Monitor your queues

You’ll turn on lifecycle events, run the GisilaRose dashboard to watch tasks and workers live, and forward failures to your own alerting.

Prerequisites

Producers and workers on Redis.

1. Turn on events in every process

Events are off by default. With GisilaQueue.redis, pass events: true in producers and workers:

import 'package:gisila_queue/gisila_queue.dart';

final app = GisilaQueue.redis(namespace: 'shop', events: true);

When building components by hand, add events: RedisEventSink(prefix: 'shop'). WorkerCli.buildApp() does not add an event sink, so build the worker app yourself (see Run workers).

2. Run the dashboard

REDIS_URL=redis://:secret@localhost:6379/0 \
  dart run gisila_queue:rose --namespace=shop --port=5555

Open http://localhost:5555. You see recent tasks with state, args, result and timings, workers with their counters and heartbeats, queue depths, and per-task-name stats. Fire, retry, revoke, pause and resume are available because the binary opens a broker and a control channel.

To expose it beyond localhost, require auth:

dart run gisila_queue:rose --namespace=shop --host=0.0.0.0 --basic-auth=admin:long-random-password

Rose refuses a public bind with write actions and no auth. --no-queues --no-control makes it read-only.

3. Use the JSON API

curl -u admin:pw localhost:5555/api/stats
curl -u admin:pw 'localhost:5555/api/tasks?state=failure&limit=20'
curl -u admin:pw -X POST -H 'Content-Type: application/json' \
  -d '{"name":"health.ping"}' localhost:5555/api/tasks/fire
curl -u admin:pw -X POST -H 'Content-Type: application/json' localhost:5555/api/queues/emails/pause

All POSTs need Content-Type: application/json. See the HTTP API table.

4. Check depths without Rose

dart run gisila_queue:gisila_queue inspect --namespace=shop --queues=default,emails,dead_letter

Alert when dead_letter grows or a queue’s depth keeps rising.

5. Forward events to your own systems

Implement QueueEventSink and combine it with the Redis sink:

class AlertSink implements QueueEventSink {
  @override
  Future<void> emit(QueueEvent event) async {
    if (event.type != QueueEventType.taskFailed) return;
    try {
      await alerts.send('${event.taskName} ${event.taskId}: ${event.fields['error']}');
    } catch (_) {
      // never throw from a sink
    }
  }

  @override
  Future<void> close() async {}
}

final app = GisilaQueue(
  broker: RedisBroker(prefix: 'shop'),
  backend: RedisResultBackend(prefix: 'shop'),
  events: MultiEventSink([RedisEventSink(prefix: 'shop'), AlertSink()]),
);

Or read the stream from a separate process with RedisEventSource(prefix: 'shop').events().

Verify it works

Enqueue a task and watch it appear in Rose as pending, then started, then success. Stop a worker and its status turns not alive about 30 seconds after its last heartbeat.