How to monitor Celery

What can go wrong with Celery

• Workers alive but stuck. A task blocks on a network call with no timeout and the worker's slots stay busy. Processes look fine; the queue grows.
• Queue backlog. Producers outpace workers after a traffic spike or a slow dependency. Users wait minutes for emails, exports or payments to process.
• celery beat not running, or running twice. No beat means no periodic tasks at all, with no error. Two beats means every periodic task runs twice.
• Errors swallowed. A task that raises is logged on the worker and retried or dropped; unless something collects the exception, nobody sees it.
• Redis visibility timeout. With the Redis broker, a task not acknowledged within visibility_timeout (1 hour by default) is delivered again, so long tasks with acks_late can run twice.

The signals that matter, with alert thresholds

Thresholds are starting points. Tune them after a week of data: an alert nobody acts on should be raised or removed.

How to check Celery by hand

celery -A proj inspect ping
celery -A proj inspect active

celery -A proj inspect reserved

redis-cli -n 0 llen celery

sudo rabbitmqctl list_queues name messages consumers

systemctl status celerybeat     # or: pgrep -af 'celery.*beat'

How to watch Celery with Vigil

1. Add the host under Servers and run the one-line installer it shows (curl -fsSL https://puzaricloud.in/install-server-agent.sh | sh -s -- --token <token>). The agent is one Python file with no dependencies; CPU, memory, disk, inode and load rules start working with the first report. Add the worker service and beat to the systemd services to watch.
2. Add the commands below: missing workers (limit celery_workers_missing = 1) and queue length (limit e.g. celery_queue = 500). Commands run every interval with a 10-second timeout.
3. For each periodic task that matters, create a Heartbeat in New monitor with the beat schedule, and call the ping URL (e.g. with httpx) at the end of the task. If beat stops, the pings stop.
4. Create a project under Errors and use its DSN with sentry_sdk.init(dsn=…); the Sentry SDK enables its Celery integration automatically, so task exceptions arrive grouped with arguments and breadcrumbs.
5. For task traces, use opentelemetry-instrumentation-celery and send spans to the OTLP endpoint (quick start).

{
  "commands": {
    "celery_workers_missing": "cd /srv/app && echo $(( 4 - $(.venv/bin/celery -A proj inspect ping 2>/dev/null | grep -c pong) ))",
    "celery_queue": "redis-cli -n 0 llen celery"
  }
}

Every alert goes to the channels you pick: email, Slack, Telegram, Discord, Microsoft Teams, Pushover, webhooks or on-call paging. Failures are confirmed by a second check before anyone is woken up.

Celery monitoring FAQ