Per github example, to enable autoscaling on RayCluster CRD we need to tweak autoscalerOptions and enableInTreeAutoscaling fields, however I can’t find similar fields either in kuberay-apiserver create cluster request payload, or in ray-cluster helm chart values. It also seems the config spec is inconsistent between those three (RayCluster, kuberay-apiserver requests payloads, ray-cluster helm chart). Is it intentional? Is it possible to enable autoscaling via apiserver/helm chart?
Autoscaling can be enabled via Helm chart in master.
For the API server, feel free to open a GitHub issue and tag me (@DmitriGekhtman).
I can assign the relevant parties.