This is available if you use vSphere as a hypervisor for Openstack.

Otherwise you can write your own automation to detect when a HV fails and restart the instances on another HV. Shared storage will be required between compute nodes to make this happen.

Openstack has an evacuate feature:

Evacuate instances

If a cloud compute node fails due to a hardware malfunction or another reason, you can evacuate instances to make them available again. You can optionally include the target host on the evacuate command. If you omit the host, the scheduler determines the target host.

To preserve user data on server disk, you must configure shared storage on the target host. Also, you must validate that the current VM host is down; otherwise, the evacuation fails with an error.

http://docs.openstack.org/user-guide-...