mlflow
mlflow copied to clipboard
[BUG] Memory usage 400MB (very high) - can it be reduced?
Hi,
I installed MLflow in my one node Kubernetes cluster. It needs 400MB of RAM after fresh install and without any load. That is a lot...
A fresh Postgres install needs 34Mi and MinIO only 115Mi.
I am using MLflow 1-23-1 with external Postgres and MinIO. I am starting it with:
command:
- mlflow
- server
- -h
- "0.0.0.0"
- --backend-store-uri
- "postgresql://$(DB_NAME):$(DB_PASSWORD)@$(DB_HOST)/$(DB_NAME)"
- --default-artifact-root
- "s3://mlflow/"
Is there any way you could reduce memory usage or tell me how I can tune it?
@PhilipMay Thanks for reporting this, could you investigate what's increasing the memory usage?
@harupy it is nothing - just a fresh install of a "naked" MLflow without any usage or data. See here:

Any solutions? Mine is consuming 500+ Mi