Windows 2008 R2 VM Guest live migration issue. KVM on RHEL6.1

Latest response

Hi All,

 

I'm very new to this forum. Apologies if its not the right forum for the question - please move the topic for me if that is the case.

 

Basically, I have 2 RHEL6.1 machines with drbd blocks where I have VM guests installed.

Drbd replicates guests images. I've installed Win2k3-R2-32bit and Win2k8-R2-64bit guests. Both installed and run normally. However when I try consequent live migration of those guests between machines (qemu+ssh, passwordless authentication), host1->host2->host1->host2->host1, etc. ,Win2k3 guest migrates with no problems, whereas Win2k8 guest migrations fail intermittently, and virtual machine stops with error:

 

error: cannot send monitor command '{"execute":"query-migrate"}': Connection reset by peer
 

I have installed latest version of virtio drivers too (but they look exactly the same in windows - version 6.0.209.605, only the modification date is different). virtio-win-1.3.3-0.el6.noarch.

That hasn't helped.

I have got exactly the same error for both live and test environments, for live environment I have a couple of Dell R610s and for test environment I have a couple of Dell PCs. Any ideas ? I can provide more details if required.

 

Regards,

Alex

 

 

 

 

 

 

Responses

Hi David,

 

Do you see anything in the libvirt logs?

What about the VM, when migration fails, does it end up turned off, crashed, or still running on the source host?

The error you reported, do you see it on the source or destination host?

Can you post the VM details - vcpu settings, memory sizes, disk types etc

 

Cheers,

Dan

Hi Dan,

 

It appears to be the problem with Netgear switches where we have hosts connected to. So I don't think it has anything to do with KVM at all as I tested the issue on the other two hosts connected via crossover - never hit any issues, whereas having them connected via switches showed the same occasional issues.

 

many thanks for your help anyway

 

Alex

Thanks for the update,

 

Can you please confirm that with a higher end, more enterprisey switch, migrations work fine? Just to be certain

 

Thanks,

Dan

Heh... those Netgears are of GSM7352 type - the most enterprisey-ish Gigabit layer 3 switches  Netgear produces :) So if you want NOT to have any issues - don't touch Netgear :) unless its for home :) but even then :)

Unfortunatelly, I cannot test it with any Cisco at the moment as this company doesn't have any :(

Well, I wouldn't put it down to the switches being at fault until I've tested with different switches. And this can be a good reason to push for some hardware upgrades. Or to get in touch with Netgear support and have them analyze their net gear, maybe there's a firmware update or an internal setting that could make everything alright.