reschedule failed because port still in use

[Liusheng]

When booting an instance with a port specified, if the instance spawn faild for libvit error, the instance will be rescheduled, and will raise a "PortInUse" exception.

To reproduce, we can add "raise Exception" after spawn in _build_instance() and restart nova-compute.

more details plz see:
Traceback (most recent call last):
  File "/usr/lib64/python2.6/site-packages/nova/compute/manager.py", line 1043, in _build_instance
    set_access_ip=set_access_ip)
  File "/usr/lib64/python2.6/site-packages/nova/compute/manager.py", line 1426, in _spawn
    LOG.exception(_('Instance failed to spawn'), instance=instance)
  File "/usr/lib64/python2.6/site-packages/nova/compute/manager.py", line 1423, in _spawn
    block_device_info)
  File "/usr/lib64/python2.6/site-packages/nova/virt/libvirt/driver.py", line 2083, in spawn
    admin_pass=admin_password)
  File "/usr/lib64/python2.6/site-packages/nova/virt/libvirt/driver.py", line 2480, in _create_image
    net = netutils.get_injected_network_template(network_info)
  File "/usr/lib64/python2.6/site-packages/nova/virt/netutils.py", line 74, in get_injected_network_
    if not (network_info and template):
  File "/usr/lib64/python2.6/site-packages/nova/network/model.py", line 379, in __len__
    return self._sync_wrapper(fn, *args, **kwargs)
  File "/usr/lib64/python2.6/site-packages/nova/network/model.py", line 366, in _sync_wrapper
    self.wait()
  File "/usr/lib64/python2.6/site-packages/nova/network/model.py", line 398, in wait
    self[:] = self._gt.wait()
  File "/usr/lib64/python2.6/site-packages/eventlet/greenthread.py", line 168, in wait
    return self._exit_event.wait()
  File "/usr/lib64/python2.6/site-packages/eventlet/event.py", line 120, in wait
    current.throw(*self._exc)
  File "/usr/lib64/python2.6/site-packages/eventlet/greenthread.py", line 194, in main
    result = function(*args, **kwargs)
  File "/usr/lib64/python2.6/site-packages/nova/compute/manager.py", line 1244, in _allocate_network
    dhcp_options=dhcp_options)
  File "/usr/lib64/python2.6/site-packages/nova/network/neutronv2/api.py", line 243, in allocate_for
    raise exception.PortInUse(port_id=port_id)
PortInUse: Port faf3aa64-11f8-4fc7-81bc-084098014f4a is still in use.

[Openstack Gerrit (openstack-gerrit)]

Fix proposed to branch: master
Review: https://review.openstack.org/87915

[Openstack Gerrit (openstack-gerrit)]

Fix proposed to branch: stable/havana
Review: https://review.openstack.org/89505

[Chris Behrens (cbehrens)]

actually, this bug may no longer be valid... because I think the build process doesn't deallocate and attempt to reallocate on reschedules anymore (unless the driver is baremetal/ironic)... Can you check the latest code from master to see if this is still a problem?

[Chris Behrens (cbehrens)]

Ignore that last comment.

[Chris Behrens (cbehrens)]

So I think the real problem here is that the device_id was never unassigned before the instance rescheduled. I suspect this review that landed 5 days ago happens to fix this bug also:

https://review.openstack.org/#/c/99400/

[OpenStack Infra (hudson-openstack)]

Change abandoned by liusheng (<email address hidden>) on branch: master
Review: https://review.openstack.org/87915
Reason: abandon, the https://review.openstack.org/#/c/99400/ has fixed this bug :(

[Liusheng (liusheng)]

Chris Behrens: thanks, it is. this bug has been fixed in https://review.openstack.org/#/c/99400/.

[Chris Behrens (cbehrens)]

Confirmed that review above does fix this problem too. Marking this as duplicate as 1321381 and closing.

[Chris Behrens (cbehrens)]

nevermind... was changed to invalid already.