1.高可用集群
1.1简介
Keepalived是Linux下一个轻量级别的高可用解决方案。高可用:广义来讲,是指整个系统的高可用行;狭义的来讲就是主机的冗余和接管。
Keepalived起初是为LVS设计的,专门用来监控集群系统中各个服务节点的状态,它根据TCP/IP参考模型的第三、第四层、第五层交换机制检测每个服务节点的状态,如果某个服务器节点出现异常,或者工作出现故障,Keepalived将检测到,并将出现的故障的服务器节点从集群系统中剔除,这些工作全部是自动完成的,不需要人工干涉,需要人工完成的只是修复出现故障的服务节点。
后来Keepalived又加入了VRRP的功能,VRRP(VritrualRouterRedundancyProtocol,虚拟路由冗余协议)出现的目的是解决静态路由出现的单点故障问题,通过VRRP可以实现网络不间断稳定运行,因此Keepalvied一方面具有服务器状态检测和故障隔离功能,另外一方面也有HAcluster功能。
健康检查和失败切换是keepalived的两大核心功能。所谓的健康检查,就是采用tcp三次握手,icmp请求,http请求,udp
echo请求等方式对负载均衡器后面的实际的服务器(通常是承载真实业务的服务器)进行保活;而失败切换主要是应用于配置了主备模式的负载均衡器,利用VRRP维持主备负载均衡器的心跳,当主负载均衡器出现问题时,由备负载均衡器承载对应的业务,从而在最大限度上减少流量损失,并提供服务的稳定性。
1.2原理
Keepalived对服务器运行状态和故障隔离的工作原理:
Keepalived工作在TCP/IP参考模型的三层、四层、五层(物理层,链路层):
-
网络层(3):Keepalived通过ICMP协议向服务器集群中的每一个节点发送一个ICMP数据包(有点类似与Ping的功能),如果某个节点没有返回响应数据包,那么认为该节点发生了故障,Keepalived将报告这个节点失效,并从服务器集群中剔除故障节点。
-
传输层(4):Keepalived在传输层里利用了TCP协议的端口连接和扫描技术来判断集群节点的端口是否正常,比如对于常见的WEB服务器80端口。或者SSH服务22端口,Keepalived一旦在传输层探测到这些端口号没有数据响应和数据返回,就认为这些端口发生异常,然后强制将这些端口所对应的节点从服务器集群中剔除掉。
-
应用层(5):,Keepalived的运行方式也更加全面化和复杂化,用户可以通过自定义Keepalived工作方式,例如:可以通过编写程序或者脚本来运行Keepalived,而Keepalived将根据用户的设定参数检测各种程序或者服务是否允许正常,如果Keepalived的检测结果和用户设定的不一致时,Keepalived将把对应的服务器从服务器集群中剔除。
1.3 集群类型
- LB:Load Balance 负载均衡
LVS/HAProxy/nginx(http/upstream, stream/upstream) - HA:High Availability 高可用集群
数据库、Redis - SPoF: Single Point of Failure,解决单点故障
HPC:High Performance Computing 高性能集群
1.4实现高可用
提升系统高用性的解决方案:降低MTTR- Mean Time To Repair(平均故障时间)
解决方案:建立冗余机制
- active/passive 主/备
- active/active 双主
- active --> HEARTBEAT --> passive
- active <–> HEARTBEAT <–> active
1.5VRRP:Virtual Router Redundancy Protocol
虚拟路由冗余协议,解决静态网关单点风险
物理层:路由器、三层交换机
软件层:keepalived
1.5.1 VRRP 相关术语
虚拟路由器:Virtual Router
虚拟路由器标识:VRID(0-255),唯一标识虚拟路由器
VIP:Virtual IP
VMAC:Virutal MAC (00-00-5e-00-01-VRID)
物理路由器:
master:主设备
backup:备用设备
priority:优先级
1.5.2VRRP 相关技术
通告:心跳,优先级等;周期性
工作方式:抢占式,非抢占式
安全认证:
无认证
简单字符认证:预共享密钥
MD5
工作模式:
主/备:单虚拟路由器
主/主:主/备(虚拟路由器1),备/主(虚拟路由器2)
2.实验
2.1keepalived环境部署
功能 | ip |
---|---|
KA1 | 172.25.254.10 |
KA2 | 172.25.254.20 |
realserver1 | 172.25.254.110 |
realserver2 | 172.252.54.120 |
#realserver1、realserver2
yum install httpd -y
[root@realserver1 ~]# echo realserver1 - 172.25.254.110 > /var/www/html/index.html
[root@realserver1 ~]# systemctl enable --now httpd
[root@realserver2 ~]# echo realserver2 - 172.25.254.120 > /var/www/html/index.html
[root@realserver2 ~]# systemctl enable --now httpd
#ka1、ka2访问
curl 172.25.254.110
realserver1
curl 172.25.254.120
realserver2
