[SAMZA-1302] Improve Zookeeper error handling

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 0.13.1
• Assignee: Shanthoosh Venkataraman
• Reporter: Shanthoosh Venkataraman
• Created: 19/May/17 23:30
• Updated: 20/Jul/17 18:45
• Resolved: 20/Jul/17 18:45

Description

In samza standalone, Zookeeper is used for accomplishing the following.

A) For storing and retrieving the job model.
B) Leader election amongst available stream processors.
C) Propagating job model expiry to stream processors.
D) Barrier for job model version upgrade.

ZooKeeper Exceptions from all of the above scenarios are not propagated back to LocalApplicationRunner.waitForFinish.

Incase of exception, an application will wait indefinitely if it is not bubbled up correctly.

Potentially could cause stalled processors and duplicate stream processors.

Issue Links

• is related to: SAMZA-1064 Standalone Samza with Zookeeper for Coordination