[HIVE-18959] Avoid creating extra pool of threads within LLAP

• Type: Task
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0
• Component/s: Druid integration

Environment

Kerberos Cluster

• Assignee: Slim Bouguerra Assign to me
• Reporter: Slim Bouguerra
• Created: 14/Mar/18 18:45
• Updated: 22/May/18 23:14
• Resolved: 16/Mar/18 20:35

Description

The current Druid-Kerberos-Http client is using an external single threaded pool to handle retry auth calls (eg when a cookie expire or other transient auth issues). 

First, this is not buying us anything since all the Druid Task is executed as one synchronous task.

Second, this can cause a major issue if an exception occurs that leads to shutting down the LLAP main thread with an exception like:

 org.apache.hadoop.hive.llap.daemon.impl.LlapDaemon: Thread Thread[KerberosHttpClient... threw an Exception. Shutting down now...

 Thus to fix this we should avoid using an external thread pool and handle retrying in a synchronous way.

Issue Links

• relates to: HIVE-18966 LLAP should not shut down when some random thread goes down