Line data Source code
1 : //! This module manages synchronizing local FS with remote storage.
2 : //!
3 : //! # Overview
4 : //!
5 : //! * [`RemoteTimelineClient`] provides functions related to upload/download of a particular timeline.
6 : //! It contains a queue of pending uploads, and manages the queue, performing uploads in parallel
7 : //! when it's safe to do so.
8 : //!
9 : //! * Stand-alone function, [`list_remote_timelines`], to get list of timelines of a tenant.
10 : //!
11 : //! These functions use the low-level remote storage client, [`remote_storage::RemoteStorage`].
12 : //!
13 : //! # APIs & How To Use Them
14 : //!
15 : //! There is a [RemoteTimelineClient] for each [Timeline][`crate::tenant::Timeline`] in the system,
16 : //! unless the pageserver is configured without remote storage.
17 : //!
18 : //! We allocate the client instance in [Timeline][`crate::tenant::Timeline`], i.e.,
19 : //! either in [`crate::tenant::mgr`] during startup or when creating a new
20 : //! timeline.
21 : //! However, the client does not become ready for use until we've initialized its upload queue:
22 : //!
23 : //! - For timelines that already have some state on the remote storage, we use
24 : //! [`RemoteTimelineClient::init_upload_queue`] .
25 : //! - For newly created timelines, we use
26 : //! [`RemoteTimelineClient::init_upload_queue_for_empty_remote`].
27 : //!
28 : //! The former takes the remote's [`IndexPart`] as an argument, possibly retrieved
29 : //! using [`list_remote_timelines`]. We'll elaborate on [`IndexPart`] in the next section.
30 : //!
31 : //! Whenever we've created/updated/deleted a file in a timeline directory, we schedule
32 : //! the corresponding remote operation with the timeline's [`RemoteTimelineClient`]:
33 : //!
34 : //! - [`RemoteTimelineClient::schedule_layer_file_upload`] when we've created a new layer file.
35 : //! - [`RemoteTimelineClient::schedule_index_upload_for_metadata_update`] when we've updated the timeline metadata file.
36 : //! - [`RemoteTimelineClient::schedule_index_upload_for_file_changes`] to upload an updated index file, after we've scheduled file uploads
37 : //! - [`RemoteTimelineClient::schedule_layer_file_deletion`] when we've deleted one or more layer files.
38 : //!
39 : //! Internally, these functions create [`UploadOp`]s and put them in a queue.
40 : //!
41 : //! There are also APIs for downloading files.
42 : //! These are not part of the aforementioned queuing and will not be discussed
43 : //! further here, except in the section covering tenant attach.
44 : //!
45 : //! # Remote Storage Structure & [`IndexPart`] Index File
46 : //!
47 : //! The "directory structure" in the remote storage mirrors the local directory structure, with paths
48 : //! like `tenants/<tenant_id>/timelines/<timeline_id>/<layer filename>`.
49 : //! Yet instead of keeping the `metadata` file remotely, we wrap it with more
50 : //! data in an "index file" aka [`IndexPart`], containing the list of **all** remote
51 : //! files for a given timeline.
52 : //! If a file is not referenced from [`IndexPart`], it's not part of the remote storage state.
53 : //!
54 : //! Having the `IndexPart` also avoids expensive and slow `S3 list` commands.
55 : //!
56 : //! # Consistency
57 : //!
58 : //! To have a consistent remote structure, it's important that uploads and
59 : //! deletions are performed in the right order. For example, the index file
60 : //! contains a list of layer files, so it must not be uploaded until all the
61 : //! layer files that are in its list have been successfully uploaded.
62 : //!
63 : //! The contract between client and its user is that the user is responsible of
64 : //! scheduling operations in an order that keeps the remote consistent as
65 : //! described above.
66 : //!
67 : //! From the user's perspective, the operations are executed sequentially.
68 : //! Internally, the client knows which operations can be performed in parallel,
69 : //! and which operations act like a "barrier" that require preceding operations
70 : //! to finish. The calling code just needs to call the schedule-functions in the
71 : //! correct order, and the client will parallelize the operations in a way that
72 : //! is safe. For more details, see `UploadOp::can_bypass`.
73 : //!
74 : //! All of this relies on the following invariants:
75 : //!
76 : //! - We rely on read-after write consistency in the remote storage.
77 : //! - Layer files are immutable.
78 : //!
79 : //! NB: Pageserver assumes that it has exclusive write access to the tenant in remote
80 : //! storage. Different tenants can be attached to different pageservers, but if the
81 : //! same tenant is attached to two pageservers at the same time, they will overwrite
82 : //! each other's index file updates, and confusion will ensue. There's no interlock or
83 : //! mechanism to detect that in the pageserver, we rely on the control plane to ensure
84 : //! that that doesn't happen.
85 : //!
86 : //! ## Implementation Note
87 : //!
88 : //! The *actual* remote state lags behind the *desired* remote state while
89 : //! there are in-flight operations.
90 : //! We keep track of the desired remote state in [`UploadQueueInitialized::dirty`].
91 : //! It is initialized based on the [`IndexPart`] that was passed during init
92 : //! and updated with every `schedule_*` function call.
93 : //! All this is necessary necessary to compute the future [`IndexPart`]s
94 : //! when scheduling an operation while other operations that also affect the
95 : //! remote [`IndexPart`] are in flight.
96 : //!
97 : //! # Retries & Error Handling
98 : //!
99 : //! The client retries operations indefinitely, using exponential back-off.
100 : //! There is no way to force a retry, i.e., interrupt the back-off.
101 : //! This could be built easily.
102 : //!
103 : //! # Cancellation
104 : //!
105 : //! The operations execute as plain [`task_mgr`] tasks, scoped to
106 : //! the client's tenant and timeline.
107 : //! Dropping the client will drop queued operations but not executing operations.
108 : //! These will complete unless the `task_mgr` tasks are cancelled using `task_mgr`
109 : //! APIs, e.g., during pageserver shutdown, timeline delete, or tenant detach.
110 : //!
111 : //! # Completion
112 : //!
113 : //! Once an operation has completed, we update [`UploadQueueInitialized::clean`] immediately,
114 : //! and submit a request through the DeletionQueue to update
115 : //! [`UploadQueueInitialized::visible_remote_consistent_lsn`] after it has
116 : //! validated that our generation is not stale. It is this visible value
117 : //! that is advertized to safekeepers as a signal that that they can
118 : //! delete the WAL up to that LSN.
119 : //!
120 : //! The [`RemoteTimelineClient::wait_completion`] method can be used to wait
121 : //! for all pending operations to complete. It does not prevent more
122 : //! operations from getting scheduled.
123 : //!
124 : //! # Crash Consistency
125 : //!
126 : //! We do not persist the upload queue state.
127 : //! If we drop the client, or crash, all unfinished operations are lost.
128 : //!
129 : //! To recover, the following steps need to be taken:
130 : //! - Retrieve the current remote [`IndexPart`]. This gives us a
131 : //! consistent remote state, assuming the user scheduled the operations in
132 : //! the correct order.
133 : //! - Initiate upload queue with that [`IndexPart`].
134 : //! - Reschedule all lost operations by comparing the local filesystem state
135 : //! and remote state as per [`IndexPart`]. This is done in
136 : //! [`Tenant::timeline_init_and_sync`].
137 : //!
138 : //! Note that if we crash during file deletion between the index update
139 : //! that removes the file from the list of files, and deleting the remote file,
140 : //! the file is leaked in the remote storage. Similarly, if a new file is created
141 : //! and uploaded, but the pageserver dies permanently before updating the
142 : //! remote index file, the new file is leaked in remote storage. We accept and
143 : //! tolerate that for now.
144 : //! Note further that we cannot easily fix this by scheduling deletes for every
145 : //! file that is present only on the remote, because we cannot distinguish the
146 : //! following two cases:
147 : //! - (1) We had the file locally, deleted it locally, scheduled a remote delete,
148 : //! but crashed before it finished remotely.
149 : //! - (2) We never had the file locally because we haven't on-demand downloaded
150 : //! it yet.
151 : //!
152 : //! # Downloads
153 : //!
154 : //! In addition to the upload queue, [`RemoteTimelineClient`] has functions for
155 : //! downloading files from the remote storage. Downloads are performed immediately
156 : //! against the `RemoteStorage`, independently of the upload queue.
157 : //!
158 : //! When we attach a tenant, we perform the following steps:
159 : //! - create `Tenant` object in `TenantState::Attaching` state
160 : //! - List timelines that are present in remote storage, and for each:
161 : //! - download their remote [`IndexPart`]s
162 : //! - create `Timeline` struct and a `RemoteTimelineClient`
163 : //! - initialize the client's upload queue with its `IndexPart`
164 : //! - schedule uploads for layers that are only present locally.
165 : //! - After the above is done for each timeline, open the tenant for business by
166 : //! transitioning it from `TenantState::Attaching` to `TenantState::Active` state.
167 : //! This starts the timelines' WAL-receivers and the tenant's GC & Compaction loops.
168 : //!
169 : //! # Operating Without Remote Storage
170 : //!
171 : //! If no remote storage configuration is provided, the [`RemoteTimelineClient`] is
172 : //! not created and the uploads are skipped.
173 : //!
174 : //! [`Tenant::timeline_init_and_sync`]: super::Tenant::timeline_init_and_sync
175 : //! [`Timeline::load_layer_map`]: super::Timeline::load_layer_map
176 :
177 : pub(crate) mod download;
178 : pub mod index;
179 : pub mod manifest;
180 : pub(crate) mod upload;
181 :
182 : use std::collections::{HashMap, HashSet, VecDeque};
183 : use std::ops::DerefMut;
184 : use std::sync::atomic::{AtomicU32, Ordering};
185 : use std::sync::{Arc, Mutex, OnceLock};
186 : use std::time::Duration;
187 :
188 : use anyhow::Context;
189 : use camino::Utf8Path;
190 : use chrono::{NaiveDateTime, Utc};
191 : pub(crate) use download::{
192 : download_index_part, download_initdb_tar_zst, download_tenant_manifest, is_temp_download_file,
193 : list_remote_tenant_shards, list_remote_timelines,
194 : };
195 : use index::GcCompactionState;
196 : pub(crate) use index::LayerFileMetadata;
197 : use pageserver_api::models::{RelSizeMigration, TimelineArchivalState, TimelineVisibilityState};
198 : use pageserver_api::shard::{ShardIndex, TenantShardId};
199 : use regex::Regex;
200 : use remote_storage::{
201 : DownloadError, GenericRemoteStorage, ListingMode, RemotePath, TimeoutOrCancel,
202 : };
203 : use scopeguard::ScopeGuard;
204 : use tokio_util::sync::CancellationToken;
205 : use tracing::{Instrument, debug, error, info, info_span, instrument, warn};
206 : pub(crate) use upload::upload_initdb_dir;
207 : use utils::backoff::{
208 : self, DEFAULT_BASE_BACKOFF_SECONDS, DEFAULT_MAX_BACKOFF_SECONDS, exponential_backoff,
209 : };
210 : use utils::id::{TenantId, TimelineId};
211 : use utils::lsn::Lsn;
212 : use utils::pausable_failpoint;
213 : use utils::shard::ShardNumber;
214 :
215 : use self::index::IndexPart;
216 : use super::config::AttachedLocationConfig;
217 : use super::metadata::MetadataUpdate;
218 : use super::storage_layer::{Layer, LayerName, ResidentLayer};
219 : use super::timeline::import_pgdata;
220 : use super::upload_queue::{NotInitialized, SetDeletedFlagProgress};
221 : use super::{DeleteTimelineError, Generation};
222 : use crate::config::PageServerConf;
223 : use crate::context::RequestContext;
224 : use crate::deletion_queue::{DeletionQueueClient, DeletionQueueError};
225 : use crate::metrics::{
226 : MeasureRemoteOp, REMOTE_ONDEMAND_DOWNLOADED_BYTES, REMOTE_ONDEMAND_DOWNLOADED_LAYERS,
227 : RemoteOpFileKind, RemoteOpKind, RemoteTimelineClientMetrics,
228 : RemoteTimelineClientMetricsCallTrackSize,
229 : };
230 : use crate::task_mgr::{BACKGROUND_RUNTIME, TaskKind, shutdown_token};
231 : use crate::tenant::metadata::TimelineMetadata;
232 : use crate::tenant::remote_timeline_client::download::download_retry;
233 : use crate::tenant::storage_layer::AsLayerDesc;
234 : use crate::tenant::upload_queue::{
235 : Delete, OpType, UploadOp, UploadQueue, UploadQueueInitialized, UploadQueueStopped,
236 : UploadQueueStoppedDeletable, UploadTask,
237 : };
238 : use crate::tenant::{TIMELINES_SEGMENT_NAME, debug_assert_current_span_has_tenant_and_timeline_id};
239 : use crate::{TENANT_HEATMAP_BASENAME, task_mgr};
240 :
241 : // Occasional network issues and such can cause remote operations to fail, and
242 : // that's expected. If a download fails, we log it at info-level, and retry.
243 : // But after FAILED_DOWNLOAD_WARN_THRESHOLD retries, we start to log it at WARN
244 : // level instead, as repeated failures can mean a more serious problem. If it
245 : // fails more than FAILED_DOWNLOAD_RETRIES times, we give up
246 : pub(crate) const FAILED_DOWNLOAD_WARN_THRESHOLD: u32 = 3;
247 : pub(crate) const FAILED_REMOTE_OP_RETRIES: u32 = 10;
248 :
249 : // Similarly log failed uploads and deletions at WARN level, after this many
250 : // retries. Uploads and deletions are retried forever, though.
251 : pub(crate) const FAILED_UPLOAD_WARN_THRESHOLD: u32 = 3;
252 :
253 : pub(crate) const INITDB_PATH: &str = "initdb.tar.zst";
254 :
255 : pub(crate) const INITDB_PRESERVED_PATH: &str = "initdb-preserved.tar.zst";
256 :
257 : /// Default buffer size when interfacing with [`tokio::fs::File`].
258 : pub(crate) const BUFFER_SIZE: usize = 32 * 1024;
259 :
260 : /// Doing non-essential flushes of deletion queue is subject to this timeout, after
261 : /// which we warn and skip.
262 : const DELETION_QUEUE_FLUSH_TIMEOUT: Duration = Duration::from_secs(10);
263 :
264 : pub enum MaybeDeletedIndexPart {
265 : IndexPart(IndexPart),
266 : Deleted(IndexPart),
267 : }
268 :
269 : #[derive(Debug, thiserror::Error)]
270 : pub enum PersistIndexPartWithDeletedFlagError {
271 : #[error("another task is already setting the deleted_flag, started at {0:?}")]
272 : AlreadyInProgress(NaiveDateTime),
273 : #[error("the deleted_flag was already set, value is {0:?}")]
274 : AlreadyDeleted(NaiveDateTime),
275 : #[error(transparent)]
276 : Other(#[from] anyhow::Error),
277 : }
278 :
279 : #[derive(Debug, thiserror::Error)]
280 : pub enum WaitCompletionError {
281 : #[error(transparent)]
282 : NotInitialized(NotInitialized),
283 : #[error("wait_completion aborted because upload queue was stopped")]
284 : UploadQueueShutDownOrStopped,
285 : }
286 :
287 : #[derive(Debug, thiserror::Error)]
288 : #[error("Upload queue either in unexpected state or hasn't downloaded manifest yet")]
289 : pub struct UploadQueueNotReadyError;
290 :
291 : #[derive(Debug, thiserror::Error)]
292 : pub enum ShutdownIfArchivedError {
293 : #[error(transparent)]
294 : NotInitialized(NotInitialized),
295 : #[error("timeline is not archived")]
296 : NotArchived,
297 : }
298 :
299 : /// Behavioral modes that enable seamless live migration.
300 : ///
301 : /// See docs/rfcs/028-pageserver-migration.md to understand how these fit in.
302 : struct RemoteTimelineClientConfig {
303 : /// If this is false, then update to remote_consistent_lsn are dropped rather
304 : /// than being submitted to DeletionQueue for validation. This behavior is
305 : /// used when a tenant attachment is known to have a stale generation number,
306 : /// such that validation attempts will always fail. This is not necessary
307 : /// for correctness, but avoids spamming error statistics with failed validations
308 : /// when doing migrations of tenants.
309 : process_remote_consistent_lsn_updates: bool,
310 :
311 : /// If this is true, then object deletions are held in a buffer in RemoteTimelineClient
312 : /// rather than being submitted to the DeletionQueue. This behavior is used when a tenant
313 : /// is known to be multi-attached, in order to avoid disrupting other attached tenants
314 : /// whose generations' metadata refers to the deleted objects.
315 : block_deletions: bool,
316 : }
317 :
318 : /// RemoteTimelineClientConfig's state is entirely driven by LocationConf, but we do
319 : /// not carry the entire LocationConf structure: it's much more than we need. The From
320 : /// impl extracts the subset of the LocationConf that is interesting to RemoteTimelineClient.
321 : impl From<&AttachedLocationConfig> for RemoteTimelineClientConfig {
322 924 : fn from(lc: &AttachedLocationConfig) -> Self {
323 924 : Self {
324 924 : block_deletions: !lc.may_delete_layers_hint(),
325 924 : process_remote_consistent_lsn_updates: lc.may_upload_layers_hint(),
326 924 : }
327 924 : }
328 : }
329 :
330 : /// A client for accessing a timeline's data in remote storage.
331 : ///
332 : /// This takes care of managing the number of connections, and balancing them
333 : /// across tenants. This also handles retries of failed uploads.
334 : ///
335 : /// Upload and delete requests are ordered so that before a deletion is
336 : /// performed, we wait for all preceding uploads to finish. This ensures sure
337 : /// that if you perform a compaction operation that reshuffles data in layer
338 : /// files, we don't have a transient state where the old files have already been
339 : /// deleted, but new files have not yet been uploaded.
340 : ///
341 : /// Similarly, this enforces an order between index-file uploads, and layer
342 : /// uploads. Before an index-file upload is performed, all preceding layer
343 : /// uploads must be finished.
344 : ///
345 : /// This also maintains a list of remote files, and automatically includes that
346 : /// in the index part file, whenever timeline metadata is uploaded.
347 : ///
348 : /// Downloads are not queued, they are performed immediately.
349 : pub(crate) struct RemoteTimelineClient {
350 : conf: &'static PageServerConf,
351 :
352 : runtime: tokio::runtime::Handle,
353 :
354 : tenant_shard_id: TenantShardId,
355 : timeline_id: TimelineId,
356 : generation: Generation,
357 :
358 : upload_queue: Mutex<UploadQueue>,
359 :
360 : pub(crate) metrics: Arc<RemoteTimelineClientMetrics>,
361 :
362 : storage_impl: GenericRemoteStorage,
363 :
364 : deletion_queue_client: DeletionQueueClient,
365 :
366 : /// Subset of tenant configuration used to control upload behaviors during migrations
367 : config: std::sync::RwLock<RemoteTimelineClientConfig>,
368 :
369 : cancel: CancellationToken,
370 : }
371 :
372 : impl Drop for RemoteTimelineClient {
373 40 : fn drop(&mut self) {
374 40 : debug!("dropping RemoteTimelineClient");
375 40 : }
376 : }
377 :
378 : impl RemoteTimelineClient {
379 : ///
380 : /// Create a remote storage client for given timeline
381 : ///
382 : /// Note: the caller must initialize the upload queue before any uploads can be scheduled,
383 : /// by calling init_upload_queue.
