nsfinalizer

Documentation

Index

• func NewFinalizerController(namespaceName string, ...) factory.Controller

Functions

func NewFinalizerController

func NewFinalizerController(
	namespaceName string,
	kubeInformersForTargetNamespace kubeinformers.SharedInformerFactory,
	namespaceGetter v1.NamespacesGetter,
	eventRecorder events.Recorder,
) factory.Controller

NewFinalizerController is here because When running an aggregated API on the platform, you delete the namespace hosting the aggregated API. Doing that the namespace controller starts by doing complete discovery and then deleting all objects, but pods have a grace period, so it deletes the rest and requeues. The ns controller starts again and does a complete discovery and.... fails. The failure means it refuses to complete the cleanup. Now, we don't actually want to delete the resoruces from our aggregated API, only the server plus config if we remove the apiservices to unstick it, GC will start cleaning everything. For now, we can unbork 4.0, but clearing the finalizer after the pod and daemonset we created are gone.