfg(target_os = "linux")]
-            Metadata::TokioEpollUring(statx) => statx.stx_size,
-        }
-    }
-}

pgxn/neon/control_plane_connector.c

@@ -35,7 +35,6 @@
 #include "utils/memutils.h"
 #include "utils/jsonb.h"
 
-#include "control_plane_connector.h"
 #include "neon_utils.h"
 
 static ProcessUtility_hook_type PreviousProcessUtilityHook = NULL;
@@ -114,8 +113,6 @@ ConstructDeltaMessage()
 	if (RootTable.db_table)
 	{
 		JsonbValue	dbs;
-		HASH_SEQ_STATUS status;
-		DbEntry    *entry;
 
 		dbs.type = jbvString;
 		dbs.val.string.val = "dbs";
@@ -123,6 +120,9 @@ ConstructDeltaMessage()
 		pushJsonbValue(&state, WJB_KEY, &dbs);
 		pushJsonbValue(&state, WJB_BEGIN_ARRAY, NULL);
 
+		HASH_SEQ_STATUS status;
+		DbEntry    *entry;
+
 		hash_seq_init(&status, RootTable.db_table);
 		while ((entry = hash_seq_search(&status)) != NULL)
 		{
@@ -168,9 +168,8 @@ ConstructDeltaMessage()
 #else
 				const char *logdetail;
 #endif
-				char	   *encrypted_password;
 				PushKeyValue(&state, "password", (char *) entry->password);
-				encrypted_password = get_role_password(entry->name, &logdetail);
+				char	   *encrypted_password = get_role_password(entry->name, &logdetail);
 
 				if (encrypted_password)
 				{
@@ -832,7 +831,7 @@ NeonProcessUtility(
 	}
 }
 
-void
+extern void
 InitControlPlaneConnector()
 {
 	PreviousProcessUtilityHook = ProcessUtility_hook;

pgxn/neon/control_plane_connector.h

@@ -1,6 +1,6 @@
 #ifndef CONTROL_PLANE_CONNECTOR_H
 #define CONTROL_PLANE_CONNECTOR_H
 
-void		InitControlPlaneConnector(void);
+void		InitControlPlaneConnector();
 
 #endif

pgxn/neon/extension_server.c

@@ -14,7 +14,6 @@
 
 #include "utils/guc.h"
 
-#include "extension_server.h" 
 #include "neon_utils.h"
 
 static int	extension_server_port = 0;

pgxn/neon/extension_server.h

@@ -1,17 +0,0 @@
-/*-------------------------------------------------------------------------
- *
- * extension_server.h
- *	  Request compute_ctl to download extension files.
- *
- * IDENTIFICATION
- *	 contrib/neon/extension_server.h
- *
- *-------------------------------------------------------------------------
- */
-
-#ifndef EXTENSION_SERVER_H
-#define EXTENSION_SERVER_H
-
-void pg_init_extension_server(void);
-
-#endif							/* EXTENSION_SERVER_H */

pgxn/neon/neon.c

@@ -29,7 +29,6 @@
 #include "utils/guc.h"
 #include "utils/wait_event.h"
 
-#include "extension_server.h"
 #include "neon.h"
 #include "walproposer.h"
 #include "pagestore_client.h"

pgxn/neon/neon.h

@@ -25,11 +25,12 @@ extern int	wal_acceptor_connection_timeout;
 extern void pg_init_libpagestore(void);
 extern void pg_init_walproposer(void);
 
+extern void pg_init_extension_server(void);
+
 extern uint64 BackpressureThrottlingTime(void);
 extern void replication_feedback_get_lsns(XLogRecPtr *writeLsn, XLogRecPtr *flushLsn, XLogRecPtr *applyLsn);
 
 extern void PGDLLEXPORT WalProposerSync(int argc, char *argv[]);
 extern void PGDLLEXPORT WalProposerMain(Datum main_arg);
-PGDLLEXPORT void LogicalSlotsMonitorMain(Datum main_arg);
 
 #endif							/* NEON_H */

pgxn/neon/neon_utils.c

@@ -6,7 +6,6 @@
 
 #include "postgres.h"
 
-#include "neon_utils.h"
 #include "lib/stringinfo.h"
 #include "libpq/pqformat.h"
 
@@ -15,7 +14,7 @@
  *
  * Returns -1 if the character is not a hexadecimal digit.
  */
-static int
+int
 HexDecodeChar(char c)
 {
 	if (c >= '0' && c <= '9')

pgxn/neon/neon_utils.h

@@ -12,7 +12,7 @@ uint32		pq_getmsgint32_le(StringInfo msg);
 uint64		pq_getmsgint64_le(StringInfo msg);
 void		pq_sendint32_le(StringInfo buf, uint32 i);
 void		pq_sendint64_le(StringInfo buf, uint64 i);
-void        disable_core_dump(void);
+extern void disable_core_dump();
 
 #ifndef WALPROPOSER_LIB
 

pgxn/neon/walproposer.c

@@ -1460,7 +1460,7 @@ RecvAppendResponses(Safekeeper *sk)
 }
 
 /* Parse a PageserverFeedback message, or the PageserverFeedback part of an AppendResponse */
-static void
+void
 ParsePageserverFeedbackMessage(WalProposer *wp, StringInfo reply_message, PageserverFeedback *rf)
 {
 	uint8		nkeys;
@@ -1590,9 +1590,9 @@ GetAcknowledgedByQuorumWALPosition(WalProposer *wp)
 Safekeeper *
 GetDonor(WalProposer *wp, XLogRecPtr *donor_lsn)
 {
+	*donor_lsn = InvalidXLogRecPtr;
 	Safekeeper *donor = NULL;
 	int			i;
-	*donor_lsn = InvalidXLogRecPtr;
 
 	if (wp->n_votes < wp->quorum)
 	{

pgxn/neon/walproposer_pg.c

@@ -398,7 +398,7 @@ walprop_pg_get_shmem_state(WalProposer *wp)
 	return walprop_shared;
 }
 
-static void
+void
 replication_feedback_set(PageserverFeedback *rf)
 {
 	SpinLockAcquire(&walprop_shared->mutex);

proxy/Cargo.toml

@@ -30,10 +30,6 @@ hostname.workspace = true
 humantime.workspace = true
 hyper-tungstenite.workspace = true
 hyper.workspace = true
-hyper1 = { package = "hyper", version = "1.2", features = ["server", "http1", "http2"] }
-hyper-util = { version = "0.1", features = ["tokio"] }
-http1 = { package = "http", version = "1" }
-http-body-util = { version = "0.1" }
 ipnet.workspace = true
 itertools.workspace = true
 lasso = { workspace = true, features = ["multi-threaded"] }
@@ -72,6 +68,7 @@ task-local-extensions.workspace = true
 thiserror.workspace = true
 tikv-jemallocator.workspace = true
 tikv-jemalloc-ctl = { workspace = true, features = ["use_std"] }
+tls-listener.workspace = true
 tokio-postgres.workspace = true
 tokio-rustls.workspace = true
 tokio-util.workspace = true

proxy/src/auth/backend/link.rs

@@ -102,8 +102,6 @@ pub(super) async fn authenticate(
 
     ctx.set_user(db_info.user.into());
     ctx.set_project(db_info.aux.clone());
-    let cold_start_info = db_info.aux.cold_start_info.clone().unwrap_or_default();
-    info!(?cold_start_info, "woken up a compute node");
 
     // Backwards compatibility. pg_sni_proxy uses "--" in domain names
     // while direct connections do not. Once we migrate to pg_sni_proxy

proxy/src/bin/pg_sni_router.rs

@@ -10,7 +10,6 @@ use itertools::Itertools;
 use proxy::config::TlsServerEndPoint;
 use proxy::context::RequestMonitoring;
 use proxy::proxy::run_until_cancelled;
-use rustls::pki_types::PrivateKeyDer;
 use tokio::net::TcpListener;
 
 use anyhow::{anyhow, bail, ensure, Context};
@@ -77,40 +76,37 @@ async fn main() -> anyhow::Result<()> {
         (Some(key_path), Some(cert_path)) => {
             let key = {
                 let key_bytes = std::fs::read(key_path).context("TLS key file")?;
-
-                let mut keys =
-                    rustls_pemfile::pkcs8_private_keys(&mut &key_bytes[..]).collect_vec();
+                let mut keys = rustls_pemfile::pkcs8_private_keys(&mut &key_bytes[..])
+                    .context(format!("Failed to read TLS keys at '{key_path}'"))?;
 
                 ensure!(keys.len() == 1, "keys.len() = {} (should be 1)", keys.len());
-                PrivateKeyDer::Pkcs8(
-                    keys.pop()
-                        .unwrap()
-                        .context(format!("Failed to read TLS keys at '{key_path}'"))?,
-                )
+                keys.pop().map(rustls::PrivateKey).unwrap()
             };
 
             let cert_chain_bytes = std::fs::read(cert_path)
                 .context(format!("Failed to read TLS cert file at '{cert_path}.'"))?;
 
-            let cert_chain: Vec<_> = {
+            let cert_chain = {
                 rustls_pemfile::certs(&mut &cert_chain_bytes[..])
-                .try_collect()
-                .with_context(|| {
-                    format!("Failed to read TLS certificate chain from bytes from file at '{cert_path}'.")
-                })?
+                    .context(format!(
+                        "Failed to read TLS certificate chain from bytes from file at '{cert_path}'."
+                    ))?
+                    .into_iter()
+                    .map(rustls::Certificate)
+                    .collect_vec()
             };
 
             // needed for channel bindings
             let first_cert = cert_chain.first().context("missing certificate")?;
             let tls_server_end_point = TlsServerEndPoint::new(first_cert)?;
 
-            let tls_config = rustls::ServerConfig::builder_with_protocol_versions(&[
-                &rustls::version::TLS13,
-                &rustls::version::TLS12,
-            ])
-            .with_no_client_auth()
-            .with_single_cert(cert_chain, key)?
-            .into();
+            let tls_config = rustls::ServerConfig::builder()
+                .with_safe_default_cipher_suites()
+                .with_safe_default_kx_groups()
+                .with_protocol_versions(&[&rustls::version::TLS13, &rustls::version::TLS12])?
+                .with_no_client_auth()
+                .with_single_cert(cert_chain, key)?
+                .into();
 
             (tls_config, tls_server_end_point)
         }
@@ -175,7 +171,7 @@ async fn task_main(
                     .context("failed to set socket option")?;
 
                 info!(%peer_addr, "serving");
-                let ctx = RequestMonitoring::new(session_id, peer_addr, "sni_router", "sni");
+                let ctx = RequestMonitoring::new(session_id, peer_addr.ip(), "sni_router", "sni");
                 handle_client(ctx, dest_suffix, tls_config, tls_server_end_point, socket).await
             }
             .unwrap_or_else(|e| {

proxy/src/config.rs

@@ -1,10 +1,6 @@
 use crate::{auth, rate_limiter::RateBucketInfo, serverless::GlobalConnPoolOptions};
 use anyhow::{bail, ensure, Context, Ok};
-use itertools::Itertools;
-use rustls::{
-    crypto::ring::sign,
-    pki_types::{CertificateDer, PrivateKeyDer},
-};
+use rustls::{sign, Certificate, PrivateKey};
 use sha2::{Digest, Sha256};
 use std::{
     collections::{HashMap, HashSet},
@@ -92,14 +88,14 @@ pub fn configure_tls(
 
     let cert_resolver = Arc::new(cert_resolver);
 
-    // allow TLS 1.2 to be compatible with older client libraries
-    let config = rustls::ServerConfig::builder_with_protocol_versions(&[
-        &rustls::version::TLS13,
-        &rustls::version::TLS12,
-    ])
-    .with_no_client_auth()
-    .with_cert_resolver(cert_resolver.clone())
-    .into();
+    let config = rustls::ServerConfig::builder()
+        .with_safe_default_cipher_suites()
+        .with_safe_default_kx_groups()
+        // allow TLS 1.2 to be compatible with older client libraries
+        .with_protocol_versions(&[&rustls::version::TLS13, &rustls::version::TLS12])?
+        .with_no_client_auth()
+        .with_cert_resolver(cert_resolver.clone())
+        .into();
 
     Ok(TlsConfig {
         config,
@@ -137,14 +133,14 @@ pub enum TlsServerEndPoint {
 }
 
 impl TlsServerEndPoint {
-    pub fn new(cert: &CertificateDer) -> anyhow::Result<Self> {
+    pub fn new(cert: &Certificate) -> 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,
             oid_registry::OID_PKCS1_SHA256WITHRSA,
         ];
 
-        let pem = x509_parser::parse_x509_certificate(cert)
+        let pem = x509_parser::parse_x509_certificate(&cert.0)
             .context("Failed to parse PEM object from cerficiate")?
             .1;
 
@@ -154,7 +150,8 @@ impl TlsServerEndPoint {
         let oid = pem.signature_algorithm.oid();
         let alg = reg.get(oid);
         if sha256_oids.contains(oid) {
-            let tls_server_end_point: [u8; 32] = Sha256::new().chain_update(cert).finalize().into();
+            let tls_server_end_point: [u8; 32] =
+                Sha256::new().chain_update(&cert.0).finalize().into();
             info!(subject = %pem.subject, signature_algorithm = alg.map(|a| a.description()), tls_server_end_point = %base64::encode(tls_server_end_point), "determined channel binding");
             Ok(Self::Sha256(tls_server_end_point))
         } else {
@@ -168,7 +165,7 @@ impl TlsServerEndPoint {
     }
 }
 
