something changed in the kvm audio subsystem, so a virt that was suspended during shutdown wouldn't resume after reboot. the logs had complaints of:
savevm: unsupported version 2 for 'hda-audio' v1 load of migration failed
i poked around on the rh bz site and found a few hits, but they seemed to be related to older versions. i forced a clean boot of my vm and it was happy.
On 03.03.2013 19:21, Joe Pruett wrote:
something changed in the kvm audio subsystem, so a virt that was suspended during shutdown wouldn't resume after reboot. the logs had complaints of:
savevm: unsupported version 2 for 'hda-audio' v1 load of migration failed
i poked around on the rh bz site and found a few hits, but they seemed to be related to older versions. i forced a clean boot of my vm and it was happy.
Sounds like expected behaviour to me; because the kvm in 6.4 is different in significant ways it's very problematic to resume. I would imagine live migration between the previous version in 6.3 and the one in 6.4 will also not work. This is not something new; this kind of stuff requires the exact same version of KVM when playing with the guest's memory. This is all very IMHO and AFAIK.
HTH Lucian
On 2013-03-03 12:05, Nux! wrote:
On 03.03.2013 19:21, Joe Pruett wrote:
something changed in the kvm audio subsystem, so a virt that was suspended during shutdown wouldn't resume after reboot. the logs had complaints of:
savevm: unsupported version 2 for 'hda-audio' v1 load of migration failed
i poked around on the rh bz site and found a few hits, but they seemed to be related to older versions. i forced a clean boot of my vm and it was happy.
Sounds like expected behaviour to me; because the kvm in 6.4 is different in significant ways it's very problematic to resume. I would imagine live migration between the previous version in 6.3 and the one in 6.4 will also not work. This is not something new; this kind of stuff requires the exact same version of KVM when playing with the guest's memory. This is all very IMHO and AFAIK.
it wasn't a huge deal for me since it was just a windows 7 vm on my desktop, but i would be quite cranky about this on a server with lots of client vms. maybe they wouldn't have audio and thus be unaffected, but forcing a reboot of all guests to facilitate an upgrade of the host isn't very friendly of our upstream overlords.
On Sun, Mar 03, 2013 at 12:24:48PM -0800, Joe Pruett wrote:
it wasn't a huge deal for me since it was just a windows 7 vm on my desktop, but i would be quite cranky about this on a server with lots of client vms. maybe they wouldn't have audio and thus be unaffected, but forcing a reboot of all guests to facilitate an upgrade of the host isn't very friendly of our upstream overlords.
If this was a server with lots of VMs one would hope that proper testing in a proper testing environment would weed out these sorts of problems before live / production instances are in any way affected. This is an issue if this was not noted in the releases notes, otherwise it's just the cost of doing business.
John
-
Joe Pruett -
John R. Dennison -
Nux!