[MAPREDUCE-7056] Ensure that mapreduce.job.reduces is not negative

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: client
• Assignee: Peter Bacsko
• Reporter: Peter Bacsko
• Created: 20/Feb/18 15:46
• Updated: 22/Feb/19 11:22

Description

Recently we've seen a strange problem that was related to mapreduce.job.reduces being set to -1. If this value is negative, two things can happen:

1. If we use the old API, then the mappers will pass, but the number of reducers will be recorded as "-1" if we open it later from JHS. This can confuse Hadoop users.

2. If use the new API, then we'll see a not-so-obvious stack trace:

2018-02-20 06:37:35,493 INFO [main] org.apache.hadoop.mapred.MapTask: Starting flush of map output
2018-02-20 06:37:35,507 INFO [main] org.apache.hadoop.mapred.MapTask: Starting flush of map output
2018-02-20 06:37:35,507 INFO [main] org.apache.hadoop.mapred.MapTask: kvbuffer is null. Skipping flush.
2018-02-20 06:37:35,508 WARN [main] org.apache.hadoop.mapred.YarnChild: Exception running child : java.lang.IllegalArgumentException
	at java.nio.ByteBuffer.allocate(ByteBuffer.java:334)
	at org.apache.hadoop.mapred.SpillRecord.<init>(SpillRecord.java:51)
	at org.apache.hadoop.mapred.MapTask$MapOutputBuffer.mergeParts(MapTask.java:1891)
	at org.apache.hadoop.mapred.MapTask$MapOutputBuffer.flush(MapTask.java:1527)
	at org.apache.hadoop.mapred.MapTask$NewOutputCollector.close(MapTask.java:735)
	at org.apache.hadoop.mapred.MapTask.runNewMapper(MapTask.java:805)
	at org.apache.hadoop.mapred.MapTask.run(MapTask.java:347)
	at org.apache.hadoop.mapred.YarnChild$2.run(YarnChild.java:174)
	at java.security.AccessController.doPrivileged(Native Method)
	at javax.security.auth.Subject.doAs(Subject.java:422)
	at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1962)
	at org.apache.hadoop.mapred.YarnChild.main(YarnChild.java:168)

and the job fails.

We should either fail if this property is negative or set to "0" to avoid this.

Issue Links

• is duplicated by: MAPREDUCE-7186 JobSubmitter#checkSpecs doesn not handle the situation when reduceNums is set less than zero.