improve instructions

Cargo.lock

@@ -570,9 +570,9 @@ dependencies = [
 
 [[package]]
 name = "aws-smithy-types"
-version = "0.56.0"
+version = "0.56.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "eed0a94eefd845a2a78677f1b72f02fa75802d38f7f59be675add140279aa8bf"
+checksum = "d90dbc8da2f6be461fa3c1906b20af8f79d14968fe47f2b7d29d086f62a51728"
 dependencies = [
  "base64-simd",
  "itoa",
@@ -2420,6 +2420,16 @@ dependencies = [
  "winapi",
 ]
 
+[[package]]
+name = "nu-ansi-term"
+version = "0.46.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "77a8165726e8236064dbb45459242600304b42a5ea24ee2948e18e023bf7ba84"
+dependencies = [
+ "overload",
+ "winapi",
+]
+
 [[package]]
 name = "num-bigint"
 version = "0.4.3"
@@ -2652,6 +2662,12 @@ version = "0.5.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4030760ffd992bef45b0ae3f10ce1aba99e33464c90d14dd7c039884963ddc7a"
 
+[[package]]
+name = "overload"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b15813163c1d831bf4a13c3610c05c0d03b39feb07f7e09fa234dac9b15aaf39"
+
 [[package]]
 name = "pagectl"
 version = "0.1.0"
@@ -3748,9 +3764,11 @@ dependencies = [
  "aws-config",
  "aws-sdk-s3",
  "aws-smithy-http",
+ "aws-smithy-types",
  "aws-types",
  "bincode",
  "bytes",
+ "camino",
  "chrono",
  "clap",
  "crc32c",
@@ -3758,9 +3776,11 @@ dependencies = [
  "futures-util",
  "hex",
  "histogram",
+ "humantime",
  "itertools",
  "pageserver",
  "rand",
+ "remote_storage",
  "reqwest",
  "serde",
  "serde_json",
@@ -4942,6 +4962,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "30a651bc37f915e81f087d86e62a18eec5f79550c7faff886f7090b4ea757c77"
 dependencies = [
  "matchers",
+ "nu-ansi-term",
  "once_cell",
  "regex",
  "serde",

pageserver/src/deletion_queue.rs

@@ -153,7 +153,7 @@ impl FlushOp {
 
 #[derive(Clone, Debug)]
 pub struct DeletionQueueClient {
-    tx: tokio::sync::mpsc::UnboundedSender<ListWriterQueueMessage>,
+    tx: tokio::sync::mpsc::Sender<ListWriterQueueMessage>,
     executor_tx: tokio::sync::mpsc::Sender<DeleterMessage>,
 
     lsn_table: Arc<std::sync::RwLock<VisibleLsnUpdates>>,
@@ -416,7 +416,7 @@ pub enum DeletionQueueError {
 impl DeletionQueueClient {
     pub(crate) fn broken() -> Self {
         // Channels whose receivers are immediately dropped.
-        let (tx, _rx) = tokio::sync::mpsc::unbounded_channel();
+        let (tx, _rx) = tokio::sync::mpsc::channel(1);
         let (executor_tx, _executor_rx) = tokio::sync::mpsc::channel(1);
         Self {
             tx,
@@ -428,12 +428,12 @@ impl DeletionQueueClient {
     /// This is cancel-safe.  If you drop the future before it completes, the message
     /// is not pushed, although in the context of the deletion queue it doesn't matter: once
     /// we decide to do a deletion the decision is always final.
-    fn do_push<T>(
+    async fn do_push<T>(
         &self,
-        queue: &tokio::sync::mpsc::UnboundedSender<T>,
+        queue: &tokio::sync::mpsc::Sender<T>,
         msg: T,
     ) -> Result<(), DeletionQueueError> {
-        match queue.send(msg) {
+        match queue.send(msg).await {
             Ok(_) => Ok(()),
             Err(e) => {
                 // This shouldn't happen, we should shut down all tenants before
@@ -445,7 +445,7 @@ impl DeletionQueueClient {
         }
     }
 
-    pub(crate) fn recover(
+    pub(crate) async fn recover(
         &self,
         attached_tenants: HashMap<TenantId, Generation>,
     ) -> Result<(), DeletionQueueError> {
@@ -453,6 +453,7 @@ impl DeletionQueueClient {
             &self.tx,
             ListWriterQueueMessage::Recover(RecoverOp { attached_tenants }),
         )
+        .await
     }
 
     /// When a Timeline wishes to update the remote_consistent_lsn that it exposes to the outside
@@ -525,21 +526,6 @@ impl DeletionQueueClient {
             return self.flush_immediate().await;
         }
 
-        self.push_layers_sync(tenant_id, timeline_id, current_generation, layers)
-    }
-
-    /// When a Tenant has a generation, push_layers is always synchronous because
-    /// the ListValidator channel is an unbounded channel.
-    ///
-    /// This can be merged into push_layers when we remove the Generation-less mode
-    /// support (`<https://github.com/neondatabase/neon/issues/5395>`)
-    pub(crate) fn push_layers_sync(
-        &self,
-        tenant_id: TenantId,
-        timeline_id: TimelineId,
-        current_generation: Generation,
-        layers: Vec<(LayerFileName, Generation)>,
-    ) -> Result<(), DeletionQueueError> {
         metrics::DELETION_QUEUE
             .keys_submitted
             .inc_by(layers.len() as u64);
@@ -553,16 +539,17 @@ impl DeletionQueueClient {
                 objects: Vec::new(),
             }),
         )
+        .await
     }
 
     /// This is cancel-safe.  If you drop the future the flush may still happen in the background.
     async fn do_flush<T>(
         &self,
-        queue: &tokio::sync::mpsc::UnboundedSender<T>,
+        queue: &tokio::sync::mpsc::Sender<T>,
         msg: T,
         rx: tokio::sync::oneshot::Receiver<()>,
     ) -> Result<(), DeletionQueueError> {
-        self.do_push(queue, msg)?;
+        self.do_push(queue, msg).await?;
         if rx.await.is_err() {
             // This shouldn't happen if tenants are shut down before deletion queue.  If we
             // encounter a bug like this, then a flusher will incorrectly believe it has flushed
@@ -583,18 +570,6 @@ impl DeletionQueueClient {
             .await
     }
 