384 : ///
385 904 : pub(crate) fn new(
386 904 : remote_storage: GenericRemoteStorage,
387 904 : deletion_queue_client: DeletionQueueClient,
388 904 : conf: &'static PageServerConf,
389 904 : tenant_shard_id: TenantShardId,
390 904 : timeline_id: TimelineId,
391 904 : generation: Generation,
392 904 : location_conf: &AttachedLocationConfig,
393 904 : ) -> RemoteTimelineClient {
394 904 : RemoteTimelineClient {
395 904 : conf,
396 904 : runtime: if cfg!(test) {
397 : // remote_timeline_client.rs tests rely on current-thread runtime
398 904 : tokio::runtime::Handle::current()
399 : } else {
400 0 : BACKGROUND_RUNTIME.handle().clone()
401 : },
402 904 : tenant_shard_id,
403 904 : timeline_id,
404 904 : generation,
405 904 : storage_impl: remote_storage,
406 904 : deletion_queue_client,
407 904 : upload_queue: Mutex::new(UploadQueue::Uninitialized),
408 904 : metrics: Arc::new(RemoteTimelineClientMetrics::new(
409 904 : &tenant_shard_id,
410 904 : &timeline_id,
411 904 : )),
412 904 : config: std::sync::RwLock::new(RemoteTimelineClientConfig::from(location_conf)),
413 904 : cancel: CancellationToken::new(),
414 904 : }
415 904 : }
416 :
417 : /// Initialize the upload queue for a remote storage that already received
418 : /// an index file upload, i.e., it's not empty.
419 : /// The given `index_part` must be the one on the remote.
420 12 : pub fn init_upload_queue(&self, index_part: &IndexPart) -> anyhow::Result<()> {
421 12 : // Set the maximum number of inprogress tasks to the remote storage concurrency. There's
422 12 : // certainly no point in starting more upload tasks than this.
423 12 : let inprogress_limit = self
424 12 : .conf
425 12 : .remote_storage_config
426 12 : .as_ref()
427 12 : .map_or(0, |r| r.concurrency_limit());
428 12 : let mut upload_queue = self.upload_queue.lock().unwrap();
429 12 : upload_queue.initialize_with_current_remote_index_part(index_part, inprogress_limit)?;
430 12 : self.update_remote_physical_size_gauge(Some(index_part));
431 12 : info!(
432 0 : "initialized upload queue from remote index with {} layer files",
433 0 : index_part.layer_metadata.len()
434 : );
435 12 : Ok(())
436 12 : }
437 :
438 : /// Initialize the upload queue for the case where the remote storage is empty,
439 : /// i.e., it doesn't have an `IndexPart`.
440 : ///
441 : /// `rel_size_v2_status` needs to be carried over during branching, and that's why
442 : /// it's passed in here.
443 892 : pub fn init_upload_queue_for_empty_remote(
444 892 : &self,
445 892 : local_metadata: &TimelineMetadata,
446 892 : rel_size_v2_status: Option<RelSizeMigration>,
447 892 : ) -> anyhow::Result<()> {
448 892 : // Set the maximum number of inprogress tasks to the remote storage concurrency. There's
449 892 : // certainly no point in starting more upload tasks than this.
450 892 : let inprogress_limit = self
451 892 : .conf
452 892 : .remote_storage_config
453 892 : .as_ref()
454 892 : .map_or(0, |r| r.concurrency_limit());
455 892 : let mut upload_queue = self.upload_queue.lock().unwrap();
456 892 : let initialized_queue =
457 892 : upload_queue.initialize_empty_remote(local_metadata, inprogress_limit)?;
458 892 : initialized_queue.dirty.rel_size_migration = rel_size_v2_status;
459 892 : self.update_remote_physical_size_gauge(None);
460 892 : info!("initialized upload queue as empty");
461 892 : Ok(())
462 892 : }
463 :
464 : /// Initialize the queue in stopped state. Used in startup path
465 : /// to continue deletion operation interrupted by pageserver crash or restart.
466 0 : pub fn init_upload_queue_stopped_to_continue_deletion(
467 0 : &self,
468 0 : index_part: &IndexPart,
469 0 : ) -> anyhow::Result<()> {
470 : // FIXME: consider newtype for DeletedIndexPart.
471 0 : let deleted_at = index_part.deleted_at.ok_or(anyhow::anyhow!(
472 0 : "bug: it is responsibility of the caller to provide index part from MaybeDeletedIndexPart::Deleted"
473 0 : ))?;
474 0 : let inprogress_limit = self
475 0 : .conf
476 0 : .remote_storage_config
477 0 : .as_ref()
478 0 : .map_or(0, |r| r.concurrency_limit());
479 0 :
480 0 : let mut upload_queue = self.upload_queue.lock().unwrap();
481 0 : upload_queue.initialize_with_current_remote_index_part(index_part, inprogress_limit)?;
482 0 : self.update_remote_physical_size_gauge(Some(index_part));
483 0 : self.stop_impl(&mut upload_queue);
484 0 :
485 0 : upload_queue
486 0 : .stopped_mut()
487 0 : .expect("stopped above")
488 0 : .deleted_at = SetDeletedFlagProgress::Successful(deleted_at);
489 0 :
490 0 : Ok(())
491 0 : }
492 :
493 : /// Notify this client of a change to its parent tenant's config, as this may cause us to
494 : /// take action (unblocking deletions when transitioning from AttachedMulti to AttachedSingle)
495 0 : pub(super) fn update_config(&self, location_conf: &AttachedLocationConfig) {
496 0 : let new_conf = RemoteTimelineClientConfig::from(location_conf);
497 0 : let unblocked = !new_conf.block_deletions;
498 0 :
499 0 : // Update config before draining deletions, so that we don't race with more being
500 0 : // inserted. This can result in deletions happening our of order, but that does not
501 0 : // violate any invariants: deletions only need to be ordered relative to upload of the index
502 0 : // that dereferences the deleted objects, and we are not changing that order.
503 0 : *self.config.write().unwrap() = new_conf;
504 0 :
505 0 : if unblocked {
506 : // If we may now delete layers, drain any that were blocked in our old
507 : // configuration state
508 0 : let mut queue_locked = self.upload_queue.lock().unwrap();
509 :
510 0 : if let Ok(queue) = queue_locked.initialized_mut() {
511 0 : let blocked_deletions = std::mem::take(&mut queue.blocked_deletions);
512 0 : for d in blocked_deletions {
513 0 : if let Err(e) = self.deletion_queue_client.push_layers(
514 0 : self.tenant_shard_id,
515 0 : self.timeline_id,
516 0 : self.generation,
517 0 : d.layers,
518 0 : ) {
519 : // This could happen if the pageserver is shut down while a tenant
520 : // is transitioning from a deletion-blocked state: we will leak some
521 : // S3 objects in this case.
522 0 : warn!("Failed to drain blocked deletions: {}", e);
523 0 : break;
524 0 : }
525 : }
526 0 : }
527 0 : }
528 0 : }
529 :
530 : /// Returns `None` if nothing is yet uplodaded, `Some(disk_consistent_lsn)` otherwise.
531 0 : pub fn remote_consistent_lsn_projected(&self) -> Option<Lsn> {
532 0 : match &mut *self.upload_queue.lock().unwrap() {
533 0 : UploadQueue::Uninitialized => None,
534 0 : UploadQueue::Initialized(q) => q.get_last_remote_consistent_lsn_projected(),
535 0 : UploadQueue::Stopped(UploadQueueStopped::Uninitialized) => None,
536 0 : UploadQueue::Stopped(UploadQueueStopped::Deletable(q)) => q
537 0 : .upload_queue_for_deletion
538 0 : .get_last_remote_consistent_lsn_projected(),
539 : }
540 0 : }
541 :
542 0 : pub fn remote_consistent_lsn_visible(&self) -> Option<Lsn> {
543 0 : match &mut *self.upload_queue.lock().unwrap() {
544 0 : UploadQueue::Uninitialized => None,
545 0 : UploadQueue::Initialized(q) => Some(q.get_last_remote_consistent_lsn_visible()),
546 0 : UploadQueue::Stopped(UploadQueueStopped::Uninitialized) => None,
547 0 : UploadQueue::Stopped(UploadQueueStopped::Deletable(q)) => Some(
548 0 : q.upload_queue_for_deletion
549 0 : .get_last_remote_consistent_lsn_visible(),
550 0 : ),
551 : }
552 0 : }
553 :
554 : /// Returns true if this timeline was previously detached at this Lsn and the remote timeline
555 : /// client is currently initialized.
556 0 : pub(crate) fn is_previous_ancestor_lsn(&self, lsn: Lsn) -> bool {
557 0 : self.upload_queue
558 0 : .lock()
559 0 : .unwrap()
560 0 : .initialized_mut()
561 0 : .map(|uq| uq.clean.0.lineage.is_previous_ancestor_lsn(lsn))
562 0 : .unwrap_or(false)
563 0 : }
564 :
565 : /// Returns whether the timeline is archived.
566 : /// Return None if the remote index_part hasn't been downloaded yet.
567 0 : pub(crate) fn is_archived(&self) -> Option<bool> {
568 0 : self.upload_queue
569 0 : .lock()
570 0 : .unwrap()
571 0 : .initialized_mut()
572 0 : .map(|q| q.clean.0.archived_at.is_some())
573 0 : .ok()
574 0 : }
575 :
576 : /// Returns true if the timeline is invisible in synthetic size calculations.
577 0 : pub(crate) fn is_invisible(&self) -> Option<bool> {
578 0 : self.upload_queue
579 0 : .lock()
580 0 : .unwrap()
581 0 : .initialized_mut()
582 0 : .map(|q| q.clean.0.marked_invisible_at.is_some())
583 0 : .ok()
584 0 : }
585 :
586 : /// Returns `Ok(Some(timestamp))` if the timeline has been archived, `Ok(None)` if the timeline hasn't been archived.
587 : ///
588 : /// Return Err(_) if the remote index_part hasn't been downloaded yet, or the timeline hasn't been stopped yet.
589 4 : pub(crate) fn archived_at_stopped_queue(
590 4 : &self,
591 4 : ) -> Result<Option<NaiveDateTime>, UploadQueueNotReadyError> {
592 4 : self.upload_queue
593 4 : .lock()
594 4 : .unwrap()
595 4 : .stopped_mut()
596 4 : .map(|q| q.upload_queue_for_deletion.clean.0.archived_at)
597 4 : .map_err(|_| UploadQueueNotReadyError)
598 4 : }
599 :
600 3859 : fn update_remote_physical_size_gauge(&self, current_remote_index_part: Option<&IndexPart>) {
601 3859 : let size: u64 = if let Some(current_remote_index_part) = current_remote_index_part {
602 2967 : current_remote_index_part
603 2967 : .layer_metadata
604 2967 : .values()
605 35178 : .map(|ilmd| ilmd.file_size)
606 2967 : .sum()
607 : } else {
608 892 : 0
609 : };
610 3859 : self.metrics.remote_physical_size_gauge.set(size);
611 3859 : }
612 :
613 4 : pub fn get_remote_physical_size(&self) -> u64 {
614 4 : self.metrics.remote_physical_size_gauge.get()
615 4 : }
616 :
617 : //
618 : // Download operations.
619 : //
620 : // These don't use the per-timeline queue. They do use the global semaphore in
621 : // S3Bucket, to limit the total number of concurrent operations, though.
622 : //
623 :
624 : /// Download index file
625 40 : pub async fn download_index_file(
626 40 : &self,
627 40 : cancel: &CancellationToken,
628 40 : ) -> Result<MaybeDeletedIndexPart, DownloadError> {
629 40 : let _unfinished_gauge_guard = self.metrics.call_begin(
630 40 : &RemoteOpFileKind::Index,
631 40 : &RemoteOpKind::Download,
632 40 : crate::metrics::RemoteTimelineClientMetricsCallTrackSize::DontTrackSize {
633 40 : reason: "no need for a downloads gauge",
634 40 : },
635 40 : );
636 :
637 40 : let (index_part, index_generation, index_last_modified) = download::download_index_part(
638 40 : &self.storage_impl,
639 40 : &self.tenant_shard_id,
640 40 : &self.timeline_id,
641 40 : self.generation,
642 40 : cancel,
643 40 : )
644 40 : .measure_remote_op(
645 40 : RemoteOpFileKind::Index,
646 40 : RemoteOpKind::Download,
647 40 : Arc::clone(&self.metrics),
648 40 : )
649 40 : .await?;
650 :
651 : // Defense in depth: monotonicity of generation numbers is an important correctness guarantee, so when we see a very
652 : // old index, we do extra checks in case this is the result of backward time-travel of the generation number (e.g.
653 : // in case of a bug in the service that issues generation numbers). Indices are allowed to be old, but we expect that
654 : // when we load an old index we are loading the _latest_ index: if we are asked to load an old index and there is
655 : // also a newer index available, that is surprising.
656 : const INDEX_AGE_CHECKS_THRESHOLD: Duration = Duration::from_secs(14 * 24 * 3600);
657 40 : let index_age = index_last_modified.elapsed().unwrap_or_else(|e| {
658 0 : if e.duration() > Duration::from_secs(5) {
659 : // We only warn if the S3 clock and our local clock are >5s out: because this is a low resolution
660 : // timestamp, it is common to be out by at least 1 second.
661 0 : tracing::warn!("Index has modification time in the future: {e}");
662 0 : }
663 0 : Duration::ZERO
664 40 : });
665 40 : if index_age > INDEX_AGE_CHECKS_THRESHOLD {
666 0 : tracing::info!(
667 : ?index_generation,
668 0 : age = index_age.as_secs_f64(),
669 0 : "Loaded an old index, checking for other indices..."
670 : );
671 :
672 : // Find the highest-generation index
673 0 : let (_latest_index_part, latest_index_generation, latest_index_mtime) =
674 0 : download::download_index_part(
675 0 : &self.storage_impl,
676 0 : &self.tenant_shard_id,
677 0 : &self.timeline_id,
678 0 : Generation::MAX,
679 0 : cancel,
680 0 : )
681 0 : .await?;
682 :
683 0 : if latest_index_generation > index_generation {
684 : // Unexpected! Why are we loading such an old index if a more recent one exists?
685 : // We will refuse to proceed, as there is no reasonable scenario where this should happen, but
686 : // there _is_ a clear bug/corruption scenario where it would happen (controller sets the generation
687 : // backwards).
688 0 : tracing::error!(
689 : ?index_generation,
690 : ?latest_index_generation,
691 : ?latest_index_mtime,
692 0 : "Found a newer index while loading an old one"
693 : );
694 0 : return Err(DownloadError::Fatal(
695 0 : "Index age exceeds threshold and a newer index exists".into(),
696 0 : ));
697 0 : }
698 40 : }
699 :
700 40 : if index_part.deleted_at.is_some() {
701 0 : Ok(MaybeDeletedIndexPart::Deleted(index_part))
702 : } else {
703 40 : Ok(MaybeDeletedIndexPart::IndexPart(index_part))
704 : }
705 40 : }
706 :
707 : /// Download a (layer) file from `path`, into local filesystem.
708 : ///
709 : /// 'layer_metadata' is the metadata from the remote index file.
710 : ///
711 : /// On success, returns the size of the downloaded file.
712 28 : pub async fn download_layer_file(
713 28 : &self,
714 28 : layer_file_name: &LayerName,
715 28 : layer_metadata: &LayerFileMetadata,
716 28 : local_path: &Utf8Path,
717 28 : gate: &utils::sync::gate::Gate,
718 28 : cancel: &CancellationToken,
719 28 : ctx: &RequestContext,
720 28 : ) -> Result<u64, DownloadError> {
721 28 : let downloaded_size = {
722 28 : let _unfinished_gauge_guard = self.metrics.call_begin(
723 28 : &RemoteOpFileKind::Layer,
724 28 : &RemoteOpKind::Download,
725 28 : crate::metrics::RemoteTimelineClientMetricsCallTrackSize::DontTrackSize {
726 28 : reason: "no need for a downloads gauge",
727 28 : },
728 28 : );
729 28 : download::download_layer_file(
730 28 : self.conf,
731 28 : &self.storage_impl,
732 28 : self.tenant_shard_id,
733 28 : self.timeline_id,
734 28 : layer_file_name,
735 28 : layer_metadata,
736 28 : local_path,
737 28 : gate,
738 28 : cancel,
739 28 : ctx,
740 28 : )
741 28 : .measure_remote_op(
742 28 : RemoteOpFileKind::Layer,
743 28 : RemoteOpKind::Download,
744 28 : Arc::clone(&self.metrics),
745 28 : )
746 28 : .await?
747 : };
748 :
749 28 : REMOTE_ONDEMAND_DOWNLOADED_LAYERS.inc();
750 28 : REMOTE_ONDEMAND_DOWNLOADED_BYTES.inc_by(downloaded_size);
751 28 :
752 28 : Ok(downloaded_size)
753 28 : }
754 :
755 : //
756 : // Upload operations.
757 : //
758 :
759 : /// Launch an index-file upload operation in the background, with
760 : /// fully updated metadata.
761 : ///
762 : /// This should only be used to upload initial metadata to remote storage.
763 : ///
764 : /// The upload will be added to the queue immediately, but it
765 : /// won't be performed until all previously scheduled layer file
766 : /// upload operations have completed successfully. This is to
767 : /// ensure that when the index file claims that layers X, Y and Z
768 : /// exist in remote storage, they really do. To wait for the upload
769 : /// to complete, use `wait_completion`.
770 : ///
771 : /// If there were any changes to the list of files, i.e. if any
772 : /// layer file uploads were scheduled, since the last index file
773 : /// upload, those will be included too.
774 460 : pub fn schedule_index_upload_for_full_metadata_update(
775 460 : self: &Arc<Self>,
776 460 : metadata: &TimelineMetadata,
777 460 : ) -> anyhow::Result<()> {
778 460 : let mut guard = self.upload_queue.lock().unwrap();
779 460 : let upload_queue = guard.initialized_mut()?;
780 :
781 : // As documented in the struct definition, it's ok for latest_metadata to be
782 : // ahead of what's _actually_ on the remote during index upload.
783 460 : upload_queue.dirty.metadata = metadata.clone();
784 460 :
785 460 : self.schedule_index_upload(upload_queue);
786 460 :
787 460 : Ok(())
788 460 : }
789 :
790 : /// Launch an index-file upload operation in the background, with only parts of the metadata
791 : /// updated.
