Open hatfieldbrian opened 1 year ago
Seems like you describe the solution in the description, but the problem is not clear. How do you reproduce it? How does it look when inspecting the clusters?
@nirs Problem description added
Labeled low priority since this is an uncommon use case.
Problem
Disabling and reenabling a VRG's
KubeObjectProtection
does not delete its existing backups, nor the pointer to the latest backup stored in the VRG status in the S3 store(s). If the VRG were recreated, its old backups would be restored. This may be ok, but may not be. If a VRG were deleted its backups would be deleted.Approach 1 - Define a separate CRD for
KubeObjectProtection
so it can have its own finalizer to handle backup deletionVRG references it by name. If it is deleted but still referenced, then assume Kube objects need not be protected, or require reference be nil? VRG controller watches it and reconciles it.