I would not advise to do this - only if you have no other options and you are sure there are classpath problems (this suggests you had that - likely not having set up the Spark service dependency with Hive). Always check the HS2 logs first what is your problem. With including all the hadoop jars to SPARK_DIST_CLASSPATH you submit and upload lots of jars to the container classpath unnecessarily, which will slow down the job submission time.  ... View more