HA Cluster配置前提:
    1、本机的主机名,要与hostname(uname -n)获得的名称保持一致;
        CentOS 6: /etc/sysconfig/network
        CentOS 7: hostnamectl set-hostname HOSTNAME
        各节点要能互相解析主机名;一般建议通过hosts文件进行解析(防止DNS服务无法访问);
    2、各节点时间同步;
    3、确保iptables及selinux不会成为服务阻碍;

keepalived是vrrp协议在Linux主机上以守护进程方式的实现,能够根据配置文件自动生成ipvs规则;
可以对各RS做健康状态检测;
    配置文件的组成部分:keepalived.conf文件
        1.GLOBAL CONFIGURATION
        2.VRRPD CONFIGURATION
            vrrp instance
            vrrp synchonization group
        3.LVS CONFIGURATION

获取帮助:man keepalived.conf


keepalived默认不输出日志解决:

编辑/etc/sysconfig/keepalived修改为

KEEPALIVED_OPTIONS="-D -S 3"

编辑/etc/rsyslog.conf添加一行

local3.*    /var/log/keepalived.log

重启keepalived服务和rsyslog服务即可

systemctl restart rsyslog.service

systemctl restart keepalived.service


使用systemctl status keepalived可以查看详细


示例:

global_defs {
   notification_email {
     [email protected]
     [email protected]   #指明多个服务监控中收集的信息发送给哪些收件人
     [email protected]
   }
   notification_email_from [email protected]#指明发件人
   smtp_server 127.0.0.1
   smtp_connect_timeout 30
   router_id LVS_DEVEL   
   #vrrp_mcast_group1 224.0.0.100  #指定节点间传递心跳的多播地址,如果多实例运行,不要全局指定,在每个vrrp实例中指定或者不指定 
}
 
vrrp_instance VI_1 {
    state MASTER    #初始状态  master和backup两种
    interface eth0    #流动ip绑定于那块网卡上
    #use_vmac  指定虚拟mac地址,可选
    virtual_router_id 51    #虚拟路由组自己的ID号,用于区分多个虚拟路由组,须唯一
    priority 100    #抢占模式下,即使state定义为master,而自身优先级不高还是会被抢占
    advert_int 1    #每隔多少秒向外发送一次心跳信息
    authentication {
        auth_type PASS  认证方式PASS表示简单字符认证,还有MD5认证
        auth_pass 1111    认证密码 (可以使用openssl命令生成Openssl rand -hex 4)
    }
    virtual_ipaddress {
        192.168.200.16
        192.168.200.17    #虚拟ip地址,下面的示例表示可以给出详细地址信息,如别名,设备等
        192.168.200.18
    }
        nopreempt  #非抢占模式;默认为抢占模式;
}
    virtual_ipaddress {
        / brd  dev  scope  label 

不重启服务,手动让主节点成为备节点:

#在vrrp上下文之外定义一个脚本实现:
vrrp_script chk_mantaince_down {
   script "[[ -f /etc/keepalived/down ]] && exit 1 || exit 0"
   interval 1    #多久检测一次
   weight -2    #检测到出现down文件时自身优先级减少几
}
#在vrrp上下文中调用
track_script {   
        chk_mantaince_down
    }

同步组定义:当我们基于keepalive做高可用,又为keepalive本身提供了负载均衡,这时需要定义两个虚拟路由分别负责外网和内网,外网ip移动到另一台主机时,内网ip也需要流动,内网ip移动到另一主机时,外网同理也需要移动。最后将两个虚拟路由归并到一个组中,如下图LVS-NAT模型负载均衡就需要这样做

keepalived构建高可用集群_第1张图片

vrrp_sync_group VG_1 {
    group {
    VI_1   # name of vrrp_instance (below)
    VI_2  # One for each moveable IP.
           }
        }
vrrp_instance VI_1 {
    eth0    #外网网卡
    vip
        }

vrrp_instance VI_2 {
    eth1    #内网网卡
    dip
        }

在virtual instance中的主机状态发生改变时发送通知:

