Troubleshooting Kerberos Errors

This topic describes some common Kerberos issues and their recommended solutions.

HDFS commands fail with Kerberos errors even though Kerberos authentication is successful in the web application

If Kerberos authentication is successful in the web application, and the output of klist in the engine reveals a valid-looking TGT, but commands such as hdfs dfs -ls / still fail with a Kerberos error, it is possible that your cluster is missing the Java Cryptography Extension (JCE) Unlimited Strength Jurisdiction Policy File. The JCE policy file is required when Red Hat uses AES-256 encryption. This library should be installed on each cluster host and will live under $JAVA_HOME. For more information, see Using AES-256 Encryption.

Cannot find renewable Kerberos TGT

Cloudera Machine Learning runs its own Kerberos TGT renewer which produces non-renewable TGT. However, this confuses Hadoop's renewer which looks for renewable TGTs. If the Spark 2 logging level is set to WARN or lower, you may see exceptions such as:
16/12/24 16:38:40 WARN security.UserGroupInformation: Exception encountered while running the renewal command. Aborting renew thread. ExitCodeException exitCode=1: kinit: Resource temporarily unavailable while renewing credentials

16/12/24 16:41:23 WARN security.UserGroupInformation: PriviledgedActionException as:user@CLOUDERA.LOCAL (auth:KERBEROS) GSS initiate failed [Caused by GSSException: No valid credentials provided (Mechanism level: Failed to find any Kerberos tgt)]

This is not a bug. Spark 2 workloads will not be affected by this. Access to Kerberized resources should also work as expected.