[STORM-3656] Change handling of Hadoop TGT renewal exception

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.2.0
• Fix Version/s: 2.3.0
• Component/s: storm-client
• Assignee: Aaron Gresch
• Reporter: Aaron Gresch
• Created: 24/Jun/20 14:23
• Updated: 01/Jul/20 14:05
• Resolved: 25/Jun/20 18:16

Description

STORM-3606 identified an issue where Hadoop's TGT auto renewal thread causes an exception and worker restart.  The fix involved a lot of reflection calls to emulate the Hadoop code while avoiding launching this thread.

 

It's possible Hadoop could change their code, causing this reflection to fail.   To handle this case, we could instead allow Hadoop to launch this autorenewal thread, and have the worker catch the specific NPE from Hadoop in the exception handler.  

Issue Links

• relates to: STORM-3606 AutoTGT shouldn't invoke TGT renewal thread (from UserGroupInformation.loginUserFromSubject)