You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-dev@hadoop.apache.org by "Piotr Kołaczkowski (JIRA)" <ji...@apache.org> on 2013/04/04 12:17:15 UTC

[jira] [Created] (MAPREDUCE-5126) Add possibility to set a custom system classloader for mapred child processes, separate from mapred.child.java.opts

Piotr Kołaczkowski created MAPREDUCE-5126:
---------------------------------------------

             Summary: Add possibility to set a custom system classloader for mapred child processes, separate from mapred.child.java.opts
                 Key: MAPREDUCE-5126
                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5126
             Project: Hadoop Map/Reduce
          Issue Type: New Feature
            Reporter: Piotr Kołaczkowski
            Priority: Minor


Some third party frameworks / systems based on Hadoop might want to set a custom classloader for loading classes of their jobs to better resolve conflicts with their libraries. 

While it is possible to set a custom classloader using the mapred.child.java.opts, this field is often overriden by users in their job configuration. So in order to change e.g. heap sizes the user would need to remember also to include the custom classloader property from the framework-defaults or otherwise he would break the framework.

This small patch introduces another parameter: mapred.child.java.class.loader that allows to set the classloader separately. This gives custom frameworks built on top of Hadoop more flexibility to supply their own classloader, without need to force users to adjust any settings.

--
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