CentOS 7下Hadoop完全分布式集群搭建

共计 10736 个字符，预计需要花费 27 分钟才能阅读完成。

Hadoop 集群部署，就是以 Cluster mode 方式进行部署。本文是基于 JDK1.7.0_79，hadoop2.7.5。

1.Hadoop 的节点构成如下：

HDFS daemon：NameNode, SecondaryNameNode, DataNode

YARN damones：ResourceManager, NodeManager, WebAppProxy

MapReduce Job History Server

本次测试的分布式环境为：Master 1 台 (test166)，Slave 1 台 (test167)

2.1 安装 JDK 及下载解压 hadoop

JDK 安装可参考：http://www.linuxidc.com/Linux/2017-01/139874.htm 或 CentOS7.2 安装 JDK1.7 http://www.linuxidc.com/Linux/2016-11/137398.htm

从官网下载 Hadoop 最新版 2.7.5

[hadoop@hadoop-master ~]$ su – hadoop
[hadoop@hadoop-master ~]$ cd /usr/hadoop/
[hadoop@hadoop-master ~]$ wget http://mirrors.shu.edu.cn/apache/hadoop/common/hadoop-2.7.5/hadoop-2.7.5.tar.gz

将 hadoop 解压到 /usr/hadoop/ 下

[hadoop@hadoop-master ~]$ tar zxvf /root/hadoop-2.7.5.tar.gz

结果：

[hadoop@hadoop-master ~]$ ll
total 211852
drwxr-xr-x.  2 hadoop hadoop        6 Jan 31 23:41 Desktop
drwxr-xr-x.  2 hadoop hadoop        6 Jan 31 23:41 Documents
drwxr-xr-x.  2 hadoop hadoop        6 Jan 31 23:41 Downloads
drwxr-xr-x. 10 hadoop hadoop      4096 Feb 22 01:36 hadoop-2.7.5
-rw-rw-r–.  1 hadoop hadoop 216929574 Dec 16 12:03 hadoop-2.7.5.tar.gz
drwxr-xr-x.  2 hadoop hadoop        6 Jan 31 23:41 Music
drwxr-xr-x.  2 hadoop hadoop        6 Jan 31 23:41 Pictures
drwxr-xr-x.  2 hadoop hadoop        6 Jan 31 23:41 Public
drwxr-xr-x.  2 hadoop hadoop        6 Jan 31 23:41 Templates
drwxr-xr-x.  2 hadoop hadoop        6 Jan 31 23:41 Videos
[hadoop@hadoop-master ~]$

2.2 在各节点上设置主机名及创建 hadoop 组和用户

所有节点（master,slave）

1 [root@hadoop-master ~]#  su – root
2 [root@hadoop-master ~]#  vi /etc/hosts
3 10.86.255.166 hadoop-master
4 10.86.255.167 slave1
5 注意：修改 hosts 中，是立即生效的，无需 source 或者.。

先使用

建立 hadoop 用户组

新建用户，useradd -d /usr/hadoop -g hadoop -m hadoop（新建用户 hadoop 指定用户主目录 /usr/hadoop 及所属组 hadoop）

passwd hadoop 设置 hadoop 密码（这里设置密码为 hadoop）

[root@hadoop-master ~]# groupadd hadoop
[root@hadoop-master ~]# useradd -d /usr/hadoop -g hadoop -m hadoop
[root@hadoop-master ~]# passwd hadoop

