Celery memory leak in Tribe results in OOM warm restart

[acherla]

Celery containers appear to be hitting a memory leak, setting container limits to 5GB or 10GB appears to result in all memory being used by the container even when MAX_WORKERS=1 and when process is completely idle. This is likely due to a bug in celery version 5.4.0 which is resulting in a memory leak that causes memory usage to increase incrementally until hitting an OOM.

[acherla]

testing this on python3.11 to see if its only impacting the python 3.10 image

[StreetLamb]

Thanks, will take a look. I dont think its a celery issue. The memory spike seems to be happening right when documents are uploaded to the qdrant collection.

[acherla]

I would normally agree, but this seems to be occuring even on brand new fresh clusters with no documents.

[StreetLamb]

I pushed #133 which helps to alleviate the memory leak issue for me. Let me know if it addresses yours too.

[acherla]

Thanks we did the same thing on our side