03-06-2018 02:53 AM - last edited on 03-06-2018 05:42 AM by cjervis
I am trying to install Zeppelin on a CDH installation that has Kerberos and Sentry configured.
So far I am facing the following problems:
1. Zeppelin submits spark jobs to the cluster using the user that started the service (zeppelin-daemon.sh). In Zeppelin interpreters, there are some configuration for Kerberos auth but haven't got them work. As a workaround, I can make a configuration to the edge nodes so that each user starts his own instance of Zeppelin in a different port. Can you advice for anything better, is there any other solution?
2. I edited the %jdbc interpreter to define hive connection (hive.url, hive.driver), but I get errors about proxy and anonymous user. Is there any advice to make this work with Kerberos?