Hadoop分布式HA的安装部署

前言

单机版的Hadoop环境只有一个namenode，一般namenode出现问题，整个系统也就无法使用，所以高可用主要指的是namenode的高可用，即存在两个namenode节点，一个为active状态，一个为standby状态。如下图：

说明如下：
HDFS的HA，指的是在一个集群中存在两个NameNode，分别运行在独立的物理节点上。在任何时间点，只有一个NameNodes是处于Active状态，另一种是在Standby状态。 Active NameNode负责所有的客户端的操作，而Standby NameNode用来同步Active NameNode的状态信息，以提供快速的故障恢复能力。
为了保证Active NN与Standby NN节点状态同步，即元数据保持一致。除了DataNode需要向两个NN发送block位置信息外，还构建了一组独立的守护进程”JournalNodes”,用来同步FsEdits信息。当Active NN执行任何有关命名空间的修改，它需要持久化到一半以上的JournalNodes上。而Standby NN负责观察JNs的变化，读取从Active NN发送过来的FsEdits信息，并更新自己内部的命名空间。一旦ActiveNN遇到错误，Standby NN需要保证从JNs中读出了全部的FsEdits，然后切换成Active状态。
使用HA的时候，不能启动
SecondaryNameNode，会出错。

集群的规划

        ip                      基本的软件                       运行的进程       
        uplooking01             jdk、zk、hadoop                   NameNode、zkfc、zk、journalNode        
        uplooking02             jdk、zk、hadoop                   NameNode、zkfc、zk、journalNode、datanode、ResourceManager、NodeManager
        uplooking03             jdk、zk、hadoop                   zk、journalNode、datanode、ResourceManager、NodeManager

zookeeper集群搭建

    1、解压：
        [uplooking@uplooking01 ~]$ tar -zxvf soft/zookeeper-3.4.6.tar.gz -C app/
    2、重命名
        [uplooking@uplooking01 ~]$ mv app/zookeeper-3.4.6 app/zookeeper
    3、配置文件重命名
        [uplooking@uplooking01 zookeeper]$ cp conf/zoo_sample.cfg conf/zoo.cfg
    4、修改配置文件$ZOOKEEPER_HOME/conf/zoo.cfg
        dataDir=/home/uplooking/app/zookeeper/data
        dataLogDir=/home/uplooking/logs/zookeeper

        server.101=uplooking01:2888:3888
        server.102=uplooking02:2888:3888
        server.103=uplooking03:2888:3888

        启动server表示当前节点就是zookeeper集群中的一个server节点
        server后面的.数字(不能重复)是当前server节点在该zk集群中的唯一标识
        =后面则是对当前server的说明，用":"分隔开，
        第一段是当前server所在机器的主机名
        第二段和第三段以及2818端口
            2181--->zookeeper服务器开放给client连接的端口
            2888--->zookeeper服务器之间进行通信的端口
            3888--->zookeeper和外部进程进行通信的端口
    5、在dataDir=/home/uplooking/app/zookeeper/data下面创建一个文件myid
        uplooking01机器对应的server.后面的101
        uplooking02机器对应的server.后面的102
        uplooking03机器对应的server.后面的103
    6、需要将在uplooking01上面的zookeeper拷贝之uplooking02和uplooking03，这里使用scp远程拷贝
        scp -r app/zookeeper uplooking@uplooking02:/home/uplooking/app
        scp -r app/zookeeper uplooking@uplooking03:/home/uplooking/app
        在拷贝的过程中需要设置ssh免密码登录
            在uplooking02和uplooking03上面生成ssh密钥
            ssh-keygen -t rsa
            将密钥拷贝授权文件中
            uplooking02：
                ssh-keygen -t rsa
                ssh-copy-id -i uplooking@uplooking02
            uplooking03：
                ssh-keygen -t rsa
                ssh-copy-id -i uplooking@uplooking03
            uplooking01：
                ssh-copy-id -i uplooking@uplooking03
    7、修改myid文件          
        [uplooking@uplooking02 ~]$ echo 102 > app/zookeeper/data/myid 
        [uplooking@uplooking03 ~]$ echo 103 > app/zookeeper/data/myid 
    8、同步环境变量文件
        [uplooking@uplooking01 ~]$ scp .bash_profile uplooking@uplooking02:/home/uplooking/
        [uplooking@uplooking01 ~]$ scp .bash_profile uplooking@uplooking03:/home/uplooking/
    9、启动
        在1、2、3分别执行zkServer.sh start