虚拟路由管理
#ka1、ka2
[root@ka1 ~]# yum install keepalived -y
[root@ka2 ~]# yum install keepalived -y
#全局配置
[root@ka1 ~]# vim /etc/keepalived/keepalived.conf
global_defs {
notification_email {
2727584440@qq.com
}
notification_email_from keepalived@timinglee.org
smtp_server 127.0.0.1
smtp_connect_timeout 30
router_id ka1.timinglee.org
vrrp_skip_check_adv_addr
vrrp_strict
vrrp_garp_interval 0
vrrp_gna_interval 0
vrrp_mcast_group4 224.0.0.18
}
#虚拟路由器配置
vrrp_instance VI_1 {
state MASTER
interface eth0
virtual_router_id 100
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:1
}
}
[root@ka1 ~]# systemctl enable --now keepalived.service
[root@ka1 ~]# ifconfig
[root@ka1 ~]# scp /etc/keepalived/keepalived.conf root@172.25.254.20:/etc/keepalived/keepalived.conf
# 在ka2上修改这两条,先不重新启动,查看是那个被用
vrrp_instance VI_1 {
state BACKUP
priority 80
#测试
[root@ka1 ~]# tcpdump -i eth0 -nn host 224.0.0.18
23:04:18.883958 IP 172.25.254.10 > 224.0.0.18: VRRPv2, Advertisement, vrid 100, prio 100, authtype simple, intvl 1s, length 20
#关掉ka1的keepalived,会出现另一台的
23:04:05.872730 IP 172.25.254.20 > 224.0.0.18: VRRPv2, Advertisement, vrid 100, prio 80, authtype simple, intvl 1s, length 20
启用keepalived日志功能
[root@ka1 ~]# vim /etc/sysconfig/keepalived
KEEPALIVED_OPTIONS="-D -S 6"
[root@ka1 ~]# vim /etc/rsyslog.conf
local6.* /var/log/keepa
lived.log
[root@ka1 ~]# systemctl restart keepalived.service
[root@ka1 ~]# systemctl restart rsyslog.service
[root@ka1 ~]# ll /var/log/keepalived.log
独立子配置文件
[root@ka1 ~]# vim /etc/keepalived/keepalived.conf
#vrrp_instance VI_1 {
# state MASTER
# interface eth0
# virtual_router_id 100
# priority 100
# advert_int 1
# authentication {
# auth_type PASS
# auth_pass 1111
# }
# virtual_ipaddress {
# 172.25.254.100/24 dev eth0 label eth0:1
# }
#}
include "/etc/keepalived/conf.d/*.conf"
[root@ka1 ~]# mkdir -p /etc/keepalived/conf.d
[root@ka1 ~]# vim /etc/keepalived/conf.d/172.25.254.100.conf
vrrp_instance VI_1 {
state MASTER
interface eth0
virtual_router_id 100
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:1
}
}
[root@ka1 ~]# systemctl restart keepalived.service
[root@ka1 ~]# ifconfig
2.2抢占模式和非抢占模式
2.2.1非抢占模式
非抢占模式 nopreempt
默认为抢占模式preempt,即当高优先级的主机恢复在线后,会抢占低先级的主机的master角色,
这样会使vip在KA主机中来回漂移,造成网络抖动,
建议设置为非抢占模式 nopreempt ,即高优先级主机恢复后,并不会抢占低优先级主机的master角色
非抢占模块下,如果原主机down机, VIP迁移至的新主机, 后续也发生down时,仍会将VIP迁移回原主机
注意:要关闭 VIP抢占,必须将各 keepalived 服务器state配置为BACKUP
#ka1主机配置
vrrp_instance VI_1 {
state BACKUP
interface eth0
virtual_router_id 20
