Route celery task to specific queue

Question

I have two separate celeryd processes running on my server, managed by supervisor . They are set to listen on separate queues as such:

[program:celeryd1]
command=/path/to/celeryd --pool=solo --queues=queue1
...

[program:celeryd2]
command=/path/to/celeryd --pool=solo --queues=queue2
...

And my celeryconfig looks something like this:

from celery.schedules import crontab

BROKER_URL = "amqp://guest:guest@localhost:5672//"

CELERY_DISABLE_RATE_LIMITS = True
CELERYD_CONCURRENCY = 1
CELERY_IGNORE_RESULT = True

CELERY_DEFAULT_QUEUE = 'default'
CELERY_QUEUES = {
    'default': {
        "exchange": "default",
        "binding_key": "default",
    },
    'queue1': {
        'exchange': 'queue1',
        'routing_key': 'queue1',
    },
    'queue2': {
        'exchange': 'queue2',
        'routing_key': 'queue2',
    },
}

CELERY_IMPORTS = ('tasks', )

CELERYBEAT_SCHEDULE = {
    'first-queue': {
        'task': 'tasks.sync',
        'schedule': crontab(hour=02, minute=00),
        'kwargs': {'client': 'client_1'},
        'options': {'queue': 'queue1'},
    },
    'second-queue': {
        'task': 'tasks.sync',
        'schedule': crontab(hour=02, minute=00),
        'kwargs': {'client': 'client_2'},
        'options': {'queue': 'queue1'},
    },
}

All tasks.sync tasks must be routed to a specific queue (and therefore celeryd progress). But when I try to run the task manually with sync.apply_async(kwargs={'client': 'value'}, queue='queue1') both celery workers pick up the task. How can I make the task route to the correct queue and only be run by the worker that is bound to the queue?

Answer 1

You are only running one celerybeat instance right?

Maybe you have old queue bindings that clash with this? Try running rabbitmqctl list_queues and rabbitmqctl list_bindings , maybe reset the data in the broker to start from scratch.

The example you have here should work, and is working for me when I just tried it.

Tip: Since you are using the same exchange and binding_key value as the queue name, you don't have to explicitly list them in CELERY_QUEUES. When CELERY_CREATE_MISSING_QUEUES is on (which it is by default) the queues will be automatically created exactly like you have if you just do celeryd -Q queue1 or send a task to a queue that is undefined.