Disaster recovery fails

Yesterday I updated the snap, so LXD went from 5.11 to 5.14, but I lost access to the ZFS filesystem. I’m runnig on Debian 10/Buster, ZFS 0.7, I found that 5.11 was the latest version for ZFS 0.7, so I removed 5.14 and installed the package from cache with “snap install lxd_24483.snap”.
So when I run “lxd recover” I get this error message:

Error: Failed validation request: Failed checking volumes on pool “default”: Failed to run: zfs set mountpoint=legacy canmount=noauto default/custom/default_emails: exit status 1 (cannot set property for ‘default/custom/default_emails’: ‘mountpoint’ does not apply to datasets of this type)

Sometimes it shows “backup.yaml” not found for “default/custom” or “default/virtual-machines”…

Can someone help me please?


Junior Polegato

Hrm I’ve not got an entirely clear picture of what has been going on here.

So upgrading from LXD 5.11 to LXD 5.14 would have involved several DB schema changes which would have prevented starting LXD when you downgraded again. Unless as part of the downgrade, you removed LXD’s database.

As LXD requires a ZFS kernel module >= 0.8 the preferred solution for Debian 10 is to switch to the newer ZFS kernel modules, see LXD 5.12: ZFS stopped working in lxd - Error: Required tool 'zpool' is missing when kernel ZFS module version < 0.8 - #16 by tomp

Rather than not upgrading, as that will mean you cannot get security and bug fixes.

Anyway, what is done is done, so going back to your recovery question:

I’m going to assume you totally removed LXD and that is what allowed you to go back to LXD 5.11 albeit it without a database containing your ZFS data.

Please can you show me the full transcript of your lxd recover session as well as the output of sudo zfs list -t all