三节点Hadoop集群搭建

Posted 子墨言良

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了三节点Hadoop集群搭建相关的知识,希望对你有一定的参考价值。

1. 基础环境搭建

    新建3个CentOS6.5操作系统的虚拟机,命名(可自定)为masternode、slavenode1和slavenode2。该过程参考上一篇博文CentOS6.5安装配置详解

2.Hadoop集群搭建(以下操作中三个节点相同的地方就只给出主节点的截图,不同的才给出所有节点的截图)

  2.1 系统时间同步

    使用date命令查看当前系统时间

技术分享图片

    系统时间同步

[[email protected] ~]# cd /usr/share/zoneinfo/
[[email protected] zoneinfo]# ls		                                                //找到Asia
[[email protected] zoneinfo]# cd Asia/		                                        //进入Asia目录
[[email protected] Asia]# ls		                                                //找到Shanghai
[[email protected] Asia]# cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime		//当前时区替换为上海

    我们可以同步当前系统时间和日期与NTP(网络时间协议)一致。

[[email protected] Asia]# yum install ntp                                                 //如果ntp命令不存在,在线安装ntp
[[email protected] Asia]# ntpdate pool.ntp.org		                                //执行此命令同步日期时间

  分别在masternode、slavenode1和slavenode2节点内新建hadoop用户组和用户,专用于Hadoop集群的操作和管理。命令如下:

[[email protected] ~]# groupadd hadoop
[[email protected] ~]# useradd -g hadoop hadoop

    创建结果如下:

技术分享图片

   然后执行命令创建密码。注意:此过程你所输入的内容是不可见的,但其实已经输入了。还有,不可以删除。

[[email protected] hadoop]# passwd hadoop

  2.2 目录规划

    下面首先为这三台机器分配IP地址及相应的角色

    192.168.86.135-----master,namenode,jobtracker

    192.168.86.136-----slave1,datanode,tasktracker

    192.168.86.137-----slave2,datanode,tasktracker

    在所有节点的hosts文件中添加静态IP与hostname的映射配置信息。

[[email protected] ~]# vi /etc/hosts

技术分享图片

    然后依次对master、slave1、slave2进行目录规划。

                        名称                                                               路径       

                        所有集群安装的软件目录                                  /home/hadoop/app/

                        所有临时目录                                                   /tmp

    系统默认的临时目录是在/tmp下,而这个目录在每次重启后都会被删掉,必须重新执行format才行,否则会出错。

  2.3 禁用防火墙

    所有节点的防火墙都要关闭。查看防火墙状态:

[[email protected] ~]# service iptables status
iptables: Firewall is not running.

    如果不是上面的关闭状态,则需要关闭防火墙。

[[email protected] ~]# chkconfig iptables off		//永久关闭防火墙
[[email protected] ~]# service iptables stop		//临时关闭防火墙

技术分享图片

  2.4 SSH免密通信配置

[[email protected] ~]$ su root                                            //切换到hadoop用户下
Password: 
[[email protected] hadoop]# su hadoop                                       //切换到hadoop用户目录
[[email protected] ~]$ mkdir .ssh
mkdir: cannot create directory `.ssh‘: File exists                        //我的已经存在,不影响,继续下面的操作
[[email protected] ~]$ ssh-keygen -t rsa                                  //执行命令一路回车,生成秘钥
Generating public/private rsa key pair.
Enter file in which to save the key (/home/hadoop/.ssh/id_rsa): 
Enter passphrase (empty for no passphrase): 
Enter same passphrase again: 
Your identification has been saved in /home/hadoop/.ssh/id_rsa.
Your public key has been saved in /home/hadoop/.ssh/id_rsa.pub.
The key fingerprint is:
4b:a8:30:35:0e:cc:82:3f:1b:78:81:9c:e2:ee:ca:7d [email protected]
The key‘s randomart image is:
+--[ RSA 2048]----+                                                       //生成的密钥图像
|                 |
|o+.              |
|=o= o            |
|o+ = . .         |
|..B . . S        |
|.. * . . .       |
| .. .   .        |
|o .  E           |
|oo ..            |
+-----------------+
[[email protected] ~]$ cd .ssh
[[email protected] .ssh]$ ls
id_rsa  id_rsa.pub
[[email protected] .ssh]$ cat id_rsa.pub >> authorized_keys               //将公钥保存到authorized_keys认证文件中
[[email protected] .ssh]$ ls
authorized_keys  id_rsa  id_rsa.pub
[[email protected] .ssh]$ cd ..
[[email protected] ~]$ chmod 700 .ssh
[[email protected] ~]$ chmod 600 .ssh/*
[[email protected] ~]$ ssh masternode
The authenticity of host ‘masternode (192.168.86.135)‘ can‘t be established.
RSA key fingerprint is 45:13:ab:81:3a:53:44:2b:59:8f:06:fb:56:2f:b6:d8.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added ‘masternode,192.168.86.135‘ (RSA) to the list of known hosts.
Last login: Tue Apr 17 14:16:46 2018 from 192.168.86.1
[[email protected] ~]$ ssh masternode
Last login: Tue Apr 17 15:45:44 2018 from masternode