priority 100 #优先级高
nopreempt #非抢占模式
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:0
}
}
#KA2主机配置
vrrp_instance VI_1 {
state BACKUP
interface eth0
virtual_router_id 20
priority 80 #优先级低
advert_int 1
nopreempt #非抢占模式
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:0
}
}
2.2.2抢占延迟模式 preempt_delay
抢占延迟模式,即优先级高的主机恢复后,不会立即抢回VIP,而是延迟一段时间(默认300s)再抢回VIP
preempt_delay # #指定抢占延迟时间为#s,默认延迟300s
注意:需要各keepalived服务器state为BACKUP,并且不要启用 vrrp_strict
#ka1主机配置
vrrp_instance VI_1 {
state BACKUP
interface eth0
virtual_router_id 20
priority 100 #优先级高
preempt_delay 10s #抢占延迟10s
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:0
}
}
#KA2主机配置
vrrp_instance VI_1 {
state BACKUP
interface eth0
virtual_router_id 20
priority 80 #优先级低
advert_int 1
preempt_delay 10s #抢占延迟10S
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:0
}
}
2.3VIP的单播配置
默认keepalived主机之间利用多播相互通告消息,会造成网络拥塞,可以替换成单播,减少网络流量
#在所有节点vrrp_instance语句块中设置对方主机的IP,建议设置为专用于对应心跳线网络的地址,而非使用业务网络
master主机配置
[root@KA1 ~]# vim /etc/keepalived/keepalived.conf
! Configuration File for keepalived
global_defs {
notification_email {
594233887@qq.com
}
notification_email_from keepalived@KA1.timinglee.org
smtp_server 127.0.0.1
smtp_connect_timeout 30
router_id KA1.timinglee.org
vrrp_skip_check_adv_addr
#vrrp_strict #注释此参数,与vip单播模式冲突
vrrp_garp_interval 0
vrrp_gna_interval 0
vrrp_ipsets keepalived
}
vrrp_instance VI_1 {
state MASTER
interface eth0
virtual_router_id 20
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:0
}
unicast_src_ip 172.25.254.20 #本机IP
unicast_peer {
172.25.254.30 #指向对方主机IP
#如果有多个keepalived,再加其它节点的IP
}
}
salve主机配置
[root@KA2 ~]# vim /etc/keepalived/keepalived.conf
! Configuration File for keepalived
global_defs {
notification_email {
594233887@qq.com
}
notification_email_from keepalived@KA1.timinglee.org
smtp_server 127.0.0.1
smtp_connect_timeout 30
router_id KA1.timinglee.org
vrrp_skip_check_adv_addr
#vrrp_strict #注释此参数,与vip单播模式冲突
vrrp_garp_interval 0
vrrp_gna_interval 0
vrrp_ipsets keepalived
}
vrrp_instance VI_1 {
state BACKUP
interface eth0
virtual_router_id 20
priority 80
advert_int 1
preempt_delay 60
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:0
}
unicast_src_ip 172.25.254.30 #本机ip
unicast_peer {
172.25.254.20 #对端主机IP
}
}
测试:抓包查看单播效果
[root@KA1 ~]# tcpdump -i eth0 -nn src host 172.25.254.20 and dst 172.25.254.30
dropped privs to tcpdump
tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), snapshot length 262144 bytes
00:20:16.150917 IP 172.25.254.10 > 172.25.254.20: VRRPv2, Advertisement, vrid 20,
