[MAPREDUCE-6689] MapReduce job can infinitely increase number of reducer resource requests

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Fix Version/s: 2.8.0, 2.7.3, 2.6.5, 3.0.0-alpha1
• Target Version/s: 2.8.0, 2.7.3, 2.6.4
• Hadoop Flags: Reviewed
• Assignee: Wangda Tan
• Reporter: Wangda Tan
• Created: 05/May/16 00:38
• Updated: 30/Aug/16 01:13
• Resolved: 06/May/16 22:41

Description

We have seen this issue from one of our clusters: when running terasort map-reduce job, some mappers failed after reducer started, and then MR AM tries to preempt reducers to schedule these failed mappers.

After that, MR AM enters an infinite loop, for every RMContainerAllocator#heartbeat run, it:

• In preemptReducesIfNeeded, it cancels all scheduled reducer requests. (total scheduled reducers = 1024)
• Then, in scheduleReduces, it ramps up all reducers (total = 1024).

As a result, we can see total #requested-containers increased 1024 for every MRAM-RM heartbeat (1 sec per heartbeat). The AM is hanging for 18+ hours, so we get 18 * 3600 * 1024 ~ 66M+ requested containers in RM side.

And this bug also triggered YARN-4844, which makes RM stop scheduling anything.

Thanks to sidharta-s for helping with analysis.

Issue Links

• is broken by: MAPREDUCE-6302 Preempt reducers after a configurable timeout irrespective of headroom
• is related to: YARN-4844 Add getMemorySize/getVirtualCoresSize to o.a.h.y.api.records.Resource