It's been one of those weeks getting lost in acronyms again :)
We recently deployed a pair of HP ProLiant DL380p Gen8's, sat on a a pair EqualLogic load balanced SAN's. Combined resource available of 4 E5-2620 with 512GB RAM and about 50TB per SAN.
We went with Hyper-V Core 2012 and a smattering of VM's from Win7/8, Server 2008r2 and Server 2012 with SQL 2005/08/12 (legacy applications). Each node/host sits well below the 50% allocated resource as we built it to be a failover cluster.
Now I've not actually tested pulling the power cables from one of the nodes, nor shall I, but I have live migrated all VM's to 1 node with no downtime and users reported no speed issues.
My questions, should a node fail;
How long will it take for VM's to spin up on another node.
Should we cluster SQL too, or is the HA failover tolerance enough to cope?
I understand if a VM with a SQL instance becomes corrupt we'd lose it (to our last good backup), but being transactional I'd need to check if we could host the 3 DB's 3 times over throughout SQL, or is that not how SQL clusters go? Or would good practice have just 1 SQL VM per Node holding all 3 DB's just twice?
Many thanks, coffee time.