2.3 在各节点上设置 SSH 无密码登录

最终达到目的：即在 master: 节点执行  ssh hadoop@salve1 不需要密码，此处只需配置 master 访问 slave1 免密。

su – hadoop

进入~/.ssh 目录

执行：ssh-keygen -t rsa，一路回车

生成两个文件，一个私钥，一个公钥，在 master1 中执行：cp id_rsa.pub authorized_keys

[hadoop@hadoop-master ~]$ su – hadoop
[hadoop@hadoop-master ~]$ pwd
/usr/hadoop
[hadoop@hadoop-master ~]$ cd .ssh
[hadoop@hadoop-master .ssh]$ ssh-keygen -t rsa
Generating public/private rsa key pair.
Enter file in which to save the key (/usr/hadoop/.ssh/id_rsa):
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
Your identification has been saved in /usr/hadoop/.ssh/id_rsa.
Your public key has been saved in /usr/hadoop/.ssh/id_rsa.pub.
The key fingerprint is:
11:b2:23:8c:e7:32:1d:4c:2f:00:32:1a:15:43:bb:de hadoop@hadoop-master
The key’s randomart image is:
+–[RSA 2048]—-+
|=+*.. . .        |
|oo O . o .      |
|. o B + .        |
|  = + . .      |
|  + o  S        |
| . +            |
|  . E            |
|                |
|                |
+—————–+
[hadoop@hadoop-master .ssh]$
[hadoop@hadoop-master .ssh]$ cp id_rsa.pub authorized_keys
[hadoop@hadoop-master .ssh]$ ll
total 16
-rwx——. 1 hadoop hadoop 1230 Jan 31 23:27 authorized_keys
-rwx——. 1 hadoop hadoop 1675 Feb 23 19:07 id_rsa
-rwx——. 1 hadoop hadoop  402 Feb 23 19:07 id_rsa.pub
-rwx——. 1 hadoop hadoop  874 Feb 13 19:40 known_hosts
[hadoop@hadoop-master .ssh]$

2.3.1：本机无密钥登录

[hadoop@hadoop-master ~]$ pwd
/usr/hadoop
[hadoop@hadoop-master ~]$ chmod -R 700 .ssh
[hadoop@hadoop-master ~]$ cd .ssh
[hadoop@hadoop-master .ssh]$ chmod 600 authorized_keys
[hadoop@hadoop-master .ssh]$ ll
total 16
-rwx——. 1 hadoop hadoop 1230 Jan 31 23:27 authorized_keys
-rwx——. 1 hadoop hadoop 1679 Jan 31 23:26 id_rsa
-rwx——. 1 hadoop hadoop  410 Jan 31 23:26 id_rsa.pub
-rwx——. 1 hadoop hadoop  874 Feb 13 19:40 known_hosts

验证：
没有提示输入密码则表示本机无密钥登录成功，如果此步不成功，后续启动 hdfs 脚本会要求输入密码

[hadoop@hadoop-master ~]$ ssh hadoop@hadoop-master
Last login: Fri Feb 23 18:54:59 2018 from hadoop-master
[hadoop@hadoop-master ~]$

2.3.2：master 与其他节点无密钥登录     

(若已有 authorized_keys，则执行 ssh-copy-id ssh-copy-id -i ~/.ssh/id_rsa.pub hadoop@slave1 上面命令的功能 ssh-copy-id 将 pub 值写入远程机器的~/.ssh/authorized_key 中)     

从 master 中把 authorized_keys 分发到各个结点上 (会提示输入密码，输入 slave1 相应的密码即可)：

scp /usr/hadoop/.ssh/authorized_keys hadoop@slave1:/home/master/.ssh
/usr/bin/ssh-copy-id: INFO: attempting to log in with the new key(s), to filter out any that are already installed
/usr/bin/ssh-copy-id: INFO: 1 key(s) remain to be installed — if you are prompted now it is to install the new keys
hadoop@slave1’s password:

 Number of key(s) added: 1

Now try logging into the machine, with:  “ssh ‘hadoop@slave1′” and check to make sure that only the key(s) you wanted were added.

[hadoop@hadoop-master .ssh]$

然后在各个节点对 authorized_keys 执行 (一定要执行该步，否则会报错)：chmod 600 authorized_keys

保证.ssh 700，.ssh/authorized_keys 600 权限

测试如下（第一次 ssh 时会提示输入 yes/no，输入 yes 即可）：

[hadoop@hadoop-master ~]$ ssh hadoop@slave1
Last login: Fri Feb 23 18:40:10 2018
[hadoop@slave1 ~]$
[hadoop@slave1 ~]$ exit
logout
Connection to slave1 closed.
[hadoop@hadoop-master ~]$

2.4 设置 Hadoop 的环境变量

Master 及 slave1 都需操作

[root@hadoop-master ~]#  su – root

[root@hadoop-master ~]#  vi /etc/profile  末尾添加，保证任何路径下可执行 hadoop 命令

Java_HOME=/usr/java/jdk1.7.0_79

CLASSPATH=.:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar

PATH=/usr/hadoop/hadoop-2.7.5/bin:$JAVA_HOME/bin:$PATH

让设置生效

[root@hadoop-master ~]# source /etc/profile

或者

[root@hadoop-master ~]# . /etc/profile

Master 设置 hadoop 环境

su – hadoop
1 # vi etc/hadoop/hadoop-env.sh  新增以下内容
2 export  JAVA_HOME=/usr/java/jdk1.7.0_79
3 export  HADOOP_HOME=/usr/hadoop/hadoop-2.7.5

