Hadoop HA的搭建

1.首先添加hosts文件

vim /etc/hosts

192.168.0.1  MSJTVL-DSJC-H01
192.168.0.2  MSJTVL-DSJC-H03
192.168.0.3  MSJTVL-DSJC-H05
192.168.0.4  MSJTVL-DSJC-H02
192.168.0.5  MSJTVL-DSJC-H04

2.几台机器做互信

Setup passphraseless ssh

Now check that you can ssh to the localhost without a passphrase:

  $ ssh localhost
If you cannot ssh to localhost without a passphrase, execute the following commands:

  $ ssh-keygen -t dsa -P '' -f ~/.ssh/id_dsa
  $ cat ~/.ssh/id_dsa.pub >> ~/.ssh/authorized_keys

把其他几台机器的秘钥文件复制到MSJTVL-DSJC-H01的authorized_keys文件中

[hadoop@MSJTVL-DSJC-H01 .ssh]$ scp hadoop@MSJTVL-DSJC-H02:/hadoop/.ssh/id_dsa.pub ./id_dsa.pub2
[hadoop@MSJTVL-DSJC-H01 .ssh]$ scp hadoop@MSJTVL-DSJC-H03:/hadoop/.ssh/id_dsa.pub ./id_dsa.pub3
[hadoop@MSJTVL-DSJC-H01 .ssh]$ scp hadoop@MSJTVL-DSJC-H04:/hadoop/.ssh/id_dsa.pub ./id_dsa.pub4
[hadoop@MSJTVL-DSJC-H01 .ssh]$ scp hadoop@MSJTVL-DSJC-H05:/hadoop/.ssh/id_dsa.pub ./id_dsa.pub5

[hadoop@MSJTVL-DSJC-H01 .ssh]$ cat ~/.ssh/id_dsa.pub2 >> ~/.ssh/authorized_keys
[hadoop@MSJTVL-DSJC-H01 .ssh]$ cat ~/.ssh/id_dsa.pub3 >> ~/.ssh/authorized_keys
[hadoop@MSJTVL-DSJC-H01 .ssh]$ cat ~/.ssh/id_dsa.pub4 >> ~/.ssh/authorized_keys
[hadoop@MSJTVL-DSJC-H01 .ssh]$ cat ~/.ssh/id_dsa.pub5 >> ~/.ssh/authorized_keys

以上操作实现了MSJTVL-DSJC-H02,3,4,5对MSJTVL-DSJC-H01的无密码登录

要是实现MSJTVL-DSJC-H01-5的全部互信则把MSJTVL-DSJC-H01上的authorized_keys文件COPY到其他机器上去

[hadoop@MSJTVL-DSJC-H02 ~]$ scp hadoop@MSJTVL-DSJC-H01:/hadoop/.ssh/authorized_keys /hadoop/.ssh/authorized_keys

　

下载相应的tar包

wget http://apache.fayea.com/hadoop/common/hadoop-2.6.4/hadoop-2.6.4.tar.gz

解压tar包并且建立相应的软链接

[hadoop@MSJTVL-DSJC-H01 ~]$ tar -zxvf hadoop-2.6.4.tar.gz
[hadoop@MSJTVL-DSJC-H01 ~]$ ln -sf hadoop-2.6.4 hadoop

进到hadoop相应的配置文件路径，修改hadoop-env.sh的内容

[hadoop@MSJTVL-DSJC-H01 ~]$ cd hadoop/etc/hadoop/
[hadoop@MSJTVL-DSJC-H01 hadoop]$ vim hadoop-env.sh

修改hadoop-env.sh里java_home的参数信息

接下来修改hdfs-site.xml中的相关内容，来源http://hadoop.apache.org/docs/r2.6.4/hadoop-project-dist/hadoop-hdfs/HDFSHighAvailabilityWithQJM.html

首先配置一个匿名服务dfs.nameservices

[hadoop@MSJTVL-DSJC-H01 hadoop]$ vim hdfs-site.xml
<configuration>
//配置服务的名称，可以进行相应的修改
<property>
  <name>dfs.nameservices</name>
  <value>mycluster</value>
</property>

//配置namenode的名称,mycluster需要和前面的保持一致，nn1和nn2只是名称无所谓叫啥
<property>
  <name>dfs.ha.namenodes.mycluster</name>
  <value>nn1,nn2</value>
</property>

