hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Lars George (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-1791) Timeout in IndexRecordWriter
Date Tue, 25 Aug 2009 17:10:59 GMT

    [ https://issues.apache.org/jira/browse/HBASE-1791?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12747490#action_12747490

Lars George commented on HBASE-1791:


I attribute this to the new Hadoop mapreduce package and the new context class layout. There
is an old documentation that refers to setting the status to report progress but I am not
surprised that the new classes rather use the proper progress() call. The unit test is too
small to notice this change, so good that Bradford found it!

> Timeout in IndexRecordWriter
> ----------------------------
>                 Key: HBASE-1791
>                 URL: https://issues.apache.org/jira/browse/HBASE-1791
>             Project: Hadoop HBase
>          Issue Type: Bug
>    Affects Versions: 0.20.0, 0.20.1, 0.21.0
>         Environment: 19 HBase nodes, 8 cores, 8 GB RAM, CentOS
>            Reporter: Bradford Stephens
>            Priority: Minor
>             Fix For: 0.20.0
>         Attachments: hbase-1791.patch
>   Original Estimate: 3h
>  Remaining Estimate: 3h
> A MapReduce job to generate Lucene Indexes from HBase will fail on sufficiently large
tables. After the indexing finished, the close() method of IndexRecordWriter is called.  The
 writer.optimize() call in this method can take many minutes, forcing most MapReduce tasks
to timeout. There is a HeartBeatsThread, but it does not seem to send progress updates. 
> A suggested fix may be to add context.progress(); in the HeardbeatsThread run() method,
after the context.setStatus call. Not sure why context.setStatus is not "good enough". 

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

View raw message