Description of problem:
Backup process seem to be broken for active VM when Cinder when NFS backend is used for cinder-volume and cinder-backup services:

• at first a separate volume is created from VM's disk
• then separate CoW image is created on top of original disk
• then original disk is deleted
• then a separate volume created during first step is deleted as well

As a result, VM is left with CoW disk image without backing image and Nova generates the following error when instance is rebooted:

2023-11-16 17:52:16.219 2 ERROR nova.virt.libvirt.guest libvirt.libvirtError: Cannot access backing file '/var/lib/nova/mnt/MNT-ID/volume-UUID1' of storage file '/var/lib/nova/mnt/MNT-ID/volume-UUID1.UUID2' (as uid:107, gid:107): No such file or directory

Version-Release number of selected component (if applicable):
RHOSP 17.0 and RHOSP 17.1.1

How reproducible:

• ask cinder-backup to create a backup of running instance in environment with NFS backend used for cinder and cinder-backup

Actual results:
VM's original disk image is removed while CoW image created on top of it still points to original image.

Expected results:
Original image is either left or CoW image should point to another disk.

Additional info:
Log extracts will be provided privately.