nova instances which are deleted from database remain on the hypervisor and consume disk space in Red Hat OpenStack Platform 10

Issue

A compute node is shut down forcefully and instances on this compute node are deleted while the hypervisor is down. Under specific circumstances, instances will be deleted from nova, but they remain visible on the hypervisor when it is brought back online.

[root@overcloud-compute-0 heat-admin]# virsh list --all
 Id    Name                           State
----------------------------------------------------
 -     instance-0004708a              shut off
 -     instance-0004708d              shut off
 -     instance-00047090              shut off
 -     instance-00047093              shut off
(...)

The problem may become more serious due to the fact that the ephemeral disks of these instance remain on the hypervisor and eventually fill up a significant amount of disk space. This can cause issues with the scheduling of new instances on the hypervisor.