//配置RPC协议的端口，两个namenode的RPC协议和端口，需要修改servicesname和value中的主机名称，MSJTVL-DSJC-H01和MSJTVL-DSJC-H02是两个namenode的主机名称
<property>
  <name>dfs.namenode.rpc-address.mycluster.nn1</name>
  <value>MSJTVL-DSJC-H01:8020</value>
</property>
<property>
  <name>dfs.namenode.rpc-address.mycluster.nn2</name>
  <value>MSJTVL-DSJC-H02:8020</value>
</property>

//配置下面是http的主机和端口
<property>
  <name>dfs.namenode.http-address.mycluster.nn1</name>
  <value>MSJTVL-DSJC-H01:50070</value>
</property>
<property>
  <name>dfs.namenode.http-address.mycluster.nn2</name>
  <value>MSJTVL-DSJC-H02:50070</value>
</property>

//接下来配置的是JournalNodes的URL地址
<property>
  <name>dfs.namenode.shared.edits.dir</name>
  <value>qjournal://MSJTVL-DSJC-H03:8485;MSJTVL-DSJC-H04:8485;MSJTVL-DSJC-H05:8485/mycluster</value>
</property>

//然后是固定的一个客户端使用的类（需要修改serversname的名称），客户端通过这个类找到
<property>
  <name>dfs.client.failover.proxy.provider.mycluster</name>
  <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
</property>

//sshfence - SSH to the Active NameNode and kill the process，注意为hadoop下.ssh目录中生成的秘钥文件
<property>
  <name>dfs.ha.fencing.methods</name>
  <value>sshfence</value>
</property>
<property>
  <name>dfs.ha.fencing.ssh.private-key-files</name>
  <value>/hadoop/.ssh/id_dsa</value>
</property>

//JournalNodes的工作目录
<property>
  <name>dfs.journalnode.edits.dir</name>
  <value>/hadoop/jn/data</value>
</property>

//开启自动切换namenode
<property>
   <name>dfs.ha.automatic-failover.enabled</name>
   <value>true</value>
 </property>
</configuration>

接下来编辑core-site.xml的配置文件

//首先配置namenode的入口，同样注意serversname的名称
<property>
  <name>fs.defaultFS</name>
  <value>hdfs://mycluster</value>
</property>

//配置zookeeper的集群
<property>
   <name>ha.zookeeper.quorum</name>
   <value>MSJTVL-DSJC-H03:2181,MSJTVL-DSJC-H04:2181,MSJTVL-DSJC-H05:2181</value>
</property>

//hadoop的临时目录
<property>
  <name>hadoop.tmp.dir</name>
  <value>/hadoop/tmp</value>
</property>

 配置slaves

MSJTVL-DSJC-H03
MSJTVL-DSJC-H04
MSJTVL-DSJC-H05

安装zookeeper

直接解压

修改相应的配置文件

[zookeeper@MSJTVL-DSJC-H03 conf]$ vim zoo.cfg
//修改dataDir=/opt/zookeeper/data，不要放到tmp下
dataDir=/opt/zookeeper/data

#autopurge.purgeInterval=1
server.1=MSJTVL-DSJC-H03:2888:3888
server.2=MSJTVL-DSJC-H04:2888:3888
server.3=MSJTVL-DSJC-H05:2888:3888

在/opt/zookeeper/data下建立myid里面存储跟server一样的数字

启动zookeeper（zkServer.sh start），jps查看启动状态

启动HA集群

1.首先启动JournalNodes,到sbin目录下

 ./hadoop-daemon.sh start journalnode

[hadoop@MSJTVL-DSJC-H03 sbin]$ ./hadoop-daemon.sh start journalnode
starting journalnode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-journalnode-MSJTVL-DSJC-H03.out
[hadoop@MSJTVL-DSJC-H03 sbin]$ jps
3204 JournalNode
3252 Jps
[hadoop@MSJTVL-DSJC-H03 sbin]$ 