-    /// Issue a flush without waiting for it to complete.  This is useful on advisory flushes where
-    /// the caller wants to avoid the risk of waiting for lots of enqueued work, such as on tenant
-    /// detach where flushing is nice but not necessary.
-    ///
-    /// This function provides no guarantees of work being done.
-    pub fn flush_advisory(&self) {
-        let (flush_op, _) = FlushOp::new();
-
-        // Transmit the flush message, ignoring any result (such as a closed channel during shutdown).
-        drop(self.tx.send(ListWriterQueueMessage::FlushExecute(flush_op)));
-    }
-
     // Wait until all previous deletions are executed
     pub(crate) async fn flush_execute(&self) -> Result<(), DeletionQueueError> {
         debug!("flush_execute: flushing to deletion lists...");
@@ -611,7 +586,9 @@ impl DeletionQueueClient {
         // Flush any immediate-mode deletions (the above backend flush will only flush
         // the executor if deletions had flowed through the backend)
         debug!("flush_execute: flushing execution...");
-        self.flush_immediate().await?;
+        let (flush_op, rx) = FlushOp::new();
+        self.do_flush(&self.executor_tx, DeleterMessage::Flush(flush_op), rx)
+            .await?;
         debug!("flush_execute: finished flushing execution...");
         Ok(())
     }
@@ -666,10 +643,8 @@ impl DeletionQueue {
     where
         C: ControlPlaneGenerationsApi + Send + Sync,
     {
-        // Unbounded channel: enables non-async functions to submit deletions.  The actual length is
-        // constrained by how promptly the ListWriter wakes up and drains it, which should be frequent
-        // enough to avoid this taking pathologically large amount of memory.
-        let (tx, rx) = tokio::sync::mpsc::unbounded_channel();
+        // Deep channel: it consumes deletions from all timelines and we do not want to block them
+        let (tx, rx) = tokio::sync::mpsc::channel(16384);
 
         // Shallow channel: it carries DeletionLists which each contain up to thousands of deletions
         let (backend_tx, backend_rx) = tokio::sync::mpsc::channel(16);
@@ -982,7 +957,7 @@ mod test {
         // Basic test that the deletion queue processes the deletions we pass into it
         let ctx = setup("deletion_queue_smoke").expect("Failed test setup");
         let client = ctx.deletion_queue.new_client();
-        client.recover(HashMap::new())?;
+        client.recover(HashMap::new()).await?;
 
         let layer_file_name_1: LayerFileName = "000000000000000000000000000000000000-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF__00000000016B59D8-00000000016B5A51".parse().unwrap();
         let tenant_id = ctx.harness.tenant_id;
@@ -1050,7 +1025,7 @@ mod test {
     async fn deletion_queue_validation() -> anyhow::Result<()> {
         let ctx = setup("deletion_queue_validation").expect("Failed test setup");
         let client = ctx.deletion_queue.new_client();
-        client.recover(HashMap::new())?;
+        client.recover(HashMap::new()).await?;
 
         // Generation that the control plane thinks is current
         let latest_generation = Generation::new(0xdeadbeef);
@@ -1107,7 +1082,7 @@ mod test {
         // Basic test that the deletion queue processes the deletions we pass into it
         let mut ctx = setup("deletion_queue_recovery").expect("Failed test setup");
         let client = ctx.deletion_queue.new_client();
-        client.recover(HashMap::new())?;
+        client.recover(HashMap::new()).await?;
 
         let tenant_id = ctx.harness.tenant_id;
 
@@ -1170,7 +1145,9 @@ mod test {
         drop(client);
         ctx.restart().await;
         let client = ctx.deletion_queue.new_client();
-        client.recover(HashMap::from([(tenant_id, now_generation)]))?;
+        client
+            .recover(HashMap::from([(tenant_id, now_generation)]))
+            .await?;
 
         info!("Flush-executing");
         client.flush_execute().await?;
@@ -1196,7 +1173,7 @@ pub(crate) mod mock {
     };
 
     pub struct ConsumerState {
-        rx: tokio::sync::mpsc::UnboundedReceiver<ListWriterQueueMessage>,
+        rx: tokio::sync::mpsc::Receiver<ListWriterQueueMessage>,
         executor_rx: tokio::sync::mpsc::Receiver<DeleterMessage>,
     }
 
@@ -1273,7 +1250,7 @@ pub(crate) mod mock {
     }
 
     pub struct MockDeletionQueue {
-        tx: tokio::sync::mpsc::UnboundedSender<ListWriterQueueMessage>,
+        tx: tokio::sync::mpsc::Sender<ListWriterQueueMessage>,
         executor_tx: tokio::sync::mpsc::Sender<DeleterMessage>,
         executed: Arc<AtomicUsize>,
         remote_storage: Option<GenericRemoteStorage>,
@@ -1283,7 +1260,7 @@ pub(crate) mod mock {
 
     impl MockDeletionQueue {
         pub fn new(remote_storage: Option<GenericRemoteStorage>) -> Self {
-            let (tx, rx) = tokio::sync::mpsc::unbounded_channel();
+            let (tx, rx) = tokio::sync::mpsc::channel(16384);
             let (executor_tx, executor_rx) = tokio::sync::mpsc::channel(16384);
 
             let executed = Arc::new(AtomicUsize::new(0));

pageserver/src/deletion_queue/deleter.rs

@@ -12,6 +12,7 @@ use remote_storage::MAX_KEYS_PER_DELETE;
 use std::time::Duration;
 use tokio_util::sync::CancellationToken;
 use tracing::info;
+use tracing::warn;
 
 use crate::metrics;
 
