[
https://issues.apache.org/jira/browse/YARN-673?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13655073#comment-13655073
]
Karthik Kambatla commented on YARN-673:
---------------------------------------
While I understand that having two different default values in the default.xml
and code could be error-prone, I believe we should be careful about removing
yarn-default.xml altogether. Particularly because moving it further from the
code would make it even harder to ensure the documentation reflects the same
default as the code. Also, it is a lot easier to grep through default.xml for
defaults than documentation.
> Remove yarn-default.xml
> -----------------------
>
> Key: YARN-673
> URL: https://issues.apache.org/jira/browse/YARN-673
> Project: Hadoop YARN
> Issue Type: Improvement
> Reporter: Siddharth Seth
>
> The default configuration files serve 2 purposes
> 1. Documenting available config parameters, and their default values.
> 2. Specifying default values for these parameters.
> An xml file hidden inside a jar is not necessarily the best way to document
> parameters. This could be moved into the documentation itself.
> Default values already exist in code for most parameters. There's no need to
> specify them in two places. We need to make sure defaults exist for all
> parameters before attempting this.
> Having default configuration files just bloats job conf files; over 450
> parameters, out of which <20 are likely job specific params. JobConf files
> end up being rather big, and the memory footprint of the conf object is large
> (>300KB last I checked).
--
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
