While starting my flume-agent I did use these options:- java -Xms512m -Xmx2048m. But that did not help. What should I tweak in flume-env.sh? Not sure?
I noticed that the source for fileChannel was included in flume 1.2. So I tried compiling that & using that jar; but got the exact same results with fileChannel - the process stopped after approx 9.5 hrs (after having written approx 9.5 million records).
Any help would be appreciated. Thanks.
--- On Sat, 11/3/12, Alexander Lorenz <[EMAIL PROTECTED]> wrote:
From: Alexander Lorenz <[EMAIL PROTECTED]>
Subject: Re: ExecSource tail -F stops after a while
To: [EMAIL PROTECTED]
Date: Saturday, November 3, 2012, 1:55 AM
Please see the WARNING ;)
Anyhow, this error indicates that you've to less memory configured for too many events. Tune flume-env.sh and tweak JAVA_OPTS into "-Xms100m -Xmx200m" or something else
FileChannel is available in from up to 1.3
On Nov 3, 2012, at 1:21 AM, Parag Hukeri <[EMAIL PROTECTED]> wrote:
> I tried to use the FileChannel; but am getting an exception -
> org.apache.flume.FlumeException: Unable to load channel type: file, class: org.apache.flume.channel.file.FileChannel
> Caused by: java.lang.ClassNotFoundException: org.apache.flume.channel.file.FileChannel
> I am using apache-flume-1.2.0. What version of flume is needed for FileChannel?
> --- On Fri, 11/2/12, Josh West <[EMAIL PROTECTED]> wrote:
> From: Josh West <[EMAIL PROTECTED]>
> Subject: Re: ExecSource tail -F stops after a while
> To: [EMAIL PROTECTED]
> Date: Friday, November 2, 2012, 1:48 PM
> This happened to me consistently when buffering tail -F contents into the Memory Channel. Try switching to the File Channel. This solved the issue for me.
> On 11/02/2012 07:15 PM, Parag Hukeri wrote:
> I am using apache-flume-1.2.0. I have set up an ExecSource that tails a log file & gets the entries from this log file. The log file size is approx 4 GB (approx 12 million lines written over a period of 12 hrs). As the process starts everything seems to work fine for approx 9.5 hrs. (approx 9.5 million lines of log data is successfully retrieved by ExecSource). It then stops reading the output from tail. I don't see any exceptions being thrown. I don't think the problem is with the actual tail command. I was able to successfully manually tail the log file & all 12 million records were retrieved by tail. Running on a Linux system. Not sure what the issue might be with ExecSource? Any help would be appreciated. Thanks.
> Josh West
> Lead Systems Administrator
> One.com, [EMAIL PROTECTED]
German Hadoop LinkedIn Group: http://goo.gl/N8pCF