in _build_instance UnexpectedTaskStateError does not deallocate the network

[Joshua Harlow]

It appears that when a build & spawn fails with a UnexpectedTaskStateError there is no attempt to deallocate the network, this might be a feature or flaw (not entirely clear). If its a flaw, its likely leaving the allocated network orphaned.

[Dan Smith (danms)]

Please provide a way to actually reproduce this so I can take a look.

[Joshua Harlow (harlowja)]

Will see if I can get a way to reproduce that isn't by just looking at the code.

Here is the snippet that I was talking about. Comments about what might not be right marked with <<<<

         instance = self._spawn(context, instance, image_meta,
                                       network_info, block_device_info,
                                       injected_files, admin_password,
                                       set_access_ip=set_access_ip)
        except exception.InstanceNotFound:
            # the instance got deleted during the spawn
            with excutils.save_and_reraise_exception():
                try:
                   <<<< Deallocated network if somehow the instance went poof before we could finish spawn()
                    self._deallocate_network(context, instance)
                except Exception:
                    msg = _('Failed to dealloc network '
                            'for deleted instance')
                    LOG.exception(msg, instance=instance)
        except exception.UnexpectedTaskStateError as e:
            <<<< No deallocation of network here :(
            actual_task_state = e.kwargs.get('actual', None)
            if actual_task_state == 'deleting':
                msg = _('Instance was deleted during spawn.')
                LOG.debug(msg, instance=instance)
                raise exception.BuildAbortException(
                        instance_uuid=instance['uuid'], reason=msg)
            else:
                raise
        except Exception:

[Chris Behrens (cbehrens)]

That is hit when you do a quick boot/delete.. ie, delete while instance is being scheduled.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/33460

[Andrew Laski (alaski)]

The code in question here is no longer in use. So without further confirmation that this issue is still occuring I'm marking this incomplete for now.