Greetings:

I recently moved all my LXD containers to a new host. The old host was running Debian 11 and was using LXD with a btrfs storage pool (the host was also using btrfs as the underlying filesystem). The new host is running Debian 11 also using btrfs as the underlying file system and the new LXD installation using a btrfs storage pool for storage.

I transferred the containers via the rather ham-fisted approach of tar.bz’ing the /var/snap/lxd folder and subfolders on the old host. I then deleted the same /var/snap/lxd folder and subfolders on the new host and replaced them with the same from the old host. Everything spun up wonderfully until I noticed that snapshots were not being taken. I then attempted to copy a container and received the following error:

Quote:
Error: Create instance from copy: Failed to run: btrfs subvolume snapshot /var/snap/lxd/common/lxd/storage-pools/dauntless/containers/graylog /var/snap/lxd/common/lxd/storage-pools/dauntless/containers/graylog-broken: ERROR: Not a Btrfs subvolume: Invalid argument
"Dauntless" is the name of the btrfs storage pool:https://omegle.site


I suspect this is because I was rather inelegant in moving containers over to the new host. Any suggestions on how to fix this issue so that I can regain full container functionality (i.e. snapshots, creating copies, and presumably other storage-related features I have not yet discovered to be not working)?

So why am I not posting this in the Containers forum? Well, I am not sure this is a LXD issue but, rather, a btrfs issue. I think there must be a permissions issue (flag issue? I don't know much about btrfs) that is cause the "Not a Btrfs subvolume" error.

Any help is appreciated. Thank you!

You did not migrate the volumes (subvolumes) you migrated the contents. On the target you now have file and folder structures (possibly ON a single BTRFS volume) but do not have them as BTRFS structure.

Not, BTW, a bad thing in itself as long as you do not require BTRFS snapshots to work properly. The same process on EXT4, XFS, or JFS2 might have been just fine.

I have only been really using BTRFS for a year, and have researched some local migration issues. (I get faster performance for my applications using EXT4/RAID-5. BTRFS is catching up, but is not as performant. Yet.) None of them involved cloning subvolums to a different server. It would be easier to clone a discrete BTRFS device, but I am not finding a clear and obvious solution to clone a single or small group of subvolumes.

If someone has solved that, I hope they pipe up, I would like to see how that would work.