[GOBBLIN-310] Skip rerunning completed tasks on mapper reattempts

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.12.0
• Sprint: Apache Gobblin 170905
• Assignee: Hung Tran
• Reporter: Hung Tran
• Created: 09/Nov/17 17:49
• Updated: 03/Jan/18 06:35
• Resolved: 10/Nov/17 00:32

Description

Subsequent executions of a failed mapper will rerun completed tasks. This can result in duplicate data or errors due to collisions when publishing.

The state of completed mappers should be recorded and completed mappers should be skipped on subsequent attemps.