-#[derive(Default, Debug)]
+#[derive(Default)]
 pub struct CertResolver {
     certs: HashMap<String, (Arc<rustls::sign::CertifiedKey>, TlsServerEndPoint)>,
     default: Option<(Arc<rustls::sign::CertifiedKey>, TlsServerEndPoint)>,
@@ -188,14 +185,11 @@ impl CertResolver {
         let priv_key = {
             let key_bytes = std::fs::read(key_path)
                 .context(format!("Failed to read TLS keys at '{key_path}'"))?;
-            let mut keys = rustls_pemfile::pkcs8_private_keys(&mut &key_bytes[..]).collect_vec();
+            let mut keys = rustls_pemfile::pkcs8_private_keys(&mut &key_bytes[..])
+                .context(format!("Failed to parse TLS keys at '{key_path}'"))?;
 
             ensure!(keys.len() == 1, "keys.len() = {} (should be 1)", keys.len());
-            PrivateKeyDer::Pkcs8(
-                keys.pop()
-                    .unwrap()
-                    .context(format!("Failed to parse TLS keys at '{key_path}'"))?,
-            )
+            keys.pop().map(rustls::PrivateKey).unwrap()
         };
 
         let cert_chain_bytes = std::fs::read(cert_path)
@@ -203,10 +197,14 @@ impl CertResolver {
 
         let cert_chain = {
             rustls_pemfile::certs(&mut &cert_chain_bytes[..])
-                .try_collect()
                 .with_context(|| {
-                    format!("Failed to read TLS certificate chain from bytes from file at '{cert_path}'.")
+                    format!(
+                    "Failed to read TLS certificate chain from bytes from file at '{cert_path}'."
+                )
                 })?
+                .into_iter()
+                .map(rustls::Certificate)
+                .collect()
         };
 
         self.add_cert(priv_key, cert_chain, is_default)
@@ -214,15 +212,15 @@ impl CertResolver {
 
     pub fn add_cert(
         &mut self,
-        priv_key: PrivateKeyDer<'static>,
-        cert_chain: Vec<CertificateDer<'static>>,
+        priv_key: PrivateKey,
+        cert_chain: Vec<Certificate>,
         is_default: bool,
     ) -> anyhow::Result<()> {
         let key = sign::any_supported_type(&priv_key).context("invalid private key")?;
 
         let first_cert = &cert_chain[0];
         let tls_server_end_point = TlsServerEndPoint::new(first_cert)?;
-        let pem = x509_parser::parse_x509_certificate(first_cert)
+        let pem = x509_parser::parse_x509_certificate(&first_cert.0)
             .context("Failed to parse PEM object from cerficiate")?
             .1;
 

proxy/src/console/messages.rs

@@ -101,10 +101,9 @@ pub struct MetricsAuxInfo {
     pub cold_start_info: Option<ColdStartInfo>,
 }
 
-#[derive(Debug, Default, Serialize, Deserialize, Clone)]
+#[derive(Debug, Serialize, Deserialize, Clone)]
 #[serde(rename_all = "snake_case")]
 pub enum ColdStartInfo {
-    #[default]
     Unknown = 0,
     Warm = 1,
     PoolHit = 2,

proxy/src/console/provider/neon.rs

@@ -134,7 +134,9 @@ impl Api {
             let start = Instant::now();
             let response = self.endpoint.execute(request).await?;
             info!(duration = ?start.elapsed(), "received http response");
+            info!(response = ?response, "this is response");
             let body = parse_body::<WakeCompute>(response).await?;
+            info!(body = ?body, "parsed response body");
 
             // Unfortunately, ownership won't let us use `Option::ok_or` here.
             let (host, port) = match parse_host_port(&body.address) {
@@ -259,9 +261,6 @@ impl super::Api for Api {
         }
 
         let node = self.do_wake_compute(ctx, user_info).await?;
-        ctx.set_project(node.aux.clone());
-        let cold_start_info = node.aux.cold_start_info.clone().unwrap_or_default();
-        info!(?cold_start_info, "woken up a compute node");
         let (_, cached) = self.caches.node_info.insert(key.clone(), node);
         info!(key = &*key, "created a cache entry for compute node info");
 

proxy/src/context.rs

@@ -3,7 +3,7 @@
 use chrono::Utc;
 use once_cell::sync::OnceCell;
 use smol_str::SmolStr;
-use std::net::{IpAddr, SocketAddr};
+use std::net::IpAddr;
 use tokio::sync::mpsc;
 use tracing::{field::display, info_span, Span};
 use uuid::Uuid;
@@ -62,7 +62,7 @@ pub enum AuthMethod {
 impl RequestMonitoring {
     pub fn new(
         session_id: Uuid,
-        peer_addr: SocketAddr,
+        peer_addr: IpAddr,
         protocol: &'static str,
         region: &'static str,
     ) -> Self {
@@ -75,7 +75,7 @@ impl RequestMonitoring {
         );
 
         Self {
-            peer_addr: peer_addr.ip(),
+            peer_addr,
             session_id,
             protocol,
             first_packet: Utc::now(),
@@ -100,12 +100,7 @@ impl RequestMonitoring {
 
     #[cfg(test)]
     pub fn test() -> Self {
-        RequestMonitoring::new(
-            Uuid::now_v7(),
-            ([127, 0, 0, 1], 5432).into(),
-            "test",
-            "test",
-        )
+        RequestMonitoring::new(Uuid::now_v7(), [127, 0, 0, 1].into(), "test", "test")
     }
 
     pub fn console_application_name(&self) -> String {

proxy/src/metrics.rs

@@ -4,7 +4,7 @@ use ::metrics::{
     register_int_gauge_vec, Histogram, HistogramVec, HyperLogLogVec, IntCounterPairVec,
     IntCounterVec, IntGauge, IntGaugeVec,
 };
-use metrics::{register_int_counter, register_int_counter_pair, IntCounter, IntCounterPair};
+use metrics::{register_int_counter_pair, IntCounterPair};
 
 use once_cell::sync::Lazy;
 use tokio::time;
@@ -303,20 +303,3 @@ pub static ENDPOINT_ERRORS_BY_KIND: Lazy<HyperLogLogVec<32>> = Lazy::new(|| {
     )
     .unwrap()
 });
-
-pub static REDIS_BROKEN_MESSAGES: Lazy<IntCounterVec> = Lazy::new(|| {
-    register_int_counter_vec!(
-        "proxy_redis_errors_total",
-        "Number of errors by a given classification",
-        &["channel"],
-    )
-    .unwrap()
-});
-
-pub static TLS_HANDSHAKE_FAILURES: Lazy<IntCounter> = Lazy::new(|| {
-    register_int_counter!(
-        "proxy_tls_handshake_failures",
-        "Number of TLS handshake failures",
-    )
-    .unwrap()
-});

proxy/src/protocol2.rs

@@ -1,7 +1,8 @@
 //! Proxy Protocol V2 implementation
 
 use std::{
-    future::{poll_fn, Future},
+    future::poll_fn,
+    future::Future,
     io,
     net::SocketAddr,
     pin::{pin, Pin},
@@ -9,13 +10,13 @@ use std::{
 };
 
 use bytes::{Buf, BytesMut};
-use hyper::server::conn::AddrIncoming;
+use hyper::server::conn::{AddrIncoming, AddrStream};
 use pin_project_lite::pin_project;
+use tls_listener::AsyncAccept;
 use tokio::io::{AsyncRead, AsyncReadExt, AsyncWrite, ReadBuf};
 
 pub struct ProxyProtocolAccept {
     pub incoming: AddrIncoming,
-    pub protocol: &'static str,
 }
 
 pin_project! {
@@ -325,6 +326,24 @@ impl<T: AsyncRead> AsyncRead for WithClientIp<T> {
     }
 }
 
+impl AsyncAccept for ProxyProtocolAccept {
+    type Connection = WithClientIp<AddrStream>;
+
+    type Error = io::Error;
+
+    fn poll_accept(
+        mut self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Option<Result<Self::Connection, Self::Error>>> {
+        let conn = ready!(Pin::new(&mut self.incoming).poll_accept(cx)?);
+        let Some(conn) = conn else {
+            return Poll::Ready(None);
+        };
+
+        Poll::Ready(Some(Ok(WithClientIp::new(conn))))
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use std::pin::pin;

proxy/src/proxy.rs

@@ -24,7 +24,6 @@ use crate::{
 };
 use futures::TryFutureExt;
 use itertools::Itertools;
-use metrics::IntCounterPairGuard;
 use once_cell::sync::OnceCell;
 use pq_proto::{BeMessage as Be, StartupMessageParams};
 use regex::Regex;
@@ -79,20 +78,15 @@ pub async fn task_main(
     {
         let (socket, peer_addr) = accept_result?;
 
-        let conn_gauge = NUM_CLIENT_CONNECTION_GAUGE
-            .with_label_values(&["tcp"])
-            .guard();
-
         let session_id = uuid::Uuid::new_v4();
         let cancellation_handler = Arc::clone(&cancellation_handler);
         let endpoint_rate_limiter = endpoint_rate_limiter.clone();
 
-        tracing::info!(protocol = "tcp", %session_id, "accepted new TCP connection");
-
         connections.spawn(async move {
             let mut socket = WithClientIp::new(socket);
-            let peer_addr = match socket.wait_for_addr().await {
-                Ok(Some(addr)) => addr,
+            let mut peer_addr = peer_addr.ip();
+            match socket.wait_for_addr().await {
+                Ok(Some(addr)) => peer_addr = addr.ip(),
                 Err(e) => {
                     error!("per-client task finished with an error: {e:#}");
                     return;
@@ -101,8 +95,8 @@ pub async fn task_main(
                     error!("missing required client IP");
                     return;
                 }
-                Ok(None) => peer_addr
-            };
+                Ok(None) => {}
+            }
 
             match socket.inner.set_nodelay(true) {
                 Ok(()) => {},
@@ -122,7 +116,6 @@ pub async fn task_main(
                 socket,
                 ClientMode::Tcp,
                 endpoint_rate_limiter,
-                conn_gauge,
             )
             .instrument(span.clone())
             .await;
@@ -236,11 +229,13 @@ pub async fn handle_client<S: AsyncRead + AsyncWrite + Unpin>(
     stream: S,
     mode: ClientMode,
     endpoint_rate_limiter: Arc<EndpointRateLimiter>,
-    conn_gauge: IntCounterPairGuard,
 ) -> Result<Option<ProxyPassthrough<S>>, ClientRequestError> {
     info!("handling interactive connection from client");
 
     let proto = ctx.protocol;
+    let _client_gauge = NUM_CLIENT_CONNECTION_GAUGE
+        .with_label_values(&[proto])
+        .guard();
     let _request_gauge = NUM_CONNECTION_REQUESTS_GAUGE
         .with_label_values(&[proto])
         .guard();
@@ -330,7 +325,7 @@ pub async fn handle_client<S: AsyncRead + AsyncWrite + Unpin>(
         aux: node.aux.clone(),
         compute: node,
         req: _request_gauge,
-        conn: conn_gauge,
+        conn: _client_gauge,
         cancel: session,
     }))
 }
@@ -379,11 +374,6 @@ impl NeonOptions {
         Self::parse_from_iter(StartupMessageParams::parse_options_raw(options))
     }
 
-    pub fn is_ephemeral(&self) -> bool {
-        // Currently, neon endpoint options are all reserved for ephemeral endpoints.
-        !self.0.is_empty()
-    }
-
     fn parse_from_iter<'a>(options: impl Iterator<Item = &'a str>) -> Self {
         let mut options = options
             .filter_map(neon_option)

proxy/src/proxy/tests.rs

@@ -20,7 +20,6 @@ use crate::{http, sasl, scram};
 use anyhow::{bail, Context};
 use async_trait::async_trait;
 use rstest::rstest;
-use rustls::pki_types;
 use tokio_postgres::config::SslMode;
 use tokio_postgres::tls::{MakeTlsConnect, NoTls};
 use tokio_postgres_rustls::{MakeRustlsConnect, RustlsStream};
@@ -29,11 +28,7 @@ use tokio_postgres_rustls::{MakeRustlsConnect, RustlsStream};
 fn generate_certs(
     hostname: &str,
     common_name: &str,
-) -> anyhow::Result<(
-    pki_types::CertificateDer<'static>,
-    pki_types::CertificateDer<'static>,
-    pki_types::PrivateKeyDer<'static>,
-)> {
+) -> anyhow::Result<(rustls::Certificate, rustls::Certificate, rustls::PrivateKey)> {
     let ca = rcgen::Certificate::from_params({
         let mut params = rcgen::CertificateParams::default();
         params.is_ca = rcgen::IsCa::Ca(rcgen::BasicConstraints::Unconstrained);
@@ -50,9 +45,9 @@ fn generate_certs(
     })?;
 
     Ok((
-        pki_types::CertificateDer::from(ca.serialize_der()?),
-        pki_types::CertificateDer::from(cert.serialize_der_with_signer(&ca)?),
-        pki_types::PrivateKeyDer::Pkcs8(cert.serialize_private_key_der().into()),
+        rustls::Certificate(ca.serialize_der()?),
+        rustls::Certificate(cert.serialize_der_with_signer(&ca)?),
+        rustls::PrivateKey(cert.serialize_private_key_der()),
     ))
 }
 
