conductor-sdk 0.2.0

Rust SDK for Netflix Conductor workflow orchestration
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
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
// Copyright {{.Year}} Conductor OSS
// Licensed under the Apache License, Version 2.0. See LICENSE in the project root for license information.

use std::sync::Arc;
use std::time::{Duration, Instant};
use tokio::task::JoinHandle;
use tracing::{debug, error, info, warn};

use crate::client::{ConductorClient, MetadataClient, SchemaClient, TaskClient};
use crate::configuration::{resolve_worker_config, Configuration, WorkerConfig};
use crate::error::{ConductorError, Result};
use crate::events::{EventDispatcher, TaskRunnerEventsListener};
use crate::http::ApiClient;
use crate::metrics::{MetricsCollector, MetricsSettings};
use crate::models::{SchemaDef, TaskDef};

use super::{TaskRunner, Worker};

/// Task handler for managing multiple workers.
pub struct TaskHandler {
    #[expect(dead_code)] // Stored for potential future use (e.g., reconfiguration)
    config: Configuration,
    api_client: ApiClient,
    event_dispatcher: EventDispatcher,
    workers: Vec<Arc<dyn Worker>>,
    runners: Vec<Arc<TaskRunner>>,
    handles: Vec<JoinHandle<()>>,
    metrics_collector: Option<Arc<MetricsCollector>>,
    metrics_handle: Option<JoinHandle<()>>,
}

impl TaskHandler {
    /// Create a new task handler with the given configuration.
    ///
    /// # Errors
    ///
    /// Returns [`crate::error::ConductorError::Http`] if the underlying `reqwest` client fails to build (see [`ApiClient::new`]).
    pub fn new(config: Configuration) -> Result<Self> {
        let api_client = ApiClient::new(config.clone())?;
        let event_dispatcher = EventDispatcher::new();

        Ok(Self {
            config,
            api_client,
            event_dispatcher,
            workers: Vec::new(),
            runners: Vec::new(),
            handles: Vec::new(),
            metrics_collector: None,
            metrics_handle: None,
        })
    }

    /// Create a builder for more flexible configuration.
    #[must_use]
    pub fn builder(config: Configuration) -> TaskHandlerBuilder {
        TaskHandlerBuilder::new(config)
    }

    /// Add a worker to the handler.
    ///
    /// Registration is always explicit -- there is no scan-a-package auto-discovery here.
    /// Rust's callers already write out every `TaskDef`/`AgentDef`/`ToolDef` by hand at a
    /// well-typed call site, so scanning for them would only add indirection, not remove real
    /// boilerplate. A crate-level `inventory` dependency was carried for a while in anticipation
    /// of a `#[worker]`-style attribute macro doing `inventory::submit!`-based collection
    /// instead, but was removed unused: besides never having a caller, that approach has a real
    /// sharp edge `inventory` itself documents -- registrations only get linked in if the
    /// registering crate is a *direct* dependency of the final binary, which link-time
    /// dead-code elimination can silently defeat depending on build settings.
    /// `add_worker`/`add_workers` (and [`TaskHandlerBuilder::worker`]/
    /// [`TaskHandlerBuilder::workers`]) staying explicit sidesteps that class of bug entirely.
    pub fn add_worker(&mut self, worker: impl Worker + 'static) {
        self.workers.push(Arc::new(worker));
    }

    /// Add multiple workers.
    pub fn add_workers(&mut self, workers: impl IntoIterator<Item = Arc<dyn Worker>>) {
        self.workers.extend(workers);
    }

    /// Register an event listener.
    pub fn add_event_listener(&self, listener: Arc<dyn TaskRunnerEventsListener>) {
        self.event_dispatcher.register(listener);
    }

    /// Enable metrics collection and (optionally) the HTTP scrape endpoint.
    ///
    /// Creates a [`MetricsCollector`] with the given settings and wires it
    /// as both the [`TaskRunnerEventsListener`] (for task/workflow counters
    /// and histograms) and the
    /// [`HttpMetricsObserver`](crate::http::HttpMetricsObserver) (for
    /// `http_api_client_request_seconds`).
    ///
    /// Must be called **before** [`start`](Self::start). Clients vended
    /// after this call (via [`conductor_client`](Self::conductor_client),
    /// [`task_client`](Self::task_client), etc.) will share the observer.
    pub fn enable_metrics(&mut self, settings: MetricsSettings) {
        let collector = Arc::new(MetricsCollector::new(settings));
        self.event_dispatcher
            .register(Arc::clone(&collector) as Arc<dyn TaskRunnerEventsListener>);
        self.api_client
            .set_http_observer(Arc::clone(&collector) as Arc<dyn crate::http::HttpMetricsObserver>);
        self.metrics_collector = Some(collector);
    }