prio 100, authtype simple, intvl 1s, length 20
00:20:17.151569 IP 172.25.254.10 > 172.25.254.20: VRRPv2, Advertisement, vrid 20,
prio 100, authtype simple, intvl 1s, length 20
00:20:18.151754 IP 172.25.254.10 > 172.25.254.20: VRRPv2, Advertisement, vrid 20,
prio 100, authtype simple, intvl 1s, length 20
00:20:19.152290 IP 172.25.254.10 > 172.25.254.20: VRRPv2, Advertisement, vrid 20,
prio 100, authtype simple, intvl 1s, length 20
[root@KA2 ~]# tcpdump -i eth0 -nn src 172.25.254.20 and dst 172.25.254.10
dropped privs to tcpdump
tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), snapshot length 262144 bytes
00:20:50.853174 IP 172.25.254.20 > 172.25.254.10: VRRPv2, Advertisement, vrid 30,
prio 80, authtype simple, intvl 1s, length 20
00:20:51.853798 IP 172.25.254.20 > 172.25.254.10: VRRPv2, Advertisement, vrid 30,
prio 80, authtype simple, intvl 1s, length 20
2.4Keepalived 通知脚本配置
当keepalived的状态变化时,可以自动触发脚本的执行,比如:发邮件通知用户
默认以用户keepalived_script身份执行脚本
如果此用户不存在,以root执行脚本可以用下面指令指定脚本执行用户的身份
2.4.1邮件配置
# 安装邮件发送工具
[root@KA1 ~]# yum install mailx -y
[root@KA1 ~]# vim /etc/mail.rc
#######mail set##########
set from=2567419484@qq.com
set smtp=smtp.qq.com
set smtp-auth-user=2567419484@qq.com
set smtp-auth-password=angelababybeau #自己邮箱的授权码
set smtp-auth=login
set ssl-verify=ignore
#发送邮件
echo hello word | mail -s test 2567419484@qq.com
[root@ka1 ~]# vim /etc/keepalived/mail.sh
chmod +x /etc/keepalived/mail.sh
[root@KA1 +K2 ~]#vim /etc/keepalived/keepalived.conf
vrrp_instance VI_1 {
state MASTER
interface eth0
virtual_router_id 20
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:0
}
unicast_src_ip 172.25.254.20
unicast_peer {
172.25.254.30
}
notify_master "/etc/keepalived/mail.sh master"
notify_backup "/etc/keepalived/mail.sh backup"
notify_fault "/etc/keepalived/mail.sh fault"
}
#模拟master故障
[root@ka1-centos8 ~]#killall keepalived
测试
2.5实现 master/master 的 Keepalived 双主架构
master/slave的单主架构,同一时间只有一个Keepalived对外提供服务,此主机繁忙,而另一台主机却很空闲,利用率低下,可以使用master/master的双主架构,解决此问题。
master/master 的双主架构:
即将两个或以上VIP分别运行在不同的keepalived服务器,以实现服务器并行提供web访问的目的,提高服务器资源利用率
# ka1配置
vrrp_instance VI_1 {
state MASTER
interface eth0
virtual_router_id 100
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:1
}
unicast_src_ip 172.25.254.10
unicast_peer {
172.25.254.20
}
track_script {
check_haproxy
}
}
vrrp_instance VI_2 {
state BACKUP
interface eth0
virtual_router_id 200
priority 80
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.200/24 dev eth0 label eth0:2
