At least for version 2.6.3 and above the section "Running import script on kerberized cluster" is wrong. You don't need to provide any of the options (properties) indicated (except maybe the debug one If you want to have debug output) because they are automatically detected and included in the script. Also at least in 2.6.5 a direct execution of the script in a Kerberized cluster will fail because of the CLASSPATH generated into the script. I had to edit this replacing many single JAR files by a glob inside their parent folder in order for the command to run without error. If you have this problem see answer o "Atlas can't see Hive tables" question. ... View more