HDFS的守护进程分别是NameNode, SecondaryNameNode,DataNode.
YARN的守护进程分别是ResourceManager, NodeManager,WebAppProxy.

一、安装前准备：

1.hosts解析：

三台机器配置

cat >> /etc/hosts <<EOF
10.0.0.150    hadoop-01
10.0.0.151    hadoop-02
10.0.0.152    hadoop-03
EOF

2.配置免密登陆

# 创建用户
useradd hadoop
su hadoop
ssh-keygen -P '' -t rsa -f ~/.ssh/id_rsa
cat ~/.ssh/id_rsa.pub >> authorized_keys
chmod 600 ~/.ssh/authorized_keys
# 发送配置到其他机器
for i in {151..152};do scp -rp ~/.ssh/ root@10.0.0.$i:/home/hadoop/;done
for i in {151..152};do ssh root@10.0.0.$i chown -R hadoop:hadoop /home/hadoop/;done

3.准备安装包：

三台机器配置

mkdir  /server/packages -p
cd /server/packages
[root@hadoop-01 packages]# ls /server/packages/
hadoop-3.2.1.tar.gz  jdk-8u221-linux-x64.tar.gz
# 发送软件包到其他节点
for i in {151..152};do scp ./* hadoop@10.0.0.$i:`pwd`;done

二、安装jdk：

三台机器配置

tar xf jdk-8u221-linux-x64.tar.gz
mv jdk1.8.0_221/ /usr/local/jdk

三、安装hadoop

三台机器配置

tar xf hadoop-3.2.1.tar.gz
mv hadoop-3.2.1 /usr/local/hadoop
chown -R hadoop.hadoop /usr/local/hadoop

四、配置环境变量

三台机器配置

echo 'export JAVA_HOME=/usr/local/jdk' >> /etc/profile
echo 'export HADOOP_HOME=/usr/local/hadoop' >> /etc/profile
echo 'export PATH=$PATH:$JAVA_HOME/bin:$HADOOP_HOME/bin' >> /etc/profile
source /etc/profile

五、配置hadoop

在每个节点配置JAVA_HOME

su hadoop
cd /usr/local/hadoop
sed -i 's@# export JAVA_HOME=@export JAVA_HOME=/usr/local/jdk@' etc/hadoop/hadoop-env.sh

etc/hadoop/core-site.xml

<configuration>
    <property>
        <!--使用默认文件系统-->
        <name>fs.defaultFS</name>
        <!--指定文件系统的IP和端口-->
        <value>hdfs://hadoop-01:9000</value>
    </property>
</configuration>

etc/hadoop/hdfs-site.xml

<!--默认值3,hdfs存储数据的副本数-->
<property>
    <name>dfs.replication</name>
    <value>3</value>
</property>
 
<!--dfs namenode web界面的监听地址-->
<property>
    <name>dfs.namenode.http-address</name>
    <value>hadoop-01:9870</value>
</property>

<!-- dfs保存namenode数据的路径-->
<property>
    <name>dfs.namenode.name.dir</name>
    <value>/data/hadoop/hdfs/name</value>
</property>  

<!-- dfs保存datanode数据的路径-->
<property>
    <name>dfs.datanode.data.dir</name>
    <value>/data/hadoop/hdfs/data</value>
</property>

etc/hadoop/yarn-site.xml

配置ResourceManager和NodeManager:

<property>
    <name>yarn.resourcemanager.hostname</name>
    <value>hadoop-01</value>
</property>

etc/hadoop/mapred-site.xml

配置MapReduce

<property>
    <!--执行框架设置为YARN-->
    <name>mapreduce.framework.name</name>
    <value>yarn</value>
</property>

添加工作节点信息

cat > etc/hadoop/workers <<EOF
hadoop-01
hadoop-02
hadoop-03
EOF

复制配置文件

cd /usr/local/hadoop
for i in {151..152};do scp -rp etc/hadoop/* 10.0.0.$i:`pwd`/etc/hadoop;done

6.启动hdfs

# 格式化文件系统
hdfs namenode -format
# 启动nodename
hdfs --daemon start namenode
hdfs --daemon start datanode

hadoop-02，hadoop-03启动datanode

hdfs --daemon start datanode

7.启动yarn

hadoop-01启动resourcemanager、nodemanager

yarn --daemon start resourcemanager
yarn --daemon start nodemsnager

hadoop-02、hadoop-03启动nodemanager

yarn --daemon start nodemanager

hdfs web界面：
http://hadoop-01:9870

yarn web界面：
http://hadoop-01:8088/cluster/nodes