I am looking for best practices regarding setting up volumes for VMs running on Hyper-V R2/Windows 2008 R2 Hyper-V clusters. In the old days I was running 1 VM per volume. I would snapshot the volumes and replicate them. If I needed to recover a VM, I could just revert the volume to an earlier version. If I wanted to clone a VM, I could clone the volume or mount a snapshot. Pretty easy.

Now I am runnung VMs in a 4 node Hyper-V Server 2008 R2 cluster. Each host has 2 NICs connecting to my iSCSI SAN. I have two members in my storage group. Each volume takes up (two nics per host * two group members * 4 nodes) 16 connections to the storage pool. EqualLogic has a hard limit of 512 connections to a pool, so with my current configuration I can have only 32 fully clustered volumes. Adding member arrays actually reduces the number of volumes I can have.

Fortunately, Microsoft released Cluster Shared Volumes, so I have consolidated most of my virtual machines to a single large volume. CSV makes some things easier, like making sure everything gets snapshotted and replicated, but I can't just clone a single VM, and I shudder to think about trying to restore a single VM. It will probably take an hour or so just to mount up a snapshot of all of my VMs and then copy a 20-30 GB VHD file from a temporary volume back to the main volume. How do other people manage their VMs and volumes in a cluster? I have 45 running VMs and 83 volumes in use.

Interesting topic.

Regarding cloning a VM I would highly suggest using SCVMM 2008 R2 (System Center Virtual Machine Manager) which I hope you're already using? No Hyper-V cluster should be really be used without it.

I'm currently planning to convert my hardware from a lab cluster to setting it as a production Hyper-V R2 cluster, entirely from scratch. Hoping to document as much as possible to that I can post it here on Dell TechCenter. Regarding volumes I have also pondered about how I will set it up. I'm going to use CSV and will most likely create a couple of fairly large volumes (maybe around 1TB) but I haven't found any best practices about this.

I'm not going to replicate but instead backup my VMs using SCDPM 2010 (System Center Data Protection Manager).

Hopefully someone else can chime in on this thread with suggestions about how to go about this. Maybe JeffS can ask EqualLogic if they have some best practices document in the works?

I forgot to ask about placement of the pagefiles for VMs. I have found that leaving the pagefile in the default location for a Windows machine can increase be several times the about of change on that VHD file that ends up getting replicated. A VM with 100 MB of daily change may end up with 1 GB of daily change when the pagefile is included. I have ended up creating a new CSV that is not replicated, and putting a VHD for each VM on the new volume and dedicating those VHD files to the pagefile. This way I can migrate VMs within the cluster, but I still would need to recreate the VHD file on the DR cluster, and then regreate the pagefile on that new VHD file should I ever stand up the VM in a new cluster. I wonder who else has tried to tackle this. It certainly makes automated recovery a little bit more challenging.

khsieh,

The Dell VSE team has done testing and validation around this. They will be launching a Business Ready Config around Hyper-V R2 and EQL storage arrays. Those docs should contain the best practices from Dell PG engineering. The docs should be live in the next few weeks.

Is there an updated date when the docs will be available? Is equallogic planning on bulding ASM integration with SCVMM as opposed to installing ASM on every hyper-v host (ie VMWare VCenter ASM)?

Any update on the status of these docs?

Thx

We're very interested in these docs. Are these available yet?

This document would be really helpfull! Please post a link if it is available.

http://www.equallogic.com/partnerships/default.aspx?id=8677 has some good references

and in particular try looking at this technical paper - http://www.equallogic.com/resourcecenter/assetview.aspx?id=7787 (sorry you have to register to get the pdf)

I've looked at that document and quite honestly it doesn't provide any details about all the differences with R2 and CSV which is exactly what we are all looking for... how the CSV stuff has impacts how we should configure our EQ units in conjunction with clustering the system.

The biggest question I have is specifically surrounding VSS snapshots and CSV's with a PS array. Using a single LUN - can we still have Hyper-V VSS snapshots that are stored on the array?

EqualLogic arrays fully support Hyper-V clusters using CSV volumes. The Auto-Snapshot manager tool in our Host Integration tools for Microsoft does not yet support CSV volumes due to the Microsoft supplied VSS writer for CSV being unable to quiese all of the virtual machines in a Hyper-V cluster simultaneously. We are working with Microsoft to improve their CSV VSS writer

When is the update on the support for ASM and CSV in Windows 2008 R2 -- We are now almost 5 months past RTM.

"When is the update on the support for ASM and CSV in Windows 2008 R2 -- We are now almost 5 months past RTM.
"

Yes I am also eager to know when we will have support for ASM and CSV. It's been a while since Windows 2008 R2 has been released.

The support for CSV with ASM is currently in beta and will release around the Q2 timeframe.