    集群所有节点都要行上面的操作,然后将所有节点中的共钥id_ras.pub拷贝到masternode中的authorized_keys文件中。

[[email protected] ~]$ cat ~/.ssh/id_rsa.pub | ssh [email protected] ‘cat >> ~/.ssh/authorized_keys‘         
    //所有节点都需要执行这条命令

        再将masternode中的authorized_keys文件分发到所有节点上面。

[[email protected] ~]$ cd .ssh
[[email protected] .ssh]$ ls
authorized_keys  id_rsa  id_rsa.pub  known_hosts
[[email protected] .ssh]$ scp -r authorized_keys [email protected]:~/.ssh/
[email protected]‘s password: 
authorized_keys                                                100% 1596     1.6KB/s   00:00    
[[email protected] .ssh]$ scp -r authorized_keys [email protected]:~/.ssh/
[email protected]‘s password: 
authorized_keys                                                100% 1596     1.6KB/s   00:00    

    到此,集群的SSH免密通信就配置完成了。

  2.5 脚本工具

    在masternode节点上创建/home/hadoop/tools目录。

[[email protected] ~]$ mkdir /home/hadoop/tools
[[email protected] ~]$ cd /home/hadoop/tools

    将本地脚本文件上传至/home/hadoop/tools目录下。这些脚本大家如果能看懂也可以自己写, 如果看不懂直接使用就可以,后面慢慢补补Linux相关的知识。

    先创建脚本文件,然后分别填入下面内容:

[[email protected] tools]$ touch deploy.conf
[[email protected] ~]$ vi deploy.conf
    masternode,all,namenode,zookeeper,resourcemanager,
    slavenode1,all,slave,namenode,zookeeper,resourcemanager,
    slavenode2,all,slave,datanode,zookeeper,
[[email protected] tools]$ touch deploy.sh
[[email protected] ~]$ vi deploy.sh
#!/bin/bash
#set -x

if [ $# -lt 3 ]
then 
  echo "Usage: ./deply.sh srcFile(or Dir) descFile(or Dir) MachineTag"
  echo "Usage: ./deply.sh srcFile(or Dir) descFile(or Dir) MachineTag confFile"
  exit 
fi

src=$1
dest=$2
tag=$3
if [ ‘a‘$4‘a‘ == ‘aa‘ ]
then
  confFile=/home/hadoop/tools/deploy.conf
else 
  confFile=$4
fi

if [ -f $confFile ]
then
  if [ -f $src ]
  then
    for server in `cat $confFile|grep -v ‘^#‘|grep ‘,‘$tag‘,‘|awk -F‘,‘ ‘{print $1}‘` 
    do
       scp $src $server":"${dest}
    done 
  elif [ -d $src ]
  then
    for server in `cat $confFile|grep -v ‘^#‘|grep ‘,‘$tag‘,‘|awk -F‘,‘ ‘{print $1}‘` 
    do
       scp -r $src $server":"${dest}
    done 
  else
      echo "Error: No source file exist"
  fi

else
  echo "Error: Please assign config file or run deploy.sh command with deploy.conf in same directory"
fi

