Comment 2 for bug 1586065

Pavel, could you please clarify what version of libvirt was used? What image did you use for booting VMs?

I checked the attached logs ( http://mos-scale-share.mirantis.com/small_logs.tar.bz2 ), but unfortunately these won't be enough, as we also need libvirtd and qemu logs: just based on the traceback in Nova logs nova-compute timed out waiting for a disk to be detached from a libvirt domain.

Could you please attach logs from /var/log/libvirtd directory on the affected compute nodes?

Note, that we already retry the detach call multiple times in Nova - https://github.com/openstack/nova/blob/master/nova/virt/libvirt/guest.py#L299-L346 (waiting up to 30s), so it's not clear why qemu did not manage to detach the device within that time.