My organization has two identical servers, HP ProLiant DL380 G5. We do not have a SAN device. Each server has 8x 146GB HDD, and 48GB RAM.

One server is currently configured as an Exchange server. The other server is configured as a virtual host, with VMs for ePO, WSUS, SQL, File server, DHCP, and a few other services.

Obviously, an exchange server running with 48GB RAM is a bit wasteful, and we have zero machine redundancy for the virtual machines. We would like to fix that by making exchange a virtual machine, and using both physical servers as redundant machines.

In my last organization, we did this using a SAN device. The physical servers did not host the VMs or VHDs, they just pointed to the SAN.

My idea was to set up DFS-R across both physical servers, and then set up failover clustering, but I have no idea if this is anywhere near the right direction.

Can anyone please help me out? Thanks

Edit: After doing a bit of reading (namely this article: http://blogs.technet.com/b/josebda/archive/2008/07/16/failover-clustering-for-hyper-v-with-file-server-storage.aspx) is it possible for me to set up a cluster between the machines, making a highly-available file server, utilizing both machines, and then the highly-available VMs, utilizing both machines?


DFS-R doesn't help until the files in question are closed, and the virtual hard disk files will never be closed.

In order to get true redundancy with Hyper-V, you need to cluster the Hyper-V hosts. And that cluster must have an odd number of "witnesses" so that there is never a tie when it comes to deciding which set of machines is in charge when something goes wrong. When you use a SAN, you create a "witness LUN" which, along with the two hosts, which are also witnesses, makes an odd number. Windows Clusters can be configured so that a file server is used as a witness, but that's probably not going to meet your needs.

Even if you do get a cluster working, you have to consider how to make the storage reliable, too. A SAN is usually a cluster in its own right. It has redundant power domains, disks, processing, memory, etc., so that no one failure will cause it to fail. There are various software packages that can emulate that behavior, often by scavenging disk space on your available machines and replicating writes to multiple hosts. These might work for you, but they probably won't have the performance characteristics that you would want given only the two machines that you described.

If you can wait for Windows Server 2012, you might be satisfied with Hyper-V Replica. It will allow you to run virtual machines on each of your hosts and it will replicate the contents of the virtual hard disk to the other host, asynchronously. If a host crashes, you can restart the VM on the other host, having lost some amount of data that hadn't yet been replicated.

| improve this answer | |
  • This is a government network, and it will take years for Server 2012 to be approved. I guess I will wait until the next refresh of the equipment, which will have a SAN device. But that may be years.... – Mike Christiansen Jun 15 '12 at 6:48

You can try to cobble together some sort of solution where one of the servers would become a "shared storage device" that would allow the other to access the same data, but then you can only lose the other server.

If you want real clustering and failover, you need shared storage.

| improve this answer | |
  • I know it is probably not the best, but would DFS-R fit this bill? Both servers would contain copies of all of the data.... – Mike Christiansen Jun 14 '12 at 7:41
  • 1
    No it wouldn't, as DFS-R only works when the VHD files are closed. – Jake Oshins Jun 14 '12 at 21:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.