libvirt users - Oct 2011 - Live Migration using Libvirt migrate()

Hi,
This is a follow up of the mail regd migration I'd posted a few days ago. My
scenario:

I'm creating two virtual machines images from an Ubuntu 10.10 iso of RAM
size 128 and 700 MB.

Using the migrate() , I migrate it from the current physical machine to the
destination machine via a 100 Mbps  switch LAN connection. BOTH migrations
take the same amount of time.

Checking out the eth0's transmitted and received bytes after and before the
migration, I find that there's NO change in the readings. Using any other
method to migrate(virsh,Qemu) takes a substantial amount of time(close to 10
seconds) for the 128 and 700 MB images.

How can this drastic change in migration time be accounted for?

Are there any other ways to check if the VM has been successfully migrated
to the destination(i tried using virsh console after migrate() but for some
reason it just hangs. I tried out the suggested patches on various sites but
to no avail)?

Is pinging the migrated VM from its PM a sure test for successful migration?

Thanks guys.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://listman.redhat.com/archives/libvirt-users/attachments/20111010/a5cfccf5/attachment.htm>

On 10/10/2011 04:28 AM, PREETHI RAMESH wrote:
> Hi,
> This is a follow up of the mail regd migration I'd posted a few days
ago. My
> scenario:
>
> I'm creating two virtual machines images from an Ubuntu 10.10 iso of
RAM
> size 128 and 700 MB.
>
> Using the migrate()
Which language binding?  Show the actual code snippet you used to do the 
migration (whether by 'virsh migrate ...' or whatever language binding 
you were using).

 > , I migrate it from the current physical machine to
the
> destination machine via a 100 Mbps  switch LAN connection. BOTH migrations
> take the same amount of time.
>
> Checking out the eth0's transmitted and received bytes after and before
the
> migration, I find that there's NO change in the readings.
As in no packets transmitted, or as in the same amount of packets 
transmitted for both migration memory sizes?

 > Using any other
> method to migrate(virsh,Qemu) takes a substantial amount of time(close to
10
> seconds) for the 128 and 700 MB images.
>
> How can this drastic change in migration time be accounted for?
Most likely, if things are happening super-fast with no change in packet 
counts, then you aren't using the migration API correctly.
>
> Are there any other ways to check if the VM has been successfully migrated
> to the destination(i tried using virsh console after migrate() but for some
> reason it just hangs. I tried out the suggested patches on various sites
but
> to no avail)?
>
> Is pinging the migrated VM from its PM a sure test for successful
migration?
Pinging is a sure test that the guest is still running, but not a sure 
test of which host is running the guest.

I'm not sure you have quite presented enough details to make your setup 
reproducible by others, which is probably why no one has been able to 
give more specific answers yet.

-- 
Eric Blake   eblake at redhat.com    +1-801-349-2682
Libvirt virtualization library http://libvirt.org