There's an issue at the moment (or maybe it was always there?) that means our shutdown_celery_worker() use of .broadcast("shutdown") doesn't work with gevent
celery/celery#4019 celery/celery#5209
Celery Execution Pools: What is it all about? https://www.distributedpython.com/2018/10/26/celery-execution-pool/
Maybe it could be a Kombu thing? That came up a lot in older discussions about this issue.
WORKAROUND: Went back to the old method of killing and restarting tasks (rather than trying to kill the whole worker)
This is a tricky one! Not sure what's causing it.
This old forum thread has a good examination of the same issue: https://groups.google.com/g/celery-users/c/W0Qf09ahjas?pli=1 (In their case it was Flower restoring unacked tasks)
WORKAROUND: is_a_matching_fill_missing_tweets_task_already_running() now kills all duplicate tasks and we've added a button to manually trigger task_fill_missing_tweets()
Fixed in https://github.com/tweepy/tweepy/commit/68e19cc6b9b23d72369ca1520093770eb18a5a9f and slated for the 4.0 release.
Relevant issues: tweepy/tweepy#448 tweepy/tweepy#237
WORKAROUND: Ignore it for now and wait for Tweepy 4.0
Django 3 is partway through migrating to use Python 3's new async functionality. Unfortunately for us, the ORM is hasn't been migrated yet - so we're sometimes seeing these 'sync to async' errors being thrown when Celery tasks call the ORM.
We tried to opt out of seeing warnings by setting DJANGO_ALLOW_ASYNC_UNSAFE=true per https://docs.djangoproject.com/en/3.1/topics/async/#envvar-DJANGO_ALLOW_ASYNC_UNSAFE, but that didn't work and we still kept seeing the errors.
You cannot use AsyncToSync in the same thread as an async event loop - just await the async function directly.
c.f. see the comments here: https://stackoverflow.com/a/43325237/7368493
WORKAROUND: Rollback to the latest in the Django 2.x series