> I noticed that the "bin/hadoop jar" command doesn't add the jar being
> executed to the classpath. Is this deliberate and what is the reasoning? The
> result is that resources in the jar are not accessible from the system class
> loader. Rather they are only available from the thread context class loader
> and the class loader of the main class.
In map and reduce tasks' jvm, job libraries are added to the system
classloader. However for others only framework code is present in system
classloader. If you are seeing this as a problem in your client side code, you
can use Configuration#getClassByName(String name) instead of Class.forName()
for loading your job related classes.