2.在一台namenode上进行格式化

[hadoop@MSJTVL-DSJC-H01 bin]$ ./hdfs namenode -format

初始化之后会在/hadoop/tmp/dfs/name/current下产生相应的元数据文件

[hadoop@MSJTVL-DSJC-H01 ~]$ cd tmp/
[hadoop@MSJTVL-DSJC-H01 tmp]$ ll
总用量 4
drwxr-xr-x. 3 hadoop hadoop 4096 9月   6 16:54 dfs
[hadoop@MSJTVL-DSJC-H01 tmp]$ cd dfs/
[hadoop@MSJTVL-DSJC-H01 dfs]$ ll
总用量 4
drwxr-xr-x. 3 hadoop hadoop 4096 9月   6 16:54 name
[hadoop@MSJTVL-DSJC-H01 dfs]$ cd name/
[hadoop@MSJTVL-DSJC-H01 name]$ ll
总用量 4
drwxr-xr-x. 2 hadoop hadoop 4096 9月   6 16:54 current
[hadoop@MSJTVL-DSJC-H01 name]$ cd current/
[hadoop@MSJTVL-DSJC-H01 current]$ ll
总用量 16
-rw-r--r--. 1 hadoop hadoop 352 9月   6 16:54 fsimage_0000000000000000000
-rw-r--r--. 1 hadoop hadoop  62 9月   6 16:54 fsimage_0000000000000000000.md5
-rw-r--r--. 1 hadoop hadoop   2 9月   6 16:54 seen_txid
-rw-r--r--. 1 hadoop hadoop 201 9月   6 16:54 VERSION
[hadoop@MSJTVL-DSJC-H01 current]$ pwd
/hadoop/tmp/dfs/name/current
[hadoop@MSJTVL-DSJC-H01 current]$ 

3.把初始化的元数据文件COPY到其他的namenode上去，COPY之前需要先启动格式化的namenode

[hadoop@MSJTVL-DSJC-H01 sbin]$ ./hadoop-daemon.sh start namenode
starting namenode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-namenode-MSJTVL-DSJC-H01.out
[hadoop@MSJTVL-DSJC-H01 sbin]$ jps
3324 NameNode
3396 Jps
[hadoop@MSJTVL-DSJC-H01 sbin]$ 

然后在没有格式化的namenode上执行hdfs namenode -bootstrapStandby，执行完后查看元数据文件是一样的表示成功。

[hadoop@MSJTVL-DSJC-H02 bin]$ hdfs namenode -bootstrapStandby

　　

4.初始化ZKFC,在任意一台机器上执行hdfs zkfc -formatZK初始化ZKFC

5.重启整个HDFS集群

[hadoop@MSJTVL-DSJC-H01 sbin]$ ./start-dfs.sh 
16/09/06 17:10:25 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Starting namenodes on [MSJTVL-DSJC-H01 MSJTVL-DSJC-H02]
MSJTVL-DSJC-H02: starting namenode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-namenode-MSJTVL-DSJC-H02.out
MSJTVL-DSJC-H01: starting namenode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-namenode-MSJTVL-DSJC-H01.out
MSJTVL-DSJC-H03: starting datanode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-datanode-MSJTVL-DSJC-H03.out
MSJTVL-DSJC-H04: starting datanode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-datanode-MSJTVL-DSJC-H04.out
MSJTVL-DSJC-H05: starting datanode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-datanode-MSJTVL-DSJC-H05.out
Starting journal nodes [MSJTVL-DSJC-H03 MSJTVL-DSJC-H04 MSJTVL-DSJC-H05]
MSJTVL-DSJC-H03: starting journalnode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-journalnode-MSJTVL-DSJC-H03.out
MSJTVL-DSJC-H04: starting journalnode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-journalnode-MSJTVL-DSJC-H04.out
MSJTVL-DSJC-H05: starting journalnode, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-journalnode-MSJTVL-DSJC-H05.out
16/09/06 17:10:43 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Starting ZK Failover Controllers on NN hosts [MSJTVL-DSJC-H01 MSJTVL-DSJC-H02]
MSJTVL-DSJC-H02: starting zkfc, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-zkfc-MSJTVL-DSJC-H02.out
MSJTVL-DSJC-H01: starting zkfc, logging to /hadoop/hadoop-2.6.4/logs/hadoop-hadoop-zkfc-MSJTVL-DSJC-H01.out
[hadoop@MSJTVL-DSJC-H01 sbin]$ jps
4345 Jps
4279 DFSZKFailoverController
3993 NameNode

