[MAPREDUCE-4951] Container preemption interpreted as task failure

• Type: Sub-task
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.0.2-alpha
• Fix Version/s: 2.1.0-beta
• Component/s: applicationmaster, mr-am, mrv2
• Hadoop Flags: Reviewed
• Assignee: Sandy Ryza
• Reporter: Sandy Ryza
• Created: 21/Jan/13 03:00
• Updated: 21/May/14 21:30
• Resolved: 21/Feb/13 12:05

Description

When YARN reports a completed container to the MR AM, it always interprets it as a failure. This can lead to a job failing because too many of its tasks failed, when in fact they only failed because the scheduler preempted them.

MR needs to recognize the special exit code value of -100 and interpret it as a container being killed instead of a container failure.

Issue Links

• is depended upon by: MAPREDUCE-5900 Container preemption interpreted as task failures and eventually job failures
• is related to: YARN-352 Inconsistent picture of how a container was killed when querying RM and NM in case of preemption
• is related to: MAPREDUCE-4788 Job are marking as FAILED even if there are no failed tasks in it