此时 hadoop 安装已完成，可执行 hadoop 命令，后续步骤为集群部署

[hadoop@hadoop-master ~]$ hadoop
Usage: hadoop [–config confdir] [COMMAND | CLASSNAME]
  CLASSNAME            run the class named CLASSNAME
 or
  where COMMAND is one of:
  fs                  run a generic filesystem user client
  version              print the version
  jar <jar>            run a jar file
                      note: please use “yarn jar” to launch
                            YARN applications, not this command.
  checknative [-a|-h]  check native hadoop and compression libraries availability
  distcp <srcurl> <desturl> copy file or directories recursively
  archive -archiveName NAME -p <parent path> <src>* <dest> create a hadoop archive
  classpath            prints the class path needed to get the
  credential          interact with credential providers
                      Hadoop jar and the required libraries
  daemonlog            get/set the log level for each daemon
  trace                view and modify Hadoop tracing settings

Most commands print help when invoked w/o parameters.
[hadoop@hadoop-master ~]$

2.5 Hadoop 设定

2.5.0 开放端口 50070

注：centos7 版本对防火墙进行 加强, 不再使用原来的 iptables, 启用 firewall

Master 节点:

su – root

firewall-cmd –state 查看状态（若关闭，则先开启 systemctl start firewalld）

firewall-cmd –list-ports 查看已开放的端口

开启 8000 端口：firewall-cmd –zone=public(作用域) –add-port=8000/tcp(端口和访问类型) –permanent(永久生效)

firewall-cmd –zone=public –add-port=1521/tcp –permanent

firewall-cmd –zone=public –add-port=3306/tcp –permanent

firewall-cmd –zone=public –add-port=50070/tcp –permanent

firewall-cmd –zone=public –add-port=8088/tcp –permanent

firewall-cmd –zone=public –add-port=19888/tcp –permanent

firewall-cmd –zone=public –add-port=9000/tcp –permanent

firewall-cmd –zone=public –add-port=9001/tcp –permanent

firewall-cmd –reload - 重启防火墙

firewall-cmd –list-ports 查看已开放的端口

systemctl stop firewalld.service 停止防火墙

systemctl disable firewalld.service 禁止防火墙开机启动

关闭端口：firewall-cmd –zone= public –remove-port=8000/tcp –permanent

Slave1 节点：

su – root
systemctl stop firewalld.service 停止防火墙

systemctl disable firewalld.service 禁止防火墙开机启动

2.5.1 在 Master 节点的设定文件中指定 Slave 节点

[hadoop@hadoop-master hadoop]$ pwd
/usr/hadoop/hadoop-2.7.5/etc/hadoop
[hadoop@hadoop-master hadoop]$ vi slaves
slave1

2.5.2 在各节点指定 HDFS 文件存储的位置（默认是 /tmp）

Master 节点：namenode

创建目录并赋予权限

Su – root

# mkdir -p  /usr/local/hadoop-2.7.5/tmp/dfs/name

# chmod -R 777  /usr/local/hadoop-2.7.5/tmp

# chown  -R hadoop:hadoop /usr/local/hadoop-2.7.5

Slave 节点：datanode

创建目录并赋予权限，改变所有者

Su – root

# mkdir -p  /usr/local/hadoop-2.7.5/tmp/dfs/data

# chmod -R 777  /usr/local/hadoop-2.7.5/tmp

# chown -R  hadoop:hadoop /usr/local/hadoop-2.7.5

2.5.3 在 Master 中设置配置文件 (包括 yarn)