@@ -87,8 +82,9 @@ fn generate_tls_config<'a>(
 
     let tls_config = {
         let config = rustls::ServerConfig::builder()
+            .with_safe_defaults()
             .with_no_client_auth()
-            .with_single_cert(vec![cert.clone()], key.clone_key())?
+            .with_single_cert(vec![cert.clone()], key.clone())?
             .into();
 
         let mut cert_resolver = CertResolver::new();
@@ -105,9 +101,10 @@ fn generate_tls_config<'a>(
 
     let client_config = {
         let config = rustls::ClientConfig::builder()
+            .with_safe_defaults()
             .with_root_certificates({
                 let mut store = rustls::RootCertStore::empty();
-                store.add(ca)?;
+                store.add(&ca)?;
                 store
             })
             .with_no_client_auth();

proxy/src/redis/notifications.rs

@@ -10,7 +10,6 @@ use crate::{
     cache::project_info::ProjectInfoCache,
     cancellation::{CancelMap, CancellationHandler, NotificationsCancellationHandler},
     intern::{ProjectIdInt, RoleNameInt},
-    metrics::REDIS_BROKEN_MESSAGES,
 };
 
 const CPLANE_CHANNEL_NAME: &str = "neondb-proxy-ws-updates";
@@ -116,9 +115,6 @@ impl<
         let msg: Notification = match serde_json::from_str(&payload) {
             Ok(msg) => msg,
             Err(e) => {
-                REDIS_BROKEN_MESSAGES
-                    .with_label_values(&[msg.get_channel_name()])
-                    .inc();
                 tracing::error!("broken message: {e}");
                 return Ok(());
             }

proxy/src/serverless.rs

@@ -4,45 +4,45 @@
 
 mod backend;
 mod conn_pool;
-mod http_auto;
 mod json;
 mod sql_over_http;
 mod websocket;
 
-use bytes::Bytes;
 pub use conn_pool::GlobalConnPoolOptions;
 
-use anyhow::Context;
-use futures::future::{select, Either};
-use http1::{Method, Response, StatusCode};
-use http_body_util::Full;
-use hyper1::body::Incoming;
+use anyhow::bail;
+use hyper::StatusCode;
+use metrics::IntCounterPairGuard;
 use rand::rngs::StdRng;
 use rand::SeedableRng;
 pub use reqwest_middleware::{ClientWithMiddleware, Error};
 pub use reqwest_retry::{policies::ExponentialBackoff, RetryTransientMiddleware};
-use serde::Serialize;
-use tokio::time::timeout;
 use tokio_util::task::TaskTracker;
 
 use crate::context::RequestMonitoring;
-use crate::metrics::{NUM_CLIENT_CONNECTION_GAUGE, TLS_HANDSHAKE_FAILURES};
-use crate::protocol2::WithClientIp;
-use crate::proxy::run_until_cancelled;
+use crate::metrics::NUM_CLIENT_CONNECTION_GAUGE;
+use crate::protocol2::{ProxyProtocolAccept, WithClientIp};
 use crate::rate_limiter::EndpointRateLimiter;
 use crate::serverless::backend::PoolingBackend;
-use crate::serverless::http_auto::Rewind;
 use crate::{cancellation::CancellationHandler, config::ProxyConfig};
+use futures::StreamExt;
+use hyper::{
+    server::{
+        accept,
+        conn::{AddrIncoming, AddrStream},
+    },
+    Body, Method, Request, Response,
+};
 
 use std::convert::Infallible;
-use std::net::SocketAddr;
-use std::pin::pin;
-use std::sync::Arc;
-use std::time::Duration;
+use std::net::IpAddr;
+use std::task::Poll;
+use std::{future::ready, sync::Arc};
+use tls_listener::TlsListener;
 use tokio::net::TcpListener;
 use tokio_util::sync::CancellationToken;
 use tracing::{error, info, warn, Instrument};
-use utils::http::error::ApiError;
+use utils::http::{error::ApiError, json::json_response};
 
 pub const SERVERLESS_DRIVER_SNI: &str = "api";
 
@@ -94,221 +94,121 @@ pub async fn task_main(
     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 mut addr_incoming = AddrIncoming::from_listener(ws_listener)?;
+    let _ = addr_incoming.set_nodelay(true);
+    let addr_incoming = ProxyProtocolAccept {
+        incoming: addr_incoming,
+    };
+
     let ws_connections = tokio_util::task::task_tracker::TaskTracker::new();
     ws_connections.close(); // allows `ws_connections.wait to complete`
 
-    let http_connections = tokio_util::task::task_tracker::TaskTracker::new();
-    http_connections.close();
-
-    let server = http_auto::Builder::new();
-
-    loop {
-        let Some(res) = run_until_cancelled(ws_listener.accept(), &cancellation_token).await else {
-            break;
-        };
-
-        let (conn, mut peer_addr) = res.context("could not accept TCP stream")?;
-        if let Err(e) = conn.set_nodelay(true) {
-            tracing::error!("could not set nodolay: {e}");
-            continue;
+    let tls_listener = TlsListener::new(tls_acceptor, addr_incoming).filter(|conn| {
+        if let Err(err) = conn {
+            error!("failed to accept TLS connection for websockets: {err:?}");
+            ready(false)
+        } else {
+            ready(true)
         }
-        let cancellation_token = cancellation_token.child_token();
+    });
 
-        let tls = tls_acceptor.clone();
+    let make_svc = hyper::service::make_service_fn(
+        |stream: &tokio_rustls::server::TlsStream<WithClientIp<AddrStream>>| {
+            let (io, _) = stream.get_ref();
+            let client_addr = io.client_addr();
+            let remote_addr = io.inner.remote_addr();
+            let backend = backend.clone();
+            let ws_connections = ws_connections.clone();
+            let endpoint_rate_limiter = endpoint_rate_limiter.clone();
+            let cancellation_handler = cancellation_handler.clone();
+            async move {
+                let peer_addr = match client_addr {
+                    Some(addr) => addr,
+                    None if config.require_client_ip => bail!("missing required client ip"),
+                    None => remote_addr,
+                };
+                Ok(MetricService::new(hyper::service::service_fn(
+                    move |req: Request<Body>| {
+                        let backend = backend.clone();
+                        let ws_connections = ws_connections.clone();
+                        let endpoint_rate_limiter = endpoint_rate_limiter.clone();
+                        let cancellation_handler = cancellation_handler.clone();
 
-        let backend = backend.clone();
-        let ws_connections = ws_connections.clone();
-        let endpoint_rate_limiter = endpoint_rate_limiter.clone();
-        let cancellation_handler = cancellation_handler.clone();
-        let server = server.clone();
-
-        http_connections.spawn(async move {
-            let _gauge = NUM_CLIENT_CONNECTION_GAUGE
-                .with_label_values(&["http"])
-                .guard();
-
-            // handle PROXY protocol
-            let mut conn = WithClientIp::new(conn);
-            let peer = match conn.wait_for_addr().await {
-                Ok(peer) => peer,
-                Err(e) => {
-                    tracing::error!(
-                        "failed to accept TCP connection: invalid PROXY protocol V2 header: {e:#}"
-                    );
-                    return;
-                }
-            };
-
-            if let Some(peer) = peer {
-                peer_addr = peer;
-            }
-            info!(%peer_addr, protocol = "http", "accepted new TCP connection");
-
-            let accept = tls.accept(conn);
-            let conn = match timeout(Duration::from_secs(10), accept).await {
-                Ok(Ok(conn)) => {
-                    info!(%peer_addr, protocol = "http", "accepted new TLS connection");
-                    conn
-                }
-                // The handshake failed, try getting another connection from the queue
-                Ok(Err(e)) => {
-                    TLS_HANDSHAKE_FAILURES.inc();
-                    warn!(%peer_addr, protocol = "http", "failed to accept TLS connection: {e:?}");
-                    return;
-                }
-                // The handshake timed out, try getting another connection from the queue
-                Err(_) => {
-                    TLS_HANDSHAKE_FAILURES.inc();
-                    warn!(%peer_addr, protocol = "http", "failed to accept TLS connection: timeout");
-                    return;
-                }
-            };
-
-            let (version, conn) = match conn.get_ref().1.alpn_protocol() {
-                Some(b"http/1.1") => (http_auto::Version::H1, Rewind::new(conn)),
-                Some(b"h2") => (http_auto::Version::H2, Rewind::new(conn)),
-                _ => {
-                    tracing::debug!("HTTP: no ALPN negotiated");
-                    let conn = timeout(Duration::from_secs(10), http_auto::read_version(conn)).await;
-                    match conn {
-                        Ok(Ok(v)) => v,
-                        Ok(Err(e)) => {
-                            tracing::warn!("HTTP connection error: {e}");
-                            return;
-                        },
-                        Err(_) => {
-                            tracing::warn!("HTTP connection error: timeout determining http version");
-                            return;
-                        }
-                    }
-                }
-            };
-
-            let conn = server.serve_connection_with_upgrades(
-                conn,
-                version,
-                hyper1::service::service_fn(move |req: hyper1::Request<Incoming>| {
-                    let backend = backend.clone();
-                    let ws_connections = ws_connections.clone();
-                    let endpoint_rate_limiter = endpoint_rate_limiter.clone();
-                    let cancellation_handler = cancellation_handler.clone();
-
-                    async move {
-                        Ok::<_, Infallible>(
-                            request_handler(
-                                req,
-                                config,
-                                backend,
-                                ws_connections,
-                                cancellation_handler,
-                                peer_addr,
-                                endpoint_rate_limiter,
+                        async move {
+                            Ok::<_, Infallible>(
+                                request_handler(
+                                    req,
+                                    config,
+                                    backend,
+                                    ws_connections,
+                                    cancellation_handler,
+                                    peer_addr.ip(),
+                                    endpoint_rate_limiter,
+                                )
+                                .await
+                                .map_or_else(|e| e.into_response(), |r| r),
                             )
-                            .await
-                            .map_or_else(api_error_into_response, |r| r),
-                        )
-                    }
-                })
-            );
-
-
-            let cancel = pin!(cancellation_token.cancelled());
-            let conn = pin!(conn);
-            let res = match select(cancel, conn).await {
-                Either::Left((_cancelled, mut conn)) => {
-                    conn.as_mut().graceful_shutdown();
-                    conn.await
-                }
-                Either::Right((res, _)) => res,
-            };
-
-            match res {
-                Ok(()) => {}
-                Err(e) => {
-                    tracing::warn!("HTTP connection error {e}")
-                }
+                        }
+                    },
+                )))
             }
-        });
-    }
+        },
+    );
+
+    hyper::Server::builder(accept::from_stream(tls_listener))
+        .serve(make_svc)
+        .with_graceful_shutdown(cancellation_token.cancelled())
+        .await?;
 
     // await websocket connections
-    http_connections.wait().await;
     ws_connections.wait().await;
 
     Ok(())
 }
 
-fn api_error_into_response(this: ApiError) -> Response<Full<Bytes>> {
-    match this {
-        ApiError::BadRequest(err) => HttpErrorBody::response_from_msg_and_status(
-            format!("{err:#?}"), // use debug printing so that we give the cause
-            StatusCode::BAD_REQUEST,
-        ),
-        ApiError::Forbidden(_) => {
-            HttpErrorBody::response_from_msg_and_status(this.to_string(), StatusCode::FORBIDDEN)
+struct MetricService<S> {
+    inner: S,
+    _gauge: IntCounterPairGuard,
+}
+
+impl<S> MetricService<S> {
+    fn new(inner: S) -> MetricService<S> {
+        MetricService {
+            inner,
+            _gauge: NUM_CLIENT_CONNECTION_GAUGE
+                .with_label_values(&["http"])
+                .guard(),
         }
-        ApiError::Unauthorized(_) => {
-            HttpErrorBody::response_from_msg_and_status(this.to_string(), StatusCode::UNAUTHORIZED)
-        }
-        ApiError::NotFound(_) => {
-            HttpErrorBody::response_from_msg_and_status(this.to_string(), StatusCode::NOT_FOUND)
-        }
-        ApiError::Conflict(_) => {
-            HttpErrorBody::response_from_msg_and_status(this.to_string(), StatusCode::CONFLICT)
-        }
-        ApiError::PreconditionFailed(_) => HttpErrorBody::response_from_msg_and_status(
-            this.to_string(),
-            StatusCode::PRECONDITION_FAILED,
-        ),
-        ApiError::ShuttingDown => HttpErrorBody::response_from_msg_and_status(
-            "Shutting down".to_string(),
-            StatusCode::SERVICE_UNAVAILABLE,
-        ),
-        ApiError::ResourceUnavailable(err) => HttpErrorBody::response_from_msg_and_status(
-            err.to_string(),
-            StatusCode::SERVICE_UNAVAILABLE,
-        ),
-        ApiError::Timeout(err) => HttpErrorBody::response_from_msg_and_status(
-            err.to_string(),
-            StatusCode::REQUEST_TIMEOUT,
-        ),
-        ApiError::InternalServerError(err) => HttpErrorBody::response_from_msg_and_status(
-            err.to_string(),
-            StatusCode::INTERNAL_SERVER_ERROR,
-        ),
     }
 }
 
-#[derive(Serialize)]
-struct HttpErrorBody {
-    pub msg: String,
-}
+impl<S, ReqBody> hyper::service::Service<Request<ReqBody>> for MetricService<S>
+where
+    S: hyper::service::Service<Request<ReqBody>>,
+{
+    type Response = S::Response;
+    type Error = S::Error;
+    type Future = S::Future;
 
-impl HttpErrorBody {
-    pub fn response_from_msg_and_status(msg: String, status: StatusCode) -> Response<Full<Bytes>> {
-        HttpErrorBody { msg }.to_response(status)
+    fn poll_ready(&mut self, cx: &mut std::task::Context<'_>) -> Poll<Result<(), Self::Error>> {
+        self.inner.poll_ready(cx)
     }
 
