[ClusterLabs] Gracefully Failing Live Migrations

[Ken Gaillot]

On Thu, 2024-02-01 at 10:20 -0600, Billy Croan wrote:
> Sometimes I've tried to move a resource from one node to another, and
> it migrates live without a problem.  Other times I get 
> > Failed Resource Actions:
> > * vm_myvm_migrate_to_0 on node1 'unknown error' (1): call=102,
> > status=complete, exitreason='myvm: live migration to node2 failed:
> > 1',
> >     last-rc-change='Sat Jan 13 09:13:31 2024', queued=1ms,
> > exec=35874ms
> > 
> 
> And I find out the live part of the migration failed, when the vm
> reboots and an (albeit minor) outage occurs.
> 
> Is there a way to configure pacemaker, so that if it is unable to
> migrate live it simply does not migrate at all?
> 

No. Pacemaker automatically replaces a required stop/start sequence
with live migration when possible. If there is a live migration
attempted, by definition the resource must move one way or another.
Also, live migration involves three steps, and if one of them fails,
the resource is in an unknown state, so it must be restarted anyway.
-- 
Ken Gaillot <kgaillot at redhat.com>