@@ -87,10 +88,7 @@ impl Deleter {
                     self.accumulator.clear();
                 }
                 Err(e) => {
-                    // The RemoteStorage interface doesn't discriminate between
-                    // real errors and 503/429 responses, so we log at INFO level
-                    // to avoid propagating spurious error-severity logs.
-                    info!("DeleteObjects request failed: {e:#}, will retry");
+                    warn!("DeleteObjects request failed: {e:#}, will retry");
                     metrics::DELETION_QUEUE
                         .remote_errors
                         .with_label_values(&["execute"])

pageserver/src/deletion_queue/list_writer.rs

@@ -85,7 +85,7 @@ pub(super) struct ListWriter {
     conf: &'static PageServerConf,
 
     // Incoming frontend requests to delete some keys
-    rx: tokio::sync::mpsc::UnboundedReceiver<ListWriterQueueMessage>,
+    rx: tokio::sync::mpsc::Receiver<ListWriterQueueMessage>,
 
     // Outbound requests to the backend to execute deletion lists we have composed.
     tx: tokio::sync::mpsc::Sender<ValidatorQueueMessage>,
@@ -111,7 +111,7 @@ impl ListWriter {
 
     pub(super) fn new(
         conf: &'static PageServerConf,
-        rx: tokio::sync::mpsc::UnboundedReceiver<ListWriterQueueMessage>,
+        rx: tokio::sync::mpsc::Receiver<ListWriterQueueMessage>,
         tx: tokio::sync::mpsc::Sender<ValidatorQueueMessage>,
         cancel: CancellationToken,
     ) -> Self {

pageserver/src/http/routes.rs

@@ -575,14 +575,9 @@ async fn tenant_detach_handler(
 
     let state = get_state(&request);
     let conf = state.conf;
-    mgr::detach_tenant(
-        conf,
-        tenant_id,
-        detach_ignored.unwrap_or(false),
-        &state.deletion_queue_client,
-    )
-    .instrument(info_span!("tenant_detach", %tenant_id))
-    .await?;
+    mgr::detach_tenant(conf, tenant_id, detach_ignored.unwrap_or(false))
+        .instrument(info_span!("tenant_detach", %tenant_id))
+        .await?;
 
     json_response(StatusCode::OK, ())
 }
@@ -1039,7 +1034,7 @@ async fn put_tenant_location_config_handler(
     // The `Detached` state is special, it doesn't upsert a tenant, it removes
     // its local disk content and drops it from memory.
     if let LocationConfigMode::Detached = request_data.config.mode {
-        mgr::detach_tenant(conf, tenant_id, true, &state.deletion_queue_client)
+        mgr::detach_tenant(conf, tenant_id, true)
             .instrument(info_span!("tenant_detach", %tenant_id))
             .await?;
         return json_response(StatusCode::OK, ());

pageserver/src/tenant.rs

@@ -45,7 +45,6 @@ use std::sync::{Mutex, RwLock};
 use std::time::{Duration, Instant};
 
 use self::config::AttachedLocationConfig;
-use self::config::AttachmentMode;
 use self::config::LocationConf;
 use self::config::TenantConf;
 use self::delete::DeleteTenantFlow;
@@ -2077,15 +2076,6 @@ impl Tenant {
             }
         }
     }
-
-    pub(crate) fn get_attach_mode(&self) -> AttachmentMode {
-        self.tenant_conf
-            .read()
-            .unwrap()
-            .location
-            .attach_mode
-            .clone()
-    }
 }
 
 /// Given a Vec of timelines and their ancestors (timeline_id, ancestor_id),

pageserver/src/tenant/mgr.rs

@@ -24,7 +24,7 @@ use crate::control_plane_client::{
 };
 use crate::deletion_queue::DeletionQueueClient;
 use crate::task_mgr::{self, TaskKind};
-use crate::tenant::config::{AttachmentMode, LocationConf, LocationMode, TenantConfOpt};
+use crate::tenant::config::{LocationConf, LocationMode, TenantConfOpt};
 use crate::tenant::delete::DeleteTenantFlow;
 use crate::tenant::{
     create_tenant_files, AttachedTenantConf, CreateTenantFilesMode, Tenant, TenantState,
@@ -206,7 +206,8 @@ async fn init_load_generations(
     if resources.remote_storage.is_some() {
         resources
             .deletion_queue_client
-            .recover(generations.clone())?;
+            .recover(generations.clone())
+            .await?;
     }
 
     Ok(Some(generations))
@@ -694,18 +695,6 @@ pub(crate) async fn upsert_location(
 
     if let Some(tenant) = shutdown_tenant {
         let (_guard, progress) = utils::completion::channel();
-
-        match tenant.get_attach_mode() {
-            AttachmentMode::Single | AttachmentMode::Multi => {
-                // Before we leave our state as the presumed holder of the latest generation,
-                // flush any outstanding deletions to reduce the risk of leaking objects.
-                deletion_queue_client.flush_advisory()
-            }
-            AttachmentMode::Stale => {
-                // If we're stale there's not point trying to flush deletions
-            }
-        };
-
         info!("Shutting down attached tenant");
         match tenant.shutdown(progress, false).await {
             Ok(()) => {}
@@ -860,16 +849,8 @@ pub async fn detach_tenant(
     conf: &'static PageServerConf,
     tenant_id: TenantId,
     detach_ignored: bool,
-    deletion_queue_client: &DeletionQueueClient,
 ) -> Result<(), TenantStateError> {
-    let tmp_path = detach_tenant0(
-        conf,
-        &TENANTS,
-        tenant_id,
-        detach_ignored,
-        deletion_queue_client,
-    )
-    .await?;
+    let tmp_path = detach_tenant0(conf, &TENANTS, tenant_id, detach_ignored).await?;
     // Although we are cleaning up the tenant, this task is not meant to be bound by the lifetime of the tenant in memory.
     // After a tenant is detached, there are no more task_mgr tasks for that tenant_id.
     let task_tenant_id = None;
@@ -894,7 +875,6 @@ async fn detach_tenant0(
     tenants: &tokio::sync::RwLock<TenantsMap>,
     tenant_id: TenantId,
     detach_ignored: bool,
-    deletion_queue_client: &DeletionQueueClient,
 ) -> Result<Utf8PathBuf, TenantStateError> {
     let tenant_dir_rename_operation = |tenant_id_to_clean| async move {
         let local_tenant_directory = conf.tenant_path(&tenant_id_to_clean);
@@ -906,10 +886,6 @@ async fn detach_tenant0(
     let removal_result =
         remove_tenant_from_memory(tenants, tenant_id, tenant_dir_rename_operation(tenant_id)).await;
 
