Hello,

I run a relatively vanilla unRAID setup.  3x4TB drives with two 1TB SSDs as a ZFS cache pool.  Once ZFS was available, I converted the cache pool from BTRFS to ZFS due to continued corruption issues every few months.  I'm happy to report that it has been perfect so far.  I am seeing a message in the logs that I wanted to ask here for any insights.  Ever time the mover runs, I get the following set of error messages:

Jan 20 00:40:01 vader shfs: /usr/sbin/zfs destroy 'cache/Backups'
Jan 20 00:40:01 vader root: cannot destroy 'cache/Backups': dataset is busy
Jan 20 00:40:01 vader shfs: retval: 1 attempting 'destroy'
Jan 20 00:40:01 vader shfs: /usr/sbin/zfs mount 'cache/Backups'
Jan 20 00:40:41 vader shfs: /usr/sbin/zfs unmount 'cache/Media'
Jan 20 00:40:41 vader shfs: /usr/sbin/zfs destroy 'cache/Media'
Jan 20 00:40:41 vader root: cannot destroy 'cache/Media': dataset is busy
Jan 20 00:40:41 vader shfs: retval: 1 attempting 'destroy'
Jan 20 00:40:41 vader shfs: /usr/sbin/zfs mount 'cache/Media'
Jan 20 00:40:42 vader shfs: /usr/sbin/zfs unmount 'cache/Proxback'
Jan 20 00:40:42 vader shfs: /usr/sbin/zfs destroy 'cache/Proxback'
Jan 20 00:40:42 vader root: cannot destroy 'cache/Proxback': dataset is busy
Jan 20 00:40:42 vader shfs: retval: 1 attempting 'destroy'
Jan 20 00:40:42 vader shfs: /usr/sbin/zfs mount 'cache/Proxback'

I have tried stopping and starting the array, as well as a complete reboot, but these messages persist.  I do have the "Backups" and "Proxback" shares available on my network as NFS shares.  But not "Media".  However, I do have an arr stack and Plex dockers that have the /mnt/user/Media directory referenced (direct on server, not via SMB/CIFS).  Is this an expected error message?

vader-diagnostics-20240120-1320.zip

It is if something is using the datasets when the mover is attempting to destroy them.

On 1/21/2024 at 4:30 AM, JorgeB said:

It is if something is using the datasets when the mover is attempting to destroy them.

I also get this
"shfs: /usr/sbin/zfs unmount 'cache/*' 2>&1

shfs: /usr/sbin/zfs destroy 'cache/*' 2>&1

shfs: cannot destroy 'cache/*': dataset is busy

shfs: /usr/sbin/zfs mount 'cache/*' 2>&1"
spammed in my logs when mover is running! Its bothered me, but I've always believed it was due to people having folders/files open, overnight, whilst the mover is trying to do its job; so its not able to remove the empty folder from the cache pool. I have a large setup with many pools with over 100TB total storage! Its used in a production environment so these pools are filled/emptied every day!

Since this is an expected behaviour, and the shares use the "cache" all the time (so the empty folder will, most likely, just be created again the next day), we need a setting to tune out these empty folder deletes whilst mover is running! I suggest this so that the logs are not flooded with pointless messages, burying possible more important ones (i'm one of those weirdo daily log readers). Is there already a way to set the mover to "if share uses this cache pool, leave root folder in pool"? If not, i think it would be a great addition (and i assume being able to turn it on or off would appease some who want the folder to go away every move operation).

If I'm not missing this setting somewhere, I will post this in feature requests as well!