I am having some troubles in understanding how the whole stuff works.. Compiling with ant works ok and I am able to compile a jar which is afterwards deployed to the cluster. On the cluster I've set the HADOOP_CLASSPATH variable to point just to jar files in the lib folder ($HD_HOME/lib/*.jar), where I put the new compiled hadoop-core-myversion.jar.
Before deploying I guarantee that in the $HD_HOME folder and $HD_HOME/lib there are no previous version of hadoop-core-xxx.jar or core-3.3.1.jar . The problem is that I suspect that hadoop is picking the wrong hadoop-core jars so I am interested how the whole mechanism works and what is the purpose of the $HD_HOME/share/hadoop folder where I can locate other hadoop-core jars and which is included in the classpath in hadoop-env.sh? My last question is what is the easiest way to see that your build is up and running? Maybe from the release tag in JT? Thanks you..