hadoop依赖jdk

从后往前安装

jdk篇

要避免各种麻烦请从binary安装

wget http://download.oracle.com/otn-pub/java/jdk/6u25-b06/jdk-6u25-linux-x64.bin

将上述文件mv至期望的安装目录并cd过去

./jdk-6u25-linux-x64.bin

export PATH=$PATH:path_of_javac

并将path_of_javac加到.bashrc

javac即安装成功

并且自带了jdb用来调试

hadoop篇

同样从binary按教程安装

注意如果一台机器上有多个hadoop实例

就需要改9个端口 在mapred-site.xml core-site.xml hdfs-site.xml

且不能用ip，只能域名，不然reduce一直是0%

可能需要清datanode(rm -rf path_of_dfs.data.dir)并重启namenode

参考资料：

hadoop Quick Start

Hadoop Troubleshooting

hadoop FAQ

错误fix

Name node is in safe mode?

bin/hadoop dfsadmin -safemode leave

can't access single columns?

<property>
  <name>hive.aux.jars.path</name>
  <value>file:path_to_hive_lib/hive-contrib-0.8.0.jar</value>
  <description>These JAR file are available to all users for alljobs</description>
</property>