[jira] [Created] (ZEPPELIN-993) Two Spark interpreters with different kerberos authentication

["Partha Pratim Ghosh (JIRA)" <ji...@apache.org>]

Partha Pratim Ghosh created ZEPPELIN-993:
--------------------------------------------

             Summary: Two Spark interpreters with different kerberos authentication
                 Key: ZEPPELIN-993
                 URL: https://issues.apache.org/jira/browse/ZEPPELIN-993
             Project: Zeppelin
          Issue Type: Bug
            Reporter: Partha Pratim Ghosh


Spark authorization can be brought about by kerberos authentication. Now if we want to have two groups of users working on Spark contexts authorized by different ids, how to go about doing that?

If we create 2 spark interpreters and have 2 kinits with different ids before starting the interpreters then both the Spark contexts start fine. However, once started, the Spark context started by the latest kinit only runs. The other one does not get kerberos support and fails. Its of little use to do kinit before every Spark paragraph run as 2 paragraphs can run in parallel and the ensuing race condition may (shall) lead to failure in many cases. 

Any idea then how 2 Spark interpreters running on 2 different kerberos ids can be maintained in same zeppelin instance, as this shall be necessary to maintain Spark authorization from zeppelin instance? alternatively,  any idea how Spark authorization can work with zeppelin.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)