Line data Source code
1 : //!
2 : //! This module provides centralized handling of tokio tasks in the Page Server.
3 : //!
4 : //! We provide a few basic facilities:
5 : //! - A global registry of tasks that lists what kind of tasks they are, and
6 : //! which tenant or timeline they are working on
7 : //!
8 : //! - The ability to request a task to shut down.
9 : //!
10 : //!
11 : //! # How it works?
12 : //!
13 : //! There is a global hashmap of all the tasks (`TASKS`). Whenever a new
14 : //! task is spawned, a PageServerTask entry is added there, and when a
15 : //! task dies, it removes itself from the hashmap. If you want to kill a
16 : //! task, you can scan the hashmap to find it.
17 : //!
18 : //! # Task shutdown
19 : //!
20 : //! To kill a task, we rely on co-operation from the victim. Each task is
21 : //! expected to periodically call the `is_shutdown_requested()` function, and
22 : //! if it returns true, exit gracefully. In addition to that, when waiting for
23 : //! the network or other long-running operation, you can use
24 : //! `shutdown_watcher()` function to get a Future that will become ready if
25 : //! the current task has been requested to shut down. You can use that with
26 : //! Tokio select!().
27 : //!
28 : //! TODO: This would be a good place to also handle panics in a somewhat sane way.
29 : //! Depending on what task panics, we might want to kill the whole server, or
30 : //! only a single tenant or timeline.
31 : //!
32 :
33 : use std::collections::HashMap;
34 : use std::fmt;
35 : use std::future::Future;
36 : use std::num::NonZeroUsize;
37 : use std::panic::AssertUnwindSafe;
38 : use std::str::FromStr;
39 : use std::sync::atomic::{AtomicU64, Ordering};
40 : use std::sync::{Arc, Mutex};
41 :
42 : use futures::FutureExt;
43 : use pageserver_api::shard::TenantShardId;
44 : use tokio::task::JoinHandle;
45 : use tokio::task_local;
46 : use tokio_util::sync::CancellationToken;
47 :
48 : use tracing::{debug, error, info, warn};
49 :
50 : use once_cell::sync::Lazy;
51 :
52 : use utils::env;
53 : use utils::id::TimelineId;
54 :
55 : use crate::metrics::set_tokio_runtime_setup;
56 :
57 : //
58 : // There are four runtimes:
59 : //
60 : // Compute request runtime
61 : // - used to handle connections from compute nodes. Any tasks related to satisfying
62 : // GetPage requests, base backups, import, and other such compute node operations
63 : // are handled by the Compute request runtime
64 : // - page_service.rs
65 : // - this includes layer downloads from remote storage, if a layer is needed to
66 : // satisfy a GetPage request
67 : //
68 : // Management request runtime
69 : // - used to handle HTTP API requests
70 : //
71 : // WAL receiver runtime:
72 : // - used to handle WAL receiver connections.
73 : // - and to receiver updates from storage_broker
74 : //
75 : // Background runtime
76 : // - layer flushing
77 : // - garbage collection
78 : // - compaction
79 : // - remote storage uploads
80 : // - initial tenant loading
81 : //
82 : // Everything runs in a tokio task. If you spawn new tasks, spawn it using the correct
83 : // runtime.
84 : //
85 : // There might be situations when one task needs to wait for a task running in another
86 : // Runtime to finish. For example, if a background operation needs a layer from remote
87 : // storage, it will start to download it. If a background operation needs a remote layer,
88 : // and the download was already initiated by a GetPage request, the background task
89 : // will wait for the download - running in the Page server runtime - to finish.
90 : // Another example: the initial tenant loading tasks are launched in the background ops
91 : // runtime. If a GetPage request comes in before the load of a tenant has finished, the
92 : // GetPage request will wait for the tenant load to finish.
93 : //
94 : // The core Timeline code is synchronous, and uses a bunch of std Mutexes and RWLocks to
95 : // protect data structures. Let's keep it that way. Synchronous code is easier to debug
96 : // and analyze, and there's a lot of hairy, low-level, performance critical code there.
97 : //
98 : // It's nice to have different runtimes, so that you can quickly eyeball how much CPU
99 : // time each class of operations is taking, with 'top -H' or similar.
