On 08/21/2012 09:38 AM, Akendo wrote:
Most likely, this happened because you upgraded libvirt versions, and the file in /var/run/libvirt/qemu corresponding to the disappearing guest contained information placed by the older libvirt that the newer libvirt does not understand.
CentOS 6 ships a version of libvirt based on RHEL, and RHEL 6.3 only ships with libvirt 0.9.10 + patches. Some of those patches are important to preserve if you want a clean upgrade path.
Thank you very much for you reply. I upgraded the CentOS 6.3 some days ago, but this is a default version of CentOS.
It may be default CentOS, but the fact still remains that CentOS doesn't ship libvirt 0.9.13; it trails RHEL 6.3 which ships libvirt 0.9.10+patches.
I restarted the libvirtd without success of finding my VM. But very interesting is the fact there is a $dom.xml for my VM. Also is a $DOM.pid file.
Yes, the domain is still running, it's just that libvirtd couldn't figure out how to manage it because of the invalid XML.
But i found another problem: our system has 12 cores i allowed to use 8 cores which raise this error message: error : virDomainDefParseXML:8871 : Maximum CPUs greater than topology limit
Yep, that would explain it. Older libvirt silently ignored bugs like this in the XML, while newer libvirt rejects it; once the XML is rejected, the VM is no longer listed by libvirt.
This error message has for some reason suppress in the syslog (lor i just didn't read right...) i reduced the amount of CPU to 4. Now it's working again.
Yep, that's the correct fix. Glad we found it for you. -- Eric Blake eblake@redhat.com +1-919-301-3266 Libvirt virtualization library http://libvirt.org