You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-issues@hadoop.apache.org by "Jason Dere (JIRA)" <ji...@apache.org> on 2014/04/22 22:46:17 UTC

[jira] [Commented] (MAPREDUCE-5853) ChecksumFileSystem.getContentSummary() including contents for crc files

    [ https://issues.apache.org/jira/browse/MAPREDUCE-5853?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13977383#comment-13977383 ] 

Jason Dere commented on MAPREDUCE-5853:
---------------------------------------

It looks like FileSystem's implementation of getContentSummary() really just uses getFileStatus()/listStatus().  If we get rid of the overridden version of getContentSummary() in FilterFileSystem and just fall back to the FileSystem implementation, would this work correctly, since FilterFileSystem does have overridden versions of getFileStatus()/listStatus()?

> ChecksumFileSystem.getContentSummary() including contents for crc files 
> ------------------------------------------------------------------------
>
>                 Key: MAPREDUCE-5853
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5853
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>            Reporter: Jason Dere
>
> Trying to track down some differences in Hive statistics between hadoop-1/hadoop-2.  It looks like although ChecksumFileSystem.listStatus() filters out CRC files, getContentSummary() falls back to using the FilterFileSystem.getContentSummary() implementation, which calls fs.getContentSummary().  The underlying fs may not have the same filters as the ChecksumFileSystem and so the CRC files can get included in the content summary.



--
This message was sent by Atlassian JIRA
(v6.2#6252)