    /// Get the metrics collector.
    #[must_use]
    pub fn metrics_collector(&self) -> Option<&Arc<MetricsCollector>> {
        self.metrics_collector.as_ref()
    }

    /// Get the event dispatcher.
    #[must_use]
    pub fn event_dispatcher(&self) -> &EventDispatcher {
        &self.event_dispatcher
    }

    /// Get a task client.
    #[must_use]
    pub fn task_client(&self) -> TaskClient {
        TaskClient::new(self.api_client.clone())
    }

    /// Get a metadata client.
    #[must_use]
    pub fn metadata_client(&self) -> MetadataClient {
        MetadataClient::new(self.api_client.clone())
    }

    /// Get the Conductor client wired to this handler's event dispatcher.
    ///
    /// Sharing the dispatcher means `WorkflowStarted` /
    /// `WorkflowStartFailure` events emitted by the returned client's
    /// `WorkflowClient` flow into the same `MetricsCollector` that
    /// [`enable_metrics`](Self::enable_metrics) installed.
    #[must_use]
    pub fn conductor_client(&self) -> ConductorClient {
        ConductorClient::from_api_client(self.api_client.clone())
            .with_event_dispatcher(self.event_dispatcher.clone())
    }

    /// Get a schema client.
    #[must_use]
    pub fn schema_client(&self) -> SchemaClient {
        SchemaClient::new(self.api_client.clone())
    }

    /// Start all workers.
    ///
    /// # Errors
    ///
    /// Returns [`crate::error::ConductorError::Worker`] if no workers are registered.
    pub async fn start(&mut self) -> Result<()> {
        if self.workers.is_empty() {
            return Err(ConductorError::worker("No workers registered"));
        }

        info!(worker_count = self.workers.len(), "Starting task handler");

        // Start metrics HTTP server if configured
        if let Some(ref collector) = self.metrics_collector {
            self.metrics_handle = collector.start_http_server().await;
        }

        // Register task definitions if configured
        let metadata_client = self.metadata_client();
        for worker in &self.workers {
            // Resolve worker configuration
            let config = self.resolve_worker_config(worker.as_ref());

            // Register task definition if configured
            if config.register_task_def {
                if let Err(e) = self
                    .register_task_definition(&metadata_client, worker.as_ref(), &config)
                    .await
                {
                    warn!(
                        task_type = %worker.task_definition_name(),
                        error = %e,
                        "Failed to register task definition (worker will still start)"
                    );
                }
            }

            debug_log_worker(worker.as_ref(), &config);
        }

        // Create and start task runners
        let task_client = self.task_client();

        for worker in &self.workers {
            let runner = Arc::new(TaskRunner::new(
                Arc::clone(worker),
                task_client.clone(),
                self.event_dispatcher.clone(),
            ));

            self.runners.push(Arc::clone(&runner));

            let handle = tokio::spawn(async move {
                runner.run().await;
            });

            self.handles.push(handle);
        }

        info!(
            runner_count = self.runners.len(),
            "Task handler started successfully"
        );

        Ok(())
    }

    /// Verify that every registered worker's polling task actually started and completed at
    /// least one real poll attempt against the server, within `timeout`. Call this right after
    /// [`TaskHandler::start`].
    ///
    /// This crate's workers are `tokio::spawn`ed futures in the same process, where spawning
    /// itself essentially never fails silently -- so the real risk isn't "did the task get
    /// scheduled," it's "did this worker's task actually reach and complete its first poll,"
    /// which catches an early panic during setup (before or shortly into [`TaskRunner::run`])
    /// or a spawned task that's starved and never reaches the network call.
    /// [`TaskRunner::poll_attempt_count`] is the signal used to tell that apart from
    /// [`TaskRunner::is_running`] merely being `true`, which happens before the first poll is
    /// even attempted.
    ///
    /// No-op (returns `Ok(())` immediately) if [`TaskHandler::start`] hasn't been called yet, or
    /// registered no workers -- there's nothing to verify.
    ///
    /// # Errors
    ///
    /// Returns [`crate::error::ConductorError::Worker`] naming every task type that either
    /// exited before its first poll (its spawned task ended, e.g. via an early panic) or simply
    /// hadn't completed one within `timeout`.
    pub async fn verify_workers_started(&self, timeout: Duration) -> Result<()> {
        if self.runners.is_empty() {
            return Ok(());
        }