792 : ///
793 : /// This is the regular way of updating metadata on layer flushes or Gc.
794 : ///
795 : /// Using this lighter update mechanism allows for reparenting and detaching without changes to
796 : /// `index_part.json`, while being more clear on what values update regularly.
797 2456 : pub(crate) fn schedule_index_upload_for_metadata_update(
798 2456 : self: &Arc<Self>,
799 2456 : update: &MetadataUpdate,
800 2456 : ) -> anyhow::Result<()> {
801 2456 : let mut guard = self.upload_queue.lock().unwrap();
802 2456 : let upload_queue = guard.initialized_mut()?;
803 :
804 2456 : upload_queue.dirty.metadata.apply(update);
805 2456 :
806 2456 : // Defense in depth: if we somehow generated invalid metadata, do not persist it.
807 2456 : upload_queue
808 2456 : .dirty
809 2456 : .validate()
810 2456 : .map_err(|e| anyhow::anyhow!(e))?;
811 :
812 2456 : self.schedule_index_upload(upload_queue);
813 2456 :
814 2456 : Ok(())
815 2456 : }
816 :
817 : /// Launch an index-file upload operation in the background, with only the `archived_at` field updated.
818 : ///
819 : /// Returns whether it is required to wait for the queue to be empty to ensure that the change is uploaded,
820 : /// so either if the change is already sitting in the queue, but not commited yet, or the change has not
821 : /// been in the queue yet.
822 4 : pub(crate) fn schedule_index_upload_for_timeline_archival_state(
823 4 : self: &Arc<Self>,
824 4 : state: TimelineArchivalState,
825 4 : ) -> anyhow::Result<bool> {
826 4 : let mut guard = self.upload_queue.lock().unwrap();
827 4 : let upload_queue = guard.initialized_mut()?;
828 :
829 : /// Returns Some(_) if a change is needed, and Some(true) if it's a
830 : /// change needed to set archived_at.
831 8 : fn need_change(
832 8 : archived_at: &Option<NaiveDateTime>,
833 8 : state: TimelineArchivalState,
834 8 : ) -> Option<bool> {
835 8 : match (archived_at, state) {
836 : (Some(_), TimelineArchivalState::Archived)
837 : | (None, TimelineArchivalState::Unarchived) => {
838 : // Nothing to do
839 0 : tracing::info!("intended state matches present state");
840 0 : None
841 : }
842 8 : (None, TimelineArchivalState::Archived) => Some(true),
843 0 : (Some(_), TimelineArchivalState::Unarchived) => Some(false),
844 : }
845 8 : }
846 4 : let need_upload_scheduled = need_change(&upload_queue.dirty.archived_at, state);
847 :
848 4 : if let Some(archived_at_set) = need_upload_scheduled {
849 4 : let intended_archived_at = archived_at_set.then(|| Utc::now().naive_utc());
850 4 : upload_queue.dirty.archived_at = intended_archived_at;
851 4 : self.schedule_index_upload(upload_queue);
852 4 : }
853 :
854 4 : let need_wait = need_change(&upload_queue.clean.0.archived_at, state).is_some();
855 4 : Ok(need_wait)
856 4 : }
857 :
858 0 : pub(crate) fn schedule_index_upload_for_timeline_invisible_state(
859 0 : self: &Arc<Self>,
860 0 : state: TimelineVisibilityState,
861 0 : ) -> anyhow::Result<()> {
862 0 : let mut guard = self.upload_queue.lock().unwrap();
863 0 : let upload_queue = guard.initialized_mut()?;
864 :
865 0 : fn need_change(
866 0 : marked_invisible_at: &Option<NaiveDateTime>,
867 0 : state: TimelineVisibilityState,
868 0 : ) -> Option<bool> {
869 0 : match (marked_invisible_at, state) {
870 0 : (Some(_), TimelineVisibilityState::Invisible) => Some(false),
871 0 : (None, TimelineVisibilityState::Invisible) => Some(true),
872 0 : (Some(_), TimelineVisibilityState::Visible) => Some(false),
873 0 : (None, TimelineVisibilityState::Visible) => Some(true),
874 : }
875 0 : }
876 :
877 0 : let need_upload_scheduled = need_change(&upload_queue.dirty.marked_invisible_at, state);
878 :
879 0 : if let Some(marked_invisible_at_set) = need_upload_scheduled {
880 0 : let intended_marked_invisible_at =
881 0 : marked_invisible_at_set.then(|| Utc::now().naive_utc());
882 0 : upload_queue.dirty.marked_invisible_at = intended_marked_invisible_at;
883 0 : self.schedule_index_upload(upload_queue);
884 0 : }
885 :
886 0 : Ok(())
887 0 : }
888 :
889 : /// Shuts the timeline client down, but only if the timeline is archived.
890 : ///
891 : /// This function and [`Self::schedule_index_upload_for_timeline_archival_state`] use the
892 : /// same lock to prevent races between unarchival and offloading: unarchival requires the
893 : /// upload queue to be initialized, and leaves behind an upload queue where either dirty
894 : /// or clean has archived_at of `None`. offloading leaves behind an uninitialized upload
895 : /// queue.
896 4 : pub(crate) async fn shutdown_if_archived(
897 4 : self: &Arc<Self>,
898 4 : ) -> Result<(), ShutdownIfArchivedError> {
899 4 : {
900 4 : let mut guard = self.upload_queue.lock().unwrap();
901 4 : let upload_queue = guard
902 4 : .initialized_mut()
903 4 : .map_err(ShutdownIfArchivedError::NotInitialized)?;
904 :
905 4 : match (
906 4 : upload_queue.dirty.archived_at.is_none(),
907 4 : upload_queue.clean.0.archived_at.is_none(),
908 4 : ) {
909 : // The expected case: the timeline is archived and we don't want to unarchive
910 4 : (false, false) => {}
911 : (true, false) => {
912 0 : tracing::info!("can't shut down timeline: timeline slated for unarchival");
913 0 : return Err(ShutdownIfArchivedError::NotArchived);
914 : }
915 0 : (dirty_archived, true) => {
916 0 : tracing::info!(%dirty_archived, "can't shut down timeline: timeline not archived in remote storage");
917 0 : return Err(ShutdownIfArchivedError::NotArchived);
918 : }
919 : }
920 :
921 : // Set the shutting_down flag while the guard from the archival check is held.
922 : // This prevents a race with unarchival, as initialized_mut will not return
923 : // an upload queue from this point.
924 : // Also launch the queued tasks like shutdown() does.
925 4 : if !upload_queue.shutting_down {
926 4 : upload_queue.shutting_down = true;
927 4 : upload_queue.queued_operations.push_back(UploadOp::Shutdown);
928 4 : // this operation is not counted similar to Barrier
929 4 : self.launch_queued_tasks(upload_queue);
930 4 : }
931 : }
932 :
933 4 : self.shutdown().await;
934 :
935 4 : Ok(())
936 4 : }
937 :
938 : /// Launch an index-file upload operation in the background, setting `import_pgdata` field.
939 0 : pub(crate) fn schedule_index_upload_for_import_pgdata_state_update(
940 0 : self: &Arc<Self>,
941 0 : state: Option<import_pgdata::index_part_format::Root>,
942 0 : ) -> anyhow::Result<()> {
943 0 : let mut guard = self.upload_queue.lock().unwrap();
944 0 : let upload_queue = guard.initialized_mut()?;
945 0 : upload_queue.dirty.import_pgdata = state;
946 0 : self.schedule_index_upload(upload_queue);
947 0 : Ok(())
948 0 : }
949 :
950 : /// Launch an index-file upload operation in the background, setting `gc_compaction_state` field.
951 0 : pub(crate) fn schedule_index_upload_for_gc_compaction_state_update(
952 0 : self: &Arc<Self>,
953 0 : gc_compaction_state: GcCompactionState,
954 0 : ) -> anyhow::Result<()> {
955 0 : let mut guard = self.upload_queue.lock().unwrap();
956 0 : let upload_queue = guard.initialized_mut()?;
957 0 : upload_queue.dirty.gc_compaction = Some(gc_compaction_state);
958 0 : self.schedule_index_upload(upload_queue);
959 0 : Ok(())
960 0 : }
961 :
962 : /// Launch an index-file upload operation in the background, setting `rel_size_v2_status` field.
963 0 : pub(crate) fn schedule_index_upload_for_rel_size_v2_status_update(
964 0 : self: &Arc<Self>,
965 0 : rel_size_v2_status: RelSizeMigration,
966 0 : ) -> anyhow::Result<()> {
967 0 : let mut guard = self.upload_queue.lock().unwrap();
968 0 : let upload_queue = guard.initialized_mut()?;
969 0 : upload_queue.dirty.rel_size_migration = Some(rel_size_v2_status);
970 0 : // TODO: allow this operation to bypass the validation check because we might upload the index part
971 0 : // with no layers but the flag updated. For now, we just modify the index part in memory and the next
972 0 : // upload will include the flag.
973 0 : // self.schedule_index_upload(upload_queue);
974 0 : Ok(())
975 0 : }
976 :
977 : ///
978 : /// Launch an index-file upload operation in the background, if necessary.
979 : ///
980 : /// Use this function to schedule the update of the index file after
981 : /// scheduling file uploads or deletions. If no file uploads or deletions
982 : /// have been scheduled since the last index file upload, this does
983 : /// nothing.
984 : ///
985 : /// Like schedule_index_upload_for_metadata_update(), this merely adds
986 : /// the upload to the upload queue and returns quickly.
987 736 : pub fn schedule_index_upload_for_file_changes(self: &Arc<Self>) -> Result<(), NotInitialized> {
988 736 : let mut guard = self.upload_queue.lock().unwrap();
989 736 : let upload_queue = guard.initialized_mut()?;
990 :
991 736 : if upload_queue.latest_files_changes_since_metadata_upload_scheduled > 0 {
992 28 : self.schedule_index_upload(upload_queue);
993 708 : }
994 :
995 736 : Ok(())
996 736 : }
997 :
998 : /// Only used in the `patch_index_part` HTTP API to force trigger an index upload.
999 0 : pub fn force_schedule_index_upload(self: &Arc<Self>) -> Result<(), NotInitialized> {
1000 0 : let mut guard = self.upload_queue.lock().unwrap();
1001 0 : let upload_queue = guard.initialized_mut()?;
1002 0 : self.schedule_index_upload(upload_queue);
1003 0 : Ok(())
1004 0 : }
1005 :
1006 : /// Launch an index-file upload operation in the background (internal function)
1007 3092 : fn schedule_index_upload(self: &Arc<Self>, upload_queue: &mut UploadQueueInitialized) {
1008 3092 : let disk_consistent_lsn = upload_queue.dirty.metadata.disk_consistent_lsn();
1009 3092 : // fix up the duplicated field
1010 3092 : upload_queue.dirty.disk_consistent_lsn = disk_consistent_lsn;
1011 3092 :
1012 3092 : // make sure it serializes before doing it in perform_upload_task so that it doesn't
1013 3092 : // look like a retryable error
1014 3092 : let void = std::io::sink();
1015 3092 : serde_json::to_writer(void, &upload_queue.dirty).expect("serialize index_part.json");
1016 3092 :
1017 3092 : let index_part = &upload_queue.dirty;
1018 3092 :
1019 3092 : info!(
1020 0 : "scheduling metadata upload up to consistent LSN {disk_consistent_lsn} with {} files ({} changed)",
1021 0 : index_part.layer_metadata.len(),
1022 : upload_queue.latest_files_changes_since_metadata_upload_scheduled,
1023 : );
1024 :
1025 3092 : let op = UploadOp::UploadMetadata {
1026 3092 : uploaded: Box::new(index_part.clone()),
1027 3092 : };
1028 3092 : self.metric_begin(&op);
1029 3092 : upload_queue.queued_operations.push_back(op);
1030 3092 : upload_queue.latest_files_changes_since_metadata_upload_scheduled = 0;
1031 3092 :
1032 3092 : // Launch the task immediately, if possible
1033 3092 : self.launch_queued_tasks(upload_queue);
1034 3092 : }
1035 :
1036 : /// Reparent this timeline to a new parent.
1037 : ///
1038 : /// A retryable step of timeline ancestor detach.
1039 0 : pub(crate) async fn schedule_reparenting_and_wait(
1040 0 : self: &Arc<Self>,
1041 0 : new_parent: &TimelineId,
1042 0 : ) -> anyhow::Result<()> {
1043 0 : let receiver = {
1044 0 : let mut guard = self.upload_queue.lock().unwrap();
1045 0 : let upload_queue = guard.initialized_mut()?;
1046 :
1047 0 : let Some(prev) = upload_queue.dirty.metadata.ancestor_timeline() else {
1048 0 : return Err(anyhow::anyhow!(
1049 0 : "cannot reparent without a current ancestor"
1050 0 : ));
1051 : };
1052 :
1053 0 : let uploaded = &upload_queue.clean.0.metadata;
1054 0 :
1055 0 : if uploaded.ancestor_timeline().is_none() && !uploaded.ancestor_lsn().is_valid() {
1056 : // nothing to do
1057 0 : None
1058 : } else {
1059 0 : upload_queue.dirty.metadata.reparent(new_parent);
1060 0 : upload_queue.dirty.lineage.record_previous_ancestor(&prev);
1061 0 :
1062 0 : self.schedule_index_upload(upload_queue);
1063 0 :
1064 0 : Some(self.schedule_barrier0(upload_queue))
1065 : }
1066 : };
1067 :
1068 0 : if let Some(receiver) = receiver {
1069 0 : Self::wait_completion0(receiver).await?;
1070 0 : }
1071 0 : Ok(())
1072 0 : }
1073 :
1074 : /// Schedules uploading a new version of `index_part.json` with the given layers added,
1075 : /// detaching from ancestor and waits for it to complete.
1076 : ///
1077 : /// This is used with `Timeline::detach_ancestor` functionality.
1078 0 : pub(crate) async fn schedule_adding_existing_layers_to_index_detach_and_wait(
1079 0 : self: &Arc<Self>,
1080 0 : layers: &[Layer],
1081 0 : adopted: (TimelineId, Lsn),
1082 0 : ) -> anyhow::Result<()> {
1083 0 : let barrier = {
1084 0 : let mut guard = self.upload_queue.lock().unwrap();
1085 0 : let upload_queue = guard.initialized_mut()?;
1086 :
1087 0 : if upload_queue.clean.0.lineage.detached_previous_ancestor() == Some(adopted) {
1088 0 : None
1089 : } else {
1090 0 : upload_queue.dirty.metadata.detach_from_ancestor(&adopted);
1091 0 : upload_queue.dirty.lineage.record_detaching(&adopted);
1092 :
1093 0 : for layer in layers {
1094 0 : let prev = upload_queue
1095 0 : .dirty
1096 0 : .layer_metadata
1097 0 : .insert(layer.layer_desc().layer_name(), layer.metadata());
1098 0 : assert!(prev.is_none(), "copied layer existed already {layer}");
1099 : }
1100 :
1101 0 : self.schedule_index_upload(upload_queue);
1102 0 :
1103 0 : Some(self.schedule_barrier0(upload_queue))
1104 : }
1105 : };
1106 :
1107 0 : if let Some(barrier) = barrier {
1108 0 : Self::wait_completion0(barrier).await?;
1109 0 : }
1110 0 : Ok(())
1111 0 : }
1112 :
1113 : /// Adds a gc blocking reason for this timeline if one does not exist already.
1114 : ///
1115 : /// A retryable step of timeline detach ancestor.
1116 : ///
1117 : /// Returns a future which waits until the completion of the upload.
1118 0 : pub(crate) fn schedule_insert_gc_block_reason(
1119 0 : self: &Arc<Self>,
1120 0 : reason: index::GcBlockingReason,
1121 0 : ) -> Result<impl std::future::Future<Output = Result<(), WaitCompletionError>>, NotInitialized>
1122 0 : {
1123 0 : let maybe_barrier = {
1124 0 : let mut guard = self.upload_queue.lock().unwrap();
1125 0 : let upload_queue = guard.initialized_mut()?;
1126 :
1127 0 : if let index::GcBlockingReason::DetachAncestor = reason {
1128 0 : if upload_queue.dirty.metadata.ancestor_timeline().is_none() {
1129 0 : drop(guard);
1130 0 : panic!("cannot start detach ancestor if there is nothing to detach from");
1131 0 : }
1132 0 : }
1133 :
1134 0 : let wanted = |x: Option<&index::GcBlocking>| x.is_some_and(|x| x.blocked_by(reason));
1135 :
1136 0 : let current = upload_queue.dirty.gc_blocking.as_ref();
1137 0 : let uploaded = upload_queue.clean.0.gc_blocking.as_ref();
1138 0 :
1139 0 : match (current, uploaded) {
1140 0 : (x, y) if wanted(x) && wanted(y) => None,
1141 0 : (x, y) if wanted(x) && !wanted(y) => Some(self.schedule_barrier0(upload_queue)),
1142 : // Usual case: !wanted(x) && !wanted(y)
1143 : //
1144 : // Unusual: !wanted(x) && wanted(y) which means we have two processes waiting to
1145 : // turn on and off some reason.
1146 0 : (x, y) => {
1147 0 : if !wanted(x) && wanted(y) {
1148 : // this could be avoided by having external in-memory synchronization, like
1149 : // timeline detach ancestor
1150 0 : warn!(
1151 : ?reason,
1152 : op = "insert",
1153 0 : "unexpected: two racing processes to enable and disable a gc blocking reason"
1154 : );
1155 0 : }
1156 :
1157 : // at this point, the metadata must always show that there is a parent
1158 0 : upload_queue.dirty.gc_blocking = current
1159 0 : .map(|x| x.with_reason(reason))
1160 0 : .or_else(|| Some(index::GcBlocking::started_now_for(reason)));
1161 0 : self.schedule_index_upload(upload_queue);
1162 0 : Some(self.schedule_barrier0(upload_queue))
1163 : }
1164 : }
1165 : };
1166 :
1167 0 : Ok(async move {
1168 0 : if let Some(barrier) = maybe_barrier {
1169 0 : Self::wait_completion0(barrier).await?;
1170 0 : }
1171 0 : Ok(())
1172 0 : })
1173 0 : }
1174 :
1175 : /// Removes a gc blocking reason for this timeline if one exists.
1176 : ///
1177 : /// A retryable step of timeline detach ancestor.
1178 : ///
1179 : /// Returns a future which waits until the completion of the upload.
