Comment 1 for bug 1443438

With access to a failed environment I found that compute nodes seem to be dying after/while launching VMs. I manually brought all nodes online and attempted to run OSTF smoke tests [0]. Test 9 failed and at almost the exact same time the compute node where those VMs were scheduled went away. There was no network connectivity and I wasn't able to access the host over VNC. The compute node VM was still running according to libvirt. I have also attached a diagnostic snapshot.

[0] http://paste.openstack.org/show/204907/