taskvisor 0.5.0

Task supervisor for Tokio: restarts background tasks on failure with exponential backoff and jitter, graceful shutdown, and lifecycle events
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
//! # Task completion outcomes.
//!
//! This module defines:
//! - [`TaskOutcome`]: the final result of one supervised task run,
//! - [`TaskWaiter`]: an awaitable receiver for that result.
//!
//! Outcomes are delivered on the completion plane, not through the event bus.
//! This means they are not affected by broadcast lag or dropped subscriber events.
//!
//! ## Completion Plane
//!
//! ```text
//! add_and_watch(spec)
//!   creates oneshot
//!   sends Add(id, spec, done) to registry
//!   returns TaskWaiter after registration is confirmed
//!
//! submit_and_watch(spec)
//!   creates oneshot
//!   parks sender in controller until admission or rejection
//!   returns TaskWaiter immediately after controller accepts the submission
//!
//! actor finishes
//!   registry joins actor
//!   registry maps actor exit to TaskOutcome
//!   registry resolves oneshot
//! ```
//!
//! ## Resolution Rules
//!
//! - One waiter observes one task identity.
//! - The outcome is resolved through a `oneshot` channel.
//! - The sender has one owner at a time.
//! - Dropping [`TaskWaiter`] is safe. Sending the outcome then becomes a no-op.
//! - The outcome describes the final actor result, after retries are finished.
//! - Per-attempt progress is reported by lifecycle events.
//!
//! ## Rejection
//!
//! [`TaskOutcome::Rejected`] means the task body never ran.
//!
//! It is normally observed from `submit_and_watch` (feature = `controller`), when the controller rejects a submission before it becomes a running task.
//! A duplicate-name registration failure is also resolved as `Rejected` inside the registry, but direct [`add_and_watch`](crate::SupervisorHandle::add_and_watch)
//! returns [`RuntimeError::TaskAlreadyExists`](crate::RuntimeError::TaskAlreadyExists) before handing the waiter to the caller.

use std::sync::Arc;

use tokio::sync::oneshot;

use crate::error::{RuntimeError, SharedError};
use crate::identity::TaskId;

/// Final result of a supervised task run.
///
/// This is the authoritative terminal result for a watched task.
/// It is delivered after the actor retry loop has ended and the registry has joined the actor.
///
/// Use events for live progress.
/// Use `TaskOutcome` when you need the final result.
///
/// ## Outcome vs Events
///
/// Events are best-effort observability.
/// They may be dropped under load, and one event kind can represent several final outcomes.
/// For example,
/// `ActorExhausted` can mean success, retry exhaustion, or task-reported cancellation depending on its reason.
/// `TaskOutcome` is delivered through a `oneshot` channel and is not affected by event-bus lag.
///
/// | Outcome                              | Typical event path | Meaning |
/// |--------------------------------------|------------------------------------------------------|------------------------------------------------------|
/// | [`Completed`](Self::Completed)       | `TaskStopped`, `ActorExhausted`, `TaskRemoved`       | success under `Never` or `OnFailure`                 |
/// | [`Failed`](Self::Failed)             | `TaskFailed`, `ActorExhausted`, `TaskRemoved`        | retryable/non-fatal error stopped after policy limit |
/// | [`Fatal`](Self::Fatal)               | `TaskFailed`, `ActorDead`, `TaskRemoved`             | fatal task error stopped immediately                 |
/// | [`Canceled`](Self::Canceled)         | `TaskCanceled`, `TaskRemoved`                        | cooperative cancellation                             |
/// | [`ForceAborted`](Self::ForceAborted) | `TaskRemoved(reason="force_terminated_after_grace")` | task did not stop in time                            |
/// | [`Panicked`](Self::Panicked)         | `ActorDead(reason="actor_panic")`, `TaskRemoved`     | actor task panicked                                  |
/// | [`Rejected`](Self::Rejected)         | `ControllerRejected` or `TaskAddFailed`              | task body never ran                                  |
///
/// # Also
///
/// - [`SupervisorHandle::add_and_watch`](crate::SupervisorHandle::add_and_watch) - direct watched task add
#[cfg_attr(
    feature = "controller",
    doc = "- [`SupervisorHandle::submit_and_watch`](crate::SupervisorHandle::submit_and_watch) - controller watched submission"
)]
/// - [`TaskWaiter`] - awaitable handle that returns this outcome
/// - [`EventKind`](crate::EventKind) - live observability events
#[non_exhaustive]
#[derive(Debug, Clone)]
pub enum TaskOutcome {
    /// Final attempt succeeded and restart policy stopped the actor.
    Completed,

    /// Final attempt failed and the actor stopped without a fatal error.
    ///
    /// Occurs when:
    /// - `RestartPolicy::Never` does not allow a retry,
    /// - the error is not retryable,
    /// - the retry budget is used up.
    #[non_exhaustive]
    Failed {
        /// Final failure message. Same text as the `ActorExhausted` event reason.
        reason: Arc<str>,
        /// Numeric exit code from a process-like task, if any.
        exit_code: Option<i32>,
        /// Original error source from the final [`TaskError`](crate::TaskError), if any.
        source: Option<SharedError>,
    },

