Any early feedback on ReFS?

[Stilez]

Coverage of the Server 2012 filing system "ReFS" seems to be two kinds - people saying how useful it is for data resilience compared to NTFS, and people saying it's not yet proven and a new filing system can have obscure bugs so leave it for now. (Ignoring feature issues such as permissions and streams, as these aren't affecting resilience)

Both seem to be sane advice, but what I'm not seeing is whether there is any early anecdotal evidence of adopters reporting issues with ReFS that impact data loss or other problems, or if it seems to be a good job and tentatively reasonably reliable. I've tried to find this in Microsoft's forums and Google, but no real luck.

Is anyone else tracking ReFS feedback, or trying ReFS, and if so what has your experience been so far? Have there been any hints of issues you've heard about? How trustworthy does it seem from a "random data loss" perspective? Is anyone out there using it heavily as an early adopter, or finding problems?

I'm looking for actual posts and comments (or general blogosphere silence which might show "lack of reported issues"?) showing people's real experience who have tried it or are using it, or posted about trying it.

Thanks!

[aexphoric]

I've been playing with Windows Server 2012 Storage Pools with ReFS since September last year and although I haven't benchmarked or stress tested it, it's been running on my HP Microserver with no issues whatsoever.

[Eric Veteran]

[Thread cleaned]

Please keep the discussion of thread topics here. (Do not crosslink to other sites, etc.)

[+Fahim S. MVC]

This is all I have seen in the way of negative feedback:

https://www.neowin.net/forum/topic/1136796-just-learned-the-hard-way-not-to-use-new-file-system-technologies/page__p__595524650#entry595524650

[TPreston]

I have a 10TB ReFS volume on a 3ware RAID card, No issues the only improvement ive noticed are greatly increased speed when replacing permissions on an entire drive.

The only issue I have is the lack of support in practically every application out there from Diskeeper to Paragon Hard Disk Manager (even though its supposed to now support Server 2012), File recovery programs and Windows 8 (Which is a PITA! Since we are ment to use it for admin'ing server 2012 although this can be remedied with a package released online bring support to windows 8).

So in summery I use it but know that you will not be able to use any of the above applications.

[aexphoric]

Ah one problem I did have was that you cannot store Windows Deployment Services data on it, (annoyingly). Has to be NTFS.

[SirEvan]

Yeah... DON'T USE IT. SERIOUSLY. When I upgraded my home file server, I made three RAID arrays. a 10TB RAID-5 for Media, a 3TB RAID-10 for Hyper-V VMs, and a 2TB RAID-1 for pc backups. Every File System was formatted to NTFS, except the RAID-10, I decided to try ReFS for that, thinking it would be more resilient and better for VMs. Then one day, the power connector on my SAS Expander had come loose, or a pin didn't touch...something happened, and when I fixed it, and rebooted the box, My "E:" drive was gone from windows. Did some checking, did some searching, and low and behold the answer on technet to a missing RAID drive with ReFS? "Rebuild it from backups".

Something happened that caused either the file system itself or windows to think the array was empty, as in nothing used (said 3/3TB free). Fortunately I had backups from a week or two prior, so I didn't loose much, except a week of email (exchange VM), however I will NEVER use ReFS again until it's been around for a few years like NTFS, and can prove itself. If I were you...I'd steer clear. It does not handle power loss AT ALL. The two NTFS volumes? not a scratch on them..they were just fine.

[TPreston]

^ I don't get this, My write cache is enabled and I have hit the reset button plenty of times with no issues. How would this be a file system issue ? If you've got redundancy surely the file system would be happy.

[articuno1au]

I did a lot of research into it before I rebuilt my media server (running Windows Server 2012 Datacenter).

I don't believe SirEvans issue is was related to the ReFS. Apart from anything else, NTFS can die in the same situation.. Not much you can do about faulty install/hardware issues.