[[email protected] tools]$ touch runRemoteCmd.sh
[[email protected] ~]$ vi runRemoteCmd.sh#!/bin/bash
#set -x

if [ $# -lt 2 ]
then 
  echo "Usage: ./runRemoteCmd.sh Command MachineTag"
  echo "Usage: ./runRemoteCmd.sh Command MachineTag confFile"
  exit 
fi

cmd=$1
tag=$2
if [ ‘a‘$3‘a‘ == ‘aa‘ ]
then
  
  confFile=/home/hadoop/tools/deploy.conf
else 
  confFile=$3
fi

if [ -f $confFile ]
then
    for server in `cat $confFile|grep -v ‘^#‘|grep ‘,‘$tag‘,‘|awk -F‘,‘ ‘{print $1}‘` 
    do
       echo "*******************$server***************************"
       ssh $server "source /etc/profile; $cmd"
    done 
else
  echo "Error: Please assign config file or run deploy.sh command with deploy.conf in same directory"
fi

    如果我们想直接使用脚本,还需要给脚本添加执行权限。

[[email protected] tools]$ chmod u+x deploy.sh
[[email protected] tools]$ chmod u+x runRemoteCmd.sh

    此时,我们需要将/home/hadoop/tools目录配置到PATH路径中,并使配置文件生效。

[[email protected] tools]$ su root
Password:
[[email protected] tools]# vi /etc/profile
    PATH=/home/hadoop/tools:$PATH
    export PATH
[[email protected] app]# source /etc/profile 

    我们在masternode节点上,通过runRemoteCmd.sh脚本,一键创建所有节点的软件安装目录/home/hadoop/app。

[[email protected] tools]$ runRemoteCmd.sh "mkdir /home/hadoop/app" all

    我们可以在所有节点查看到/home/hadoop/app目录已经创建成功。

  2.6 JDK安装与配置

    将本地下载好的jdk1.7,上传至hadoop11节点下的/home/hadoop/app目录并解压。

[[email protected] ~]$ cd /home/hadoop/app/
[[email protected] app]$ rz

[[email protected] app]$ ls
jdk-8u60-linux-x64.tar.gz
[[email protected] app]$ tar zxvf jdk-8u60-linux-x64.tar.gz             //解压
[[email protected] app]$ ls
jdk1.8.0_60  jdk-8u60-linux-x64.tar.gz
[[email protected] app]$ rm -f jdk-8u60-linux-x64.tar.gz                //删除安装包

    然后,添加JDK环境变量。

[[email protected] app]$ su root
Password: 
[[email protected] app]# vi /etc/profile
    JAVA_HOME=/home/hadoop/app/jdk1.8.0_60
    CLASSPATH=.:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar
    PATH=$JAVA_HOME/bin:/home/hadoop/tools:$PATH                        //黑色字体为2.5中脚本工具的配置信息
    export JAVA_HOME CLASSPATH PATH
[[email protected] app]# source /etc/profile
[[email protected] app]# java -version
java version "1.8.0_60"
Java(TM) SE Runtime Environment (build 1.8.0_60-b27)
Java HotSpot(TM) 64-Bit Server VM (build 25.60-b23, mixed mode)  

    slavenode1和slavenode2节点重复masternode节点上的jdk配置即可。

  2.7 Zookeeper安装与配置

    将本地下载好的zookeeper-3.4.6.tar.gz安装包,上传至masternode节点下的/home/hadoop/app目录下。

[[email protected] app]# su hadoop                                      
[[email protected] app]$ rz                                            //选择本地下载好的zookeeper-3.4.6.tar.gz

[[email protected] app]$ ls
jdk1.8.0_60  zookeeper-3.4.5-cdh5.10.0.tar.gz    
[[email protected] app]$ //重命名
[[email protected] app]$ ls
jdk1.8.0_60  zookeeper

    修改Zookeeper中的配置文件,一定注意将下面配置信息中的所有中文注释去掉,否则编码会出错导致无法启动zookeeper,以后也是,配置中尽量不要出现中午和字符(空格,tab等)!

