[FLINK-10311] HA end-to-end/Jepsen tests for standby Dispatchers

• Type: Improvement
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.7.0
• Fix Version/s: 1.6.2, 1.7.0
• Component/s: Tests
• Labels: pull-request-available
• Assignee: Gary Yao
• Reporter: Till Rohrmann
• Created: 10/Sep/18 07:57
• Updated: 29/Sep/18 11:05
• Resolved: 29/Sep/18 11:05

Description

We should add end-to-end or Jepsen tests to verify the HA behaviour when using multiple standby Dispatchers. In particular, we should verify that jobs get properly cleaned up after they finished successfully (see FLINK-10255 and FLINK-10011):

1. Test that a standby Dispatcher does not affect job execution and resource cleanup
2. Test that a standby Dispatcher recovers all submitted jobs after the leader loses leadership

Issue Links

• links to: GitHub Pull Request #6712