test_delete_saving_image fails because image hasn't transitioned to SAVING

Bug #1713163 reported by Ihar Hrachyshka on 2017-08-25
16
This bug affects 2 people
Affects Status Importance Assigned to Milestone
tempest
Medium
Unassigned

Bug Description

http://logs.openstack.org/33/440933/4/check/gate-tempest-dsvm-neutron-linuxbridge-ubuntu-xenial/ebfa453/logs/testr_results.html.gz

Traceback (most recent call last):
  File "tempest/api/compute/images/test_images.py", line 48, in test_delete_saving_image
    wait_until='SAVING')
  File "tempest/api/compute/base.py", line 301, in create_image_from_server
    image_id, wait_until)
  File "tempest/common/waiters.py", line 179, in wait_for_image_status
    raise lib_exc.TimeoutException(message)
tempest.lib.exceptions.TimeoutException: Request timed out
Details: (ImagesTestJSON:test_delete_saving_image) Image 93090479-409d-49f9-b2d3-2bc11a40f967 failed to reach SAVING state (current state ACTIVE) within the required time (196 s).

I briefly looked at the test case, and it seems to me that it may happen when the image snapshot is produced quicker than tempest being able to catch the intermittent transition. If so, the test case may need a refinement to check that a snapshot was produced and not that the state was in SAVING for some time.

tags: added: gate-failure
jichenjc (jichenjc) wrote :

accordingly to avoid description, looks like this is a tempest issue?

affects: nova → tempest
Changed in tempest:
status: New → Confirmed
assignee: nobody → chandan kumar (chkumar246)

As far as I understand, this test shouldn't even exist in scope of tempest, because you can't guarantee that backend will be slow to process the request. This kind of interaction would be better served by a test with better execution flow control. I am not sure if glance have those, but in neutron realm it would be a functional test with backend persistence artificially slowed down.

Added glance to the list of affected projects because I believe they will need to port the tempest test to somewhere else.

Brian Rosmaita (brian-rosmaita) wrote :

Marking as invalid for Glance. The problem is occurring in a Compute API (nova) test. I think your analysis is correct, that this is a timing problem and that this may be too fine-grained a test; the test should use a different success criteria.

Changed in glance:
status: New → Invalid
Matt Riedemann (mriedem) on 2019-02-01
Changed in tempest:
assignee: chandan kumar (chkumar246) → nobody
importance: Undecided → Medium
no longer affects: glance
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Duplicates of this bug

Other bug subscribers