su – hadoop
# vi  etc/hadoop/core-site.xml

<configuration>

<property>

      <name>fs.default.name</name>

      <value>hdfs://hadoop-master:9000</value>

  </property>

<property>

        <name>hadoop.tmp.dir</name>

        <value>/usr/local/hadoop-2.7.5/tmp</value>

  </property>

</configuration>

# vi  etc/hadoop/hdfs-site.xml

<configuration>

<property>

      <name>dfs.replication</name>

    <value>3</value>

</property>

<property>

    <name>dfs.name.dir</name>

      <value>/usr/local/hadoop-2.7.5/tmp/dfs/name</value>

</property>

  <property>

    <name>dfs.data.dir</name>

    <value>/usr/local/hadoop-2.7.5/tmp/dfs/data</value>

  </property>

</configuration>

#cp mapred-site.xml.template  mapred-site.xml

# vi  etc/hadoop/mapred-site.xml

<configuration>

  <property>

      <name>mapreduce.framework.name</name>

      <value>yarn</value>

  </property>

</configuration>

YARN 设定

yarn 的组成（Master 节点：resourcemanager，Slave 节点：nodemanager）

以下仅在 master 操作, 后面步骤会统一分发至 salve1。

# vi  etc/hadoop/yarn-site.xml

<configuration>

  <property>

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

      <value>hadoop-master</value>

  </property>

  <property>

      <name>yarn.nodemanager.aux-services</name>

      <value>mapreduce_shuffle</value>

  </property>

</configuration>

2.5.4 将 Master 的文件分发至 slave1 节点。

cd /usr/hadoop

scp -r  hadoop-2.7.5 hadoop@hadoop-master:/usr/hadoop

2.5.5 Master 上启动 job history server，Slave 节点上指定

此步 2.5.5 可跳过

Mater:

启动 jobhistory daemon

# sbin/mr-jobhistory-daemon.sh start historyserver

确认

# jps

访问 Job History Server 的 web 页面

http://localhost:19888/

Slave 节点：

# vi  etc/hadoop/mapred-site.xml

  <property>

      <name>mapreduce.jobhistory.address</name>

      <value>hadoop-master:10020</value>

  </property>

2.5.6 格式化 HDFS（Master）

# hadoop namenode -format

Master 结果：

2.5.7 在 Master 上启动 daemon，Slave 上的服务会一起启动

启动:

[hadoop@hadoop-master hadoop-2.7.5]$ pwd
/usr/hadoop/hadoop-2.7.5[hadoop@hadoop-master hadoop-2.7.5]$ sbin/start-all.sh
This script is Deprecated. Instead use start-dfs.sh and start-yarn.sh
Starting namenodes on [hadoop-master]
hadoop-master: starting namenode, logging to /usr/hadoop/hadoop-2.7.5/logs/hadoop-hadoop-namenode-hadoop-master.out
slave1: starting datanode, logging to /usr/hadoop/hadoop-2.7.5/logs/hadoop-hadoop-datanode-slave1.out
Starting secondary namenodes [0.0.0.0]
0.0.0.0: starting secondarynamenode, logging to /usr/hadoop/hadoop-2.7.5/logs/hadoop-hadoop-secondarynamenode-hadoop-master.out
starting yarn daemons
starting resourcemanager, logging to /usr/hadoop/hadoop-2.7.5/logs/yarn-hadoop-resourcemanager-hadoop-master.out
slave1: starting nodemanager, logging to /usr/hadoop/hadoop-2.7.5/logs/yarn-hadoop-nodemanager-slave1.out
[hadoop@hadoop-master hadoop-2.7.5]$

确认

Master 节点：

[hadoop@hadoop-master hadoop-2.7.5]$ jps
81209 NameNode
81516 SecondaryNameNode
82052 Jps
81744 ResourceManager

Slave 节点：

[hadoop@slave1 ~]$ jps
58913 NodeManager
59358 Jps
58707 DataNode

停止（需要的时候再停止，后续步骤需 running 状态）:

[hadoop@hadoop-master hadoop-2.7.5]$ sbin/stop-all.sh
This script is Deprecated. Instead use stop-dfs.sh and stop-yarn.sh
Stopping namenodes on [hadoop-master]
hadoop-master: stopping namenode
slave1: stopping datanode
Stopping secondary namenodes [0.0.0.0]
0.0.0.0: stopping secondarynamenode
stopping yarn daemons
stopping resourcemanager
slave1: stopping nodemanager
no proxyserver to stop

2.5.8 创建 HDFS

# hdfs dfs -mkdir /user

# hdfs dfs -mkdir  /user/test22

2.5.9 拷贝 input 文件到 HDFS 目录下

# hdfs dfs -put etc/hadoop/*.sh  /user/test22/input

查看

# hdfs dfs -ls /user/test22/input

2.5.10 执行 hadoop job

统计单词的例子，此时的 output 是 hdfs 中的目录，hdfs dfs -ls 可查看

# hadoop jar  share/hadoop/mapreduce/hadoop-mapreduce-examples-2.7.5.jar wordcount  /user/test22/input  output

确认执行结果

# hdfs dfs -cat output/*

2.5.11 查看错误日志

注：日志在 salve1 的 *.log 中而不是在 master 或 *.out 中

2.6 Q&A

1. hdfs dfs -put 报错如下，解决关闭 master&salve 防火墙

hdfs.DFSClient: Exception in createBlockOutputStream

java.net.NoRouteToHostException: No route to host