[ 
https://issues.apache.org/jira/browse/HADOOP-4631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12653727#action_12653727
 ] 

Sharad Agarwal commented on HADOOP-4631:
----------------------------------------

bq. It would be good to get this into 0.20, no?
Apart from documentation, I think there are few things which needs to ironed 
out before this gets into:
- backward compatibility: should we keep hadoop-site.xml as well in conf and 
add it in Configuration class as a default resource before core-site.xml ? It 
should be sufficient for the backward compatibility, right?  Or keeping it will 
add to the confusion, lets live with the incompatible change ?
hadoop-default.xml, anyways people should not care about, so we don't need to 
keep it. Moving of defaults to src folder should not have any impact on 
applications.

- Applications can provide conf file via bin/hadoop -conf option. The provided 
conf file should override the defaults. In the current code, if hdfs or mapred 
code is linked later on, the default resource are added in the end of the 
Configuration#resources . This would make default override the user provided 
config file, which is not correct.  We can do :
{code}
private void loadResources(Properties properties,
                             ArrayList resources,
                             boolean quiet) {
    if(loadDefaults) {
      for (String resource : defaultResources) {
        loadResource(properties, resource, quiet);
      }
    }
    for (Object resource : resources) {
      loadResource(properties, resource, quiet);
    }
  }
{code}

- currently contrib projects like hod generates the hadoop-site.xml and uses it 
in various scripts. I am not sure what impact this jira have on such projects. 
Contrib test cases are very limited and not sufficient to catch problems. So 
wherever hadoop-site.xml is being referred in the code has to be analyzed 
closely.

> Split the default configurations into 3 parts
> ---------------------------------------------
>
>                 Key: HADOOP-4631
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4631
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: conf
>            Reporter: Owen O'Malley
>            Assignee: Sharad Agarwal
>             Fix For: 0.20.0
>
>         Attachments: 4631_v1.patch
>
>
> We need to split hadoop-default.xml into core-default.xml, hdfs-default.xml 
> and mapreduce-default.xml. That will enable us to split the project into 3 
> parts that have the defaults distributed with each component.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to