摘要:
早上发现一个任务有20个reduce,但是只有四个正常完成,剩余16个等待了8个小时才分配执行(集群槽位资源充足)解决方法:查看了集群的log,发现有这种warn:2013-09-11 01:56:38,620 WARN org.apache.hadoop.mapred.JobInProgress: No room for reduce task. Node tracker_slave028.hadoop.sjs.org:localhost/127.0.0.1:54676 has 91520086016 bytes free; but we expect reduce input to tak 阅读全文