Details
-
Bug
-
Resolution: Unresolved
-
Minor
-
None
-
Lustre 2.13.0, Lustre 2.14.0, Lustre 2.12.5
-
3
-
9223372036854775807
Description
We see, in LU-13759, that a sanityn test that is run as part of sanity-dom can fail and the failure is not propagated to a level that Maloo recognizes it as a failure. There are several problems that are also seen with sanity-dom logs; logs for the failed tests are not collected and client log for successful tests are not collected. Note that this seems like an issue with the test framework and not with Maloo.
For failure at https://testing.whamcloud.com/test_sets/08c6fa9d-e2a3-457b-a1ed-b4318dbf166a, we can see that the sanity-dom/sanityn test 20 failure is recognized, but this does not lead to the sanity-dom test suite being marked at failed. In the results file at https://testing.whamcloud.com/test_sessions/b31ca578-dd4f-4725-9a0c-5e19f3031c69/show_results, we see that the fail is registered, but
- name: test_sanityn
- name: test_1
status: PASS
duration: 4
return_code: 0
error:
…
- name: test_19
status: SKIP
duration: 2
return_code: 0
error: not cache-capable obdfilter
- name: test_20
status: FAIL
duration: 5
return_code: 1
error: 1 page left in cache after lock cancel
- name: test_23
status: PASS
duration: 65
return_code: 0
error:
…
- name: test_51d
status: PASS
duration: 435
return_code: 0
error:
duration: 1129
status: PASS
We had a related issue where, when a sanity-dom/sanityn test failed, the failure would trigger the whole test suite to fail, but Maloo thinks the last test in sanityn failed which is false; see LU-10589 and one example failure at https://testing.whamcloud.com/test_sets/870c7a78-467f-11e9-9646-52540065bddc. In this case it looks like all logs are not collected.
Patrick Farrell produced a patch which clears the problem with a sub test of a sub suite failure not triggering the suite to be marked as fail at https://review.whamcloud.com/#/c/34186/ . I think the no log collected/displayed issue still exists with this proposed solution.