Generally this issue is caused by the Qppid or rabbitmq not synchronizing messages across the cluster , it can be debugged by looking into the Qppid messages stack , also you may try the simple fix below .
Make sure all compute services are up and running
Check the compute services are running on horizon dashboard
or you may execute the following command on the controller node
nova-manager service list
if all the compute services are up and running try restarting the Qppid service on the controller node , keep in mind that restarting Qppid would only reset the service and not flush the Qppid stack .
Information you should know about this article
Openstack Version : Havana
Operating System : Centos 6.5