        let deadline = Instant::now() + timeout;
        let poll_interval = Duration::from_millis(50);

        loop {
            let not_ready: Vec<String> = self
                .runners
                .iter()
                .zip(&self.handles)
                .filter_map(|(runner, handle)| {
                    if handle.is_finished() {
                        Some(format!(
                            "{} (task exited before its first poll)",
                            runner.task_type()
                        ))
                    } else if runner.poll_attempt_count() == 0 {
                        Some(runner.task_type().to_owned())
                    } else {
                        None
                    }
                })
                .collect();

            if not_ready.is_empty() {
                return Ok(());
            }
            if Instant::now() >= deadline {
                return Err(ConductorError::worker(format!(
                    "Worker startup verification failed: worker(s) never completed a first \
                     poll within {timeout:?}: [{}]. This usually means a worker panicked during \
                     setup, or the async runtime is starved. Check logs and retry start().",
                    not_ready.join(", "),
                )));
            }
            tokio::time::sleep(poll_interval).await;
        }
    }

    /// Stop all workers gracefully.
    ///
    /// This method:
    /// 1. Signals all runners to stop polling
    /// 2. Waits for in-flight tasks to complete (up to 30 seconds)
    /// 3. Stops the metrics server
    ///
    /// # Errors
    ///
    /// Currently always returns `Ok(())`; the `Result` is reserved for future shutdown failures (e.g. a worker task panic is logged, not propagated).
    pub async fn stop(&mut self) -> Result<()> {
        info!("Stopping task handler");

        // Signal all runners to stop
        for runner in &self.runners {
            runner.stop();
        }

        // Wait for all runner handles to complete (with timeout)
        // The runners themselves wait for their spawned tasks
        let shutdown_timeout = std::time::Duration::from_secs(30);
        let start = std::time::Instant::now();

        // Take handles using futures::future::join_all for cleaner handling
        let handles: Vec<_> = std::mem::take(&mut self.handles);

        if !handles.is_empty() {
            let wait_result =
                tokio::time::timeout(shutdown_timeout, futures::future::join_all(handles)).await;

            match wait_result {
                Ok(results) => {
                    for result in results {
                        if let Err(e) = result {
                            error!(error = %e, "Worker task panicked during shutdown");
                        }
                    }
                    debug!(
                        elapsed_ms = start.elapsed().as_millis(),
                        "All workers stopped gracefully"
                    );
                }
                Err(_) => {
                    warn!(
                        timeout_secs = shutdown_timeout.as_secs(),
                        "Timeout waiting for workers to stop, some tasks may still be running"
                    );
                }
            }
        }

        // Stop metrics server
        if let Some(handle) = self.metrics_handle.take() {
            handle.abort();
        }

        info!("Task handler stopped");
        Ok(())
    }

    /// Wait for all workers to complete.
    ///
    /// # Errors
    ///
    /// Currently always returns `Ok(())`; a worker task panic is logged, not propagated.
    pub async fn join(&mut self) -> Result<()> {
        for handle in self.handles.drain(..) {
            if let Err(e) = handle.await {
                error!(error = %e, "Worker task panicked");
            }
        }
        Ok(())
    }

    /// Get the number of registered workers.
    #[must_use]
    pub fn worker_count(&self) -> usize {
        self.workers.len()
    }

    /// Check if all workers are running.
    #[must_use]
    pub fn is_running(&self) -> bool {
        !self.runners.is_empty() && self.runners.iter().all(|r| r.is_running())
    }

    /// Pause a specific worker by task type.
    pub fn pause_worker(&self, task_type: &str) {
        for runner in &self.runners {
            if runner.task_type() == task_type {
                runner.pause();
            }
        }
    }

    /// Resume a specific worker by task type.
    pub fn resume_worker(&self, task_type: &str) {
        for runner in &self.runners {
            if runner.task_type() == task_type {
                runner.resume();
            }
        }
    }

    /// Pause all workers.
    pub fn pause_all(&self) {
        for runner in &self.runners {
            runner.pause();
        }
    }

    /// Resume all workers.
    pub fn resume_all(&self) {
        for runner in &self.runners {
            runner.resume();
        }
    }

