KubeRay cluster config questions

How severe does this issue affect your experience of using Ray?

  • Low: It annoys or frustrates me for a moment.

So I was trying to figure out how exactly the configuration goes for KubeRay Ray clusters. Right now I have started by using kuberay/ray-cluster.autoscaler.yaml at master · ray-project/kuberay · GitHub which works. I have a setup of 5 computers with 16 physical cores each and 128GB (each) ram with various GPUs. Is there anything special I should do to ensure that the cluster spins up the appropriate amount of workers? Also, what should be a good gauge for determining these numbers?


improving the docs is on the docket for the immediate future.

I’d recommend allocating a whole Computer for each Ray pod – when sizing the pod leave a bit of room for system daemons and such

you’ll also need to specify num_gpus: 1 or num_gpus: 2 or however many GPUs you have under rayStartParams for your GPU pods

Please let me know if you hit any questions or hit any issues.

Hi @cloudhaxor, I will mark this question as resolved. Feel free to respond or create a new post for if you need more guidance.