2

I'm a bit of a newbie when it comes to storage, so I'd be grateful for any pointers!

I'm trying to plan out a small Windows Server 2012R2 Hyper-V cluster for an SMB consisting of 2 servers and since I wanted to include a failover option, should one of the hosts die, I wanted to include some sort of shared storage to enable the VMs to fail over to the second host if needed.

After doing some research, it would seem a direct-attached-storage box might be the best solution, or at least the best compromise between cost and performance.

If at all possible, I'd like to have some sort of redundancy for the storage and of course RAID comes to mind. Unfortunately according to this, it would seem RAID is not supported:

"The clustered storage pool MUST be comprised of Serial Attached SCSI (SAS) connected physical disks. Layering any form of storage subsystem, whether an internal RAID card or an external RAID box, regardless of being directly connected or connected via a storage fabric, is not supported."

...and this is where I get confused. My understanding is that whatever the DAS storage solution does internally (e.g. set up a RAID volume on a few disks and give the hosts access to said volume) should be completely transparent to the hosts themselves. Said hosts should be then able to use this volume to create a Windows Failover Cluster and a Hyper-V cluster after that.

So, on to my questions:

  1. does the article only apply to WFC storage configured using Storage Space? i.e. did I completely misunderstand it?
  2. Will I be able to use an SAS DAS box with a RAID volume (with SAS HBA cards for the hosts) to configure my cluster?

Thanks in advance!

  • Check the guide here: petri.com/create-windows-server-2012-r2-cluster-shared-volume – Overmind Jul 5 '18 at 9:08
  • @Overmind, that would seem to confirm my initial understanding that the LUNs can be created on top of a RAID pool, so the hosts shouldn't care or even be able to tell what's below the LUN. Is that right? – Alberto Jul 5 '18 at 9:16
  • Share storage- so you just replaced having independent server with having servers that rely on a single point of failure. Well done - total failure. – TomTom Jul 5 '18 at 12:06
-2

No, hosts shouldn't care what's below the LUN, but there are a few things you should consider:

  1. Organizing your data.

Consider a physical server for which you would organize the disks and files as follows: System files, including a page file, on one physical disk; Data files on another physical disk.

For an equivalent clustered VM, you should organize the volumes and files in a similar way: System files, including a page file, in a VHD file on one CSV, Data files in a VHD file on another CSV.

Try to keep the same rules when/if you add new VM hosts.

  1. Adding any disks to Available Storage

In Failover Cluster Manager, in the console tree, expand the name of the cluster, and then expand Storage. Right-click Disks, and then select Add Disk. A list appears showing the disks that can be added for use in a failover cluster. Select the LUN disk or disks you want to add, and then select OK. The disks are now assigned to the Available Storage group.

The disks can be the LUNs, they do not have to be physical disks.

You do not even have to use pools (depending on how you planned things to be done).

Practically, as long as you managed to create the LUN, what type of storage configuration is behind them is irrelevant. In my case, I use a Dell storage with SAS SSD for the high speed requirements and another Dell storage with HDDs as secondary and backup.

  • Both points make sense, thanks @Overmind, I'll definitely make sure to use "tiered storage" depending on the specific requirements. And great to have my suspicion confirmed that whatever is below a LUN is irrelevant to the host(s). – Alberto Jul 5 '18 at 12:58
2

You effetivtly shoot yourself into the foot. Total failure.

You replace the issue of one of the two servers failing with a single point of failure - the shared storage.

If you really want to go that way, your only way is to use a system that creates shared storage that can handle one server being down.

One way to do that - extensive outside it being R&D and thus covered by MSDN - is S2D, Storage Spaces Direct, preferable with a modern server, not something that outdated. Note that Storage Spaces DIRECT is NOT STorage Spaces - the later is in Windows Standard and doesn not handle cluster/failure scenarios. It also lacks the amazing NVME ssd cache capabilities.

A cheaper solution is to use something like Starwind. They have a software that does exactly what you want. And there are others.

  • Yes, I am aware that I have now just moved the point of failure from one host to one storage device. I was hoping to mitigate that using RAID and adding a second storage box in the near future and enable mirroring between the two. Still, thanks for pointing that out. – Alberto Jul 5 '18 at 13:02
  • 2
    Check starwind. – TomTom Jul 5 '18 at 13:04

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.