-    // Flush pending deletions, so that they have a good chance of passing validation
-    // before this tenant is potentially re-attached elsewhere.
-    deletion_queue_client.flush_advisory();
-
     // Ignored tenants are not present in memory and will bail the removal from memory operation.
     // Before returning the error, check for ignored tenant removal case — we only need to clean its local files then.
     if detach_ignored && matches!(removal_result, Err(TenantStateError::NotFound(_))) {

proxy/src/http/sql_over_http.rs

@@ -47,7 +47,6 @@ enum Payload {
 
 const MAX_RESPONSE_SIZE: usize = 10 * 1024 * 1024; // 10 MiB
 const MAX_REQUEST_SIZE: u64 = 10 * 1024 * 1024; // 10 MiB
-const HTTP_CONNECTION_TIMEOUT: tokio::time::Duration = tokio::time::Duration::from_secs(15);
 
 static RAW_TEXT_OUTPUT: HeaderName = HeaderName::from_static("neon-raw-text-output");
 static ARRAY_MODE: HeaderName = HeaderName::from_static("neon-array-mode");
@@ -190,44 +189,27 @@ pub async fn handle(
     conn_pool: Arc<GlobalConnPool>,
     session_id: uuid::Uuid,
 ) -> Result<Response<Body>, ApiError> {
-    let result = tokio::time::timeout(
-        HTTP_CONNECTION_TIMEOUT,
-        handle_inner(request, sni_hostname, conn_pool, session_id),
-    )
-    .await;
+    let result = handle_inner(request, sni_hostname, conn_pool, session_id).await;
+
     let mut response = match result {
-        Ok(r) => match r {
-            Ok(r) => r,
-            Err(e) => {
-                let message = format!("{:?}", e);
-                let code = e.downcast_ref::<tokio_postgres::Error>().and_then(|e| {
-                    e.code()
-                        .map(|s| serde_json::to_value(s.code()).unwrap_or_default())
-                });
-                let code = match code {
-                    Some(c) => c,
+        Ok(r) => r,
+        Err(e) => {
+            let message = format!("{:?}", e);
+            let code = match e.downcast_ref::<tokio_postgres::Error>() {
+                Some(e) => match e.code() {
+                    Some(e) => serde_json::to_value(e.code()).unwrap(),
                     None => Value::Null,
-                };
-                error!(
-                    ?code,
-                    "sql-over-http per-client task finished with an error: {e:#}"
-                );
-                // TODO: this shouldn't always be bad request.
-                json_response(
-                    StatusCode::BAD_REQUEST,
-                    json!({ "message": message, "code": code }),
-                )?
-            }
-        },
-        Err(_) => {
-            let message = format!(
-                "HTTP-Connection timed out, execution time exeeded {} seconds",
-                HTTP_CONNECTION_TIMEOUT.as_secs()
+                },
+                None => Value::Null,
+            };
+            error!(
+                ?code,
+                "sql-over-http per-client task finished with an error: {e:#}"
             );
-            error!(message);
+            // TODO: this shouldn't always be bad request.
             json_response(
-                StatusCode::GATEWAY_TIMEOUT,
-                json!({ "message": message, "code": StatusCode::GATEWAY_TIMEOUT.as_u16() }),
+                StatusCode::BAD_REQUEST,
+                json!({ "message": message, "code": code }),
             )?
         }
     };

proxy/src/proxy.rs

@@ -7,7 +7,6 @@ use crate::{
     compute::{self, PostgresConnection},
     config::{ProxyConfig, TlsConfig},
     console::{self, errors::WakeComputeError, messages::MetricsAuxInfo, Api},
-    http::StatusCode,
     metrics::{Ids, USAGE_METRICS},
     protocol2::WithClientIp,
     stream::{PqStream, Stream},
@@ -76,15 +75,6 @@ static NUM_CONNECTION_FAILURES: Lazy<IntCounterVec> = Lazy::new(|| {
     .unwrap()
 });
 
-static NUM_WAKEUP_FAILURES: Lazy<IntCounterVec> = Lazy::new(|| {
-    register_int_counter_vec!(
-        "proxy_connection_failures_breakdown",
-        "Number of wake-up failures (per kind).",
-        &["retry", "kind"],
-    )
-    .unwrap()
-});
-
 static NUM_BYTES_PROXIED_COUNTER: Lazy<IntCounterVec> = Lazy::new(|| {
     register_int_counter_vec!(
         "proxy_io_bytes_per_client",
@@ -407,46 +397,6 @@ impl ConnectMechanism for TcpMechanism<'_> {
     }
 }
 