100 : //
101 : // It's also good to avoid hogging all threads that would be needed to process
102 : // other operations, if the upload tasks e.g. get blocked on locks. It shouldn't
103 : // happen, but still.
104 : //
105 :
106 498 : pub(crate) static TOKIO_WORKER_THREADS: Lazy<NonZeroUsize> = Lazy::new(|| {
107 498 : // replicates tokio-1.28.1::loom::sys::num_cpus which is not available publicly
108 498 : // tokio would had already panicked for parsing errors or NotUnicode
109 498 : //
110 498 : // this will be wrong if any of the runtimes gets their worker threads configured to something
111 498 : // else, but that has not been needed in a long time.
112 498 : NonZeroUsize::new(
113 498 : std::env::var("TOKIO_WORKER_THREADS")
114 498 : .map(|s| s.parse::<usize>().unwrap())
115 498 : .unwrap_or_else(|_e| usize::max(2, num_cpus::get())),
116 498 : )
117 498 : .expect("the max() ensures that this is not zero")
118 498 : });
119 :
120 : enum TokioRuntimeMode {
121 : SingleThreaded,
122 : MultiThreaded { num_workers: NonZeroUsize },
123 : }
124 :
125 : impl FromStr for TokioRuntimeMode {
126 : type Err = String;
127 :
128 0 : fn from_str(s: &str) -> Result<Self, Self::Err> {
129 0 : match s {
130 0 : "current_thread" => Ok(TokioRuntimeMode::SingleThreaded),
131 0 : s => match s.strip_prefix("multi_thread:") {
132 0 : Some("default") => Ok(TokioRuntimeMode::MultiThreaded {
133 0 : num_workers: *TOKIO_WORKER_THREADS,
134 0 : }),
135 0 : Some(suffix) => {
136 0 : let num_workers = suffix.parse::<NonZeroUsize>().map_err(|e| {
137 0 : format!(
138 0 : "invalid number of multi-threaded runtime workers ({suffix:?}): {e}",
139 0 : )
140 0 : })?;
141 0 : Ok(TokioRuntimeMode::MultiThreaded { num_workers })
142 : }
143 0 : None => Err(format!("invalid runtime config: {s:?}")),
144 : },
145 : }
146 0 : }
147 : }
148 :
149 498 : static TOKIO_THREAD_STACK_SIZE: Lazy<NonZeroUsize> = Lazy::new(|| {
150 498 : env::var("NEON_PAGESERVER_TOKIO_THREAD_STACK_SIZE")
151 498 : // the default 2MiB are insufficent, especially in debug mode
152 498 : .unwrap_or_else(|| NonZeroUsize::new(4 * 1024 * 1024).unwrap())
153 498 : });
154 :
155 498 : static ONE_RUNTIME: Lazy<Option<tokio::runtime::Runtime>> = Lazy::new(|| {
156 498 : let thread_name = "pageserver-tokio";
157 498 : let Some(mode) = env::var("NEON_PAGESERVER_USE_ONE_RUNTIME") else {
158 : // If the env var is not set, leave this static as None.
159 498 : set_tokio_runtime_setup(
160 498 : "multiple-runtimes",
161 498 : NUM_MULTIPLE_RUNTIMES
162 498 : .checked_mul(*TOKIO_WORKER_THREADS)
163 498 : .unwrap(),
164 498 : );
165 498 : return None;
166 : };
167 0 : Some(match mode {
168 : TokioRuntimeMode::SingleThreaded => {
169 0 : set_tokio_runtime_setup("one-runtime-single-threaded", NonZeroUsize::new(1).unwrap());
170 0 : tokio::runtime::Builder::new_current_thread()
171 0 : .thread_name(thread_name)
172 0 : .enable_all()
173 0 : .thread_stack_size(TOKIO_THREAD_STACK_SIZE.get())
174 0 : .build()
175 0 : .expect("failed to create one single runtime")
176 : }
177 0 : TokioRuntimeMode::MultiThreaded { num_workers } => {
178 0 : set_tokio_runtime_setup("one-runtime-multi-threaded", num_workers);
179 0 : tokio::runtime::Builder::new_multi_thread()
180 0 : .thread_name(thread_name)
181 0 : .enable_all()
182 0 : .worker_threads(num_workers.get())
183 0 : .thread_stack_size(TOKIO_THREAD_STACK_SIZE.get())
184 0 : .build()
185 0 : .expect("failed to create one multi-threaded runtime")
186 : }
187 : })
188 498 : });
189 :
190 : /// Declare a lazy static variable named `$varname` that will resolve
191 : /// to a tokio runtime handle. If the env var `NEON_PAGESERVER_USE_ONE_RUNTIME`
192 : /// is set, this will resolve to `ONE_RUNTIME`. Otherwise, the macro invocation
193 : /// declares a separate runtime and the lazy static variable `$varname`
194 : /// will resolve to that separate runtime.
