In a cluster where the Spark and/or Spark2 Livy servers and Zeppelin are installed and Kerberos is enabled, it is expected that that livy-conf/livy.superusers and livy2-conf/livy.superusers contain the principal name of the Zeppelin user. However, this value is not always set, depending on the order in which the services were installed, when Kerberos was enabled, and whether an Ambari or stack upgrade was involved. And if it is set, the value may be incorrect since the Kerberos descriptor assumes the Zeppelin principal is zeppelin-<clustername>
The solution is to move the logic to set the livy-conf/livy.superusers and livy2-conf/livy.superusers to the stack advisor to the appropriate value can be added as needed. Also while upgrading to Ambari 2.5.2, the value(s) should be fixed if necessary.