So a few days ago I was having an issue during a parity sync with a drive that was failing (see my post about it here). While trying to get the data off that drive, and replace it, the parity disk dropped out again. So I am now left with a disabled parity drive and an unmountable data disk. My limited knowledge tells me to just bite the bullet and set up a new config, leaving out the dead drive and parity disk, restart the array, resync parity, and be on my way. When I stop the array to do so, I get a message about too many wrong or missing disks, which seems correct (a dead drive and a wonky parity that's disabled).

 

In preparation for this I shut down the server, reseated the SATA and power cables, rebooted and started a memtest going overnight (no errors found). This morning when the server rebooted I noticed an odd message during the boot sequence, which I had been seeing occasionally over the past little while.

Tower login: XFS (md2): metadata I/O error: block 0xaed3fbe7 ("xlog_bread_noalign") error 5 numblks 1

So my questions are:

 

1. Does this message just indicate that UnRAID is unhappy with the unmountable, dead drive?

 

2. Would it be worth it to jump into maintenance mode and play around with the unmountable drive, with the hopes of making it mountable and pulling some more data off,  before going the new config route?

 

3. Being new to 'new config' is it really as simple as reassigning all the remaining data disks, leaving off the dead disk and parity disk and restarting the array?

 

4. Anything else I am missing here?

 

Thank you for your patience, if you're reading this far,  my knowledge is limited and I really appreciate any help.

1).  This might be why the disk is being shown as unmountable

 

2).  Definitely worth trying.    You want to run a file system check, and assuming a problem is reported then rerun it with the option to only check removed so repairs are attempted.  If a drive has not physically failed this normally means the drive becomes mountable again.

 

3).  With New Config it is easiest to select the option to retain all current settings, and then go back to the Main screen to remove any drives you no longer wish to keep assigned before starting the array.   Be aware that this process removes the possibility of using Parity to recover a disk, but I think in you case this is not a requirement?

Since that disk in known to have pending sectors xfs_repair will most likely fail, best chance of recovering anything is to clone that disk to a new disk, then run xfs_repair.

39 minutes ago, itimpi said:

1).  This might be why the disk is being shown as unmountable

 

Thats what I was thinking, but I am not fluent in log reading.

 

Quote

2).  Definitely worth trying.    You want to run a file system check, and assuming a problem is reported then rerun it with the option to only check removed so repairs are attempted.  If a drive has not physically failed this normally means the drive becomes mountable again.

 

I assume that the option to only checked removed is on the disk page in UnRAID when in maintenance mode, as I don't see mention of that option on the fsck manpage. note: I am working up the patience to stop that array and want all ducks in a row before I do so, last time I did it I could not start it again due to too many wrong/missing disks error and had to reboot.

 

Quote

3).  With New Config it is easiest to select the option to retain all current settings, and then go back to the Main screen to remove any drives you no longer wish to keep assigned before starting the array.   Be aware that this process removes the possibility of using Parity to recover a disk, but I think in you case this is not a requirement?

 

That's correct, at this point I just need to get the remaining disks up and running so I can resync parity. I was hoping to mount the failed disk out of the array and try to pull whatever info I could off it, but it seems like that may not happen.

Edited September 7, 20178 yr by ratmice

23 minutes ago, johnnie.black said:

Since that disk in known to have pending sectors xfs_repair will most likely fail, best chance of recovering anything is to clone that disk to a new disk, then run xfs_repair.

 

How would one go about cloning an unmountable disk? I'm feeling pretty ignorant, right about now.

18 minutes ago, ratmice said:

 

How would one go about cloning an unmountable disk? I'm feeling pretty ignorant, right about now.

 

You can use dd:

 

dd if=/dev/sdX of=/dev/sdY bs=4k conv=noerror,sync status=progress

sdX = source

sdY = destination

Edited September 7, 20178 yr by johnnie.black

So, I stopped the array, checked maintenance mode, pressed start and array unable to start in maintenance mode due to too many wrong/missing disks. AS I see it the options are set up a new config either just removing the parity disk, or removing both parity and the dead disk. If I take the dead disk out of the new config, will I still be able to access it outside the array to run fsck and/or clone it? I'm in over my head. LOL.

Since parity is invalid and the disk is disable it's normal for it to be unmountable, parity should still be mostly correct, for the best chance of recovering your data you can try two options:

 

1) Tools -> New Config -> Retain current configuration: All -> Apply
-assign any missing disk(s) if needed
-check both "parity is already valid" and "maintenance mode" before starting the array
-start the array
-stop array, unassign disk2
-start array, check emulated disk mounts and contents look correct

 

If it mounts you can rebuild to a new disk or copy data from the emulated disk to another disk or PC, if it doesn't mount you can run xfs_repair on the emulated disk to see if it's fixable, this will only work if parity doesn't drop again

 

2) clone bad disk with dd (array can be stopped)

-try to mount it on a different pc or change the UUID to mount on the same server.

-if it doesn't mount run xfs_repair on it.

-copy everything you can.

Between both options you should recover most of the data.

Thanks to all that replied on this thread. I am now up and running, albeit having lost some data, but all in all not too bad