The most important thing in job.jar is the Mapper class and Reducer class. Because the tasktracker do not have these two classes. So hadoop will help you create the job.jar and then transfer it to tasktracker. This is hadoop's philosophy: move computation rather than moving data.
On Mon, Mar 15, 2010 at 11:47 PM, psdc1978 <[email protected]> wrote: > Hi, > > When I'm running an hadoop example: > $ bin/hadoop jar build/hadoop-0.20.2-dev-examples.jar wordcount gutenberg > gutenberg-output > > I've noticed that it's created a job.jar file with classes of the > hadoop-0.20.2-dev-examples jar. Why job.jar is created? > > In the "Hadoop-Definitive Guide" book, it exists several examples that use > the job.jar. For example: > $ hadoop jar job.jar v3.MaxTemperatureDriver -conf conf/hadoop-cluster.xml > input/ncdc/all max-temp > > Is it possible to create a job.jar file in the bash command line? > > > PS: > I've put some posts in the MR mailing list that weren't answered. These > posts can be viewed by other users? > > > Regards > -- > Pedro > -- Best Regards Jeff Zhang
