job执行太长hadoop hadoop

hadoop执行总是有一个reduce失败，求解？这取决于要回滚的减速器级别。map成功执行后，Hadoop会启动整个作业对应的reducer，会有reduce阶段。如...