Weaviate on EKS in AWS - Running Hot!

General
1

Hello,
I have set up weaviate on EKS within AWS. My pod is running on a m7g.8xlarge instance which has 128GiB of RAM. Weaviate is performing very well as inserts and queries are lightning fast, so I am very happy with that, however, whenever I ssh into my instance and run the “top” command I can see that my weaviate pod is consistently utilizing 85%-95% of the machine’s memory. I was shocked to see this given how large my instance is. Has anyone seen this before, or have any suggestions on how to use less memory?

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
XXX XXXX 20 0 249.8g 118.4g 762720 S 1.0 95.8 372:51.04 weaviate

2

Hi!

Check out this article: GOMEMLIMIT is a game changer for high-memory applications | Weaviate - vector database

Do you already have some data and activity going on in Weaviate? Go memory can be very oportunistic, so it may be leveraging the available memory.

Let me know if this helps!

Thanks!

1 Like
3

Thanks for the information @DudaNogueira , I updated my GOMEMLIMIT to 32GiB, but my pod is still utilizing ~95% of the 128GiB of memory. I have been really impressed with the speed and responsiveness of Weaviate. Is it always indexing/charding/running algorithms behind the scenes to optimizie searching and data retrieval?