Hadoop分布式HA的部署

    1、解压
        [uplooking@uplooking01 ~]$ tar -zvxf soft/hadoop-2.6.4.tar.gz -C app/
    2、重命名
        [uplooking@uplooking01 ~]$ mv app/hadoop-2.6.4/ app/hadoop
    3、修改配置文件
        hadoop-env.sh、yarn-env.sh、hdfs-site.xml、core-site.xml、mapred-site.xml、yarn-site.xml、slaves
        1°、hadoop-env.sh
            export JAVA_HOME=/opt/jdk
        2°、yarn-env.sh
            export JAVA_HOME=/opt/jdk
        3°、slaves
            uplooking02
            uplooking03
        4°、hdfs-site.xml
            <configuration>
                <!--指定hdfs的nameservice为ns1，需要和core-site.xml中的保持一致 -->
                <property>
                    <name>dfs.nameservices</name>
                    <value>ns1</value>
                </property>
                <!-- ns1下面有两个NameNode，分别是nn1，nn2 -->
                <property>
                    <name>dfs.ha.namenodes.ns1</name>
                    <value>nn1,nn2</value>
                </property>
                <!-- nn1的RPC通信地址 -->
                <property>
                    <name>dfs.namenode.rpc-address.ns1.nn1</name>
                    <value>uplooking01:9000</value>
                </property>
                <!-- nn1的http通信地址 -->
                <property>
                    <name>dfs.namenode.http-address.ns1.nn1</name>
                    <value>uplooking01:50070</value>
                </property>
                <!-- nn2的RPC通信地址 -->
                <property>
                    <name>dfs.namenode.rpc-address.ns1.nn2</name>
                    <value>uplooking02:9000</value>
                </property>
                <!-- nn2的http通信地址 -->
                <property>
                    <name>dfs.namenode.http-address.ns1.nn2</name>
                    <value>uplooking02:50070</value>
                </property>
                <!-- 指定NameNode的元数据在JournalNode上的存放位置 -->
                <property>
                    <name>dfs.namenode.shared.edits.dir</name>
                    <value>qjournal://uplooking01:8485;uplooking02:8485;uplooking03:8485/ns1</value>
                </property>
                <!-- 指定JournalNode在本地磁盘存放数据的位置 -->
                <property>
                    <name>dfs.journalnode.edits.dir</name>
                    <value>/home/uplooking/data/hadoop/journal</value>
                </property>
                <property>  
                    <name>dfs.namenode.name.dir</name>  
                    <value>/home/uplooking/data/hadoop/name</value>  
                </property>  
                <property>  
                    <name>dfs.datanode.data.dir</name>  
                    <value>/home/uplooking/data/hadoop/data</value>  
                </property> 
                <!-- 开启NameNode失败自动切换 -->
                <property>
                    <name>dfs.ha.automatic-failover.enabled</name>
                    <value>true</value>
                </property>
                <!-- 配置失败自动切换实现方式 -->
                <property>
                    <name>dfs.client.failover.proxy.provider.ns1</name>
                    <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
                </property>
                <!-- 配置隔离机制方法，多个机制用换行分割，即每个机制暂用一行-->
                <property>
                    <name>dfs.ha.fencing.methods</name>
                    <value>
                    sshfence
                    shell(/bin/true)
                    </value>
                </property>
                <!-- 使用sshfence隔离机制时需要ssh免登陆 -->
                <property>
                    <name>dfs.ha.fencing.ssh.private-key-files</name>
                    <value>/home/uplooking/.ssh/id_rsa</value>
                </property>
                <!-- 配置sshfence隔离机制超时时间 -->
                <property>
                    <name>dfs.ha.fencing.ssh.connect-timeout</name>
                    <value>30000</value>
                </property>
            </configuration>
        5°、core-site.xml    
            <configuration>
                <!-- 指定hdfs的nameservice为ns1 -->
                <property>
                    <name>fs.defaultFS</name>
                    <value>hdfs://ns1</value>
                </property>
                <!-- 指定hadoop临时目录 -->
                <property>
                    <name>hadoop.tmp.dir</name>
                    <value>/home/uplooking/data/hadoop/tmp</value>
                </property>
                <!-- 指定zookeeper地址 -->
                <property>
                    <name>ha.zookeeper.quorum</name>
                    <value>uplooking01:2181,uplooking02:2181,uplooking03:2181</value>
                </property>
            </configuration>
        6°、mapred-site.xml  
            <configuration>
                <!-- mr依赖的框架名称 yarn-->
                <property>
                    <name>mapreduce.framework.name</name>
                    <value>yarn</value>
                </property>
                <!-- mr转化历史任务的rpc通信地址-->
                <property>  
                    <name>mapreduce.jobhistory.address</name>  
                    <value>uplooking02:10020</value>  
                </property>
                <!-- mr转化历史任务的http通信地址-->
                <property>  
                    <name>mapreduce.jobhistory.webapp.address</name>  
                    <value>uplooking02:19888</value>  
                </property>
                <!-- 会在hdfs的根目录下面创建一个history的文件夹，存放历史任务的相关运行情况-->
                <property>
                    <name>yarn.app.mapreduce.am.staging-dir</name>
                    <value>/history</value>
                </property>
                <!-- map和reduce的日志级别-->
                <property>
                    <name>mapreduce.map.log.level</name>
                    <value>INFO</value>
                </property>
                <property>
                    <name>mapreduce.reduce.log.level</name>
                    <value>INFO</value>
                </property>
            </configuration>
        7°、yarn-site.xml    
            <configuration>
                <!-- 开启RM高可靠 -->
                <property>
                    <name>yarn.resourcemanager.ha.enabled</name>
                    <value>true</value>
                </property>
                <!-- 指定RM的cluster id -->
                <property>
                    <name>yarn.resourcemanager.cluster-id</name>
                    <value>yrc</value>
                </property>
                <!-- 指定RM的名字 -->
                <property>
                    <name>yarn.resourcemanager.ha.rm-ids</name>
                    <value>rm1,rm2</value>
                </property>
                <!-- 分别指定RM的地址 -->
                <property>
                    <name>yarn.resourcemanager.hostname.rm1</name>
                    <value>uplooking02</value>
                </property>
                <property>
                    <name>yarn.resourcemanager.hostname.rm2</name>
                    <value>uplooking03</value>
                </property>
                <!-- 指定zk集群地址 -->
                <property>
                    <name>yarn.resourcemanager.zk-address</name>
                    <value>uplooking01:2181,uplooking02:2181,uplooking03:2181</value>
                </property>
                <property>
                    <name>yarn.nodemanager.aux-services</name>
                    <value>mapreduce_shuffle</value>
                </property>
            </configuration>
    4、准备hadoop所需要的几个目录
        [uplooking@uplooking01 hadoop]$ mkdir -p /home/uplooking/data/hadoop/journal
        [uplooking@uplooking01 hadoop]$ mkdir -p /home/uplooking/data/hadoop/name
        [uplooking@uplooking01 hadoop]$ mkdir -p /home/uplooking/data/hadoop/data
        [uplooking@uplooking01 hadoop]$ mkdir -p /home/uplooking/data/hadoop/tmp
    5、同步到uplooking02和uplooking03
            [uplooking@uplooking01 ~]$ scp -r data/hadoop uplooking@uplooking02:/home/uplooking/data/
            [uplooking@uplooking01 ~]$ scp -r data/hadoop uplooking@uplooking03:/home/uplooking/data/