-const fn bool_to_str(x: bool) -> &'static str {
-    if x {
-        "true"
-    } else {
-        "false"
-    }
-}
-
-fn report_error(e: &WakeComputeError, retry: bool) {
-    use crate::console::errors::ApiError;
-    let retry = bool_to_str(retry);
-    let kind = match e {
-        WakeComputeError::BadComputeAddress(_) => "bad_compute_address",
-        WakeComputeError::ApiError(ApiError::Transport(_)) => "api_transport_error",
-        WakeComputeError::ApiError(ApiError::Console {
-            status: StatusCode::LOCKED,
-            ref text,
-        }) if text.contains("written data quota exceeded")
-            || text.contains("the limit for current plan reached") =>
-        {
-            "quota_exceeded"
-        }
-        WakeComputeError::ApiError(ApiError::Console {
-            status: StatusCode::LOCKED,
-            ..
-        }) => "api_console_locked",
-        WakeComputeError::ApiError(ApiError::Console {
-            status: StatusCode::BAD_REQUEST,
-            ..
-        }) => "api_console_bad_request",
-        WakeComputeError::ApiError(ApiError::Console { status, .. })
-            if status.is_server_error() =>
-        {
-            "api_console_other_server_error"
-        }
-        WakeComputeError::ApiError(ApiError::Console { .. }) => "api_console_other_error",
-    };
-    NUM_WAKEUP_FAILURES.with_label_values(&[retry, kind]).inc();
-}
-
 /// Try to connect to the compute node, retrying if necessary.
 /// This function might update `node_info`, so we take it by `&mut`.
 #[tracing::instrument(skip_all)]
@@ -490,12 +440,10 @@ where
         match handle_try_wake(wake_res, num_retries) {
             Err(e) => {
                 error!(error = ?e, num_retries, retriable = false, "couldn't wake compute node");
-                report_error(&e, false);
                 return Err(e.into());
             }
             // failed to wake up but we can continue to retry
             Ok(ControlFlow::Continue(e)) => {
-                report_error(&e, true);
                 warn!(error = ?e, num_retries, retriable = true, "couldn't wake compute node");
             }
             // successfully woke up a compute node and can break the wakeup loop

s3_scrubber/Cargo.toml

@@ -33,9 +33,13 @@ reqwest = { workspace = true, default-features = false, features = ["rustls-tls"
 aws-config = { workspace = true, default-features = false, features = ["rustls", "credentials-sso"] }
 
 pageserver = { path = "../pageserver" }
+remote_storage = { path = "../libs/remote_storage" }
 
 tracing.workspace = true
-tracing-subscriber.workspace = true
+tracing-subscriber = { version = "0.3.17", features = ["ansi"] }
 clap.workspace = true
 tracing-appender = "0.2"
 histogram = "0.7"
+humantime.workspace = true
+camino.workspace = true
+aws-smithy-types = "0.56.1"

s3_scrubber/src/lib.rs

@@ -4,7 +4,6 @@ pub mod delete_batch_producer;
 pub mod metadata_stream;
 mod s3_deletion;
 pub mod scan_metadata;
-
 use std::env;
 use std::fmt::Display;
 use std::time::Duration;

s3_scrubber/src/main.rs

@@ -16,6 +16,8 @@ use tracing::{info, warn};
 
 use clap::{Parser, Subcommand, ValueEnum};
 
+mod restore_tenant_from_object_versioning;
+
 #[derive(Parser)]
 #[command(author, version, about, long_about = None)]
 #[command(arg_required_else_help(true))]
@@ -59,6 +61,9 @@ enum Command {
         skip_validation: bool,
     },
     ScanMetadata {},
+    RestoreTenantFromObjectVersioningMostRecentIndexPart(
+        restore_tenant_from_object_versioning::Command,
+    ),
 }
 
 async fn tidy(
@@ -247,5 +252,14 @@ async fn main() -> anyhow::Result<()> {
                 }
             }
         },
+        Command::RestoreTenantFromObjectVersioningMostRecentIndexPart(arg) => {
+            match restore_tenant_from_object_versioning::doit(arg).await {
+                Err(e) => {
+                    tracing::error!("Failed: {e:?}");
+                    Err(e)
+                }
+                Ok(()) => Ok(()),
+            }
+        }
     }
 }