[[email protected] app]$ cd /home/hadoop/app/zookeeper/conf/
[[email protected] conf]$ ls
configuration.xsl  log4j.properties  zoo_sample.cfg
[[email protected] conf]$ cp zoo_sample.cfg zoo.cfg                //复制生成zoo.cfg文件
[[email protected] conf]$ vi zoo.cfg 
    dataDir=/home/hadoop/data/zookeeper/zkdata                     //数据文件目录
    dataLogDir=/home/hadoop/data/zookeeper/zkdatalog                 //日志目录
    # the port at which the clients will connect
    clientPort=2181                                                  //默认端口号
    #server.服务编号=主机名称:Zookeeper不同节点之间同步和通信的端口:选举端口(选举leader)
    server.1=masternode:2888:3888
    server.2=slavenode1:2888:3888
    server.3=slavenode2:2888:3888

    通过远程命令远程拷贝命令scp将Zookeeper安装目录拷贝到其他节点上面。

[[email protected] zookeeper]# scp -r zookeeper slavenode1:/home/hadoop/app
[[email protected] zookeeper]# scp -r zookeeper slavenode2:/home/hadoop/app

    通过远程命令runRemoteCmd.sh在所有的节点上面创建目录:

[[email protected] app]$ runRemoteCmd.sh "mkdir -p /home/hadoop/data/zookeeper/zkdata" all
*******************masternode***************************
*******************slavenode1***************************
mkdir: cannot create directory `/home/hadoop/data/zookeeper‘: Permission denied
*******************slavenode2***************************
mkdir: cannot create directory `/home/hadoop/data/zookeeper‘: Permission denied

      结果出现访问拒绝命令,这是由于用户组权限问题,应该是我们之前创建的data/目录是属于root用户组的,我们需要赋予权限给hadoop用户组。

[[email protected] hadoop]$ chown -R hadoop:hadoop data

    接下来就可以成功创建目录了:

[[email protected] tools]$ runRemoteCmd.sh "mkdir -p /home/hadoop/data/zookeeper/zkdata" all
*******************masternode***************************
*******************slavenode1***************************
*******************slavenode2***************************
[hado[email protected] tools]$ runRemoteCmd.sh "mkdir -p /home/hadoop/data/zookeeper/zkdatalog" all
*******************masternode***************************
*******************slavenode1***************************
*******************slavenode2***************************

    然后分别在masternode、slavenode1和slavenode2上面,进入zkdata目录下,创建文件myid,里面的内容分别填充为:1、2、3, 这里我们以masternode为例。

[[email protected] tools]$ cd /home/hadoop/data/zookeeper/zkdata
[[email protected] zkdata]$ vi myid
    1

    配置Zookeeper环境变量。

[[email protected] zkdata]$ su root
Password: 
[[email protected] zookeeper]# vi /etc/profile
    TOOL_HOME=/home/hadoop/tools
    JAVA_HOME=/home/hadoop/app/jdk1.8.0_60
    ZOOKEEPER_HOME=/home/hadoop/app/zookeeper
    CLASSPATH=.:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar
    PATH=$JAVA_HOME/bin:$ZOOKEEPER_HOME/bin:$TOOLO_HOME:$PATH
    export JAVA_HOME CLASSPATH PATH ZOOKEEPER_HOME
[[email protected] zookeeper]# source /etc/profile                            //使配置生效

    在masternode节点上面启动所有节点的Zookeeper并查看状态。

[[email protected] ~]$ cd /home/hadoop/tools/
[[email protected] tools]$ runRemoteCmd.sh "/home/hadoop/app/zookeeper/bin/zkServer.sh start" zookeeper
*******************masternode***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
*******************slavenode1***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
*******************slavenode2***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
[[email protected] tools]$ runRemoteCmd.sh "/home/hadoop/app/zookeeper/bin/zkServer.sh status" zookeeper
*******************masternode***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Mode: follower
*******************slavenode1***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Mode: leader                                                            //leader节点是通过zookeeper的leader选举算法决定的,和启动顺序有关,
                                                                        //正常启动时第一个启动的就是leade;如果该节点挂掉则根据算法再选举另一个节点作为leader节点。
*******************slavenode2***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Mode: follower

   2.8 Hadoop环境配置

    将下载好的apache hadoop-2.6.0.tar.gz安装包,上传至masternode节点下的/home/hadoop/app目录下,然后解压。

[[email protected] tools]$ cd /home/hadoop/app                 //将本地的hadoop-2.6.0.tar.gz安装包上传至当前目录
[[email protected] app]$ rz