    /// Resolve worker configuration from environment variables.
    fn resolve_worker_config(&self, worker: &dyn Worker) -> WorkerConfig {
        let defaults = WorkerConfig {
            task_definition_name: worker.task_definition_name().to_owned(),
            poll_interval: std::time::Duration::from_millis(worker.poll_interval_millis()),
            domain: worker.domain().map(std::borrow::ToOwned::to_owned),
            worker_id: worker.identity(),
            thread_count: worker.thread_count(),
            ..Default::default()
        };

        resolve_worker_config(worker.task_definition_name(), defaults)
    }

    /// Register a task definition for a worker.
    async fn register_task_definition(
        &self,
        metadata_client: &MetadataClient,
        worker: &dyn Worker,
        config: &WorkerConfig,
    ) -> Result<()> {
        let task_name = worker.task_definition_name();

        // Check if task definition exists
        let exists = metadata_client.task_def_exists(task_name).await?;

        if exists && !config.overwrite_task_def {
            debug!(
                task_type = %task_name,
                "Task definition already exists, skipping registration (overwrite_task_def=false)"
            );
            return Ok(());
        }

        // Create task definition from worker
        let task_def = TaskDef::new(task_name)
            .with_description("Task registered by Rust SDK worker")
            .with_runtime_metadata(worker.declared_credentials());

        if exists {
            info!(
                task_type = %task_name,
                "Updating existing task definition"
            );
            metadata_client.update_task_def(&task_def).await?;
        } else {
            info!(
                task_type = %task_name,
                "Registering new task definition"
            );
            metadata_client.register_task_def(&task_def).await?;
        }

        // Register input/output schemas if provided
        self.register_worker_schemas(worker, task_name).await?;

        Ok(())
    }

    /// Register input and output schemas for a worker.
    async fn register_worker_schemas(&self, worker: &dyn Worker, task_name: &str) -> Result<()> {
        let schema_client = self.schema_client();

        // Register input schema if provided
        if let Some(input_schema) = worker.input_schema() {
            let schema_def = SchemaDef::new(format!("{task_name}_input"), 1, input_schema);

            match schema_client.register_schema(&schema_def).await {
                Ok(()) => {
                    info!(
                        task_type = %task_name,
                        schema_name = %schema_def.name,
                        "Registered input schema"
                    );
                }
                Err(e) => {
                    warn!(
                        task_type = %task_name,
                        error = %e,
                        "Failed to register input schema (continuing anyway)"
                    );
                }
            }
        }

        // Register output schema if provided
        if let Some(output_schema) = worker.output_schema() {
            let schema_def = SchemaDef::new(format!("{task_name}_output"), 1, output_schema);

            match schema_client.register_schema(&schema_def).await {
                Ok(()) => {
                    info!(
                        task_type = %task_name,
                        schema_name = %schema_def.name,
                        "Registered output schema"
                    );
                }
                Err(e) => {
                    warn!(
                        task_type = %task_name,
                        error = %e,
                        "Failed to register output schema (continuing anyway)"
                    );
                }
            }
        }

        Ok(())
    }
}

fn debug_log_worker(worker: &dyn Worker, config: &WorkerConfig) {
    info!(
        task_type = %worker.task_definition_name(),
        worker_id = %config.worker_id,
        thread_count = config.thread_count,
        poll_interval_ms = config.poll_interval.as_millis(),
        domain = ?config.domain,
        paused = config.paused,
        register_task_def = config.register_task_def,
        "Conductor Worker registered"
    );
}

/// Builder for `TaskHandler`.
pub struct TaskHandlerBuilder {
    config: Configuration,
    workers: Vec<Arc<dyn Worker>>,
    event_listeners: Vec<Arc<dyn TaskRunnerEventsListener>>,
    metrics_settings: Option<MetricsSettings>,
}

impl TaskHandlerBuilder {
    /// Create a new builder.
    #[must_use]
    pub fn new(config: Configuration) -> Self {
        Self {
            config,
            workers: Vec::new(),
            event_listeners: Vec::new(),
            metrics_settings: None,
        }
    }

    /// Add a worker.
    #[must_use]
    pub fn worker(mut self, worker: impl Worker + 'static) -> Self {
        self.workers.push(Arc::new(worker));
        self
    }

    /// Add multiple workers.
    #[must_use]
    pub fn workers(mut self, workers: impl IntoIterator<Item = Arc<dyn Worker>>) -> Self {
        self.workers.extend(workers);
        self
    }

    /// Add an event listener.
    #[must_use]
    pub fn event_listener(mut self, listener: impl TaskRunnerEventsListener + 'static) -> Self {
        self.event_listeners.push(Arc::new(listener));
        self
    }

