Hadoop分布式集群搭建

Hadoop版本：hadoop-0.20.205.0-1.i386.rpm
下载地址：http://www.fayea.com/apache-mirror/hadoop/common/hadoop-0.20.205.0/
jdk版本：jdk-6u35-linux-i586-rpm.bin 

下载地址：http://www.oracle.com/technetwork/java/javase/downloads/jdk6u35-downloads-1836443.html

环境为RedHat6.2 32bit
master: 192.169.1.133
slave1: 192.169.1.134
slave2: 192.169.1.135

总体的步骤：
1.修改主机名/etc/hosts（虚拟机拷贝后若不一致要从新修改，从新分发）
2.创建一个普通账户（hadoop），hadoop以此账户运行。
2.root安装jdk
3.修改环境变量
4.安装hadoop，修改配置文件
5.将虚拟机拷贝2份，分别作为slave1，slave2
6.配置ssh，使两两之间，自己登陆自己都免密码
7.用普通账户格式化namenode
8.启动，并观察是否正常运行了

注意两个错误：
1.Warning: $HADOOP_HOME is deprecated. 关闭
解决方法：将export HADOOP_HOME_WARN_SUPPRESS=TRUE添加到每个节点的/etc/hadoop/hadoop-env.sh配置文件中。
2.提示不能创建虚拟机错误
#[root@master ~]# /usr/bin/start-all.sh
namenode running as process 26878. Stop it first.
slave2: starting datanode, logging to /var/log/hadoop/root/hadoop-root-datanode-slave2.out
slave1: starting datanode, logging to /var/log/hadoop/root/hadoop-root-datanode-slave1.out
slave2: Unrecognized option: -jvm
slave2: Could not create the Java virtual machine.
slave1: Unrecognized option: -jvm
slave1: Could not create the Java virtual machine.
master: secondarynamenode running as process 26009. Stop it first.
jobtracker running as process 25461. Stop it first.
slave2: starting tasktracker, logging to /var/log/hadoop/root/hadoop-root-tasktracker-slave2.out
slave1: starting tasktracker, logging to /var/log/hadoop/root/hadoop-root-tasktracker-slave1.out

解决方法：root不能启动hadoop，需要用普通账户启动。