[Bug] Unable to deploy standalone ray-operator from stable version v0.2.0

[jeroen-vankelecom]

trafficstars

Search before asking

• [ ] I searched the issues and found no similar issues.

KubeRay Component

ray-operator

What happened + What you expected to happen

Bug

We are installing the ray-operator following the instructions from the README - but we would like to pin the version on the stable version 0.2.0.

$ kubectl apply -k "github.com/ray-project/kuberay/ray-operator/config/default?ref=v0.2.0&timeout=90s"

Expected behaviour

Expecting to see v0.2.0 of the operator launched on the cluster, but it seems it is using the nightly build:

$ kubectl get pods --namespace ray-system -o jsonpath="{.items[*].spec.containers[*].image}"
kuberay/operator:nightly

The image tag is wrongly set in the default Kustomization for the ray-operator.

Workaround?

Deploying ray-operator through the base Kustomization seems to work fine, but it also deploys the api server (which we do not need).

So our questions:

1. Is there a workaround to override the tag, or to disable the api-server deployment?
2. Could you pin the operator version?

Reproduction script

kubectl get pods --namespace ray-system -o jsonpath="{.items[*].spec.containers[*].image}"
kubectl apply -k "github.com/ray-project/kuberay/ray-operator/config/default?ref=v0.2.0&timeout=90s"

Anything else

No response

Are you willing to submit a PR?

• [ ] Yes I am willing to submit a PR!

[jeroen-vankelecom]

Currently working around the issue by using another Kustomization:

apiVersion: kustomize.config.k8s.io/v1beta1
kind: Kustomization

resources:
- [email protected]:ray-project/kuberay/ray-operator/config/default?ref=v0.2.0&timeout=90s

images:
- name: kuberay/operator
  newName: kuberay/operator
  newTag: v0.2.0

[kevin85421]

The release v0.4.0 is almost done. Close this issue.