    /// Task returned [`TaskError::Fatal`](crate::TaskError::Fatal).
    ///
    /// Fatal errors are not retried.
    #[non_exhaustive]
    Fatal {
        /// Fatal error message. Same text as the `ActorDead` event reason.
        reason: Arc<str>,
        /// Numeric exit code from a process-like task, if any.
        exit_code: Option<i32>,
        /// Original error source from the fatal [`TaskError`](crate::TaskError), if any.
        source: Option<SharedError>,
    },

    /// Task stopped because of cooperative cancellation.
    ///
    /// This can come from shutdown, explicit remove/cancel, or the task itself returning [`TaskError::Canceled`](crate::TaskError::Canceled).
    Canceled,

    /// Task ignored cooperative cancellation and was aborted after the grace period.
    ForceAborted,

    /// The actor itself panicked.
    ///
    /// This is a runtime bug guard.
    /// Panics inside the task body are caught by `run_once` and become retryable task failures instead.
    Panicked,

    /// The task body never ran.
    ///
    /// Common reasons:
    /// - controller slot was busy under `DropIfRunning`,
    /// - controller queue was full,
    /// - queued submission was replaced,
    /// - queued submission was removed,
    /// - controller was shutting down,
    /// - registration failed because the task name already existed.
    Rejected {
        /// Why the submission was rejected.
        reason: Arc<str>,
    },
}

impl TaskOutcome {
    /// Returns `true` only for [`Completed`](Self::Completed).
    #[must_use]
    pub fn is_success(&self) -> bool {
        matches!(self, TaskOutcome::Completed)
    }

    /// Creates a [`Failed`](Self::Failed) outcome for tests.
    ///
    /// Real outcomes normally come from the runtime.
    /// The `Failed` and `Fatal` variants are `#[non_exhaustive]`;
    /// other crates cannot build them directly.
    ///
    /// This helper lets tests cover code that handles failed outcomes; `source` is `None`.
    ///
    /// ```rust
    /// use taskvisor::TaskOutcome;
    ///
    /// let outcome = TaskOutcome::failed_for_tests("boom", Some(3));
    /// assert!(!outcome.is_success());
    /// ```
    #[cfg(feature = "test-util")]
    #[cfg_attr(docsrs, doc(cfg(feature = "test-util")))]
    #[must_use]
    pub fn failed_for_tests(reason: impl Into<Arc<str>>, exit_code: Option<i32>) -> Self {
        Self::Failed {
            reason: reason.into(),
            exit_code,
            source: None,
        }
    }

    /// Creates a [`Fatal`](Self::Fatal) outcome for tests.
    ///
    /// See [`failed_for_tests`](Self::failed_for_tests) for why this helper exists; `source` is `None`.
    #[cfg(feature = "test-util")]
    #[cfg_attr(docsrs, doc(cfg(feature = "test-util")))]
    #[must_use]
    pub fn fatal_for_tests(reason: impl Into<Arc<str>>, exit_code: Option<i32>) -> Self {
        Self::Fatal {
            reason: reason.into(),
            exit_code,
            source: None,
        }
    }

    /// Returns the original error source for [`Failed`](Self::Failed) or [`Fatal`](Self::Fatal).
    ///
    /// Returns `None` when the outcome has no source error.
    /// This allows callers to use `downcast_ref` or error-reporting crates on the completion plane.
    #[must_use]
    pub fn source(&self) -> Option<&(dyn std::error::Error + 'static)> {
        match self {
            TaskOutcome::Failed { source, .. } | TaskOutcome::Fatal { source, .. } => {
                source.as_ref().map(|e| {
                    let e: &(dyn std::error::Error + 'static) = e.as_ref();
                    e
                })
            }
            _ => None,
        }
    }

    /// Returns a stable machine-readable label.
    ///
    /// Useful for logs, metrics, and telemetry.
    #[must_use]
    pub fn as_label(&self) -> &'static str {
        match self {
            TaskOutcome::Completed => "outcome_completed",
            TaskOutcome::Failed { .. } => "outcome_failed",
            TaskOutcome::Fatal { .. } => "outcome_fatal",
            TaskOutcome::Canceled => "outcome_canceled",
            TaskOutcome::ForceAborted => "outcome_force_aborted",
            TaskOutcome::Panicked => "outcome_panicked",
            TaskOutcome::Rejected { .. } => "outcome_rejected",
        }
    }
}

