Support dynamically watching objects based on namespaces labels

[tabarafael]

Inside "sigs.k8s.io/controller-runtime/pkg/cache".Options The option for DefaultNamespaces should select the namespaces in which the cache watches.

When setting the cache to watch dynamic namespaces that should be identified through a label selector the cache refuses to sync properly, when attempting to get a secret from such namespace, it returns the error:

failed to get secret: /v1, Kind=Secret is not cached.

code example:

	labelRequirements, err := labels.NewRequirement(
		"key", selection.In, []string{"value1", "value2"})
	
        cacheOpts := cache.Options{
		DefaultNamespaces: map[string]cache.Config{
			cache.AllNamespaces: cache.Config{
				LabelSelector: labels.NewSelector().Add(*labelRequirements),
			},
		},
		ReaderFailOnMissingInformer: true,
		SyncPeriod:                  &syncTimer,
	}
....

When replacing the cache.AllNamespaces string with the name of the namespaces and removing the label selector it works properly.

[alvaroaleman]

The labelselector is not a labelselector for namespaces, it is a labelselector for objects inside the namespace

[tabarafael]

So selecting a namespace by the label instead of the name is just not a supported feature? instead of a bug.

You see, I need to cache the objects inside these dynamic namespaces with random (uuid) names, the objects don't have any label that I can go off of to select them. I think I have exhausted my options already.

[alvaroaleman]

So selecting a namespace by the label instead of the name is just not a supported feature

It is not supported, no.

[sbueringer]

You can potentially just watch cluster-wide and then filter out ConfigMaps in namespaces you don't care about via predicate or within Reconcile

[tabarafael]

That just has the problem in that it requires cluster-wide permission for get/list, instead of only namespaced permission.

[alvaroaleman]

/rename Support dynamically watching objects based on namespaces labels

Renaming the issue to clarify the problem.

What you are asking for isn't supported well by either the server side nor the client libs, you would need some sort of controller that dynamically starts and stops informers based on the labeling of namespaces

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale