Harsh J created MAPREDUCE-4694:
----------------------------------
Summary: Inconsistency in reduce record and counters between the
stable and evolving APIs
Key: MAPREDUCE-4694
URL: https://issues.apache.org/jira/browse/MAPREDUCE-4694
Project: Hadoop Map/Reduce
Issue Type: Bug
Components: client
Affects Versions: 2.0.0-alpha
Reporter: Harsh J
In the stable (mapred) API execution, if the values iterator is skipped by a
user, the records underneath it aren't counted in the "Reduce input records"
counter as the key progresses to the next unique one. In the evolving API
(mapreduce) API execution, if the values iterator is skipped by a user, the
records underneath it is still counted as the key progresses to the next unique
one.
This behavior comes to me as a faulty one in the old API. A "Reduce input
records" counter must always define all the records that have been passed into
a reducer (cause they are read regardless of skipping), and both API's record
counting despite user applications must be consistent.
I'll post a test case illustrating this shortly.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