}
unicast_src_ip 172.25.254.10
unicast_peer {
172.25.254.20
}
}
# ka2配置
vrrp_instance VI_1 {
state BACKUP
interface eth0
virtual_router_id 100
priority 80
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:1
}
unicast_src_ip 172.25.254.20
unicast_peer {
172.25.254.10
}
}
vrrp_instance VI_2 {
state MASTER
interface eth0
virtual_router_id 200
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.200/24 dev eth0 label eth0:2
}
unicast_src_ip 172.25.254.20
unicast_peer {
172.25.254.10
}
}
2.6实现IPVS的高可用性
# 在realserver1/2
[root@realserver2 ~]# ip a a 172.25.254.100/32 dev lo
[root@realserver2 ~]# vim /etc/sysctl.d/arp.conf
net.ipv4.conf.all.arp_ignore=1
net.ipv4.conf.all.arp_announce=2
net.ipv4.conf.lo.arp_ignore=1
net.ipv4.conf.lo.arp_announce=2
[root@realserver2 ~]# sysctl --system
# ka1/2
[root@ka1 ~]# yum install ipvsadm -y
[root@ka1 ~]# vim /etc/keepalived/keepalived.conf
virtual_server 172.25.254.100 80 {
delay_loop 6
lb_algo wrr
lb_kind DR
#persistence_timeout 50
protocol TCP
real_server 172.25.254.110 80 {
weight 1
HTTP_GET {
url {
path /
status_code 200
}
connect_timeout 3
nb_get_retry 2
delay_before_retry 2
}
}
real_server 172.25.254.120 80 {
weight 1
HTTP_GET {
url {
path /
status_code 200
}
connect_timeout 3
nb_get_retry 2
delay_before_retry 2
}
}
}
[root@ka1 ~]# systemctl restart keepalived.service
[root@ka1 ~]# ipvsadm -Ln
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
-> RemoteAddress:Port Forward Weight ActiveConn InActConn
[root@ka2 ~]# ipvsadm -A -t 172.255.254.100:80 -s wrr
[root@ka2 ~]# ipvsadm -Ln
[root@ka1 ~]# ipvsadm -Ln
# 测试
[root@ka2 ~]# for i in {1..6}; do curl 172.25.254.100; done
realserver2 - 172.25.254.120
realserver1 - 172.25.254.110
realserver2 - 172.25.254.120
realserver1 - 172.25.254.110
realserver2 - 172.25.254.120
2.7实现其它应用的高可用性VRRP Script
keepalived利用 VRRP Script 技术,可以调用外部的辅助脚本进行资源监控,并根据监控的结果实现优先动态调整,从而实现其它应用的高可用性功能
分两步实现:
1、定义脚本
vrrp_script:自定义资源监控脚本,vrrp实例根据脚本返回值,公共定义,可被多个实例调用,定义在vrrp实例之外的独立配置块,一般放在global_defs设置块之后。
通常此脚本用于监控指定应用的状态。一旦发现应用的状态异常,则触发对MASTER节点的权重减至低于SLAVE节点,从而实现 VIP 切换到 SLAVE 节点
vrrp_script <SCRIPT_NAME> {
script <STRING>|<QUOTED-STRING> #此脚本返回值为非0时,会触发下面OPTIONS执行
OPTIONS
}
2、调用脚本
track_script:调用vrrp_script定义的脚本去监控资源,定义在VRRP实例之内,调用事先定义的vrrp_script
track_script {
SCRIPT_NAME_1
SCRIPT_NAME_2
}
2.7.1实战:利用脚本实现主从角色切换
# 在ka1上
[root@ka1 ~]# vim /etc/keepalived/test.sh
#!/bin/bash
[ ! -f /mnt/lee ]
[root@ka1 ~]# sh /etc/keepalived/test.sh
[root@ka1 ~]# echo $?
