[FLINK-2287] Implement JobManager high availability

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.10.0
• Component/s: Runtime / Coordination
• Assignee: Unassigned
• Reporter: Ufuk Celebi
• Created: 29/Jun/15 08:51
• Updated: 28/Feb/19 14:01
• Resolved: 20/Oct/15 12:56

Description

The problem: The JobManager (JM) is a single point of failure. When it crashes, TaskManagers (TM) fail all running jobs and try to reconnect to the same JM. A failed JM looses all state and can not resume the running jobs; even if it recovers and the TMs reconnect.

Solution: implement JM fault tolerance/high availability by having multiple JM instances running with one as leader and the other(s) in standby. The exact coordination and state update protocol between JM, TM, and clients is covered in sub-tasks/issues.

Related Wiki: https://cwiki.apache.org/confluence/display/FLINK/JobManager+High+Availability

Issue Links

• is depended upon by: FLINK-2340 Provide standalone mode for web interface of the JobManager
• is superceded by: FLINK-7106 Make SubmittedJobGraphStore implementation configurable

Sub-Tasks

• Setup ZooKeeper for distributed coordination: Closed
• Use ZooKeeper to elect JobManager leader and send information to TaskManagers: Closed
• Refactor RPCs from within the ExecutionGraph: Closed
• Assign session IDs to JobManager and TaskManager messages: Closed
• Recover running jobs on JobManager failure: Resolved
• Add high availability support for Yarn: Resolved