While I should be on holiday, and not thinking about all things network, the boss ok'ed a SAN on Friday(which one is still to be decided, that will go to scrutiny in the new year).
Currently we run 2 main hosts, both identical R710's on Server 2008R2, one purely on local storage and the other on a PERC attached MD1000 drive shelf. Each one has around 5 VM's on it. No clustering.
The SAN is going to be storage for the VM's, as local storage is currently dwindling.
The plan is to cluster the 2 servers and load balance the VM's over the 2 servers and provide enough oomph on each to failover to single node if needed. That has pretty much been decided, it is the how that I am scratching my head with.
I would like to go with 2012R2 going forward, and the servers are on 2008R2 so would it be best to export the VM's to a share on the new SAN(shared on a different SAN attached server to the hosts) and then completely wipe and rebuild the 2 hosts from scratch as a 2012R2 cluster and then import the VM's back in to the new cluster? That seems to be the best way to do it to me, but am I right?
Not being an expert in HyperV I can only tell you what I would do if it were vSphere.
1. First, I would configure the existing servers to both use the new SAN in a cluster
2. Then I would migrate the VMs to use the SAN.
3. Taking backups at this point of the VMs as well as the HyperV servers just in case would be next.
4. Now, remove one of the servers from the cluster after moving all of its VMs to the other server
5. Install 2012R2 on the server you've removed from the cluster and configure HyperV.
6. Next, join the new HyperV server to the cluster (I'm assuming this can be done, it should), and test that the VMs work fine on it (upgrade templates if needed).
7. Move all of the VMs to the new 2012 server and then remove the 2008 server from the cluster
8. Install 2012R2 on the second server and configure HyperV.
9. Join the second server to the cluster and test.
10. Configure failover, load balancing, etc.
11. Backup everything again after you've finished the config.
My opinion is that it's a bad idea to just blow away two VM hosts if not necessary and it also means extensive downtime, especially of there are problems. The method above means little or no downtime and very little risk.