I have a number of VMs with raw disk files on an NFS datastore. I created drbd mirror to other host in proxmox cluster. Put LVM on top of that with shared set to true. This way, as I understand it, each vm disk is an LVM. Anyway, all but two restored just fine (e.g. I did backup from NFS raw disk => NFS backup, then restore => LVM/drbd). Two of the restores failed at the very start with an assertion error because the raw disk images were apparently not a multiple of some expected size. I don't remember the exact error (I was unable to get a screenshot at the time.) I have no idea how the files got grown that way - possibly from when I migrated them from xen? Anyway, my 'fix' was to use qemu-img to grow each disk to the next GB multiple, repeat the backup/restore steps and all was well. Definitely scary, and seems like a bug to me - even if the rawfile is not a multiple of X bytes, you can certainly round it up before writing to the LVM?