yahoo-eng-team team mailing list archive

[OpenStack Infra <1639894@xxxxxxxxxxxxxxxxxx>]

Reviewed:  https://review.openstack.org/394585
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=2e32ed122b415009722456716c47d8374e9d449c
Submitter: Jenkins
Branch:    master

commit 2e32ed122b415009722456716c47d8374e9d449c
Author: Matt Riedemann <mriedem@xxxxxxxxxx>
Date:   Wed Nov 23 16:18:18 2016 -0500

    Fix race in test_volume_swap_server_with_error
    
    This functional test is waiting for a flag to be set in
    the CinderFixture which is called when swap_volume calls
    volume_api.unreserve_volume after the swap_volume error
    occurs. That happens after the two swap-volume notifications
    are sent (for start and error).
    
    The test was asserting a 5th notification for the generic
    instance fault, which happens, but might not happen before
    the CinderFixture flag is set and the test checks for 5
    notifications.
    
    This test really only cares that there are the two swap-volume
    notifications, so to avoid a race this change just makes
    the test assert that there are at least 4 versioned notifications
    by the time the swap-volume error occurs, which is all it was
    checking after that anyway.
    
    Change-Id: I8fd44b083f109c034958a305a4e41f9d810c4389
    Closes-Bug: #1639894


** Changed in: nova
       Status: In Progress => Fix Released

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1639894

Title:
  TestInstanceNotificationSample.test_volume_swap_server_with_error is
  racy

Status in OpenStack Compute (nova):
  Fix Released

Bug description:
  This failed on an unrelated change today:

  http://logs.openstack.org/24/394524/1/check/gate-nova-tox-db-
  functional-ubuntu-
  xenial/01a5cce/console.html#_2016-11-07_17_27_14_569696

  https://github.com/openstack/nova/blob/0132cc8c2663843a891e054d9185e6ba2fd589ad/nova/tests/functional/notification_sample_tests/test_instance.py#L547

  That says it expects 3 notifications, but it really only cares about
  2. Based on when the compute.exception happens, and when
  self._wait_until_swap_volume_error() returns True, the 3rd
  compute.exception notification might not have happened.

  The swap_error flag is set in the cinder fixture here:

  https://github.com/openstack/nova/blob/0132cc8c2663843a891e054d9185e6ba2fd589ad/nova/tests/fixtures.py#L868

  That happens here:

  https://github.com/openstack/nova/blob/0132cc8c2663843a891e054d9185e6ba2fd589ad/nova/compute/manager.py#L4936

  Which is after the swap-volume error notification is sent.

  The compute.exception comes from the instance fault handler here:

  https://github.com/openstack/nova/blob/0132cc8c2663843a891e054d9185e6ba2fd589ad/nova/compute/manager.py#L4961

  Which is after cinder.swap_error is set to true.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1639894/+subscriptions