hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Harsh J (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HADOOP-9223) support specify config items through system property
Date Fri, 18 Jan 2013 11:22:13 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-9223?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13557150#comment-13557150

Harsh J commented on HADOOP-9223:

Got it and what you propose to add is good. The only worry of mine is that this may unintentionally
bloat up the config object size which inadvertently affects MR and other systems that do not
really require this feature.

Perhaps we can add it with a switch that has it turned off by default and only enabled if
an advanced user needs it?

Also, some tests would be good to have in the added patch, that also demonstrate order of
preference (I assume it is to be most preferred if the value is from a sys-prop rather than
a file), etc..

P.s. Fix Version is to be set only after it has been committed into a branch. Please use the
Target Version alone to track the version where you want it to be committed.
> support specify config items through system property
> ----------------------------------------------------
>                 Key: HADOOP-9223
>                 URL: https://issues.apache.org/jira/browse/HADOOP-9223
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: conf
>    Affects Versions: 2.0.0-alpha
>            Reporter: Zesheng Wu
>            Priority: Minor
>              Labels: configuration, hadoop
>         Attachments: HADOOP-9223.patch
>   Original Estimate: 24h
>  Remaining Estimate: 24h
> The current hadoop config items are mainly interpolated from the *-site.xml files. In
our production environment, we need a mechanism that can specify config items through system
properties, which is something like the gflags in system built with C++, it's really very
> The main purpose of this patch is to improve the convenience of hadoop systems, especially
when people do testing or perf tuning, which always need to modify the *-site.xml files
> If this patch is applied, then people can start hadoop programs in this way: java -cp
$class_path -Dhadoop.property.$name=$value $program

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

View raw message