/// Awaitable handle for one task outcome.
///
/// Created by:
/// - [`SupervisorHandle::add_and_watch`](crate::SupervisorHandle::add_and_watch)
#[cfg_attr(
    feature = "controller",
    doc = "- [`SupervisorHandle::submit_and_watch`](crate::SupervisorHandle::submit_and_watch)"
)]
///
/// A waiter is consumed by [`wait`](Self::wait).
/// It resolves once the watched task reaches a terminal outcome, or returns an error if the sender is dropped before an outcome is produced.
///
/// ## Example
///
/// ```rust,no_run
/// # use taskvisor::prelude::*;
/// # #[tokio::main] async fn main() -> Result<(), Box<dyn std::error::Error>> {
/// # let sup = Supervisor::new(SupervisorConfig::default(), vec![]);
/// # let handle = sup.serve();
/// let job: TaskRef = TaskFn::arc("job", |_ctx| async {
///     Ok(())
/// });
///
/// let (id, waiter) = handle
///     .add_and_watch(TaskSpec::once(job))
///     .await?;
///
/// match waiter.wait().await? {
///     TaskOutcome::Completed => println!("{id} completed"),
///     other => eprintln!("{id} ended with {other:?}"),
/// }
/// # Ok(()) }
/// ```
#[derive(Debug)]
#[must_use = "a TaskWaiter does nothing unless awaited via `.wait()`"]
pub struct TaskWaiter {
    id: TaskId,
    rx: oneshot::Receiver<TaskOutcome>,
}

impl TaskWaiter {
    /// Creates a waiter for one task identity.
    pub(crate) fn new(id: TaskId, rx: oneshot::Receiver<TaskOutcome>) -> Self {
        Self { id, rx }
    }

    /// Returns the task identity observed by this waiter.
    #[must_use]
    pub fn id(&self) -> TaskId {
        self.id
    }

    /// Waits until the task reaches a final outcome.
    ///
    /// During normal operation, this returns a [`TaskOutcome`].
    /// During normal shutdown, tasks resolve to [`TaskOutcome::Canceled`] or [`TaskOutcome::ForceAborted`].
    ///
    /// # Errors
    ///
    /// Returns [`RuntimeError::ShuttingDown`] if the runtime drops the sending half before producing an outcome.
    /// This means the watched task was never resolved by the registry/controller path.
    pub async fn wait(self) -> Result<TaskOutcome, RuntimeError> {
        self.rx.await.map_err(|_| RuntimeError::ShuttingDown)
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[cfg(feature = "test-util")]
    #[test]
    fn test_constructors_build_the_terminal_failure_variants() {
        let failed = TaskOutcome::failed_for_tests("boom", Some(3));
        assert!(matches!(
            &failed,
            TaskOutcome::Failed { reason, exit_code: Some(3), .. } if reason.as_ref() == "boom"
        ));
        assert!(failed.source().is_none(), "test outcomes carry no source");

        let fatal = TaskOutcome::fatal_for_tests("bad config", None);
        assert!(matches!(
            &fatal,
            TaskOutcome::Fatal { reason, exit_code: None, .. } if reason.as_ref() == "bad config"
        ));
    }

    #[test]
    fn labels_are_stable_and_distinct() {
        let outcomes = [
            TaskOutcome::Completed,
            TaskOutcome::Failed {
                reason: Arc::from("x"),
                exit_code: None,
                source: None,
            },
            TaskOutcome::Fatal {
                reason: Arc::from("x"),
                exit_code: Some(1),
                source: None,
            },
            TaskOutcome::Canceled,
            TaskOutcome::ForceAborted,
            TaskOutcome::Panicked,
            TaskOutcome::Rejected {
                reason: Arc::from("x"),
            },
        ];
        let labels: std::collections::HashSet<&str> =
            outcomes.iter().map(|o| o.as_label()).collect();
        assert_eq!(labels.len(), outcomes.len(), "labels must be distinct");
    }

    #[test]
    fn only_completed_is_success() {
        assert!(TaskOutcome::Completed.is_success());
        assert!(!TaskOutcome::Canceled.is_success());
        assert!(!TaskOutcome::Panicked.is_success());
    }

    #[test]
    fn failed_outcome_exposes_downcastable_source() {
        let io = std::io::Error::new(std::io::ErrorKind::PermissionDenied, "denied");
        let outcome = TaskOutcome::Failed {
            reason: Arc::from("denied"),
            exit_code: None,
            source: Some(Arc::new(io)),
        };

        let src = outcome
            .source()
            .expect("a Failed outcome with a cause must expose its source");
        assert_eq!(
            src.downcast_ref::<std::io::Error>().unwrap().kind(),
            std::io::ErrorKind::PermissionDenied
        );
    }

    #[test]
    fn sourceless_outcomes_report_no_source() {
        assert!(TaskOutcome::Completed.source().is_none());
        assert!(
            TaskOutcome::Failed {
                reason: Arc::from("plain"),
                exit_code: Some(1),
                source: None,
            }
            .source()
            .is_none()
        );
    }

    #[tokio::test]
    async fn waiter_resolves_with_sent_outcome() {
        let (tx, rx) = oneshot::channel();
        let waiter = TaskWaiter::new(TaskId::next(), rx);
        tx.send(TaskOutcome::Completed).unwrap();
        assert!(matches!(
            waiter.wait().await.unwrap(),
            TaskOutcome::Completed
        ));
    }

    #[tokio::test]
    async fn waiter_maps_dropped_sender_to_shutting_down() {
        let (tx, rx) = oneshot::channel::<TaskOutcome>();
        let waiter = TaskWaiter::new(TaskId::next(), rx);
        drop(tx);
        assert!(matches!(
            waiter.wait().await,
            Err(RuntimeError::ShuttingDown)
        ));
    }
}