You are viewing a plain text version of this content. The canonical link for it is here.

Posted to common-dev@hadoop.apache.org by "Sharad Agarwal (JIRA)" <ji...@apache.org> on 2009/05/05 08:03:30 UTC

[jira] Commented: (HADOOP-5710) Counter MAP_INPUT_BYTES missing from new mapreduce api.

    [ https://issues.apache.org/jira/browse/HADOOP-5710?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12705915#action_12705915 ] 

Sharad Agarwal commented on HADOOP-5710:
----------------------------------------

I propose that we remove MAP_INPUT_BYTES from the framework counters as it is specific to file based input formats only. With new api, the framework no more calculates this. It is getting calculated in lib. So we can move  MAP_INPUT_BYTES to FileInputFormat class. That seems to be its more logical place.

> Counter MAP_INPUT_BYTES missing from new mapreduce api.
> -------------------------------------------------------
>
>                 Key: HADOOP-5710
>                 URL: https://issues.apache.org/jira/browse/HADOOP-5710
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: mapred
>            Reporter: Amareshwari Sriramadasu
>            Assignee: Amareshwari Sriramadasu
>             Fix For: 0.21.0
>
>         Attachments: patch-5710.txt
>
>
> MapTask, with current new mapreduce api, does not maintain MAP_INPUT_BYTES counter. Since RecordReader doesnot have getPos() api, it is not possible for the map task to maintain. Individual record readers (LineRecordReader, SequenceFileRecordReader) could be modified to maintain the same.

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