There is no way to delete the Lease programatically

[MathieuCesbron]

Bug Report

With this config, a Lease is created and a leader is chosen. I installed my operator and all is working well. The issue is that everytime I undeploy the operator the Lease is not deleted. Is there a way to add a owner ownerReferences to the Lease in the code ? Or something that will clean the Lease on undeploy.

mgr, err := ctrl.NewManager(ctrl.GetConfigOrDie(), ctrl.Options{
    Scheme:                  scheme,
    MetricsBindAddress:      metricsAddr,
    Port:                    9443,
    HealthProbeBindAddress:  probeAddr,
    LeaderElection:          true,
    LeaderElectionID:        "operator-lock",
    LeaderElectionNamespace: "default",
})

Thanks,

[jberkhahn]

What exactly do you mean by 'undeploying' your operator? Specifically, are you running the controller inside or outside the cluster?

[MathieuCesbron]

I am running the controller inside the cluster. By "undeploying" the operator I mean to use the make undeploy of the Makefile.

Adding a LeaderElection: true will create a Lease object on deployment of the cluster. I am looking for a way to garbage collect this Lease when the pods it references are deleted.

[acornett21]

Hi @MathieuCesbron I had a chance to test this, and the Lease object is cleaned up by K8s after the reference is gone.

This was tested on the below:

oc version
Client Version: 4.14.11
Kustomize Version: v5.0.1
Server Version: 4.14.1
Kubernetes Version: v1.27.6+f67aeb3

$ make deploy 
...
...
deployment.apps/certification-operator-controller-manager created


$ oc get lease
NAME                  HOLDER                                                                                           AGE
ef59679f.redhat.com   certification-operator-controller-manager-8cc6589ff-2rjv7_55c951d2-c54f-45db-896a-8d01e87af0f7   3m44s

$ make undeploy
...
...
deployment.apps "certification-operator-controller-manager" deleted

$ oc get lease
No resources found in certification-operator-system namespace.

The value for LeaderElectionID is autogenerated via kubebuilder based on how the operator is scaffolded, so the only thing I can think of is that there is a conflict in your cluster with the name operator-lock. Can you try replacing with the autogenerated value?

[MathieuCesbron]

Thanks for testing it. There must be something I am doing wrong then, I will close the issue.