Description of problem:

When an OpenShift cluster uses VMWare as the backing storage for it's workloads, persistent volumes are created as VMDK files on the VMWare backing storage. When the persistant volume is removed from OpenShift, it removes the reference from OCP so there is no longer a pointer to the correlating VMDK file, but the file is not removed from the VMWare storage. This is resulting in orphaned files that can no longer be identified as belonging to any specific workload (vm, pod, etc.), so it is impossible to even manually clean up these orphaned files from the VMWare side, as you do not know where they came from and which ones are in use and which ones aren't.

This happens irrespective of the reclaimPolicy defined in the storageclass. The behavior is the same whether it is set to Retain or Delete.

Version-Release number of selected component (if applicable): 4.12.z, 4.14.z, 4,16,z

How reproducible: Consistent

Steps to Reproduce:

    1. Create a storageclass that uses VMWare as the backing store and the VMware CSI driver with a reclaim policy of Delete
    2. Create a PV using this storageclass
    3. Observe the creation of a VMDK file on the backing store
    4. Create some PVCs in this PV
    4. Delete the PVCs
    5. Delete the PV

Actual results: VMDK file on the backing store persists.

Expected results: VMDK file on the backing store should be deleted.