[ 
https://issues.apache.org/jira/browse/YARN-4281?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Chang Li updated YARN-4281:
---------------------------
    Description: 
2.7 RM app page is broken by the cherry pick of YARN-3248 on 23/Sep. It broke 
the work around 2.7 patch of YARN-3544 to let it still use container report. 
Currently, our cluster's 2.7 RM app page is completely broken due to 500 error, 
which is caused by when user UGI is null, completed app can not retrieve its 
container report, and in that code path, it doesn't catch 
ContainerNotFoundException, but throw the exception, therefore cause the 500 
error.
 Running app is also broken because of the way it construct containerID by 
{code} "ContainerId.newContainerId(
              appAttemptReport.getApplicationAttemptId(), 1)" 
{code}, 
which will not include epoch number in ID, so it will also get 
ContainerNotFoundException and throw 500 error.
Propose to use the branch-2 version of YARN-3544, instead of the work around 
2.7 patch because branch 2 patch on 2.7 is no longer blocked.

  was:
2.7 RM app page is broken by the cherry pick of YARN-3248 on 23/Sep. It broke 
the work around 2.7 patch of YARN-3544 to let it still use container report. 
Currently, our cluster's 2.7 RM app page is completely broken due to 500 error, 
which is caused by when user UGI is null, completed app can not retrieve its 
container report, and in that code path, it doesn't catch 
ContainerNotFoundException, but throw the exception, therefore cause the 500 
error.
 Running app is also broken because of the way it construct containerID by 
{code} "ContainerId.newContainerId(
              appAttemptReport.getApplicationAttemptId(), 1)" 
{code}, 
which will not include epoch number, so it will also get 
ContainerNotFoundException and throw 500 error.
Propose to use the branch-2 version of YARN-3544, instead of the work around 
2.7 patch because branch 2 patch on 2.7 is no longer blocked.


> 2.7 RM app page is broken
> -------------------------
>
>                 Key: YARN-4281
>                 URL: https://issues.apache.org/jira/browse/YARN-4281
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Chang Li
>            Assignee: Chang Li
>             Fix For: 2.7.2
>
>         Attachments: YARN-4281-branch-2.7.patch, YARN-4281.2.7.modify.patch
>
>
> 2.7 RM app page is broken by the cherry pick of YARN-3248 on 23/Sep. It broke 
> the work around 2.7 patch of YARN-3544 to let it still use container report. 
> Currently, our cluster's 2.7 RM app page is completely broken due to 500 
> error, which is caused by when user UGI is null, completed app can not 
> retrieve its container report, and in that code path, it doesn't catch 
> ContainerNotFoundException, but throw the exception, therefore cause the 500 
> error.
>  Running app is also broken because of the way it construct containerID by 
> {code} "ContainerId.newContainerId(
>               appAttemptReport.getApplicationAttemptId(), 1)" 
> {code}, 
> which will not include epoch number in ID, so it will also get 
> ContainerNotFoundException and throw 500 error.
> Propose to use the branch-2 version of YARN-3544, instead of the work around 
> 2.7 patch because branch 2 patch on 2.7 is no longer blocked.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to