# notify scripts, alert as above
    notify_master | #当前主机转换为主节点时发送通知
    notify_backup | #当前主机转换为备节点时发送通知
    notify_fault |  #当前主机故障时发送通知
    notify |        #自行指明
    smtp_alert

例如:(在VI上下文中定义)
notify_master "/etc/keepalived/notify.sh master"
notify_backup "/etc/keepalived/notify.sh backup"
notify_fault "/etc/keepalived/notify.sh fault"

脚本简单示例:

vip=172.16.20.100
contact='root@localhost'

notify() {
    mailsubject="`hostname` to be $1: $vip floating"
    mailbody="`date '+%F %H:%M:%S'`: vrrp transition, `hostname` changed to be $1"
    echo $mailbody | mail -s "$mailsubject" $contact
}

case "$1" in
    master)
        notify master
#        /etc/rc.d/init.d/keepalived start
        exit 0
    ;;
    backup)
        notify backup
#        /etc/rc.d/init.d/keepalived stop
        exit 0
    ;;
    fault)
        notify fault
#        /etc/rc.d/init.d/keepalived stop
        exit 0
    ;;
    *)
        echo 'Usage: `basename $0` {master|backup|fault}'
        exit 1
    ;;
esac

案例1:lvs-dr+keepalived实现负载均衡和高可用

keepalived构建高可用集群_第2张图片

1.初始化两个real server配置:在r1和r2上分别执行./lvs.sh start
#!/bin/bash
#
case $1 in
start)
        echo 1 > /proc/sys/net/ipv4/conf/all/arp_ignore
        echo 1 > /proc/sys/net/ipv4/conf/lo/arp_ignore
        echo 2 > /proc/sys/net/ipv4/conf/all/arp_announce
        echo 2 > /proc/sys/net/ipv4/conf/all/arp_announce
        ;;
stop)
        echo 0 > /proc/sys/net/ipv4/conf/all/arp_ignore
        echo 0 > /proc/sys/net/ipv4/conf/lo/arp_ignore
        echo 0 > /proc/sys/net/ipv4/conf/all/arp_announce
        echo 0 > /proc/sys/net/ipv4/conf/all/arp_announce
        ;;
esac
2.为两个real server配置vip,添加路由:
[root@node1 ~]# ifconfig lo:0 192.168.20.100 netmask 255.255.255.255 broadcast 192.168.20.100 up
[root@node1 ~]# route add -host 192.168.20.100 dev lo:0

测试阶段:使用keepalived前先测试lvs是好用的
其中一个director上配置vip:
ip addr add 192.168.20.100/32 dev eno50332208 
[root@node3 ~]# ip addr list eno50332208 | grep 100
4: eno50332208:  mtu 1500 qdisc pfifo_fast state UP qlen 1000
    inet 192.168.20.100/32 scope global eno50332208
director上添加规则:
[root@node3 ~]# ipvsadm -A -t 192.168.20.100:80 -s rr
[root@node3 ~]# ipvsadm -a -t 192.168.20.100:80 -r 192.168.20.7 -g -w 1
[root@node3 ~]# ipvsadm -a -t 192.168.20.100:80 -r 192.168.20.8 -g -w 1
另一台主机访问虚拟ip发现以轮询:
[root@node4 ~]# curl 192.168.20.100
httpd on node1
[root@node4 ~]# curl 192.168.20.100
httpd on node3

3.在两个director上安装httpd作为sorry server
yum install httpd
配置sorry server页面
echo "sorry , maintannancing,here is director1" > /var/www/html/index.html
echo "sorry , maintannancing,here is director2" > /var/www/html/index.html