            [uplooking@uplooking01 ~]$ scp -r app/hadoop uplooking@uplooking02:/home/uplooking/app/
            [uplooking@uplooking01 ~]$ scp -r app/hadoop uplooking@uplooking03:/home/uplooking/app/     
    6、格式化&启动
        1°、启动zk
        2°、启动jouralnode
            hadoop-deamon.sh start journalnode
        3°、在uplooking01或者uplooking02中的一台机器上面格式化hdfs
            hdfs namenode -format
                18/03/02 11:16:20 INFO common.Storage: Storage directory /home/uplooking/data/hadoop/name has been successfully formatted.
                说明格式化成功
            将格式化后的namenode的元数据信息拷贝到另外一台namenode之上就可以了
            将uplooking01上面产生的namenode的元数据信息，拷贝到uplooking02上面，
            scp -r /home/uplooking/data/hadoop/name uplooking@uplooking02:/home/uplooking/data/hadoop/
        4°、格式化zkfc
            hdfs zkfc -formatZK
            实际上是在zookeeper中创建一个目录节点/hadoop-ha/ns1
        5°、启动hdfs
            在uplooking01机器上面或者uplooking02上面启动、start-dfs.sh
        6、启动yarn
            在yarn配置的机器上面启动start-yarn.sh
            在uplooking02上面启动start-yarn.sh
            在uplooking03上面启动脚本
            yarn-daemon.sh start resourcemanager（在3上没有resourcemanager进程，需要手动启动一下）
            (hadoop的bug，在u2上启动yarn后，2上是有resourcemanager进程的，但是3上是没有的，所以3上面是需要手动启动的)
        7°、要启动hdfs中某一个节点，使用脚本hadoop-daemon.sh start 节点进程名

