Open slesru opened 6 years ago
You could create an agent which monitors the config directory, and make the VM resource dependent on that agent. That way it won't be able to migrate if the config directory isn't present.
Hello!
Sorry, I can't understand- pacemaker manages cluster resources, i.e., if it is mount point, then this resource can be moved between nodes. In your suggestion we need to have directory mounted on all nodes, so this is local resource, not cluster. If we'll create 3 resources in 3 nodes cluster and make VM dependent on all 3 , then if only 2 machines are available, then VM will be not be able to move... I think that what I need should be implemented in VirtialDomain resource agent.
Hello!
We use gluster as VM shared storage using gfapi and as domain definition (xml files) storage , mounted. Problem here is that sometimes for some reason config files may not be mounted, but, because libvirt uses gfapi, VM's storage is available. So, if we move VM to node , where there is no config directory mounted, then VM will be moved, but pacemaker can't monitor it, so it can be started twice and VM storage will be damaged.
Is it possible to add config presence check in pacemaker before doing live migration?
Thank you!