The current deployment of the NFS server introduces an issue related to a scenario where NFS loopback is created - the VM under test is scheduled on the same node where the NFS server resides. Therefore the local PV of the VM creates the loopback while exporting the NFS share. 

We suspect that this scenario introduces high IO wait that causes nfsd kernel panics.

We want to explore an alternative solution where the NFS server is created in a dedicated container like the local registry container, and the dnsmasq to provide a consistent DNS name for the NFS, and the IP setup to be the same as for the registry container.