lxc copy --refresh in ZFS deletes CT target volume after network interruption

[alepaes1975]

Required information

• Distribution: Ubuntu
• Distribution version: 22.04
  • LXD version: 5.14
  • Storage backend in use: ZFS

Issue description

I use copy --refresh to backup containers daily. LXD 5.14 in both servers. I have autosnapshots in source cts with a 7 day retention.

After a network failure during a lxc copy --refresh, my target instance was deleted on ZFS. Repeating the command:

lxc copy cb2:zammad zammad-backup --refresh Error: Failed instance creation: Error transferring instance data: Failed migration on target: Failed creating instance on target: Volume exists in database but not on storage

The only way to correct is removing this target machine and copy all again.

Steps to reproduce

1. lxc copy sourceserver:ct ct-backup --refresh
2. disable network to cause a transfer interrupt
3. An error occurs and taget volume was deleted, but machine exists in db

[tomponline]

https://github.com/lxc/lxd/pull/11823#issuecomment-1604646813

[Doublefull]

Have same issue on Ubuntu 22.04 and LXD 6.1 with ZFS storage:

1. Execute lxc copy --refresh ct target-host:ct on source server
2. Simulate network interrupt: disable network,reboot target server, etc
3. Previously successful backup from target server deleted. Container is present in output of lxc list but volume has gone. On source server following error:
   

Error: Failed instance creation: Error transferring instance data: Failed migration on target: Failed creating instance on target: Volume exists in database but not on storage