#celery

hi guys, any idea how can i link on another queue that the send_task

Does `autoretry_for` only work for function based tasks? It looks like it doesn't apply to class based tasks.

xcm: Have you tried getting the task traceback? Initialize an AsyncResult with the task id and try to access AsyncResult.traceback

ahadi: Usually concurrency is set at the worker level not so much at the task level. What exactly are you trying to do?

ahadi: autoretry_for should also work for function based tasks. You have to give it the correct tuple of exceptions

josuebc: I have some tasks talking to a specific api and the concurrent connections to this API are limited. Thats why I wanted to regulate how many tasks can run concurrently. My solution for now is to have a separate queue with one specific worker that has the right amount of concurrency set.

josuebc: I know that it works for function based tasks, but it doesn't for class based tasks

I have a workaround for that already, as written here: https://github.com/celery/celery/issues/4684

Phrk_: Freneticks: You should be able to create a signature with the routing options: http://docs.celeryproject.org/en/latest/usergui...

ahadi: Ah, there's a way to specify the task rate on the task definition: http://docs.celeryproject.org/en/latest/usergui...

ahadi: Oh, so it's not working for class based views. Is it possible to take a look at the code?

josuebc: I already found the problem. It looks like that the PR that would bring the autoretry_for for class based views was merged after the release of 4.1.0

It should be fixed for 4.2.0

But thank you anyways!

ahadi: Ah, good you found the problem.

josuebc: first of all, thanks so much for reminding me about this issue because i didn't even put it in my bugtracker. i didn't know about traceback, so i'll try it next time this issue crops up

ppp/win 90

xcm: Sure thing, hopefully you can get it resolved

How How can I ensure a revoke(id, terminate=True) call actually closed the task?

If some task fails to close within reasonable time (whatever "reasonable" is), I'm expecting celery to kill it

it looks like it does not do the job. From time to time, the revoke() call is blocking, probably because the task never returns, and that paralyses my task manager permanently

With Py3's multiprocessing library, I sent SIGINT, then join()ed the process with some timeout, and finally send SIGKILL it needed ; the same thing with Py2's multiprocessing can be done via a thread monitoring the termination of the process and sending SIGKILL if necessary, however I was expecting from a library like celery to do that work for me. How can it help handle these cases?