Dataset cannot be destroyed with error "dataset already exists"

[SimonR]

Hi all,

after a bad sync with zfs-autobackup I cannot delete 2 ZFS-datasets on my server.

There are no snapshots or clones existing for these datasets, but it still shows USEDCHILD-space, and I'm wondering why.



I'm able to rename the datasets. I can snapshot these 2 datasets too, and I can delete the snapshots. But I'm not able to delete the datasets itself.



The properties of the dataset are:

root@pvezfsham:~# zfs get all rpool/crypt/vm-100-disk-0-alt
NAME PROPERTY VALUE SOURCE
rpool/crypt/vm-100-disk-0-alt type volume -
rpool/crypt/vm-100-disk-0-alt creation Tue Nov 8 12:09 2022 -
rpool/crypt/vm-100-disk-0-alt used 5.03G -
rpool/crypt/vm-100-disk-0-alt available 14.1T -
rpool/crypt/vm-100-disk-0-alt referenced 4.62G -
rpool/crypt/vm-100-disk-0-alt compressratio 1.54x -
rpool/crypt/vm-100-disk-0-alt reservation none default
rpool/crypt/vm-100-disk-0-alt volsize 200G local
rpool/crypt/vm-100-disk-0-alt volblocksize 8K default
rpool/crypt/vm-100-disk-0-alt checksum on default
rpool/crypt/vm-100-disk-0-alt compression gzip-9 inherited from rpool
rpool/crypt/vm-100-disk-0-alt readonly off default
rpool/crypt/vm-100-disk-0-alt createtxg 83488 -
rpool/crypt/vm-100-disk-0-alt copies 1 default
rpool/crypt/vm-100-disk-0-alt refreservation none default
rpool/crypt/vm-100-disk-0-alt guid 9378425283866405324 -
rpool/crypt/vm-100-disk-0-alt primarycache all default
rpool/crypt/vm-100-disk-0-alt secondarycache all default
rpool/crypt/vm-100-disk-0-alt usedbysnapshots 0B -
rpool/crypt/vm-100-disk-0-alt usedbydataset 4.62G -
rpool/crypt/vm-100-disk-0-alt usedbychildren 424M -
rpool/crypt/vm-100-disk-0-alt usedbyrefreservation 0B -
rpool/crypt/vm-100-disk-0-alt logbias latency default
rpool/crypt/vm-100-disk-0-alt objsetid 83735 -
rpool/crypt/vm-100-disk-0-alt dedup off default
rpool/crypt/vm-100-disk-0-alt mlslabel none default
rpool/crypt/vm-100-disk-0-alt sync standard inherited from rpool
rpool/crypt/vm-100-disk-0-alt refcompressratio 1.50x -
rpool/crypt/vm-100-disk-0-alt written 4.62G -
rpool/crypt/vm-100-disk-0-alt logicalused 5.32G -
rpool/crypt/vm-100-disk-0-alt logicalreferenced 4.77G -
rpool/crypt/vm-100-disk-0-alt volmode default default
rpool/crypt/vm-100-disk-0-alt snapshot_limit none default
rpool/crypt/vm-100-disk-0-alt snapshot_count none default
rpool/crypt/vm-100-disk-0-alt snapdev hidden default
rpool/crypt/vm-100-disk-0-alt context none default
rpool/crypt/vm-100-disk-0-alt fscontext none default
rpool/crypt/vm-100-disk-0-alt defcontext none default
rpool/crypt/vm-100-disk-0-alt rootcontext none default
rpool/crypt/vm-100-disk-0-alt redundant_metadata all default
rpool/crypt/vm-100-disk-0-alt encryption aes-256-gcm -
rpool/crypt/vm-100-disk-0-alt keylocation none default
rpool/crypt/vm-100-disk-0-alt keyformat passphrase -
rpool/crypt/vm-100-disk-0-alt pbkdf2iters 350000 -
rpool/crypt/vm-100-disk-0-alt encryptionroot rpool/crypt -
rpool/crypt/vm-100-disk-0-alt keystatus available -
rpool/crypt/vm-100-disk-0-alt autobackup:dusber false local
rpool/crypt/vm-100-disk-0-alt autobackup:dusham false local



Has anyone a hint, what I can do to destroy these datasets?

I think the problem is the space usedbychildren, but there are no children of these datasets.

The datasets are not mounted anywhere else, they exist in an encrypted zfs. There are no vm's on the the proxmox server configured.

Kind regards

 

[Richard]

I think the problem is the space usedbychildren, but there are no children of these datasets.

Probably yes, try to add -r option:

zfs destroy -r rpool/crypt/vm-100-disk-0-alt

 

[SimonR]

Probably yes, try to add -r option:

zfs destroy -r rpool/crypt/vm-100-disk-0-alt

No chance, it's a ZFS Bug, really, they will implement and release the fix soon, I think. But the zfs vol will stay there without any chance of deleting it. It will live forever now with childdata without child.

https://github.com/openzfs/zfs/pull/14119#issuecomment-1331985760