s3_scrubber/src/restore_tenant_from_object_versioning.rs

@@ -0,0 +1,513 @@
+//! Restore pageserver state from S3 object versioning.
+//!
+//! This sub-cmmand allows restoring a tenant's pageserver S3 state from S3 object versioning.
+//!
+//! # Instructions
+//!
+//! - Run command
+//!   ```
+//!   SSO_ACCOUNT_ID=... REGION=... \
+//!   BUCKET=neon-{prod,staging}-storage-... \
+//!     cargo run -p s3_scrubber \
+//!         restore-tenant-from-object-versioning-most-recent-index-part \
+//!         TENANT_TO_RESTORE \
+//!         ./restore
+//!         timeline-list TIMELINE_TO_RESTORE TIMELINE_TO_RESTORE ...
+//!   ```
+//! - `./restore` now contains the timeline state referenced by the latest `index_part.json`s of the
+//!   specified timelines in the `timeline-list`` argument
+//! - Use `cargo neon` to start a pageserver
+//!     - `rm -rf .neon`
+//!     - `cargo neon init`
+//!     - `sed -i 's/\(.*control_plane_api.*\)/#\1/' .neon/config`
+//!     - `sed -i 's/\(.*control_plane_api.*\)/#\1/' .neon/pageserver_1/pageserver.toml`
+//!     - configure the pageserver remote storage config to point to the restore directory.
+//!       Use your text editor to edit the TOML file: `.neon/pageserver_1/pageserver.toml`.
+//!       ```
+//!       [remote_storage]
+//!       local_path = "/path/to/restore/pageserver/v1"
+//!       ````
+//!    - `cargo neon start`
+//!    - make sure attaching the tenant works
+//!      - `curl -X POST localhost:9898/v1/tenant/TENANT_TO_RESTORE/attach`
+//!      - check `curl -X GET localhost:9898/v1/tenant/TENANT_TO_RESTORE | jq`
+//!     - for each timeline $timeline_id to restore:
+//!         - `cargo neon mappings map --branch-name restore-$timeline_id --tenant-id TENANT_TO_RESTORE --timeline-id $timeline_id`
+//!         - `cargo neon endpoint create --tenant-id TENANT_TO_RESTORE --branch-name restore-$timeline_id ep-restore-$timeline_id`
+//!         - `cargo neon endpoint start --tenant-id TENANT_TO_RESTORE ep-restore-$timeline_id`
+//!             - it prints a connection string, looking like `postgresql://cloud_admin@127.0.0.1:PORT/DB`
+//!         - dump database contents using postgres tools
+//!             - determine PG version `$restore_pg_version` using
+//!               ```
+//!               curl -s -X GET localhost:9898/v1/tenant/TENANT_TO_RESTORE/timeline/$timeline_id | jq .pg_version
+//!               ```
+//!             - pg_dumpall
+//!               ```
+//!               ./pg_install/$restore_pg_version/bin/pg_dumpall -d THE_CONNECTION_STRING/postgres  > ./restore/pg_dumpall.out
+//!               ```
+//!            - pg_dump a specific database
+//!              ```
+//!              ./pg_install/v15/bin/pg_dump -d 'THE_CONNECTION_STRING/THEDBTODUMP' > ./restore/pg_dump_THEDBTODUMP.out
+//!              ```
+//!         - `cargo neon endpoint stop --tenant-id TENANT_TO_RESTORE restore-$timeline_id`
+//!
+//! - Use the pg_dump files to restore the database into a new Neon project.
+//!
+//! # Limitations & Future Work
+//!
+//! Just restoring Pageserver S3 state restores a consistent state at an LSN that is NOT THE LAST COMMIT LSN.
+//! The reason is that Pageserver uploads layers to S3 with implementation-specific delays that are optimized for day-to-day operation.
+//!
+//! If we still had the Safekeeper WAL, we could restore the Safekeeper S3 state in a similar way.
+//! In that case, we wouldn't need the `pg_dump` step.
+//! We would simply attach the tenant to safekeepers and pageservers.
+//! When attaching to Safekeeper, we would need to tell it that PS remote_consistent_lsn is the restore-point-LSN,
+//! i.e., the LSNs in the restored index_part.json's in Pageserver S3 state.
+//! Pageserver attach would pick up the restored state from S3, and the Safekeeper & Pageserver would
+//! resume normal operation as if the clock had been wound back to restore-point-LSN.
+
+use std::{
+    collections::{HashMap, HashSet},
+    num::{NonZeroU32, NonZeroUsize},
+    sync::Arc,
+};
+
+use anyhow::Context;
+
+use aws_sdk_s3::operation::list_object_versions::ListObjectVersionsOutput;
+use aws_types::region::Region;
+use camino::Utf8PathBuf;
+use pageserver::tenant::{
+    remote_timeline_client::index::LayerFileMetadata, TENANTS_SEGMENT_NAME,
+    TENANT_DELETED_MARKER_FILE_NAME, TIMELINES_SEGMENT_NAME,
+};
+use remote_storage::{GenericRemoteStorage, RemotePath, RemoteStorageConfig, RemoteStorageKind};
+use s3_scrubber::{init_logging, init_s3_client, BucketConfig};
+use tokio::io::AsyncReadExt;
+use tracing::{debug, info, info_span, Instrument};
+use utils::id::{TenantId, TimelineId};
+
+#[derive(Debug, Clone, clap::Subcommand)]
+enum ResurrectTimelines {
+    TimelineList { timeline_ids: Vec<TimelineId> },
+    // AllTimelinesDeletedAfter { timestamp: humantime::Timestamp },
+}
+
+#[derive(clap::Args)]
+pub(crate) struct Command {
+    tenant_id: TenantId,
+    dest_dir: Utf8PathBuf,
+    #[clap(short, long)]
+    dry_run: bool,
+    #[clap(subcommand)]
+    timelines: ResurrectTimelines,
+}
+
+pub(crate) async fn doit(args: Command) -> anyhow::Result<()> {
+    let _logging_guard = {
+        let log_prefix = format!("restore_tenant_from_object_versioning_{}", args.tenant_id);
+        let dry_suffix = if args.dry_run { "__dry" } else { "" };
+        let file_name = {
+            format!(
+                "{}_{}{}.log",
+                log_prefix,
+                chrono::Utc::now().format("%Y_%m_%d__%H_%M_%S"),
+                dry_suffix,
+            )
+        };
+        init_logging(&file_name)
+    };
+
+    let restore_dst = if tokio::fs::try_exists(&args.dest_dir).await? {
+        anyhow::bail!("destination directory already exists: {}", args.dest_dir,);
+    } else {
+        GenericRemoteStorage::from_config(&RemoteStorageConfig {
+            max_concurrent_syncs: NonZeroUsize::new(100).unwrap(),
+            max_sync_errors: NonZeroU32::new(1).unwrap(), // ???? would want so specify 0
+            storage: RemoteStorageKind::LocalFs(args.dest_dir.clone()),
+        })
+        .context("instantiate restore destination")?
+    };
+
+    let bucket_config = BucketConfig::from_env()?;
+
+    let bucket_region = Region::new(bucket_config.region);
+    let delimiter = "/".to_string();
+    let s3_client = Arc::new(init_s3_client(bucket_config.sso_account_id, bucket_region));
+    let tenant_root = [
+        "pageserver",
+        "v1",
+        TENANTS_SEGMENT_NAME,
+        &args.tenant_id.to_string(),
+    ]
+    .join(&delimiter);
+
+    let tenant_delete_marker = [&tenant_root, TENANT_DELETED_MARKER_FILE_NAME].join(&delimiter);
+
+    // - Ensure the prefix is empty when ignoring existence of versions.
+    // - Ensure the tenant delete marker key is part of `DeleteMarkers`. If it isn't, the tenant hasn't finished deletion yet and we should let pageservers complete it first.
+    // - Restore each index_part.json based on the version in DeleteMarkers as well as the layers it references. For the layers, also use the version in DeleteMarkers and ensure it is the latest.
+    // - Remove the deleted_at mark for the specified timelines.
+    //
+    // Notes:
+    //   - The restore will happen in-place because it's hard to change tenant/timeline ids.
+    //   - The restore could be interrupted mid-way.
+    //   - Hence, separate plan-making and plan-execution.