195 : ///
196 : /// The result is is that `$varname.spawn()` will use `ONE_RUNTIME` if
197 : /// `NEON_PAGESERVER_USE_ONE_RUNTIME` is set, and will use the separate runtime
198 : /// otherwise.
199 : macro_rules! pageserver_runtime {
200 : ($varname:ident, $name:literal) => {
201 528 : pub static $varname: Lazy<&'static tokio::runtime::Runtime> = Lazy::new(|| {
202 528 : if let Some(runtime) = &*ONE_RUNTIME {
203 0 : return runtime;
204 528 : }
205 528 : static RUNTIME: Lazy<tokio::runtime::Runtime> = Lazy::new(|| {
206 528 : tokio::runtime::Builder::new_multi_thread()
207 528 : .thread_name($name)
208 528 : .worker_threads(TOKIO_WORKER_THREADS.get())
209 528 : .enable_all()
210 528 : .thread_stack_size(TOKIO_THREAD_STACK_SIZE.get())
211 528 : .build()
212 528 : .expect(std::concat!("Failed to create runtime ", $name))
213 528 : });
214 528 : &*RUNTIME
215 528 : });
216 : };
217 : }
218 :
219 : pageserver_runtime!(COMPUTE_REQUEST_RUNTIME, "compute request worker");
220 : pageserver_runtime!(MGMT_REQUEST_RUNTIME, "mgmt request worker");
221 : pageserver_runtime!(WALRECEIVER_RUNTIME, "walreceiver worker");
222 : pageserver_runtime!(BACKGROUND_RUNTIME, "background op worker");
223 : // Bump this number when adding a new pageserver_runtime!
224 : // SAFETY: it's obviously correct
225 : const NUM_MULTIPLE_RUNTIMES: NonZeroUsize = unsafe { NonZeroUsize::new_unchecked(4) };
226 :
227 : #[derive(Debug, Clone, Copy)]
228 : pub struct PageserverTaskId(u64);
229 :
230 : impl fmt::Display for PageserverTaskId {
231 0 : fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
232 0 : self.0.fmt(f)
233 0 : }
234 : }
235 :
236 : /// Each task that we track is associated with a "task ID". It's just an
237 : /// increasing number that we assign. Note that it is different from tokio::task::Id.
238 : static NEXT_TASK_ID: AtomicU64 = AtomicU64::new(1);
239 :
240 : /// Global registry of tasks
241 : static TASKS: Lazy<Mutex<HashMap<u64, Arc<PageServerTask>>>> =
242 516 : Lazy::new(|| Mutex::new(HashMap::new()));
243 :
244 : task_local! {
245 : // This is a cancellation token which will be cancelled when a task needs to shut down. The
246 : // root token is kept in the global registry, so that anyone can send the signal to request
247 : // task shutdown.
248 : static SHUTDOWN_TOKEN: CancellationToken;
249 :
250 : // Each task holds reference to its own PageServerTask here.
251 : static CURRENT_TASK: Arc<PageServerTask>;
252 : }
253 :
254 : ///
255 : /// There are many kinds of tasks in the system. Some are associated with a particular
256 : /// tenant or timeline, while others are global.
257 : ///
258 : /// Note that we don't try to limit how many task of a certain kind can be running
259 : /// at the same time.