6.创建一个目录

./hdfs dfs -mkdir -p /usr/file
./hdfs dfs -put /hadoop/tian.txt /usr/file

放上一个文件可以在网页中查看相应的文件。

MR高可用

配置yarn-site.xml

<configuration> 
    		<!--启用RM高可用--> 

    <property> 

    		<name>yarn.resourcemanager.ha.enabled</name> 

    		<value>true</value> 

    </property> 

    		<!--RM集群标识符--> 

    <property> 

    		<name>yarn.resourcemanager.cluster-id</name> 

    		<value>rm-cluster</value> 

    </property> 

    <property> 

    		<!--指定两台RM主机名标识符--> 

    		<name>yarn.resourcemanager.ha.rm-ids</name> 

    		<value>rm1,rm2</value> 

    </property> 

    		<!--RM故障自动切换--> 

    <property> 

        <name>yarn.resourcemanager.ha.automatic-failover.recover.enabled</name> 

        <value>true</value> 

    </property> 

        <!--RM故障自动恢复--> 

    <property> 

        <name>yarn.resourcemanager.recovery.enabled</name>  

        <value>true</value>  

    </property> --> 

        <!--RM主机1--> 

    <property> 

        <name>yarn.resourcemanager.hostname.rm1</name> 

        <value>MSJTVL-DSJC-H01</value> 

    </property> 

        <!--RM主机2--> 

    <property> 

        <name>yarn.resourcemanager.hostname.rm2</name> 

        <value>MSJTVL-DSJC-H02</value> 

    </property> 

        <!--RM状态信息存储方式,一种基于内存(MemStore),另一种基于ZK(ZKStore)--> 

    <property> 

        <name>yarn.resourcemanager.store.class</name> 

        <value>org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore</value> 

    </property> 

        <!--使用ZK集群保存状态信息--> 

    <property> 

        <name>yarn.resourcemanager.zk-address</name> 

        <value>MSJTVL-DSJC-H03:2181,MSJTVL-DSJC-H04:2181,MSJTVL-DSJC-H05:2181</value> 

    </property> 

        <!--向RM调度资源地址--> 

    <property> 

        <name>yarn.resourcemanager.scheduler.address.rm1</name> 

        <value>MSJTVL-DSJC-H01:8030</value> 

    </property> 

    <property> 

        <name>yarn.resourcemanager.scheduler.address.rm2</name> 

        <value>MSJTVL-DSJC-H02:8030</value> 

    </property> 

        <!--NodeManager通过该地址交换信息--> 

    <property> 

        <name>yarn.resourcemanager.resource-tracker.address.rm1</name> 

        <value>MSJTVL-DSJC-H01:8031</value> 

    </property> 

    <property> 

        <name>yarn.resourcemanager.resource-tracker.address.rm2</name> 

        <value>MSJTVL-DSJC-H02:8031</value> 

    </property> 

        <!--客户端通过该地址向RM提交对应用程序操作--> 

    <property> 

        <name>yarn.resourcemanager.address.rm1</name> 

        <value>MSJTVL-DSJC-H01:8032</value> 

    </property> 

    <property> 

        <name>yarn.resourcemanager.address.rm2</name> 

        <value>MSJTVL-DSJC-H02:8032</value> 

    </property> 

        <!--管理员通过该地址向RM发送管理命令--> 

    <property> 

        <name>yarn.resourcemanager.admin.address.rm1</name> 

        <value>MSJTVL-DSJC-H01:8033</value> 

    </property> 

    <property> 

        <name>yarn.resourcemanager.admin.address.rm2</name> 

        <value>MSJTVL-DSJC-H02:8033</value> 

    </property> 

        <!--RM HTTP访问地址,查看集群信息--> 

    <property> 

        <name>yarn.resourcemanager.webapp.address.rm1</name> 

        <value>MSJTVL-DSJC-H01:8088</value> 

    </property> 

    <property> 

        <name>yarn.resourcemanager.webapp.address.rm2</name> 

        <value>MSJTVL-DSJC-H02:8088</value> 

    </property> 

</configuration> 

　　

配置mapred-site.xml

//指定mr框架为yarn方式
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>

standby的MR需要手动启动

[hadoop@MSJTVL-DSJC-H02 sbin]$ yarn-daemon.sh start resourcemanager
starting resourcemanager, logging to /hadoop/hadoop-2.6.4/logs/yarn-hadoop-resourcemanager-MSJTVL-DSJC-H02.out
[hadoop@MSJTVL-DSJC-H02 sbin]$ jps
3000 ResourceManager
2812 NameNode
3055 Jps
2922 DFSZKFailoverController
[hadoop@MSJTVL-DSJC-H02 sbin]$