Hadoop集群搭建

1.创建三台虚拟机，本次使用的是centos7，关闭所有机器的防火墙。

关闭防火墙：

[hadoop@localhost ~]$ systemctl stop firewalld.service

修改主机名，方便对虚拟机进行区分。

主节点名称设为master，其他两个节点名称设置为slave1，slave2。

查看主机名并修改：
```
[hadoop@localhost ~]$ hostname
localhost.localdomain
[hadoop@localhost ~]$ hostnamectl set-hostname master
[hadoop@localhost ~]$ hostname
master
```
修改完重启机器：
```
[hadoop@localhost ~]$ reboot
```

2.编辑ip与hostname的映射表 /etc/hosts

将所有机器的ip与hostname的对应关系添加到hosts文件中，所有的节点都要添加，相当于DNS

172.16.46.161	master
172.16.46.163	slave1
172.16.46.162	slave2

3.ssh免密登陆

请查看ssh免密登陆

4.安装jdk

请查看jdk安装

5.安装hadoop

下载地址hadoop，下载.tar.gz格式的包

解压

[hadoop@master ~]$ tar -zxvf hadoop-2.9.2.tar.gz

设置环境变量，在/etc/profile最下方添加

export HADOOP_HOME=/home/hadoop/hadoop-2.9.2
export PATH=.:$HADOOP_HOME/bin:$PATH

加载环境变量

source /etc/profile

验证hadoop是否安装成功

[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
                       Hadoop jar and the required libraries
  credential           interact with credential providers
  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.

出现上述输出，证明安装成功。

如果未输出，则可以尝试重新启动机器加载环境变量。

6.配置hadoop

进入hadoop安装目录

6.1配置etc/hadoop/hadoop-env.sh

修改JAVA_HOME为jdk安装目录的绝对路径

6.2配置etc/hadoop/core-site.xml

设置hdfs的Namenode地址，设置hadoop运行时临时文件的存储路径

<configuration>
    <property>
        <name>fs.defaultFS</name>
       <value>hdfs://172.16.46.161:9000</value>
    </property>
    <property>
        <name>hadoop.tmp.dir</name>
       <value>file:/home/hadoop/hadoop-2.9.2/tmp</value>
    </property>
    <property>
        <name>io.file.buffer.size</name>
        <value>131702</value>
    </property>
</configuration>

如果没有配置hadoop.tmp.dir，默认存储在/tmp/hadoop-username目录下

6.3配置etc/hadoop/hdfs-site.xml

<configuration>
  <property>
    <name>dfs.replication</name>
    <value>4</value>
  </property>
  <property>
    <name>dfs.namenode.name.dir</name>
    <value>file:/home/hadoop/hadoop-2.9.2/hdfs/name</value>
    <final>true</final>
  </property>
  <property>
    <name>dfs.datanode.data.dir</name>
    <value>file:/home/hadoop/hadoop-2.9.2/hdfs/data</value>
    <final>true</final>
  </property>
  <property>
    <name>dfs.namenode.secondary.http-address</name>
    <value>172.16.46.161:9001</value>
  </property>
  <property>
    <name>dfs.webhdfs.enabled</name>
    <value>true</value>
  </property>
  <property>
    <name>dfs.permissions</name>
    <value>false</value>
  </property>
</configuration>

6.4配置etc/hadoop/mapred-site.xml

将mapper-site.xml.template重命名为mapper-site.xml

[hadoop@master hadoop-2.9.2]$ mv etc/hadoop/mapred-site.xml.template etc/hadoop/mapred-site.xml

修改mapper-site.xml

<configuration>
    <property>
        <name>mapreduce.framework.name</name>
        <value>yarn</value>
    </property>
</configuration>

指定mr运行在yarn上

6.5配置etc/hadoop/slaves

删除原有内容，写入所有节点的ip地址

172.16.46.161
172.16.46.163
172.16.46.162

6.6配置etc/hadoop/yarn-env.sh和etc/hadoop/mapred-env.sh

将JAVA_HOME配置成jdk安装目录的绝对路径

6.7配置etc/hadoop/yarn-site.xml

<configuration>

<!-- Site specific YARN configuration properties -->
  <property>
    <name>yarn.resourcemanager.address</name>
    <value>172.16.46.161:18040</value>
  </property>
  <property>
    <name>yarn.resourcemanager.scheduler.address</name>
    <value>172.16.46.161:18030</value>
  </property>
  <property>
    <name>yarn.resourcemanager.webapp.address</name>
    <value>172.16.46.161:18088</value>
  </property>
  <property>
    <name>yarn.resourcemanager.resource-tracker.address</name>
    <value>172.16.46.161:18025</value>
  </property>
  <property>
    <name>yarn.resourcemanager.admin.address</name>
    <value>172.16.46.161:18141</value>
  </property>
  <property>
    <name>yarn.nodemanager.aux-services</name>
    <value>mapreduce_shuffle</value>
  </property>
  <property>
    <name>yarn.nodemanager.auxservices.mapreduce.shuffle.class</name>
    <value>org.apache.hadoop.mapred.ShuffleHandler</value>
  </property>
</configuration>