[[email protected] app]$ tar zvxf hadoop-2.6.0.tar.gz          //解压
[[email protected] app]$ ls
hadoop-2.6.0  hadoop-2.6.0.tar.gz  jdk1.8.0_60  zookeeper
[[email protected] app]$ rm -f hadoop-2.6.0.tar.gz             //删除安装包
[[email protected] app]$ mv hadoop-2.6.0/ hadoop               //重命名
[[email protected] app]$ ls
hadoop  jdk1.8.0_60  zookeeper

    配置HDFS 

    切换到/home/hadoop/app/hadoop/etc/hadoop/目录下,修改配置文件。

[[email protected] app]$ cd /home/hadoop/app/hadoop/etc/hadoop/
[[email protected] hadoop]$ ls
capacity-scheduler.xml      httpfs-env.sh            mapred-env.sh
configuration.xsl           httpfs-log4j.properties  mapred-queues.xml.template
container-executor.cfg      httpfs-signature.secret  mapred-site.xml.template
core-site.xml               httpfs-site.xml          slaves
hadoop-env.cmd              kms-acls.xml             ssl-client.xml.example
hadoop-env.sh               kms-env.sh               ssl-server.xml.example
hadoop-metrics2.properties  kms-log4j.properties     yarn-env.cmd
hadoop-metrics.properties   kms-site.xml             yarn-env.sh
hadoop-policy.xml           log4j.properties         yarn-site.xml
hdfs-site.xml               mapred-env.cmd

    配置hadoop-env.sh文件

[[email protected] hadoop]$ vi hadoop-env.sh
 
export JAVA_HOME=/home/hadoop/app/jdk1.8.0_60

    配置core-site.xml文件

[[email protected] hadoop]$ vi core-site.xml 
<configuration> <property> <name>fs.defaultFS</name> <value>hdfs://cluster1</value> </property>     
        < 这里的值指的是默认的HDFS路径 ,取名为cluster1>
      <property> <name>hadoop.tmp.dir</name>
                <value>/home/hadoop/data/tmp</value>
         </property>
         < hadoop的临时目录,如果需要配置多个目录,需要逗号隔开,data目录需要我们自己创建>
         <property>
             <name>ha.zookeeper.quorum</name>
             <value>masternode:2181,slavenode1:2181,slavenode2:2181</value>
         </property>
       < 配置Zookeeper 管理HDFS>
</configuration>

    配置hdfs-site.xml文件

[[email protected] hadoop]$ vi hdfs-site.xm
<configuration>
	<property>
		<name>dfs.replication</name>
		<value>3</value>
    </property>
	< 数据块副本数为3>
	<property>
		<name>dfs.permissions</name>
		<value>false</value>
	</property>
	<property>
		<name>dfs.permissions.enabled</name>
		<value>false</value>
	</property>
	< 权限默认配置为false>
	<property>
		<name>dfs.nameservices</name>
		<value>cluster1</value>
	</property>
	< 命名空间,它的值与fs.defaultFS的值要对应,namenode高可用之后有两个namenode,cluster1是对外提供的统一入口>
	<property>
		<name>dfs.ha.namenodes.cluster1</name>
		<value>masternode,slavenode1</value>
	</property>
	< 指定 nameService 是 cluster1 时的nameNode有哪些,这里的值也是逻辑名称,名字随便起,相互不重复即可>
	<property>
		<name>dfs.namenode.rpc-address.cluster1.masternode</name>
		<value>masternode:9000</value>
	</property>
	< masternode rpc地址>
	<property>
		<name>dfs.namenode.http-address.cluster1.masternode</name>
		<value>masternode:50070</value>
	</property>
	< masternode http地址>
	<property>
		<name>dfs.namenode.rpc-address.cluster1.slavenode1</name>
		<value>slavenode1:9000</value>
	</property>
	< slavenode1 rpc地址>
	<property>
		<name>dfs.namenode.http-address.cluster1.slavenode1</name>
		<value>slavenode1:50070</value>
	</property>
	< slavenode2 http地址>
	<property>
		<name>dfs.ha.automatic-failover.enabled</name>
		<value>true</value>
    </property>
	< 启动故障自动恢复>
	<property>
		<name>dfs.namenode.shared.edits.dir</name>
		<value>qjournal://masternode:8485;slavenode1:8485;slavenode2:8485/cluster1</value>
	</property>
	< 指定journal>
	<property>
		<name>dfs.client.failover.proxy.provider.cluster1</name>
		<value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
    </property>
	< 指定 cluster1 出故障时,哪个实现类负责执行故障切换>
    <property>
		<name>dfs.journalnode.edits.dir</name>
		<value>/home/hadoop/data/journaldata/jn</value>
    </property>
	< 指定JournalNode集群在对nameNode的目录进行共享时,自己存储数据的磁盘路径 >
	<property>
		<name>dfs.ha.fencing.methods</name>
		<value>shell(/bin/true)</value>
    </property>
    <property>
        <name>dfs.ha.fencing.ssh.private-key-files</name>
        <value>/home/hadoop/.ssh/id_rsa</value>
    </property>
	<property>
        <name>dfs.ha.fencing.ssh.connect-timeout</name>
        <value>10000</value>
    </property>
	< 脑裂默认配置>
    <property>
		<name>dfs.namenode.handler.count</name>
		<value>100</value>
    </property>
</configuration>

