[ 
https://issues.apache.org/jira/browse/HADOOP-8573?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13408061#comment-13408061
 ] 

Robert Joseph Evans commented on HADOOP-8573:
---------------------------------------------

The only real way to fix this is to cache the contents of the InputStream, or a 
parsed version of it.  The question is how do we reduce the potential memory 
impact of this?  We could change configuration so that it is a list of 
properties instead of a single combined properties object, but this is a very 
large change to fix this issue. Because of this I am inclined to not really 
worry about the memory impact right now.  This feature is not that commonly 
used, and most configuration objects tend to be shared a lot so I don't see the 
impact being that large.
                
> Configuration tries to read from an inputstream resource multiple times. 
> -------------------------------------------------------------------------
>
>                 Key: HADOOP-8573
>                 URL: https://issues.apache.org/jira/browse/HADOOP-8573
>             Project: Hadoop Common
>          Issue Type: Bug
>          Components: conf
>    Affects Versions: 1.0.2, 0.23.3, 2.0.1-alpha, 3.0.0
>            Reporter: Robert Joseph Evans
>            Assignee: Robert Joseph Evans
>
> If someone calls Configuration.addResource(InputStream) and then 
> reloadConfiguration is called for any reason, Configruation will try to 
> reread the contents of the InputStream, after it has already closed it.
> This never showed up in 1.0 because the framework itself does not call 
> addResource with an InputStream, and typically by the time user code starts 
> running that might call this, all of the default and site resources have 
> already been loaded.
> In 0.23 mapreduce is now a client library, and mapred-site.xml and 
> mapred-default.xml are loaded much later in the process.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to