260 : ///
261 : #[derive(
262 : Debug,
263 : // NB: enumset::EnumSetType derives PartialEq, Eq, Clone, Copy
264 0 : enumset::EnumSetType,
265 : enum_map::Enum,
266 : serde::Serialize,
267 0 : serde::Deserialize,
268 8736 : strum_macros::IntoStaticStr,
269 0 : strum_macros::EnumString,
270 : )]
271 : pub enum TaskKind {
272 : // Pageserver startup, i.e., `main`
273 : Startup,
274 :
275 : // libpq listener task. It just accepts connection and spawns a
276 : // PageRequestHandler task for each connection.
277 : LibpqEndpointListener,
278 :
279 : // HTTP endpoint listener.
280 : HttpEndpointListener,
281 :
282 : // Task that handles a single connection. A PageRequestHandler task
283 : // starts detached from any particular tenant or timeline, but it can be
284 : // associated with one later, after receiving a command from the client.
285 : PageRequestHandler,
286 :
287 : /// Manages the WAL receiver connection for one timeline.
288 : /// It subscribes to events from storage_broker and decides which safekeeper to connect to.
289 : /// Once the decision has been made, it establishes the connection using the `tokio-postgres` library.
290 : /// There is at most one connection at any given time.
291 : ///
292 : /// That `tokio-postgres` library represents a connection as two objects: a `Client` and a `Connection`.
293 : /// The `Client` object is what library users use to make requests & get responses.
294 : /// Internally, `Client` hands over requests to the `Connection` object.
295 : /// The `Connection` object is responsible for speaking the wire protocol.
296 : ///
297 : /// Walreceiver uses a legacy abstraction called `TaskHandle` to represent the activity of establishing and handling a connection.
298 : /// The `WalReceiverManager` task ensures that this `TaskHandle` task does not outlive the `WalReceiverManager` task.
299 : /// For the `RequestContext` that we hand to the TaskHandle, we use the [`WalReceiverConnectionHandler`] task kind.
300 : ///
301 : /// Once the connection is established, the `TaskHandle` task spawns a
302 : /// [`WalReceiverConnectionPoller`] task that is responsible for polling
303 : /// the `Connection` object.
304 : /// A `CancellationToken` created by the `TaskHandle` task ensures
305 : /// that the [`WalReceiverConnectionPoller`] task will cancel soon after as the `TaskHandle` is dropped.
306 : ///
307 : /// [`WalReceiverConnectionHandler`]: Self::WalReceiverConnectionHandler
308 : /// [`WalReceiverConnectionPoller`]: Self::WalReceiverConnectionPoller
309 : WalReceiverManager,
310 :
311 : /// The `TaskHandle` task that executes `handle_walreceiver_connection`.
312 : /// See the comment on [`WalReceiverManager`].
313 : ///
314 : /// [`WalReceiverManager`]: Self::WalReceiverManager
315 : WalReceiverConnectionHandler,
316 :
317 : /// The task that polls the `tokio-postgres::Connection` object.
318 : /// Spawned by task [`WalReceiverConnectionHandler`](Self::WalReceiverConnectionHandler).
319 : /// See the comment on [`WalReceiverManager`](Self::WalReceiverManager).
320 : WalReceiverConnectionPoller,
321 :
322 : // Garbage collection worker. One per tenant
323 : GarbageCollector,
324 :
325 : // Compaction. One per tenant.
326 : Compaction,
327 :
328 : // Eviction. One per timeline.
329 : Eviction,
330 :
331 : // Ingest housekeeping (flushing ephemeral layers on time threshold or disk pressure)
332 : IngestHousekeeping,
333 :
334 : /// See [`crate::disk_usage_eviction_task`].
335 : DiskUsageEviction,
336 :
337 : /// See [`crate::tenant::secondary`].
338 : SecondaryDownloads,
339 :
340 : /// See [`crate::tenant::secondary`].