    配置slave文件

[[email protected] hadoop]$ vi slaves
    slavenode2

    向所有节点分发hadoop安装包。

[[email protected] app]# scp -r zookeeper slavenode1:/home/hadoop/app
[[email protected] app]# scp -r zookeeper slavenode2:/home/hadoop/app

   hdfs配置完毕后的启动顺序

    1)启动所有节点上面的Zookeeper进程

[[email protected] app]$ cd /home/hadoop/tools/
[[email protected] tools]$ runRemoteCmd.sh "/home/hadoop/app/zookeeper/bin/zkServer.sh start" zookeeper
*******************masternode***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
*******************slavenode1***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
*******************slavenode2***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
[[email protected] tools]$ runRemoteCmd.sh "/home/hadoop/app/zookeeper/bin/zkServer.sh status" zookeeper
*******************masternode***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Mode: follower
*******************slavenode1***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Mode: follower
*******************slavenode2***************************
JMX enabled by default
Using config: /home/hadoop/app/zookeeper/bin/../conf/zoo.cfg
Mode: leader
[[email protected] hadoop]$ jps   6560 Jps
6459 QuorumPeerMain

     其中,QuorumPeerMain对应zookeeper的进程。

     2)启动所有节点上面的journalnode进程

[[email protected] tools]$ runRemoteCmd.sh "/home/hadoop/app/hadoop/sbin/hadoop-daemon.sh start journalnode" all
*******************masternode***************************
starting journalnode, logging to /home/hadoop/app/hadoop/logs/hadoop-hadoop-journalnode-masternode.out
*******************slavenode1***************************
starting journalnode, logging to /home/hadoop/app/hadoop/logs/hadoop-hadoop-journalnode-slavenode1.out
*******************slavenode2***************************
starting journalnode, logging to /home/hadoop/app/hadoop/logs/hadoop-hadoop-journalnode-slavanode2.out
[[email protected] tools]$ jps
6672 Jps
6624 JournalNode
6459 QuorumPeerMain

    或者在每个节点上使用以下命令分别启动

[[email protected] hadoop]$ sbin/hadoop-daemon.sh start journalnode

    3)首先在主节点上(比如,masterndoe)执行格式化并启动Namenode

[[email protected] hadoop]$ bin/hdfs namenode -format              //namenode 格式化
[[email protected] hadoop]$ bin/hdfs zkfc -formatZK                //格式化高可用
[[email protected] hadoop]$ bin/hdfs namenode                      //启动namenode
		

    4)与此同时,需要在备节点(比如,slavenode1)上执行数据同步

[[email protected] hadoop]$ bin/hdfs namenode -bootstrapStandby	    //同步主节点和备节点之间的元数据,

    5)slavenode1同步完数据后,紧接着在masterndoe节点上,按下ctrl+c来结束namenode进程。 然后关闭所有节点上面的journalnode进程

[[email protected] hadoop]$ runRemoteCmd.sh "/home/hadoop/app/hadoop/sbin/hadoop-daemon.sh stop journalnode" all	
                                                                    
//然后停掉各节点的journalnode
[[email protected] hadoop]$ jps
6842 Jps
6459 QuorumPeerMain

