So, I screwed up.

I've been moving my server from an ITX case to a Poweredge R730xd. I was initially having trouble with drives cutting out, and one of my data drives wound up with a corrupted, unmountable file system.

I tried several times to repair it, but it would always fail at the end. After a day or two of trying to get it mountable (the drive itself passed diagnostics, the XFS filesystem was just completely hosed), I figured that since the emulated drive was working, I'd boot into a different Linux, wipe that drive, and let Unraid treat it like a new replacement and rebuild.

My mistake was not getting the serials of the drives first. On the five drives, I saw four XFS volumes and one unmountable unknown format. The four XFS drives seemed to pass a check, so I figured the bad drive was the unknown, and deleted the partition. Rebooted to Unraid and discovered it was the parity drive, and my array was now totally shot.

I haven't mounted any of the drives since. I'm currently running TestDisk on the parity drive. It's found deleted FAT32, HPFS - NTFS, and FAT16 partition info so far, but it's still got 25% or so to go on the quick scan.

Is there any hope it'll find the correct partition and let me restore it? If so, what format is the partition in? I assume the filesystem is something not normally readable, and as I recall gparted said it was a Linux partition.

Failing that, is there another way to recreate the partition since the data should still be intact? I didn't format the drive, only remove the partition.

45 minutes ago, TVsIan said:

since the emulated drive was working

I thought you said it was unmountable. Rebuilding from parity can't fix unmountable.

Parity has no filesystem and it has no data, just parity bits.

Assign the other disks as data disks, leave parity out for now, start the array, then

Attach Diagnostics ZIP to your NEXT post in this thread.

Mounting the drives outside the array invalidated parity anyway.

13 minutes ago, trurl said:

I thought you said it was unmountable.

And you didn't actually mention that it was disabled or missing, which it would have to be if it was being emulated. How did it become emulated?

14 minutes ago, trurl said:

I thought you said it was unmountable. Rebuilding from parity can't fix unmountable.

Parity has no filesystem and it has no data, just parity bits.

Assign the other disks as data disks, leave parity out for now, start the array, then

Attach Diagnostics ZIP to your NEXT post in this thread.

The actual drive was unmountable, but the emulated drive was running in its place. If I started the array in normal mode, my data was available, but that drive had a red X and on the array function screen it said it was unmountable. I tried an xfs repair both from the UI (while in maintenance mode) and the command line, but it kept failing.

I knew rebuilding wouldn't fix that, which is why I didn't do a rebuild, and was trying to reset that drive so I could replace it with itself, basically.

I didn't actually mount the other drives, though if I can get the bad data drive repaired and mounted, then I can rebuild the parity, hopefully.

2 minutes ago, TVsIan said:

The actual drive was unmountable, but the emulated drive was running in its place. If I started the array in normal mode, my data was available, but that drive had a red X and on the array function screen it said it was unmountable. I tried an xfs repair both from the UI (while in maintenance mode) and the command line, but it kept failing.

If Array Devices said the drive was disabled and unmountable, then you couldn't access its data because it wasn't able to mount the emulated drive. Data from other drives would be available since each disk in the Unraid array is an independent filesystem.

What exactly do you mean by "the actual drive"? Did you try to mount it outside the array with Unassigned Devices?

You should post the diagnostics to see the current array status

10 hours ago, trurl said:

If Array Devices said the drive was disabled and unmountable, then you couldn't access its data because it wasn't able to mount the emulated drive. Data from other drives would be available since each disk in the Unraid array is an independent filesystem.

What exactly do you mean by "the actual drive"? Did you try to mount it outside the array with Unassigned Devices?

I did not attempt to mount the drive anywhere else. Before I saw that it was unmountable, I started the array in normal mode intending to rebuild. The rebuild started automatically, but there was a warning that Disk 2 was unmountable, and on the array status that drive had a red X rather than the green dot. Hovering over the drive said that the contents were emulated, and my system appeared to be functioning normally otherwise, my docker containers were running without problems.

At that point, I cancelled the rebuild since I knew it being unmountable would not let it complete successfully. I stopped the array and restarted it in maintenance mode. The SMART diagnostics for Disk 2 were fine, so I ran the XFS file check. It attempted to fix it but failed. I tried again from the command line and got the same failure. In maintenance mode, there was no error message about the drive being unmountable, but it came back when I switched back to normal mode.

My initial thinking at that point was format the drive and let it rebuild. But it did warn that formatting would lose the data, so I didn't proceed with that.

When that wasn't getting anywhere, I shut down and rebooted from a Linux Mint image. I hoped that if I formatted the drive outside of Unraid, it would be treated like a new disk and I could add it in place of the now-missing Disk 2 to rebuild, since the contents would have been emulated. None of the drives were mounted while I was in Mint. I ran xfs_repair -n on the four drives showing XFS filesystems. None showed any errors that should have made them unmountable, and Gnome's disk manager showed an enabled mount button for all 4 (which I did not hit).

My mistake was thinking the parity drive was also xfs and the one drive showing an unknown file system and no mount button was Disk 2 of my Unraid array, since I didn't see any reason it would be unmountable in Unraid but (theoretically) mountable in Mint.

At this point I'm thinking I'll have to recreate the array with a couple of smaller disks that I have available, copy data via mounting in unassigned devices, then add the now-empty disk to the array and move onto the next. If I'm understanding the way Unraid stores files, I should be able to get at least 3/4 of my data from the 3 good drives and cache SSD, and maybe the rest if the failed drive can be mounted.