7.将修改后的配置拷贝到其他节点

scp -r etc/ hadoop@slave1:~/hadoop-2.9.2/

8.启动集群

8.1格式化namenode

集群搭建好了，将磁盘格式化一下，后面要存数据，避免有脏数据，同时创建一些东西。

只有第一次启动需要格式化

namenode设置在哪个节点上就在哪个节点上执行下面的命令

bin/hdfs namenode -format

8.2启动集群前必须保证namenode和datanode已经启动

单节点启动namenode

[hadoop@master hadoop-2.9.2]# sbin/hadoop-daemon.sh start namenode
starting namenode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-root-namenode-master.out
[hadoop@master hadoop-2.9.2]# jps
3877 NameNode
3947 Jps

单节点启动datanode

[hadoop@master hadoop-2.9.2]# sbin/hadoop-daemon.sh start datanode
starting datanode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-root-datanode-master.out
[hadoop@master hadoop-2.9.2]# jps
3877 NameNode
4060 Jps
3982 DataNode

在其他节点依次启动 datanode

这样启动hdfs很麻烦，而且发现SecondaryNameNode并没有启动，素有hadoop提供了其他的启动方式

一步启动hdfs集群：Namenode、Datanode、SecondaryNameNode

[hadoop@master hadoop-2.9.2]$ sbin/start-all.sh
This script is Deprecated. Instead use start-dfs.sh and start-yarn.sh
Starting namenodes on [master]
master: starting namenode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-hadoop-namenode-master.out
172.16.46.162: starting datanode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-hadoop-datanode-slave2.out
172.16.46.161: starting datanode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-hadoop-datanode-master.out
172.16.46.163: starting datanode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-hadoop-datanode-slave1.out
Starting secondary namenodes [master]
master: starting secondarynamenode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-hadoop-secondarynamenode-master.out
starting yarn daemons
starting resourcemanager, logging to /home/hadoop/hadoop-2.9.2/logs/yarn-hadoop-resourcemanager-master.out
172.16.46.163: starting nodemanager, logging to /home/hadoop/hadoop-2.9.2/logs/yarn-hadoop-nodemanager-slave1.out
172.16.46.162: starting nodemanager, logging to /home/hadoop/hadoop-2.9.2/logs/yarn-hadoop-nodemanager-slave2.out
172.16.46.161: starting nodemanager, logging to /home/hadoop/hadoop-2.9.2/logs/yarn-hadoop-nodemanager-master.out
[hadoop@master hadoop-2.9.2]$ jps
4192 Jps
3237 NameNode
3543 SecondaryNameNode
3374 DataNode

8.3启动yarn

看yarn要设置在哪个节点，就在哪个节点执行下面的命令。

[hadoop@master hadoop-2.9.2]# sbin/start-yarn.sh 
starting yarn daemons
starting resourcemanager, logging to /home/hadoop/hadoop-2.9.2/logs/yarn-root-resourcemanager-master.out
172.16.46.162: starting datanode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-hadoop-datanode-slave2.out
172.16.46.161: starting datanode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-hadoop-datanode-master.out
172.16.46.163: starting datanode, logging to /home/hadoop/hadoop-2.9.2/logs/hadoop-hadoop-datanode-slave1.out
[hadoop@master hadoop-2.9.2]$ jps
4192 Jps
3237 NameNode
3814 NodeManager
3543 SecondaryNameNode
3374 DataNode
3695 ResourceManager

ResourceManager和NodeManager都启动了。

8.4hadoop集群启动成功，包括hdfs、yarn、mapreduce

上面这种启动方式很麻烦，hadoop还提供了一键启动和一键关闭。

sbin/start-all.sh 
sbin/stop-all.sh

9.远程访问hadoop集群

访问hdfs：http://172.16.46.161:50070/

10.简单测试

在hdfs文件系统中创建目录，两种方式。

bin/hdfs dfs -mkdir -p /usr/input
bin/hadoop fs -mkdir -p /usr/output

集群部署规划

上述步骤已经将hadoop集群搭建完成，但是我们将Namenode、SecondaryNamenode、ResourceManager都部署到一台机器上。

这样会增大服务器的压力，而且组件的资源都被压缩了。所以可以部署到三台机器。

	hadoop11	hadoop12	hadoop13
HDFS	NameNode、DataNode	DataNode	SecondaryNameNode
YARN	NodeManager	ResourceManager、NodeManager	NodeManager

三个核心组件分布到三台机器。

异常记录

找不到jps

jps是查看java进程的，找不到说明java没有装好，需要设置java环境变量
重启后无法启动datanode

通常在第一次搭建时可以成功，但是重启后不能成功，datanode 无法启动，原因是 datanode 无法被 namenode 识别。

namenode 在 format 时会形成两个标识，blockPoolId 和 clusterId；

当有 datanode 加入时，会获取这两个标识作为从属这个 namenode 的标识，这样才能组成集群；

一旦 namenode 被重新 format，会更新这两个标识；

然而 datanode 还拿原来的标识过来接头，自然被拒之门外

解决方法：删除所有节点的数据，即 tmp，包括 namenode 的数据，重新格式化，再启动

各种操作都会有如下警告

WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable

无需理会，只是警告，确实想解决，参考解决办法