-    pub fn to_response(&self, status: StatusCode) -> Response<Full<Bytes>> {
-        Response::builder()
-            .status(status)
-            .header(http1::header::CONTENT_TYPE, "application/json")
-            // we do not have nested maps with non string keys so serialization shouldn't fail
-            .body(Full::new(Bytes::from(serde_json::to_string(self).unwrap())))
-            .unwrap()
+    fn call(&mut self, req: Request<ReqBody>) -> Self::Future {
+        self.inner.call(req)
     }
 }
 
 #[allow(clippy::too_many_arguments)]
 async fn request_handler(
-    mut request: hyper1::Request<Incoming>,
+    mut request: Request<Body>,
     config: &'static ProxyConfig,
     backend: Arc<PoolingBackend>,
     ws_connections: TaskTracker,
     cancellation_handler: Arc<CancellationHandler>,
-    peer_addr: SocketAddr,
+    peer_addr: IpAddr,
     endpoint_rate_limiter: Arc<EndpointRateLimiter>,
-) -> Result<Response<Full<Bytes>>, ApiError> {
+) -> Result<Response<Body>, ApiError> {
     let session_id = uuid::Uuid::new_v4();
 
     let host = request
@@ -347,14 +247,14 @@ async fn request_handler(
 
         // Return the response so the spawned future can continue.
         Ok(response)
-    } else if request.uri().path() == "/sql" && *request.method() == Method::POST {
+    } else if request.uri().path() == "/sql" && request.method() == Method::POST {
         let ctx = RequestMonitoring::new(session_id, peer_addr, "http", &config.region);
         let span = ctx.span.clone();
 
         sql_over_http::handle(config, ctx, request, backend)
             .instrument(span)
             .await
-    } else if request.uri().path() == "/sql" && *request.method() == Method::OPTIONS {
+    } else if request.uri().path() == "/sql" && request.method() == Method::OPTIONS {
         Response::builder()
             .header("Allow", "OPTIONS, POST")
             .header("Access-Control-Allow-Origin", "*")
@@ -364,24 +264,9 @@ async fn request_handler(
             )
             .header("Access-Control-Max-Age", "86400" /* 24 hours */)
             .status(StatusCode::OK) // 204 is also valid, but see: https://developer.mozilla.org/en-US/docs/Web/HTTP/Methods/OPTIONS#status_code
-            .body(Full::new(Bytes::new()))
+            .body(Body::empty())
             .map_err(|e| ApiError::InternalServerError(e.into()))
     } else {
         json_response(StatusCode::BAD_REQUEST, "query is not supported")
     }
 }
-
-fn json_response<T: Serialize>(
-    status: StatusCode,
-    data: T,
-) -> Result<Response<Full<Bytes>>, ApiError> {
-    let json = serde_json::to_string(&data)
-        .context("Failed to serialize JSON response")
-        .map_err(ApiError::InternalServerError)?;
-    let response = Response::builder()
-        .status(status)
-        .header(http1::header::CONTENT_TYPE, "application/json")
-        .body(Full::new(Bytes::from(json)))
-        .map_err(|e| ApiError::InternalServerError(e.into()))?;
-    Ok(response)
-}

proxy/src/serverless/conn_pool.rs

@@ -43,13 +43,8 @@ impl ConnInfo {
         (self.dbname.clone(), self.user_info.user.clone())
     }
 
-    pub fn endpoint_cache_key(&self) -> Option<EndpointCacheKey> {
-        // We don't want to cache http connections for ephemeral endpoints.
-        if self.user_info.options.is_ephemeral() {
-            None
-        } else {
-            Some(self.user_info.endpoint_cache_key())
-        }
+    pub fn endpoint_cache_key(&self) -> EndpointCacheKey {
+        self.user_info.endpoint_cache_key()
     }
 }
 
@@ -365,11 +360,8 @@ impl<C: ClientInnerExt> GlobalConnPool<C> {
         conn_info: &ConnInfo,
     ) -> Result<Option<Client<C>>, HttpConnError> {
         let mut client: Option<ClientInner<C>> = None;
-        let Some(endpoint) = conn_info.endpoint_cache_key() else {
-            return Ok(None);
-        };
 
-        let endpoint_pool = self.get_or_create_endpoint_pool(&endpoint);
+        let endpoint_pool = self.get_or_create_endpoint_pool(&conn_info.endpoint_cache_key());
         if let Some(entry) = endpoint_pool
             .write()
             .get_conn_entry(conn_info.db_and_user())
@@ -463,10 +455,8 @@ pub fn poll_client<C: ClientInnerExt>(
     span.in_scope(|| {
         info!(%conn_info, %session_id, "new connection");
     });
-    let pool = match conn_info.endpoint_cache_key() {
-        Some(endpoint) => Arc::downgrade(&global_pool.get_or_create_endpoint_pool(&endpoint)),
-        None => Weak::new(),
-    };
+    let pool =
+        Arc::downgrade(&global_pool.get_or_create_endpoint_pool(&conn_info.endpoint_cache_key()));
     let pool_clone = pool.clone();
 
     let db_user = conn_info.db_and_user();
@@ -733,9 +723,8 @@ mod tests {
             dbname: "dbname".into(),
             password: "password".as_bytes().into(),
         };
-        let ep_pool = Arc::downgrade(
-            &pool.get_or_create_endpoint_pool(&conn_info.endpoint_cache_key().unwrap()),
-        );
+        let ep_pool =
+            Arc::downgrade(&pool.get_or_create_endpoint_pool(&conn_info.endpoint_cache_key()));
         {
             let mut client = Client::new(create_inner(), conn_info.clone(), ep_pool.clone());
             assert_eq!(0, pool.get_global_connections_count());
@@ -791,9 +780,8 @@ mod tests {
             dbname: "dbname".into(),
             password: "password".as_bytes().into(),
         };
-        let ep_pool = Arc::downgrade(
-            &pool.get_or_create_endpoint_pool(&conn_info.endpoint_cache_key().unwrap()),
-        );
+        let ep_pool =
+            Arc::downgrade(&pool.get_or_create_endpoint_pool(&conn_info.endpoint_cache_key()));
         {
             let mut client = Client::new(create_inner(), conn_info.clone(), ep_pool.clone());
             client.do_drop().unwrap()();

proxy/src/serverless/http_auto.rs

@@ -1,316 +0,0 @@
-//! [`hyper-util`] offers an 'auto' connection to detect whether the connection should be HTTP1 or HTTP2.
-//! There's a bug in this implementation where graceful shutdowns are not properly respected.
-
-use futures::ready;
-use hyper1::body::Body;
-use hyper1::service::HttpService;
-use hyper_util::rt::{TokioExecutor, TokioIo, TokioTimer};
-use std::future::Future;
-use std::marker::PhantomPinned;
-use std::pin::Pin;
-use std::task::{Context, Poll};
-use std::{error::Error as StdError, io, marker::Unpin};
-use tokio::io::{AsyncRead, AsyncWrite, ReadBuf};
-
-use ::http1::{Request, Response};
-use bytes::Bytes;
-use hyper1::{body::Incoming, service::Service};
-
-use hyper1::server::conn::http1;
-use hyper1::{rt::bounds::Http2ServerConnExec, server::conn::http2};
-
-use pin_project_lite::pin_project;
-
-type Error = Box<dyn std::error::Error + Send + Sync>;
-
-type Result<T> = std::result::Result<T, Error>;
-
-const H2_PREFACE: &[u8] = b"PRI * HTTP/2.0\r\n\r\nSM\r\n\r\n";
-
-/// Http1 or Http2 connection builder.
-#[derive(Clone, Debug)]
-pub struct Builder {
-    http1: http1::Builder,
-    http2: http2::Builder<TokioExecutor>,
-}
-
-impl Builder {
-    /// Create a new auto connection builder.
-    pub fn new() -> Self {
-        let mut builder = Self {
-            http1: http1::Builder::new(),
-            http2: http2::Builder::new(TokioExecutor::new()),
-        };
-
-        builder.http1.timer(TokioTimer::new());
-        builder.http2.timer(TokioTimer::new());
-
-        builder
-    }
-
-    /// Bind a connection together with a [`Service`], with the ability to
-    /// handle HTTP upgrades. This requires that the IO object implements
-    /// `Send`.
-    pub fn serve_connection_with_upgrades<I, S, B>(
-        &self,
-        io: Rewind<I>,
-        version: Version,
-        service: S,
-    ) -> UpgradeableConnection<I, S>
-    where
-        S: Service<Request<Incoming>, Response = Response<B>>,
-        S::Future: 'static,
-        S::Error: Into<Box<dyn StdError + Send + Sync>>,
-        B: Body + 'static,
-        B::Error: Into<Box<dyn StdError + Send + Sync>>,
-        I: AsyncRead + AsyncWrite + Unpin + Send + 'static,
-        TokioExecutor: Http2ServerConnExec<S::Future, B>,
-    {
-        match version {
-            Version::H1 => {
-                let conn = self
-                    .http1
-                    .serve_connection(TokioIo::new(io), service)
-                    .with_upgrades();
-                UpgradeableConnection {
-                    state: UpgradeableConnState::H1 { conn },
-                }
-            }
-            Version::H2 => {
-                let conn = self.http2.serve_connection(TokioIo::new(io), service);
-                UpgradeableConnection {
-                    state: UpgradeableConnState::H2 { conn },
-                }
-            }
-        }
-    }
-}
-
-#[derive(Copy, Clone)]
-pub(crate) enum Version {
-    H1,
-    H2,
-}
-
-pub(crate) fn read_version<I>(io: I) -> ReadVersion<I>
-where
-    I: AsyncRead + Unpin,
-{
-    ReadVersion {
-        io: Some(io),
-        buf: [0; 24],
-        filled: 0,
-        version: Version::H2,
-        _pin: PhantomPinned,
-    }
-}
-
-pin_project! {
-    pub(crate) struct ReadVersion<I> {
-        io: Option<I>,
-        buf: [u8; 24],
-        // the amount of `buf` thats been filled
-        filled: usize,
-        version: Version,
-        // Make this future `!Unpin` for compatibility with async trait methods.
-        #[pin]
-        _pin: PhantomPinned,
-    }
-}
-
-impl<I> Future for ReadVersion<I>
-where
-    I: AsyncRead + Unpin,
-{
-    type Output = io::Result<(Version, Rewind<I>)>;
-
-    fn poll(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
-        let this = self.project();
-
-        let mut buf = ReadBuf::new(&mut *this.buf);
-        buf.set_filled(*this.filled);
-
-        // We start as H2 and switch to H1 as soon as we don't have the preface.
-        while buf.filled().len() < H2_PREFACE.len() {
-            let len = buf.filled().len();
-            ready!(Pin::new(this.io.as_mut().unwrap()).poll_read(cx, &mut buf))?;
-            *this.filled = buf.filled().len();
-
-            // We starts as H2 and switch to H1 when we don't get the preface.
-            if buf.filled().len() == len
-                || buf.filled()[len..] != H2_PREFACE[len..buf.filled().len()]
-            {
-                *this.version = Version::H1;
-                break;
-            }
-        }
-
-        let io = this.io.take().unwrap();
-        let buf = buf.filled().to_vec();
-        Poll::Ready(Ok((
-            *this.version,
-            Rewind::new_buffered(io, Bytes::from(buf)),
-        )))
-    }
-}
-
-pin_project! {
-    /// Connection future.
-    pub struct UpgradeableConnection<I, S>
-    where
-        S: HttpService<Incoming>,
-    {
-        #[pin]
-        state: UpgradeableConnState<I, S>,
-    }
-}
-
-type Http1UpgradeableConnection<I, S> =
-    hyper1::server::conn::http1::UpgradeableConnection<TokioIo<Rewind<I>>, S>;
-type Http2Connection<I, S> =
-    hyper1::server::conn::http2::Connection<TokioIo<Rewind<I>>, S, TokioExecutor>;
-
-pin_project! {
-    #[project = UpgradeableConnStateProj]
-    enum UpgradeableConnState<I, S>
-    where
-        S: HttpService<Incoming>,
-    {
-        H1 {
-            #[pin]
-            conn: Http1UpgradeableConnection<I, S>,
-        },
-        H2 {
-            #[pin]
-            conn: Http2Connection<I, S>,
-        },
-    }
-}
-
-impl<I, S, B> UpgradeableConnection<I, S>
-where
-    S: HttpService<Incoming, ResBody = B>,
-    S::Error: Into<Box<dyn StdError + Send + Sync>>,
-    I: AsyncRead + AsyncWrite + Unpin,
-    B: Body + 'static,
-    B::Error: Into<Box<dyn StdError + Send + Sync>>,
-    TokioExecutor: Http2ServerConnExec<S::Future, B>,
-{
-    /// Start a graceful shutdown process for this connection.
-    ///
-    /// This `UpgradeableConnection` should continue to be polled until shutdown can finish.
-    ///
-    /// # Note
-    ///
-    /// This should only be called while the `Connection` future is still nothing. pending. If
-    /// called after `UpgradeableConnection::poll` has resolved, this does nothing.
-    pub fn graceful_shutdown(self: Pin<&mut Self>) {
-        match self.project().state.project() {
-            UpgradeableConnStateProj::H1 { conn } => conn.graceful_shutdown(),
-            UpgradeableConnStateProj::H2 { conn } => conn.graceful_shutdown(),
-        }
-    }
-}
-
-impl<I, S, B> Future for UpgradeableConnection<I, S>
-where
-    S: Service<Request<Incoming>, Response = Response<B>>,
-    S::Future: 'static,
-    S::Error: Into<Box<dyn StdError + Send + Sync>>,
-    B: Body + 'static,
-    B::Error: Into<Box<dyn StdError + Send + Sync>>,
-    I: AsyncRead + AsyncWrite + Unpin + Send + 'static,
-    TokioExecutor: Http2ServerConnExec<S::Future, B>,
-{
-    type Output = Result<()>;
-
-    fn poll(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
-        let mut this = self.as_mut().project();
-        match this.state.as_mut().project() {
-            UpgradeableConnStateProj::H1 { conn } => conn.poll(cx).map_err(Into::into),
-            UpgradeableConnStateProj::H2 { conn } => conn.poll(cx).map_err(Into::into),
-        }
-    }
-}
-
-/// Combine a buffer with an IO, rewinding reads to use the buffer.
-#[derive(Debug)]
-pub(crate) struct Rewind<T> {
-    pre: Option<Bytes>,
-    inner: T,
-}
-
-impl<T> Rewind<T> {
-    pub(crate) fn new(io: T) -> Self {
-        Rewind {
-            pre: None,
-            inner: io,
-        }
-    }
-
-    pub(crate) fn new_buffered(io: T, buf: Bytes) -> Self {
-        Rewind {
-            pre: Some(buf),
-            inner: io,
-        }
-    }
-}
-
-impl<T> AsyncRead for Rewind<T>
-where
-    T: AsyncRead + Unpin,
-{
-    fn poll_read(
-        mut self: Pin<&mut Self>,
-        cx: &mut Context<'_>,
-        buf: &mut ReadBuf<'_>,
-    ) -> Poll<io::Result<()>> {
-        if let Some(prefix) = self.pre.take() {
-            // If there are no remaining bytes, let the bytes get dropped.
-            if !prefix.is_empty() {
-                let copy_len = std::cmp::min(prefix.len(), buf.remaining());
-                buf.put_slice(&prefix[..copy_len]);
-                // Put back what's left
-                if !prefix.is_empty() {
-                    self.pre = Some(prefix);
-                }
-
-                return Poll::Ready(Ok(()));
-            }
-        }
-        Pin::new(&mut self.inner).poll_read(cx, buf)
-    }
-}
-
-impl<T> AsyncWrite for Rewind<T>
-where
-    T: AsyncWrite + Unpin,
-{
-    fn poll_write(
-        mut self: Pin<&mut Self>,
-        cx: &mut Context<'_>,
-        buf: &[u8],
-    ) -> Poll<io::Result<usize>> {
-        Pin::new(&mut self.inner).poll_write(cx, buf)
-    }
-
-    fn poll_write_vectored(
-        mut self: Pin<&mut Self>,
-        cx: &mut Context<'_>,
-        bufs: &[io::IoSlice<'_>],
-    ) -> Poll<io::Result<usize>> {
-        Pin::new(&mut self.inner).poll_write_vectored(cx, bufs)
-    }
-
-    fn poll_flush(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
-        Pin::new(&mut self.inner).poll_flush(cx)
-    }
-
-    fn poll_shutdown(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
-        Pin::new(&mut self.inner).poll_shutdown(cx)
-    }
-
-    fn is_write_vectored(&self) -> bool {
-        self.inner.is_write_vectored()
-    }
-}

