libvirt users - Oct 2012 - Migrating fails with "Timed out during operation: cannot acquire state change lock"

Hi,

After recently upgrading to 0.9.11 (as shipped with Fedora 17), I was just 
trying to migrate a qemu domain from one host to another (using 
virDomainMigrateToURI()), when I got this error message:

Unsafe migration: Migration may lead to data corruption if disks use cache != 
none

Okay, this is explainable (if a bit disappointing - I would have hoped that 
qemu would be able to disable disk caches before migration and then reenable 
them afterwards...). However, the next time I tried migrating the same domain, 
I got this error message after about 30 seconds:

Timed out during operation: cannot acquire state change lock

This looks like a bug to me - granted not a high profile one if it only 
happens when someone is retrying a failed migration, but still...

	Guido

Hi.

On Tue, Oct 16, 2012 at 16:20:59 +0200, Guido Winkelmann
wrote:
> After recently upgrading to 0.9.11 (as shipped with Fedora 17), I was just 
> trying to migrate a qemu domain from one host to another (using 
> virDomainMigrateToURI()), when I got this error message:
> 
> Unsafe migration: Migration may lead to data corruption if disks use cache
!=
> none
> 
> Okay, this is explainable (if a bit disappointing - I would have hoped that
> qemu would be able to disable disk caches before migration and then
reenable
> them afterwards...). However, the next time I tried migrating the same
domain,
> I got this error message after about 30 seconds:
> 
> Timed out during operation: cannot acquire state change lock
Oops, that's a longstanding bug which hasn't been noticed until now.
Thanks
for reporting it. Luckily, the fix is trivial:
https://www.redhat.com/archives/libvir-list/2012-October/msg00891.html

Jirka