4.配置keepalived
yum install keepalived
keepalived.conf文件:
! Configuration File for keepalived

global_defs {
   notification_email {
       root@localhost
    }
   notification_email_from leeha@localhost
   smtp_server 127.0.0.1
   smtp_connect_timeout 30
   router_id LVS_DEVEL
}
vrrp_script chk_mantaince_down {
   script "[[ -f /etc/keepalived/down ]] && exit 1 || exit 0"
   interval 1
   weight -2
}
vrrp_instance VI_1 {
    state MASTER    #第二个director上定义为BACKUP
    interface eno50332208
    virtual_router_id 51
    priority 100 #第二个director上定以为99
    advert_int 1
    authentication {
        auth_type PASS
        auth_pass 1111
    }
    virtual_ipaddress {
        192.168.20.100 dev eno50332208
    }
    track_script {
        chk_mantaince_down
}
        notify_master "/etc/keepalived/notify.sh master"
        notify_backup "/etc/keepalived/notify.sh backup"
        notify_fault "/etc/keepalived/notify.sh fault"
}

#ipvs配置:
virtual_server 192.168.20.100 80 {
    delay_loop 6
    lb_algo wrr
    lb_kind DR
    nat_mask 255.255.0.0
    persistence_timeout 0
    protocol TCP
     sorry_server 127.0.0.1 80  #配置sorry server
#real server健康监测,使用HTTP_GET
    real_server 192.168.20.7 80 {
        weight 1
        HTTP_GET {
            url {
              path /
              status_code 200
                }
           # url {
          &n`sp; #  path /mrtg/
            #  digest 9b3a0c85a887a256d6939da88aabd8cd
            #}
            connect_timeout 3
            nb_get_retry 3
            delay_before_retry 3
    }
}
#real server健康监测
    real_server 192.168.20.8 80 {
        weight 1
        HTTP_GET {
            url {
              path /
              status_code 200
                }
           # url {
            #  path /mrtg/
            #  digest 9b3a0c85a887a256d6939da88aabd8cd
            #}
            connect_timeout 3
            nb_get_retry 3
            delay_before_retry 3
    }
}
}

TIPS:健康监测也可以用tcp_check
##      TCP_CHECK {
##         connect_timeout 3
##      }


给予notify脚本:
使用上文中脚本简单示例给出的即可
两个director上启动keepalived

测试:
1.抓包查看:
tcpdump -i eno50332208 -nn host 192.168.20.1

2.lvs规则自动根据keepalived中配置生成了
[root@node3 keepalived]# ipvsadm -L -n
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  192.168.20.100:80 wrr
  -> 192.168.20.7:80              Route   1      0          0         
  -> 192.168.20.8:80              Route   1      0          0   
3.vip自动配置上去了
[root@node3 keepalived]# ip addr list | grep 100/32
    inet 192.168.20.100/32 scope global eno50332208
4.通知邮件收到:
Message 23:
From [email protected]  Wed Oct 21 02:23:16 2015
Return-Path: 
X-Original-To: root@localhost
Delivered-To: [email protected]
Date: Wed, 21 Oct 2015 02:23:16 -0700
To: [email protected]
Subject: node3.lee.com to be master: 192.168.20.100 floating
User-Agent: Heirloom mailx 12.5 7/5/10
Content-Type: text/plain; charset=us-ascii
From: [email protected] (root)
Status: RO

2015-10-21 02:23:16: vrrp transition, node3.lee.com changed to be master

5.让192.168.20.7这个real server下线:service httpd stop
在director上查看发现规则被删除
[root@node3 keepalived]# ipvsadm -L -n
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  192.168.20.100:80 wrr
  -> 192.168.20.8:80              Route   1      0          0         
You have new mail in /var/spool/mail/root

6.让第一个director下线,测试
在director1的/etc/keepalived/下创建down文件,发现地址转移到第二个director上,访问real server成功
7.让两个real server都下线,看看sorry server是否生效
[root@node3 keepalived]# ipvsadm -L -n
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  192.168.20.100:80 wrr
  -> 127.0.0.1:80                 Route   1      0          0