proxy/src/serverless/sql_over_http.rs

@@ -1,19 +1,15 @@
 use std::sync::Arc;
 
-use super::json_response;
 use anyhow::bail;
-use bytes::Bytes;
+use futures::pin_mut;
 use futures::StreamExt;
-use http_body_util::BodyExt;
-use http_body_util::Full;
-use hyper1::body::Body;
-use hyper1::body::Incoming;
-use hyper1::header;
-use hyper1::http::HeaderName;
-use hyper1::http::HeaderValue;
-use hyper1::Response;
-use hyper1::StatusCode;
-use hyper1::{HeaderMap, Request};
+use hyper::body::HttpBody;
+use hyper::header;
+use hyper::http::HeaderName;
+use hyper::http::HeaderValue;
+use hyper::Response;
+use hyper::StatusCode;
+use hyper::{Body, HeaderMap, Request};
 use serde_json::json;
 use serde_json::Value;
 use tokio::try_join;
@@ -27,6 +23,7 @@ use tracing::error;
 use tracing::info;
 use url::Url;
 use utils::http::error::ApiError;
+use utils::http::json::json_response;
 
 use crate::auth::backend::ComputeUserInfo;
 use crate::auth::endpoint_sni;
@@ -195,9 +192,9 @@ fn get_conn_info(
 pub async fn handle(
     config: &'static ProxyConfig,
     mut ctx: RequestMonitoring,
-    request: Request<Incoming>,
+    request: Request<Body>,
     backend: Arc<PoolingBackend>,
-) -> Result<Response<Full<Bytes>>, ApiError> {
+) -> Result<Response<Body>, ApiError> {
     let result = tokio::time::timeout(
         config.http_config.request_timeout,
         handle_inner(config, &mut ctx, request, backend),
@@ -304,18 +301,19 @@ pub async fn handle(
         }
     };
 
-    response
-        .headers_mut()
-        .insert("Access-Control-Allow-Origin", HeaderValue::from_static("*"));
+    response.headers_mut().insert(
+        "Access-Control-Allow-Origin",
+        hyper::http::HeaderValue::from_static("*"),
+    );
     Ok(response)
 }
 
 async fn handle_inner(
     config: &'static ProxyConfig,
     ctx: &mut RequestMonitoring,
-    request: Request<Incoming>,
+    request: Request<Body>,
     backend: Arc<PoolingBackend>,
-) -> anyhow::Result<Response<Full<Bytes>>> {
+) -> anyhow::Result<Response<Body>> {
     let _request_gauge = NUM_CONNECTION_REQUESTS_GAUGE
         .with_label_values(&[ctx.protocol])
         .guard();
@@ -372,12 +370,9 @@ async fn handle_inner(
     }
 
     let fetch_and_process_request = async {
-        let body = request
-            .into_body()
-            .collect()
+        let body = hyper::body::to_bytes(request.into_body())
             .await
-            .map_err(anyhow::Error::from)?
-            .to_bytes();
+            .map_err(anyhow::Error::from)?;
         info!(length = body.len(), "request payload read");
         let payload: Payload = serde_json::from_slice(&body)?;
         Ok::<Payload, anyhow::Error>(payload) // Adjust error type accordingly
@@ -496,7 +491,7 @@ async fn handle_inner(
     let body = serde_json::to_string(&result).expect("json serialization should not fail");
     let len = body.len();
     let response = response
-        .body(Full::new(Bytes::from(body)))
+        .body(Body::from(body))
         // only fails if invalid status code or invalid header/values are given.
         // these are not user configurable so it cannot fail dynamically
         .expect("building response payload should not fail");
@@ -536,12 +531,13 @@ async fn query_to_json<T: GenericClient>(
 ) -> anyhow::Result<(ReadyForQueryStatus, Value)> {
     info!("executing query");
     let query_params = data.params;
-    let mut row_stream = std::pin::pin!(client.query_raw_txt(&data.query, query_params).await?);
+    let row_stream = client.query_raw_txt(&data.query, query_params).await?;
     info!("finished executing query");
 
     // Manually drain the stream into a vector to leave row_stream hanging
     // around to get a command tag. Also check that the response is not too
     // big.
+    pin_mut!(row_stream);
     let mut rows: Vec<tokio_postgres::Row> = Vec::new();
     while let Some(row) = row_stream.next().await {
         let row = row?;

proxy/src/serverless/websocket.rs

@@ -3,7 +3,6 @@ use crate::{
     config::ProxyConfig,
     context::RequestMonitoring,
     error::{io_error, ReportableError},
-    metrics::NUM_CLIENT_CONNECTION_GAUGE,
     proxy::{handle_client, ClientMode},
     rate_limiter::EndpointRateLimiter,
 };
@@ -139,10 +138,6 @@ pub async fn serve_websocket(
     endpoint_rate_limiter: Arc<EndpointRateLimiter>,
 ) -> anyhow::Result<()> {
     let websocket = websocket.await?;
-    let conn_gauge = NUM_CLIENT_CONNECTION_GAUGE
-        .with_label_values(&["ws"])
-        .guard();
-
     let res = handle_client(
         config,
         &mut ctx,
@@ -150,7 +145,6 @@ pub async fn serve_websocket(
         WebSocketRw::new(websocket),
         ClientMode::Websockets { hostname },
         endpoint_rate_limiter,
-        conn_gauge,
     )
     .await;
 

proxy/src/stream.rs

@@ -1,6 +1,5 @@
 use crate::config::TlsServerEndPoint;
 use crate::error::{ErrorKind, ReportableError, UserFacingError};
-use crate::metrics::TLS_HANDSHAKE_FAILURES;
 use bytes::BytesMut;
 
 use pq_proto::framed::{ConnectionError, Framed};
@@ -225,10 +224,7 @@ impl<S: AsyncRead + AsyncWrite + Unpin> Stream<S> {
     /// If possible, upgrade raw stream into a secure TLS-based stream.
     pub async fn upgrade(self, cfg: Arc<ServerConfig>) -> Result<TlsStream<S>, StreamUpgradeError> {
         match self {
-            Stream::Raw { raw } => Ok(tokio_rustls::TlsAcceptor::from(cfg)
-                .accept(raw)
-                .await
-                .inspect_err(|_| TLS_HANDSHAKE_FAILURES.inc())?),
+            Stream::Raw { raw } => Ok(tokio_rustls::TlsAcceptor::from(cfg).accept(raw).await?),
             Stream::Tls { .. } => Err(StreamUpgradeError::AlreadyTls),
         }
     }

s3_scrubber/src/checks.rs

@@ -11,7 +11,7 @@ use utils::id::TimelineId;
 use crate::cloud_admin_api::BranchData;
 use crate::metadata_stream::stream_listing;
 use crate::{download_object_with_retries, RootTarget, TenantShardTimelineId};
-use futures_util::StreamExt;
+use futures_util::{pin_mut, StreamExt};
 use pageserver::tenant::remote_timeline_client::parse_remote_index_path;
 use pageserver::tenant::storage_layer::LayerFileName;
 use pageserver::tenant::IndexPart;
@@ -285,7 +285,8 @@ pub(crate) async fn list_timeline_blobs(
     let mut index_parts: Vec<ObjectIdentifier> = Vec::new();
     let mut initdb_archive: bool = false;
 
-    let mut stream = std::pin::pin!(stream_listing(s3_client, &timeline_dir_target));
+    let stream = stream_listing(s3_client, &timeline_dir_target);
+    pin_mut!(stream);
     while let Some(obj) = stream.next().await {
         let obj = obj?;
         let key = obj.key();

s3_scrubber/src/garbage.rs

@@ -12,7 +12,7 @@ use aws_sdk_s3::{
     types::{Delete, ObjectIdentifier},
     Client,
 };
-use futures_util::TryStreamExt;
+use futures_util::{pin_mut, TryStreamExt};
 use pageserver_api::shard::TenantShardId;
 use serde::{Deserialize, Serialize};
 use tokio_stream::StreamExt;
@@ -199,12 +199,12 @@ async fn find_garbage_inner(
             }
         }
     });
-    let mut tenants_checked =
-        std::pin::pin!(tenants_checked.try_buffer_unordered(CONSOLE_CONCURRENCY));
+    let tenants_checked = tenants_checked.try_buffer_unordered(CONSOLE_CONCURRENCY);
 
     // Process the results of Tenant checks.  If a Tenant is garbage, it goes into
     // the `GarbageList`.  Else it goes into `active_tenants` for more detailed timeline
     // checks if they are enabled by the `depth` parameter.
+    pin_mut!(tenants_checked);
     let mut garbage = GarbageList::new(node_kind, bucket_config);
     let mut active_tenants: Vec<TenantShardId> = vec![];
     let mut counter = 0;
@@ -267,10 +267,10 @@ async fn find_garbage_inner(
                 .map(|r| (ttid, r))
         }
     });
-    let mut timelines_checked =
-        std::pin::pin!(timelines_checked.try_buffer_unordered(CONSOLE_CONCURRENCY));
+    let timelines_checked = timelines_checked.try_buffer_unordered(CONSOLE_CONCURRENCY);
 
     // Update the GarbageList with any timelines which appear not to exist.
+    pin_mut!(timelines_checked);
     while let Some(result) = timelines_checked.next().await {
         let (ttid, console_result) = result?;
         if garbage.maybe_append(GarbageEntity::Timeline(ttid), console_result) {
@@ -425,9 +425,9 @@ pub async fn purge_garbage(
             }
         }
     });
-    let mut get_objects_results =
-        std::pin::pin!(get_objects_results.try_buffer_unordered(S3_CONCURRENCY));
+    let get_objects_results = get_objects_results.try_buffer_unordered(S3_CONCURRENCY);
 
+    pin_mut!(get_objects_results);
     let mut objects_to_delete = Vec::new();
     while let Some(result) = get_objects_results.next().await {
         let mut object_list = result?;

s3_scrubber/src/scan_metadata.rs

@@ -7,7 +7,7 @@ use crate::checks::{
 use crate::metadata_stream::{stream_tenant_timelines, stream_tenants};
 use crate::{init_remote, BucketConfig, NodeKind, RootTarget, TenantShardTimelineId};
 use aws_sdk_s3::Client;
-use futures_util::{StreamExt, TryStreamExt};
+use futures_util::{pin_mut, StreamExt, TryStreamExt};
 use histogram::Histogram;
 use pageserver::tenant::remote_timeline_client::remote_layer_path;
 use pageserver::tenant::IndexPart;
@@ -226,7 +226,7 @@ pub async fn scan_metadata(
         Ok((ttid, data))
     }
     let timelines = timelines.map_ok(|ttid| report_on_timeline(&s3_client, &target, ttid));
-    let mut timelines = std::pin::pin!(timelines.try_buffered(CONCURRENCY));
+    let timelines = timelines.try_buffered(CONCURRENCY);
 
