[
https://issues.apache.org/jira/browse/YARN-585?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zhijie Shen updated YARN-585:
-----------------------------
Attachment: YARN-585.1.patch
After YARN-514, storing application becomes a state of an RMApp's life cycle.
RMStateStore needs the applicationId from ApplicationSubmissionContext, it
cannot be null. Therefore, we need to make sure applicationId is set in
ApplicationSubmissionContext when the RMApp's transitions are tested.
A specific problem with TestFairScheduler#testNotAllowSubmitApplication is that
due to one more state before RMApp enters SUBMITTED, APP_REJECTED is already
sent to the attempt. It results in an undefined transition. So before
generating the APP_REJECTED event, main thread should be hung on for 100ms and
yield the process to asyncDispatcher to move MRApp to SUBMITTED, where the
attempt is instantiated and the START event is sent. Therefore, the attempt
gets START, moves from NEW to SUBMITTED, where the transition from SUBMITTED to
FAILED on APP_REJECTED is defined.
> TestFairScheduler#testNotAllowSubmitApplication is broken due to YARN-514
> -------------------------------------------------------------------------
>
> Key: YARN-585
> URL: https://issues.apache.org/jira/browse/YARN-585
> Project: Hadoop YARN
> Issue Type: Bug
> Reporter: Zhijie Shen
> Assignee: Zhijie Shen
> Attachments: YARN-585.1.patch
>
>
> TestFairScheduler#testNotAllowSubmitApplication is broken due to YARN-514.
> See the discussions in YARN-514.
--
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
