When get evacuated instance of host through API filtering will cause the main nova process to block for an extended amount of time.

When destory the evacuated instances from a host,first it will get the host's instance which are found on the hypervisor.
If the driver doesn't support uuids listing, it will get all the instances on openstack.It will cause the main nova process to block for an extended amount of time.
like below.
https://github.com/openstack/nova/blob/master/nova/compute/manager.py
 def _destroy_evacuated_instances(self, context):
  ...
        filters = {'deleted': False}
        local_instances = self._get_instances_on_driver(context, filters)
  ...

 def _get_instances_on_driver(self, context, filters=None):
   ...
        # The driver doesn't support uuids listing, so we'll have
        # to brute force.
        driver_instances = self.driver.list_instances()
        instances = objects.InstanceList.get_by_filters(context, filters,
                                                        use_slave=True)

So when get instances on the host which should be destoryed,it should filtering by uuids which were evacuated from current host.
 def _destroy_evacuated_instances(self, context):
   ...
        filters = {'deleted': False,
                     'uuid':[mig.instance_uuid for mig in evacuations]
                    }
        local_instances = self._get_instances_on_driver(context, filters)
        evacuations = {mig.instance_uuid: mig for mig in evacuations}
        evacuated = [inst for inst in local_instances
                            if inst.uuid in evacuations]
        for instance in evacuated:
   ...