I am in the process of creating a Hyper-V 2012 R2 Cluster. I have 3 physical hosts (128GB RAM , dual Hex-core and 12 NICs each) and a SAN to play with. The SAN is a Virtual Storage system (Datacore), serves its virtual disks through iSCSI and I have the ability to create as many virtual disks (i.e. LUNS) as I require (and that I have capacity for !). We have used the Datacore SAN successfully for several years with an ESX cluster, but we are moving to Hyper-V due to licensing costs and the fact that is now offering features on par with what we currently use with ESX (we have also run a a couple of standalone Hyper-V servers for some years as a backup, so are so familiar with that technology too).
So this question is specifically about mixing clustering roles in 2012 R2
I have already created the Hyper-V cluster which use CSVs ( to store VM files) on several iSCSI LUNs from the SAN that all the hosts can see, but my next step is to configure some Highly available File Servers for general user use. I should point out that I do know the difference between using CSVs which are active-active and principally designed for application use (like Hyper-V) as opposed to shared storage as used by a file server cluster which is active-passive and I don't propose to do anything different from that: however there are a couple of different ways I could implement file servers.
The 1st main way I could do this is create a 'guest cluster' - that is 2 VMs as the cluster file server nodes (on different hosts). Their OS volumes would obviously be separate, however there are 2 ways I could get them to share their file storage volumes (i.e. where the shares reside)
a. They both share a common VHDX that obviously resides on the Hyper-V CSV
b. Or they use their iSCSI initiators and directly access a dedicated LUN on the SAN
The 2nd way is to not use Hyper-V at all for the File server nodes, but create a File server ( general Purpose) role (alongside the Hyper-V role) on the same hosts in the cluster : these would then access the shared storage on dedicated (non-CSV) LUNs presented by the SAN.
What are the pitfalls of each ? My feeling is that the second option actually has less overhead ( no Virtual layer and no VM required), but it does mean the hosts in the cluster are providing both a Clustered Hyper-V and a Clustered File Server service at the same time - is that going to be an issue ? I also think I could even leverage in some crude load-balancing, by splitting my File shares across 3 File Server roles on the cluster each set to principally run on one node ( when all is working !), and each using separate LUNS.
I appreciate the answer will rely on just how many VMs and such like that I plan to run, but just assume I will keep an eye on resources ( e.g. making sure the VMs don't take up all the host RAM) and I manage the NIC allocation properly so there are no bandwidth issues.
Is there any technical reason, at all, why I can't do option 2 ? Many thanks !