I am researching a new Windows Server 2012 server deployment utilizing failover clustering with cluster shared volumes for Hyper-V on a small network. From what I've read, most two node cluster configurations also require a third server to host the SAN. The two clustered host computers would then be able to live migrate VMs from one hypervisor to another with minimal disruption. However, it seems to me then that the SAN is the single point of failure between the two.
What I am wondering is if I can eliminate the dedicated SAN server altogether and create two iSCSI target roles on each of the Hyper-V host computers. Then, I could make those discs available for cluster shared volumes. Ideally, the VMs run out of that, and the config/vhd files are duplicated on both machines ready for failover. However, I have no idea if it works like this or not.
So, can you run CSV in a two-node cluster without a dedicated SAN box? If so, are the virtual machines mirrored between both nodes, but only go "live" when the preferred node goes offline? If this doesn't work in a two node cluster, would three make any difference?
-Thomas