Robert Kanter created OOZIE-1332:
------------------------------------
Summary: Flakey test
TestActionCheckXCommand.testActionCheckTransientDuringMRAction
Key: OOZIE-1332
URL: https://issues.apache.org/jira/browse/OOZIE-1332
Project: Oozie
Issue Type: Bug
Components: tests
Affects Versions: 3.3.2, trunk, 4.0.0
Reporter: Robert Kanter
Assignee: Robert Kanter
Fix For: trunk, 4.0.0
We've been seeing
TestActionCheckXCommand.testActionCheckTransientDuringMRAction as flakey on our
CI tests.
This incorrect sequence of events was occurring:
{noformat}
...
- ActionCheckXCommand start
- suspend WF
- ActionCheckXCommand end
- ActionCheckXCommand start
- JT up
- unknown hadoop job error
- fail WF
- ActionCheckXCommand end
- ResumeXCommand cannot start (because WF fail)
- test finish
{noformat}
when this correct sequence of events should have been occuring:
{noformat}
...
- ActionCheckXCommand start
- suspend WF
- ActionCheckXCommand end
- JT Up
- ResumeXCommand start
- ResumeXCommand end
- ActionStartXCommand start
- ActionStartXCommadn end
- ActionCheckXCommand start
- ActionCheckXCommand end
...
- test finish
{noformat}
It turns out that the ActionCheckerService was triggering an extra
ActionCheckXCommand at just the wrong moment. We should disable the
ActionCheckerService during this test to prevent this issue (When I originally
wrote the test, I incorrectly thought it was needed and that it had to run more
frequently to make the test faster).
As a pre-emptive measure, we should also do this for
TestActionCheckXCommand.testActionCheckTransientDuringLauncher, which is very
similar even though we didn't see any flakiness.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