(
    Note：在保证已经格式化hdfs和zkfc后，可以直接使用start-dfs.sh start来启动，这时会依次启动：namenode datanode journalnode zkfc
Starting namenodes on [uplooking01 uplooking02]
uplooking01: starting namenode, logging to /home/uplooking/app/hadoop/logs/hadoop-uplooking-namenode-uplooking01.out
uplooking02: starting namenode, logging to /home/uplooking/app/hadoop/logs/hadoop-uplooking-namenode-uplooking02.out
uplooking03: starting datanode, logging to /home/uplooking/app/hadoop/logs/hadoop-uplooking-datanode-uplooking03.out
uplooking02: starting datanode, logging to /home/uplooking/app/hadoop/logs/hadoop-uplooking-datanode-uplooking02.out
Starting journal nodes [uplooking01 uplooking02 uplooking03]
uplooking03: starting journalnode, logging to /home/uplooking/app/hadoop/logs/hadoop-uplooking-journalnode-uplooking03.out
uplooking02: starting journalnode, logging to /home/uplooking/app/hadoop/logs/hadoop-uplooking-journalnode-uplooking02.out
uplooking01: starting journalnode, logging to /home/uplooking/app/hadoop/logs/hadoop-uplooking-journalnode-uplooking01.out
18/03/04 01:00:17 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 [uplooking01 uplooking02]
uplooking02: starting zkfc, logging to /home/uplooking/app/hadoop/logs/hadoop-uplooking-zkfc-uplooking02.out
uplooking01: starting zkfc, logging to /home/uplooking/app/hadoop/logs/hadoop-uplooking-zkfc-uplooking01.out
)

    7、访问和验证
        1°、访问
            web
                hdfs
                    http://uplooking01:50070
                    http://uplooking02:50070
                    其中一个是active，一个是standby
                yarn
                    http://uplooking02:8088
                    http://uplooking03:8088
                    在浏览的时候standby会重定向跳转之active对应的页面
            shell
                我们是无法操作standby对应的hdfs的，只能操作active的namenode
                    Operation category READ is not supported in state standby
        2、ha的验证
            NameNode HA
                访问：
                    uplooking01:50070
                    uplooking02:50070
                    其中一个active的状态，一个是StandBy的状态
                当访问standby的namenode时候：
                    Operation category READ is not supported in state standby

                    主备切换验证：
                        在uplooking01上kill -9 namenode的进程
                        这时访问uplooking02:50070发现变成了active的
                        然后在uplooking01上重新启动namenode，发现启动后状态变成standby的

            Yarn HA
                web访问：默认端口是8088
                    uplooking02:8088
                    uplooking03:8088
                        This is standby RM. Redirecting to the current active RM: http://uplooking02:8088/

                    主备切换验证：
                        在uplooking02上kill -9 resourcemanager的进程
                        这时可以访问uplooking03:8088
                        然后在uplooking02上重新启动resourcemanager，再访问时就是跳转到uplooking03:8088
            主备切换结论：
                原来的主再恢复时，为了系统的稳定性，不会再进行主备的切换。

        3、简单操作
            cd /home/uplooking/app/hadoop/share/hadoop/mapreduce
            [uplooking@uplooking01 mapreduce]$ yarn jar hadoop-mapreduce-examples-2.6.4.jar wordcount /hello /output/mr/wc

在启动hdfs时，会发现datanode一直起不来，查看日志，会有类似如下的输出：

INFO org.apache.hadoop.hdfs.server.datanode.DataNode: Removed Block pool <registering> (Datanode Uuid unassigned)

那是因为datanode的clusterID 和 namenode的clusterID 不匹配。
显然是在格式化hdfs时出现了问题，可以考虑把相关data目录下的相关数据删除掉，重新格式化。当然针对这个问题，也有下面的解决方法：
将name/current下的VERSION中的clusterID复制到data/current下的VERSION中，覆盖掉原来的clusterID