flower
flower copied to clipboard
Max tasks question
Is the behavior when max_tasks is reached that old tasks are discarded and new ones recorded, or that no new tasks are recorded and the existing state is preserved until a restart?
I'm seeing the 2nd behavior but wondering if this is a bug. Happy to provide more info, but I'm basically just running https://github.com/mavenclinic/celery-flower-redis with basic auth in the environment so nothing too exotic. Thanks!
It depends on celery.events.state.State.
I am experiencing the same issue as @thekantian. After a certain number of tasks are executed, no more tasks are executing. (At least not that I can see from the Flower GUI)
@mozillazg, would you please expound on the celery.events.state.State config? How can we configure our code so that Celery discards any finished tasks over the MAX_TASKS option?
@mozillazg thanks for the pointer.
@jacobbridges to be clear - my celery workers are still executing tasks, just the flower interface does not register them. for now, I solved this with a cron job to restart flower often enough for my purposes - I have other logs that I keep so I can live with the trade-off until I understand the State more clearly - I haven't grokked it yet.
@thekantian ah.. I am running my project within a Docker container so I could not easily check that. Thank you for the update and the temporary solution. Restarting flower with cron, eh? Docker challenge accepted :sunglasses:
Did you try sorting tasks by received/started?
Yeah, and tried filtering as well...
Facing the same issue with flower 0.8.2..
We are running into this issue as well... flower 0.8.2 and redis.
I'm not able to reproduce this issue. Can someone setup a sample project which demonstrates the issue?
+1
it seems i got the same issue.
after read flower docs, i know all the data i see from flower dashbord is maintained by it in memory. and i start wonder how many guys will enable persistent
feature in product env?
is this feature will make query slower and ulimit disk usage growing?