1180 0 : pub(crate) fn schedule_remove_gc_block_reason(
1181 0 : self: &Arc<Self>,
1182 0 : reason: index::GcBlockingReason,
1183 0 : ) -> Result<impl std::future::Future<Output = Result<(), WaitCompletionError>>, NotInitialized>
1184 0 : {
1185 0 : let maybe_barrier = {
1186 0 : let mut guard = self.upload_queue.lock().unwrap();
1187 0 : let upload_queue = guard.initialized_mut()?;
1188 :
1189 0 : if let index::GcBlockingReason::DetachAncestor = reason {
1190 0 : if !upload_queue.clean.0.lineage.is_detached_from_ancestor() {
1191 0 : drop(guard);
1192 0 : panic!("cannot complete timeline_ancestor_detach while not detached");
1193 0 : }
1194 0 : }
1195 :
1196 0 : let wanted = |x: Option<&index::GcBlocking>| {
1197 0 : x.is_none() || x.is_some_and(|b| !b.blocked_by(reason))
1198 0 : };
1199 :
1200 0 : let current = upload_queue.dirty.gc_blocking.as_ref();
1201 0 : let uploaded = upload_queue.clean.0.gc_blocking.as_ref();
1202 0 :
1203 0 : match (current, uploaded) {
1204 0 : (x, y) if wanted(x) && wanted(y) => None,
1205 0 : (x, y) if wanted(x) && !wanted(y) => Some(self.schedule_barrier0(upload_queue)),
1206 0 : (x, y) => {
1207 0 : if !wanted(x) && wanted(y) {
1208 0 : warn!(
1209 : ?reason,
1210 : op = "remove",
1211 0 : "unexpected: two racing processes to enable and disable a gc blocking reason (remove)"
1212 : );
1213 0 : }
1214 :
1215 0 : upload_queue.dirty.gc_blocking =
1216 0 : current.as_ref().and_then(|x| x.without_reason(reason));
1217 0 : assert!(wanted(upload_queue.dirty.gc_blocking.as_ref()));
1218 0 : self.schedule_index_upload(upload_queue);
1219 0 : Some(self.schedule_barrier0(upload_queue))
1220 : }
1221 : }
1222 : };
1223 :
1224 0 : Ok(async move {
1225 0 : if let Some(barrier) = maybe_barrier {
1226 0 : Self::wait_completion0(barrier).await?;
1227 0 : }
1228 0 : Ok(())
1229 0 : })
1230 0 : }
1231 :
1232 : /// Launch an upload operation in the background; the file is added to be included in next
1233 : /// `index_part.json` upload.
1234 2760 : pub(crate) fn schedule_layer_file_upload(
1235 2760 : self: &Arc<Self>,
1236 2760 : layer: ResidentLayer,
1237 2760 : ) -> Result<(), NotInitialized> {
1238 2760 : let mut guard = self.upload_queue.lock().unwrap();
1239 2760 : let upload_queue = guard.initialized_mut()?;
1240 :
1241 2760 : self.schedule_layer_file_upload0(upload_queue, layer);
1242 2760 : self.launch_queued_tasks(upload_queue);
1243 2760 : Ok(())
1244 2760 : }
1245 :
1246 3484 : fn schedule_layer_file_upload0(
1247 3484 : self: &Arc<Self>,
1248 3484 : upload_queue: &mut UploadQueueInitialized,
1249 3484 : layer: ResidentLayer,
1250 3484 : ) {
1251 3484 : let metadata = layer.metadata();
1252 3484 :
1253 3484 : upload_queue
1254 3484 : .dirty
1255 3484 : .layer_metadata
1256 3484 : .insert(layer.layer_desc().layer_name(), metadata.clone());
1257 3484 : upload_queue.latest_files_changes_since_metadata_upload_scheduled += 1;
1258 3484 :
1259 3484 : info!(
1260 : gen=?metadata.generation,
1261 : shard=?metadata.shard,
1262 0 : "scheduled layer file upload {layer}",
1263 : );
1264 :
1265 3484 : let op = UploadOp::UploadLayer(layer, metadata, None);
1266 3484 : self.metric_begin(&op);
1267 3484 : upload_queue.queued_operations.push_back(op);
1268 3484 : }
1269 :
1270 : /// Launch a delete operation in the background.
1271 : ///
1272 : /// The operation does not modify local filesystem state.
1273 : ///
1274 : /// Note: This schedules an index file upload before the deletions. The
1275 : /// deletion won't actually be performed, until all previously scheduled
1276 : /// upload operations, and the index file upload, have completed
1277 : /// successfully.
1278 16 : pub fn schedule_layer_file_deletion(
1279 16 : self: &Arc<Self>,
1280 16 : names: &[LayerName],
1281 16 : ) -> anyhow::Result<()> {
1282 16 : let mut guard = self.upload_queue.lock().unwrap();
1283 16 : let upload_queue = guard.initialized_mut()?;
1284 :
1285 16 : let with_metadata =
1286 16 : self.schedule_unlinking_of_layers_from_index_part0(upload_queue, names.iter().cloned());
1287 16 :
1288 16 : self.schedule_deletion_of_unlinked0(upload_queue, with_metadata);
1289 16 :
1290 16 : // Launch the tasks immediately, if possible
1291 16 : self.launch_queued_tasks(upload_queue);
1292 16 : Ok(())
1293 16 : }
1294 :
1295 : /// Unlinks the layer files from `index_part.json` but does not yet schedule deletion for the
1296 : /// layer files, leaving them dangling.
1297 : ///
1298 : /// The files will be leaked in remote storage unless [`Self::schedule_deletion_of_unlinked`]
1299 : /// is invoked on them.
1300 8 : pub(crate) fn schedule_gc_update(
1301 8 : self: &Arc<Self>,
1302 8 : gc_layers: &[Layer],
1303 8 : ) -> Result<(), NotInitialized> {
1304 8 : let mut guard = self.upload_queue.lock().unwrap();
1305 8 : let upload_queue = guard.initialized_mut()?;
1306 :
1307 : // just forget the return value; after uploading the next index_part.json, we can consider
1308 : // the layer files as "dangling". this is fine, at worst case we create work for the
1309 : // scrubber.
1310 :
1311 8 : let names = gc_layers.iter().map(|x| x.layer_desc().layer_name());
1312 8 :
1313 8 : self.schedule_unlinking_of_layers_from_index_part0(upload_queue, names);
1314 8 :
1315 8 : self.launch_queued_tasks(upload_queue);
1316 8 :
1317 8 : Ok(())
1318 8 : }
1319 :
1320 : /// Update the remote index file, removing the to-be-deleted files from the index,
1321 : /// allowing scheduling of actual deletions later.
1322 176 : fn schedule_unlinking_of_layers_from_index_part0<I>(
1323 176 : self: &Arc<Self>,
1324 176 : upload_queue: &mut UploadQueueInitialized,
1325 176 : names: I,
1326 176 : ) -> Vec<(LayerName, LayerFileMetadata)>
1327 176 : where
1328 176 : I: IntoIterator<Item = LayerName>,
1329 176 : {
1330 176 : // Decorate our list of names with each name's metadata, dropping
1331 176 : // names that are unexpectedly missing from our metadata. This metadata
1332 176 : // is later used when physically deleting layers, to construct key paths.
1333 176 : let with_metadata: Vec<_> = names
1334 176 : .into_iter()
1335 1024 : .filter_map(|name| {
1336 1024 : let meta = upload_queue.dirty.layer_metadata.remove(&name);
1337 :
1338 1024 : if let Some(meta) = meta {
1339 1024 : upload_queue.latest_files_changes_since_metadata_upload_scheduled += 1;
1340 1024 : Some((name, meta))
1341 : } else {
1342 : // This can only happen if we forgot to to schedule the file upload
1343 : // before scheduling the delete. Log it because it is a rare/strange
1344 : // situation, and in case something is misbehaving, we'd like to know which
1345 : // layers experienced this.
1346 0 : info!("Deleting layer {name} not found in latest_files list, never uploaded?");
1347 0 : None
1348 : }
1349 1024 : })
1350 176 : .collect();
1351 :
1352 : #[cfg(feature = "testing")]
1353 1200 : for (name, metadata) in &with_metadata {
1354 1024 : let gen_ = metadata.generation;
1355 1024 : if let Some(unexpected) = upload_queue.dangling_files.insert(name.to_owned(), gen_) {
1356 0 : if unexpected == gen_ {
1357 0 : tracing::error!("{name} was unlinked twice with same generation");
1358 : } else {
1359 0 : tracing::error!(
1360 0 : "{name} was unlinked twice with different generations {gen_:?} and {unexpected:?}"
1361 : );
1362 : }
1363 1024 : }
1364 : }
1365 :
1366 : // after unlinking files from the upload_queue.latest_files we must always schedule an
1367 : // index_part update, because that needs to be uploaded before we can actually delete the
1368 : // files.
1369 176 : if upload_queue.latest_files_changes_since_metadata_upload_scheduled > 0 {
1370 144 : self.schedule_index_upload(upload_queue);
1371 144 : }
1372 :
1373 176 : with_metadata
1374 176 : }
1375 :
1376 : /// Schedules deletion for layer files which have previously been unlinked from the
1377 : /// `index_part.json` with [`Self::schedule_gc_update`] or [`Self::schedule_compaction_update`].
1378 1025 : pub(crate) fn schedule_deletion_of_unlinked(
1379 1025 : self: &Arc<Self>,
1380 1025 : layers: Vec<(LayerName, LayerFileMetadata)>,
1381 1025 : ) -> anyhow::Result<()> {
1382 1025 : let mut guard = self.upload_queue.lock().unwrap();
1383 1025 : let upload_queue = guard.initialized_mut()?;
1384 :
1385 1025 : self.schedule_deletion_of_unlinked0(upload_queue, layers);
1386 1025 : self.launch_queued_tasks(upload_queue);
1387 1025 : Ok(())
1388 1025 : }
1389 :
1390 1037 : fn schedule_deletion_of_unlinked0(
1391 1037 : self: &Arc<Self>,
1392 1037 : upload_queue: &mut UploadQueueInitialized,
1393 1037 : mut with_metadata: Vec<(LayerName, LayerFileMetadata)>,
1394 1037 : ) {
1395 1037 : // Filter out any layers which were not created by this tenant shard. These are
1396 1037 : // layers that originate from some ancestor shard after a split, and may still
1397 1037 : // be referenced by other shards. We are free to delete them locally and remove
1398 1037 : // them from our index (and would have already done so when we reach this point
1399 1037 : // in the code), but we may not delete them remotely.
1400 1037 : with_metadata.retain(|(name, meta)| {
1401 1025 : let retain = meta.shard.shard_number == self.tenant_shard_id.shard_number
1402 1025 : && meta.shard.shard_count == self.tenant_shard_id.shard_count;
1403 1025 : if !retain {
1404 0 : tracing::debug!(
1405 0 : "Skipping deletion of ancestor-shard layer {name}, from shard {}",
1406 : meta.shard
1407 : );
1408 1025 : }
1409 1025 : retain
1410 1037 : });
1411 :
1412 2062 : for (name, meta) in &with_metadata {
1413 1025 : info!(
1414 0 : "scheduling deletion of layer {}{} (shard {})",
1415 0 : name,
1416 0 : meta.generation.get_suffix(),
1417 : meta.shard
1418 : );
1419 : }
1420 :
1421 : #[cfg(feature = "testing")]
1422 2062 : for (name, meta) in &with_metadata {
1423 1025 : let gen_ = meta.generation;
1424 1025 : match upload_queue.dangling_files.remove(name) {
1425 1017 : Some(same) if same == gen_ => { /* expected */ }
1426 0 : Some(other) => {
1427 0 : tracing::error!("{name} was unlinked with {other:?} but deleted with {gen_:?}");
1428 : }
1429 : None => {
1430 8 : tracing::error!("{name} was unlinked but was not dangling");
1431 : }
1432 : }
1433 : }
1434 :
1435 : // schedule the actual deletions
1436 1037 : if with_metadata.is_empty() {
1437 : // avoid scheduling the op & bumping the metric
1438 12 : return;
1439 1025 : }
1440 1025 : let op = UploadOp::Delete(Delete {
1441 1025 : layers: with_metadata,
1442 1025 : });
1443 1025 : self.metric_begin(&op);
1444 1025 : upload_queue.queued_operations.push_back(op);
1445 1037 : }
1446 :
1447 : /// Schedules a compaction update to the remote `index_part.json`.
1448 : ///
1449 : /// `compacted_from` represent the L0 names which have been `compacted_to` L1 layers.
1450 152 : pub(crate) fn schedule_compaction_update(
1451 152 : self: &Arc<Self>,
1452 152 : compacted_from: &[Layer],
1453 152 : compacted_to: &[ResidentLayer],
1454 152 : ) -> Result<(), NotInitialized> {
1455 152 : let mut guard = self.upload_queue.lock().unwrap();
1456 152 : let upload_queue = guard.initialized_mut()?;
1457 :
1458 876 : for layer in compacted_to {
1459 724 : self.schedule_layer_file_upload0(upload_queue, layer.clone());
1460 724 : }
1461 :
1462 1012 : let names = compacted_from.iter().map(|x| x.layer_desc().layer_name());
1463 152 :
1464 152 : self.schedule_unlinking_of_layers_from_index_part0(upload_queue, names);
1465 152 : self.launch_queued_tasks(upload_queue);
1466 152 :
1467 152 : Ok(())
1468 152 : }
1469 :
1470 : /// Wait for all previously scheduled uploads/deletions to complete
1471 440 : pub(crate) async fn wait_completion(self: &Arc<Self>) -> Result<(), WaitCompletionError> {
1472 440 : let receiver = {
1473 440 : let mut guard = self.upload_queue.lock().unwrap();
1474 440 : let upload_queue = guard
1475 440 : .initialized_mut()
1476 440 : .map_err(WaitCompletionError::NotInitialized)?;
1477 440 : self.schedule_barrier0(upload_queue)
1478 440 : };
1479 440 :
1480 440 : Self::wait_completion0(receiver).await
1481 440 : }
1482 :
1483 440 : async fn wait_completion0(
1484 440 : mut receiver: tokio::sync::watch::Receiver<()>,
1485 440 : ) -> Result<(), WaitCompletionError> {
1486 440 : if receiver.changed().await.is_err() {
1487 0 : return Err(WaitCompletionError::UploadQueueShutDownOrStopped);
1488 440 : }
1489 440 :
1490 440 : Ok(())
1491 440 : }
1492 :
1493 12 : pub(crate) fn schedule_barrier(self: &Arc<Self>) -> anyhow::Result<()> {
1494 12 : let mut guard = self.upload_queue.lock().unwrap();
1495 12 : let upload_queue = guard.initialized_mut()?;
1496 12 : self.schedule_barrier0(upload_queue);
1497 12 : Ok(())
1498 12 : }
1499 :
1500 452 : fn schedule_barrier0(
1501 452 : self: &Arc<Self>,
1502 452 : upload_queue: &mut UploadQueueInitialized,
1503 452 : ) -> tokio::sync::watch::Receiver<()> {
1504 452 : let (sender, receiver) = tokio::sync::watch::channel(());
1505 452 : let barrier_op = UploadOp::Barrier(sender);
1506 452 :
1507 452 : upload_queue.queued_operations.push_back(barrier_op);
1508 452 : // Don't count this kind of operation!
1509 452 :
1510 452 : // Launch the task immediately, if possible
1511 452 : self.launch_queued_tasks(upload_queue);
1512 452 :
1513 452 : receiver
1514 452 : }
1515 :
1516 : /// Wait for all previously scheduled operations to complete, and then stop.
1517 : ///
1518 : /// Not cancellation safe
1519 20 : pub(crate) async fn shutdown(self: &Arc<Self>) {
1520 20 : // On cancellation the queue is left in ackward state of refusing new operations but
1521 20 : // proper stop is yet to be called. On cancel the original or some later task must call
1522 20 : // `stop` or `shutdown`.
1523 20 : let sg = scopeguard::guard((), |_| {
1524 0 : tracing::error!(
1525 0 : "RemoteTimelineClient::shutdown was cancelled; this should not happen, do not make this into an allowed_error"
1526 : )
1527 20 : });
1528 :
1529 16 : let fut = {
1530 20 : let mut guard = self.upload_queue.lock().unwrap();
1531 20 : let upload_queue = match &mut *guard {
1532 : UploadQueue::Stopped(_) => {
1533 4 : scopeguard::ScopeGuard::into_inner(sg);
1534 4 : return;
1535 : }
1536 : UploadQueue::Uninitialized => {
1537 : // transition into Stopped state
1538 0 : self.stop_impl(&mut guard);
1539 0 : scopeguard::ScopeGuard::into_inner(sg);
1540 0 : return;
1541 : }
1542 16 : UploadQueue::Initialized(init) => init,
1543 16 : };
1544 16 :
1545 16 : // if the queue is already stuck due to a shutdown operation which was cancelled, then
1546 16 : // just don't add more of these as they would never complete.
1547 16 : //
1548 16 : // TODO: if launch_queued_tasks were to be refactored to accept a &mut UploadQueue
1549 16 : // in every place we would not have to jump through this hoop, and this method could be
1550 16 : // made cancellable.
1551 16 : if !upload_queue.shutting_down {
1552 12 : upload_queue.shutting_down = true;
1553 12 : upload_queue.queued_operations.push_back(UploadOp::Shutdown);
1554 12 : // this operation is not counted similar to Barrier
1555 12 :
1556 12 : self.launch_queued_tasks(upload_queue);
1557 12 : }
1558 :
1559 16 : upload_queue.shutdown_ready.clone().acquire_owned()
1560 : };
1561 :
1562 16 : let res = fut.await;
1563 :
1564 16 : scopeguard::ScopeGuard::into_inner(sg);
1565 16 :
1566 16 : match res {
1567 0 : Ok(_permit) => unreachable!("shutdown_ready should not have been added permits"),
1568 16 : Err(_closed) => {
1569 16 : // expected
1570 16 : }
1571 16 : }
1572 16 :
1573 16 : self.stop();
1574 20 : }
1575 :
1576 : /// Set the deleted_at field in the remote index file.
1577 : ///
1578 : /// This fails if the upload queue has not been `stop()`ed.
1579 : ///
1580 : /// The caller is responsible for calling `stop()` AND for waiting
1581 : /// for any ongoing upload tasks to finish after `stop()` has succeeded.
1582 : /// Check method [`RemoteTimelineClient::stop`] for details.
