Open Stack

Hey,

I am having trouble live-migrating relatively large VMs (16 cores, 64G
ram). I am seeing a constant 50-60MBps traffic between source and
destination hypervisors, but the migration wasn't done even after 10+
hours. I am using cinder with rbd backend.

People on libvirt mailing list recommended me to suspend the VM in
question, since the migration could never finish if memory changes faster
than network could transfer. I did "virsh suspend <domain>" on the source
host, and the migration finished right away.

Unfortunately, nova-compute log shows an "Unauthorized" error on the
loopingcall (waiting for migration status via libvirt?) Even though the VM
migration finished successfully (and VM being reachable), the VM is stuck
in migrating state on the original host according to nova.

A few questions:
1) If in the future I need to interfere with VM migration again (perform
manual suspend), what is the right way to do so in order not to upset
nova-compute?
2) What's the recommended way to recover from my current state?

Thanks.
-Simon
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20150122/68255485/attachment.html>