From everything I read (and my own limited testing) performance is pretty close. In all situations it was within error rate of NTFS, barring a few blips. Permissions are managed differently on the drive (as observed above) so they can be updated much more rapidly. There are however, other issues. I couldn't get MSSQL to talk to my ReFS hosted database. It didn't like it one bit. There's also a lot of anecdotal evidence talking about media systems bombing out with it (in particular PLEX). This is in part due to limited MS support atm, and some NTFS specific code in some apps.

Whilst it is more redundant (and pretty much does away with check disk :D (so much hooray)), it's just not ready for the prime time in my opinion due to lack of support. I'll be pleased when it's fully supported and usable. It seems to be an enormous technical improvement :)

I considered using it on our external backup media, but it gives no benefit over NTFS in this case as the backup drives are snapshotted, so they're never undergoing constant change (which is the biggest benefit scenario for ReFS).

tl;dr - Avoid for now :)

[SirEvan]

^ I don't get this, My write cache is enabled and I have hit the reset button plenty of times with no issues. How would this be a file system issue ? If you've got redundancy surely the file system would be happy.

http://social.techne...d-8a507ff6fcef/

^^ Read that. Basically the same issue I experienced

We have a 100GB ReFS volume. The volume was filled to capacity and now the volume is unaccesible. If you click on the drive letter in Windows Explorer it displays the following "W:\ is not accessible. The volume repair was not successful." There are several events in the system event log for event ID 133. "The file system detected a checksum error and was not able to correct it. The name of the file or folder is "<unable to determine file name>"." I have tried added additional space to the drive via Disk Management as well but it failed with the following error. "The volume cannot be extended because the volume does not contain a recognized file system."

Anyone have any ideas on how to fix an ReFS volume?

So far I am not too impressed with this "resilient" file system...

Hi,

Yes, the same space of total space and free space is the reason why drive could not be restored correctly and we recieved the error message. For RAID 5 corruption, currently there is no fix to fix the issue except rebuilding. Thanks.

After electrical break at shutdown process 13Tb ReFS volume becomes unavailable with same symptoms as in Nick case. Volume filled at ~8Tb.

Volume based on Logical Disk hosted on LSI Hardware RAID controller. Level RAID-5. Data is very sensitive and should be restored.

My case was similar, after an electrical break, the ReFS formatted volume (mirrored on two hard drives) did not survived. What is funny, NTFS formatted volume located on the same hard drives survived (chkdisk corrected all errors)

After the incident, both hard drives are healthy (no reallocated, pending or unstable sectors).

I agree that the concept of self-healing file system is the future, but as we see, the ReFS is not ready yet. Also I could not find any data recovery option on the local market yet.

In my opinion, people should be warned. The documentation should contain huge warning "[experimental, potential data loss possible]"

I emphasized the relevant post from a Microsoft employee which basically says "Sorry..you lost your data.. bye" Not very resilient if you ask me.

[rodger91]

I have been playing around with Windows 2012 storage in general.

My experience with ReFS has been good, and I've run volumes out of space repeatedly.

I have a 45TB volume in service at the minute.

My feedback would be -

Performance is 15-20% better over NTFS in our workflow. (Principally moving around large media files).

Use an industrial strength Array Controller i.e. equivalent of HP P822. You NEED the fast rebuild time and ease of managing failures compared with Storage Spaces.

[HawkMan]

http://social.techne...d-8a507ff6fcef/

^^ Read that. Basically the same issue I experienced

I emphasized the relevant post from a Microsoft employee which basically says "Sorry..you lost your data.. bye" Not very resilient if you ask me.

None of that means the same couldn't have happened with NTFS though.

[SirEvan]

None of that means the same couldn't have happened with NTFS though.

True, but all I'm saying from first hand experience is that I had 3 Arrays on the same controller. 2x NTFS, 1x ReFS. The ReFS was the only one that was obliterated by the power loss to the expander sitting between the controller and the drives, and all other factors were equal, except NTFS vs. ReFS.

I'll wait a long time before venturing back into the ReFS area, I seriously doubt it's resiliency after nearly losing a ton of data.

[TPreston]

Its gotta be a hardware issue because my 2 REFS arrays have been through many a rebuild and are good as new.

The only thing I miss is the ability to use Diskeeper.