1583 : #[instrument(skip_all)]
1584 : pub(crate) async fn persist_index_part_with_deleted_flag(
1585 : self: &Arc<Self>,
1586 : ) -> Result<(), PersistIndexPartWithDeletedFlagError> {
1587 : let index_part_with_deleted_at = {
1588 : let mut locked = self.upload_queue.lock().unwrap();
1589 :
1590 : // We must be in stopped state because otherwise
1591 : // we can have inprogress index part upload that can overwrite the file
1592 : // with missing is_deleted flag that we going to set below
1593 : let stopped = locked.stopped_mut()?;
1594 :
1595 : match stopped.deleted_at {
1596 : SetDeletedFlagProgress::NotRunning => (), // proceed
1597 : SetDeletedFlagProgress::InProgress(at) => {
1598 : return Err(PersistIndexPartWithDeletedFlagError::AlreadyInProgress(at));
1599 : }
1600 : SetDeletedFlagProgress::Successful(at) => {
1601 : return Err(PersistIndexPartWithDeletedFlagError::AlreadyDeleted(at));
1602 : }
1603 : };
1604 : let deleted_at = Utc::now().naive_utc();
1605 : stopped.deleted_at = SetDeletedFlagProgress::InProgress(deleted_at);
1606 :
1607 : let mut index_part = stopped.upload_queue_for_deletion.dirty.clone();
1608 : index_part.deleted_at = Some(deleted_at);
1609 : index_part
1610 : };
1611 :
1612 0 : let undo_deleted_at = scopeguard::guard(Arc::clone(self), |self_clone| {
1613 0 : let mut locked = self_clone.upload_queue.lock().unwrap();
1614 0 : let stopped = locked
1615 0 : .stopped_mut()
1616 0 : .expect("there's no way out of Stopping, and we checked it's Stopping above");
1617 0 : stopped.deleted_at = SetDeletedFlagProgress::NotRunning;
1618 0 : });
1619 :
1620 : pausable_failpoint!("persist_deleted_index_part");
1621 :
1622 : backoff::retry(
1623 0 : || {
1624 0 : upload::upload_index_part(
1625 0 : &self.storage_impl,
1626 0 : &self.tenant_shard_id,
1627 0 : &self.timeline_id,
1628 0 : self.generation,
1629 0 : &index_part_with_deleted_at,
1630 0 : &self.cancel,
1631 0 : )
1632 0 : },
1633 0 : |_e| false,
1634 : 1,
1635 : // have just a couple of attempts
1636 : // when executed as part of timeline deletion this happens in context of api call
1637 : // when executed as part of tenant deletion this happens in the background
1638 : 2,
1639 : "persist_index_part_with_deleted_flag",
1640 : &self.cancel,
1641 : )
1642 : .await
1643 0 : .ok_or_else(|| anyhow::Error::new(TimeoutOrCancel::Cancel))
1644 0 : .and_then(|x| x)?;
1645 :
1646 : // all good, disarm the guard and mark as success
1647 : ScopeGuard::into_inner(undo_deleted_at);
1648 : {
1649 : let mut locked = self.upload_queue.lock().unwrap();
1650 :
1651 : let stopped = locked
1652 : .stopped_mut()
1653 : .expect("there's no way out of Stopping, and we checked it's Stopping above");
1654 : stopped.deleted_at = SetDeletedFlagProgress::Successful(
1655 : index_part_with_deleted_at
1656 : .deleted_at
1657 : .expect("we set it above"),
1658 : );
1659 : }
1660 :
1661 : Ok(())
1662 : }
1663 :
1664 0 : pub(crate) fn is_deleting(&self) -> bool {
1665 0 : let mut locked = self.upload_queue.lock().unwrap();
1666 0 : locked.stopped_mut().is_ok()
1667 0 : }
1668 :
1669 0 : pub(crate) async fn preserve_initdb_archive(
1670 0 : self: &Arc<Self>,
1671 0 : tenant_id: &TenantId,
1672 0 : timeline_id: &TimelineId,
1673 0 : cancel: &CancellationToken,
1674 0 : ) -> anyhow::Result<()> {
1675 0 : backoff::retry(
1676 0 : || async {
1677 0 : upload::preserve_initdb_archive(&self.storage_impl, tenant_id, timeline_id, cancel)
1678 0 : .await
1679 0 : },
1680 0 : TimeoutOrCancel::caused_by_cancel,
1681 0 : FAILED_DOWNLOAD_WARN_THRESHOLD,
1682 0 : FAILED_REMOTE_OP_RETRIES,
1683 0 : "preserve_initdb_tar_zst",
1684 0 : &cancel.clone(),
1685 0 : )
1686 0 : .await
1687 0 : .ok_or_else(|| anyhow::Error::new(TimeoutOrCancel::Cancel))
1688 0 : .and_then(|x| x)
1689 0 : .context("backing up initdb archive")?;
1690 0 : Ok(())
1691 0 : }
1692 :
1693 : /// Uploads the given layer **without** adding it to be part of a future `index_part.json` upload.
1694 : ///
1695 : /// This is not normally needed.
1696 0 : pub(crate) async fn upload_layer_file(
1697 0 : self: &Arc<Self>,
1698 0 : uploaded: &ResidentLayer,
1699 0 : cancel: &CancellationToken,
1700 0 : ) -> anyhow::Result<()> {
1701 0 : let remote_path = remote_layer_path(
1702 0 : &self.tenant_shard_id.tenant_id,
1703 0 : &self.timeline_id,
1704 0 : uploaded.metadata().shard,
1705 0 : &uploaded.layer_desc().layer_name(),
1706 0 : uploaded.metadata().generation,
1707 0 : );
1708 0 :
1709 0 : backoff::retry(
1710 0 : || async {
1711 0 : upload::upload_timeline_layer(
1712 0 : &self.storage_impl,
1713 0 : uploaded.local_path(),
1714 0 : &remote_path,
1715 0 : uploaded.metadata().file_size,
1716 0 : cancel,
1717 0 : )
1718 0 : .await
1719 0 : },
1720 0 : TimeoutOrCancel::caused_by_cancel,
1721 0 : FAILED_UPLOAD_WARN_THRESHOLD,
1722 0 : FAILED_REMOTE_OP_RETRIES,
1723 0 : "upload a layer without adding it to latest files",
1724 0 : cancel,
1725 0 : )
1726 0 : .await
1727 0 : .ok_or_else(|| anyhow::Error::new(TimeoutOrCancel::Cancel))
1728 0 : .and_then(|x| x)
1729 0 : .context("upload a layer without adding it to latest files")
1730 0 : }
1731 :
1732 : /// Copies the `adopted` remote existing layer to the remote path of `adopted_as`. The layer is
1733 : /// not added to be part of a future `index_part.json` upload.
1734 0 : pub(crate) async fn copy_timeline_layer(
1735 0 : self: &Arc<Self>,
1736 0 : adopted: &Layer,
1737 0 : adopted_as: &Layer,
1738 0 : cancel: &CancellationToken,
1739 0 : ) -> anyhow::Result<()> {
1740 0 : let source_remote_path = remote_layer_path(
1741 0 : &self.tenant_shard_id.tenant_id,
1742 0 : &adopted
1743 0 : .get_timeline_id()
1744 0 : .expect("Source timeline should be alive"),
1745 0 : adopted.metadata().shard,
1746 0 : &adopted.layer_desc().layer_name(),
1747 0 : adopted.metadata().generation,
1748 0 : );
1749 0 :
1750 0 : let target_remote_path = remote_layer_path(
1751 0 : &self.tenant_shard_id.tenant_id,
1752 0 : &self.timeline_id,
1753 0 : adopted_as.metadata().shard,
1754 0 : &adopted_as.layer_desc().layer_name(),
1755 0 : adopted_as.metadata().generation,
1756 0 : );
1757 0 :
1758 0 : backoff::retry(
1759 0 : || async {
1760 0 : upload::copy_timeline_layer(
1761 0 : &self.storage_impl,
1762 0 : &source_remote_path,
1763 0 : &target_remote_path,
1764 0 : cancel,
1765 0 : )
1766 0 : .await
1767 0 : },
1768 0 : TimeoutOrCancel::caused_by_cancel,
1769 0 : FAILED_UPLOAD_WARN_THRESHOLD,
1770 0 : FAILED_REMOTE_OP_RETRIES,
1771 0 : "copy timeline layer",
1772 0 : cancel,
1773 0 : )
1774 0 : .await
1775 0 : .ok_or_else(|| anyhow::Error::new(TimeoutOrCancel::Cancel))
1776 0 : .and_then(|x| x)
1777 0 : .context("remote copy timeline layer")
1778 0 : }
1779 :
1780 0 : async fn flush_deletion_queue(&self) -> Result<(), DeletionQueueError> {
1781 0 : match tokio::time::timeout(
1782 0 : DELETION_QUEUE_FLUSH_TIMEOUT,
1783 0 : self.deletion_queue_client.flush_immediate(),
1784 0 : )
1785 0 : .await
1786 : {
1787 0 : Ok(result) => result,
1788 0 : Err(_timeout) => {
1789 0 : // Flushing remote deletions is not mandatory: we flush here to make the system easier to test, and
1790 0 : // to ensure that _usually_ objects are really gone after a DELETE is acked. However, in case of deletion
1791 0 : // queue issues (https://github.com/neondatabase/neon/issues/6440), we don't want to wait indefinitely here.
1792 0 : tracing::warn!(
1793 0 : "Timed out waiting for deletion queue flush, acking deletion anyway"
1794 : );
1795 0 : Ok(())
1796 : }
1797 : }
1798 0 : }
1799 :
1800 : /// Prerequisites: UploadQueue should be in stopped state and deleted_at should be successfuly set.
1801 : /// The function deletes layer files one by one, then lists the prefix to see if we leaked something
1802 : /// deletes leaked files if any and proceeds with deletion of index file at the end.
1803 0 : pub(crate) async fn delete_all(self: &Arc<Self>) -> Result<(), DeleteTimelineError> {
1804 0 : debug_assert_current_span_has_tenant_and_timeline_id();
1805 :
1806 0 : let layers: Vec<RemotePath> = {
1807 0 : let mut locked = self.upload_queue.lock().unwrap();
1808 0 : let stopped = locked.stopped_mut().map_err(DeleteTimelineError::Other)?;
1809 :
1810 0 : if !matches!(stopped.deleted_at, SetDeletedFlagProgress::Successful(_)) {
1811 0 : return Err(DeleteTimelineError::Other(anyhow::anyhow!(
1812 0 : "deleted_at is not set"
1813 0 : )));
1814 0 : }
1815 0 :
1816 0 : debug_assert!(stopped.upload_queue_for_deletion.no_pending_work());
1817 :
1818 0 : stopped
1819 0 : .upload_queue_for_deletion
1820 0 : .dirty
1821 0 : .layer_metadata
1822 0 : .drain()
1823 0 : .filter(|(_file_name, meta)| {
1824 0 : // Filter out layers that belonged to an ancestor shard. Since we are deleting the whole timeline from
1825 0 : // all shards anyway, we _could_ delete these, but
1826 0 : // - it creates a potential race if other shards are still
1827 0 : // using the layers while this shard deletes them.
1828 0 : // - it means that if we rolled back the shard split, the ancestor shards would be in a state where
1829 0 : // these timelines are present but corrupt (their index exists but some layers don't)
1830 0 : //
1831 0 : // These layers will eventually be cleaned up by the scrubber when it does physical GC.
1832 0 : meta.shard.shard_number == self.tenant_shard_id.shard_number
1833 0 : && meta.shard.shard_count == self.tenant_shard_id.shard_count
1834 0 : })
1835 0 : .map(|(file_name, meta)| {
1836 0 : remote_layer_path(
1837 0 : &self.tenant_shard_id.tenant_id,
1838 0 : &self.timeline_id,
1839 0 : meta.shard,
1840 0 : &file_name,
1841 0 : meta.generation,
1842 0 : )
1843 0 : })
1844 0 : .collect()
1845 0 : };
1846 0 :
1847 0 : let layer_deletion_count = layers.len();
1848 0 : self.deletion_queue_client
1849 0 : .push_immediate(layers)
1850 0 : .await
1851 0 : .map_err(|_| DeleteTimelineError::Cancelled)?;
1852 :
1853 : // Delete the initdb.tar.zst, which is not always present, but deletion attempts of
1854 : // inexistant objects are not considered errors.
1855 0 : let initdb_path =
1856 0 : remote_initdb_archive_path(&self.tenant_shard_id.tenant_id, &self.timeline_id);
1857 0 : self.deletion_queue_client
1858 0 : .push_immediate(vec![initdb_path])
1859 0 : .await
1860 0 : .map_err(|_| DeleteTimelineError::Cancelled)?;
1861 :
1862 : // Do not delete index part yet, it is needed for possible retry. If we remove it first
1863 : // and retry will arrive to different pageserver there wont be any traces of it on remote storage
1864 0 : let timeline_storage_path = remote_timeline_path(&self.tenant_shard_id, &self.timeline_id);
1865 0 :
1866 0 : // Execute all pending deletions, so that when we proceed to do a listing below, we aren't
1867 0 : // taking the burden of listing all the layers that we already know we should delete.
1868 0 : self.flush_deletion_queue()
1869 0 : .await
1870 0 : .map_err(|_| DeleteTimelineError::Cancelled)?;
1871 :
1872 0 : let cancel = shutdown_token();
1873 :
1874 0 : let remaining = download_retry(
1875 0 : || async {
1876 0 : self.storage_impl
1877 0 : .list(
1878 0 : Some(&timeline_storage_path),
1879 0 : ListingMode::NoDelimiter,
1880 0 : None,
1881 0 : &cancel,
1882 0 : )
1883 0 : .await
1884 0 : },
1885 0 : "list remaining files",
1886 0 : &cancel,
1887 0 : )
1888 0 : .await
1889 0 : .context("list files remaining files")?
1890 : .keys;
1891 :
1892 : // We will delete the current index_part object last, since it acts as a deletion
1893 : // marker via its deleted_at attribute
1894 0 : let latest_index = remaining
1895 0 : .iter()
1896 0 : .filter(|o| {
1897 0 : o.key
1898 0 : .object_name()
1899 0 : .map(|n| n.starts_with(IndexPart::FILE_NAME))
1900 0 : .unwrap_or(false)
1901 0 : })
1902 0 : .filter_map(|o| {
1903 0 : parse_remote_index_path(o.key.clone()).map(|gen_| (o.key.clone(), gen_))
1904 0 : })
1905 0 : .max_by_key(|i| i.1)
1906 0 : .map(|i| i.0.clone())
1907 0 : .unwrap_or(
1908 0 : // No generation-suffixed indices, assume we are dealing with
1909 0 : // a legacy index.
1910 0 : remote_index_path(&self.tenant_shard_id, &self.timeline_id, Generation::none()),
1911 0 : );
1912 0 :
1913 0 : let remaining_layers: Vec<RemotePath> = remaining
1914 0 : .into_iter()
1915 0 : .filter_map(|o| {
1916 0 : if o.key == latest_index || o.key.object_name() == Some(INITDB_PRESERVED_PATH) {
1917 0 : None
1918 : } else {
1919 0 : Some(o.key)
1920 : }
1921 0 : })
1922 0 : .inspect(|path| {
1923 0 : if let Some(name) = path.object_name() {
1924 0 : info!(%name, "deleting a file not referenced from index_part.json");
1925 : } else {
1926 0 : warn!(%path, "deleting a nameless or non-utf8 object not referenced from index_part.json");
1927 : }
1928 0 : })
1929 0 : .collect();
1930 0 :
1931 0 : let not_referenced_count = remaining_layers.len();
1932 0 : if !remaining_layers.is_empty() {
1933 0 : self.deletion_queue_client
1934 0 : .push_immediate(remaining_layers)
1935 0 : .await
1936 0 : .map_err(|_| DeleteTimelineError::Cancelled)?;
1937 0 : }
1938 :
1939 0 : fail::fail_point!("timeline-delete-before-index-delete", |_| {
1940 0 : Err(DeleteTimelineError::Other(anyhow::anyhow!(
1941 0 : "failpoint: timeline-delete-before-index-delete"
1942 0 : )))?
1943 0 : });
1944 :
1945 0 : debug!("enqueuing index part deletion");
1946 0 : self.deletion_queue_client
1947 0 : .push_immediate([latest_index].to_vec())
1948 0 : .await
1949 0 : .map_err(|_| DeleteTimelineError::Cancelled)?;
1950 :
1951 : // Timeline deletion is rare and we have probably emitted a reasonably number of objects: wait
1952 : // for a flush to a persistent deletion list so that we may be sure deletion will occur.
1953 0 : self.flush_deletion_queue()
1954 0 : .await
1955 0 : .map_err(|_| DeleteTimelineError::Cancelled)?;
1956 :
1957 0 : fail::fail_point!("timeline-delete-after-index-delete", |_| {
1958 0 : Err(DeleteTimelineError::Other(anyhow::anyhow!(
1959 0 : "failpoint: timeline-delete-after-index-delete"
1960 0 : )))?
1961 0 : });
1962 :
1963 0 : info!(prefix=%timeline_storage_path, referenced=layer_deletion_count, not_referenced=%not_referenced_count, "done deleting in timeline prefix, including index_part.json");
1964 :
1965 0 : Ok(())
1966 0 : }
1967 :
1968 : /// Pick next tasks from the queue, and start as many of them as possible without violating
1969 : /// the ordering constraints.
1970 : ///
1971 : /// TODO: consider limiting the number of in-progress tasks, beyond what remote_storage does.
1972 : /// This can launch an unbounded number of queued tasks. `UploadQueue::next_ready()` also has
1973 : /// worst-case quadratic cost in the number of tasks, and may struggle beyond 10,000 tasks.
1974 14200 : fn launch_queued_tasks(self: &Arc<Self>, upload_queue: &mut UploadQueueInitialized) {
1975 21782 : while let Some((mut next_op, coalesced_ops)) = upload_queue.next_ready() {
1976 7582 : debug!("starting op: {next_op}");
1977 :
1978 : // Prepare upload.
1979 7582 : match &mut next_op {
1980 3484 : UploadOp::UploadLayer(layer, meta, mode) => {
1981 3484 : if upload_queue
1982 3484 : .recently_deleted
1983 3484 : .remove(&(layer.layer_desc().layer_name().clone(), meta.generation))
1984 0 : {
1985 0 : *mode = Some(OpType::FlushDeletion);
1986 0 : } else {
1987 3484 : *mode = Some(OpType::MayReorder)
1988 : }
1989 : }
1990 3005 : UploadOp::UploadMetadata { .. } => {}
1991 641 : UploadOp::Delete(Delete { layers }) => {
1992 1282 : for (name, meta) in layers {
1993 641 : upload_queue
1994 641 : .recently_deleted
1995 641 : .insert((name.clone(), meta.generation));
1996 641 : }
1997 : }
1998 452 : UploadOp::Barrier(sender) => {
1999 452 : sender.send_replace(());
2000 452 : continue;
2001 : }
2002 0 : UploadOp::Shutdown => unreachable!("shutdown is intentionally never popped off"),
2003 : };
2004 :
2005 : // Assign unique ID to this task
2006 7130 : upload_queue.task_counter += 1;
2007 7130 : let upload_task_id = upload_queue.task_counter;
2008 7130 :
2009 7130 : // Add it to the in-progress map
2010 7130 : let task = Arc::new(UploadTask {
2011 7130 : task_id: upload_task_id,
2012 7130 : op: next_op,
2013 7130 : coalesced_ops,
2014 7130 : retries: AtomicU32::new(0),
2015 7130 : });
2016 7130 : upload_queue
2017 7130 : .inprogress_tasks
2018 7130 : .insert(task.task_id, Arc::clone(&task));
2019 7130 :
2020 7130 : // Spawn task to perform the task
2021 7130 : let self_rc = Arc::clone(self);
2022 7130 : let tenant_shard_id = self.tenant_shard_id;
2023 7130 : let timeline_id = self.timeline_id;
2024 7130 : task_mgr::spawn(
2025 7130 : &self.runtime,
2026 7130 : TaskKind::RemoteUploadTask,
2027 7130 : self.tenant_shard_id,
2028 7130 : Some(self.timeline_id),
2029 7130 : "remote upload",
2030 6974 : async move {
2031 6974 : self_rc.perform_upload_task(task).await;
2032 6683 : Ok(())
2033 6683 : }
2034 7130 : .instrument(info_span!(parent: None, "remote_upload", tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), %timeline_id, %upload_task_id)),
2035 : );
2036 :
2037 : // Loop back to process next task
2038 : }
2039 14200 : }
2040 :
2041 : ///
2042 : /// Perform an upload task.
