Why manifests/base/service.yaml does not include webhook server port (443) in version 1.7.0~1.5.0?

[yeonhooy]

I wonder why manifests/base/service.yaml does not include webhook server port (443) in version 1.7.0~1.4.0?

I tried to use training-operator version 1.7.0 or 1.5.0 (stable version) instead of master version.
However, I found that there is no webhook port for training operator, which seems to cause this error:
Internal error occurred: failed calling webhook "validator.tfjob.training-operator.kubeflow.org": Post "https://training-operator.kubeflow.svc:443/validate-kubeflow-org-v1-tfjob?timeout=10s": no service port 443 found for service "training-operator"

Thanks

[show981111]

Same problem.

Edit) I deleted my minikube cluster and restarted. Then seems like working again.

[yeonhooy]

Thanks for sharing :)

It may not be a big issue with missing the webhook server port in previous versions.
Because they did not require webook validation, I guess (maybe I`m wrong).

I`m now just using a master version that does not cause any problems.

[tenzen-y]

@yeonhooy @show981111 Hi, thank you for creating this issue.
How to install training-operator? I guess that you performed kubectl apply -k "github.com/kubeflow/training-operator/manifests/overlays/standalone", right?

[yeonhooy]

@tenzen-y Right. I tried to use this command kubectl apply -k "github.com/kubeflow/training-operator/manifests/overlays/standalone?ref=v1.7.0" for stable version.
But I failed due to the missing of webhook server port.

I'm now using master version without ValidatingWebhookConfiguration.

[tenzen-y]

@tenzen-y Right. I tried to use this command kubectl apply -k "github.com/kubeflow/training-operator/manifests/overlays/standalone?ref=v1.7.0" for stable version. But I failed due to the missing of webhook server port.

I'm now using master version without ValidatingWebhookConfiguration.

It's curious. In v1.7.0, the training-operator doesn't have any webhook validations. So, the above error should not happen.
But, we introduced webhook validations since v1.8.0-rc0 (as well as the master branch). So, I suspect that you accidentally performed the kubectl apply -k "github.com/kubeflow/training-operator/manifests/overlays/standalone (the master branch) before.

[tenzen-y]

But, we introduced webhook validations since v1.8.0-rc0 (as well as the master branch).

In other words, we never provided the manifests for the validator.tfjob.training-operator.kubeflow.org until v1.8.0.

[yeonhooy]

I see. May be I performed the master version before performing version v1.7.
My curiosity is resolved :)

Thanks!!