341 : SecondaryUploads,
342 :
343 : // Initial logical size calculation
344 : InitialLogicalSizeCalculation,
345 :
346 : OndemandLogicalSizeCalculation,
347 :
348 : // Task that flushes frozen in-memory layers to disk
349 : LayerFlushTask,
350 :
351 : // Task that uploads a file to remote storage
352 : RemoteUploadTask,
353 :
354 : // task that handles the initial downloading of all tenants
355 : InitialLoad,
356 :
357 : // task that handles attaching a tenant
358 : Attach,
359 :
360 : // Used mostly for background deletion from s3
361 : TimelineDeletionWorker,
362 :
363 : // task that handhes metrics collection
364 : MetricsCollection,
365 :
366 : // task that drives downloading layers
367 : DownloadAllRemoteLayers,
368 : // Task that calculates synthetis size for all active tenants
369 : CalculateSyntheticSize,
370 :
371 : // A request that comes in via the pageserver HTTP API.
372 : MgmtRequest,
373 :
374 : DebugTool,
375 :
376 : EphemeralFilePreWarmPageCache,
377 :
378 : LayerDownload,
379 :
380 : #[cfg(test)]
381 : UnitTest,
382 :
383 : DetachAncestor,
384 : }
385 :
386 : #[derive(Default)]
387 : struct MutableTaskState {
388 : /// Handle for waiting for the task to exit. It can be None, if the
389 : /// the task has already exited.
390 : join_handle: Option<JoinHandle<()>>,
391 : }
392 :
393 : struct PageServerTask {
394 : task_id: PageserverTaskId,
395 :
396 : kind: TaskKind,
397 :
398 : name: String,
399 :
400 : // To request task shutdown, just cancel this token.
401 : cancel: CancellationToken,
402 :
403 : /// Tasks may optionally be launched for a particular tenant/timeline, enabling
404 : /// later cancelling tasks for that tenant/timeline in [`shutdown_tasks`]
405 : tenant_shard_id: TenantShardId,
406 : timeline_id: Option<TimelineId>,
407 :
408 : mutable: Mutex<MutableTaskState>,
409 : }
410 :
411 : /// Launch a new task
412 : /// Note: if shutdown_process_on_error is set to true failure
413 : /// of the task will lead to shutdown of entire process
414 10774 : pub fn spawn<F>(
415 10774 : runtime: &tokio::runtime::Handle,
416 10774 : kind: TaskKind,
417 10774 : tenant_shard_id: TenantShardId,
418 10774 : timeline_id: Option<TimelineId>,
419 10774 : name: &str,
420 10774 : future: F,
421 10774 : ) -> PageserverTaskId
422 10774 : where
423 10774 : F: Future<Output = anyhow::Result<()>> + Send + 'static,
424 10774 : {
425 10774 : let cancel = CancellationToken::new();
426 10774 : let task_id = NEXT_TASK_ID.fetch_add(1, Ordering::Relaxed);
427 10774 : let task = Arc::new(PageServerTask {
428 10774 : task_id: PageserverTaskId(task_id),
429 10774 : kind,
430 10774 : name: name.to_string(),
431 10774 : cancel: cancel.clone(),
432 10774 : tenant_shard_id,
433 10774 : timeline_id,
434 10774 : mutable: Mutex::new(MutableTaskState { join_handle: None }),
435 10774 : });
436 10774 :
437 10774 : TASKS.lock().unwrap().insert(task_id, Arc::clone(&task));
438 10774 :
439 10774 : let mut task_mut = task.mutable.lock().unwrap();
440 10774 :
441 10774 : let task_name = name.to_string();
442 10774 : let task_cloned = Arc::clone(&task);
443 10774 : let join_handle = runtime.spawn(task_wrapper(
444 10774 : task_name,
445 10774 : task_id,
446 10774 : task_cloned,
447 10774 : cancel,
448 10774 : future,
449 10774 : ));
450 10774 : task_mut.join_handle = Some(join_handle);
451 10774 : drop(task_mut);
452 10774 :
453 10774 : // The task is now running. Nothing more to do here
454 10774 : PageserverTaskId(task_id)
455 10774 : }
456 :
457 : /// This wrapper function runs in a newly-spawned task. It initializes the
458 : /// task-local variables and calls the payload function.
