[libvirt] Overview of libvirt incremental backup API, part 2 (incremental/differential pull mode)

Michael Ablassmeier abi at grinser.de
Thu Nov 22 09:07:08 UTC 2018


hi,

after watching Johns slides from the kvm forum (thanks for that) i had
some quick look at the backup-v3 branch. Just to provide some feebdack
for you guys, and some questions.

My main question is about the part of the NBD backup. By default what
you get from reading all the NDB data is a thick provisioned image of
the domains disk. One can use the `qemu-img map' function to get a
detailed information about the used blocks in the image, in case one
wants to create a thin provisioned backup.

As a third party backup vendor you cannot allways depend on qemu tools,
because you might not even install any software on the host you are
taking a backup from. So is, or will there be any way to get an output
that represents the same information from the map function in the backup
XML description via the libvirt API?  Would it make sense to provide
this information in the `backup-dumpxml' output?

>From what i know in the Citrix XEN implementation, they provide a way to read
this information via the API, because they do not want the backup vendor to
install any component on the host systems.

Another thing i came across is that libvirt currently seems to forget about the
running backup job if a domain is destroyed and started after a backup job was
created:

[root at x ~]# virsh backup-begin centos backup-pull.xml 
Backup id 1 started
backup used description from 'backup-pull.xml'
[root at x ~]# virsh destroy centos && virsh start centos
[root at x ~]# virsh backup-end --id 1 centos
error: Requested operation is not valid: No active block job 'tmp-hda'
[root at x ~]# virsh backup-dumpxml --id 1 centos
<domainbackup mode='pull' id='1'>
  <server transport='tcp' name='localhost' port='10809'/>
  <disks>
    <disk name='hda' type='file'>
      <driver type='qcow2'/>
      <scratch file='/vm/base.img.1542876138'/>
    </disk>
  </disks>
</domainbackup>


thanks for your hard work on this!

bye,
    - michael





More information about the libvir-list mailing list