    6)如果上面操作没有问题,我们可以一键启动hdfs所有相关进程

[[email protected] hadoop]$ sbin/start-dfs.sh
[[email protected]masternode hadoop]$ jps
8640 DFSZKFailoverController
8709 Jps
6459 QuorumPeerMain
8283 NameNode
8476 JournalNode
[[email protected]slavenode1 hadoop]$ jps
5667 DFSZKFailoverController
5721 Jps
5562 JournalNode
4507 QuorumPeerMain
5487 NameNode
[[email protected]slavanode2 hadoop]$ jps
5119 Jps
5040 JournalNode
5355 DataNode
4485 QuorumPeerMain

     以上masternode和slavenode1是作为NameNode的,而slavenode2则作为DataNode。

    验证是否启动成功,通过web界面查看namenode启动情况。

http://masternode:50070

  技术分享图片

技术分享图片

    如图,masternode节点状态为active,而slavenode1节点状态为standby。

    7)测试集群是否可以正常使用

    使用以下命令在HDFS中新建一个文件夹,然后再网页上通过文件系统查看。

[[email protected] hadoop]$ hdfs dfs -mkdir /test

技术分享图片

     还可以上传文件到文件夹内,这个可以自己下去测试一下。

    在这里,我想说的是,哪个是active,哪个是standby是随机的 ,这是由选举决定的。

    下面我们来试一下将slavenode1节点变为active。

    首先kill掉masternode节点的Namenode,然后刷新网页看看有声明变化。

[[email protected] hadoop]$ jps
8640 DFSZKFailoverController
8901 Jps
6459 QuorumPeerMain
8283 NameNode
8476 JournalNode
[[email protected] hadoop]$ kill -9 8283
[[email protected]masternode hadoop]$ jps
8640 DFSZKFailoverController
8916 Jps
6459 QuorumPeerMain
8476 JournalNode
[[email protected]slavenode1 hadoop]$ jps
5986 Jps
5667 DFSZKFailoverController
5562 JournalNode
4507 QuorumPeerMain
5487 NameNode

技术分享图片

    如图,slavenode1节点变为了active状态!刚才将masternode的Namenode kill掉了,所以根据选举算法,slavenode1节点被选举为Namenode节点,所以状态为active。

   2.9 YARN安装配置

    配置mapred-site.xml

[[email protected] hadoop]$ cp mapred-site.xml.template mapred-site.xml
[[email protected] hadoop]$ vi mapred-site.xml
<configuration>
	<property>
		<name>mapreduce.framework.name</name>
		<value>yarn</value>
	</property>
	<指定运行mapreduce的环境是Yarn,与hadoop1不同的地方>
</configuration>

    配置yarn-site.xml

[[email protected] hadoop]$ vi yarn-site.xml
<configuration> <property> <name>yarn.resourcemanager.connect.retry-interval.ms</name> <value>2000</value> </property> < 超时的周期> <property> <name>yarn.resourcemanager.ha.enabled</name> <value>true</value> </property> < 打开高可用> <property> <name>yarn.resourcemanager.ha.automatic-failover.enabled</name> <value>true</value> </property> <启动故障自动恢复> <property> <name>yarn.resourcemanager.ha.automatic-failover.embedded</name> <value>true</value> </property> <failover使用内部的选举算法> <property> <name>yarn.resourcemanager.cluster-id</name> <value>yarn-rm-cluster</value> </property> <给yarn cluster 取个名字yarn-rm-cluster> <property> <name>yarn.resourcemanager.ha.rm-ids</name> <value>rm1,rm2</value> </property> <给ResourceManager 取个名字 rm1,rm2> <property> <name>yarn.resourcemanager.hostname.rm1</name> <value>masternode</value> </property> <配置ResourceManagerrm1hostname> <property> <name>yarn.resourcemanager.hostname.rm2</name> <value>slavenode1</value> </property> <配置ResourceManagerrm2hostname> <property> <name>yarn.resourcemanager.recovery.enabled</name> <value>true</value> </property> <启用resourcemanager 自动恢复> <property> <name>yarn.resourcemanager.zk.state-store.address</name> <value>masternode:2181,slavenode1:2181,slavenode2:2181</value> </property> <配置Zookeeper地址> <property> <name>yarn.resourcemanager.zk-address</name> <value>masternode:2181,slavenode1:2181,slavenode2:2181</value> </property> <配置Zookeeper地址> <property> <name>yarn.resourcemanager.address.rm1</name> <value>masternode:8032</value> </property> < rm1端口号> <property> <name>yarn.resourcemanager.scheduler.address.rm1</name> <value>masternode:8034</value> </property> < rm1调度器的端口号> <property> <name>yarn.resourcemanager.webapp.address.rm1</name> <value>masternode:8088</value> </property> < rm1webapp端口号> <property> <name>yarn.resourcemanager.address.rm2</name> <value>slavenode1:8032</value> </property> < rm2端口号> <property> <name>yarn.resourcemanager.scheduler.address.rm2</name> <value>slavenode1:8034</value> </property> < rm2调度器的端口号> <property> <name>yarn.resourcemanager.webapp.address.rm2</name> <value>slavenode1:8088</value> </property> < rm2webapp端口号> <property> <name>yarn.nodemanager.aux-services</name> <value>mapreduce_shuffle</value> </property> <property> <name>yarn.nodemanager.aux-services.mapreduce_shuffle.class</name> <value>org.apache.hadoop.mapred.ShuffleHandler</value> </property> <执行MapReduce需要配置的shuffle过程> </configuration>

