OpenAPI validation for Katib CRs - prevent invalid specs

[jlewi]

/kind feature

I'm getting customer reports that invalid Experiment specs lead to experiments being stuck.

This is a common failure mode with custom resources; if the resource spec is invalid the resource will get created but when the controller tries to parse it the parsing will fail.

I think the new/recommended way of handling this might be to use OpenAPI schema validation in the CRD.
https://kubernetes.io/docs/tasks/access-kubernetes-api/custom-resources/custom-resource-definitions/#specifying-a-structural-schema

It looks like Katib CRDs might be missing this
https://github.com/kubeflow/manifests/blob/master/katib/katib-crds/base/experiment-crd.yaml

[issue-label-bot]

Issue Label Bot is not confident enough to auto-label this issue.
See dashboard for more details.

[johnugeorge]

@jlewi Can you explain more on the invalid experiment specs that get created? Katib already has a validating webhook

[gaocegege]

Kubebuilder supports generating OpenAPI schema for CRDs. I think we can support it in the next API version. (v1beta1, maybe)

[jlewi]

I believe the customer was passing names which were too long for the experiments and this caused the controller to stop working.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale]

This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.