Move All Instances From Hypervisor in Case of Hardware Failure

Problem

In case of hardware failure for a hypervisor, the instances need to be moved off the host.

Environment

• Platform9 Managed OpenStack - All Versions
• Nova
• Shared storage ( mandatory)

Procedure

Warning: this method should not be used if the host still has instances up and running. This will also fail if the compute service on the host is running.

1. To find a host for the evacuated instances, list all hosts.

None

$ openstack host list

1. Evacuate all instances from FAILED_HOST to TARGET_HOST.

None

$ nova host-evacuate --target_host TARGET_HOST FAILED_HOST

Note: The option --target_host is optional and can be omitted to let the scheduler decide where to place the instances.

Additional Information

• The argument FAILED_HOST can also be a pattern to search for instead of an exact hypervisor hostname but it is recommended to use a fully qualified domain name to make sure no hypervisor host is getting evacuated by mistake. As long as you are not using a pattern you might want to use the --strict flag.