Description of problem:
We have two compute nodes with the same configuration. One of them fails to start guest instances after hypervisor restart.

The following KCS was been created for this issue:
https://access.redhat.com/solutions/2979711

Though it works in our reproducer environment, it still can't fix the issue on customer's site.


Steps to Reproduce:
1. Start an instance on hypervisor
2. Reboot nova node

Actual results:
All instances are in shutdown state

Expected results:
Instances (with running state before hypervisor reboot) should be resumed

Additional info:
This has been tested in our reproducer env with different libvirt-guests/nova configuration options with only one combination leading to have guests resuming after nova reboot. But this doesn't seem to work in customer env

~~~
resume_guests_state_on_host_boot=true in nova.conf
libvirt-guests.service enabled
ON_BOOT=ignore, ON_SHUTDOWN=shutdown in /etc/sysconfig/libvirt-guests
~~~