2

I have a single physical server with directly attached SAS storage that I run Hyper-V on and within it all of the machines on my lab (AD, Exchange, SQL, SCCM, SCOM, File Servers, etc).

I am trying to create a guest file server cluster with two nodes, so that I can easily patch the guests without having to worry about disconnecting any applications that are potentially using storage on the file servers.

For this purpose, I am looking to use shared VHDX files between the two file servers, but it seems that the host storage needs to be CSV for that, which in turn requires a one node cluster.

However, the issue I am encountering is that even though one node clusters are possible and supported, it seems that creating a cluster (of any kind) always requires the cluster nodes to be joined to AD (even when creating a AD Detached Cluster, new to Server 2012 R2). So in my particular case this appears to be a chicken and egg sort of situation:

The AD server is running as a VM on the physical server, but I can't have it running there before creating the CSV and to do so I need to have the physical server joined to the domain.

Any thoughts on this matter and is this something someone had done before?

By the way, I am really keen on trying the shared VHDX files so am not really looking for third party software alternatives like StarWind iSCSI software or so.

Thanks in advance guys.

BaronSamedi1958
  • 13,676
  • 1
  • 21
  • 53
JohnUbuntu
  • 123
  • 2

1 Answers1

1

You can install the SVHDX filter on a single Hyper-V host, without CSV. This, of course, defeats much of the point of Shared VHDX, as you have a single point of failure. And because of this, you're in unsupported territory. But since your question clearly spells out that you want a file server cluster on an unclustered host, this is probably the solution for you. The following blog post outlines the steps:

http://www.aidanfinn.com/?p=15145

Jake Oshins
  • 5,146
  • 18
  • 15
  • That is absolutely what I was after, thank you. One thing to note (which was implied from what the article said) is that the command needs to be re-run every time the host starts up, so I've created a scheduled task for it. This is something the article probably could have included. – JohnUbuntu Dec 28 '15 at 00:53