@Vjeran , running hadoop command over java adds the Hadoop libraries and their dependencies to the classpath and brings in all the Hadoop configuration details, the former does not . There's an excellent book on Hadoop call the Hadoop the Definitive Guide by Tom White, everything is explained in that book.
Regards Niranjan Singh On Sun, Apr 14, 2013 at 1:49 AM, Vjeran Marcinko < [email protected]> wrote: > Hello,**** > > ** ** > > I am complete Hadoop and MR newbiew, so please help me with following…**** > > ** ** > > I can see that primary way to submit Hadoop MR job is via following > command (wordcount example):**** > > ** ** > > hadoop jar wordcount.jar org.mycompany.WordCount**** > > ** ** > > **1. **Although, looking at all MR examples out there, I see this > „hadooo jar“ command used for submitting MR jobs, but actually it has > nothing specific with MR job submissions, it just calls static „main“ > method, similar to plain „java“ command, and this main method can just > print „Hello world“ on console and have no business with MR framework, > right?**** > > **2. **If above is true, and I assume it is, what is the difference > with using plain java –jar in place of „hadoop jar“? Both call static > „main“ method, and this method could submit MR job via JobClient class, > right?**** > > **3. **On Hadoop docs ( > http://hadoop.apache.org/docs/r1.0.4/commands_manual.html#Generic+Options ), > I see that –libjars options is only present in „hadoop job“ command, but > not in „hadoop jar“, and later is usually used (for some unknown reason > because former also has –submit ooption?) for submitting jobs, so my > question is does that mean that when using „jar“ command I should only > priovde thrid-party libs via „fat jar“?**** > > ** ** > > Regards,**** > > Vjeran **** >