   启动YARN

   1) 将yarn-site.xml文件发送到slavenode1和slavenode2节点上。

[[email protected] hadoop]$ scp yarn-site.xml slavenode1:/home/hadoop/app/hadoop/etc/hadoop/
yarn-site.xml                                                  100% 2782     2.7KB/s   00:00    
[[email protected] hadoop]$ scp yarn-site.xml slavenode2:/home/hadoop/app/hadoop/etc/hadoop/
yarn-site.xml                                                  100% 2782     2.7KB/s   00:00    

    2)在masternode节点上执行。

[[email protected] hadoop]$ sbin/start-yarn.sh 
starting yarn daemons
starting resourcemanager, logging to /home/hadoop/app/hadoop/logs/yarn-hadoop-resourcemanager-masternode.out
slavenode2: starting nodemanager, logging to /home/hadoop/app/hadoop/logs/yarn-hadoop-nodemanager-slavanode2.out
[[email protected] hadoop]$ jps
8640 DFSZKFailoverController
8969 ResourceManager
6459 QuorumPeerMain
8476 JournalNode
9054 Jps

      YARN对应的进程为ResourceManager。

    3)在slavenode1节点上执行。

[[email protected] hadoop]$ sbin/start-yarn.sh 
starting yarn daemons
starting resourcemanager, logging to /home/hadoop/app/hadoop/logs/yarn-hadoop-resourcemanager-slavenode1.out
[[email protected] hadoop]$ jps
5667 DFSZKFailoverController
5562 JournalNode
4507 QuorumPeerMain
6059 ResourceManager
6127 Jps
5487 NameNode

    同时打开以下web界面。

http://hadoop11:8088
http://hadoop12:8088
		

    技术分享图片

技术分享图片

     检查一下ResourceManager状态

[[email protected] hadoop]$ bin/yarn rmadmin -getServiceState rm1
18/04/20 16:58:26 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
standby
[[email protected] hadoop]$ bin/yarn rmadmin -getServiceState rm2
18/04/20 16:58:33 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
active

    即slavenode1的ResourceManager,即rm1,是active;masternode的ResourceManager,即rm2,是standby;结论也与web页面吻合。关闭其中一个resourcemanager,然后再启动,同Namenode也能使两节点状态交换。

    那么,到此hadoop的3节点集群搭建完毕,我们使用zookeeper来管理hadoop集群,同时,实现了namenode热备和ResourceManager热备。

 

以上就是博主为大家介绍的这一板块的主要内容,这都是博主自己的学习过程,希望能给大家带来一定的指导作用,有用的还望大家点个支持,如果对你没用也望包涵,有错误烦请指出。如有期待可关注博主以第一时间获取更新哦,谢谢! 

 版权声明:本文为博主原创文章,未经博主允许不得转载。

 

 

以上是关于三节点Hadoop集群搭建的主要内容,如果未能解决你的问题,请参考以下文章

大数据之一:Hadoop2.6.5+centos7.5三节点大数据集群部署搭建

Hadoop完全分布式环境搭建

Hadoop集群搭建

搭建Hadoop分布式集群的详细教程

如何基于Docker快速搭建多节点Hadoop集群

Hadoop完全分布式集群搭建