案例二:keepalived+nginx实现高可用负载均衡web

负载均衡nginx
配置两个节点nginx的实现后端主机负载均衡
upstreamupservers {
        server 192.168.20.7 weight=1;
        server 192.168.20.8 weight=2;
    }
调用:
location/ {
            proxy_pass http://upservers/;
            index  index.html index.htm;
            proxy_set_header Host $host;
            proxy_set_header x-Real-IP$remote_addr;
         }
 
tips:Killall -0 nginx 可以判断某个进程是否在线
keepalived.conf配置:
! Configuration File for keepalived
 
global_defs {
   notification_email {
       root@localhost
    }
   notification_email_from leeha@localhost
   smtp_server 127.0.0.1
   smtp_connect_timeout 30
   router_id LVS_DEVEL
}
vrrp_script chk_mantaince_down {
   script "[[ -f /etc/keepalived/down ]] && exit 1 || exit 0"
   interval 1
   weight -2
}
#写一个脚本使keepalived监控nginx服务
vrrp_script chk_nginx {
        script "killall -0 nginx &> /dev/null"
        interval 1
        weight -10
}
vrrp_instance VI_1 {
    state MASTER  #第二个节点改为BACKUP
    interface eno50332208
    virtual_router_id 51
    priority 100 #第二个节点改成99
    advert_int 1
    authentication {
        auth_type PASS
        auth_pass 1111
    }
    virtual_ipaddress {
        192.168.20.100 dev eno50332208 label eno50332208:0
    }
    track_script {
        chk_mantaince_down
}
#调用上面那个检测nginx的脚本
    track_script {
        chk_nginx
        }
        notify_master "/etc/keepalived/notify.sh master"
        notify_backup "/etc/keepalived/notify.sh backup"
        notify_fault "/etc/keepalived/notify.sh fault"
}

如果使用双主则再加一个VI,这样就实现两个节点nginx都在提供服务,万一其中一个down了另一个就承载两个vip
双主情况下在notify脚本中就不能定义systemctl restart nginx.service,因为这样备节点ginx挂了会影响主节点nginx正常提供服务
vrrp_instance VI_2 {
    state BACKUP
    interface eno50332208
    virtual_router_id 61
    priority 99
    advert_int 1
    authentication {
        auth_type PASS
        auth_pass 2222
    }
    virtual_ipaddress {
        192.168.20.111 dev eno50332208 label eno50332208:1
    }
    track_script {
        chk_mantaince_down
}
    track_script {
        chk_nginx
        }
        notify_master "/etc/keepalived/notify.sh master"
        notify_backup "/etc/keepalived/notify.sh backup"
        notify_fault "/etc/keepalived/notify.sh fault"
}

定义好后重启keepalived和nginx服务,主节点上有了vip 192.168.20.100,因为上面脚本设置了监控nginx服务当主节点的nginx服务down了,vip转移到备节点
配置notify脚本:
#!/bin/bash
                # Author: MageEdu 
                # description: An example of notify script
                # 
 
                vip=192.168.20.100
                contact='root@localhost'
 
                notify() {
                    mailsubject="`hostname` to be $1: $vip floating"
                    mailbody="`date '+%F %H:%M:%S'`: vrrp transition, `hostname` changed to be $1"
                    echo $mailbody | mail -s "$mailsubject" $contact
                }

                case "$1" in
                    master)
                        notify master
                        #systemctl restart nginx.service  #这个是作用只要主节点只要在线就一定使用主节点
                        exit 0
                    ;;
                    backup)
                        notify backup
                        #systemctl restart nginx.service
                        exit 0
                    ;;
                    fault)
                        notify fault
                        exit 0
                    ;;
                    *)
                        echo 'Usage: `basename $0` {master|backup|fault}'
                        exit 1
                    ;;
                esac