[
https://issues.apache.org/jira/browse/HADOOP-9642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13682277#comment-13682277
]
Daryn Sharp commented on HADOOP-9642:
-------------------------------------
Generally sounds reasonable to me. Although you'd probably want additional
syntax such as $\{ENV:default-value\} if the env isn't defined.
One concern is introduction of non-deterministic behavior. An example being
the "works for me" scenario where user A and user B are puzzled why each gets
different results from the same command because one of them forgot to set
additional envs beyond HADOOP_HOME.
Along those lines, it does introduce additional logistics if server process
configs reference envs. It becomes easier to make mistakes if there's a mix of
interactive/non-interactive shells commands to start processes and the most
appropriate rc (profile/bashrc/etc) isn't updated. Although it could be argued
the admin brought the pain upon himself...
> Configuration to resolve environment variables via ${env.VARIABLE} references
> -----------------------------------------------------------------------------
>
> Key: HADOOP-9642
> URL: https://issues.apache.org/jira/browse/HADOOP-9642
> Project: Hadoop Common
> Issue Type: Improvement
> Components: conf
> Affects Versions: 3.0.0, 2.1.0-beta
> Reporter: Steve Loughran
> Priority: Minor
>
> We should be able to get env variables from Configuration files, rather than
> just system properties. I propose using the traditional {{env}} prefix
> {{${env.PATH}}} to make it immediately clear to people reading a conf file
> that it's an env variable -and to avoid any confusion with system properties
> and existing configuration properties.
--
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
