Attention should be given to your jar(s) which happen to be of newer version of the same jar(s) hadoop uses out of box. The functionality in the newer version jar(s) won't be picked up by task trackers.
On Fri, Aug 27, 2010 at 3:23 PM, Benjamin Campbell <[email protected] > wrote: > On Fri, Aug 27, 2010 at 3:35 PM, Venkatesh <[email protected]> wrote: > > > > > > > > > The mapreduce job code I have (java app) depends on other libraries. It > > runs fine when the job is run locally > > & but when I'm running on a true distributed setup..it's failing on > > dependencies..Do I have to put all the > > libraies, propertty files (dependent) of my application in > HADOOP_CLASSPATH > > ..for the mapreduce to run in a cluster? > > > > thanks > > venkatesh > > > > > > > I believe you can pass additional jar libraries via the -libjars switch. > > -- > Thanks > > Ben Campbell >