459 10774 : async fn task_wrapper<F>(
460 10774 : task_name: String,
461 10774 : task_id: u64,
462 10774 : task: Arc<PageServerTask>,
463 10774 : shutdown_token: CancellationToken,
464 10774 : future: F,
465 10774 : ) where
466 10774 : F: Future<Output = anyhow::Result<()>> + Send + 'static,
467 10774 : {
468 10283 : debug!("Starting task '{}'", task_name);
469 :
470 : // wrap the future so we log panics and errors
471 10283 : let tenant_shard_id = task.tenant_shard_id;
472 10283 : let timeline_id = task.timeline_id;
473 10283 : let fut = async move {
474 : // We use AssertUnwindSafe here so that the payload function
475 : // doesn't need to be UnwindSafe. We don't do anything after the
476 : // unwinding that would expose us to unwind-unsafe behavior.
477 170779 : let result = AssertUnwindSafe(future).catch_unwind().await;
478 8677 : match result {
479 : Ok(Ok(())) => {
480 8677 : debug!("Task '{}' exited normally", task_name);
481 : }
482 0 : Ok(Err(err)) => {
483 0 : error!(
484 0 : "Task '{}' tenant_shard_id: {:?}, timeline_id: {:?} exited with error: {:?}",
485 : task_name, tenant_shard_id, timeline_id, err
486 : );
487 : }
488 0 : Err(err) => {
489 0 : error!(
490 0 : "Task '{}' tenant_shard_id: {:?}, timeline_id: {:?} panicked: {:?}",
491 : task_name, tenant_shard_id, timeline_id, err
492 : );
493 : }
494 : }
495 8677 : };
496 :
497 : // add the task-locals
498 10283 : let fut = CURRENT_TASK.scope(task, fut);
499 10283 : let fut = SHUTDOWN_TOKEN.scope(shutdown_token, fut);
500 10283 :
501 10283 : // poll future to completion
502 170779 : fut.await;
503 :
504 : // Remove our entry from the global hashmap.
505 8677 : TASKS
506 8677 : .lock()
507 8677 : .unwrap()
508 8677 : .remove(&task_id)
509 8677 : .expect("no task in registry");
510 8677 : }
511 :
512 0 : pub async fn exit_on_panic_or_error<T, E>(
513 0 : task_name: &'static str,
514 0 : future: impl Future<Output = Result<T, E>>,
515 0 : ) -> T
516 0 : where
517 0 : E: std::fmt::Debug,
518 0 : {
519 : // We use AssertUnwindSafe here so that the payload function
520 : // doesn't need to be UnwindSafe. We don't do anything after the
521 : // unwinding that would expose us to unwind-unsafe behavior.
522 0 : let result = AssertUnwindSafe(future).catch_unwind().await;
523 0 : match result {
524 0 : Ok(Ok(val)) => val,
525 0 : Ok(Err(err)) => {
526 0 : error!(
527 : task_name,
528 0 : "Task exited with error, exiting process: {err:?}"
529 : );
530 0 : std::process::exit(1);
531 : }
532 0 : Err(panic_obj) => {
533 0 : error!(task_name, "Task panicked, exiting process: {panic_obj:?}");
534 0 : std::process::exit(1);
535 : }
536 : }
537 0 : }
538 :
539 : /// Signal and wait for tasks to shut down.