2043 : ///
2044 : /// The task is in the `inprogress_tasks` list. This function will try to
2045 : /// execute it, retrying forever. On successful completion, the task is
2046 : /// removed it from the `inprogress_tasks` list, and any next task(s) in the
2047 : /// queue that were waiting by the completion are launched.
2048 : ///
2049 : /// The task can be shut down, however. That leads to stopping the whole
2050 : /// queue.
2051 : ///
2052 6974 : async fn perform_upload_task(self: &Arc<Self>, task: Arc<UploadTask>) {
2053 6974 : let cancel = shutdown_token();
2054 : // Loop to retry until it completes.
2055 : loop {
2056 : // If we're requested to shut down, close up shop and exit.
2057 : //
2058 : // Note: We only check for the shutdown requests between retries, so
2059 : // if a shutdown request arrives while we're busy uploading, in the
2060 : // upload::upload:*() call below, we will wait not exit until it has
2061 : // finished. We probably could cancel the upload by simply dropping
2062 : // the Future, but we're not 100% sure if the remote storage library
2063 : // is cancellation safe, so we don't dare to do that. Hopefully, the
2064 : // upload finishes or times out soon enough.
2065 6974 : if cancel.is_cancelled() {
2066 0 : info!("upload task cancelled by shutdown request");
2067 0 : self.stop();
2068 0 : return;
2069 6974 : }
2070 6974 :
2071 6974 : // Assert that we don't modify a layer that's referenced by the current index.
2072 6974 : if cfg!(debug_assertions) {
2073 6974 : let modified = match &task.op {
2074 3354 : UploadOp::UploadLayer(layer, layer_metadata, _) => {
2075 3354 : vec![(layer.layer_desc().layer_name(), layer_metadata)]
2076 : }
2077 641 : UploadOp::Delete(delete) => {
2078 641 : delete.layers.iter().map(|(n, m)| (n.clone(), m)).collect()
2079 : }
2080 : // These don't modify layers.
2081 2979 : UploadOp::UploadMetadata { .. } => Vec::new(),
2082 0 : UploadOp::Barrier(_) => Vec::new(),
2083 0 : UploadOp::Shutdown => Vec::new(),
2084 : };
2085 6974 : if let Ok(queue) = self.upload_queue.lock().unwrap().initialized_mut() {
2086 10957 : for (ref name, metadata) in modified {
2087 3995 : debug_assert!(
2088 3995 : !queue.clean.0.references(name, metadata),
2089 8 : "layer {name} modified while referenced by index",
2090 : );
2091 : }
2092 4 : }
2093 0 : }
2094 :
2095 6966 : let upload_result: anyhow::Result<()> = match &task.op {
2096 3354 : UploadOp::UploadLayer(layer, layer_metadata, mode) => {
2097 : // TODO: check if this mechanism can be removed now that can_bypass() performs
2098 : // conflict checks during scheduling.
2099 3354 : if let Some(OpType::FlushDeletion) = mode {
2100 0 : if self.config.read().unwrap().block_deletions {
2101 : // Of course, this is not efficient... but usually the queue should be empty.
2102 0 : let mut queue_locked = self.upload_queue.lock().unwrap();
2103 0 : let mut detected = false;
2104 0 : if let Ok(queue) = queue_locked.initialized_mut() {
2105 0 : for list in queue.blocked_deletions.iter_mut() {
2106 0 : list.layers.retain(|(name, meta)| {
2107 0 : if name == &layer.layer_desc().layer_name()
2108 0 : && meta.generation == layer_metadata.generation
2109 : {
2110 0 : detected = true;
2111 0 : // remove the layer from deletion queue
2112 0 : false
2113 : } else {
2114 : // keep the layer
2115 0 : true
2116 : }
2117 0 : });
2118 0 : }
2119 0 : }
2120 0 : if detected {
2121 0 : info!(
2122 0 : "cancelled blocked deletion of layer {} at gen {:?}",
2123 0 : layer.layer_desc().layer_name(),
2124 : layer_metadata.generation
2125 : );
2126 0 : }
2127 : } else {
2128 : // TODO: we did not guarantee that upload task starts after deletion task, so there could be possibly race conditions
2129 : // that we still get the layer deleted. But this only happens if someone creates a layer immediately after it's deleted,
2130 : // which is not possible in the current system.
2131 0 : info!(
2132 0 : "waiting for deletion queue flush to complete before uploading layer {} at gen {:?}",
2133 0 : layer.layer_desc().layer_name(),
2134 : layer_metadata.generation
2135 : );
2136 : {
2137 : // We are going to flush, we can clean up the recently deleted list.
2138 0 : let mut queue_locked = self.upload_queue.lock().unwrap();
2139 0 : if let Ok(queue) = queue_locked.initialized_mut() {
2140 0 : queue.recently_deleted.clear();
2141 0 : }
2142 : }
2143 0 : if let Err(e) = self.deletion_queue_client.flush_execute().await {
2144 0 : warn!(
2145 0 : "failed to flush the deletion queue before uploading layer {} at gen {:?}, still proceeding to upload: {e:#} ",
2146 0 : layer.layer_desc().layer_name(),
2147 : layer_metadata.generation
2148 : );
2149 : } else {
2150 0 : info!(
2151 0 : "done flushing deletion queue before uploading layer {} at gen {:?}",
2152 0 : layer.layer_desc().layer_name(),
2153 : layer_metadata.generation
2154 : );
2155 : }
2156 : }
2157 3354 : }
2158 3354 : let local_path = layer.local_path();
2159 3354 :
2160 3354 : // We should only be uploading layers created by this `Tenant`'s lifetime, so
2161 3354 : // the metadata in the upload should always match our current generation.
2162 3354 : assert_eq!(layer_metadata.generation, self.generation);
2163 :
2164 3354 : let remote_path = remote_layer_path(
2165 3354 : &self.tenant_shard_id.tenant_id,
2166 3354 : &self.timeline_id,
2167 3354 : layer_metadata.shard,
2168 3354 : &layer.layer_desc().layer_name(),
2169 3354 : layer_metadata.generation,
2170 3354 : );
2171 3354 :
2172 3354 : upload::upload_timeline_layer(
2173 3354 : &self.storage_impl,
2174 3354 : local_path,
2175 3354 : &remote_path,
2176 3354 : layer_metadata.file_size,
2177 3354 : &self.cancel,
2178 3354 : )
2179 3354 : .measure_remote_op(
2180 3354 : RemoteOpFileKind::Layer,
2181 3354 : RemoteOpKind::Upload,
2182 3354 : Arc::clone(&self.metrics),
2183 3354 : )
2184 3354 : .await
2185 : }
2186 2979 : UploadOp::UploadMetadata { uploaded } => {
2187 2979 : let res = upload::upload_index_part(
2188 2979 : &self.storage_impl,
2189 2979 : &self.tenant_shard_id,
2190 2979 : &self.timeline_id,
2191 2979 : self.generation,
2192 2979 : uploaded,
2193 2979 : &self.cancel,
2194 2979 : )
2195 2979 : .measure_remote_op(
2196 2979 : RemoteOpFileKind::Index,
2197 2979 : RemoteOpKind::Upload,
2198 2979 : Arc::clone(&self.metrics),
2199 2979 : )
2200 2979 : .await;
2201 2955 : if res.is_ok() {
2202 2955 : self.update_remote_physical_size_gauge(Some(uploaded));
2203 2955 : let mention_having_future_layers = if cfg!(feature = "testing") {
2204 2955 : uploaded
2205 2955 : .layer_metadata
2206 2955 : .keys()
2207 34744 : .any(|x| x.is_in_future(uploaded.metadata.disk_consistent_lsn()))
2208 : } else {
2209 0 : false
2210 : };
2211 2955 : if mention_having_future_layers {
2212 : // find rationale near crate::tenant::timeline::init::cleanup_future_layer
2213 143 : tracing::info!(
2214 0 : disk_consistent_lsn = %uploaded.metadata.disk_consistent_lsn(),
2215 0 : "uploaded an index_part.json with future layers -- this is ok! if shutdown now, expect future layer cleanup"
2216 : );
2217 2812 : }
2218 0 : }
2219 2955 : res
2220 : }
2221 633 : UploadOp::Delete(delete) => {
2222 633 : if self.config.read().unwrap().block_deletions {
2223 0 : let mut queue_locked = self.upload_queue.lock().unwrap();
2224 0 : if let Ok(queue) = queue_locked.initialized_mut() {
2225 0 : queue.blocked_deletions.push(delete.clone());
2226 0 : }
2227 0 : Ok(())
2228 : } else {
2229 633 : pausable_failpoint!("before-delete-layer-pausable");
2230 632 : self.deletion_queue_client
2231 632 : .push_layers(
2232 632 : self.tenant_shard_id,
2233 632 : self.timeline_id,
2234 632 : self.generation,
2235 632 : delete.layers.clone(),
2236 632 : )
2237 632 : .map_err(|e| anyhow::anyhow!(e))
2238 : }
2239 : }
2240 0 : unexpected @ UploadOp::Barrier(_) | unexpected @ UploadOp::Shutdown => {
2241 : // unreachable. Barrier operations are handled synchronously in
2242 : // launch_queued_tasks
2243 0 : warn!("unexpected {unexpected:?} operation in perform_upload_task");
2244 0 : break;
2245 : }
2246 : };
2247 :
2248 0 : match upload_result {
2249 : Ok(()) => {
2250 6683 : break;
2251 : }
2252 0 : Err(e) if TimeoutOrCancel::caused_by_cancel(&e) => {
2253 0 : // loop around to do the proper stopping
2254 0 : continue;
2255 : }
2256 0 : Err(e) => {
2257 0 : let retries = task.retries.fetch_add(1, Ordering::SeqCst);
2258 0 :
2259 0 : // Uploads can fail due to rate limits (IAM, S3), spurious network problems,
2260 0 : // or other external reasons. Such issues are relatively regular, so log them
2261 0 : // at info level at first, and only WARN if the operation fails repeatedly.
2262 0 : //
2263 0 : // (See similar logic for downloads in `download::download_retry`)
2264 0 : if retries < FAILED_UPLOAD_WARN_THRESHOLD {
2265 0 : info!(
2266 0 : "failed to perform remote task {}, will retry (attempt {}): {:#}",
2267 0 : task.op, retries, e
2268 : );
2269 : } else {
2270 0 : warn!(
2271 0 : "failed to perform remote task {}, will retry (attempt {}): {:?}",
2272 0 : task.op, retries, e
2273 : );
2274 : }
2275 :
2276 : // sleep until it's time to retry, or we're cancelled
2277 0 : exponential_backoff(
2278 0 : retries,
2279 0 : DEFAULT_BASE_BACKOFF_SECONDS,
2280 0 : DEFAULT_MAX_BACKOFF_SECONDS,
2281 0 : &cancel,
2282 0 : )
2283 0 : .await;
2284 : }
2285 : }
2286 : }
2287 :
2288 6683 : let retries = task.retries.load(Ordering::SeqCst);
2289 6683 : if retries > 0 {
2290 0 : info!(
2291 0 : "remote task {} completed successfully after {} retries",
2292 0 : task.op, retries
2293 : );
2294 : } else {
2295 6683 : debug!("remote task {} completed successfully", task.op);
2296 : }
2297 :
2298 : // The task has completed successfully. Remove it from the in-progress list.
2299 6683 : let lsn_update = {
2300 6683 : let mut upload_queue_guard = self.upload_queue.lock().unwrap();
2301 6683 : let upload_queue = match upload_queue_guard.deref_mut() {
2302 0 : UploadQueue::Uninitialized => panic!(
2303 0 : "callers are responsible for ensuring this is only called on an initialized queue"
2304 0 : ),
2305 0 : UploadQueue::Stopped(_stopped) => None,
2306 6683 : UploadQueue::Initialized(qi) => Some(qi),
2307 : };
2308 :
2309 6683 : let upload_queue = match upload_queue {
2310 6683 : Some(upload_queue) => upload_queue,
2311 : None => {
2312 0 : info!("another concurrent task already stopped the queue");
2313 0 : return;
2314 : }
2315 : };
2316 :
2317 6683 : upload_queue.inprogress_tasks.remove(&task.task_id);
2318 :
2319 6683 : let lsn_update = match task.op {
2320 3096 : UploadOp::UploadLayer(_, _, _) => None,
2321 2955 : UploadOp::UploadMetadata { ref uploaded } => {
2322 2955 : // the task id is reused as a monotonicity check for storing the "clean"
2323 2955 : // IndexPart.
2324 2955 : let last_updater = upload_queue.clean.1;
2325 2955 : let is_later = last_updater.is_some_and(|task_id| task_id < task.task_id);
2326 2955 : let monotone = is_later || last_updater.is_none();
2327 :
2328 2955 : assert!(
2329 2955 : monotone,
2330 0 : "no two index uploads should be completing at the same time, prev={last_updater:?}, task.task_id={}",
2331 0 : task.task_id
2332 : );
2333 :
2334 : // not taking ownership is wasteful
2335 2955 : upload_queue.clean.0.clone_from(uploaded);
2336 2955 : upload_queue.clean.1 = Some(task.task_id);
2337 2955 :
2338 2955 : let lsn = upload_queue.clean.0.metadata.disk_consistent_lsn();
2339 2955 : self.metrics
2340 2955 : .projected_remote_consistent_lsn_gauge
2341 2955 : .set(lsn.0);
2342 2955 :
2343 2955 : if self.generation.is_none() {
2344 : // Legacy mode: skip validating generation
2345 0 : upload_queue.visible_remote_consistent_lsn.store(lsn);
2346 0 : None
2347 2955 : } else if self
2348 2955 : .config
2349 2955 : .read()
2350 2955 : .unwrap()
2351 2955 : .process_remote_consistent_lsn_updates
2352 : {
2353 2955 : Some((lsn, upload_queue.visible_remote_consistent_lsn.clone()))
2354 : } else {
2355 : // Our config disables remote_consistent_lsn updates: drop it.
2356 0 : None
2357 : }
2358 : }
2359 632 : UploadOp::Delete(_) => None,
2360 0 : UploadOp::Barrier(..) | UploadOp::Shutdown => unreachable!(),
2361 : };
2362 :
2363 : // Launch any queued tasks that were unblocked by this one.
2364 6683 : self.launch_queued_tasks(upload_queue);
2365 6683 : lsn_update
2366 : };
2367 :
2368 6683 : if let Some((lsn, slot)) = lsn_update {
2369 : // Updates to the remote_consistent_lsn we advertise to pageservers
2370 : // are all routed through the DeletionQueue, to enforce important
2371 : // data safety guarantees (see docs/rfcs/025-generation-numbers.md)
2372 2955 : self.deletion_queue_client
2373 2955 : .update_remote_consistent_lsn(
2374 2955 : self.tenant_shard_id,
2375 2955 : self.timeline_id,
2376 2955 : self.generation,
2377 2955 : lsn,
2378 2955 : slot,
2379 2955 : )
2380 2955 : .await;
2381 3728 : }
2382 :
2383 6683 : self.metric_end(&task.op);
2384 6683 : for coalesced_op in &task.coalesced_ops {
2385 11 : self.metric_end(coalesced_op);
2386 11 : }
2387 6683 : }
2388 :
2389 14311 : fn metric_impl(
2390 14311 : &self,
2391 14311 : op: &UploadOp,
2392 14311 : ) -> Option<(
2393 14311 : RemoteOpFileKind,
2394 14311 : RemoteOpKind,
2395 14311 : RemoteTimelineClientMetricsCallTrackSize,
2396 14311 : )> {
2397 : use RemoteTimelineClientMetricsCallTrackSize::DontTrackSize;
2398 14311 : let res = match op {
2399 6580 : UploadOp::UploadLayer(_, m, _) => (
2400 6580 : RemoteOpFileKind::Layer,
2401 6580 : RemoteOpKind::Upload,
2402 6580 : RemoteTimelineClientMetricsCallTrackSize::Bytes(m.file_size),
2403 6580 : ),
2404 6058 : UploadOp::UploadMetadata { .. } => (
2405 6058 : RemoteOpFileKind::Index,
2406 6058 : RemoteOpKind::Upload,
2407 6058 : DontTrackSize {
2408 6058 : reason: "metadata uploads are tiny",
2409 6058 : },
2410 6058 : ),
2411 1657 : UploadOp::Delete(_delete) => (
2412 1657 : RemoteOpFileKind::Layer,
2413 1657 : RemoteOpKind::Delete,
2414 1657 : DontTrackSize {
2415 1657 : reason: "should we track deletes? positive or negative sign?",
2416 1657 : },
2417 1657 : ),
2418 : UploadOp::Barrier(..) | UploadOp::Shutdown => {
2419 : // we do not account these
2420 16 : return None;
2421 : }
2422 : };
2423 14295 : Some(res)
2424 14311 : }
2425 :
2426 7601 : fn metric_begin(&self, op: &UploadOp) {
2427 7601 : let (file_kind, op_kind, track_bytes) = match self.metric_impl(op) {
2428 7601 : Some(x) => x,
2429 0 : None => return,
2430 : };
2431 7601 : let guard = self.metrics.call_begin(&file_kind, &op_kind, track_bytes);
2432 7601 : guard.will_decrement_manually(); // in metric_end(), see right below
2433 7601 : }
2434 :
2435 6710 : fn metric_end(&self, op: &UploadOp) {
2436 6710 : let (file_kind, op_kind, track_bytes) = match self.metric_impl(op) {
2437 6694 : Some(x) => x,
2438 16 : None => return,
2439 : };
2440 6694 : self.metrics.call_end(&file_kind, &op_kind, track_bytes);
2441 6710 : }
2442 :
2443 : /// Close the upload queue for new operations and cancel queued operations.
2444 : ///
2445 : /// Use [`RemoteTimelineClient::shutdown`] for graceful stop.
2446 : ///
2447 : /// In-progress operations will still be running after this function returns.
