[
https://issues.apache.org/jira/browse/HIVE-14888?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16809957#comment-16809957
]
David McGinnis commented on HIVE-14888:
---------------------------------------
Looks like we are safe in the assumption that the login will either be simple
or kerberos. Anything else is rejected because the login type is not set. Error
thrown is from the line below:
[https://github.com/apache/hadoop/blob/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/security/UserGroupInformation.java#L1474]
Once the build issues in the pre-commit build are resolved, I'll resubmit the
patch and continue with this submission.
> SparkClientImpl checks for "kerberos" string in hiveconf only when
> determining whether to use keytab file.
> ----------------------------------------------------------------------------------------------------------
>
> Key: HIVE-14888
> URL: https://issues.apache.org/jira/browse/HIVE-14888
> Project: Hive
> Issue Type: Bug
> Components: spark-branch
> Affects Versions: 2.1.0
> Reporter: Thomas Rega
> Assignee: David McGinnis
> Priority: Major
> Attachments: HIVE-14888.1-spark.patch, HIVE-14888.2.patch
>
> Original Estimate: 5m
> Remaining Estimate: 5m
>
> The SparkClientImpl will only provide a principal and keytab argument if the
> HADOOP_SECURITY_AUTHENTICATION in hive conf is set to "kerberos". This will
> not work on clusters with Hadoop security enabled that are not configured as
> "kerberos", for example, a cluster which is configured for "ldap".
> The solution is to call UserGroupInformation.isSecurityEnabled() instead.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
