What is the best solution for making your VMs able to automatically
boot up on another working host when something goes wrong (gluster
problem, non responsive host etc)? Would you enable the Affinity
Manager and enforce some policies or would you set the VMs you want
as Highly Available?
High Availability and Host fencing are the 2 parts that you need to
ensure that the VM will be restarted after a failure.
If you storage domain goes bad, the VMs will be paused and
theoretically they will be resumed automatically when the storage
domain is back.