Help with Clustered Shared Volumes

jay-garrick · 2018-08-29T15:16:02.000Z

Hi All,

I have a question regarding Clustered Shared Volumes.

I created a new cluster and have been testing failovers with test hyper-v virtual machines to make sure they function properly. I have used the process where creating the hyper V disk with a drive letter, set it up on the Cluster, create the virtual machine, then move it with Quick and Live Migrations to see if the failover works. I have tried without using Cluster Shared Volumes and with it.

We have a 2 node failover cluster – Server 1 and Server 2 with a Hyper-V virtual machine on it. Typically, from what I have been told where I work, when a failover happens, the role of the hyper-V moves from Server 1 to Server 2 or vice versa and the disk moves with the role. For Example:

Failover from Server 1 to Server 2

Role and Disk – Server 1 → Server 2

Failback from Server 2 back to Server 1

Role and Disk – Server 2 → Server 1

Using Cluster Shared Volumes, the disk does not follow the role ( and vice versa) when it fails over. For Example:

Failover from Server 1 to Server 2

Role – Server 1 → Server 2

Disk – Stays with Server 1

Failback from Server 2 back to Server 1

Role – Server 2 → Server 1

Disk – Still on Server 1

I can also failover the disk from Server 1 to Server 2 and the Role does not failover with the disk.

If I shut Server 1 down and either the Role or the Disk or Both are on Server 1, they failover to Server 2, so it is functioning the way I would expect it to.

It is also interesting to note that I can have duplicate Disk Numbers in the Failover Cluster Manager as long as the disks are owned by different nodes.

I wanted to get some opinions from others who have more experience with Clustered Shared Volumes than I do and let me know if what I’m seeing is what is supposed to happen with the Roles and Disks

kevinhsieh · 2018-08-29T15:41:37.000Z

With CSV, the owner of the disk and owner of the VM is independent, which is pretty nice. They also don’t consume any drive letters, which is also nice.

You can have a reasonable number of VMs on s single CSV. I like to keep my CSV at less than 2 TB each. I wouldn’t put 100 high IO VMs on a CSV. No issue with 20 light IO ones, as long as CSV size isn’t too large.

The behavior you are seeing is correct.

I like to rename my CSV, so instead of c:\ClusterStorage\Volume 1, I rename it to match my SAN volume name, like C:\ClusterStorage\VMs1. It can be done via standard directory rename via Windows Explorer or command prompt from the host that owns the disk.

Congratulations on moving to CSV. You’ve now catching up to 2009!

bruceforbes7564 · 2018-08-31T10:52:29.000Z

I agree. We have a 3 node Hyper-V cluster and have 2 CSVs attached. We have about 40 VMs running. We also as a rule keep our CSV limit at the 2TB mark.

I have no issues moving a CSV from one host to another, live migrating a VM from one host to another and I have even live migrated quite a few VMs from one CSV to the other. Other than a very slight pause the VMs continue to run and nobody notices anything.

Cluster shared volumes are the way to go for Hyper-V clusters. It just makes life easier.