[FALCON-1810] Instance status response is KILLED instead of FAILED

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: trunk, 0.9
• Component/s: client, oozie, scheduler
• Assignee: Praveen Rachabattuni
• Reporter: Pragya Mittal
• Created: 02/Feb/16 12:17
• Updated: 08/Feb/16 12:24

Description

In workflow if one of the action fails then the intance status is KILLED instead of FAILED.

2016-02-02T11:34Z 	ProcessMultipleClustersTest-corp-3a97a54e	-	KILLED	2016-02-02T11:57Z	2016-02-02T11:57Z	-	http://8RPCG32.corp.inmobi.com:11000/oozie?job=0000938-160202121302340-oozie-oozi-W
actions:
 hdfscommands	OK	-
 aggregator	FAILED/KILLED	http://8RPCG32.corp.inmobi.com:8088/proxy/application_1454048976510_6891/

Such instances retry after some time (given retry is enabled)

This leads to inconsistencies like :
1. Manual kill will assume this instance to be killed and would not kill it. Although retry will kick off in future and the instance will run while user expects the instance to be killed forever.
2. Manual suspend will assume this instance to be killed and would not suspend it (KILLED instance cannot be suspended). Although retry will kick off in future and the instance will run while user expects the instance to be suspended until someone resumes it.