2448 : /// Use `task_mgr::shutdown_tasks(Some(TaskKind::RemoteUploadTask), Some(self.tenant_shard_id), Some(timeline_id))`
2449 : /// to wait for them to complete, after calling this function.
2450 36 : pub(crate) fn stop(&self) {
2451 36 : // Whichever *task* for this RemoteTimelineClient grabs the mutex first will transition the queue
2452 36 : // into stopped state, thereby dropping all off the queued *ops* which haven't become *tasks* yet.
2453 36 : // The other *tasks* will come here and observe an already shut down queue and hence simply wrap up their business.
2454 36 : let mut guard = self.upload_queue.lock().unwrap();
2455 36 : self.stop_impl(&mut guard);
2456 36 : }
2457 :
2458 36 : fn stop_impl(&self, guard: &mut std::sync::MutexGuard<UploadQueue>) {
2459 36 : match &mut **guard {
2460 : UploadQueue::Uninitialized => {
2461 0 : info!("UploadQueue is in state Uninitialized, nothing to do");
2462 0 : **guard = UploadQueue::Stopped(UploadQueueStopped::Uninitialized);
2463 : }
2464 : UploadQueue::Stopped(_) => {
2465 : // nothing to do
2466 16 : info!("another concurrent task already shut down the queue");
2467 : }
2468 20 : UploadQueue::Initialized(initialized) => {
2469 20 : info!("shutting down upload queue");
2470 :
2471 : // Replace the queue with the Stopped state, taking ownership of the old
2472 : // Initialized queue. We will do some checks on it, and then drop it.
2473 20 : let qi = {
2474 : // Here we preserve working version of the upload queue for possible use during deletions.
2475 : // In-place replace of Initialized to Stopped can be done with the help of https://github.com/Sgeo/take_mut
2476 : // but for this use case it doesnt really makes sense to bring unsafe code only for this usage point.
2477 : // Deletion is not really perf sensitive so there shouldnt be any problems with cloning a fraction of it.
2478 20 : let upload_queue_for_deletion = UploadQueueInitialized {
2479 20 : inprogress_limit: initialized.inprogress_limit,
2480 20 : task_counter: 0,
2481 20 : dirty: initialized.dirty.clone(),
2482 20 : clean: initialized.clean.clone(),
2483 20 : latest_files_changes_since_metadata_upload_scheduled: 0,
2484 20 : visible_remote_consistent_lsn: initialized
2485 20 : .visible_remote_consistent_lsn
2486 20 : .clone(),
2487 20 : inprogress_tasks: HashMap::default(),
2488 20 : queued_operations: VecDeque::default(),
2489 20 : #[cfg(feature = "testing")]
2490 20 : dangling_files: HashMap::default(),
2491 20 : blocked_deletions: Vec::new(),
2492 20 : shutting_down: false,
2493 20 : shutdown_ready: Arc::new(tokio::sync::Semaphore::new(0)),
2494 20 : recently_deleted: HashSet::new(),
2495 20 : };
2496 20 :
2497 20 : let upload_queue = std::mem::replace(
2498 20 : &mut **guard,
2499 20 : UploadQueue::Stopped(UploadQueueStopped::Deletable(
2500 20 : UploadQueueStoppedDeletable {
2501 20 : upload_queue_for_deletion,
2502 20 : deleted_at: SetDeletedFlagProgress::NotRunning,
2503 20 : },
2504 20 : )),
2505 20 : );
2506 20 : if let UploadQueue::Initialized(qi) = upload_queue {
2507 20 : qi
2508 : } else {
2509 0 : unreachable!("we checked in the match above that it is Initialized");
2510 : }
2511 : };
2512 :
2513 : // We don't need to do anything here for in-progress tasks. They will finish
2514 : // on their own, decrement the unfinished-task counter themselves, and observe
2515 : // that the queue is Stopped.
2516 20 : drop(qi.inprogress_tasks);
2517 :
2518 : // Tear down queued ops
2519 20 : for op in qi.queued_operations.into_iter() {
2520 16 : self.metric_end(&op);
2521 16 : // Dropping UploadOp::Barrier() here will make wait_completion() return with an Err()
2522 16 : // which is exactly what we want to happen.
2523 16 : drop(op);
2524 16 : }
2525 : }
2526 : }
2527 36 : }
2528 :
2529 : /// Returns an accessor which will hold the UploadQueue mutex for accessing the upload queue
2530 : /// externally to RemoteTimelineClient.
2531 0 : pub(crate) fn initialized_upload_queue(
2532 0 : &self,
2533 0 : ) -> Result<UploadQueueAccessor<'_>, NotInitialized> {
2534 0 : let mut inner = self.upload_queue.lock().unwrap();
2535 0 : inner.initialized_mut()?;
2536 0 : Ok(UploadQueueAccessor { inner })
2537 0 : }
2538 :
2539 16 : pub(crate) fn no_pending_work(&self) -> bool {
2540 16 : let inner = self.upload_queue.lock().unwrap();
2541 16 : match &*inner {
2542 : UploadQueue::Uninitialized
2543 0 : | UploadQueue::Stopped(UploadQueueStopped::Uninitialized) => true,
2544 16 : UploadQueue::Stopped(UploadQueueStopped::Deletable(x)) => {
2545 16 : x.upload_queue_for_deletion.no_pending_work()
2546 : }
2547 0 : UploadQueue::Initialized(x) => x.no_pending_work(),
2548 : }
2549 16 : }
2550 :
2551 : /// 'foreign' in the sense that it does not belong to this tenant shard. This method
2552 : /// is used during GC for other shards to get the index of shard zero.
2553 0 : pub(crate) async fn download_foreign_index(
2554 0 : &self,
2555 0 : shard_number: ShardNumber,
2556 0 : cancel: &CancellationToken,
2557 0 : ) -> Result<(IndexPart, Generation, std::time::SystemTime), DownloadError> {
2558 0 : let foreign_shard_id = TenantShardId {
2559 0 : shard_number,
2560 0 : shard_count: self.tenant_shard_id.shard_count,
2561 0 : tenant_id: self.tenant_shard_id.tenant_id,
2562 0 : };
2563 0 : download_index_part(
2564 0 : &self.storage_impl,
2565 0 : &foreign_shard_id,
2566 0 : &self.timeline_id,
2567 0 : Generation::MAX,
2568 0 : cancel,
2569 0 : )
2570 0 : .await
2571 0 : }
2572 : }
2573 :
2574 : pub(crate) struct UploadQueueAccessor<'a> {
2575 : inner: std::sync::MutexGuard<'a, UploadQueue>,
2576 : }
2577 :
2578 : impl UploadQueueAccessor<'_> {
2579 0 : pub(crate) fn latest_uploaded_index_part(&self) -> &IndexPart {
2580 0 : match &*self.inner {
2581 0 : UploadQueue::Initialized(x) => &x.clean.0,
2582 : UploadQueue::Uninitialized | UploadQueue::Stopped(_) => {
2583 0 : unreachable!("checked before constructing")
2584 : }
2585 : }
2586 0 : }
2587 : }
2588 :
2589 0 : pub fn remote_tenant_path(tenant_shard_id: &TenantShardId) -> RemotePath {
2590 0 : let path = format!("tenants/{tenant_shard_id}");
2591 0 : RemotePath::from_string(&path).expect("Failed to construct path")
2592 0 : }
2593 :
2594 1360 : pub fn remote_tenant_manifest_path(
2595 1360 : tenant_shard_id: &TenantShardId,
2596 1360 : generation: Generation,
2597 1360 : ) -> RemotePath {
2598 1360 : let path = format!(
2599 1360 : "tenants/{tenant_shard_id}/tenant-manifest{}.json",
2600 1360 : generation.get_suffix()
2601 1360 : );
2602 1360 : RemotePath::from_string(&path).expect("Failed to construct path")
2603 1360 : }
2604 :
2605 : /// Prefix to all generations' manifest objects in a tenant shard
2606 452 : pub fn remote_tenant_manifest_prefix(tenant_shard_id: &TenantShardId) -> RemotePath {
2607 452 : let path = format!("tenants/{tenant_shard_id}/tenant-manifest",);
2608 452 : RemotePath::from_string(&path).expect("Failed to construct path")
2609 452 : }
2610 :
2611 512 : pub fn remote_timelines_path(tenant_shard_id: &TenantShardId) -> RemotePath {
2612 512 : let path = format!("tenants/{tenant_shard_id}/{TIMELINES_SEGMENT_NAME}");
2613 512 : RemotePath::from_string(&path).expect("Failed to construct path")
2614 512 : }
2615 :
2616 0 : fn remote_timelines_path_unsharded(tenant_id: &TenantId) -> RemotePath {
2617 0 : let path = format!("tenants/{tenant_id}/{TIMELINES_SEGMENT_NAME}");
2618 0 : RemotePath::from_string(&path).expect("Failed to construct path")
2619 0 : }
2620 :
2621 60 : pub fn remote_timeline_path(
2622 60 : tenant_shard_id: &TenantShardId,
2623 60 : timeline_id: &TimelineId,
2624 60 : ) -> RemotePath {
2625 60 : remote_timelines_path(tenant_shard_id).join(Utf8Path::new(&timeline_id.to_string()))
2626 60 : }
2627 :
2628 : /// Obtains the path of the given Layer in the remote
2629 : ///
2630 : /// Note that the shard component of a remote layer path is _not_ always the same
2631 : /// as in the TenantShardId of the caller: tenants may reference layers from a different
2632 : /// ShardIndex. Use the ShardIndex from the layer's metadata.
2633 3985 : pub fn remote_layer_path(
2634 3985 : tenant_id: &TenantId,
2635 3985 : timeline_id: &TimelineId,
2636 3985 : shard: ShardIndex,
2637 3985 : layer_file_name: &LayerName,
2638 3985 : generation: Generation,
2639 3985 : ) -> RemotePath {
2640 3985 : // Generation-aware key format
2641 3985 : let path = format!(
2642 3985 : "tenants/{tenant_id}{0}/{TIMELINES_SEGMENT_NAME}/{timeline_id}/{1}{2}",
2643 3985 : shard.get_suffix(),
2644 3985 : layer_file_name,
2645 3985 : generation.get_suffix()
2646 3985 : );
2647 3985 :
2648 3985 : RemotePath::from_string(&path).expect("Failed to construct path")
2649 3985 : }
2650 :
2651 : /// Returns true if a and b have the same layer path within a tenant/timeline. This is essentially
2652 : /// remote_layer_path(a) == remote_layer_path(b) without the string allocations.
2653 : ///
2654 : /// TODO: there should be a variant of LayerName for the physical path that contains information
2655 : /// about the shard and generation, such that this could be replaced by a simple comparison.
2656 2349709 : pub fn is_same_remote_layer_path(
2657 2349709 : aname: &LayerName,
2658 2349709 : ameta: &LayerFileMetadata,
2659 2349709 : bname: &LayerName,
2660 2349709 : bmeta: &LayerFileMetadata,
2661 2349709 : ) -> bool {
2662 2349709 : // NB: don't assert remote_layer_path(a) == remote_layer_path(b); too expensive even for debug.
2663 2349709 : aname == bname && ameta.shard == bmeta.shard && ameta.generation == bmeta.generation
2664 2349709 : }
2665 :
2666 8 : pub fn remote_initdb_archive_path(tenant_id: &TenantId, timeline_id: &TimelineId) -> RemotePath {
2667 8 : RemotePath::from_string(&format!(
2668 8 : "tenants/{tenant_id}/{TIMELINES_SEGMENT_NAME}/{timeline_id}/{INITDB_PATH}"
2669 8 : ))
2670 8 : .expect("Failed to construct path")
2671 8 : }
2672 :
2673 4 : pub fn remote_initdb_preserved_archive_path(
2674 4 : tenant_id: &TenantId,
2675 4 : timeline_id: &TimelineId,
2676 4 : ) -> RemotePath {
2677 4 : RemotePath::from_string(&format!(
2678 4 : "tenants/{tenant_id}/{TIMELINES_SEGMENT_NAME}/{timeline_id}/{INITDB_PRESERVED_PATH}"
2679 4 : ))
2680 4 : .expect("Failed to construct path")
2681 4 : }
2682 :
2683 3109 : pub fn remote_index_path(
2684 3109 : tenant_shard_id: &TenantShardId,
2685 3109 : timeline_id: &TimelineId,
2686 3109 : generation: Generation,
2687 3109 : ) -> RemotePath {
2688 3109 : RemotePath::from_string(&format!(
2689 3109 : "tenants/{tenant_shard_id}/{TIMELINES_SEGMENT_NAME}/{timeline_id}/{0}{1}",
2690 3109 : IndexPart::FILE_NAME,
2691 3109 : generation.get_suffix()
2692 3109 : ))
2693 3109 : .expect("Failed to construct path")
2694 3109 : }
2695 :
2696 0 : pub(crate) fn remote_heatmap_path(tenant_shard_id: &TenantShardId) -> RemotePath {
2697 0 : RemotePath::from_string(&format!(
2698 0 : "tenants/{tenant_shard_id}/{TENANT_HEATMAP_BASENAME}"
2699 0 : ))
2700 0 : .expect("Failed to construct path")
2701 0 : }
2702 :
2703 : /// Given the key of an index, parse out the generation part of the name
2704 36 : pub fn parse_remote_index_path(path: RemotePath) -> Option<Generation> {
2705 36 : let file_name = match path.get_path().file_name() {
2706 36 : Some(f) => f,
2707 : None => {
2708 : // Unexpected: we should be seeing index_part.json paths only
2709 0 : tracing::warn!("Malformed index key {}", path);
2710 0 : return None;
2711 : }
2712 : };
2713 :
2714 36 : match file_name.split_once('-') {
2715 24 : Some((_, gen_suffix)) => Generation::parse_suffix(gen_suffix),
2716 12 : None => None,
2717 : }
2718 36 : }
2719 :
2720 : /// Given the key of a tenant manifest, parse out the generation number
2721 0 : pub fn parse_remote_tenant_manifest_path(path: RemotePath) -> Option<Generation> {
2722 : static RE: OnceLock<Regex> = OnceLock::new();
2723 0 : let re = RE.get_or_init(|| Regex::new(r".*tenant-manifest-([0-9a-f]{8}).json").unwrap());
2724 0 : re.captures(path.get_path().as_str())
2725 0 : .and_then(|c| c.get(1))
2726 0 : .and_then(|m| Generation::parse_suffix(m.as_str()))
2727 0 : }
2728 :
2729 : #[cfg(test)]
2730 : mod tests {
2731 : use std::collections::HashSet;
2732 :
2733 : use super::*;
2734 : use crate::DEFAULT_PG_VERSION;
2735 : use crate::context::RequestContext;
2736 : use crate::tenant::config::AttachmentMode;
2737 : use crate::tenant::harness::{TIMELINE_ID, TenantHarness};
2738 : use crate::tenant::storage_layer::layer::local_layer_path;
2739 : use crate::tenant::{Tenant, Timeline};
2740 :
2741 16 : pub(super) fn dummy_contents(name: &str) -> Vec<u8> {
2742 16 : format!("contents for {name}").into()
2743 16 : }
2744 :
2745 4 : pub(super) fn dummy_metadata(disk_consistent_lsn: Lsn) -> TimelineMetadata {
2746 4 : let metadata = TimelineMetadata::new(
2747 4 : disk_consistent_lsn,
2748 4 : None,
2749 4 : None,
2750 4 : Lsn(0),
2751 4 : Lsn(0),
2752 4 : Lsn(0),
2753 4 : // Any version will do
2754 4 : // but it should be consistent with the one in the tests
2755 4 : crate::DEFAULT_PG_VERSION,
2756 4 : );
2757 4 :
2758 4 : // go through serialize + deserialize to fix the header, including checksum
2759 4 : TimelineMetadata::from_bytes(&metadata.to_bytes().unwrap()).unwrap()
2760 4 : }
2761 :
2762 4 : fn assert_file_list(a: &HashSet<LayerName>, b: &[&str]) {
2763 12 : let mut avec: Vec<String> = a.iter().map(|x| x.to_string()).collect();
2764 4 : avec.sort();
2765 4 :
2766 4 : let mut bvec = b.to_vec();
2767 4 : bvec.sort_unstable();
2768 4 :
2769 4 : assert_eq!(avec, bvec);
2770 4 : }
2771 :
2772 8 : fn assert_remote_files(expected: &[&str], remote_path: &Utf8Path, generation: Generation) {
2773 8 : let mut expected: Vec<String> = expected
2774 8 : .iter()
2775 32 : .map(|x| format!("{}{}", x, generation.get_suffix()))
2776 8 : .collect();
2777 8 : expected.sort();
2778 8 :
2779 8 : let mut found: Vec<String> = Vec::new();
2780 32 : for entry in std::fs::read_dir(remote_path).unwrap().flatten() {
2781 32 : let entry_name = entry.file_name();
2782 32 : let fname = entry_name.to_str().unwrap();
2783 32 : found.push(String::from(fname));
2784 32 : }
2785 8 : found.sort();
2786 8 :
2787 8 : assert_eq!(found, expected);
2788 8 : }
2789 :
2790 : struct TestSetup {
2791 : harness: TenantHarness,
2792 : tenant: Arc<Tenant>,
2793 : timeline: Arc<Timeline>,
2794 : tenant_ctx: RequestContext,
2795 : }
2796 :
2797 : impl TestSetup {
2798 16 : async fn new(test_name: &str) -> anyhow::Result<Self> {
2799 16 : let test_name = Box::leak(Box::new(format!("remote_timeline_client__{test_name}")));
2800 16 : let harness = TenantHarness::create(test_name).await?;
2801 16 : let (tenant, ctx) = harness.load().await;
2802 :
2803 16 : let timeline = tenant
2804 16 : .create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)
2805 16 : .await?;
2806 :
2807 16 : Ok(Self {
2808 16 : harness,
2809 16 : tenant,
2810 16 : timeline,
2811 16 : tenant_ctx: ctx,
2812 16 : })
2813 16 : }
2814 :
2815 : /// Construct a RemoteTimelineClient in an arbitrary generation
2816 20 : fn build_client(&self, generation: Generation) -> Arc<RemoteTimelineClient> {
2817 20 : let location_conf = AttachedLocationConfig {
2818 20 : generation,
2819 20 : attach_mode: AttachmentMode::Single,
2820 20 : };
2821 20 : Arc::new(RemoteTimelineClient {
2822 20 : conf: self.harness.conf,
2823 20 : runtime: tokio::runtime::Handle::current(),
2824 20 : tenant_shard_id: self.harness.tenant_shard_id,
2825 20 : timeline_id: TIMELINE_ID,
2826 20 : generation,
2827 20 : storage_impl: self.harness.remote_storage.clone(),
2828 20 : deletion_queue_client: self.harness.deletion_queue.new_client(),
2829 20 : upload_queue: Mutex::new(UploadQueue::Uninitialized),
2830 20 : metrics: Arc::new(RemoteTimelineClientMetrics::new(
2831 20 : &self.harness.tenant_shard_id,
2832 20 : &TIMELINE_ID,
2833 20 : )),
2834 20 : config: std::sync::RwLock::new(RemoteTimelineClientConfig::from(&location_conf)),
2835 20 : cancel: CancellationToken::new(),
2836 20 : })
2837 20 : }
2838 :
2839 : /// A tracing::Span that satisfies remote_timeline_client methods that assert tenant_id
2840 : /// and timeline_id are present.
