6 Apr
2016
6 Apr
'16
5:49 p.m.
On 04/04/16 16:03 -0400, Tim Macy wrote:
Occasionally I have a single VM go into an "?" unknown state and the only method of recovery is to kill the PID on the host. This happens once or twice a week to 1 out of 500+ VM's. There is no pattern to this and VM operating system doesn't seem to make any difference.
Logs are attached. aus02gmovirt01 is the Engine aus02gdc1kvm02 is the VM Host
Please also attach a vdsm.log (/var/log/vdsm/vdsm.log) taken from the host where the non-responsive VM is running. If you could also collect 'journalctl -u libvirtd' that would help to see if libvirt is to blame. -- Adam Litke