0
[root@ka1 ~]# chmod +x /etc/keepalived/test.sh
[root@ka1 ~]# vim /etc/keepalived/keepalived.conf
vrrp_script check_haproxy {
script "/etc/keepalived/test.sh"
interval 1
weight -30
fall 2
rise 2
timeout 2
}
vrrp_instance VI_1 {
state MASTER
interface eth0
virtual_router_id 100
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:1
}
unicast_src_ip 172.25.254.10
unicast_peer {
172.25.254.20
}
track_script {
check_haproxy
}
}
[root@ka1 ~]# systemctl restart keepalived.service
# 测试
[root@ka1 ~]# touch /mnt/lee
[root@ka1 ~]# ifconfig
[root@ka2 ~]# ifconfig
eth0:1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 172.25.254.100 netmask 255.255.255.0 broadcast 0.0.0.0
ether 00:0c:29:44:a9:9f txqueuelen 1000 (Ethernet)
[root@ka1 ~]# rm -rf /mnt/lee
[root@ka1 ~]# systemctl restart keepalived.service
[root@ka1 ~]# ifconfig
eth0:1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 172.25.254.100 netmask 255.255.255.0 broadcast 0.0.0.0
ether 00:0c:29:42:39:a7 txqueuelen 1000 (Ethernet)
2.7.2实战:实现HAProxy高可用
#ka1、ka2
[root@ka1 ~]# yum install haproxy -y
[root@ka1 ~]# vim /etc/sysctl.conf
net.ipv4.ip_nonlocal_bind = 1
[root@ka1 ~]# sysctl -p
[root@ka1 ~]# vim /etc/haproxy/haproxy.cfg
listen webcluster
bind 172.25.254.100:80
mode http
balance roundrobin
server web1 172.25.254.110:80 check inter 3 fall 2 rise 5
server web2 172.25.254.120:80 check inter 3 fall 2 rise 5
[root@ka1 ~]# sysetemctl enable ---now haproxy
[root@ka1 ~]# netstat -antlupe | grep haproxy
[root@ka1 ~]# curl 172.25.254.100 (不能,和lvs冲突)
[root@ka1 ~]# vim /etc/keepalived/keepalived.conf
注释掉
#virtual_server 172.25.254.100 80 {
[root@ka1 ~]# systemctl restart keepalived.service
#realwebser1、2
[root@realserver2 ~]# systemctl restart network (删掉lo)
[root@realserver2 ~]# vim /etc/sysctl.d/arp.conf
net.ipv4.conf.all.arp_ignore=0
net.ipv4.conf.all.arp_announce=0
net.ipv4.conf.lo.arp_ignore=0
net.ipv4.conf.lo.arp_announce=0
[root@realserver2 ~]# sysctl --system
[root@ka1 ~]# curl 172.25.254.100
[root@ka1 ~]# curl 172.25.254.110
[root@ka1 ~]# curl 172.25.254.120
在ka1关掉haproxy,此时外面访问不了172.25.254.10
# 在ka1
[root@ka1 ~]# vim /etc/keepalived/test.sh
#!/bin/bash
killall -0 haproxy
[root@ka1 ~]# sh /etc/keepalived/test.sh
[root@ka1 ~]# echo $?
[root@ka1 ~]# vim /etc/keepalived/keepalived.conf
vrrp_script check_haproxy {
script "/etc/keepalived/test.sh"
interval 1
weight -30
fall 2
rise 2
timeout 2
}
vrrp_instance VI_1 {
state MASTER
interface eth0
virtual_router_id 100
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.25.254.100/24 dev eth0 label eth0:1
}
unicast_src_ip 172.25.254.10
unicast_peer {
172.25.254.20
}
track_script {
check_haproxy
}
}
[root@ka1 ~]# systemctl restart keepalived.service
# 测试 关掉ka1的haproxy,也能一直访问
[root@ka1 ~]# systemctl stop haproxy.service
[root@ka1 ~]# systemctl start haproxy.service
[root@ka2 ~]# while true; do curl 172.25.254.100; done