You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flume.apache.org by "Jarek Jarcec Cecho (Updated) (JIRA)" <ji...@apache.org> on 2011/11/25 08:36:40 UTC

[jira] [Updated] (FLUME-861) AvroSource is failing on ClosedChannelException

     [ https://issues.apache.org/jira/browse/FLUME-861?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jarek Jarcec Cecho updated FLUME-861:
-------------------------------------

    Attachment: hddev02.properties
                hddev01.properties
                hddev02.log
                hddev01.log
    
> AvroSource is failing on ClosedChannelException
> -----------------------------------------------
>
>                 Key: FLUME-861
>                 URL: https://issues.apache.org/jira/browse/FLUME-861
>             Project: Flume
>          Issue Type: Bug
>    Affects Versions: NG alpha 2
>            Reporter: Jarek Jarcec Cecho
>         Attachments: hddev01.log, hddev01.properties, hddev02.log, hddev02.properties
>
>
> I've noticed that flume-ng is not reconnecting avro source/sink combination after restart of the sink part.
> Steps to reproduce:
> 1) Run flume-ng on two nodes
> 2) Start sending data
> 3) Turn off sink instance (I used CTRL+C)
> 4) Turn it on again
> 5) Avro source part will fail on:
> 2011-11-25 02:22:23,720 (Thread-1) [WARN - org.apache.avro.ipc.NettyTransceiver$NettyClientAvroHandler.exceptionCaught(NettyTransceiver.java:257)] Unexpected exception from downstream.
> java.nio.channels.ClosedChannelException
> (entire error message is present in hddev01.log)
> I've used two centos boxes where first was acting as an agent, reading local log file using ExecSource. Second machine was acting as a collector, waiting for input events and storing them on HDFS. Both machines were connected using AVRO sink+source combination. You can find configuration files for both machines with their logs as well attached to this JIRA bug. 
>  I've executed both flume-ng instances using following commands: 
>  ./bin/flume-ng node --conf conf/ --classpath flume-ng.jar --f conf/configuration.properties -n hddev01 > hddev01.log 2>&1 
>  ./bin/flume-ng node --conf conf/ --classpath flume-ng.jar --f conf/configuration.properties -n hddev02 > hddev02.log 2>&1 
>  Input file was created using following small bash script (it was executed after flume-ng was successfully started): 
>  for i in `seq -w 01 180`; do echo $i; echo Yoda-$i >> /var/log/jarcec; sleep 1s; done 
>  Please note that I applied patch from FLUME-858 in order to get DataStream file type working for this test, however I was observing same behavior without it as well (e.g. using default sequence file)

--
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