First, a little background: I have a "White Lights" Mybook World Edition. I mostly use it for archiving backup drive images from other systems on the network via cifs mounts. As a result, I'm storing pretty large files on it - anywhere from 4-50 GB per image file. These are bzip2 and gzip compressed raw drive images, produced with my own script which uses dd and bzip2 or gzip to produce the image, and md5sum to perform integrity checking of the compressed image against the original device. The compression and md5 hashing runs on the client hosts, not the MBWE NAS device.
During my most recent round of backups, the compression and transfer (via cifs) seemed to complete successfully, but integrity checking failed during decompression of the compressed image. It seems the bzip2 compressed data on the NAS was corrupt. To troubleshoot, I reverted to using gzip instead of bzip2, but had similar results: gunzip would fail with a "format violated" error. I spent a couple of evenings troubleshooting this on the client side before realising the problem lies on the NAS. My first indication that this was the case was an email notification of missing default shares ("Shared Music," etc). After rebooting, the missing shares reappeared. After a couple more large file transfers, the problem occurred again. dmesg reported some XFS corruption (Sorry, I don't have the log handy). My problem seems to correspond to some known issues with XFS in some of the the 2.6.xx kernels (Again, sorry to be vague here; I don't have copies of the logs, and am having trouble finding the original forum posts that helped me determine this. I can dig this information up if need be, but it may not be relevant to my specific question - see below).
But here is my specific question: I have found a couple of sets of detailed instructions for checking and repairing XFS filesystems. Here is one of the better ones, which I am trying to follow:
(I can't post a link since I'm a low-karma user, so I've removed the http protocol from the URL) mybookworld.wikidot.com/forum/t-284602/mbwe-2t-does-not-mount-directories
So I try to follow the directions in the "Prerequisites" section: Stop unessential processes, umount the directories in /Shares/, and so on. Everything goes fine until I try to "umount /DataVolume", which fails with the following error:
[code]umount: Couldn't umount /DataVolume: Invalid argument[/code]
I also tried "umount /dev/md2", and received the same error. Getting a bit desperate, I tried to remount the filesystem read-only with "mount -t auto /dev/md2 -o ro /DataVolume", and received the following error:
[code]mount: Mounting /dev/md2 on /DataVolume failed: Device or resource busy[/code]
I then tried "umount -f /DataVolume" as a last resort, which also failed.
I'd love to take a stab at repairing the filesystem using the native xfs_* tools, but I can't even begin unless I can unmount /DataVolume. I've found posts from other people who seemed to be having similar issues with XFS corruption who had good results with xfs_repair, but none of them seemed to have any issue umounting /DataVolume. Am I making a stupid error somewhere? Or perhaps these umount failures are somehow caused by the XFS corruption? Does anyone have any ideas? I'd really like to avoid recreating the volume from scratch, since I'd have to back up almost a terabyte of data and I don't really have a good place to put it at the moment. I'm in the process of building my own RAID server to be a more robust replacement for the MBWE for backups, but it's nowhere near complete and I don't have the drives yet, so I would have to do some serious data juggling if I can't repair this volume.
Any help or suggestions are appreciated.