✨ metadata field changes for structural schema

[droot]

This PR handles generating schema for metadata as per structural schema.
https://github.com/kubernetes/enhancements/blob/master/keps/sig-api-machinery/20190425-structural-openapi.md

Partial fix for #216

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: droot

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [droot]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[droot]

/cc @DirectXMan12 @sttts @mengqiy

[sttts]

these two are possible, but not needed in a structural schema.

[sttts]

for this @DirectXMan12 brought up the point of partial ObjectMeta. There is no requirement that the metadata is complete for an embedded resource. /cc @liggitt

We either want a way to mark an embedded ObjectMeta as partial, or we relax the validation we do today (if that is possible now in beta). @liggitt wdyt?

[liggitt]

what is currently required that we would want to relax?

[liggitt]

as far as I remember, only kind/apiVersion were required

[sttts]

It's about ObjectMeta: name, namespace as far as I see.

[liggitt]

name must be optional in openapi because of generateName, and iirc our server-side implementation put a dummy name in place when validating if no name was provided

namespace must already be optional (both server-side and in openapi) because of cluster-scoped resources

[sttts]

ok, I think I misunderstood @DirectXMan12 in Slack. He was talking about TypeMeta (and I remember the dummy name value we put it to work around the check).

So the question is whether without TypeMeta it is really an embedded object.

[droot]

@sttts @DirectXMan12 @mengqiy I reevaluated my approach and thought of another approach:

By default, do not generate any validation for ObjectMeta (irrespective of it being a root or embedded). Let users specify marker // +kubebuilder:validation:embedded if we want to enable validation for embedded objects. This is how users do it for enabling any validation for the fields in the spec today, so it's inline with that. This will also avoid the partial objectMeta issue that @stts commented above where in current flattenVisitor it doesn't have a good way to track if objectMeta has sibling node TypeMeta or not. Thoughts ?

[DirectXMan12]

sgtm on the marker

[DirectXMan12]

that doesn't actually resolve the problem though -- the x-kubernetes-embedded, IIRC requires apiVersion and kind to be specified too, so won't work for things like the template field in Deployment/CronJob/etc.

@sttts @liggitt

but we can tackle that later

[k8s-ci-robot]

@droot: PR needs rebase.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[droot]

closing this in favor of #266