+
+    async {
+        info!("send request");
+        let res = s3_client
+            .list_objects_v2()
+            .bucket(&bucket_config.bucket)
+            .prefix(&tenant_root)
+            .send()
+            .await?;
+
+        info!(response=?res, "got response");
+
+        if res.key_count() > 0 {
+            anyhow::bail!("tenant prefix is not empty in S3");
+        }
+        if res.is_truncated() {
+            unimplemented!("can this even happen")
+        }
+        Ok(())
+    }
+    .instrument(info_span!("ensure prefix empty"))
+    .await
+    .context("ensure prefix is empty")?;
+
+    async {
+        info!("send request");
+        let res = s3_client
+            .list_object_versions()
+            .bucket(&bucket_config.bucket)
+            .prefix(&tenant_delete_marker)
+            .send()
+            .await?;
+        debug!(response=?res, "got response");
+
+        if res.is_truncated() {
+            unimplemented!("can this even happen")
+        }
+
+        let markers = res
+            .delete_markers()
+            .context("expected delete marker in response")?;
+
+        if markers.len() != 1 {
+            anyhow::bail!("expected exactly one delete marker because we create and delete the marker exactly once, got {}", markers.len());
+        }
+
+        if !markers[0].is_latest() {
+            anyhow::bail!("expected delete marker to have IsLatest set: {:?}", markers[0]);
+        }
+
+        Ok(())
+    }
+    .instrument(info_span!(
+        "ensure tenant delete marker exists in DeleteMarkers",
+        tenant_delete_marker,
+    ))
+    .await
+    .context("ensure tenant delete marker exists in DeleteMarkers")?;
+
+    let timelines = match args.timelines {
+        ResurrectTimelines::TimelineList { timeline_ids } => timeline_ids,
+    };
+
+    // Fetch all the information we need to execute the restore.
+    let version_responses_by_timeline = async {
+        let mut out: HashMap<TimelineId, Vec<Arc<ListObjectVersionsOutput>>> = Default::default();
+        for tl in &timelines {
+            async {
+                let timeline_prefix = [tenant_root.as_str(), TIMELINES_SEGMENT_NAME , &tl.to_string()].join(&delimiter);
+                let mut next_key_marker = None;
+                let mut next_version_id_marker = None;
+                loop {
+                    info!("sending request");
+                    let res: ListObjectVersionsOutput = s3_client.list_object_versions()
+                        .bucket(&bucket_config.bucket)
+                        .prefix(&timeline_prefix)
+                        .set_key_marker(next_key_marker.take())
+                        .set_version_id_marker(next_version_id_marker.take())
+                        .send()
+                        .await?;
+
+                    let res = Arc::new(res);
+                    out.entry(*tl).or_default().push(Arc::clone(&res));
+
+                    info!("got response");
+                    match res.versions() {
+                        Some(versions) => {
+                            for version in versions {
+                                info!("version: {:?}", version);
+                            }
+                        }
+                        None => {
+                            info!("no versions");
+                        }
+                    }
+                    match res.delete_markers() {
+                        Some(markers) => {
+                            for marker in markers {
+                                info!("delete marker: {:?}", marker);
+                            }
+                        }
+                        None => {
+                            info!("no delete markers");
+                        }
+                    }
+
+                    if !res.is_truncated() {
+                        break;
+                    }
+                    next_key_marker = res.next_key_marker().map(|s| s.to_string());
+                    next_version_id_marker = res.next_version_id_marker().map(|s| s.to_string());
+                    if let (None, None) = (&next_key_marker, &next_version_id_marker) {
+                        anyhow::bail!("s3 returned is_truncated=true but neither next_key_marker nor next_version_id_marker are set");
+                    }
+                }
+                Ok(())
+            }.instrument(info_span!("timeline", timeline_id=%tl)).await?;
+        }
+        anyhow::Ok(out)
+    }.instrument(info_span!("list all object versions and delete markers")).await?;
+    #[derive(Debug)]
+    struct LatestVersion {
+        key: String,
+        last_modified: aws_smithy_types::DateTime,
+        version_id: String,
+    }
+    let find_latest_version_based_on_delete_marker_last_modified = |tl: &TimelineId, key: &str| {
+        let restore_version_delete_marker = {
+            let mut candidates = Vec::new();
+            for res in &version_responses_by_timeline[tl] {
+                let Some(markers) = res.delete_markers() else {
+                    continue;
+                };
+                for marker in markers {
+                    if !marker.is_latest() {
+                        continue;
+                    }
+                    if marker.key().unwrap() != key {
+                        continue;
+                    }
+                    candidates.push(LatestVersion {
+                        key: marker.key().unwrap().to_owned(),
+                        last_modified: marker.last_modified().unwrap().clone(),
+                        version_id: marker.version_id().unwrap().to_owned(),
+                    });
+                }
+            }
+            info!(?candidates, "marker candidates");
+            if candidates.len() != 1 {
+                anyhow::bail!("expected exactly one IsLatest, got {}", candidates.len());
+            }
+            candidates.pop().unwrap()
+        };
+        info!(?restore_version_delete_marker, "found marker");
+
+        // There's no way to get the latest version from the delete marker.
+        // But, we observe (can't find written guarantee) that the Delete Marker's "Last Modified" is >= the latest version.
+        // So, find latest version based on that.
+        let restore_version = {
+            let mut candidates = Vec::new();
+            for res in &version_responses_by_timeline[tl] {
+                let Some(versions) = res.versions() else {
+                    continue;
+                };
+                for version in versions {
+                    if version.key().unwrap() != restore_version_delete_marker.key {
+                        continue;
+                    }
+                    candidates.push(LatestVersion {
+                        key: version.key().unwrap().to_owned(),
+                        last_modified: version.last_modified().unwrap().clone(),
+                        version_id: version.version_id().unwrap().to_owned(),
+                    });
+                }
+            }
+            candidates.sort_by_key(|v| v.last_modified.clone());
+            info!(?candidates, "version candidates");
+            if candidates.is_empty() {
+                anyhow::bail!(
+                    "expected at least one version matching the delete marker's key, got none"
+                );
+            }
