[Bug] Kuberay API Server Makefile target deploy-opeartor default image does not exist

[tedhtchang]

Search before asking

• I searched the issues and found no similar issues.

KubeRay Component

apiserver

What happened + What you expected to happen

The Makefile target deploy-operator deploys no-existing default image kuberay/operator:latest

Reproduction script

Run:

# clone the kuberay repo
cd apiserver
make deploy-operator
kubectl get pod

Got

NAME                                READY   STATUS         RESTARTS   AGE
kuberay-operator-664fc97c4c-tl5kl   0/1     ErrImagePull   0          100s

Expect:

NAME                                READY   STATUS    RESTARTS   AGE
kuberay-operator-6b864f7f7b-9crrp   1/1     Running   0          3m42s

Anything else

Recommend:
Change OPERATOR_IMAGE_TAG ?= nightly in the apiserver/Makefile and document how to specify version OPERATOR_IMAGE_TAG=v0.6.0 make deploy-operator in Development.md
Or
Tag latest or stable in the kuberay/operator repo

Are you willing to submit a PR?

• Yes I am willing to submit a PR!

[tedhtchang]

I realized the these deploy and deploy-operator commends need to follow other commends to work. For example:

# build image locally-> create KinD cluster -> load image to local KinD registry -> helm upgrade using the local registry
make docker-image cluster load-image deploy
make operator-image cluster load-operator-image deploy-operator

@z103cb Should we explicitly set helm to use KinD localhost registry in the makefile for the intention above ?

helm upgrade --install kuberay-apiserver ../helm-chart/kuberay-apiserver --wait --set image.tag=latest --set image.pullPolicy=IfNotPresent --set image.repository=localhost/kuberay/apiserver
helm upgrade --install raycluster ../helm-chart/kuberay-operator --wait --set image.tag=latest --set image.pullPolicy=IfNotPresent --set image.repository=localhost/kuberay/operator

[z103cb]

I realized the these deploy and deploy-operator commends need to follow other commends to work. For example:

# build image locally-> create KinD cluster -> load image to local KinD registry -> helm upgrade using the local registry
make docker-image cluster load-image deploy
make operator-image cluster load-operator-image deploy-operator

@z103cb Should we explicitly set helm to use KinD localhost registry in the makefile for the intention above ?

helm upgrade --install kuberay-apiserver ../helm-chart/kuberay-apiserver --wait --set image.tag=latest --set image.pullPolicy=IfNotPresent --set image.repository=localhost/kuberay/apiserver
helm upgrade --install raycluster ../helm-chart/kuberay-operator --wait --set image.tag=latest --set image.pullPolicy=IfNotPresent --set image.repository=localhost/kuberay/operator

@tedhtchang perhaps not (the targets work for me without the need to set the registry). In retrospect, I think we should set the operator image to the nightly tag and document how to build and deploy an operator image from source (perhaps add the targets to make things simpler).

Thoughts ?