TrueNAS 13-U1 5.16TB of remaining data hidden in dataset after pool rebuild

[wan2bretired]

Copied data from a 5x(2x4gb) mirrored to 4x(2x12gb) mirrored disk array. 13TB transferred successfully. removed the 4GB drives and replaced with 12TB drives and imported successfully. Data was compared before transfer and was identical. To update my pool structure I wanted to move data from a dataset subfolder to a child data set. For this I used cp -a ( changed subfolder name and created dataset with original subfolder name). I did this with 4 subfolders, and the data was identical using 2 methods with no differences. The subfolders were deleted using Samba connection to a windows pc. After deletion, all pool snapshots were deleted, and then recreated.

The used space for each new dataset is as expected, however the parent dataset is 5.13TB in excess and all is in the USEDDS on a zfs list check. Zfs list snapshot was checked after pool snapshot deletion, and for zpool is was empty. The pool is zpool and the parent dataset is zpool/PhotoLibrary, which has the usedds of 5.17TB. Finally, a scrub and a reboot were performed prior to posting.


The USEDDS size of 5.17 TB matches the size of the reported dataset size of /zpool/PhotoLibrary/Photosync on Syncthing share with backup server, see image. This would suggest a ghost image of the deleted folder that has not been fully released from /zpool/PhotoLibrary were it resided.

Asrock x470du2-2t,ryzen 3600, 124 MB 2666 ecc, HUH 12 TB 520 series x8, mirrored x2x4, SAS 3008 HBA, On board Nics. I am using Truenas core 13.0 U1.

 

[wan2bretired]

Snapshot info:

 

[anodos]

Did you have the Samba recycle bin enabled? Perhaps everything got moved to a hidden .recycle directory.

 

[wan2bretired]

Thanks for your reply. If that is turned on via auxilary parameters in Services/SMB, then no. But in case that happened how would I search for that directory to double check or would it have shown on the zfs list -t filesystem -o space?

 

[anodos]

If that is turned on via auxilary parameters in Services/SMB, then no.

It's a checkbox in the SMB share.

Easiest way to check is to use `du` from shell:

du -h -d 1 /mnt/zpool/PhotoLibrary

 

[wan2bretired]

It's a checkbox in the SMB share.

Easiest way to check is to use `du` from shell:

du -h -d 1 /mnt/zpool/PhotoLibrary

Thank you I tried the du command, and it reports correctly for zpool/PhotoLibrary at 13 TB, but Storage/Pools in GUI reports 18.36 TB. There was not an individual share of zpool/PhotoLibrary, only the pool, zpool, share. The box was not checked.

 

[wan2bretired]

It's a checkbox in the SMB share.

Easiest way to check is to use `du` from shell:

du -h -d 1 /mnt/zpool/PhotoLibrary

The USEDDS has increased from 5.17 TB to 5.40 in zpool/PhotoLibrary but no files, only child datasets

 

[wan2bretired]

It's a checkbox in the SMB share.

Easiest way to check is to use `du` from shell:

du -h -d 1 /mnt/zpool/PhotoLibrary

Hi @anodos see other submissions, any suggestions? Would creating a new zpool/dataset and zfs rename all current child datasets of zpool/PhotoLibrary to that dataset ( zpool/new) and then delete zpool/PhotoLibrary and all snapshots and then recreate dataset zpool/PhotoLibrary and the zfs rename the content of zpool/new to zpool/PhotoLibrary work?

 

[wan2bretired]

Can anyone help? @anodos

 

[wan2bretired]

Well not too much insight from people smarter than me, so I did what I proposed above and it solved it! See images. The extra 5+TB is no longer on PhotoLibrary after the rename of each child data sets to a surrogate dataset, deleting the original parent dataset and then renaming them back into the recreated Parent dataset.

 

[wan2bretired]

[mod note: misposted a thread reply as a moderation report - hacky move back into thread, crossing my fingers -JG]

Well not too many responses, so I did what I proposed above and it solved it! See images. The extra 5+TB is no longer on PhotoLibrary after the rename of each child data sets to a surrogate dataset, deleting the original parent dataset and then renaming them back into the recreated Parent dataset.