TSA - HyperV Fail

Why Does Microsoft Hyper-V Not Support NFS?

By On 2 June, 2011 · 8 Comments · In Opinion, Storage, Uncategorized, Virtualisation

I recently wrote an article covering storage for virtualisation and was surprised to find out that Microsoft does not support the use of NFS shares with Hyper-V.  At first I thought perhaps I was doing something wrong and that my configuration was in error.  But after a lot of lab testing and eventually finding some obscure forum posts, I’ve concluded that NFS cannot be used for storing Hyper-V guests.  But why?

Client for NFS

Microsoft has supported an NFS server and client within Windows for some time.  In Windows Server 2008, the Client for NFS can be added through the “Services for Network File System” role.  This creates a new MMC plugin on the start menu that allows administration of both Client and Server services for NFS.  From that point on there’s not much to configure in the GUI and all NFS shares are mapped via the command line, or within WMI for Windows Server 2008 R2 (look for future posts describing NFS installation and configuration).

Hyper-V Fail

Any attempts to create Hyper-V guests fail with the error code shown in the screenshot attached.  I’ve tried lots of options; pre-creating the VHD, importing the guest and so on, however all fail to allow a VHD to be created.  Where does this leave me?  Well, I can fall back to standard block-based Fibre Channel and iSCSI LUNs but this is potentially limiting if I’m looking to be more efficient with my Hyper-V installation.  SCVMM 2012 for instance assumes I’m storing VMs on an entire single LUN as it uses LUN snapshots to replicate virtual machines.  To make this kind of configuration work best, I’d need a storage system that does LUN level thin provisioning; this escalates my costs somewhat.

So the question is, why would Microsoft put in such a restriction?  It’s not as if network share-based guests are totally banned as I can use CIFS to store them.  Of course I wouldn’t want to do that because CIFS has some severe performance and integrity limitations that make it unsuitable.  Perhaps it’s just that Microsoft still don’t “get” storage.  After all, the latest recommendations for Exchange 2010 are to use DAS.  Redmond needs to embrace the use of storage agnostic connectivity for Hyper-V; it’s these kind of features that keep VMware ahead and for many will make ESXi a preferred option.

Tagged with: • • • • •
 
If you enjoyed this article, please consider sharing it!
Reddit Facebook Twitter Delicious Mixx StumbleUpon Digg
  • Sim Alam

    When you discount CIFS as a choice does that also include SMB2 and it’s performance improvements? Does that still have the integrity issues?

  • Carmen

    Hyper-V does not support booting virtual hard disks on CIFS or SMB 2.0 either actually. Also, SMB 2.0 (as opposed to old school CIFS 1.x) removes many of the problems you mention. SMB 2.0 would provide the same experience and performance (or greater in some scenarios) than NFS.

    The official reason is the fact that host-level backups using VSS are not possible with file streams, because VSS depends on block-level volume mapping. Hopefully the decision gets revisited in the future, since I am sure some admins would be fine with losing that capability in return for the flexibility of network-level storage.

    • http://www.brookend.com Chris Evans

      Carmen

      Thanks for that response. Ultimately, having the flexibility to use NFS isn’t an issue in terms of VSS loss as the storage array itself may offer similar functionality. I’ll certainly look more at SMB 2.0 (I am aware of the improvements) and that may appear as a future topic.

      Regards
      Chris

  • http://www.virsto.com Alex Miroshnichenko

    Chris,

    Unfortunately things are even worse than was mentioned here. With CIFS you’ll also loose an ability to cluster your VMs and do live migration. In fact the only way to support live migration was to use VHD on CSVs – and now you can do it with Virsto – than don’t need CSV and are free from the LUN/per-VM limitations and VS dependencies.

    Alex.

    • http://www.facebook.com/people/Michael-Bradley/100001591676304 Michael Bradley

      It would take an act of Congress (and you know how many acts get through Congress right now), to have MS and the Open Source community to implement 100% cross-support. Each wants to “have it my way”, just like Congress. Everyone has their own opinion on what’s best for the price. You get to choose, just like in the upcoming election…..

  • mark

    Could you imagine the uncomfortable conversations internally if MS were to support NFS but not CIFS. Support for Hyper-V using their own NAS protocol is a ways off and as Chris Evans says, the whole VSS thing is something else to be overcome.

    I bet if MS had implemented CIFS properly, back in the day, they’d support Hyper-V guests over CIFS and they’d only suck a 999 things rather than the current 1,000.

  • http://blog.davidwarburton.net DavidWarburton

    I think MS gets storage pretty well. The problem is that most of its customers don’t which caused a large amount of Exchange support calls to be around misconfigured SANs.

    By recommending DAS it makes it a cheaper implementation for many customers and easier for MS to support its own replication technology.

    I’ll have to admit to very limited HyperV experience but surely it makes sense for them to stick to VSS which is a proven and reliable technology?

    • http://www.facebook.com/people/Michael-Bradley/100001591676304 Michael Bradley

      I agree. Hopefully by now the author knows that MS was working on SMB3, which is in my opinion, far superior to NFS, even the most recent version of NFS. But MS didn’t leave VMWare users high and dry, they support with SC 2012, VMWare VMs running on NFS.

Subscribe to the Blog!

Subscribe to the Blog!

Don’t forget to subscribe to the blog for more content and monthly emails!