Shuffle Error: Exceeded MAX_FAILED_UNIQUE_FETCHES解决方案

转载自：http://hi.baidu.com/xiaomi/item/6d2cbdf35064040c84d27844

1.map 完成，但reduce阶段出错。

12/01/06 17:09:15 INFO input.FileInputFormat: Total input paths to process : 2
12/01/06 17:09:16 INFO mapred.JobClient: Running job: job_201201061706_0001
12/01/06 17:09:17 INFO mapred.JobClient:  map 0% reduce 0%
12/01/06 17:09:31 INFO mapred.JobClient:  map 50% reduce 0%
12/01/06 17:09:34 INFO mapred.JobClient:  map 100% reduce 0%
12/01/06 17:09:40 INFO mapred.JobClient:  map 100% reduce 16%
12/01/06 17:16:59 INFO mapred.JobClient:  map 100% reduce 0%
12/01/06 17:17:04 INFO mapred.JobClient: Task Id : attempt_201201061706_0001_r_000000_0, Status : FAILED
Shuffle Error: Exceeded MAX_FAILED_UNIQUE_FETCHES; bailing-out.
12/01/06 17:17:05 WARN mapred.JobClient: Error reading task outputjx2
12/01/06 17:17:05 WARN mapred.JobClient: Error reading task outputjx2
12/01/06 17:17:17 INFO mapred.JobClient:  map 100% reduce 16%

解决是：在namenode以及datanode的 /etc/hosts 添加 ip与别名。如：

192.168.5.191 jx1

192.168.5.192 jx2