Not cleaning up PVCs

[chrislovecnm]

When we delete a database we do not seem to delete the PVCs correctly

# you need a bad storage class name or request over quota in the example
# this step has to fail
./hack/apply-apply-crdb-example.sh -c test
# then delete
./hack/apply-delete-crdb-example.sh -c test

You can then see the hanging PVC

$ k get pvc
NAME             STATUS    VOLUME   CAPACITY   ACCESS MODES   STORAGECLASS   AGE
datadir-crdb-0   Pending                                      crdb-io1       3m43s

[chrislovecnm]

We are not cleaning up PVCs at all. The question is what policy do we want to use for cleanup?

[chrislovecnm]

@johnrk the operator is not deleting volumes that are attached to the database Pods when it deletes the database. This is actually a feature of Kubernetes where you have to delete by hand the storage volumes when you delete a stateful set. There reasoning is that you do not delete the storage automatically, because you might delete data. I need to know if we want the default k8s behavior or another behavior.

[johnrk-zz]

@chrisseto , do you know what behavior we want here?

[chrislovecnm]

@johnrk we talked about this on standup. Did we make a decision?

[chrislovecnm]

We need to document this as part of the operator. PVs will not be deleted by the operator.

[johnrk-zz]

We need to document this as part of the operator. PVs will not be deleted by the operator.

^ cc @taroface