Hi @BrianChan, This is a known CM issue, which incurs a bad spark-defaults.conf generated by the CM Agent after some users (or applications) logged the hosts as root and kinited as some user (for example, hdfs), and left that ticket cache around.   To avoid the issue, do the following: Navigate to Cloudera Manager > Spark > Configuration. Add the following in the Spark Client Advanced Configuration Snippet (Safety Valve) for spark-conf/spark-defaults.conf. Ensure a correct value is set: spark.eventLog.dir=hdfs://nameserviceXYZ/user/spark/applicationHistory Deploy the client configuration   If this helps to resolve the issue, please accept this as a solution. Thanks. ... View more