     // We must gather all the TenantShardTimelineId->S3TimelineBlobData for each tenant, because different
     // shards in the same tenant might refer to one anothers' keys if a shard split has happened.
@@ -309,6 +309,7 @@ pub async fn scan_metadata(
     // all results for the same tenant will be adjacent.  We accumulate these,
     // and then call `analyze_tenant` to flush, when we see the next tenant ID.
     let mut summary = MetadataSummary::new();
+    pin_mut!(timelines);
     while let Some(i) = timelines.next().await {
         let (ttid, data) = i?;
         summary.update_data(&data);

safekeeper/src/wal_service.rs

@@ -68,7 +68,7 @@ async fn handle_socket(
     // is not Unpin, and all pgbackend/framed/tokio dependencies require stream
     // to be Unpin. Which is reasonable, as indeed something like TimeoutReader
     // shouldn't be moved.
-    let socket = std::pin::pin!(socket);
+    tokio::pin!(socket);
 
     let traffic_metrics = TrafficMetrics::new();
     if let Some(current_az) = conf.availability_zone.as_deref() {

test_runner/performance/pageserver/pagebench/test_large_slru_basebackup.py

@@ -1,6 +1,5 @@
 import asyncio
 import json
-import os
 from pathlib import Path
 from typing import Any, Dict, Tuple
 
@@ -20,10 +19,6 @@ from performance.pageserver.util import (
 @pytest.mark.parametrize("n_tenants", [10])
 @pytest.mark.parametrize("get_vectored_impl", ["sequential", "vectored"])
 @pytest.mark.timeout(1000)
-@pytest.mark.skipif(
-    os.getenv("CI", "false") == "true",
-    reason="The test if flaky on CI: https://github.com/neondatabase/neon/issues/7006",
-)
 def test_basebackup_with_high_slru_count(
     neon_env_builder: NeonEnvBuilder,
     zenbenchmark: NeonBenchmarker,

test_runner/performance/test_branch_creation.py

@@ -1,5 +1,4 @@
 import random
-import re
 import statistics
 import threading
 import time
@@ -8,14 +7,11 @@ from contextlib import closing
 from typing import List
 
 import pytest
-from fixtures.benchmark_fixture import MetricReport, NeonBenchmarker
+from fixtures.benchmark_fixture import MetricReport
 from fixtures.compare_fixtures import NeonCompare
 from fixtures.log_helper import log
-from fixtures.neon_fixtures import NeonPageserver
 from fixtures.pageserver.utils import wait_for_last_record_lsn
 from fixtures.types import Lsn
-from fixtures.utils import wait_until
-from prometheus_client.samples import Sample
 
 
 def _record_branch_creation_durations(neon_compare: NeonCompare, durs: List[float]):
@@ -93,17 +89,11 @@ def test_branch_creation_heavy_write(neon_compare: NeonCompare, n_branches: int)
     _record_branch_creation_durations(neon_compare, branch_creation_durations)
 
 
-@pytest.mark.parametrize("n_branches", [500, 1024])
-@pytest.mark.parametrize("shape", ["one_ancestor", "random"])
-def test_branch_creation_many(neon_compare: NeonCompare, n_branches: int, shape: str):
-    """
-    Test measures the latency of branch creation when creating a lot of branches.
-    """
+@pytest.mark.parametrize("n_branches", [1024])
+# Test measures the latency of branch creation when creating a lot of branches.
+def test_branch_creation_many(neon_compare: NeonCompare, n_branches: int):
     env = neon_compare.env
 
-    # seed the prng so we will measure the same structure every time
-    rng = random.Random("2024-02-29")
-
     env.neon_cli.create_branch("b0")
 
     endpoint = env.endpoints.create_start("b0")
@@ -112,101 +102,15 @@ def test_branch_creation_many(neon_compare: NeonCompare, n_branches: int, shape:
     branch_creation_durations = []
 
     for i in range(n_branches):
-        if shape == "random":
-            parent = f"b{rng.randint(0, i)}"
-        elif shape == "one_ancestor":
-            parent = "b0"
-        else:
-            raise RuntimeError(f"unimplemented shape: {shape}")
-
+        # random a source branch
+        p = random.randint(0, i)
         timer = timeit.default_timer()
-        # each of these uploads to remote storage before completion
-        env.neon_cli.create_branch(f"b{i + 1}", parent)
+        env.neon_cli.create_branch("b{}".format(i + 1), "b{}".format(p))
         dur = timeit.default_timer() - timer
         branch_creation_durations.append(dur)
 
     _record_branch_creation_durations(neon_compare, branch_creation_durations)
 
-    endpoint.stop_and_destroy()
-
-    with neon_compare.record_duration("shutdown"):
-        # this sleeps 100ms between polls
-        env.pageserver.stop()
-
-    startup_line = "INFO version: git(-env)?:"
-
-    # find the first line of the log file so we can find the next start later
-    _, first_start = wait_until(5, 1, lambda: env.pageserver.assert_log_contains(startup_line))
-
-    # start without gc so we can time compaction with less noise; use shorter
-    # period for compaction so it starts earlier
-    env.pageserver.start(
-        overrides=(
-            "--pageserver-config-override=tenant_config={ compaction_period = '3s', gc_period = '0s' }",
-        ),
-        # this does print more than we want, but the number should be comparable between runs
-        extra_env_vars={
-            "RUST_LOG": f"[compaction_loop{{tenant_id={env.initial_tenant}}}]=debug,info"
-        },
-    )
-
-    _, second_start = wait_until(
-        5, 1, lambda: env.pageserver.assert_log_contains(startup_line, first_start)
-    )
-    env.pageserver.quiesce_tenants()
-
-    wait_and_record_startup_metrics(env.pageserver, neon_compare.zenbenchmark, "restart_after")
-
-    # wait for compaction to complete, which most likely has already done so multiple times
-    msg, _ = wait_until(
-        30,
-        1,
-        lambda: env.pageserver.assert_log_contains(
-            f".*tenant_id={env.initial_tenant}.*: compaction iteration complete.*", second_start
-        ),
-    )
-    needle = re.search(" elapsed_ms=([0-9]+)", msg)
-    assert needle is not None, "failed to find the elapsed time"
-    duration = int(needle.group(1)) / 1000.0
-    neon_compare.zenbenchmark.record("compaction", duration, "s", MetricReport.LOWER_IS_BETTER)
-
-
-def wait_and_record_startup_metrics(
-    pageserver: NeonPageserver, target: NeonBenchmarker, prefix: str
-):
-    """
-    Waits until all startup metrics have non-zero values on the pageserver, then records them on the target
-    """
-
-    client = pageserver.http_client()
-
-    expected_labels = set(
-        [
-            "background_jobs_can_start",
-            "complete",
-            "initial",
-            "initial_tenant_load",
-            "initial_tenant_load_remote",
-        ]
-    )
-
-    def metrics_are_filled() -> List[Sample]:
-        m = client.get_metrics()
-        samples = m.query_all("pageserver_startup_duration_seconds")
-        # we should not have duplicate labels
-        matching = [
-            x for x in samples if x.labels.get("phase") in expected_labels and x.value > 0.0
-        ]
-        assert len(matching) == len(expected_labels)
-        return matching
-
-    samples = wait_until(10, 1, metrics_are_filled)
-
-    for sample in samples:
-        phase = sample.labels["phase"]
-        name = f"{prefix}.{phase}"
-        target.record(name, sample.value, "s", MetricReport.LOWER_IS_BETTER)
-
 
 # Test measures the branch creation time when branching from a timeline with a lot of relations.
 #

test_runner/performance/test_layer_map.py

@@ -17,10 +17,10 @@ def test_layer_map(neon_env_builder: NeonEnvBuilder, zenbenchmark):
     tenant, _ = env.neon_cli.create_tenant(
         conf={
             "gc_period": "0s",
-            "checkpoint_distance": "16384",
+            "checkpoint_distance": "8192",
             "compaction_period": "1 s",
             "compaction_threshold": "1",
-            "compaction_target_size": "16384",
+            "compaction_target_size": "8192",
         }
     )
 

test_runner/pg_clients/csharp/npgsql/Dockerfile

@@ -1,4 +1,4 @@
-FROM mcr.microsoft.com/dotnet/sdk:8.0 AS build
+FROM mcr.microsoft.com/dotnet/sdk:7.0 AS build
 WORKDIR /source
 
 COPY *.csproj .
@@ -7,7 +7,7 @@ RUN dotnet restore
 COPY . .
 RUN dotnet publish -c release -o /app --no-restore
 
-FROM mcr.microsoft.com/dotnet/runtime:8.0
+FROM mcr.microsoft.com/dotnet/runtime:7.0
 WORKDIR /app
 COPY --from=build /app .
 

test_runner/pg_clients/csharp/npgsql/csharp-npgsql.csproj

@@ -2,13 +2,13 @@
 
   <PropertyGroup>
     <OutputType>Exe</OutputType>
-    <TargetFramework>net8.0</TargetFramework>
+    <TargetFramework>net7.0</TargetFramework>
     <ImplicitUsings>enable</ImplicitUsings>
     <Nullable>enable</Nullable>
   </PropertyGroup>
 
   <ItemGroup>
-    <PackageReference Include="Npgsql" Version="8.0.2" />
+    <PackageReference Include="Npgsql" Version="7.0.4" />
   </ItemGroup>
 
 </Project>

test_runner/pg_clients/java/jdbc/Dockerfile

@@ -1,10 +1,10 @@
-FROM openjdk:21
+FROM openjdk:20
 WORKDIR /source
 
 COPY . .
 
 WORKDIR /app
-RUN curl --output postgresql.jar https://jdbc.postgresql.org/download/postgresql-42.7.2.jar && \
+RUN curl --output postgresql.jar https://jdbc.postgresql.org/download/postgresql-42.6.0.jar && \
     javac -d /app /source/Example.java
 
 CMD ["java", "-cp", "/app/postgresql.jar:.", "Example"]

test_runner/pg_clients/python/asyncpg/Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.12
+FROM python:3.11
 WORKDIR /source
 
 COPY . .

test_runner/pg_clients/python/asyncpg/requirements.txt

@@ -1 +1 @@
-asyncpg==0.29.0
+asyncpg==0.27.0

test_runner/pg_clients/python/pg8000/Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.12
+FROM python:3.11
 WORKDIR /source
 
 COPY . .

test_runner/pg_clients/python/pg8000/requirements.txt

@@ -1,2 +1,2 @@
-pg8000==1.30.5
+pg8000==1.29.8
 scramp>=1.4.3

test_runner/pg_clients/rust/tokio-postgres/Cargo.lock

@@ -19,9 +19,9 @@ checksum = "f26201604c87b1e01bd3d98f8d5d9a8fcbb815e8cedb41ffccbeb4bf593a35fe"
 
 [[package]]
 name = "async-trait"
-version = "0.1.77"
+version = "0.1.74"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c980ee35e870bd1a4d2c8294d4c04d0499e67bca1e4b5cefcc693c2fa00caea9"
+checksum = "a66537f1bb974b254c98ed142ff995236e81b9d0fe4db0575f46612cb15eb0f9"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -51,9 +51,9 @@ dependencies = [
 
 [[package]]
 name = "base64"
-version = "0.21.7"
+version = "0.21.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9d297deb1925b89f2ccc13d7635fa0714f12c87adce1c75356b39ca9b7178567"
+checksum = "9ba43ea6f343b788c8764558649e08df62f86c6ef251fdaeb1ffd010a9ae50a2"
 
 [[package]]
 name = "bitflags"
@@ -63,9 +63,9 @@ checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
 
 [[package]]
 name = "bitflags"
-version = "2.4.2"
+version = "2.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ed570934406eb16438a4e976b1b4500774099c13b8cb96eec99f620f05090ddf"
+checksum = "327762f6e5a765692301e5bb513e0d9fef63be86bbc14528052b1cd3e6f03e07"
 
 [[package]]
 name = "block-buffer"
@@ -78,9 +78,9 @@ dependencies = [
 
 [[package]]
 name = "bumpalo"
-version = "3.15.3"
+version = "3.14.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8ea184aa71bb362a1157c896979544cc23974e08fd265f29ea96b59f0b4a555b"
+checksum = "7f30e7476521f6f8af1a1c4c0b8cc94f0bee37d91763d0ca2665f299b6cd8aec"
 
 [[package]]
 name = "byteorder"
@@ -96,9 +96,12 @@ checksum = "a2bd12c1caf447e69cd4528f47f94d203fd2582878ecb9e9465484c4148a8223"
 
 [[package]]
 name = "cc"
-version = "1.0.89"
+version = "1.0.83"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a0ba8f7aaa012f30d5b2861462f6708eccd49c3c39863fe083a308035f63d723"
+checksum = "f1174fb0b6ec23863f8b971027804a42614e347eafb0a95bf0b12cdae21fc4d0"
+dependencies = [
+ "libc",
+]
 
 [[package]]
 name = "cfg-if"
@@ -108,9 +111,9 @@ checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
 