540 : ///
541 : ///
542 : /// The arguments are used to select the tasks to kill. Any None arguments are
543 : /// ignored. For example, to shut down all WalReceiver tasks:
544 : ///
545 : /// shutdown_tasks(Some(TaskKind::WalReceiver), None, None)
546 : ///
547 : /// Or to shut down all tasks for given timeline:
548 : ///
549 : /// shutdown_tasks(None, Some(tenant_shard_id), Some(timeline_id))
550 : ///
551 66 : pub async fn shutdown_tasks(
552 66 : kind: Option<TaskKind>,
553 66 : tenant_shard_id: Option<TenantShardId>,
554 66 : timeline_id: Option<TimelineId>,
555 66 : ) {
556 66 : let mut victim_tasks = Vec::new();
557 66 :
558 66 : {
559 66 : let tasks = TASKS.lock().unwrap();
560 66 : for task in tasks.values() {
561 52 : if (kind.is_none() || Some(task.kind) == kind)
562 26 : && (tenant_shard_id.is_none() || Some(task.tenant_shard_id) == tenant_shard_id)
563 26 : && (timeline_id.is_none() || task.timeline_id == timeline_id)
564 18 : {
565 18 : task.cancel.cancel();
566 18 : victim_tasks.push((
567 18 : Arc::clone(task),
568 18 : task.kind,
569 18 : task.tenant_shard_id,
570 18 : task.timeline_id,
571 18 : ));
572 34 : }
573 : }
574 : }
575 :
576 66 : let log_all = kind.is_none() && tenant_shard_id.is_none() && timeline_id.is_none();
577 :
578 84 : for (task, task_kind, tenant_shard_id, timeline_id) in victim_tasks {
579 18 : let join_handle = {
580 18 : let mut task_mut = task.mutable.lock().unwrap();
581 18 : task_mut.join_handle.take()
582 : };
583 18 : if let Some(mut join_handle) = join_handle {
584 18 : if log_all {
585 : // warn to catch these in tests; there shouldn't be any
586 0 : warn!(name = task.name, tenant_shard_id = ?tenant_shard_id, timeline_id = ?timeline_id, kind = ?task_kind, "stopping left-over");
587 18 : }
588 18 : if tokio::time::timeout(std::time::Duration::from_secs(1), &mut join_handle)
589 18 : .await
590 18 : .is_err()
591 : {
592 : // allow some time to elapse before logging to cut down the number of log
593 : // lines.
594 0 : info!("waiting for task {} to shut down", task.name);
595 : // we never handled this return value, but:
596 : // - we don't deschedule which would lead to is_cancelled
597 : // - panics are already logged (is_panicked)
598 : // - task errors are already logged in the wrapper
599 0 : let _ = join_handle.await;
600 0 : info!("task {} completed", task.name);
601 18 : }
602 0 : } else {
603 0 : // Possibly one of:
604 0 : // * The task had not even fully started yet.
605 0 : // * It was shut down concurrently and already exited
606 0 : }
607 : }
608 66 : }
609 :
610 0 : pub fn current_task_kind() -> Option<TaskKind> {
611 0 : CURRENT_TASK.try_with(|ct| ct.kind).ok()
612 0 : }
613 :
614 0 : pub fn current_task_id() -> Option<PageserverTaskId> {
615 0 : CURRENT_TASK.try_with(|ct| ct.task_id).ok()
616 0 : }
617 :
618 : /// A Future that can be used to check if the current task has been requested to
619 : /// shut down.
620 0 : pub async fn shutdown_watcher() {
621 0 : let token = SHUTDOWN_TOKEN
622 0 : .try_with(|t| t.clone())
623 0 : .expect("shutdown_watcher() called in an unexpected task or thread");
624 0 :
625 0 : token.cancelled().await;
626 0 : }
627 :
628 : /// Clone the current task's cancellation token, which can be moved across tasks.
629 : ///
630 : /// When the task which is currently executing is shutdown, the cancellation token will be
631 : /// cancelled. It can however be moved to other tasks, such as `tokio::task::spawn_blocking` or
632 : /// `tokio::task::JoinSet::spawn`.
633 9059 : pub fn shutdown_token() -> CancellationToken {
634 9059 : let res = SHUTDOWN_TOKEN.try_with(|t| t.clone());
635 9059 :
636 9059 : if cfg!(test) {
637 : // in tests this method is called from non-taskmgr spawned tasks, and that is all ok.
638 9059 : res.unwrap_or_default()
639 : } else {
640 0 : res.expect("shutdown_token() called in an unexpected task or thread")
641 : }
642 9059 : }
643 :
644 : /// Has the current task been requested to shut down?
645 0 : pub fn is_shutdown_requested() -> bool {
646 0 : if let Ok(true_or_false) = SHUTDOWN_TOKEN.try_with(|t| t.is_cancelled()) {
647 0 : true_or_false
648 : } else {
649 0 : if !cfg!(test) {
650 0 : warn!("is_shutdown_requested() called in an unexpected task or thread");
651 0 : }
652 0 : false
653 : }
654 0 : }
|
