----- On Mar 9, 2018, at 4:39 PM, Bernd Lentes bernd.lentes@helmholtz-muenchen.de wrote:

Hi,

i asked already a time ago about snapshots. I have some guests, each resides in a raw file, placed on an ext3 fs which is on top of a logical volume, for each guest a dedicated lv. The raw is not a "must have", if there are obvious reasons i can convert them to a qcow2. What i want is a consistent backup of each guest taken overnight. If possible i won't have downtime. I can use the libvirt tools, but the lvm way seems to be more elegant. Before copying the file the guests resides in i take a snapshot from the respective lv. Then i mount the snapshot and transfer it via rsync on a CIFS share. Rsync seems to be the appropriate tool because i just transfer the changes in the file compared to the file from a day before. So i don't have to transfer complete and big raw files but just the difference.

The guest still can be running, and after the transfer of the snapshot file i just delete it, and the next night the same procedure.

What do you think ?

It does not work as expected :-( My lv's are clustered, but snapshotting a clustered lv requires to activate the source lv exclusively on one node, which is not possible when it's mounted and files on it are open. So i have to try it with libvirt and qemu. I'd like to create the snapshot while running the guest, take the backup, and merge (or commit) the changes after the copy procedure, still with a running guest. Is there a way to do this ? I found https://wiki.libvirt.org/page/Live-disk-backup-with-active-blockcommit , but my software seems to be too old: I have libvirt-1.2.5-23.3.1. But my virsh offers blockcommit: virsh # help blockcommit NAME blockcommit - Start a block commit operation. SYNOPSIS blockcommit <domain> <path> [<bandwidth>] [<base>] [--shallow] [<top>] [--delete] [--wait] [--verbose] [--timeout <number>] [--async] DESCRIPTION Commit changes from a snapshot down to its backing image. OPTIONS [--domain] <string> domain name, id or uuid [--path] <string> fully-qualified path of disk [--bandwidth] <string> bandwidth limit in MiB/s [--base] <string> path of base file to commit into (default bottom of chain) --shallow use backing file of top as base [--top] <string> path of top file to commit from (default top of chain) --delete delete files that were successfully committed --wait wait for job to complete --verbose with --wait, display the progress --timeout <number> with --wait, abort if copy exceeds timeout (in seconds) --async with --wait, don't wait for cancel to finish But it doesn't work ? Although help offers it ? Bernd Helmholtz Zentrum Muenchen Deutsches Forschungszentrum fuer Gesundheit und Umwelt (GmbH) Ingolstaedter Landstr. 1 85764 Neuherberg www.helmholtz-muenchen.de Aufsichtsratsvorsitzende: MinDir'in Baerbel Brumme-Bothe Geschaeftsfuehrer: Prof. Dr. Guenther Wess, Heinrich Bassler, Dr. Alfons Enhsen Registergericht: Amtsgericht Muenchen HRB 6466 USt-IdNr: DE 129521671