I'd like to set up a 2-node HA-Cluster based on Ubuntu and KVM running about 30 VM's. Both nodes are connected to a HP MSA(5TB) via Fibre Channel. Multipathing's already working and the nodes are able to access the shared storage.
There are two main requirements:
Fail-over of running VM's from one node to the other. With downtime.
Live-Migration of VM's from one node to the other. Without downtime.
I think there are several ways to achieve this. But which is the best/working? After some investigation I found solutions/ideas like:
A. Setting up a Cluster with Pacemaker, Corosync, CMAN, DLM, CLVM, GFS2 ...
B. Setting up some Cloud SW like OpenStack, OpenNebula, CloudStack, Eucalyptus ...
C. Or can I somehow put the VM's straight on the LV/LUN and use the KVM/libvirt lock manager.
D. ...
At the moment the 30 VM's are running well on a activ/passiv(DRBD) Cluster with Ubuntu, Pacemaker, heartbeat.
Thanks for any suggestion.