[libvirt-users] Live Disk Backup

[Prof. Dr. Michael Schefczyk]

Dear All,

using CentOS 7.2.1511, and libvirt from ovirt repositories (currently 1.2.17-13.el7_2.5, but without otherwise using ovirt) I am regularly backing up my VMs which are on qcow2 files. In general, I am trying to follow http://wiki.libvirt.org/page/Live-disk-backup-with-active-blockcommit

A typical backup script would be 

#!/bin/bash
dt=`date +%y%m%d`
if virsh dominfo dockers10a | grep -q -E '^Status: *laufend|^State: *running'
then
virsh snapshot-create-as --domain dockers10a dockers10a --diskspec vda,file=/home/dockers10asnap.qcow2 --disk-only --no-metadata --atomic
cp /kvm01/dockers10a.qcow2 /backup/dockers10a$dt.qcow2
virsh blockcommit dockers10a vda --active --verbose --pivot
virsh snapshot-delete dockers10a dockers10a
rm /home/dockers10asnap.qcow2
fi

I am fully aware that the third line from the end "virsh snapshot-delete ..." will fail under regular circumstances. It is just there as a precaution to delete unnecessary snapshots should a previous backup have failed.

For some time, I am noticing that from time to time backup fails in the way that the xml definition file of the VM backed up keeps the temporary file (in the example above /home/dockers10asnap.qcow2) as the source file. Then, at least upon rebooting the host, it will be unable to restart the VM. In addition, lots of other troubles can arise (following backups failing, storage issues).

I am using a similar setup on four hosts. It seems that the better the resources of the host are, the lower the likelihood of the problem occurring - but that cannot be an acceptable state.

Can someone please point me to how to avoid this?

Regards,

Michael Schefczyk