Comment 13 for bug 1436343

Note: the reproducing steps can be simplified as the following:
instead of killing a controller node, just restart the corosync and next, the pacemaker service