Replies: 4 comments 5 replies
-
any gays can response? |
Beta Was this translation helpful? Give feedback.
-
@hassant4 This worked for me: #4079 (comment) |
Beta Was this translation helpful? Give feedback.
-
The |
Beta Was this translation helpful? Give feedback.
-
I don't think you need to switch to gevent. The "prefork" pool is the default and won't block when the worker is busy. If you do want to use gevent, set --concurrency to a very high value, and definitely not to one: See https://www.distributedpython.com/2018/10/26/celery-execution-pool/ |
Beta Was this translation helpful? Give feedback.
-
Hi, I run my nodes using the following command:
CUDA_VISIBLE_DEVICES=1 poetry run celery -A ts2 worker --loglevel=INFO --pool gevent --concurrency=1 --prefetch-multiplier=1
Most of my tasks are long running hence prefetch multiplier=1 (and late acknowledge in django settings). When running
poetry run celery -A ts2 inspect ping -d celery@$HOSTNAME
I succesfully get a response ONLY if the worker is idle. If the worker is processing a task I getError: No nodes replied within time constraint
I have read that only the pool=solo mode is blocking, hence why I changed to gevent, but this is still the case. I rely on this command to let kubernetes know that the worker is live/ready (livenessProbe and readinessProbe).
Here is an equivalent stackoverflow question, I don't have enough rep to comment and ask the poster if they solved
Can anyone help (the pre submit for issue suggested asking here first)
Beta Was this translation helpful? Give feedback.
All reactions