仗贱走天涯

背景Hive query将运算好的数据写回hdfs（比如insert into语句），有时候会产生大量的小文件，如果不采用CombineHiveInputFormat就对这些小文件进行操作的话会产生大量的map task，耗费大量集群资源，而且小文件过多会对namenode造成很大压力。所以Hive在正常job执行完之后，会起一个conditional task，来判断是否需要合并小文件，

http://blog.csdn.net/yeruby/article/details/51003445

http://www.jianshu.com/p/6e59df5f2461

http://www.cnblogs.com/woxpp/p/6248370.html

https://blog.csdn.net/nsrainbow/article/details/36396007

https://blog.csdn.net/hliq5399/article/details/78193113

https://www.cnblogs.com/davidzhu/category/1344548.html