|
|
|
@ -36,6 +36,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
### 1.3.3 下载Spark安装包
|
|
|
|
|
|
|
|
|
|
到`Spark`官网:<a href="https://archive.apache.org/dist/spark/spark-2.3.4/" target="view_frame">下载地址 </a> 下载
|
|
|
|
@ -84,11 +87,88 @@ export SPARK_MASTER_IP=master节点IP
|
|
|
|
|
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 1.3.7 修改 slaves 文件
|
|
|
|
|
### 1.3.7 配置集群机器的免密登录
|
|
|
|
|
|
|
|
|
|
我们知道集群直接如果要能互相直接更方便通信,需要设置集群机器之间的免密登录,而要搭建一个完全分布式的环境也需要设置免密登录,而在Linux操作系统中免密登录一般使用的是SSH,所以接下来需要先完成免密登录的配置。
|
|
|
|
|
|
|
|
|
|
三台机器的IP与机器名分别为(IP和机器名以你的集群为主,这里的IP和机器名可以作为参考):
|
|
|
|
|
|
|
|
|
|
```
|
|
|
|
|
192.168.20.100 master
|
|
|
|
|
192.168.20.101 woker1
|
|
|
|
|
192.168.20.102 woker2
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
**第一步:分别设置三台机器的Ip映射**
|
|
|
|
|
|
|
|
|
|
修改 Hosts 文件
|
|
|
|
|
|
|
|
|
|
```
|
|
|
|
|
vim /etc/hosts
|
|
|
|
|
添加
|
|
|
|
|
192.168.20.100 master-node
|
|
|
|
|
192.168.20.101 slave1-node
|
|
|
|
|
192.168.20.102 slave2-node
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
**第二步: 配置SSH实现免密登录**
|
|
|
|
|
|
|
|
|
|
`vim /etc/ssh/sshd_config`
|
|
|
|
|
|
|
|
|
|
修改三台机器ssh配置
|
|
|
|
|
```
|
|
|
|
|
RSAAuthentication yes # 启用 RSA 认证
|
|
|
|
|
PubkeyAuthentication yes # 启用公钥私钥配对认证方式
|
|
|
|
|
AuthorizedKeysFile .ssh/authorized_keys # 公钥文件路径(和上面生成的文件同)
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
`service sshd restart`
|
|
|
|
|
|
|
|
|
|
重启`ssh`。
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
生成从机的公钥
|
|
|
|
|
|
|
|
|
|
```
|
|
|
|
|
四台机器上都执行
|
|
|
|
|
ssh-keygen -t rsa -P ""
|
|
|
|
|
slave1上执行
|
|
|
|
|
scp ~/.ssh/id_rsa.pub 192.168.20.100:~/.ssh/id_rsa.pub.s1
|
|
|
|
|
slave2上执行
|
|
|
|
|
scp ~/.ssh/id_rsa.pub 192.168.20.100:~/.ssh/id_rsa.pub.s2
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
主机上hadoop用户执行指令将从机公钥都追加到主机`authrized_keys`文件,然后将主机的`authrized_keys`文件拷贝至从机。
|
|
|
|
|
```
|
|
|
|
|
cd ~/.ssh
|
|
|
|
|
cat id_rsa.pub.s1 >> authorized_keys
|
|
|
|
|
cat id_rsa.pub.s2 >> authorized_keys
|
|
|
|
|
scp ~/.ssh/authorized_keys 192.168.20.101:~/.ssh
|
|
|
|
|
scp ~/.ssh/authorized_keys 192.168.20.102:~/.ssh
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
从机操作
|
|
|
|
|
|
|
|
|
|
`chmod 600 ~/.ssh/authorized_keys`
|
|
|
|
|
|
|
|
|
|
**第三步:测试免密登录:**
|
|
|
|
|
|
|
|
|
|
在命令行输入:`ssh 用户名@机器名`
|
|
|
|
|
|
|
|
|
|
如果互相能免密登录则表示配置ssh免密登录成功。
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
### 1.3.8 修改 slaves 文件
|
|
|
|
|
|
|
|
|
|
首先生成一份`slaves`文件
|
|
|
|
|
|
|
|
|
|
切换到`conf`目录下:
|
|
|
|
|
切换到 Spark 的 `conf`目录下:
|
|
|
|
|
执行命令:`mv slaves.template slaves`
|
|
|
|
|
修改`slaves`文件:
|
|
|
|
|
执行命令:`vi slaves`,在该文件中加入作为`worker`节点`ip`或映射主机名。
|
|
|
|
@ -99,15 +179,17 @@ worker1
|
|
|
|
|
worker2
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 1.3.8 分发安装包
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
### 1.3.9 分发安装包
|
|
|
|
|
把master节点的`spark`安装包分发到`worker1`节点和`worker2`节点(可通过`linux`的`scp`命令)。
|
|
|
|
|
|
|
|
|
|
### 1.3.9 启动spark
|
|
|
|
|
### 1.3.10 启动spark
|
|
|
|
|
切换到`master`节点安装目录的`/sbin`目录下
|
|
|
|
|
执行命令启动`Spark`集群:`./start-all.sh`
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
### 1.3.10 检查webUI界面
|
|
|
|
|
### 1.3.11 检查webUI界面
|
|
|
|
|
|
|
|
|
|
启动完成后,在浏览器输入如下地址:`http://master地址:8888/`,可见如下页面。
|
|
|
|
|
|
|
|
|
|