2841 12 : fn span(&self) -> tracing::Span {
2842 12 : tracing::info_span!(
2843 : "test",
2844 : tenant_id = %self.harness.tenant_shard_id.tenant_id,
2845 0 : shard_id = %self.harness.tenant_shard_id.shard_slug(),
2846 : timeline_id = %TIMELINE_ID
2847 : )
2848 12 : }
2849 : }
2850 :
2851 : // Test scheduling
2852 : #[tokio::test]
2853 4 : async fn upload_scheduling() {
2854 4 : // Test outline:
2855 4 : //
2856 4 : // Schedule upload of a bunch of layers. Check that they are started immediately, not queued
2857 4 : // Schedule upload of index. Check that it is queued
2858 4 : // let the layer file uploads finish. Check that the index-upload is now started
2859 4 : // let the index-upload finish.
2860 4 : //
2861 4 : // Download back the index.json. Check that the list of files is correct
2862 4 : //
2863 4 : // Schedule upload. Schedule deletion. Check that the deletion is queued
2864 4 : // let upload finish. Check that deletion is now started
2865 4 : // Schedule another deletion. Check that it's launched immediately.
2866 4 : // Schedule index upload. Check that it's queued
2867 4 :
2868 4 : let test_setup = TestSetup::new("upload_scheduling").await.unwrap();
2869 4 : let span = test_setup.span();
2870 4 : let _guard = span.enter();
2871 4 :
2872 4 : let TestSetup {
2873 4 : harness,
2874 4 : tenant: _tenant,
2875 4 : timeline,
2876 4 : tenant_ctx: _tenant_ctx,
2877 4 : } = test_setup;
2878 4 :
2879 4 : let client = &timeline.remote_client;
2880 4 :
2881 4 : // Download back the index.json, and check that the list of files is correct
2882 4 : let initial_index_part = match client
2883 4 : .download_index_file(&CancellationToken::new())
2884 4 : .await
2885 4 : .unwrap()
2886 4 : {
2887 4 : MaybeDeletedIndexPart::IndexPart(index_part) => index_part,
2888 4 : MaybeDeletedIndexPart::Deleted(_) => panic!("unexpectedly got deleted index part"),
2889 4 : };
2890 4 : let initial_layers = initial_index_part
2891 4 : .layer_metadata
2892 4 : .keys()
2893 4 : .map(|f| f.to_owned())
2894 4 : .collect::<HashSet<LayerName>>();
2895 4 : let initial_layer = {
2896 4 : assert!(initial_layers.len() == 1);
2897 4 : initial_layers.into_iter().next().unwrap()
2898 4 : };
2899 4 :
2900 4 : let timeline_path = harness.timeline_path(&TIMELINE_ID);
2901 4 :
2902 4 : println!("workdir: {}", harness.conf.workdir);
2903 4 :
2904 4 : let remote_timeline_dir = harness
2905 4 : .remote_fs_dir
2906 4 : .join(timeline_path.strip_prefix(&harness.conf.workdir).unwrap());
2907 4 : println!("remote_timeline_dir: {remote_timeline_dir}");
2908 4 :
2909 4 : let generation = harness.generation;
2910 4 : let shard = harness.shard;
2911 4 :
2912 4 : // Create a couple of dummy files, schedule upload for them
2913 4 :
2914 4 : let layers = [
2915 4 : ("000000000000000000000000000000000000-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF__00000000016B59D8-00000000016B5A51".parse().unwrap(), dummy_contents("foo")),
2916 4 : ("000000000000000000000000000000000000-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF__00000000016B59D9-00000000016B5A52".parse().unwrap(), dummy_contents("bar")),
2917 4 : ("000000000000000000000000000000000000-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF__00000000016B59DA-00000000016B5A53".parse().unwrap(), dummy_contents("baz"))
2918 4 : ]
2919 4 : .into_iter()
2920 12 : .map(|(name, contents): (LayerName, Vec<u8>)| {
2921 12 :
2922 12 : let local_path = local_layer_path(
2923 12 : harness.conf,
2924 12 : &timeline.tenant_shard_id,
2925 12 : &timeline.timeline_id,
2926 12 : &name,
2927 12 : &generation,
2928 12 : );
2929 12 : std::fs::write(&local_path, &contents).unwrap();
2930 12 :
2931 12 : Layer::for_resident(
2932 12 : harness.conf,
2933 12 : &timeline,
2934 12 : local_path,
2935 12 : name,
2936 12 : LayerFileMetadata::new(contents.len() as u64, generation, shard),
2937 12 : )
2938 12 : }).collect::<Vec<_>>();
2939 4 :
2940 4 : client
2941 4 : .schedule_layer_file_upload(layers[0].clone())
2942 4 : .unwrap();
2943 4 : client
2944 4 : .schedule_layer_file_upload(layers[1].clone())
2945 4 : .unwrap();
2946 4 :
2947 4 : // Check that they are started immediately, not queued
2948 4 : //
2949 4 : // this works because we running within block_on, so any futures are now queued up until
2950 4 : // our next await point.
2951 4 : {
2952 4 : let mut guard = client.upload_queue.lock().unwrap();
2953 4 : let upload_queue = guard.initialized_mut().unwrap();
2954 4 : assert!(upload_queue.queued_operations.is_empty());
2955 4 : assert_eq!(upload_queue.inprogress_tasks.len(), 2);
2956 4 : assert_eq!(upload_queue.num_inprogress_layer_uploads(), 2);
2957 4 :
2958 4 : // also check that `latest_file_changes` was updated
2959 4 : assert!(upload_queue.latest_files_changes_since_metadata_upload_scheduled == 2);
2960 4 : }
2961 4 :
2962 4 : // Schedule upload of index. Check that it is queued
2963 4 : let metadata = dummy_metadata(Lsn(0x20));
2964 4 : client
2965 4 : .schedule_index_upload_for_full_metadata_update(&metadata)
2966 4 : .unwrap();
2967 4 : {
2968 4 : let mut guard = client.upload_queue.lock().unwrap();
2969 4 : let upload_queue = guard.initialized_mut().unwrap();
2970 4 : assert!(upload_queue.queued_operations.len() == 1);
2971 4 : assert!(upload_queue.latest_files_changes_since_metadata_upload_scheduled == 0);
2972 4 : }
2973 4 :
2974 4 : // Wait for the uploads to finish
2975 4 : client.wait_completion().await.unwrap();
2976 4 : {
2977 4 : let mut guard = client.upload_queue.lock().unwrap();
2978 4 : let upload_queue = guard.initialized_mut().unwrap();
2979 4 :
2980 4 : assert!(upload_queue.queued_operations.is_empty());
2981 4 : assert!(upload_queue.inprogress_tasks.is_empty());
2982 4 : }
2983 4 :
2984 4 : // Download back the index.json, and check that the list of files is correct
2985 4 : let index_part = match client
2986 4 : .download_index_file(&CancellationToken::new())
2987 4 : .await
2988 4 : .unwrap()
2989 4 : {
2990 4 : MaybeDeletedIndexPart::IndexPart(index_part) => index_part,
2991 4 : MaybeDeletedIndexPart::Deleted(_) => panic!("unexpectedly got deleted index part"),
2992 4 : };
2993 4 :
2994 4 : assert_file_list(
2995 4 : &index_part
2996 4 : .layer_metadata
2997 4 : .keys()
2998 12 : .map(|f| f.to_owned())
2999 4 : .collect(),
3000 4 : &[
3001 4 : &initial_layer.to_string(),
3002 4 : &layers[0].layer_desc().layer_name().to_string(),
3003 4 : &layers[1].layer_desc().layer_name().to_string(),
3004 4 : ],
3005 4 : );
3006 4 : assert_eq!(index_part.metadata, metadata);
3007 4 :
3008 4 : // Schedule upload and then a deletion. Check that the deletion is queued
3009 4 : client
3010 4 : .schedule_layer_file_upload(layers[2].clone())
3011 4 : .unwrap();
3012 4 :
3013 4 : // this is no longer consistent with how deletion works with Layer::drop, but in this test
3014 4 : // keep using schedule_layer_file_deletion because we don't have a way to wait for the
3015 4 : // spawn_blocking started by the drop.
3016 4 : client
3017 4 : .schedule_layer_file_deletion(&[layers[0].layer_desc().layer_name()])
3018 4 : .unwrap();
3019 4 : {
3020 4 : let mut guard = client.upload_queue.lock().unwrap();
3021 4 : let upload_queue = guard.initialized_mut().unwrap();
3022 4 :
3023 4 : // Deletion schedules upload of the index file, and the file deletion itself
3024 4 : assert_eq!(upload_queue.queued_operations.len(), 2);
3025 4 : assert_eq!(upload_queue.inprogress_tasks.len(), 1);
3026 4 : assert_eq!(upload_queue.num_inprogress_layer_uploads(), 1);
3027 4 : assert_eq!(upload_queue.num_inprogress_deletions(), 0);
3028 4 : assert_eq!(
3029 4 : upload_queue.latest_files_changes_since_metadata_upload_scheduled,
3030 4 : 0
3031 4 : );
3032 4 : }
3033 4 : assert_remote_files(
3034 4 : &[
3035 4 : &initial_layer.to_string(),
3036 4 : &layers[0].layer_desc().layer_name().to_string(),
3037 4 : &layers[1].layer_desc().layer_name().to_string(),
3038 4 : "index_part.json",
3039 4 : ],
3040 4 : &remote_timeline_dir,
3041 4 : generation,
3042 4 : );
3043 4 :
3044 4 : // Finish them
3045 4 : client.wait_completion().await.unwrap();
3046 4 : harness.deletion_queue.pump().await;
3047 4 :
3048 4 : assert_remote_files(
3049 4 : &[
3050 4 : &initial_layer.to_string(),
3051 4 : &layers[1].layer_desc().layer_name().to_string(),
3052 4 : &layers[2].layer_desc().layer_name().to_string(),
3053 4 : "index_part.json",
3054 4 : ],
3055 4 : &remote_timeline_dir,
3056 4 : generation,
3057 4 : );
3058 4 : }
3059 :
3060 : #[tokio::test]
3061 4 : async fn bytes_unfinished_gauge_for_layer_file_uploads() {
3062 4 : // Setup
3063 4 :
3064 4 : let TestSetup {
3065 4 : harness,
3066 4 : tenant: _tenant,
3067 4 : timeline,
3068 4 : ..
3069 4 : } = TestSetup::new("metrics").await.unwrap();
3070 4 : let client = &timeline.remote_client;
3071 4 :
3072 4 : let layer_file_name_1: LayerName = "000000000000000000000000000000000000-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF__00000000016B59D8-00000000016B5A51".parse().unwrap();
3073 4 : let local_path = local_layer_path(
3074 4 : harness.conf,
3075 4 : &timeline.tenant_shard_id,
3076 4 : &timeline.timeline_id,
3077 4 : &layer_file_name_1,
3078 4 : &harness.generation,
3079 4 : );
3080 4 : let content_1 = dummy_contents("foo");
3081 4 : std::fs::write(&local_path, &content_1).unwrap();
3082 4 :
3083 4 : let layer_file_1 = Layer::for_resident(
3084 4 : harness.conf,
3085 4 : &timeline,
3086 4 : local_path,
3087 4 : layer_file_name_1.clone(),
3088 4 : LayerFileMetadata::new(content_1.len() as u64, harness.generation, harness.shard),
3089 4 : );
3090 4 :
3091 4 : #[derive(Debug, PartialEq, Clone, Copy)]
3092 4 : struct BytesStartedFinished {
3093 4 : started: Option<usize>,
3094 4 : finished: Option<usize>,
3095 4 : }
3096 4 : impl std::ops::Add for BytesStartedFinished {
3097 4 : type Output = Self;
3098 8 : fn add(self, rhs: Self) -> Self::Output {
3099 8 : Self {
3100 8 : started: self.started.map(|v| v + rhs.started.unwrap_or(0)),
3101 8 : finished: self.finished.map(|v| v + rhs.finished.unwrap_or(0)),
3102 8 : }
3103 8 : }
3104 4 : }
3105 12 : let get_bytes_started_stopped = || {
3106 12 : let started = client
3107 12 : .metrics
3108 12 : .get_bytes_started_counter_value(&RemoteOpFileKind::Layer, &RemoteOpKind::Upload)
3109 12 : .map(|v| v.try_into().unwrap());
3110 12 : let stopped = client
3111 12 : .metrics
3112 12 : .get_bytes_finished_counter_value(&RemoteOpFileKind::Layer, &RemoteOpKind::Upload)
3113 12 : .map(|v| v.try_into().unwrap());
3114 12 : BytesStartedFinished {
3115 12 : started,
3116 12 : finished: stopped,
3117 12 : }
3118 12 : };
3119 4 :
3120 4 : // Test
3121 4 : tracing::info!("now doing actual test");
3122 4 :
3123 4 : let actual_a = get_bytes_started_stopped();
3124 4 :
3125 4 : client
3126 4 : .schedule_layer_file_upload(layer_file_1.clone())
3127 4 : .unwrap();
3128 4 :
3129 4 : let actual_b = get_bytes_started_stopped();
3130 4 :
3131 4 : client.wait_completion().await.unwrap();
3132 4 :
3133 4 : let actual_c = get_bytes_started_stopped();
3134 4 :
3135 4 : // Validate
3136 4 :
3137 4 : let expected_b = actual_a
3138 4 : + BytesStartedFinished {
3139 4 : started: Some(content_1.len()),
3140 4 : // assert that the _finished metric is created eagerly so that subtractions work on first sample
3141 4 : finished: Some(0),
3142 4 : };
3143 4 : assert_eq!(actual_b, expected_b);
3144 4 :
3145 4 : let expected_c = actual_a
3146 4 : + BytesStartedFinished {
3147 4 : started: Some(content_1.len()),
3148 4 : finished: Some(content_1.len()),
3149 4 : };
3150 4 : assert_eq!(actual_c, expected_c);
3151 4 : }
3152 :
3153 24 : async fn inject_index_part(test_state: &TestSetup, generation: Generation) -> IndexPart {
3154 24 : // An empty IndexPart, just sufficient to ensure deserialization will succeed
3155 24 : let example_index_part = IndexPart::example();
3156 24 :
3157 24 : let index_part_bytes = serde_json::to_vec(&example_index_part).unwrap();
3158 24 :
3159 24 : let index_path = test_state.harness.remote_fs_dir.join(
3160 24 : remote_index_path(
3161 24 : &test_state.harness.tenant_shard_id,
3162 24 : &TIMELINE_ID,
3163 24 : generation,
3164 24 : )
3165 24 : .get_path(),
3166 24 : );
3167 24 :
3168 24 : std::fs::create_dir_all(index_path.parent().unwrap())
3169 24 : .expect("creating test dir should work");
3170 24 :
3171 24 : eprintln!("Writing {index_path}");
3172 24 : std::fs::write(&index_path, index_part_bytes).unwrap();
3173 24 : example_index_part
3174 24 : }
3175 :
3176 : /// Assert that when a RemoteTimelineclient in generation `get_generation` fetches its
3177 : /// index, the IndexPart returned is equal to `expected`
3178 20 : async fn assert_got_index_part(
3179 20 : test_state: &TestSetup,
3180 20 : get_generation: Generation,
3181 20 : expected: &IndexPart,
3182 20 : ) {
3183 20 : let client = test_state.build_client(get_generation);
3184 :
3185 20 : let download_r = client
3186 20 : .download_index_file(&CancellationToken::new())
3187 20 : .await
3188 20 : .expect("download should always succeed");
3189 20 : assert!(matches!(download_r, MaybeDeletedIndexPart::IndexPart(_)));
3190 20 : match download_r {
3191 20 : MaybeDeletedIndexPart::IndexPart(index_part) => {
3192 20 : assert_eq!(&index_part, expected);
3193 : }
3194 0 : MaybeDeletedIndexPart::Deleted(_index_part) => panic!("Test doesn't set deleted_at"),
3195 : }
3196 20 : }
3197 :
3198 : #[tokio::test]
3199 4 : async fn index_part_download_simple() -> anyhow::Result<()> {
3200 4 : let test_state = TestSetup::new("index_part_download_simple").await.unwrap();
3201 4 : let span = test_state.span();
3202 4 : let _guard = span.enter();
3203 4 :
3204 4 : // Simple case: we are in generation N, load the index from generation N - 1
3205 4 : let generation_n = 5;
3206 4 : let injected = inject_index_part(&test_state, Generation::new(generation_n - 1)).await;
3207 4 :
3208 4 : assert_got_index_part(&test_state, Generation::new(generation_n), &injected).await;
3209 4 :
3210 4 : Ok(())
3211 4 : }
3212 :
3213 : #[tokio::test]
3214 4 : async fn index_part_download_ordering() -> anyhow::Result<()> {
3215 4 : let test_state = TestSetup::new("index_part_download_ordering")
3216 4 : .await
3217 4 : .unwrap();
3218 4 :
3219 4 : let span = test_state.span();
3220 4 : let _guard = span.enter();
3221 4 :
3222 4 : // A generation-less IndexPart exists in the bucket, we should find it
3223 4 : let generation_n = 5;
3224 4 : let injected_none = inject_index_part(&test_state, Generation::none()).await;
3225 4 : assert_got_index_part(&test_state, Generation::new(generation_n), &injected_none).await;
3226 4 :
3227 4 : // If a more recent-than-none generation exists, we should prefer to load that
3228 4 : let injected_1 = inject_index_part(&test_state, Generation::new(1)).await;
3229 4 : assert_got_index_part(&test_state, Generation::new(generation_n), &injected_1).await;
3230 4 :
3231 4 : // If a more-recent-than-me generation exists, we should ignore it.
3232 4 : let _injected_10 = inject_index_part(&test_state, Generation::new(10)).await;
3233 4 : assert_got_index_part(&test_state, Generation::new(generation_n), &injected_1).await;
3234 4 :
3235 4 : // If a directly previous generation exists, _and_ an index exists in my own
3236 4 : // generation, I should prefer my own generation.
3237 4 : let _injected_prev =
3238 4 : inject_index_part(&test_state, Generation::new(generation_n - 1)).await;
3239 4 : let injected_current = inject_index_part(&test_state, Generation::new(generation_n)).await;
3240 4 : assert_got_index_part(
3241 4 : &test_state,
3242 4 : Generation::new(generation_n),
3243 4 : &injected_current,
3244 4 : )
3245 4 : .await;
3246 4 :
3247 4 : Ok(())
3248 4 : }
3249 : }
|
