hadoop学习(二)

The Basics of Multimachine Clusters(2nd)

hadoop配置
可以参考本连接（http://hadoop.apache.org/common/docs/current/api/org/apache/hadoop/conf/Configuration.html）
The framework will load configuration files in order, with the values defined in later files superseding those earlier definitions. The loading order is
hadoop-default.xml, hadoop-site.xml, and then any user specified resources.
配置文件中会有${text}的value配置方式， ${text}用系统中的值替换（System.getProperties(“text”)）
Three critical parameters must be configured for any Hadoop cluster: hadoop.tmp.dir,fs.default.name, and mapred.job.tracker.
Several other parameters are important to tune but not critical: mapred.tasktracker.map.tasks.maximum, mapred.tasktracker.reduce.tasks.maximum, mapred.child.java.opts, and webinterface.private.actions.
If you don’t change this default value for ${hadoop.tmp.dir}, the HDFS data will be stored in /tmp and deleted by the system /tmp cleaning service.

 

当前hadoop-site.xml配置文件被分成了三个，分别是core-site.xml , hdfs-site.xml 和mapred-site.xml （这个在0.20中就已经是这样了）。

参照一下连接：

http://blog.csdn.net/AE86_FC/archive/2010/08/27/5844869.aspx