系统环境：CentOS6.3+JDK1.7
部署资源：hadoop-2.2.0 zookeeper-3.4.5
集群规划：
主机名 IP 安装的软件 运行的进程
reagina01 192.168.8.201 jdk,hadoop NameNode,DFSZKFailoverController
reagina02 192.168.8.202 jdk,hadoop NameNode,DFSZKFailoverController
reagina03 192.168.8.203 jdk,hadoop ResourceManager
reagina04 192.168.8.204 jdk,hadoop,zookeeper DataNode,NodeManager,JournalNode,QuorumPeerMain
reagina05 192.168.8.205 jdk,hadoop,zookeeper DataNode,NodeManager,JournalNode,QuorumPeerMain
reagina06 192.168.8.206 jdk,hadoop,zookeeper DataNode,NodeManager,JournalNode,QuorumPeerMain
在reagina01节点上拆解Hadoop：
[root@reagina01 ~]# tar -zxvf hadoop-2.2.0-64bit.tar.gz -C /reagina/
修改配置文件：
1、hadoop-env.sh
2、core-site.xml
3、hdfs-site.xml
4、mapred-site.xml
5、yarn-site.xml
6、slaves
配置免密码登录：
首先要配置reagina01到reagina01、reagina02、reagina03、reagina04、reagina05、reagina06的免密码登录。
[root@reagina01 ~]# ssh-keygen -t rsa
[root@reagina01 ~]# ssh-copy-id reagina01
[root@reagina01 ~]# ssh-copy-id reagina02
[root@reagina01 ~]# ssh-copy-id reagina03
[root@reagina01 ~]# ssh-copy-id reagina04
[root@reagina01 ~]# ssh-copy-id reagina05
[root@reagina01 ~]# ssh-copy-id reagina06
然后配置reagina03到reagina03、reagina04、reagina05、reagina06的免密码登录
[root@reagina03 ~]# ssh-keygen -t rsa
[root@reagina03 ~]# ssh-copy-id reagina03
[root@reagina03 ~]# ssh-copy-id reagina04
[root@reagina03 ~]# ssh-copy-id reagina05
[root@reagina03 ~]# ssh-copy-id reagina06
注意：两个namenode之间要配置ssh免密码登录，所以还要配置reagina02到reagina01的免密码登录：
[root@reagina02 ~]# ssh-keygen -t rsa
[root@reagina02 ~]# ssh-copy-id reagina01
将配置好的Hadoop拷贝到其他节点：
[root@reagina01 ~]# scp -r /reagina/ reagina02:/
[root@reagina01 ~]# scp -r /reagina/ reagina03:/
[root@reagina01 ~]# scp -r /reagina/hadoop-2.2.0/ root@reagina04:/reagina/
[root@reagina01 ~]# scp -r /reagina/hadoop-2.2.0/ root@reagina05:/reagina/
[root@reagina01 ~]# scp -r /reagina/hadoop-2.2.0/ root@reagina06:/reagina/
reagina04、reagina05、reagina06节点上的zookeeper集群的搭建：
http://www.cnblogs.com/reagina/p/6366363.html
启动zookeeper集群：
在reagina01上启动所有journalnode：
[root@reagina03 ~]# /reagina/hadoop-2.2.0/sbin/hadoop-daemons.sh start journalnode
格式化HDFS：
[root@reagina01 ~]# /reagina/hadoop-2.2.0/bin/hdfs namenode -format
格式化后会再根据core-site.xml中的hadoop.tmp.dir配置生成一个目录，然后将这个目录拷贝到reagina02节点：
[root@reagina01 ~]# scp -r /reagina/hadoop-2.2.0/tmp/ reagina02:/reagina/hadoop-2.2.0/
格式化ZK：
[root@reagina01 ~]# /reagina/hadoop-2.2.0/bin/hdfs zkfc -formatZK
启动HDFS：
[root@reagina01 ~]# /reagina/hadoop-2.2.0/sbin/start-dfs.sh
启动YARN（注意：是在reagina03上执行start-yarn.sh，把namenode和resourcemanager分开是因为性能问题，因为它们都要占用大量资源，所以把它们分开了，它们分开了就要分别在不同的机器上启动）
[root@reagina03 ~]# /reagina/hadoop-2.2.0/sbin/start-yarn.sh
在namenode节点上（reagina01、reagina02）启动zookeeper选举制度：
[root@reagina01 ~]# /reagina/hadoop-2.2.0/sbin/hadoop-daemon.sh start zkfc
starting zkfc, logging to /reagina/hadoop-2.2.0/logs/hadoop-root-zkfc-reagina01.out
[root@reagina02 ~]# /reagina/hadoop-2.2.0/sbin/hadoop-daemon.sh start zkfc
starting zkfc, logging to /reagina/hadoop-2.2.0/logs/hadoop-root-zkfc-reagina02.out
[root@reagina02 ~]# jps
2631 DFSZKFailoverController
2691 Jps
2183 NameNode
[root@reagina02 ~]#
到此，hadoop2.2.0配置完毕，可以统计浏览器访问：
验证HDFS HA
[root@reagina01 sbin]# hadoop fs -put /etc/profile /profile
[root@reagina01 sbin]# hadoop fs -ls /
-rw-r--r-- 3 root supergroup 1924 2017-02-05 01:10 /profile
[root@reagina01 sbin]# jps
4727 NameNode
5260 DFSZKFailoverController
5491 Jps
[root@reagina01 sbin]# kill -9 4727
