[SAMZA-2711] Fix flaky TestZkLocalApplicationRunner

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: test
• Assignee: Unassigned
• Reporter: Jiazhou Gao
• Created: 23/Nov/21 23:47
• Updated: 06/Dec/21 22:53

Description

TestZkLocalApplicationRunner.shouldStopNewProcessorsJoiningGroupWhenNumContainersIsGreaterThanNumTasks asserts
A) JobModel generated before and after the addition of appRunner2 should be equal.
B) Second stream application(appRunner2) should not join the group and process any message.

Assertion A is good, but the test can not guarantee Assertion B.
The reason is that the test kill appRunner1 then kill appRunner2, but after appRunner1 got killed and before appRunner2 gets killed, ZkJobCoordinator could generate a new JobModel with appRunner 2
That leads to appRunner2 status change from New->Running->SuccessfulFinish
instead of New->UnsuccessfulFinish, which result in the test failure