At least a few dozen people have had issues with BTRFS corruption on their cache driver under 6.12.  I solved my problem by reformatting the cache drive to ZFS.

On 8/22/2023 at 9:51 AM, JorgeB said:

Type N here then type

write

and hit enter, and post output of

btrfs fi show

 

YOU SAVED MY LIFE! our many hair from falling out or getting gray Thank you, this worked for me!

 

On 11/2/2023 at 2:40 AM, RocketSLC said:

Just wanted to post saying that this saved me countless hours of backup recovery. Thank you so much @JorgeB! I was about to wipe and start fresh.

 

Thanks from me also
... this saved me:
 

On 8/21/2023 at 8:24 PM, JorgeB said:

There are no partitions on both devices, there's one thing we can try that usually works, as long as the devices were not fully trimmed, type 

sfdisk /dev/sdb

then type

2048

and hit enter, post a screenshot of the results.

On 8/22/2023 at 9:51 AM, JorgeB said:

Type N here then type

write

In my case a had to execute the sfdisk command on both of my former NVME cache pool disks, so it was
sfdisk /dev/nvme0n1 and
sfdisk /dev/nvme1n1
with the rest being the same: type '2048', found a BTRFS partition signature, choose 'N' to keep it, and 'write'


I tried to convert the RAID1 cache pool into something where I would be able to remove a drive, so did balance > convert to single, stopped the array, removed one disk from the two-disk pool and started with "Yes I want to do this".... et voila: pooof... both drives were unmountable.

I basically misinterpreted "single" as something like the regular array of unrelated disks, where it would be possible to remove a disk like it would be possible with the array. But as I have now learned, "single" means JBOD, and a pool stays a pool, where every device needs to be kept, at least until some extra footwork in the console is also accomplished.

PS: Maybe it would be a good idea to put something more alarming here, instead of the "Yes, I want do do this" checkbox, when fiddling with changes within pools – since the concept (and consequences) are very different from what one may be familiar with, when doing things like removing drives within the regular array...

On 8/23/2023 at 5:00 AM, JorgeB said:

Lets see if we can recover the other device as well, type:

sfdisk /dev/nvme0n1

then

2048

enter, if there's a btrfs signature type N to keep it and then

write

finally

btrfs fi show

once more

 

Sorry for necro'ing this post, but i just wanted to say thank you because this saved my cache drive after unraid froze on me and i had to restart it, and then i got the unmountable/no btrfs stuff.

Same here,  Unraid froze after restart no cache server-rjpe-diagnostics-20240808-0915.zip

 

Pls Help, sfdisk /dev/nvme0n1 not work

 

Thx

2 hours ago, Router1982 said:

sfdisk /dev/nvme0n1 not work

That was not the correct command to use for the issue you were having, it was a just log tree problem:

 

Failed to recover log tree

 

Not sure what you did, but looks like no btrfs filesystem is being detected now, post the output from

 

btrfs fi show

 

 

P.S. zfs is detecting data corruption on multiple disks, this is usually a RAM problem, recommend memtest
 

disk2 - Aug  8 08:21:39 Server-RJPE emhttpd: errors: 1 data errors, use '-v' for a list

disk10 - Aug  8 08:22:17 Server-RJPE emhttpd: errors: 9 data errors, use '-v' for a list

 

root@Server-RJPE:~# btrfs fi show
Label: none  uuid: ca894379-ad55-4a40-bcaa-1e5ee33d3882
        Total devices 1 FS bytes used 3.18MiB
        devid    1 size 1.00GiB used 126.38MiB path /dev/loop2

root@Server-RJPE:~# 

Looks like you wiped the cache device, post the output of

 

fdisk -l /dev/nvme0n1

and

blkid /dev/nvme0n1

 

root@Server-RJPE:~# fdisk -l /dev/nvme0n1
Disk /dev/nvme0n1: 931.51 GiB, 1000204886016 bytes, 1953525168 sectors
Disk model: WD Red SN700 1000GB                     
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0x94557592

Device         Boot Start        End    Sectors   Size Id Type
/dev/nvme0n1p1       2048 1953525167 1953523120 931.5G 83 Linux

 

 

root@Server-RJPE:~# blkid /dev/nvme0n1
/dev/nvme0n1: PTUUID="94557592" PTTYPE="dos"

 

I assume you created that partition? If yes was there a signature present?

I Run 

:

sfdisk /dev/nvme0n1

then

2048

enter, if there's a btrfs signature type N to keep it and then

write

finally

btrfs fi show

 

 

Not Working

 

Then

 

Lets see if we can recover the other device as well, type:

sfdisk /dev/nvme0n1

then

2048

enter, if there's a btrfs signature type Y to keep it and then 

write

finally

btrfs fi show

 

Y was here the Fault....

 

Parrtion not sure  of Pressent

4 hours ago, Router1982 said:

Y was here the Fault....

Yep, no chance to recover now, and like mentioned that wasn't even your problem, it should not follow advice for other users, symptom may be similar but not the same issue, only option is trying a file recovery app, like UFS explorer.

Many Thanks, I will Format the drive and restore from an old Backup. which File System ist recommend BTRFS or ZFS

Edited August 9, 20241 yr by Router1982

1 hour ago, Router1982 said:

which File System ist recommend BTRSF or ZFS

Both are good for single device or mirrors, but between the two I usually recommend zfs.

On 1/2/2024 at 7:07 PM, JorgeB said:

Fist thing you should run memtest since there's a lot of data corruption detected on both pool devices, then type:

 

btrfs rescue zero-log /dev/nvme1n1p1

 

And re-start the array

Thank you @JorgeB! That saved me as well. Cache pool was literaly 2 weeks old when this error showed. Will make backups and reformat to zfs.

Just wanted to report, that today I ran into the same issue, that after a reboot, the cache devices were marked unmountable.

I followed these steps, and got it back working. I did not have to issue the "btrfs rescue zero-log".

 

On 8/23/2023 at 12:00 PM, JorgeB said:

Lets see if we can recover the other device as well, type:

sfdisk /dev/nvme0n1

then

2048

enter, if there's a btrfs signature type N to keep it and then

write

finally

btrfs fi show

once more

 

On 8/23/2023 at 6:34 PM, JorgeB said:

Now stop array, unassign the assigned pool device, start array without any device assigned to the pool so that the pool config is reset, stop array, re-assign both pool members, start array to import the pool, post new diags if it doesn't mount.

 

Although I had a backup of the files on the cache, these explanations saved me a lot of work. This is why I want to say THANK YOU.

I also want to thank @josephsiu, who was patient enough to follow up until the problem was solved.