Jason Lowe commented on YARN-3131:

bq. We can just simply check for failToSubmitStates? Why do we also need to 
check for waitingStates?

If we only check for failToSubmitStates then we'll continue to loop waiting for 
those failed states.  We need to check for waitingStates because that's the 
typical looping condition.  We need to keep polling for a new application 
report as long as the app is in the NEW, NEW_SAVING, or SUBMITTED state since 
those states indicate the RM hasn't finished accepting the app yet.  When it's 
not in one of the waiting states, we need to check if its one of the failed 
states to decide to throw rather than just return indicating it was successful.

> YarnClientImpl should check FAILED and KILLED state in submitApplication
> ------------------------------------------------------------------------
>                 Key: YARN-3131
>                 URL: https://issues.apache.org/jira/browse/YARN-3131
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Chang Li
>            Assignee: Chang Li
>         Attachments: yarn_3131_v1.patch, yarn_3131_v2.patch, 
> yarn_3131_v3.patch, yarn_3131_v4.patch, yarn_3131_v5.patch, yarn_3131_v6.patch
> Just run into a issue when submit a job into a non-existent queue and 
> YarnClient raise no exception. Though that job indeed get submitted 
> successfully and just failed immediately after, it will be better if 
> YarnClient can handle the immediate fail situation like YarnRunner does

This message was sent by Atlassian JIRA

Reply via email to