Strange redis key growth, at least in Heroku

[mariusandra]

Bug description

The number of keys in Redis starts to grow quite quickly once an app is launched. 2.6k keys in 40min. It takes just ~1MB of memory, but what are those keys?

It levels out at 73k keys and 25MB memory in each case.

This screenshot is from the session recording PR's heroku instance.

When working on the plugins, I noticed that in an undeployed state with no app running, the keys all expired within 24h.

Expected behavior

There shouldn't be a leak.

How to reproduce

Launch an app on Heroku and have a look. Or check any redis usage of the deployed preview apps.

Environment

• At least any posthog instance running on Heroku, could be wider spread

Additional context

• I think celery saves some "Ack" / "success" messages in the queue after an event is done, to notify of task completion (e.g. await task()).

Thank you for your bug report – we love squashing them!

[mariusandra]

This is what has piled up in ~40 minutes:

ec2-3-210-254-187.compute-1.amazonaws.com:14519> get  celery-task-meta-94f158ef-c2f7-4011-9a42-4c227e0d7b8f
{"status": "SUCCESS", "result": null, "traceback": null, "children": [], "date_done": "2020-10-28T19:49:31.704216", "task_id": "94f158ef-c2f7-4011-9a42-4c227e0d7b8f"}
ec2-3-210-254-187.compute-1.amazonaws.com:14519> get celery-task-meta-3b166121-f9dc-4303-8344-0b03cd311590
{"status": "SUCCESS", "result": null, "traceback": null, "children": [], "date_done": "2020-10-28T19:53:38.840494", "task_id": "3b166121-f9dc-4303-8344-0b03cd311590"}
ec2-3-210-254-187.compute-1.amazonaws.com:14519> get celery-task-meta-c2c4d606-6f21-4bc1-8090-405e6ecbabaf
{"status": "SUCCESS", "result": null, "traceback": null, "children": [], "date_done": "2020-10-28T19:09:54.184765", "task_id": "c2c4d606-6f21-4bc1-8090-405e6ecbabaf"}