libvirt users - Aug 2015 - Live migration & storage copy broken since 1.2.17

Hi,

It seems that live migration using storage copy is broken since libvirt
1.2.17.

Here is the command line used to do the migration using virsh:

  virsh migrate --live --p2p --persistent --undefinesource --copy
-storage-all d2b545d3-db32-48d3-b7fa-f62ff3a7fa18
qemu+tcp://dest/system

XML dump of my storage:

  <pool type='logical'>
    <name>local</name>
    <uuid>276bda97-d6c2-4681-bc3f-0c8c221bd1b1</uuid>
    <capacity unit='bytes'>1024207093760</capacity>
    <allocation unit='bytes'>4294967296</allocation>
    <available unit='bytes'>1019912126464</available>
    <source>
      <name>hkvm</name>
      <format type='lvm2'/>
    </source>
    <target>
      <path>/dev/hkvm</path>
    </target>
  </pool>

When it fails, I get the following error:

  error: Storage volume not found: no storage vol with matching name
'402aea08-26aa-45b6-a46e-a3b02137ff26'

Where 402aea08-26aa-45b6-a46e-a3b02137ff26 is the volume bound to my
VM:

      <disk type='volume' device='disk'>
        <driver name='qemu' type='raw'
cache='none'/>
        <source pool='local' volume='402aea08-26aa-45b6-a46e
-a3b02137ff26'/>
       <backingStore/>
        <target dev='vda' bus='virtio'/>
        <alias name='virtio-disk0'/>
        <address type='pci' domain='0x0000'
bus='0x00' slot='0x04'
function='0x0'/>
     </disk>

Here is the test result (version is for both source and destination):

- 1.2.15 -> ok
- 1.2.16 -> ok
- 1.2.17 -> fails
- 1.2.18 -> fails

Is it a bug? Do I miss something?

Thanks,

Antoine

Hi,

I've done a git bisec between v1.2.16 and v1.2.17 to try to debug this
problem myself, and I found that the "problem" comes from the
following
commit:

93a19e283e5a147d147d84383aaaa8be63b6a68d qemu: migration: selective
block device migration

So it seems that I miss something :-).

This commit implements a --migrate-disks option on virsh migrate
command which allow to select which disks are to be migrated. By
default, disks are not migrated, this is why I got this error.

Good news, this new feature also solves my problem to migrate VMs with
local + rbd disks.

Antoine