Never mind. I figured it out. When I increased the number of reduce tasks to more than 1, the reduce-side join now relies on records from different join sources to end up in the corresponding reduce task. Since I did not implement the partitioner to control it, it was messing up reduce-side join. Once I implement the partitioner, everything is working fine now.
-- View this message in context: http://old.nabble.com/No-output-when-more-than-one-reduce-task-tp34299347p34302622.html Sent from the Hadoop core-user mailing list archive at Nabble.com.