    /// Enable metrics with the given settings.
    #[must_use]
    pub fn metrics(mut self, settings: MetricsSettings) -> Self {
        self.metrics_settings = Some(settings);
        self
    }

    /// Build the task handler.
    ///
    /// # Errors
    ///
    /// Returns [`crate::error::ConductorError::Http`] if the underlying `reqwest` client fails to build (see [`TaskHandler::new`]).
    pub fn build(self) -> Result<TaskHandler> {
        let mut handler = TaskHandler::new(self.config)?;

        handler.add_workers(self.workers);

        for listener in self.event_listeners {
            handler.add_event_listener(listener);
        }

        if let Some(settings) = self.metrics_settings {
            handler.enable_metrics(settings);
        }

        Ok(handler)
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::worker::{Worker, WorkerOutput};
    use async_trait::async_trait;

    struct TestWorker {
        name: String,
    }

    #[async_trait]
    impl Worker for TestWorker {
        fn task_definition_name(&self) -> &str {
            &self.name
        }

        async fn execute(&self, _task: &crate::models::Task) -> Result<WorkerOutput> {
            Ok(WorkerOutput::completed_with_result("test"))
        }
    }

    #[test]
    fn test_task_handler_builder() {
        let config = Configuration::new("http://localhost:8080/api");

        let handler = TaskHandlerBuilder::new(config)
            .worker(TestWorker {
                name: "task1".to_owned(),
            })
            .worker(TestWorker {
                name: "task2".to_owned(),
            })
            .build();

        assert!(handler.is_ok());
        let handler = handler.unwrap();
        assert_eq!(handler.worker_count(), 2);
    }

    #[tokio::test]
    async fn test_verify_workers_started_is_a_no_op_before_start() {
        let config = Configuration::new("http://localhost:8080/api");
        let mut handler = TaskHandler::new(config).unwrap();
        handler.add_worker(TestWorker {
            name: "never_started".to_owned(),
        });

        // start() was never called, so there are no runners/handles yet -- nothing to verify.
        handler
            .verify_workers_started(Duration::ZERO)
            .await
            .expect("no-op before start() must not error");
    }

    #[tokio::test]
    async fn test_verify_workers_started_succeeds_after_a_real_poll() {
        let mock_server = wiremock::MockServer::start().await;

        wiremock::Mock::given(wiremock::matchers::method("GET"))
            .and(wiremock::matchers::path(
                "/api/tasks/poll/batch/verify_started_ok",
            ))
            .respond_with(wiremock::ResponseTemplate::new(200).set_body_json(serde_json::json!([])))
            .mount(&mock_server)
            .await;

        let config = Configuration::new(format!("{}/api", mock_server.uri()));
        let mut handler = TaskHandler::new(config).unwrap();
        handler.add_worker(TestWorker {
            name: "verify_started_ok".to_owned(),
        });
        handler.start().await.unwrap();

        handler
            .verify_workers_started(Duration::from_secs(2))
            .await
            .expect("worker should complete a real poll well within 2s");

        // Not calling handler.stop() -- see test_resume_registers_workers_... in
        // agents::runtime for why: the #[tokio::test] runtime aborts the spawned poller when
        // this function returns, which is faster and just as clean here.
    }

    #[tokio::test]
    async fn test_verify_workers_started_times_out_when_poll_never_completes() {
        let mock_server = wiremock::MockServer::start().await;

        // A response slower than the timeout below simulates a worker whose task is legitimately
        // still running but hasn't reached (completed) its first real poll yet.
        wiremock::Mock::given(wiremock::matchers::method("GET"))
            .and(wiremock::matchers::path(
                "/api/tasks/poll/batch/verify_started_stuck",
            ))
            .respond_with(
                wiremock::ResponseTemplate::new(200)
                    .set_body_json(serde_json::json!([]))
                    .set_delay(Duration::from_secs(5)),
            )
            .mount(&mock_server)
            .await;

        let config = Configuration::new(format!("{}/api", mock_server.uri()));
        let mut handler = TaskHandler::new(config).unwrap();
        handler.add_worker(TestWorker {
            name: "verify_started_stuck".to_owned(),
        });
        handler.start().await.unwrap();

        let err = handler
            .verify_workers_started(Duration::from_millis(200))
            .await
            .expect_err("poll response is delayed well past the 200ms timeout");
        assert!(err.to_string().contains("verify_started_stuck"));
    }
}