1、创建hadoop用户

在每个机器上新建hadoop用户：

useradd -m hadoop

passwd hadoop

2、ssh免登陆设置

首先保证集群内每台datanode和namenode都能相互ping通
在每台机器

vi /etc/hosts

在namenode添加每台机器的ip+主机名

如：

192.168.1.211 namenode

192.168.1.212 datanode1

192.168.1.213 datanode2

。。。。

在datanode添加namenode的ip+主机名

如：

192.168.1.211 namenode

使配置生效

source /etc/hosts

在所有机器新建~/.ssh目录

mkdir ~/.ssh

在namenode

cd ~/.ssh

ssh-keygen -t rsa

一路回车执行默认设置

结束后在~/.ssh目录下会生成id_rsa.pub（公钥）和id_rsa（私钥）两个文件

cp id_rsa.pub authorized_keys

把authorized_keys复制到每个datanode的~/.ssh目录

scp authorized_keys datanode1:/home/hadoop/.ssh/

scp authorized_keys datanode2:/home/hadoop/.ssh/

。。。。

在每台机器

chmod 755 ../.ssh/

chmod 644 authorized_keys

在namenode上

ssh datanode1

ssh datanode2

。。。。

之后ssh登录datanode就可以不用输入密码直接登录（hadoop namenode需要此功能）

3、hadoop安装：

在所有机器
su - hadoop

在namenode上

tar -zxvf hadoop-1.0.2.tar.gz

ln -s hadoop-1.0.2 hadoop

vi vi hadoop/conf/hadoop-env.sh

修改JAVA_HOME为本机jdk位置

vi hadoop/conf/core-site.xml

添加以下内容：

fs.default.name

hdfs://localhost:9000

vi hadoop/conf/mapred-site.xml

添加以下内容：

mapred.job.tracker

localhost:9001

vi hadoop/conf/hdfs-site.xml

添加以下内容：

hdfs.replication

3

把hadoop文件夹复制到各个datanode上

scp -r hadoop-1.0.2/ datanode1:/home/hadoop/

scp -r hadoop-1.0.2/ datanode2:/home/hadoop/

。。。。

在每台datanode

cd ~

ln -s hadoop-1.0.2 hadoop

安装完毕

4、启动 hadoop

进入hadoop/bin，首先格式化文件系统：
hadoop namenode –format
在namenode

~/hadoop/bin/start-all.sh

启动日志如下：

[hadoop@test ~]$ hadoop/bin/start-all.sh

starting namenode, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-namenode-test.out

/usr/local/jdk6/bin/java

datanode1: starting datanode, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-datanode-test2.out

datanode3: starting datanode, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-datanode-jiuzhou4.out

datanode2: starting datanode, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-datanode-jiuzhou3.out

datanode4: starting datanode, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-datanode-jiuzhou1.out

datanode1: /usr/local/jdk6/bin/java

datanode3: /usr/local/jdk6/bin/java

datanode2: /usr/local/jdk6/bin/java

datanode4: /usr/local/jdk6/bin/java

hadoop@namenode's password:

namenode: secondarynamenode running as process 18954. Stop it first.

starting jobtracker, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-jobtracker-test.out

/usr/local/jdk6/bin/java

datanode3: starting tasktracker, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-tasktracker-jiuzhou4.out

datanode2: starting tasktracker, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-tasktracker-jiuzhou3.out

datanode4: starting tasktracker, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-tasktracker-jiuzhou1.out

datanode1: starting tasktracker, logging to /home/hadoop/hadoop-1.0.2/libexec/../logs/hadoop-hadoop-tasktracker-test2.out

datanode3: /usr/local/jdk6/bin/java

datanode2: /usr/local/jdk6/bin/java

datanode4: /usr/local/jdk6/bin/java

datanode1: /usr/local/jdk6/bin/java

[hadoop@test ~]$

集群配置完毕

ITPub博客

Hadoop完全分布式安装