基于zookeeper的高可用集群

1.准备zookeeper服务器

#node1,node2,node3#安装请参考http://suyanzhu.blog.51cto.com/8050189/1946580

2.准备NameNode节点

#node1,node4

3.准备JournalNode节点

#node2,node3,node4

4.准备DataNode节点

#node2,node3,node4#启动DataNode节点命令hadoop-daemon.sh start datanode

5.修改hadoop的hdfs-site.xml配置文件

                        dfs.nameservices                yunshuocluster                                dfs.ha.namenodes.yunshuocluster                nn1,nn2                                dfs.namenode.rpc-address.yunshuocluster.nn1                node1:8020                                dfs.namenode.rpc-address.yunshuocluster.nn2                node4:8020                                dfs.namenode.http-address.yunshuocluster.nn1                node1:50070                                dfs.namenode.http-address.yunshuocluster.nn2                node4:50070                                dfs.namenode.shared.edits.dir                qjournal://node2:8485;node3:8485;node4:8485/yunshuocluster                                dfs.client.failover.proxy.provider.mycluster                org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider                                dfs.ha.fencing.methods                sshfence                                dfs.ha.fencing.ssh.private-key-files                /root/.ssh/id_dsa                                dfs.journalnode.edits.dir                /opt/journalnode/                                dfs.ha.automatic-failover.enabled                true        

6.修改hadoop的core-site.xml配置文件

            fs.defaultFS        hdfs://yunshuocluster                hadoop.tmp.dir        /opt/hadoop-2.5                ha.zookeeper.quorum        node1:2181,node2:2181,node3:2181    

7.配置slaves配置文件

node2node3node4

8.启动zookeeper(node1,node2,node3)

zkServer.sh start

9.启动Journalnode(node2,node3,node4上分别执行下面的命令)

#启动命令 停止命令hadoop-daemon.sh stop journalnodehadoop-daemon.sh start journalnode

10.检查Journalnode,通过查看日志

cd /home/hadoop-2.5.1/logslstail -200 hadoop-root-journalnode-node2.log

11.格式化NameNode(两台中的一台,这里格式化node4这台NameNode节点)

hdfs namenode -formatcd /opt/hadoop-2.5#两台NameNode同步完成scp -r /opt/hadoop-2.5/* root@node1:/opt/hadoop-2.5/

12.初始化zkfc

hdfs zkfc -formatZK

13.启动服务

start-dfs.sh#stop-dfs.sh表示停止服务