Kvm guest in HOLD state in OpenNebula after import but guest is operating normally otherwise

I have a number of CentOS 7 hosts with kvm guests. All this has been working well for years, managed with virsh and virt-manager.

I’m now beginning to use OpenNebula 5.10.1 with this collection of hosts and guests.
So far I really like what I see. Kudos to the OpenNebula developers!

I have an OpenNebula front end running. No problems there so far.

I have 16 OpenNebula nodes installed and the hosts have all been created in the front end. Again, no problems there so far.

I’ve imported about 70 of my kvm guest VMs. For the most part this seems to have worked just fine. However, about 20 of the VMs are showing up with Status “HOLD” and Host “–”. The guests themselves are operating normally though.

Does this mean that the import failed somehow and OpenNebula is just confused about the true state?

What’s the best way to proceed?

Can I return the VM to the wild - the opposite of importing it from the wild - and then try to import it again? Will be that be completely non-destructive to the kvm guest?


CentOS 7, kvm, OpenNebula 5.10.1

I discovered the onevm recover options and was able to do this to remove these VMs from OpenNebula:

onevm recover --delete-db vm_name

I then tried to import these VMs again and they ended up in the same HOLD state.
Also, the LCM State is LCM_INIT.

I also discovered the VM logs and looked there for clues but the logs for these VMs are empty. Is there a configuration setting to dial up the verbosity of these logs?

I determined the source of this problem. Some of our kvm guests have a vnc graphics device defined in their xml. It is these and only these guests that are failing to import.
I will remove the vnc device from these guests and then redo the imports.