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

Posted to yarn-issues@hadoop.apache.org by "Jason Lowe (JIRA)" <ji...@apache.org> on 2014/11/25 22:46:13 UTC

[jira] [Commented] (YARN-2905) AggregatedLogsBlock page can infinitely loop if the aggregated log file is corrupted

    [ https://issues.apache.org/jira/browse/YARN-2905?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14225259#comment-14225259 ] 

Jason Lowe commented on YARN-2905:
----------------------------------

The code assumes skip will return an EOF indicator if we reach the end of the file too soon, but it returns 0 for that case.  Either AggregatedLogsBlock needs to check for this condition, just like IOUtils.skipFully does, or possibly ContainerLogsReader.skip should call IOUtils.skipFully internally.

> AggregatedLogsBlock page can infinitely loop if the aggregated log file is corrupted
> ------------------------------------------------------------------------------------
>
>                 Key: YARN-2905
>                 URL: https://issues.apache.org/jira/browse/YARN-2905
>             Project: Hadoop YARN
>          Issue Type: Bug
>    Affects Versions: 2.5.0
>            Reporter: Jason Lowe
>            Priority: Blocker
>
> If the AggregatedLogsBlock page tries to serve up a portion of a log file that has been corrupted (e.g.: like the case that was fixed by YARN-2724) then it can spin forever trying to seek to the targeted log segment.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)