9.k8s主控节点部署L4高可用反向代理服务

VIP、keepalive、L4反向代理
node10、node11、node12

相关资料&＃xff1a;
Keepalived之——Keepalived &＃43; Nginx 实现高可用 Web 负载均衡
高可用实践——Keepalived踩坑记录
Keepalived原理

k8s主控节点-部署L4反向代理服务

安装nginx

nginx 4层代理配置

nginx http段中是7L代理配置&＃xff0c;注意不要配置到http段下
vi /etc/nginx/nginx.conf

stream {upstream kube-apiserver {server 172.10.10.21:6443 max_fails&＃61;3 fail_timeout&＃61;30s;server 172.10.10.22:6443 max_fails&＃61;3 fail_timeout&＃61;30s;}server {listen 7443;proxy_connect_timeout 2s;proxy_timeout 900s;proxy_pass kube-apiserver;}
}


node11和node12上分别配置4L代理&＃xff0c;负载node21和node22的kube-apiserver

keepalived

编写检查脚本

vi /etc/keepalived/check_port.sh

#!/bin/bash
CHK_PORT&＃61;$1
if [ -n "$CHK_PORT" ];thenPORT_PROCESS&＃61;&＃96;ss -lnt|grep $CHK_PORT|wc -l&＃96;if [ $PORT_PROCESS -eq 0 ];thenecho "Port $CHK_PORT Is Not Used,End."exit 1fi
elseecho "Check Port Cant Be Empty!"
fi


该脚本用于keepalived调用&＃xff0c;通过ss命令查看7443端口是否监听&＃xff0c;判断上文nginx是否正常

安装keepalived

yum install keepalived

配置

vi /etc/keepalived/keepalived.conf

! Configuration File for keepalived
global_defs {## 标识本节点的字条串&＃xff0c;通常为 hostnamerouter_id 172.10.10.12
}
vrrp_script chk_nginx {script "/etc/keepalived/check_port.sh 7443"interval 2weight -20
}
## 定义虚拟路由&＃xff0c; VI_1 为虚拟路由的标示符&＃xff0c;自己定义名称
vrrp_instance VI_1 {state BACKUPinterface ens33virtual_router_id 251mcast_src_ip 172.10.10.12priority 90advert_int 1authentication {auth_type PASSauth_pass 11111111}track_script {chk_nginx}virtual_ipaddress {172.10.10.10}
}


注意state参数&＃xff0c;主节点配置成MASTER&＃xff0c;备节点配置成BACKUP
vrrp_script chk_nginx 配置一个名为chk_nginx的脚本&＃xff0c;在后续vrrp_instance虚拟路由中配置使用
上数配置&＃xff0c;正常情况下MASTER服务器接管VIP成为【主】&＃xff0c;当M服务器宕机后&＃xff0c;BACKUP服务器接管VIP成为【主】&＃xff0c;已达到灾备的效果&＃xff0c;如果M服务器再次恢复正常&＃xff0c;那么M服务器会重新抢回VIP成为【主】 B服务器变成【备】。
若需VIP不随意飘动需配置nopreempt参数
vrrp_instance下配置项&＃xff1a;nopreempt设置当前节点为非抢占式的
资料&＃xff1a;Keepalived 设置master故障恢复后不重新抢回VIP(解决nopreempt无效)

systemctl start keepalived 启动keepalived
nginx reload

演示

说明:

1. 通过node10访问nginx&＃xff08;node11和node12&＃xff09;
2. 初始时&＃xff0c;node11为[主]&＃xff0c;流量流向node11
3. node11停掉nginx后&＃xff0c;node12抢占VIP&＃xff0c;流量流向node12
4. node11再次启动nginx&＃xff0c;此时VIP不会被node11抢占&＃xff0c;因为此案例node11和node12的priority都设置成90了
5. node12停掉nginx后&＃xff0c;node11才会抢占VIP&＃xff0c;获取流量