kcp

kcp copied to clipboard

Describe the bug

The /clusters/foo/openapi endpoints for kcp apiserver return the kubernetes OpenAPI spec.

Steps To Reproduce

kubectl get --raw /openapi/v2 | grep ClusterWorkspace

Expected Behaviour

We generate the definitions, so I imagine we want to serve them? This would also be nice to allow downstream users to generate clients in non-Go SDKs.

Additional Context

No response

Need to make sure we do this properly so we don't overload the server with too much OpenAPI generation across potentially hundreds/thousands of workspaces

In addition to dynamically determining this based on bound APIs within a workspace, I think we can also get a lot of value today from having (semi-static) published definitions for the core kcp APIs so that devs could explore the API surface and e.g. generate clients for non-Go SDKs.

This should also fix #1449.

Adding some more detail:

We are supposed to be including the built-in k8s types in our OpenAPI doc. But this is not happening at the moment because the MiniAggregatorConfig stores a pointer to a GenericConfig that is used by both it (mini aggregator) and the k8s control plane apiserver:

https://github.com/kcp-dev/kcp/blob/e40bffc1a67e56d780bc97b1ced45d0c1d7c786e/pkg/server/config.go#L520

This is a problem because in the mini aggregator code, we tell it to skip OpenAPI installation so we can install our own aggregated handler:

https://github.com/kcp-dev/kubernetes/blob/6ebb9f064d8cd524bafba7154d5e47619303be0a/pkg/genericcontrolplane/aggregator/aggregator.go#L103-L107

Unfortunately this is the same GenericConfig, which ends up disabling OpenAPI installation for the k8s control plane server.

I was able to "fix" this by making a shallow copy of the GenericConfig and storing that in the MiniAggregatorConfig. It feels more like a band-aid, though, and I'm wondering if we could do something better.

Issues go stale after 90d of inactivity. After a furter 30 days, they will turn rotten. Mark the issue as fresh with /remove-lifecycle stale.

If this issue is safe to close now please do so with /close.

/lifecycle stale

Stale issues rot after 30d of inactivity. Mark the issue as fresh with /remove-lifecycle rotten. Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

/lifecycle rotten

Rotten issues close after 30d of inactivity. Reopen the issue with /reopen. Mark the issue as fresh with /remove-lifecycle rotten.

/close

@kcp-ci-bot: Closing this issue.

In response to this:

Rotten issues close after 30d of inactivity. Reopen the issue with /reopen. Mark the issue as fresh with /remove-lifecycle rotten.

/close

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.