+            {
+                let mut uniq = HashSet::new();
+                for v in &candidates {
+                    if !uniq.insert(v.last_modified.clone()) {
+                        anyhow::bail!("last_modified timestamps are not unique, don't know which version to pick");
+                    }
+                }
+            }
+            candidates.pop().unwrap() // we sorted ascending, so, pop() is the latest
+        };
+        anyhow::Ok(restore_version)
+    };
+
+    let latest_index_part_versions: HashMap<TimelineId, LatestVersion> = {
+        let span = info_span!("find index part version");
+        let _enter = span.enter();
+
+        // The latest index part for a deleted tenant is always a DeletedMarker
+        let mut out = HashMap::new();
+        for tl in &timelines {
+            let span = info_span!("timeline", timeline_id=%tl);
+            let _enter = span.enter();
+            let restore_version = find_latest_version_based_on_delete_marker_last_modified(
+                tl,
+                // TODO: support generation numbers
+                &[
+                    &tenant_root,
+                    TIMELINES_SEGMENT_NAME,
+                    &tl.to_string(),
+                    pageserver::tenant::IndexPart::FILE_NAME,
+                ]
+                .join(&delimiter),
+            )?;
+            out.insert(*tl, restore_version);
+        }
+        out
+    };
+
+    let index_part_contents: HashMap<TimelineId, pageserver::tenant::IndexPart> = async {
+        let mut out = HashMap::new();
+
+        for tl in &timelines {
+            async {
+                let v = &latest_index_part_versions[tl];
+
+                let mut body_buf = Vec::new();
+                info!("send request");
+                let res = s3_client
+                    .get_object()
+                    .bucket(&bucket_config.bucket)
+                    .key(&v.key)
+                    .version_id(&v.version_id)
+                    .send()
+                    .await?;
+                info!(?res, "got response header");
+
+                res.body
+                    .into_async_read()
+                    .read_to_end(&mut body_buf)
+                    .await?;
+
+                let body_buf = String::from_utf8(body_buf)?;
+                info!(body_buf, "received response body");
+
+                let mut index_part: pageserver::tenant::IndexPart =
+                    serde_json::from_str(&body_buf)?;
+                info!(?index_part, "parsed index part");
+
+                let deleted_at = index_part.deleted_at.take();
+                info!(
+                    ?deleted_at,
+                    "removing deleted_at field from index part, previous value logged here"
+                );
+
+                let updated_buf = serde_json::to_vec(&index_part)?;
+                let updated_buf_len = updated_buf.len();
+                info!("uploading modified index part to restore_dst");
+                restore_dst
+                    .upload(
+                        std::io::Cursor::new(updated_buf),
+                        updated_buf_len,
+                        &RemotePath::from_string(&v.key).unwrap(),
+                        None,
+                    )
+                    .await
+                    .context("upload modified index part to restore_dst")?;
+
+                out.insert(*tl, index_part);
+
+                anyhow::Ok(())
+            }
+            .instrument(info_span!("timeline", timeline_id=%tl))
+            .await?;
+        }
+
+        anyhow::Ok(out)
+    }
+    .instrument(info_span!("get index part contents"))
+    .await
+    .context("get index part contents")?;
+
+    async {
+        for (tl, index_part) in &index_part_contents {
+            async {
+                for (layer_file_name, layer_md) in &index_part.layer_metadata {
+                    let layer_md: LayerFileMetadata = layer_md.into();
+                    async {
+                        // TODO: support generations
+                        let layer_file_key = [
+                            &tenant_root,
+                            TIMELINES_SEGMENT_NAME,
+                            &tl.to_string(),
+                            &layer_file_name.file_name(),
+                        ]
+                        .join(&delimiter);
+
+                        // The latest index parts naturally reference the latest layers.
+                        // So, a deleted tenant's latest layers are the ones in DeleteMarkers.
+                        //
+                        // If we want to support restoring from not-latest index part, this will require more work.
+                        // The idea is to
+                        // 1. every index_part.json that we upload contains a strongly monotonically increasing sequence number
+                        // 2. every image layer that we upload is S3-metadata-tagged with the sequence number of the IndexPart
+                        //    in which it first appeared.
+                        // This allows to recover the correct layer object version, even if we have a bug that overwrites layers.
+                        let restore_version =
+                            find_latest_version_based_on_delete_marker_last_modified(
+                                tl,
+                                &layer_file_key,
+                            )?;
+
+                        // TODO: teach RemoteStorage copy operation so we can use s3_client.copy_object()
+                        async {
+                            let res = s3_client
+                                .get_object()
+                                .bucket(&bucket_config.bucket)
+                                .key(&restore_version.key)
+                                .version_id(&restore_version.version_id)
+                                .send()
+                                .await
+                                .context("get object header")?;
+                            // TODO: instead of file_size(), do actual data integrity checking.
+                            restore_dst
+                                .upload(
+                                    res.body.into_async_read(),
+                                    layer_md.file_size().try_into().unwrap(),
+                                    &RemotePath::from_string(&restore_version.key).unwrap(),
+                                    None,
+                                )
+                                .await
+                                .context("download-body-and-upload")?;
+                            anyhow::Ok(())
+                        }
+                        .instrument(info_span!("copy", layer_file_name=%layer_file_name))
+                        .await?;
+
+                        anyhow::Ok(())
+                    }
+                    .instrument(info_span!("layer", layer_file_name=%layer_file_name))
+                    .await?;
+                }
+
+                anyhow::Ok(())
+            }
+            .instrument(info_span!("timeline", timeline_id=%tl))
+            .await?;
+        }
+
+        anyhow::Ok(())
+    }
+    .instrument(info_span!("download layer files into restore_dst"))
+    .await
+    .context("download layers into restore dst")?;
+
+    Ok(())
+}