hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Konstantinos Karanasos (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-6059) Update paused container state in the state store
Date Thu, 23 Feb 2017 01:21:44 GMT

    [ https://issues.apache.org/jira/browse/YARN-6059?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15879649#comment-15879649
] 

Konstantinos Karanasos commented on YARN-6059:
----------------------------------------------

Thanks for the latest patch, [~hrsharma]. Let's fix the following:
* In the {{RecoverPausedContainerLaunch}}, why is the {{CONTAINER_EXITED_WITH_FAILURE}} message
inside the finally statement? This should not only be called only when the {{notInterrupted}}
is true, but in all cases. Let's put it just outside (after) the finally statement, like we
do in the {{RecoveredContainerLaunch}}. Also, you need to set the retCode to be the output
of the exec.signalContainer method.
* The failing test seems related to your changes: "TestNMLeveldbStateStoreService.testContainerStorage:290
expected:<LAUNCHED> but was:<PAUSED>".
* You can find the checkstyle issues by clicking on the corresponding link in the report.
You will see a few of them, which need to be fixed (the only one I would not worry about is
regarding the number of parameters). Let me know if any of them is not self-explanatory. Pasting
here too: https://builds.apache.org/job/PreCommit-YARN-Build/14990/artifact/patchprocess/diff-checkstyle-hadoop-yarn-project_hadoop-yarn_hadoop-yarn-server_hadoop-yarn-server-nodemanager.txt.
Also address the warning shown in the output of the javadoc result.

Re: the ifs in the {{NMLeveldbStateStoreService}}: when a new recovery state is stored for
a container, a new entry is added to the StateStore. Then we iterate over all of them. The
ifs that you added essentially check the previous entry found. So the code you added means
that you are allowed to get a PAUSED recovery container status immediately after a LAUNCHED,
QUEUED or REQUESTED status. If all these "transitions" are possible, then your code is correct.

> Update paused container state in the state store
> ------------------------------------------------
>
>                 Key: YARN-6059
>                 URL: https://issues.apache.org/jira/browse/YARN-6059
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>            Reporter: Hitesh Sharma
>            Assignee: Hitesh Sharma
>         Attachments: YARN-5216-YARN-6059.001.patch, YARN-6059-YARN-5972.001.patch, YARN-6059-YARN-5972.002.patch,
YARN-6059-YARN-5972.003.patch, YARN-6059-YARN-5972.004.patch, YARN-6059-YARN-5972.005.patch,
YARN-6059-YARN-5972.006.patch, YARN-6059-YARN-5972.007.patch, YARN-6059-YARN-5972.008.patch
>
>




--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org


Mime
View raw message