[task] Deleting all pending tasks in celery / rabbitmq

How can I delete all pending tasks without knowing the task_id for each task?

1. To properly purge the queue of waiting tasks you have to stop all the workers (http://celery.readthedocs.io/en/latest/faq.html#i-ve-purged-messages-but-there-are-still-messages-left-in-the-queue):

$ sudo rabbitmqctl stop

or (in case RabbitMQ/message broker is managed by Supervisor):

$ sudo supervisorctl stop all

2. ...and then purge the tasks from a specific queue:

$ cd <source_dir>
$ celery amqp queue.purge <queue name>

3. Start RabbitMQ:

$ sudo rabbitmqctl start

or (in case RabbitMQ is managed by Supervisor):

$ sudo supervisorctl start all

I found that celery purge doesn't work for my more complex celery config. I use multiple named queues for different purposes:

$ sudo rabbitmqctl list_queues -p celery name messages consumers
Listing queues ...  # Output sorted, whitespaced for readability
celery                                          0   2
[email protected]                      0   1
[email protected]                      0   1
apns                                            0   1
[email protected]                        0   1
analytics                                       1   1
[email protected]                   0   1
bcast.361093f1-de68-46c5-adff-d49ea8f164c0      0   1
bcast.a53632b0-c8b8-46d9-bd59-364afe9998c1      0   1
celeryev.c27b070d-b07e-4e37-9dca-dbb45d03fd54   0   1
celeryev.c66a9bed-84bd-40b0-8fe7-4e4d0c002866   0   1
celeryev.b490f71a-be1a-4cd8-ae17-06a713cc2a99   0   1
celeryev.9d023165-ab4a-42cb-86f8-90294b80bd1e   0   1

The first column is the queue name, the second is the number of messages waiting in the queue, and the third is the number of listeners for that queue. The queues are:

• celery - Queue for standard, idempotent celery tasks
• apns - Queue for Apple Push Notification Service tasks, not quite as idempotent
• analytics - Queue for long running nightly analytics
• *.pidbox - Queue for worker commands, such as shutdown and reset, one per worker (2 celery workers, one apns worker, one analytics worker)
• bcast.* - Broadcast queues, for sending messages to all workers listening to a queue (rather than just the first to grab it)
• celeryev.* - Celery event queues, for reporting task analytics

The analytics task is a brute force tasks that worked great on small data sets, but now takes more than 24 hours to process. Occasionally, something will go wrong and it will get stuck waiting on the database. It needs to be re-written, but until then, when it gets stuck I kill the task, empty the queue, and try again. I detect "stuckness" by looking at the message count for the analytics queue, which should be 0 (finished analytics) or 1 (waiting for last night's analytics to finish). 2 or higher is bad, and I get an email.

celery purge offers to erase tasks from one of the broadcast queues, and I don't see an option to pick a different named queue.

Here's my process:

$ sudo /etc/init.d/celeryd stop  # Wait for analytics task to be last one, Ctrl-C
$ ps -ef | grep analytics  # Get the PID of the worker, not the root PID reported by celery
$ sudo kill <PID>
$ sudo /etc/init.d/celeryd stop  # Confim dead
$ python manage.py celery amqp queue.purge analytics
$ sudo rabbitmqctl list_queues -p celery name messages consumers  # Confirm messages is 0
$ sudo /etc/init.d/celeryd start

In Celery 3+:

CLI:

$ celery -A proj purge

Programatically:

>>> from proj.celery import app
>>> app.control.purge()

http://docs.celeryproject.org/en/latest/faq.html#how-do-i-purge-all-waiting-tasks

If you want to remove all pending tasks and also the active and reserved ones to completely stop Celery, this is what worked for me:

from proj.celery import app
from celery.task.control import inspect, revoke

# remove pending tasks
app.control.purge()

# remove active tasks
i = inspect()
jobs = i.active()
for hostname in jobs:
    tasks = jobs[hostname]
    for task in tasks:
        revoke(task['id'], terminate=True)

# remove reserved tasks
jobs = i.reserved()
for hostname in jobs:
    tasks = jobs[hostname]
    for task in tasks:
        revoke(task['id'], terminate=True)

celery 4+ celery purge command to purge all configured task queues

celery -A *APPNAME* purge

programmatically:

from proj.celery import app
app.control.purge()

all pending task will be purged. Reference: celerydoc

For Celery 2.x and 3.x:

When using worker with -Q parameter to define queues, for example

celery worker -Q queue1,queue2,queue3

then celery purge will not work, because you cannot pass the queue params to it. It will only delete the default queue. The solution is to start your workers with --purge parameter like this:

celery worker -Q queue1,queue2,queue3 --purge

This will however run the worker.

Other option is to use the amqp subcommand of celery

celery amqp queue.delete queue1
celery amqp queue.delete queue2
celery amqp queue.delete queue3

In Celery 3+

http://docs.celeryproject.org/en/3.1/faq.html#how-do-i-purge-all-waiting-tasks

CLI

Purge named queue:

 celery -A proj amqp queue.purge <queue name>

Purge configured queue

celery -A proj purge

I’ve purged messages, but there are still messages left in the queue? Answer: Tasks are acknowledged (removed from the queue) as soon as they are actually executed. After the worker has received a task, it will take some time until it is actually executed, especially if there are a lot of tasks already waiting for execution. Messages that are not acknowledged are held on to by the worker until it closes the connection to the broker (AMQP server). When that connection is closed (e.g. because the worker was stopped) the tasks will be re-sent by the broker to the next available worker (or the same worker when it has been restarted), so to properly purge the queue of waiting tasks you have to stop all the workers, and then purge the tasks using celery.control.purge().

So to purge the entire queue workers must be stopped.

For celery 3.0+:

$ celery purge

To purge a specific queue:

$ celery -Q queue_name purge