[MAPREDUCE-6801] Fix flaky TestKill.testKillJob()

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0-alpha1
• Fix Version/s: 2.8.0, 3.0.0-alpha2
• Component/s: mrv2
• Target Version/s: 3.0.0-alpha2
• Hadoop Flags: Reviewed
• Assignee: Haibo Chen
• Reporter: Haibo Chen
• Created: 26/Oct/16 15:54
• Updated: 06/Jan/17 08:16
• Resolved: 18/Nov/16 20:10

Description

TestKill.testKillJob often fails for the same reason with the following error message:

1 tests failed.
FAILED:  org.apache.hadoop.mapreduce.v2.app.TestKill.testKillJob

Error Message:
Task state not correct expected:<KILLED> but was:<NEW/SCHEDULED/RUNNING>

Stack Trace:
java.lang.AssertionError: Task state not correct expected:<KILLED> but was:<NEW/SCHEDULED/RUNNING>
	at org.junit.Assert.fail(Assert.java:88)
	at org.junit.Assert.failNotEquals(Assert.java:743)
	at org.junit.Assert.assertEquals(Assert.java:118)
	at org.apache.hadoop.mapreduce.v2.app.TestKill.testKillJob(TestKill.java:84)

The root cause is that when the job is in KILLED state from an external view, TaskKillEvents and TaskAttemptKillEvents placed on the event loop queue may not have been processed by the dispatcher thread.

Issue Links

• is duplicated by: MAPREDUCE-6802 TestKill.testKillJob() fails intermittently on Power