I just noticed that another of my drives went into "unmountable" state. I just ran xfs_repair -nv and it shows multiple bad magic in block N messages. Last time it happened I lost over 600GB of data so I don't want it to happen again.

 

How should I proceed to avoid data loss?

orion2-diagnostics-20220111-0307.zip

Basically the only option you have is to run it again without -n, you should also have backups of anything important, Unraid or any RAID solution is not a backup.

3 hours ago, JorgeB said:

Basically the only option you have is to run it again without -n, you should also have backups of anything important, Unraid or any RAID solution is not a backup.

 

I did that and it fixed it for like 10 minutes and then it unmounted again. I repeated the steps, fixed it but after another 30 minutes it happened again. Same disk every time.

If the disk is healthy I would suggest re-creating that filesystem, after backing it up.

If I replace current drive with an empty one, rebuild it from parity and then recreate fs - should it work?

1 hour ago, paululibro said:

If I replace current drive with an empty one, rebuild it from parity and then recreate fs - should it work?

Rebuilding to a replacement might help if there are actual problems with the disk causing this, in that case no point in recreating fs.

 

1 hour ago, JorgeB said:

re-creating that filesystem, after backing it up

recreating it means format, so you have to back it up somewhere first.

I'm rebuilding to a new disk. Currently at 18% but it already shows "Unmountable: not mounted".

 

6 minutes ago, trurl said:

recreating it means format, so you have to back it up somewhere first.

 

All data should still be backed up on the original drive. If recreating will be necessary I can copy from there.

Just now, paululibro said:

I'm rebuilding to a new disk. Currently at 18% but it already shows "Unmountable: not mounted".

You should try xfs repair on the rebuilt disk.

 

Also, new diagnostics might tell us if the rebuild is going well.

Here are new diagnostics. Now I wait for rebuilding to complete

orion2-diagnostics-20220111-1532.zip

looks OK so far

Rebuilding finished with 0 errors so I ran xfs_repair with no modify flag and got thousands of these lines:

 

out-of-order bno btree record 344 (108441151 18) block 0/651941
block (0,108917856-108917947) multiply claimed by bno space tree, state - 1
data fork in ino 15034795729 claims free block 37702356
free space (7,69487107-69487108) only seen by one free space btree

 

And thousands of files and directories marked to be junked:

 

entry "Season 1" in shortform directory 106340182 references non-existent inode 2183048375
would have junked entry "Season 1" in directory inode 106340182

 

Then, I plugged in original disk, mounted it with Unassigned Devices and compared tree command from today and yesterday - all data is still on it.

 

Now:

1. unplug original drive

2. go to maintenance mode

3. click on new drive

4. re-create fs (the only way I see is to set it something else like btrfs and then again to xfs)

5. connect and mount original drive

6. copy all files from orignal drive to the new one

 

Is that correct?

Looks good.

It took almost 2 days to copy everything but all files are now on a new disk and everything seems to be working fine. It copied ~8TB without any I/O errors and I was getting them quiet often before. Thanks for help