Instance remains in deleting state if libvirt down during deletion

[Unmesh Gurjar]

Scenario: While terminating an instance, if the libvirt process goes down just before the 'destroy' method gets called from virt.libvirt.connection module, instance remains in 'deleting' state.

Expected Behavior: Compute must handle the exception and update the instance vm_state='error'
Actual Result: The instance remains in the following state: vm_state='active', task_state='deleting', power_state=1.

Compute manager must handle the libvirt exception and set the instance to error state in this scenario.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/27551

[Yun Mao (yunmao)]

If the intention is to delete the instance, I'm not sure setting the VM state to Error is the best if we could retry deleting later somehow, say via periodic tasks.

[Sean Dague (sdague)]

Long incomplete bug. Please reopen if still useful.