[
https://issues.apache.org/jira/browse/OOZIE-3009?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16101683#comment-16101683
]
Andras Piros edited comment on OOZIE-3009 at 7/26/17 2:46 PM:
--------------------------------------------------------------
Root cause is following:
* there are (at least following) five test classes that are at the moment
retrying the database operations for an extended time period, but they
shouldn't (since these use {{FaultInjection}} facility, or perform any other
kind of faulty {{JPAService.execute()}} call):
** {{TestBundleJobsDeleteJPAExecutor}}
** {{TestCoordJobsDeleteJPAExecutor}}
** {{TestPurgeService}}
** {{TestPurgeXCommand}}
** {{TestWorkflowJobsDeleteJPAExecutor}}
* as a consequence, these tend to timeout often
* when that happens, all the other modules are skipped because of the timeout
* the test cases of the skipped modules are not rerun on the second run
We need to extend {{PersistenceExceptionSubclassFilterRetryPredicate}} by
following:
* if the {{Throwable}} does not have a root cause, and is either a
{{RuntimeException}} (in case of {{FaultInjection}}), or a
{{JPAExecutorException}} (e.g. {{Query.getResultList()}} returned an empty
{{List}}), the predicate returns {{false}}
* for all the other cases the existing behavior kicks in
It's also needed to enhance
{{PersistenceExceptionSubclassFilterRetryPredicate}} because there are lots of
places in the codebase where causeless {{JPAExecutorException}} instances are
thrown where there is actually a reasonable cause, e.g. {{NoResultException}}.
For those cases we will not retry the database operation.
was (Author: andras.piros):
Root cause is following:
* there are (at least following) five test classes that are at the moment
retrying the database operations for an extended time period, but they
shouldn't (since these use {{FaultInjection}} facility, or perform any other
kind of faulty {{JPAService.execute()}} call):
** {{TestBundleJobsDeleteJPAExecutor}}
** {{TestCoordJobsDeleteJPAExecutor}}
** {{TestPurgeService}}
** {{TestPurgeXCommand}}
** {{TestWorkflowJobsDeleteJPAExecutor}}
* as a consequence, these tend to timeout often
* when that happens, all the other modules are skipped because of the timeout
* the test cases of the skipped modules are not rerun on the second run
We need to extend {{PersistenceExceptionSubclassFilterRetryPredicate}} by
following:
* if the {{Throwable}} does not have a root cause, and is either a
{{RuntimeException}} (in case of {{FaultInjection}}), or a
{{JPAExecutorException}} (e.g. {{Query.getResultList()}} returned an empty
{{List}}), the predicate returns {{false}}
* for all the other cases the existing behavior kicks in
> Number of Oozie tests executed dropped after OOZIE-2854
> -------------------------------------------------------
>
> Key: OOZIE-3009
> URL: https://issues.apache.org/jira/browse/OOZIE-3009
> Project: Oozie
> Issue Type: Bug
> Reporter: Attila Sasvari
> Assignee: Andras Piros
> Priority: Blocker
>
> I noticed that the number of executed tests has been significantly dropped
> after OOZIE-2854.
> - Tests run: *1080* https://issues.apache.org/jira/browse/OOZIE-2854
> Previous tests:
> - OOZIE-2371 - Tests run: *1965*
> https://issues.apache.org/jira/browse/OOZIE-2371?focusedCommentId=16076996&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16076996
> - OOZIE-2911 - Tests run: *1966*
> https://issues.apache.org/jira/browse/OOZIE-2911?focusedCommentId=16078078&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16078078
> In OOZIE-2854, we can also see that the number of test cases vary from patch
> to patch.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
