Running tasks in the background using Celery

Background tasks are run using Celery, which delegates to Redis.

The project template’s docker-compose.yml file configures redis, celery, celerybeat and celeryflower machines.

Running background tasks

See the Celery, documentation for writing and running background tasks that take place in response to an event.

In development

The Celery-related services are started when you run docker-compose run.

If you’re not running Docker (and have Redis running on your machine) then, in 3 separate tabs, run:

./ # celery task runner
./ # celery task scheduler
./ # celery inspection service

On a server

Docker uses the included docker-compose.yml file to deploy your application with all necessary services, including the Celery-related ones.

For more information, see Deploying your site.

Scheduling tasks

Tasks that need to run regularly are scheduled in crontab-like fashion using CeleryBeat.

You can define your tasks, wherever you like (we suggest in a file called For example:

from celery.task import Task
from import call_command
from icekit.tasks import one_instance

class UpdateSearchIndexTask(Task):
    def run(self, **kwargs):
        call_command('update_index', remove=True)

The one_instance decorator

The icekit.tasks.one_instance decorator uses a Redis token to ensure that only one instance of the function will run at a time. This is useful in case tasks take longer to run than the interval between calls, but means that if tasks crash that the lock won’t be released until settings.DEFAULT_ONE_INSTANCE_TIMEOUT seconds (default 7200) have elapsed.

Then you need to specify the task in a schedule, which is defined in settings.CELERYBEAT_SCHEDULE, in this format:

from celery.schedules import crontab

    'UpdateSearchIndexTask': {
        'task': 'icekit.tasks.UpdateSearchIndexTask',
        'schedule': crontab(minute='*/15'),  # Every 15 minutes.

Add your tasks to the dictionary to include them in the scheduler.


You can use Celery Flower to monitor running tasks. Visit http://localhost:5555 (or http://{{yourproject}} )to see the interface.


Restarting Flower (e.g. on deployment) will clear its records of tasks, so use with caution in production environments.

If Celery tasks fail, they will show up as failures. However, if celery tasks ‘succeed’ but with a very quick runtime (0.02 secs), the task is possibly prevented from running by a one_instance lock.

Using Docker Cloud, you can also check the log (stdout) of the Celery Docker container for actual tasks being executed. Check the logs (stdout) of the Celery Beat container to ensure the scheduled tasks process is running.

Freeing task locks

Locks are stored in Redis as keys with a TTL. You can see the locks in a Python shell with:

>>> import redis
>>> REDIS_CLIENT = redis.Redis()
>>> [x for x in REDIS_CLIENT.keys() if 'Task' in x]
['FetchCollectionDataTask', 'FindAndFetchNetxImagesTask', 'DeleteRemovedNetxImagesTask']
>>> REDIS_CLIENT.pttl('FindAndFetchNetxImagesTask') # ms to live

(note that 'Task' in x part assumes that your CELERYBEAT_SCHEDULE entries have Task in the key name. You can delete a key, thus freeing up the lock, with:

>>> REDIS_CLIENT.delete('FindAndFetchNetxImagesTask')

(a response of 1 means the lock was deleted, a response of 0 means it was not found.)