hadoop-mapreduce-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ramkumar Vadali <ramkumar.vad...@gmail.com>
Subject Re: Problems to start Hadoop through Java Classes
Date Thu, 18 Nov 2010 18:48:00 GMT
Hi Michel

Any reason why you are not using the standard bin/start-mapred.sh? Why do
you need to create the JobTracker/TaskTracker yourself?

Ram

On Thu, Nov 18, 2010 at 4:49 AM, Michel Albonico <bsd.albonico@gmail.com>wrote:

> Hi everyone,
>
> I'm having problems to creating a Java class to start Hadoop Cluster...
>
> My NameNode e all datanodes are executing OK, I can create
> directories, put files and remove them too... (bin/hadoop dfs -mkdir
> /test)...
>
> But, I can not run a Job... The JobTracker and TaskTrackers apparently
> are executing ok...
>
> Can someone help me?
>
> Below is part of my code...
>
> To start the JobTracker:
> ...
> Configuration conf = getConfMR();
> job = new JobConf(conf);
> JobTracker jobtracker = JobTracker.startTracker(job);
> ...
>
> To start the TaskTrackers:
> ...
> Configuration conf = getConfMR();
> JobConf job = new JobConf(conf);
> TaskTracker tt = new TaskTracker(job);
> ...
>
> I trying to execute the PiEstimator, but it stop on the message: "Starting
> Job".
>
> All workers are started with Java classes, they aren't start with
> shelll command "bin/hadoop jobtracker/tasktracker..."
>
> --
> Michel Albonico
> http://www.inf.ufpr.br/michela
> (49) 9913-2766
> MSN: linucs@hotmail.com
> E-mail: albonico@inf.ufpr.br
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message