For future reference if something like this happens again, if a filesystem is corrupted and unmountable, is there a non-destructive way to reset the disk? ie remove the disk from the array and start it with an emulated disk, format the disk in unassigned devices, then re-add it as if it were a replacement drive?

Also fwiw, I'm still on 7.2.0 because I had an update fail on me a while back and was holding off on trying again. So if it was a glitch that may have been fixed since then, that's also on me.

I'll post the diagnostics when I'm back in front of it and boot it up in Unraid again.

Did you actually format a disk in Mint?

13 hours ago, TVsIan said:

figured the bad drive was the unknown, and deleted the partition

Parity does have a partition. But since it isn't clear yet that any rebuild is necessary or useful, we won't worry about that for now.

31 minutes ago, TVsIan said:

if a filesystem is corrupted and unmountable

Rebuild can't fix unmountable. Check filesystem is the answer.

31 minutes ago, TVsIan said:

recreate the array with a couple of smaller disks that I have available, copy data via mounting in unassigned devices

We can probably get your array going again with the existing disks, maybe even including the problem disk.

We will wait for diagnostics before advising further.

23 minutes ago, trurl said:

Did you actually format a disk in Mint?

Removed the partition in gnome disk management, but did not format and have not mounted the disk since. I did run TestDisk, but it didn't identify anything that looked like the partition, so I did not write any changes.

Just restarted the server in Unraid, diagnostics are attached.

If the array can be restarted and the parity recovered and/or rebuilt, that would be amazing.

longserver-diagnostics-20260317-0955.zip

Post a screenshot from Main and the output from fdisk -l /dev/sdh

root@LongServer:~# fdisk -l /dev/sdh

Disk /dev/sdh: 10.91 TiB, 12000138625024 bytes, 23437770752 sectors

Disk model: ST12000NM001G-2M

Units: sectors of 1 * 512 = 512 bytes

Sector size (logical/physical): 512 bytes / 4096 bytes

I/O size (minimum/optimal): 4096 bytes / 16773120 bytes

Disklabel type: gpt

Disk identifier: 92B2D47B-E56E-4114-9AD2-2B6719D80C4D

Diagnostics the user posted Sept 2023 had those same disk assignments, except current disk4 was parity2

Still some confusion from your first post, but maybe not critical that we understand all the details.

Was disk2 the "emulated disk"?

4 minutes ago, trurl said:

Diagnostics the user posted Sept 2023 had those same disk assignments, except current disk4 was parity2

Yes, I was running out of space recently and went down to one parity drive, with the expectation of moving to a new machine that could hold more drives and going back to 2 at that point.

2 minutes ago, trurl said:

Still some confusion from your first post, but maybe not critical that we understand all the details.

Was disk2 the "emulated disk"?

Correct. It still indicates that if I hover over it on the Main tab, but obviously it's not going to work in this state.

It can't actually emulate a disk when no parity is assigned.

We'll have to wait for @JorgeB to tell us how to fix partition on parity if possible.

Ok, I'll hold off on doing anything else with it until we determine if that repair is doable or if I'll just have to manually salvage what I can.

You can create the partition using

sgdisk -o -a 8 -n 1:32K:0 /dev/sdh

If you rebooted since the last diags, check that the disk is still sdh, it can change.

Then reboot and see if the disk goes back to being assigned, if yes, unassign disk2 and post new diags after array start, if it doesn't, you may need to do a new config, but it needs specific steps to get it to emulate disk2 again, so post here.

37 minutes ago, JorgeB said:

You can create the partition using

sgdisk -o -a 8 -n 1:32K:0 /dev/sdh

If you rebooted since the last diags, check that the disk is still sdh, it can change.

Then reboot and see if the disk goes back to being assigned, if yes, unassign disk2 and post new diags after array start, if it doesn't, you may need to do a new config, but it needs specific steps to get it to emulate disk2 again, so post here.

Ok, updated diagnostics with Disk 2 removed from the array.
Main still says it's unmountable and there's no mnt/disk2 folder, though that may be normal behavior for an emulated disk.

longserver-diagnostics-20260317-1339.zip

Check filesystem on the emulated disk 2.

Running it now. It's finding a LOT of problems that it didn't when I checked it with the disk in the array.

3 minutes ago, TVsIan said:

Running it now. It's finding a LOT of problems that it didn't when I checked it with the disk in the array.

There is no difference between emulating a missing disk and emulating a disabled disk. No way to really know what you did before and what happened.

Possibly something you did changed parity or some other disk so the emulation is producing different results than before.

Emulation doesn't use the physical disk at all, so it doesn't matter if the disk is disabled, missing, or even destroyed. The contents of the disk are calculated from the parity algorithm by reading all other disks.

Parity is basically the same concept wherever it is used and however it is implemented. It is just an extra bit that allows a missing bit to be calculated from all the other bits. Understanding parity makes a lot of this make more sense. Parity is really very simple. If you are interested go to this link:

https://docs.unraid.net/unraid-os/getting-started/what-is-unraid/

then expand the Sofware-defined NAS section, and inside that, expand the How Parity Works section.

14 minutes ago, TVsIan said:

finding a LOT of problems

Post the output, or at least the last several lines, when it has finished.