I've been running Hyper-V 2012 Server core for some time and using ReFS for the disk that contains my VM / VHDX files.

Recently I've been trying to get Hyper-V Replica working.  It seems as though there is some bug in ReFS that caused me to lose a folder and all the VHDX files in that folder?  The VM that lost the folder was running Windows Home Server 2011 (which is based off Windows Server 2008).

My system has a RAID6 data drive using six Western Digital 3 Tb (WD30EFRX) drives.  (My boot drive is setup as NTFS.)  I had this one VM that didn't come up after rebooting Hyper-V.  I found that the VM Snapshots and Virtual Machines folders were still present, but the Virtual Hard Disks folder was gone.

Because I was testing Hyper-V Replica it is possible that I manually deleted the VHDX files by mistake, but had I done that I would have also deleted the Snapshots and Virtual Machines folder at the same time.  And if this VM had been deleted from the Hyper-V Manager I would have expected the Virtual Machines folder to be empty.  (I don't recall the exact filenames, but there were two files in this folder.  I think .bin and .bsv?)

The other thing that makes me think ReFS ate this folder is that if I do a DIR command I only see 5 Tb of available disk space.  I rebooted the system just in case something had to happen for the missing space to show up, but that didn't make any difference. I still see only 5 Tb of free space.  Yet if I use diskpart it shows this volume is 10 Tb in size.  I have one VM with a 1.5 Tb VHDX and there are a few VMs that have a few hundred Mb VHDX files.  But that leaves 3-4 Tb of space unaccounted for from the DIR command output.  The WHS Virtual Hard Disks folder that is missing had a 4 Tb VHDX and two smaller VHDX in it.  Although I can no longer see this folder given the amount of free space it seems that ReFS thinks that space is in use on the RAID6 drive ... and also seems to point at ReFS as the reason the folder is missing?

In checking I've found a couple people who mentioned VHDX should not be used in a clustered (CSV) configuration.  But no one else who seems to have mentioned these same symptoms.  I have not seen anything from Microsoft saying that ReFS should not work in the configuration I've been using.

Given the enhancments in ReFS I would prefer to use it over NTFS, but I'm wondering if ReFS is not stable ... or at least not stable for this type of configuration?

Anyone have any insight on this topic?

T

VR38DETT,

It could be possible that I mistakenly deleted the files and folder so I'll looking into the audit function you mentioned.

But if I had done that I would expect the DIR command should show about 8-9 Tb of free space.  Given it is only showing 5 Tb of free space (for a RAID6 drive which DISKPART shows 10 Tb) that looks to me as though ReFS has lost something?

Thanks again.

1) There are a lot of issues with ReFS. Bing to find more. We don't have it in production and don't have any plans to have one at least before R2 of Windows Server 2012 will be selling for ~6 months. 

2) In this particular case I have an impression it's more of a human / misconfiguration error then something about ReFS. With data or metadata damaged you'd see errors inside VHD but not missing content. Turning on audit will help to trace back to find who's who (next time of course not now). 

VR38DETT,

It could be possible that I mistakenly deleted the files and folder so I'll looking into the audit function you mentioned.

But if I had done that I would expect the DIR command should show about 8-9 Tb of free space.  Given it is only showing 5 Tb of free space (for a RAID6 drive which DISKPART shows 10 Tb) that looks to me as though ReFS has lost something?

Thanks again.

Check these links about Windows Server 2012 (R2) relation audit information:

http://technet.microsoft.com/en-us/library/hh831476.aspx

http://technet.microsoft.com/en-us/library/hh849642.aspx

Hope this helped a bit :)

VR38DETT,

I skimmed over those articles.  Looks like AD is required for this audit function?  That's a bit more advanced than what I can handle.  I'm just running in a WORKGROUP configuration.

I did come across a number of other people who have run into somewhat similar problems.  Several mentioned RAID and power loss as factors that could have been similar to what happened to me.

I lost a few things that were only on the ReFS drive, but most I can recover (after several hours of work).

At this point it doesn't look like ReFS should be used for critical data.

For some similar reports of ReFS problems see these links.

http://social.technet.microsoft.com/Forums/windowsserver/en-US/171a1808-157e-4ef9-b0dd-8a507ff6fcef/refs-corruption-when-filled-to-capacity

http://social.technet.microsoft.com/Forums/windowsserver/en-US/2de3eb18-a5f7-4752-bd2e-be199c5bcd41/refs-volume-lost

http://www.neowin.net/forum/topic/1144592-any-early-feedback-on-refs/

http://www.neowin.net/forum/topic/1136796-just-learned-the-hard-way-not-to-use-new-file-system-technologies/?p=595524650#entry595524650

I'll wait a little while to see if anyone else has something to contribute to this topic before closing this thread.

Thanks.

1) You can find third-party software tools for audit. I'm pretty sure they will not ask for AD.

2) Yes, they had lost the data but they don't have the healthy volumes - that's the point :)