Interesting question:
Was just interesting. In a worse case scenario you would not be able to start up a VM.
Anyone else have a method to start a VM manually using CLI?
Of the oVirt Manager GUI (web interface) is down or no longer working how can you manually start a VM on the node? The VdsClient looks promising as well as the old virsh command. But it seems once you have stopped the VM in the GUI it is not possible to start the VM again without the GUI.
I get this:
[root@virt0 ~]# vdsClient -s 0 continue 6416b49f-6045-4972-9805-cf24bb3801d0
Virtual machine does not exist
The
continue
command is for paused VMs, not VMs that are shut down. To Start a VM you need two things - storage access and VM settings. When oVirt starts a VM, it will find the disks that are attached to the VM in the database, locate a suitable host, allow the host access to the disks, generate a domain XML for the VM and start it. Without the database, you don't know which disks belong to which VM, nor do you know the VM settings. Both can be found in the master domain's set of OVF files of course, but it's a manual process. Getting the oVirt engine up and running is faster, backing it up is very simple, and running it in HA mode is also far from being rocket science.For a quick and dirty solution, on a completely botched setup, I would scan the hosts for the VM name in vdsm logs. The entire domxml is posted there, so you can use it to generate a quick and dirty libvirt config, and also extract the VM's disk addresses. I wouldn't recommend going down this path though, especially if there is any chance to bring the engine back online.
In short, make sure your engine is clustered, or at least backed up (
engine-backup.sh
is in/usr/share/ovirt-engine/bin
) and you will not have these problems. If the engine goes down, the running VMs remain running, so if you bring it up quickly, no production downtime should be expected.