[nova][resize/migrate][performance] the whole virtual size of vm's disk is being copied instead of the actual size.

Description of problem:

When resizing/migrating an instance with qcow2 image:

1) nova is converting the image to raw (in order to unite it with the backing file )

2012-07-16 14:06:41 DEBUG nova.utils [req-58761caf-98ab-4bea-9719-8f9472d86205 29af43364ced4611bf6e093a04d4c144 3f1e394b238d4f26b742e9685a9a1a57] Running cmd (subprocess): qemu-
img convert -f qcow2 -O raw /var/lib/nova/instances/instance-00000154_resize/disk /var/lib/nova/instances/instance-00000154_resize/disk_rbase from (pid=26760) execute /usr/lib/p
ython2.6/site-packages/nova/utils.py:220
2012-07-16 14:07:00 DEBUG nova.utils [req-58761caf-98ab-4bea-9719-8f9472d86205 29af43364ced4611bf6e093a04d4c144 3f1e394b238d4f26b742e9685a9a1a57] Running cmd (subprocess): scp /
var/lib/nova/instances/instance-00000154_resize/disk_rbase 10.35.77.4:/var/lib/nova/instances/instance-00000154/disk from (pid=26760) execute /usr/lib/python2.6/site-packages/no
va/utils.py:220

~20 seconds

2) scp copies the converted image to destination host

qemu-img info:
# qemu-img info david.disk
image: david.disk
file format: raw
virtual size: 10G (10737418240 bytes)
disk size: 947M

result: 10G of data is being copied instead of 947M.

# scp david.disk root@camel-vdsb:/tmp/david.disk.copy
                                                                                                                                   100% 10GB 41.1MB/s 04:09

04:09 minutes

3) afterwards in order to be able to create snapshots on this image its being converted again to qcow2 on destenation host

2012-07-16 14:10:39 DEBUG nova.utils [req-58761caf-98ab-4bea-9719-8f9472d86205 29af43364ced4611bf6e093a04d4c144 3f1e394b238d4f26b742e9685a9a1a57] Running cmd (subprocess): qemu-img convert -f raw -O qcow2 /var/lib/nova/instances/instance-00000154/disk /var/lib/nova/instances/instance-00000154/disk_qcow from (pid=17788) execute /usr/lib/python2.6/site-packages/nova/utils.py:220

2012-07-16 14:13:12 DEBUG nova.utils [req-58761caf-98ab-4bea-9719-8f9472d86205 29af43364ced4611bf6e093a04d4c144 3f1e394b238d4f26b742e9685a9a1a57] Running cmd (subprocess): mv /var/lib/nova/instances/instance-00000154/disk_qcow /var/lib/nova/instances/instance-00000154/disk from (pid=17788) execute /usr/lib/python2.6/site-packages/nova/utils.py:220

~02:33 minutes

Tested alternative:
1) convert the image from qcow2 to qcow2 (in order to unite it with the backing file )

2) rsync --sparce the file to the destination

# qemu-img info david.qcow.disk
image: david.qcow.disk
file format: qcow2
virtual size: 10G (10737418240 bytes)
disk size: 953M
cluster_size: 65536

# time rsync -S -v david.qcow.disk root@camel-vdsb:/tmp/david.disk.rsync-copy
david.qcow.disk

sent 999218356 bytes received 31 bytes 39185034.78 bytes/sec
total size is 999096320 speedup is 1.00

real 0m24.797s
user 0m20.237s
sys 0m4.998s

Expected results:
in this case 24 seconds instead of 4:09 minutes.
conversion avoided after copy. (less 02:33 minutes)