cluster deletion does not clean up pvc

[colhom]

creating, deleting, and re-creating an etcdcluster with the same name results in:

panic: persistentvolumeclaims "etcd-cluster-pvc" already exists

goroutine 1 [running]:
panic(0x147a3a0, 0xc42034d080)
        /usr/lib/go/src/runtime/panic.go:500 +0x1a1
github.com/coreos/kube-etcd-controller/pkg/controller.(*Controller).Run(0xc420386a20)
        /home/chom/gopaths/kube-etcd-controller/src/github.com/coreos/kube-etcd-controller/pkg/controller/controller.go:206 +0x6c5
main.main()
        /home/chom/gopaths/kube-etcd-controller/src/github.com/coreos/kube-etcd-controller/cmd/controller/main.go:49 +0x55

@hongchaodeng should we be delete the pvc along with the pods and services when an etcd cluster is deleted?

[colhom]

Also, I do believe in that case we would have to clean up the backup pod for the cluster as well.

[hongchaodeng]

1. We do delete all pods/services when etcd cluster is deleted. Please rebase to master.
2. Regarding PVC, we have another field CleanupBackupIfDeleted to help clean up. However, we don't delete by default because we want to keep data in case of user mistake.
3. Instead of deleting it, I think there are better solution to this: we will have an option in backup policy to specify any existing pvc to mount, and we will make pvc name more random

[hongchaodeng]

@colhom
This and #217 is not an issue.
Is it ok for you to close them?

[colhom]

@hongchaodeng yes, not an issue. I'll close them both now.

That being said- should we file an issue to randomize pvc names on creation so we can re-create controllers without cleaning up old PVCs?

[hongchaodeng]

@colhom

should we file an issue to randomize pvc names on creation

Yes. Feel free to create an issue.