Bug 1494257

Description of problem:

The mount/unmount logic is controlled by creation/deletion of the pod. But PV/PVC is controlled by pv controller. As mentioned in issue #44297, in current design, after deleting PVC, the mount point was still exist on the Kubernetes node which the Pod was deployed and also the mount point could be seen from inside the Pod.And also if other Pod reuses the PV, then both old Pod and new Pod could access same mount point.

How reproducible:

Very

Steps to Reproduce:
PreReq: PV created (pv_rwo.yaml provided below)

In Project 1:
1. Claim the persistent volume using a PVC (oc create -f pvc_rwo.yaml)
2. Deploy a pod using the PVC/PV (oc create -f pod.yaml)
3. Write some stuff to the PV (used console terminal, not necessary, but interesting later)
4. Delete the claim (oc delete -f pvc_rwo.yaml)
* Do not delete the Pod

Change into Project 2:
1. Claim the same persistent volume using a PVC (oc create -f pvc_rwo.yaml)
2. Deploy a pod using the PVC/PV (oc create -f pod.yaml)
3. Write some stuff to the PV (used console terminal)

Go back to Project 1:
1. Go to pod's terminal and ls, you'll see pod from project2's data
2. Write some stuff -succeeds


Actual results:

This works

Expected results:

For the PV to be unable to be deleted while an active pod is using it.


Description of problem:

Version-Release number of selected component (if applicable):

How reproducible:

Steps to Reproduce:
1.
2.
3.

Actual results:


Expected results:

Master Log:

Node Log (of failed PODs):

PV Dump:

PVC Dump:

StorageClass Dump (if StorageClass used by PV/PVC):

Additional info: