Jump to content
  • 0

Adding REFS drives with data to a pool results in no data showing in pooled volume


Royce Daniel

Question

Hi...

 

I have 4x 4TB drives with data on it all formatted ReFS. I mounted the drives via a folder then created a pool. I can see the files via Explorer by navigating to and browsing the mount location. See below...

 

post-2630-0-91729000-1474268186_thumb.jpg

 

But if I goto the actual pooled volume that DrivePool creates I don't see the files. I thought you could import drives into a pool without having to re-copy the files. In DrivePool it shows the drives data as "Other".

 

post-2630-0-02684800-1474268188_thumb.jpg

 

 

Link to comment
Share on other sites

Recommended Posts

  • 0

Could you upgrade to the latest version? 

http://dl.covecube.com/DrivePoolWindows/beta/download/StableBit.DrivePool_2.2.0.711_x64_BETA.exe

This actually looks to be NFS server related, and ... there is a fix related to that in the newer version already. 

 

 

And if it continues to happen, grab that crash dump as well (also, you can zip it up, and it should reduce it's size SIGNIFICANTLY). 

Link to comment
Share on other sites

  • 0

Brand new hard drives that have never been used in Storage Spaces. Yes, formatted through disk manager with ReFS but I noticed that when the pool was done being created and the volume mounted, it was NTFS. It made a pooled volume from the free space of the 4 drives.

I tried moving one directory via Explorer from the mount point directory into the DrivePool volume and then used Re-measure... my server BSOD. I'm guessing the server didn't like having 4 freshly formatted ReFS volumes with a DrivePool NTFS volume on top of it in the free space. I'm going to try and break the pool back apart and make a DrivePool volume with an empty drive all ReFS, THEN add the x4 4TB drives back into the pool. I was just wondering if someone else ran into this. Since ReFS is technically still in Beta I thought I would point this issue out.

Link to comment
Share on other sites

  • 0

hi

 

The pool will show up as ntfs under disk management even with refs disks don't no why it's something that is on the todo list for Alex but it doesn't effect anything.

 

Drive pool does not add existing data already on the disk to the pool common practice for this is to just drag and drop any files outside the pool parts folders into the pool parts folder then either force a re-measure or restart the server which will kick in the auto re-measure.

Link to comment
Share on other sites

  • 0

Then at this point my main concern is resolving the BSOD. I can't imagine having an NTFS volume on top of an ReFS volume being a good thing. Sort of defeats the purpose of leveraging ReFS; virtual volume or not. My disk array is a bank of 48 SAS disks all connected via an HBA controller. I hope I'm not running into some kind of compatibility issue. The BSODs error indicates a Kernel panic not a driver issue. I'm still going to try the steps above to see if I can get around the BSOD. Instead of dumping all 4 drives at once onto the DrivePool I'm going to add them one at a time and reboot after each addition for good measure then examine the logs.

Link to comment
Share on other sites

  • 0

Ahhhh... LOL. I miss-read that! God I was tired last night. :D
 

Pooled Disks Can Continue to Store Non-Pooled Data
  • You can continue to use any disk added to the pool, just like you did before it was part of the pool.
  • Pooled disks will continue to be exposed as a separate drive, just as they were before, and can contain non-pooled data.
  • Disk and file I/O is not filtered or altered in any way by StableBit DrivePool when you access any pooled disk directly.
  • Essentially, nothing is altered about the disk when it becomes part of the pool, aside from it now being able to store pooled files.

 

 

Link to comment
Share on other sites

  • 0

Thanks for the screenshots. I'm glad you don't appear to be having any issues. :)

 

My concern is running an NTFS volume on top of ReFS. I'm not sure what DrivePool is using - logical? or virtual? on top of physical. I assume it's using a virtual layer since the DrivePool "partition" doesn't show up in a disk manager. My assumption is that the container for the virtual DrivePool volume resides on top of an ReFS formatted logical volume. There are known performance issues with storing Hyper-V VHD's on an ReFS volume. If my assumptions are correct this is a similar situation - storing files on a virtual NTFS volume  Also, ReFS won't protect you from BitRot if your files are stored on an NTFS virtual volume even if it's stored on an ReFS logical volume.

 

I have been running Storage Spaces for a couple of years and when I did my initial build I ran some performance tests with both an ReFS logical volume with NTFS virtual volumes and compared it against an ReFS logical volume with ReFS virtual volumes. The ReFS / ReFS setup had faster read/write & random numbers. It was also faster over a network. I easily saturated my 1gb pipe - client to server & a 4gb pipe from server to server. I'm wondering if I should also setup DrivePool the same way.

Link to comment
Share on other sites

  • 0

Hi

 

No it's not possible to format the virtual pool only the individual drives in the pool I haven't suffered any ill effects from running like this but I know Alex (the developer) is hoping to take a look at refs to maybe/hopefully try implement some of its features. I believe cloud drive is priority at the moment.

 

I haven't suffered any corrupt files since using refs this maybe just coincidence but seems to be working well for me but time will tell. I get around 116mbs transfer speeds between network computers which i believe is about max for a standard spinner. I am sure you already know but the drives within drivepool remain individual which has its benefits but if your after mega fast transfer speeds you won't get them like raid solutions.

Link to comment
Share on other sites

  • 0

Well you're limited by your network bandwidth which currently most have Gigabit... I'm still waiting for 10gbps gear to come down in price. $700 for a decent switch is still kind of steep for home / media streaming / file server archive use. I'm using link aggregation between servers right now and that is more than sufficient for now... I wish they'd hurry up and fix LAG for Windows 10 clients. bah...

Link to comment
Share on other sites

  • 0

O.K. So now I understand the underlying architecture. The StableBit DrivePool is actually a "fake" volume... at least fake from the perspective of Windows Explorer and Disk Manager. It's labeled as an NTFS volume but it's really not. AND you CAN add a bunch of ReFS drives to a DrivePool with data and move that data into the DrivePool without having to re-copy any files. When you add an already formatted volume to a DrivePool it creates a hidden folder named (in my case) PoolPart.5de19635-8ccf-4b71-8cc6-e389a29406c3. Each GUID will be different on each drive that you add to the DrivePool. This PoolPart folder is your DrivePool root directory. Upon first adding a drive with data, all your data will be in the normal root directory of the hard drive (top arrow). Just move all the data into the PoolPart folder following the layout of your folder hierarchy in each drive (bottom arrow). The screenshot shows each HD mounted to a folder instead of a drive letter so your view might be different from mine. I mounted the DrivePool drive to an actual Drive Letter.

 

post-2630-0-47849600-1474336629_thumb.jpg

 

To answer my earlier question about having an NTFS virtual volume on top of an ReFS logical volume... it's not a TRUE virtual volume it's a linked directory hierarchy made to look like a drive volume - similar to DFS? It's a true ReFS volume with all the benefits therein. Unless there are some other fancy things happening under the hood that I can't see and of course I could be totally wrong here but, I moved 12TB worth of data into a fresh DrivePool in under 5 seconds and it all showed up properly in the mounted DrivePool root. Most storage nerds will tell you that it's not really moving the files around on the storage media but more or less updating the volumes metadata; that's why it's so fast compared to copying or copying/moving to another hard drive.

 

/cheers

Link to comment
Share on other sites

  • 0

Hi...

 

I have 4x 4TB drives with data on it all formatted ReFS. I mounted the drives via a folder then created a pool. I can see the files via Explorer by navigating to and browsing the mount location. See below...

 

attachicon.gifCapture.JPG

 

But if I goto the actual pooled volume that DrivePool creates I don't see the files. I thought you could import drives into a pool without having to re-copy the files. In DrivePool it shows the drives data as "Other".

 

attachicon.gifCapture2.JPG

 

 

You added the disks to the pool, but did you move the existing data into the pool folder structure? 

 

If not: 

http://wiki.covecube.com/StableBit_DrivePool_Q4142489

O.K. So now I understand the underlying architecture. The StableBit DrivePool is actually a "fake" volume... at least fake from the perspective of Windows Explorer and Disk Manager. It's labeled as an NTFS volume but it's really not. AND you CAN add a bunch of ReFS drives to a DrivePool with data and move that data into the DrivePool without having to re-copy any files. When you add an already formatted volume to a DrivePool it creates a hidden folder named (in my case) PoolPart.5de19635-8ccf-4b71-8cc6-e389a29406c3. Each GUID will be different on each drive that you add to the DrivePool. This PoolPart folder is your DrivePool root directory. Upon first adding a drive with data, all your data will be in the normal root directory of the hard drive (top arrow). Just move all the data into the PoolPart folder following the layout of your folder hierarchy in each drive (bottom arrow). The screenshot shows each HD mounted to a folder instead of a drive letter so your view might be different from mine. I mounted the DrivePool drive to an actual Drive Letter.

 

attachicon.gifCapture.JPG

 

To answer my earlier question about having an NTFS virtual volume on top of an ReFS logical volume... it's not a TRUE virtual volume it's a linked directory hierarchy made to look like a drive volume - similar to DFS? It's a true ReFS volume with all the benefits therein. Unless there are some other fancy things happening under the hood that I can't see and of course I could be totally wrong here but, I moved 12TB worth of data into a fresh DrivePool in under 5 seconds and it all showed up properly in the mounted DrivePool root. Most storage nerds will tell you that it's not really moving the files around on the storage media but more or less updating the volumes metadata; that's why it's so fast compared to copying or copying/moving to another hard drive.

 

/cheers

 

 

The GUID (the part after PoolPart.) is the poolpart ID.  It's used to identify/track the disks.

 

As for the the file system it's similar to DFS, but not.  A better analogy is that the DrivePool driver works as a reverse proxy and load balancer for the file systems. 

 

 

Right now, ReFS isn't officially supported, and is something we do plan on looking into/improving in the future.  So for now, it shows up as NTFS, rather than ReFS.  

Link to comment
Share on other sites

  • 0

For future can we make sure both types work within the pool?  I have data aware SAN's and tiered data on them that has to stay in the NTFS format to maintain compatibility with the SAN's, but those SAN's are pooled with stand-alone drives that are ReFS.  That's working as it is, but if the pool has to be of homogeneous file type, it would prevent me from operating the pool as it is, and I could never move to ReFS in the foreseeable future...

Link to comment
Share on other sites

  • 0

Right now, both types do work in the pool. 

 

However, I supposed we could add an advanced option to allow mixing for those that understand the risks, and/or need to convert the pool over (such as for myself, where it's not feasible to switch everything over at once). 

 

 

 

But as for the latest version, we've disabled "mixing" unless the pool is already mixed.  

 

 

 

 

https://stablebit.com/Admin/IssueAnalysis/27189

Link to comment
Share on other sites

  • 0

I may be missing the issue - what's the down side of having both types in the pool?  

feature parity. 

 

There are a number of NTFS features that ReFS doesn't support, for instance. 

  • NTFS compression
  • Quotas
  • hard links
  • File level encryption (EFS)

And there are things that ReFS supports that NTFS doesn't: 

  • Integrity streams
  • Copy on Write (reduces the likelihood of data loss due to power failure)
  • self healing (in multiple forms/levels) 
  • Large volumes
  • file names longer than 255 characters

 

Also, API and file system calls may not be the same, causing issues with the pool.  

 

So, there are good reasons to not mix and match.  

 

However, I am pushing for an advanced setting to enable this, for the edge cases that may need it. 

 

 

 

That said, if the pool is already mixed, that isn't an issue. It will allow continued operation.  It should just prevent you from creating a new pool with mixed file systems.

Link to comment
Share on other sites

  • 0

Yes. I moved all the files from the root of the logical volume into the DrivePoolPart folder and it showed up in my mounted DrivePool drive. Hence my reference to it only taking a few seconds lending credence that it was just a simple file move on the same drive. Of course I had to move blocks of files from the individual drive roots outside the DrivePool mount point into the DrivePoolPart folder.

 

I fully understand that ReFS isn't supported yet and the NTFS volume that's referenced in Disk Manager is only a sudo reference. It's not a true NTFS volume. I'm wondering... when I Right Click on the DrivePool volume and try to enter Properties, it causes my server to BSOD. I have to use the Shared Folders Snap-In to create network file shares. Is this what you're referring to in regards to the API not being fully linked or the fact that it "looks" like an NTFS volume so Windows is being stupid and trying to execute NTFS API's on an ReFS volume?

Link to comment
Share on other sites

  • 0

Well, it shouldn't be BSODing!  

 

Likely, that is a bug somewhere.  But it could be because of the mismatched file system. 

 

 

What OS are you on, and what version of StableBit DrivePool are you using?

And could you do this: 

http://wiki.covecube.com/StableBit_DrivePool_System_Crashes

 

And do you have any antivirus, backup or disk tools installed?

Or just list the results of this: http://wiki.covecube.com/StableBit_DrivePool_Q2159701

Link to comment
Share on other sites

  • 0

For future can we make sure both types work within the pool?  I have data aware SAN's and tiered data on them that has to stay in the NTFS format to maintain compatibility with the SAN's, but those SAN's are pooled with stand-alone drives that are ReFS.  That's working as it is, but if the pool has to be of homogeneous file type, it would prevent me from operating the pool as it is, and I could never move to ReFS in the foreseeable future...

 

Alex has added an advanced setting to manually enable mixed pools. So this is a moot point now. 

 

Specifically, it's "CoveFs_AllowMixedFilesystems", it's defaulted to "false" (disabled), but can be enabled via:

http://wiki.covecube.com/StableBit_DrivePool_2.x_Advanced_Settings

Link to comment
Share on other sites

  • 0

Well, it shouldn't be BSODing!  

 

Likely, that is a bug somewhere.  But it could be because of the mismatched file system. 

 

 

What OS are you on, and what version of StableBit DrivePool are you using?

And could you do this: 

http://wiki.covecube.com/StableBit_DrivePool_System_Crashes

 

And do you have any antivirus, backup or disk tools installed?

Or just list the results of this: http://wiki.covecube.com/StableBit_DrivePool_Q2159701

 

MEMORY.DMP - uploading

 

C:\Users\Administrator>fltmc

 

Filter Name                     Num Instances    Altitude    Frame

------------------------------  -------------  ------------  -----

DfsDriver                               1       405000         0

msnfsflt                                0       364000         0

Cbafilt                                 3       261150         0

DfsrRo                                  0       261100         0

Datascrn                                0       261000         0

Dedup                                   5       180450         0

luafv                                   1       135000         0

Quota                                   0       125000         0

npsvctrig                               1        46000         0

 

Windows 2012 R2 Essentials upgraded to Standard (build 9600)

 

StableBit DrivePool version 2.2.0.692 BETA

 

BSOD Message: KMODE_EXCEPTION_NOT_HANDLED (covefs.sys)

Link to comment
Share on other sites

  • 0

Likely, that is a bug somewhere.  But it could be because of the mismatched file system.

My DrivePool isn't comprised of a mismatched file system. All the drives were formatted ReFS then joined to the DrivePool. I only mentioned that the resulting DrivePool seemed to show up as a Sudo NTFS volume. I'm not sure the fix you proposed would work in my case?

 

post-2630-0-79566300-1474952438_thumb.jpg

 

 

 

 

 

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Answer this question...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...