Ironic compute node doesn't take over nodes with instance when the owner compute node is down

   Description
   ===========
   Ironic compute node doesn't take over nodes with instance when the owner compute node is down

   Steps to reproduce
   ==================
   1. I have two ironic compute nodes & two BM nodes
   2. BM1 is controlled by node1, BM2 is controlled by node2
   3. I boot an instance on BM1
   4. Stop nova-compute service on node1
   5. node2 doesn't take over BM1 although node1 is not available

   Expected result
   ===============
   What I expect is that when node1 id down node2 can take over BM1.

   Actual result
   =============
   node1 doesn't take over BM1.
   And all nova operate action will fail as BM1 has nova compute service on node1 is down

   Environment
   ===========
   openstack-nova-common-14.0.3-9.el7ost.noarch
   openstack-nova-novncproxy-14.0.3-9.el7ost.noarch
   openstack-nova-scheduler-14.0.3-9.el7ost.noarch
   python-novaclient-6.0.0-1.el7ost.noarch
   openstack-nova-conductor-14.0.3-9.el7ost.noarch
   openstack-nova-api-14.0.3-9.el7ost.noarch
   python-nova-14.0.3-9.el7ost.noarch
   openstack-nova-cert-14.0.3-9.el7ost.noarch
   openstack-nova-console-14.0.3-9.el7ost.noarch
   openstack-nova-compute-14.0.3-9.el7ost.noarch
   openstack-nova-common-14.0.3-9.el7ost.noarch