 [[package]]
 name = "core-foundation"
-version = "0.9.4"
+version = "0.9.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "91e195e091a93c46f7102ec7818a2aa394e1e1771c3ab4825963fa03e45afb8f"
+checksum = "194a7a9e6de53fa55116934067c844d9d749312f75c6f6d0980e8c252f8c2146"
 dependencies = [
  "core-foundation-sys",
  "libc",
@@ -118,15 +121,15 @@ dependencies = [
 
 [[package]]
 name = "core-foundation-sys"
-version = "0.8.6"
+version = "0.8.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "06ea2b9bc92be3c2baa9334a323ebca2d6f074ff852cd1d7b11064035cd3868f"
+checksum = "e496a50fda8aacccc86d7529e2c1e0892dbd0f898a6b5645b5561b89c3210efa"
 
 [[package]]
 name = "cpufeatures"
-version = "0.2.12"
+version = "0.2.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "53fe5e26ff1b7aef8bca9c6080520cfb8d9333c7568e1829cef191a9723e5504"
+checksum = "a17b76ff3a4162b0b27f354a0c87015ddad39d35f9c0c36607a3bdd175dde1f1"
 dependencies = [
  "libc",
 ]
@@ -154,12 +157,12 @@ dependencies = [
 
 [[package]]
 name = "errno"
-version = "0.3.8"
+version = "0.3.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a258e46cdc063eb8519c00b9fc845fc47bcfca4130e2f08e88665ceda8474245"
+checksum = "ac3e13f66a2f95e32a39eaa81f6b95d42878ca0e1db0c7543723dfe12557e860"
 dependencies = [
  "libc",
- "windows-sys 0.52.0",
+ "windows-sys",
 ]
 
 [[package]]
@@ -197,9 +200,9 @@ checksum = "00b0228411908ca8685dba7fc2cdd70ec9990a6e753e89b6ac91a84c40fbaf4b"
 
 [[package]]
 name = "futures"
-version = "0.3.30"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "645c6916888f6cb6350d2550b80fb63e734897a8498abe35cfb732b6487804b0"
+checksum = "23342abe12aba583913b2e62f22225ff9c950774065e4bfb61a19cd9770fec40"
 dependencies = [
  "futures-channel",
  "futures-core",
@@ -212,9 +215,9 @@ dependencies = [
 
 [[package]]
 name = "futures-channel"
-version = "0.3.30"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "eac8f7d7865dcb88bd4373ab671c8cf4508703796caa2b1985a9ca867b3fcb78"
+checksum = "955518d47e09b25bbebc7a18df10b81f0c766eaf4c4f1cccef2fca5f2a4fb5f2"
 dependencies = [
  "futures-core",
  "futures-sink",
@@ -222,15 +225,15 @@ dependencies = [
 
 [[package]]
 name = "futures-core"
-version = "0.3.30"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dfc6580bb841c5a68e9ef15c77ccc837b40a7504914d52e47b8b0e9bbda25a1d"
+checksum = "4bca583b7e26f571124fe5b7561d49cb2868d79116cfa0eefce955557c6fee8c"
 
 [[package]]
 name = "futures-executor"
-version = "0.3.30"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a576fc72ae164fca6b9db127eaa9a9dda0d61316034f33a0a0d4eda41f02b01d"
+checksum = "ccecee823288125bd88b4d7f565c9e58e41858e47ab72e8ea2d64e93624386e0"
 dependencies = [
  "futures-core",
  "futures-task",
@@ -239,15 +242,15 @@ dependencies = [
 
 [[package]]
 name = "futures-io"
-version = "0.3.30"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a44623e20b9681a318efdd71c299b6b222ed6f231972bfe2f224ebad6311f0c1"
+checksum = "4fff74096e71ed47f8e023204cfd0aa1289cd54ae5430a9523be060cdb849964"
 
 [[package]]
 name = "futures-macro"
-version = "0.3.30"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "87750cf4b7a4c0625b1529e4c543c2182106e4dedc60a2a6455e00d212c489ac"
+checksum = "89ca545a94061b6365f2c7355b4b32bd20df3ff95f02da9329b34ccc3bd6ee72"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -256,21 +259,21 @@ dependencies = [
 
 [[package]]
 name = "futures-sink"
-version = "0.3.30"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9fb8e00e87438d937621c1c6269e53f536c14d3fbd6a042bb24879e57d474fb5"
+checksum = "f43be4fe21a13b9781a69afa4985b0f6ee0e1afab2c6f454a8cf30e2b2237b6e"
 
 [[package]]
 name = "futures-task"
-version = "0.3.30"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "38d84fa142264698cdce1a9f9172cf383a0c82de1bddcf3092901442c4097004"
+checksum = "76d3d132be6c0e6aa1534069c705a74a5997a356c0dc2f86a47765e5617c5b65"
 
 [[package]]
 name = "futures-util"
-version = "0.3.30"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3d6401deb83407ab3da39eba7e33987a73c3df0c82b4bb5813ee871c19c41d48"
+checksum = "26b01e40b772d54cf6c6d721c1d1abd0647a0106a12ecaa1c186273392a69533"
 dependencies = [
  "futures-channel",
  "futures-core",
@@ -296,9 +299,9 @@ dependencies = [
 
 [[package]]
 name = "getrandom"
-version = "0.2.12"
+version = "0.2.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "190092ea657667030ac6a35e305e62fc4dd69fd98ac98631e5d3a2b1575a12b5"
+checksum = "be4136b2a15dd319360be1c07d9933517ccf0be8f16bf62a3bee4f0d618df427"
 dependencies = [
  "cfg-if",
  "libc",
@@ -307,9 +310,9 @@ dependencies = [
 
 [[package]]
 name = "gimli"
-version = "0.28.1"
+version = "0.28.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4271d37baee1b8c7e4b708028c57d816cf9d2434acb33a549475f78c181f6253"
+checksum = "6fb8d784f27acf97159b40fc4db5ecd8aa23b9ad5ef69cdd136d3bc80665f0c0"
 
 [[package]]
 name = "hmac"
@@ -322,9 +325,9 @@ dependencies = [
 
 [[package]]
 name = "js-sys"
-version = "0.3.69"
+version = "0.3.64"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "29c15563dc2726973df627357ce0c9ddddbea194836909d655df6a75d2cf296d"
+checksum = "c5f195fe497f702db0f318b07fdd68edb16955aed830df8363d837542f8f935a"
 dependencies = [
  "wasm-bindgen",
 ]
@@ -337,15 +340,15 @@ checksum = "e2abad23fbc42b3700f2f279844dc832adb2b2eb069b2df918f455c4e18cc646"
 
 [[package]]
 name = "libc"
-version = "0.2.153"
+version = "0.2.149"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9c198f91728a82281a64e1f4f9eeb25d82cb32a5de251c6bd1b5154d63a8e7bd"
+checksum = "a08173bc88b7955d1b3145aa561539096c421ac8debde8cbc3612ec635fee29b"
 
 [[package]]
 name = "linux-raw-sys"
-version = "0.4.13"
+version = "0.4.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "01cda141df6706de531b6c46c3a33ecca755538219bd484262fa09410c13539c"
+checksum = "da2479e8c062e40bf0066ffa0bc823de0a9368974af99c9f6df941d2c231e03f"
 
 [[package]]
 name = "lock_api"
@@ -359,9 +362,9 @@ dependencies = [
 
 [[package]]
 name = "log"
-version = "0.4.21"
+version = "0.4.20"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "90ed8c1e510134f979dbc4f070f87d4313098b704861a105fe34231c70a3901c"
+checksum = "b5e6163cb8c49088c2c36f57875e58ccd8c87c7427f7fbd50ea6710b2f3f2e8f"
 
 [[package]]
 name = "md-5"
@@ -375,28 +378,28 @@ dependencies = [
 
 [[package]]
 name = "memchr"
-version = "2.7.1"
+version = "2.6.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "523dc4f511e55ab87b694dc30d0f820d60906ef06413f93d4d7a1385599cc149"
+checksum = "f665ee40bc4a3c5590afb1e9677db74a508659dfd71e126420da8274909a0167"
 
 [[package]]
 name = "miniz_oxide"
-version = "0.7.2"
+version = "0.7.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9d811f3e15f28568be3407c8e7fdb6514c1cda3cb30683f15b6a1a1dc4ea14a7"
+checksum = "e7810e0be55b428ada41041c41f32c9f1a42817901b4ccf45fa3d4b6561e74c7"
 dependencies = [
  "adler",
 ]
 
 [[package]]
 name = "mio"
-version = "0.8.11"
+version = "0.8.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a4a650543ca06a924e8b371db273b2756685faae30f8487da1b56505a8f78b0c"
+checksum = "927a765cd3fc26206e66b296465fa9d3e5ab003e651c1b3c060e7956d96b19d2"
 dependencies = [
  "libc",
  "wasi",
- "windows-sys 0.48.0",
+ "windows-sys",
 ]
 
 [[package]]
@@ -419,26 +422,26 @@ dependencies = [
 
 [[package]]
 name = "object"
-version = "0.32.2"
+version = "0.32.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a6a622008b6e321afc04970976f62ee297fdbaa6f95318ca343e3eebb9648441"
+checksum = "9cf5f9dd3933bd50a9e1f149ec995f39ae2c496d31fd772c1fd45ebc27e902b0"
 dependencies = [
  "memchr",
 ]
 
 [[package]]
 name = "once_cell"
-version = "1.19.0"
+version = "1.18.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3fdb12b2476b595f9358c5161aa467c2438859caa136dec86c26fdd2efe17b92"
+checksum = "dd8b5dd2ae5ed71462c540258bedcb51965123ad7e7ccf4b9a8cafaa4a63576d"
 
 [[package]]
 name = "openssl"
-version = "0.10.64"
+version = "0.10.60"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "95a0481286a310808298130d22dd1fef0fa571e05a8f44ec801801e84b216b1f"
+checksum = "79a4c6c3a2b158f7f8f2a2fc5a969fa3a068df6fc9dbb4a43845436e3af7c800"
 dependencies = [
- "bitflags 2.4.2",
+ "bitflags 2.4.1",
  "cfg-if",
  "foreign-types",
  "libc",
@@ -466,9 +469,9 @@ checksum = "ff011a302c396a5197692431fc1948019154afc178baf7d8e37367442a4601cf"
 
 [[package]]
 name = "openssl-sys"
-version = "0.9.101"
+version = "0.9.96"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dda2b0f344e78efc2facf7d195d098df0dd72151b26ab98da807afc26c198dff"
+checksum = "3812c071ba60da8b5677cc12bcb1d42989a65553772897a7e0355545a819838f"
 dependencies = [
  "cc",
  "libc",
@@ -494,16 +497,16 @@ checksum = "4c42a9226546d68acdd9c0a280d17ce19bfe27a46bf68784e4066115788d008e"
 dependencies = [
  "cfg-if",
  "libc",
- "redox_syscall",
+ "redox_syscall 0.4.1",
  "smallvec",
- "windows-targets 0.48.5",
+ "windows-targets",
 ]
 
 [[package]]
 name = "percent-encoding"
-version = "2.3.1"
+version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e3148f5046208a5d56bcfc03053e3ca6334e51da8dfb19b6cdc8b306fae3283e"
+checksum = "9b2a4787296e9989611394c33f193f676704af1686e70b8f8033ab5ba9a35a94"
 
 [[package]]
 name = "phf"
@@ -537,9 +540,9 @@ checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"
 
 [[package]]
 name = "pkg-config"
-version = "0.3.30"
+version = "0.3.27"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d231b230927b5e4ad203db57bbcbee2802f6bce620b1e4a9024a07d94e2907ec"
+checksum = "26072860ba924cbfa98ea39c8c19b4dd6a4a25423dbdf219c1eca91aa0cf6964"
 
 [[package]]
 name = "postgres-native-tls"
@@ -591,18 +594,18 @@ checksum = "5b40af805b3121feab8a3c29f04d8ad262fa8e0561883e7653e024ae4479e6de"
 
 [[package]]
 name = "proc-macro2"
-version = "1.0.78"
+version = "1.0.69"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e2422ad645d89c99f8f3e6b88a9fdeca7fabeac836b1002371c4367c8f984aae"
+checksum = "134c189feb4956b20f6f547d2cf727d4c0fe06722b20a0eec87ed445a97f92da"
 dependencies = [
  "unicode-ident",
 ]
 
 [[package]]
 name = "quote"
-version = "1.0.35"
+version = "1.0.33"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "291ec9ab5efd934aaf503a6466c5d5251535d108ee747472c3977cc5acc868ef"
+checksum = "5267fca4496028628a95160fc423a33e8b2e6af8a5302579e322e4b520293cae"
 dependencies = [
  "proc-macro2",
 ]
@@ -637,6 +640,15 @@ dependencies = [
  "getrandom",
 ]
 
+[[package]]
+name = "redox_syscall"
+version = "0.3.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "567664f262709473930a4bf9e51bf2ebf3348f2e748ccc50dea20646858f8f29"
+dependencies = [
+ "bitflags 1.3.2",
+]
+
 [[package]]
 name = "redox_syscall"
 version = "0.4.1"
@@ -664,24 +676,24 @@ checksum = "d626bb9dae77e28219937af045c257c28bfd3f69333c512553507f5f9798cb76"
 
 [[package]]
 name = "rustix"
-version = "0.38.31"
+version = "0.38.19"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ea3e1a662af26cd7a3ba09c0297a31af215563ecf42817c98df621387f4e949"
+checksum = "745ecfa778e66b2b63c88a61cb36e0eea109e803b0b86bf9879fbc77c70e86ed"
 dependencies = [
- "bitflags 2.4.2",
+ "bitflags 2.4.1",
  "errno",
  "libc",
  "linux-raw-sys",
- "windows-sys 0.52.0",
+ "windows-sys",
 ]
 
