ReplicaSet unexpected type in v1.7.0-rc.0

[brancz]

Is this a BUG REPORT or FEATURE REQUEST?:

/kind bug

What happened:

Logs showed:

E0710 17:49:42.568957       1 reflector.go:342] k8s.io/kube-state-metrics/internal/store/builder.go:295: expected type *v1beta1.ReplicaSet, but watch event object had type *v1.ReplicaSet

What you expected to happen:

Only non-error logs.

How to reproduce it (as minimally and precisely as possible):

Always with v1.7.0-rc.0

Anything else we need to know?:

n/a

Environment:

• Kubernetes version (use kubectl version): v1.15.0 minikube
• Kube-state-metrics image version: v1.7.0-rc.0

@tariq1890 @lilic @mxinden @andyxning

[tariq1890]

Thanks for reporting @brancz. I can push the fix for this.

[lilic]

@brancz @tariq1890 Good catch and quick fix! 🎉

It would be a good idea to have a test to make sure the kube-state-metrics has 0 errors in the logs as part of the e2e test suite.

Also erroring out and terminating the pod on any errors like this, so we (in CI) and the users (in their clusters) detect problems sooner and easier. I had an RBAC problem and while kube-state-metrics was up and running, certain resources were not working and while the logs were reporting not being able to list/watch those resource, I was expecting the pod to be terminated in that case and not just log that.

WDYT?

(Happy to tackle those if everyone agrees on those)

[brancz]

No errors in logs is something we can do immediately I think (not blocking v1.7.0, but we can immediately get started working on that), go for it! :)

Erroring out needs a bit more thought and major refactorings I think, I'm not even sure it's entirely possible with the way informers work today, but I'd be happy to see this!

[brancz]

Closed by #816. @lilic feel free to create a separate ticket for what you mentioned in #815 (comment)