When some instance aren't deleted correctly and libvirt still keep the domain for the instance, the resource tracker will failed to update available resource

[Alex Xu]

When instance was deleted in the db, but it is still at compute node, the resource tracker will fail to update available resource.

2016-03-04 10:58:28.143 ERROR nova.compute.manager [req-d2f1c99b-0e81-4b6d-9361-a40bd2218141 None None] Error updating resources for node vm6.

2016-03-04 10:58:28.143 TRACE nova.compute.manager Traceback (most recent call last):
2016-03-04 10:58:28.143 TRACE nova.compute.manager File "/opt/stack/nova/nova/compute/manager.py", line 6379, in update_available_resource
2016-03-04 10:58:28.143 TRACE nova.compute.manager rt.update_available_resource(context)
2016-03-04 10:58:28.143 TRACE nova.compute.manager File "/opt/stack/nova/nova/compute/resource_tracker.py", line 491, in update_available_resource
2016-03-04 10:58:28.143 TRACE nova.compute.manager resources = self.driver.get_available_resource(self.nodename)
2016-03-04 10:58:28.143 TRACE nova.compute.manager File "/opt/stack/nova/nova/virt/libvirt/driver.py", line 5414, in get_available_resource
2016-03-04 10:58:28.143 TRACE nova.compute.manager disk_over_committed = self._get_disk_over_committed_size_total()
2016-03-04 10:58:28.143 TRACE nova.compute.manager File "/opt/stack/nova/nova/virt/libvirt/driver.py", line 7047, in _get_disk_over_committed_size_total
2016-03-04 10:58:28.143 TRACE nova.compute.manager local_instances[guest.uuid], bdms[guest.uuid])
2016-03-04 10:58:28.143 TRACE nova.compute.manager KeyError: '49505c88-b38a-4100-ab56-97958b48b533'
2016-03-04 10:58:28.143 TRACE nova.compute.manager

The available resource won't get update until periodic_task '_cleanup_running_deleted_instances' if running_deleted_instance_action is 'reap'

[Matt Riedemann (mriedem)]

Actually I think this is already fixed:

https://review.openstack.org/#/c/221162/

Now it checks:

if guest.uuid in local_instances: