这几天因为数据源更改的原因，每天一个任务运行的作业数是8w多个，每当运行到4w多的时候，整个任务就被kill掉了，追踪日志得到原因。resourceManager的配置yarn.app.mapreduce.am.resource.mb设置为1GB，造成内存溢出，而导致作业被杀死。将参数修改为3GB。
附作业错误log：
| Job Overview | |
|---|---|
| Job Name: | insert overwrite tabl...og_raw_locate_newtmp(Stage-1) |
| User Name: | root |
| Queue: | root.base |
| State: | KILLED |
| Uberized: | false |
| Submitted: | Tue Nov 11 10:20:33 CST 2014 |
| Started: | Tue Nov 11 11:45:19 CST 2014 |
| Finished: | Tue Nov 11 12:53:05 CST 2014 |
| Elapsed: | 1hrs, 7mins, 46sec |
| Diagnostics: | |
| Average Map Time | 34sec |