[FLINK-5971] JobLeaderIdService should time out registered jobs

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.3.0
• Fix Version/s: 1.3.0
• Component/s: Runtime / Coordination
• Labels: flip-6
• Assignee: Till Rohrmann
• Reporter: Till Rohrmann
• Created: 06/Mar/17 15:25
• Updated: 13/Mar/17 14:05
• Resolved: 13/Mar/17 14:04

Description

The JobLeaderIdService has no mechanism to time out inactive jobs. At the moment it relies on the RunningJobsRegistry which only gives a heuristic answer.

We should remove the RunningJobsRegistry and register instead a timeout for each job which does not have a job leader associated.

Issue Links

• links to: GitHub Pull Request #3488