Reuque question

[Lin Lin]

Hi cluster-api team

I've notice that the cluster controller would wait for infraCluster object to be deleted. But the returning Result and error are both empty, so the controller runtime will not retry Reconcile. Is there something I missed? Could this be an issue? The code is below.

https://github.com/kubernetes-sigs/cluster-api/blob/main/internal/controllers/cluster/cluster_controller.go#L457

log.Info("Cluster still has descendants - need to requeue", "controlPlaneRef", cluster.Spec.ControlPlaneRef.Name)

return ctrl.Result{}, nil

Thanks for you time.

[Alberto Garcia Lamela]

Hi Lin Lin,

The controller is watching the descendants, so further reconciliation should be triggered by events on those.

Thanks,

Alberto

To unsubscribe from this group and stop receiving emails from it, send an email to sig-cluster-life...@kubernetes.io.

[Ian Howell]

Hey Alberto,

Apologies if I'm misunderstanding, but the log message that Lin Lin mentions specifically says that the resource will be requeued, but then it clearly is _not_ requeued. Perhaps we should update the log message?

Thanks,
Ian Howell

[Alberto Garcia Lamela]

Hey Ian,

requeing should happen driven by the chain of events as descendants go through the deletion process, but please feel free to open a PR and we can follow up there on setting deleteRequeueAfter explicitly or updating logs.

Thanks,

Alberto.

[Lin Lin]

I'm happy to do it. This is the PR to set the deleteRequeueAfter explicitly.

https://github.com/kubernetes-sigs/cluster-api/pull/11822