[FLINK-9439] DispatcherTest#testJobRecovery dead locks

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.5.0
• Fix Version/s: 1.5.2, 1.6.0
• Component/s: Runtime / Coordination
• Labels: test-stability
• Assignee: Till Rohrmann
• Reporter: Piotr Nowojski
• Created: 25/May/18 12:03
• Updated: 11/Jul/18 23:29
• Resolved: 11/Jul/18 23:29

Description

Rarely DispatcherTest#testJobRecovery dead locks on master. Example deadlock in my pull request:

https://api.travis-ci.org/v3/job/383147736/log.txt

afterwards I have stripped down `flink-runtime`, looped on travis this test and it also dead locks on clean master branch:

https://github.com/pnowojski/flink/commits/loop-runtime-master

(note, that looped versions sometimes also fails with an exception from setUp: {{ akka.actor.InvalidActorNameException: actor name [dispatcher_testJobRecovery] is not unique! }} but this might be unrelated).

Example failed build:

https://travis-ci.org/pnowojski/flink/builds/383650106

Issue Links

• links to: Github Pull Request #6279