 [[package]]
 name = "schannel"
-version = "0.1.23"
+version = "0.1.22"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fbc91545643bcf3a0bbb6569265615222618bdf33ce4ffbbd13c4bbd4c093534"
+checksum = "0c3733bf4cf7ea0880754e19cb5a462007c4a8c1914bff372ccc95b464f1df88"
 dependencies = [
- "windows-sys 0.52.0",
+ "windows-sys",
 ]
 
 [[package]]
@@ -741,18 +753,18 @@ dependencies = [
 
 [[package]]
 name = "smallvec"
-version = "1.13.1"
+version = "1.11.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e6ecd384b10a64542d77071bd64bd7b231f4ed5940fba55e98c3de13824cf3d7"
+checksum = "942b4a808e05215192e39f4ab80813e599068285906cc91aa64f923db842bd5a"
 
 [[package]]
 name = "socket2"
-version = "0.5.6"
+version = "0.5.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "05ffd9c0a93b7543e062e759284fcf5f5e3b098501104bfbdde4d404db792871"
+checksum = "4031e820eb552adee9295814c0ced9e5cf38ddf1e8b7d566d6de8e2538ea989e"
 dependencies = [
  "libc",
- "windows-sys 0.52.0",
+ "windows-sys",
 ]
 
 [[package]]
@@ -774,9 +786,9 @@ checksum = "81cdd64d312baedb58e21336b31bc043b77e01cc99033ce76ef539f78e965ebc"
 
 [[package]]
 name = "syn"
-version = "2.0.52"
+version = "2.0.38"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b699d15b36d1f02c3e7c69f8ffef53de37aefae075d8488d4ba1a7788d574a07"
+checksum = "e96b79aaa137db8f61e26363a0c9b47d8b4ec75da28b7d1d614c2303e232408b"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -785,14 +797,15 @@ dependencies = [
 
 [[package]]
 name = "tempfile"
-version = "3.10.1"
+version = "3.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "85b77fafb263dd9d05cbeac119526425676db3784113aa9295c88498cbf8bff1"
+checksum = "cb94d2f3cc536af71caac6b6fcebf65860b347e7ce0cc9ebe8f70d3e521054ef"
 dependencies = [
  "cfg-if",
  "fastrand",
+ "redox_syscall 0.3.5",
  "rustix",
- "windows-sys 0.52.0",
+ "windows-sys",
 ]
 
 [[package]]
@@ -812,9 +825,9 @@ checksum = "1f3ccbac311fea05f86f61904b462b55fb3df8837a366dfc601a0161d0532f20"
 
 [[package]]
 name = "tokio"
-version = "1.36.0"
+version = "1.33.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "61285f6515fa018fb2d1e46eb21223fff441ee8db5d0f1435e8ab4f5cdb80931"
+checksum = "4f38200e3ef7995e5ef13baec2f432a6da0aa9ac495b2c0e8f3b7eec2c92d653"
 dependencies = [
  "backtrace",
  "bytes",
@@ -823,14 +836,14 @@ dependencies = [
  "pin-project-lite",
  "socket2",
  "tokio-macros",
- "windows-sys 0.48.0",
+ "windows-sys",
 ]
 
 [[package]]
 name = "tokio-macros"
-version = "2.2.0"
+version = "2.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5b8a1e28f2deaa14e508979454cb3a223b10b938b45af148bc0986de36f1923b"
+checksum = "630bdcf245f78637c13ec01ffae6187cca34625e8c63150d424b59e55af2675e"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -875,9 +888,9 @@ dependencies = [
 
 [[package]]
 name = "tokio-util"
-version = "0.7.10"
+version = "0.7.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5419f34732d9eb6ee4c3578b7989078579b7f039cbbb9ca2c4da015749371e15"
+checksum = "1d68074620f57a0b21594d9735eb2e98ab38b17f80d3fcb189fca266771ca60d"
 dependencies = [
  "bytes",
  "futures-core",
@@ -914,9 +927,9 @@ checksum = "42ff0bf0c66b8238c6f3b578df37d0b7848e55df8577b3f74f92a69acceeb825"
 
 [[package]]
 name = "unicode-bidi"
-version = "0.3.15"
+version = "0.3.13"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "08f95100a766bf4f8f28f90d77e0a5461bbdb219042e7679bebe79004fed8d75"
+checksum = "92888ba5573ff080736b3648696b70cafad7d250551175acbaa4e0385b3e1460"
 
 [[package]]
 name = "unicode-ident"
@@ -926,9 +939,9 @@ checksum = "3354b9ac3fae1ff6755cb6db53683adb661634f67557942dea4facebec0fee4b"
 
 [[package]]
 name = "unicode-normalization"
-version = "0.1.23"
+version = "0.1.22"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a56d1686db2308d901306f92a263857ef59ea39678a5458e7cb17f01415101f5"
+checksum = "5c5713f0fc4b5db668a2ac63cdb7bb4469d8c9fed047b1d0292cc7b0ce2ba921"
 dependencies = [
  "tinyvec",
 ]
@@ -951,17 +964,11 @@ version = "0.11.0+wasi-snapshot-preview1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9c8d87e72b64a3b4db28d11ce29237c246188f4f51057d65a7eab63b7987e423"
 
-[[package]]
-name = "wasite"
-version = "0.1.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b8dad83b4f25e74f184f64c43b150b91efe7647395b42289f38e50566d82855b"
-
 [[package]]
 name = "wasm-bindgen"
-version = "0.2.92"
+version = "0.2.87"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4be2531df63900aeb2bca0daaaddec08491ee64ceecbee5076636a3b026795a8"
+checksum = "7706a72ab36d8cb1f80ffbf0e071533974a60d0a308d01a5d0375bf60499a342"
 dependencies = [
  "cfg-if",
  "wasm-bindgen-macro",
@@ -969,9 +976,9 @@ dependencies = [
 
 [[package]]
 name = "wasm-bindgen-backend"
-version = "0.2.92"
+version = "0.2.87"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "614d787b966d3989fa7bb98a654e369c762374fd3213d212cfc0251257e747da"
+checksum = "5ef2b6d3c510e9625e5fe6f509ab07d66a760f0885d858736483c32ed7809abd"
 dependencies = [
  "bumpalo",
  "log",
@@ -984,9 +991,9 @@ dependencies = [
 
 [[package]]
 name = "wasm-bindgen-macro"
-version = "0.2.92"
+version = "0.2.87"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a1f8823de937b71b9460c0c34e25f3da88250760bec0ebac694b49997550d726"
+checksum = "dee495e55982a3bd48105a7b947fd2a9b4a8ae3010041b9e0faab3f9cd028f1d"
 dependencies = [
  "quote",
  "wasm-bindgen-macro-support",
@@ -994,9 +1001,9 @@ dependencies = [
 
 [[package]]
 name = "wasm-bindgen-macro-support"
-version = "0.2.92"
+version = "0.2.87"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e94f17b526d0a461a191c78ea52bbce64071ed5c04c9ffe424dcb38f74171bb7"
+checksum = "54681b18a46765f095758388f2d0cf16eb8d4169b639ab575a8f5693af210c7b"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -1007,15 +1014,15 @@ dependencies = [
 
 [[package]]
 name = "wasm-bindgen-shared"
-version = "0.2.92"
+version = "0.2.87"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "af190c94f2773fdb3729c55b007a722abb5384da03bc0986df4c289bf5567e96"
+checksum = "ca6ad05a4870b2bf5fe995117d3728437bd27d7cd5f06f13c17443ef369775a1"
 
 [[package]]
 name = "web-sys"
-version = "0.3.69"
+version = "0.3.64"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "77afa9a11836342370f4817622a2f0f418b134426d91a82dfb48f532d2ec13ef"
+checksum = "9b85cbef8c220a6abc02aefd892dfc0fc23afb1c6a426316ec33253a3877249b"
 dependencies = [
  "js-sys",
  "wasm-bindgen",
@@ -1023,12 +1030,11 @@ dependencies = [
 
 [[package]]
 name = "whoami"
-version = "1.5.0"
+version = "1.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0fec781d48b41f8163426ed18e8fc2864c12937df9ce54c88ede7bd47270893e"
+checksum = "22fc3756b8a9133049b26c7f61ab35416c130e8c09b660f5b3958b446f52cc50"
 dependencies = [
- "redox_syscall",
- "wasite",
+ "wasm-bindgen",
  "web-sys",
 ]
 
@@ -1038,16 +1044,7 @@ version = "0.48.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "677d2418bec65e3338edb076e806bc1ec15693c5d0104683f2efe857f61056a9"
 dependencies = [
- "windows-targets 0.48.5",
-]
-
-[[package]]
-name = "windows-sys"
-version = "0.52.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "282be5f36a8ce781fad8c8ae18fa3f9beff57ec1b52cb3de0789201425d9a33d"
-dependencies = [
- "windows-targets 0.52.4",
+ "windows-targets",
 ]
 
 [[package]]
@@ -1056,28 +1053,13 @@ version = "0.48.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9a2fa6e2155d7247be68c096456083145c183cbbbc2764150dda45a87197940c"
 dependencies = [
- "windows_aarch64_gnullvm 0.48.5",
- "windows_aarch64_msvc 0.48.5",
- "windows_i686_gnu 0.48.5",
- "windows_i686_msvc 0.48.5",
- "windows_x86_64_gnu 0.48.5",
- "windows_x86_64_gnullvm 0.48.5",
- "windows_x86_64_msvc 0.48.5",
-]
-
-[[package]]
-name = "windows-targets"
-version = "0.52.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7dd37b7e5ab9018759f893a1952c9420d060016fc19a472b4bb20d1bdd694d1b"
-dependencies = [
- "windows_aarch64_gnullvm 0.52.4",
- "windows_aarch64_msvc 0.52.4",
- "windows_i686_gnu 0.52.4",
- "windows_i686_msvc 0.52.4",
- "windows_x86_64_gnu 0.52.4",
- "windows_x86_64_gnullvm 0.52.4",
- "windows_x86_64_msvc 0.52.4",
+ "windows_aarch64_gnullvm",
+ "windows_aarch64_msvc",
+ "windows_i686_gnu",
+ "windows_i686_msvc",
+ "windows_x86_64_gnu",
+ "windows_x86_64_gnullvm",
+ "windows_x86_64_msvc",
 ]
 
 [[package]]
@@ -1086,80 +1068,38 @@ version = "0.48.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2b38e32f0abccf9987a4e3079dfb67dcd799fb61361e53e2882c3cbaf0d905d8"
 
-[[package]]
-name = "windows_aarch64_gnullvm"
-version = "0.52.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bcf46cf4c365c6f2d1cc93ce535f2c8b244591df96ceee75d8e83deb70a9cac9"
-
 [[package]]
 name = "windows_aarch64_msvc"
 version = "0.48.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "dc35310971f3b2dbbf3f0690a219f40e2d9afcf64f9ab7cc1be722937c26b4bc"
 
-[[package]]
-name = "windows_aarch64_msvc"
-version = "0.52.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "da9f259dd3bcf6990b55bffd094c4f7235817ba4ceebde8e6d11cd0c5633b675"
-
 [[package]]
 name = "windows_i686_gnu"
 version = "0.48.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a75915e7def60c94dcef72200b9a8e58e5091744960da64ec734a6c6e9b3743e"
 
-[[package]]
-name = "windows_i686_gnu"
-version = "0.52.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b474d8268f99e0995f25b9f095bc7434632601028cf86590aea5c8a5cb7801d3"
-
 [[package]]
 name = "windows_i686_msvc"
 version = "0.48.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8f55c233f70c4b27f66c523580f78f1004e8b5a8b659e05a4eb49d4166cca406"
 
-[[package]]
-name = "windows_i686_msvc"
-version = "0.52.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1515e9a29e5bed743cb4415a9ecf5dfca648ce85ee42e15873c3cd8610ff8e02"
-
 [[package]]
 name = "windows_x86_64_gnu"
 version = "0.48.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "53d40abd2583d23e4718fddf1ebec84dbff8381c07cae67ff7768bbf19c6718e"
 
-[[package]]
-name = "windows_x86_64_gnu"
-version = "0.52.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5eee091590e89cc02ad514ffe3ead9eb6b660aedca2183455434b93546371a03"
-
 [[package]]
 name = "windows_x86_64_gnullvm"
 version = "0.48.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0b7b52767868a23d5bab768e390dc5f5c55825b6d30b86c844ff2dc7414044cc"
 
-[[package]]
-name = "windows_x86_64_gnullvm"
-version = "0.52.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "77ca79f2451b49fa9e2af39f0747fe999fcda4f5e241b2898624dca97a1f2177"
-
 [[package]]
 name = "windows_x86_64_msvc"
 version = "0.48.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ed94fce61571a4006852b7389a063ab983c02eb1bb37b47f8272ce92d06d9538"
-
-[[package]]
-name = "windows_x86_64_msvc"
-version = "0.52.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "32b752e52a2da0ddfbdbcc6fceadfeede4c939ed16d13e648833a61dfb611ed8"

test_runner/pg_clients/rust/tokio-postgres/Cargo.toml

@@ -9,7 +9,7 @@ publish = false
 [dependencies]
 native-tls = "0.2.11"
 postgres-native-tls = "0.5.0"
-tokio = { version = "1.36", features=["rt", "macros"] }
+tokio = { version = "1.33", features=["rt", "macros"] }
 tokio-postgres = "0.7.10"