This is an urgent one for me.

In the server “kjell”, it is a ZFS pool of 10 x Samsung_SSD_870_QVO_8TB, plus special. The media library holds 95% of used space. And when the desbribed issue occurred, about 52T of 64T was occupied.

I was running a transcoding job in Unmanic on this server, and another Unmanic job on the other server (helmaxx) on a different media share on the first server. At the same time I was trying to setup a second GPU on the second server (helmaxx), controlled from Unmanic on the first server (kjell) with help from Unmanic Discord forum. The second GPU Unmanic container started to report receiving jobs, but only one or two workers actually did any work. After a while, first server (kjell) reported 0T available.

Troubleshoot was done with my good!?? help Chat, which reported ZFS “collapse”. Chat recommended to delete files to usage below 80%, and ZFS should be ok again. Several attempts, and I risk loosing the pool altogether. I deleted a lot of files, and see strange 0 files in …../filmer and at the same time 25T usage.

It is a way to solve this? That would be something…..

// frodr



root@kjell:~# zfs get -Hp used lageret/media/filmer

lageret/media/filmer used 25095390259344 -

root@kjell:~# zfs get -Hp logicalused lageret/media/filmer

lageret/media/filmer logicalused 25215583430656 -

root@kjell:~# zfs get -Hp used lageret/media/serier

lageret/media/serier used 20942021724592 -

root@kjell:~# zfs get -Hp logicalused lageret/media/serier

lageret/media/serier logicalused 20999065692160 -

root@kjell:~# ls -lh /mnt/lageret/media/filmer

total 0

NAME            SIZE  ALLOC   FREE  CKPOINT  EXPANDSZ   FRAG    CAP  DEDUP    HEALTH  ALTROOT
lageret        74.6T  61.2T  13.4T        -         -     0%    82%  1.00x    ONLINE  -
  raidz1-0     36.4T  30.5T  5.83T        -         -     0%  84.0%      -    ONLINE
    sda1       7.28T      -      -        -         -      -      -      -    ONLINE
    sdb1       7.28T      -      -        -         -      -      -      -    ONLINE
    sdc1       7.28T      -      -        -         -      -      -      -    ONLINE
    sdd1       7.28T      -      -        -         -      -      -      -    ONLINE
    sde1       7.28T      -      -        -         -      -      -      -    ONLINE
  raidz1-1     36.4T  30.6T  5.76T        -         -     0%  84.2%      -    ONLINE
    sdf1       7.28T      -      -        -         -      -      -      -    ONLINE
    sdg1       7.28T      -      -        -         -      -      -      -    ONLINE
    sdh1       7.28T      -      -        -         -      -      -      -    ONLINE
    sdi1       7.28T      -      -        -         -      -      -      -    ONLINE
    sdj1       7.28T      -      -        -         -      -      -      -    ONLINE
special            -      -      -        -         -      -      -      -         -
  mirror-2     1.86T  34.2G  1.83T        -         -    14%  1.79%      -    ONLINE
    nvme4n1p1  1.86T      -      -        -         -      -      -      -    ONLINE
    nvme5n1p1  1.86T      -      -        -         -      -      -      -    ONLINE

kjell-diagnostics-20251204-1657.zip

I'm not really following the issue. Where are you seeing the 0TB available? Also, have you rebooted since the issue?

It was showing 0T after 1-2 hours in the Unmanic transcoding. After deleting about 30T, it now reads only 10T available. In…./filmer there is 0 files, still:

root@kjell:~# zfs get -Hp used lageret/media/filmer

lageret/media/filmer used 25095390259344 -

Yes, rebooted several times.

Post the output from zfs list -t all and ls -la /mnt/lageret/media/filmer

44 minutes ago, JorgeB said:

Post the output from zfs list -t all and ls -la /mnt/lageret/media/filmer

Outputs.rtf

From what I can see, the database is completely empty, but there are still snapshots. The last one is this:

lageret/media/filmer@2025-11-27-141321 633G - 22.7T

Should have around 23TiB of data.

So …./filmer is empty, but shapshots at about 23TiB?

Would I be able to rollback with only 10TB available?

This one below is correct?

It isn´t a risk of loosing my pool, unless I screw up.

“What´s the point of troubleshooting if you never make any trouble……"

But something strange happened during transcoding since available space read 0T. Any my not so qualified friend said I had a “ZFS-collapse” and should delete files to get below 80%. I guess keeping zpools under 80% usage is a good advice.

12 hours ago, frodr said:

Would I be able to rollback with only 10TB available?

You should because the snapshots are already using the space:

lageret/media/filmer 22.8T 9.10T 120K

That dataset is using 22.8T total (from the snapshots, since there's nothing outside them, only 120K), and there are 9.1T free.

You can also browse the snapshots by going to lageret/media/filmer/.zfs/snapshots/