Mailing-List: contact issues-help@tez.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@tez.incubator.apache.org
Date: Wed, 28 Aug 2013 03:46:10 +0000 (UTC)
From: "joeyli (JIRA)" <jira@apache.org>
To: issues@tez.incubator.apache.org
Message-ID: <JIRA.12661643.1375683979778.44979.1377661570074@arcas>
In-Reply-To: <JIRA.12661643.1375683979778@arcas>
References: <JIRA.12661643.1375683979778@arcas>
Subject: [jira] [Commented] (TEZ-336) AMContainerImpl are stored even after
 container has completed
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/TEZ-336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13752080#comment-13752080 ] 

joeyli commented on TEZ-336:
----------------------------

Looks like the most heavy object ContainerLaunchContext has already been discarded in LaunchRequestTransition.
Do we still need to do this jira?
{code:title=AMContainerImpl.java|borderStyle=solid}
      container.clc = AMContainerHelpers.createContainerLaunchContext(
          container.appContext.getApplicationACLs(),
          container.getContainerId(),
          containerContext.getLocalResources(),
          containerContext.getEnvironment(),
          containerContext.getJavaOpts(),
          container.taskAttemptListener, containerContext.getCredentials(),
          event.shouldProfile(), container.appContext);

      // Registering now, so that in case of delayed NM response, the child
      // task is not told to die since the TAL does not know about the container.
      container.registerWithTAListener();
      container.sendStartRequestToNM();
      LOG.info("Sending Launch Request for Container with id: " +
          container.container.getId());
      // Forget about the clc to save resources. At some point, part of the clc
      // info may need to be exposed to the scheduler to figure out whether a
      // container can be used for a specific TaskAttempt.
      container.clc = null;
 {code}    
                
> AMContainerImpl are stored even after container has completed
> -------------------------------------------------------------
>
>                 Key: TEZ-336
>                 URL: https://issues.apache.org/jira/browse/TEZ-336
>             Project: Apache Tez
>          Issue Type: Bug
>            Reporter: Bikas Saha
>            Assignee: joeyli
>              Labels: TEZ-0.2.0
>             Fix For: 0.2.0
>
>         Attachments: TEZ-336.3.patch
>
>
> For a long running job thats allocated 100's of 1000's of containers this can be a memory usage issue. Even if AMContainerImpls are small they contain full copies of container launch context objects that look like heavy weights with multiple maps of local resources, env etc.

--
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