Weaviate memory consumption

I am using Weaviate cr.weaviate.io/semitechnologies/weaviate:1.26.3 deployed on Kubernetes with 3 replicas, the issue is:

  • Weaviate uses about 2.5GB RAM if data on disk is size 2GB, even if no operations are running on Weaviate.
  • Tried default values of the helm chart and tried changing from GOGC: 100 to GOGC: 50 but same issue.

Is there a configuration that fixes this issue?

hi @mobehairy !!

Welcome to our community :hugs:

By default, Weaviate will store your vectors in memory. This means that even without any operation, the memory will be allocated to allow the query search.

If you want to calculate the memory usage, we have some interesting documentation on that:

Let me know if this helps!