3

I'm planning a Hyper-V virtualization system what will include:

  • 2 very nice dual 16 core servers
  • Dell MD3600f fibre channel SAN

It will be running about 10 virtual machines. The plan is to setup one LUN on the SAN to store all the OS VHD files using Cluster Shared Volumes, and seperate LUNs added to the cluster (but not CSV) for each server that requires a data partition. This should allow live migration and failover if I've read everything correctly.

Now here's the catch: I only get 1 server and the SAN to start with. We will add the second server sometime down the road.

Will Server 2008R2 let me setup a cluster with only 1 server in it? Is that the best way to go, so I can have the hyper-v using cluster shared volumes from the start? And then simply add the second server when I get it?

If it's not possible (or there is a reason why I shouldn't do it that way) how do you go about migrating existing virtual machines on the SAN to a cluster shared volume?

Edit: the server also has 96gb ram with room to upgrade later on.

2
  • 2
    Side note; you mention dual 16-core servers, which is great and all, but most VM Hosts are RAM bound, not CPU bound. I highly recommend dumping as much into RAM as you can. Our next round of VM Hosts will have 96GB each, and I'm considering if I shouldn't be getting more.
    – Chris S
    Apr 6, 2012 at 13:46
  • 1
    Absolutely true statement. We've found that RAM is by far the limiting factor on our VM clusters... our new servers are Dell R610s with just middle of the road Xeon E5620s but with 96GB... we would have bought even more RAM if we could afford it :P
    – newmanth
    Apr 6, 2012 at 16:13

1 Answer 1

4

This should be possible. You will simply create a single-node cluster. When you're ready to create the second node, you'll expand the cluster via the Failover Cluster snap-in.

There are a couple of things you should be mindful of:

  1. When adding the second node, Windows will try to get you to run the analyzer. If you do, it can potentially cause an interruption of service (if you allow it to check the shared storage). I would recommend skipping this part of the analyzer. Of course, that means you'll need to be paying attention when it comes to ensuring correct configuration.

  2. Quorum configuration will be incorrect for the cluster. On a single-node cluster, the service will configure node majority by default. You will need change this to node and disk majority (and have a shared quorum volume ready) when you first configure the cluster or when you add the second node. If it's only going to be a short time before you add the second node, I would recommend configuring the quorum disk at initial setup to minimize any configuration changes when adding the second node.

2
  • As an aside, you'll need to also configure your SAN ahead of time with the expectation that a second SAN node will be added later. My shop uses iSCSI SANs (so I'm not familiar with FC configuration, but I'm guessing it has some type of functionality similar to iSCSI virtual group IPs and multipathing).
    – newmanth
    Apr 6, 2012 at 13:42
  • +1 You have to start with a cluster of 1, then add nodes to it. The warnings are helpful; most especially "you'll need to be paying attention when it comes to ensuring correct configuration".
    – Chris S
    Apr 6, 2012 at 13:44

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .