18.6 负载均衡集群介绍
18.7 LVS介绍
18.8 LVS调度算法
18.9/18.10 LVS NAT模式搭建
18.1 集群介绍
根据功能划分为两大类:高可用和负载均衡
高可用集群通常为两台服务器,一台工作,另外一台作为冗余,当提供服务的机器宕机,冗余将接替继续提供服务。大概意思是:高可用一般使用两台机器,功能,角色是一样的。当一台服务器宕机不能服务了,利用另外的服务器顶替。
实现高可用的开源软件有:heartbeat(不建议使用,切换通信速度慢,2010年停止更新,)、keepalived(建议使用,有高可用和负载均衡的功能)
负载均衡集群,需要有一台服务器作为分发器,它负责把用户的请求分发给后端的服务器处理,在这个集群里,除了分发器外,就是给用户提供服务的服务器了,这些服务器数量至少为2
原理:一个请求,分发到后端的多台服务器去处理。假如一台服务器满足不到用户的服务需求,需要扩容到多台服务器去处理服务请求。
实现负载均衡的开源软件有LVS、keepalived、haproxy、nginx,商业的有F5、Netscaler
18.2 keepalived介绍
在这里我们使用keepalived来实现高可用集群,因为heartbeat在centos6上有一些问题,影响实验效果。
keepalived通过VRRP(Virtual Router Redundancy Protocl,虚拟路由冗余协议)来实现高可用。
在这个协议里会将多台功能相同的路由器组成一个小组,这个小组里会有1个master(主)角色和N(N>=1)个backup角色。
master会通过组播的形式向各个backup发送VRRP协议的数据包,当backup收不到master发来的VRRP数据包时,就会认为master宕机了。此时就需要根据各个backup的优先级来决定谁成为新的mater。
Keepalived要有三个模块,分别是core、check和vrrp。其中core模块为keepalived的核心,负责主进程的启动、维护以及全局配置文件的加载和解析,check模块负责健康检查,vrrp模块是来实现VRRP协议的。
18.3 用keepalived配置高可用集群(上)
用keepalived配置高可用
准备工作 2台linux系统
准备两台机器128和129,128作为master,129作为backup
1 两台机器都执行yum install -y keepalived
[root@centos7-01 src]# yum install -y keepalived
[root@centos7-02 conf]# yum install -y keepalived
2 两台机器都安装nginx(负载均衡也可以用上)
在生产环境中,许多企业把Nginx作为负载均衡器来用,它的重要性很高,一旦宕机会导致整个站点不能访问,所以有必要再准备一台备用Nginx,keepalived用在这种场景非常合适。
3 编辑128上keepalived配置文件,内容从https://coding.net/u/aminglinux/p/aminglinux-book/git/blob/master/D21Z/master_keepalived.conf获取
先清空原有的配置信息
[root@centos7-01 src]# > /etc/keepalived/keepalived.conf
然后再添加如下配置信息
global_defs { notification_email { [email protected] } notification_email_from [email protected] smtp_server 127.0.0.1 smtp_connect_timeout 30 router_id LVS_DEVEL } vrrp_script chk_nginx { script "/usr/local/sbin/check_ng.sh" interval 3 } vrrp_instance VI_1 { state MASTER interface ens33 virtual_router_id 51 priority 100 advert_int 1 authentication { auth_type PASS auth_pass aminglinux>com } virtual_ipaddress { 192.168.188.100 } track_script { chk_nginx } }
参数解释:
global_defs { notification_email { [email protected] //定义接收告警的人员信息 } notification_email_from [email protected] //定义发邮件地址(实际上没用) smtp_server 127.0.0.1 smtp_connect_timeout 30 router_id LVS_DEVEL }
以上是自定义参数,
检测服务是否正常
vrrp_script chk_nginx { //chk_nginx为自定义名字,后面还会用到它 script "/usr/local/sbin/check_ng.sh" //自定义脚本,该脚本为监控nginx服务的脚本,务必记住,稍后会用上 interval 3 //每隔3s执行一次该脚本 } vrrp_instance VI_1 { state MASTER //角色为master,从的话是backup,这里参数要大写 interface ens33 //针对哪个网卡监听VIP virtual_router_id 51 //定义路由器的ID priority 100 //权重为100,master要比backup大 advert_int 1 authentication { //认证相关信息 auth_type PASS //定义认证方式,PASS表示password auth_pass aminglinux>com //定义密码,这个密码自定义 } virtual_ipaddress { //定义VIP参数 192.168.189.100 //定义VIP(当master宕机的时候,访问backup的ip地址,这个IP是公共的) } track_script { chk_nginx //定义监控脚本,这里和上面vrr_script后面的字符串要保持一致 } }
关于VIP:它的英文名字是Virtual IP,即虚拟IP,也有浮动IP的叫法。因为这个IP是由keepalived给服务器配置上的,
服务器靠这个VIP对外提供服务,当master机器宕机,VIP被分配到backup上,这样用户看来是无感知的。
4 128编辑监控脚本,内容从https://coding.net/u/aminglinux/p/aminglinux-book/git/blob/master/D21Z/master_check_ng.sh 获取
#vim /usr/local/sbin/check_ng.sh #!/bin/bash #时间变量,用于记录日志 d=`date --date today +%Y%m%d_%H:%M:%S` #计算nginx进程数量 n=`ps -C nginx --no-heading|wc -l` #如果进程为0,则启动nginx,并且再次检测nginx进程数量, #如果还为0,说明nginx无法启动,此时需要关闭keepalived if [ $n -eq "0" ]; then /etc/init.d/nginx start n2=`ps -C nginx --no-heading|wc -l` if [ $n2 -eq "0" ]; then echo "$d nginx down,keepalived will stop" >> /var/log/check_ng.log systemctl stop keepalived fi fi
5 给予脚本755权限
[root@centos7-01 src]#chmod 755 /usr/local/sbin/check_ng.sh
6 在128启动keepalived服务,然后检查keepalived和nginx的服务进程
[root@centos7-01 src]# systemctl start keepalived [root@centos7-01 src]# ps aux |grep keepalived root 2370 0.0 0.1 118608 1384 ? Ss 16:12 0:00 /usr/sbin/keepalived -D root 2371 0.0 0.3 127468 3292 ? S 16:12 0:00 /usr/sbin/keepalived -D root 2372 0.1 0.2 127408 2828 ? S 16:12 0:00 /usr/sbin/keepalived -D root 2414 0.0 0.0 112676 984 pts/0 R+ 16:12 0:00 grep --color=auto keepalived [root@centos7-01 src]# ps aux |grep nginx root 1153 0.0 0.0 24944 864 ? Ss 14:52 0:00 nginx: master process /usr/local/nginx/sbin/nginx -c /usr/local/nginx/conf/nginx.conf nobody 1156 0.0 0.3 27388 3696 ? S 14:52 0:00 nginx: worker process nobody 1157 0.0 0.3 27388 3444 ? S 14:52 0:00 nginx: worker process root 2434 0.0 0.0 112676 984 pts/0 R+ 16:12 0:00 grep --color=auto nginx
7 先停止nginx服务,随后需要用到nginx再启动起来
[root@centos7-01 src]# /etc/init.d/nginx stop
Stopping nginx (via systemctl): [ 确定 ]
可以看到停止了nginx,它还会被加载,因为keepalived检测脚本把nginx加载起来了
[root@centos7-01 src]# ps aux |grep nginx root 2799 0.0 0.0 24944 872 ? Ss 16:15 0:00 nginx: master process /usr/local/nginx/sbin/nginx -c /usr/local/nginx/conf/nginx.conf nobody 2803 0.0 0.3 27388 3452 ? S 16:15 0:00 nginx: worker process nobody 2804 0.0 0.3 27388 3452 ? S 16:15 0:00 nginx: worker process root 2812 0.0 0.0 112676 980 pts/0 R+ 16:15 0:00 grep --color=auto nginx
18.4 用keepalived配置高可用集群(中)
keepalived日志文件在/var/log/messages
1 查看VIP地址#ip add (#ifconfig是查不到VIP的)
[root@centos7-01 src]# ip add 1: lo:mtu 65536 qdisc noqueue state UNKNOWN qlen 1 link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 inet 127.0.0.1/8 scope host lo valid_lft forever preferred_lft forever inet6 ::1/128 scope host valid_lft forever preferred_lft forever 2: ens33: mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether 00:0c:29:15:53:53 brd ff:ff:ff:ff:ff:ff inet 192.168.189.128/24 brd 192.168.189.255 scope global ens33 valid_lft forever preferred_lft forever inet 192.168.189.100/32 scope global ens33//此处是VIP地址 valid_lft forever preferred_lft forever inet 192.168.189.150/24 brd 192.168.189.255 scope global secondary ens33:0 valid_lft forever preferred_lft forever inet6 fe80::243c:86d7:d85e:224d/64 scope link valid_lft forever preferred_lft forever 3: ens37: mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether 00:0c:29:15:53:5d brd ff:ff:ff:ff:ff:ff
2 配置从之前,需要确保主和从角色有无开启防火墙,iptables,firewall和selinux等等有无处于关闭状态。
如果规则没有或者没能被清空,此时需要停止firewall
[root@centos7-01 src]# iptables -nvL [root@centos7-01 src]# getenforce Disabled [root@centos7-01 src]# systemctl stop firewall.service [root@centos7-02 conf]# systemctl stop firewall.service [root@centos7-02 conf]# setenforce 0 [root@centos7-02 conf]# getenforce Permissive [root@centos7-02 conf]# iptables -nvL
3 129上编辑配置文件,内容从https://coding.net/u/aminglinux/p/aminglinux-book/git/blob/master/D21Z/backup_keepalived.conf获取
[root@centos7-02 conf]# > /etc/keepalived/keepalived.conf [root@centos7-02 conf]# vim !$ vim /etc/keepalived/keepalived.conf global_defs { notification_email { [email protected] } notification_email_from [email protected] smtp_server 127.0.0.1 smtp_connect_timeout 30 router_id LVS_DEVEL } vrrp_script chk_nginx { script "/usr/local/sbin/check_ng.sh" interval 3 } vrrp_instance VI_1 { state BACKUP interface ens33 virtual_router_id 51 priority 90 advert_int 1 authentication { auth_type PASS auth_pass aminglinux>com } virtual_ipaddress { 192.168.189.100 } track_script { chk_nginx } }
4 129上编辑监控脚本,内容从https://coding.net/u/aminglinux/p/aminglinux-book/git/blob/master/D21Z/backup_check_ng.sh获取
[root@centos7-02 conf]# vim /usr/local/sbin/check_ng.sh [root@centos7-02 conf]# !vim vim /usr/local/sbin/check_ng.sh #时间变量,用于记录日志 d=`date --date today +%Y%m%d_%H:%M:%S` #计算nginx进程数量 n=`ps -C nginx --no-heading|wc -l` #如果进程为0,则启动nginx,并且再次检测nginx进程数量, #如果还为0,说明nginx无法启动,此时需要关闭keepalived if [ $n -eq "0" ]; then /etc/init.d/nginx start #如果yum安装的nginx,用这个启动命令systemctl start nginx n2=`ps -C nginx --no-heading|wc -l` if [ $n2 -eq "0" ]; then echo "$d nginx down,keepalived will stop" >> /var/log/check_ng.log systemctl stop keepalived fi fi
5 给脚本755权限
[root@centos7-02 conf]# chmod 755 !$
chmod 755 /usr/local/sbin/check_ng.sh
6 129上也启动服务 systemctl start keepalived
[root@centos7-02 conf]# systemctl start keepalived [root@centos7-02 conf]# ps aux |grep keepalived root 8027 0.0 0.1 118652 1400 ? Ss 16:51 0:00 /usr/sbin/keepalived -D root 8028 0.0 0.3 127516 3296 ? S 16:51 0:00 /usr/sbin/keepalived -D root 8029 0.0 0.2 127456 2848 ? S 16:51 0:00 /usr/sbin/keepalived -D root 8070 0.0 0.0 112720 972 pts/0 S+ 16:51 0:00 grep --color=auto keepalived
7 区分主从的nginx
在浏览器访问128的页面192.168.189.128
在浏览器访问129的页面192.168.189.129
访问VIP 会自动跳转到master(192.168.189.128)
访问VIP,同样也是调到mater
18.5 用keepalived配置高可用集群(下)
测试高可用
1 测试1:关闭master上的nginx服务
nginx服务关闭不到3秒,又被重新启动,这是因为keepalived的检测脚本起作用了。
2 测试2:在master上增加iptabls规则,
把vvrp出去的包给封掉
[root@centos7-01 111.com]#iptables -I OUTPUT -p vrrp -j DROP
检查iptables的行为表
[root@centos7-01 111.com]# iptables -nvL Chain INPUT (policy ACCEPT 43 packets, 3932 bytes) pkts bytes target prot opt in out source destination Chain FORWARD (policy ACCEPT 0 packets, 0 bytes) pkts bytes target prot opt in out source destination Chain OUTPUT (policy ACCEPT 27 packets, 3268 bytes) pkts bytes target prot opt in out source destination 42 1680 DROP 112 -- * * 0.0.0.0/0 0.0.0.0/0
在主山查看日志记录,频繁出现vrrp类似错误的信息
[root@centos7-01 111.com]# tail /var/log/messages May 21 17:33:09 centos7-01 NetworkManager[642]:[1526895189.1381] device (ens37): Activation: starting connection '有线连接 1' (d44e77b3-03bc-3209-8d77-782475a5a763) May 21 17:33:09 centos7-01 NetworkManager[642]: [1526895189.1383] device (ens37): state change: disconnected -> prepare (reason 'none') [30 40 0] May 21 17:33:09 centos7-01 NetworkManager[642]: [1526895189.1386] device (ens37): state change: prepare -> config (reason 'none') [40 50 0] May 21 17:33:09 centos7-01 NetworkManager[642]: [1526895189.1431] device (ens37): state change: config -> ip-config (reason 'none') [50 70 0] May 21 17:33:09 centos7-01 NetworkManager[642]: [1526895189.1435] dhcp4 (ens37): activation: beginning transaction (timeout in 45 seconds) May 21 17:33:09 centos7-01 NetworkManager[642]: [1526895189.1559] dhcp4 (ens37): dhclient started with pid 12636 May 21 17:33:09 centos7-01 dhclient[12636]: DHCPDISCOVER on ens37 to 255.255.255.255 port 67 interval 8 (xid=0x5f29b3dd) May 21 17:33:17 centos7-01 dhclient[12636]: DHCPDISCOVER on ens37 to 255.255.255.255 port 67 interval 8 (xid=0x5f29b3dd) May 21 17:33:25 centos7-01 dhclient[12636]: DHCPDISCOVER on ens37 to 255.255.255.255 port 67 interval 14 (xid=0x5f29b3dd) May 21 17:33:39 centos7-01 dhclient[12636]: DHCPDISCOVER on ens37 to 255.255.255.255 port 67 interval 16 (xid=0x5f29b3dd) 在从上查看日志,可以看到与其他机器(master)进行交互数据 [root@centos7-02 conf]# tail -10 /var/log/messages May 21 17:34:12 centos7-02 dhclient[11315]: DHCPDISCOVER on ens37 to 255.255.255.255 port 67 interval 15 (xid=0x20d5f4a4) May 21 17:34:27 centos7-02 dhclient[11315]: DHCPDISCOVER on ens37 to 255.255.255.255 port 67 interval 14 (xid=0x20d5f4a4) May 21 17:34:36 centos7-02 NetworkManager[537]: [1526895276.4276] dhcp4 (ens37): request timed out May 21 17:34:36 centos7-02 NetworkManager[537]: [1526895276.4277] dhcp4 (ens37): state changed unknown -> timeout May 21 17:34:36 centos7-02 NetworkManager[537]: [1526895276.4295] dhcp4 (ens37): canceled DHCP transaction, DHCP client pid 11315 May 21 17:34:36 centos7-02 NetworkManager[537]: [1526895276.4296] dhcp4 (ens37): state changed timeout -> done May 21 17:34:36 centos7-02 NetworkManager[537]: [1526895276.4297] device (ens37): state change: ip-config -> failed (reason 'ip-config-unavailable') [70 120 5] May 21 17:34:36 centos7-02 NetworkManager[537]: [1526895276.4299] policy: disabling autoconnect for connection '有线连接 1'. May 21 17:34:36 centos7-02 NetworkManager[537]: [1526895276.4300] device (ens37): Activation: failed for connection '有线连接 1' May 21 17:34:36 centos7-02 NetworkManager[537]: [1526895276.4302] device (ens37): state change: failed -> disconnected (reason 'none') [120 30 0]
在浏览上访问VIP,还是没有问题的。
实验证明,master虽然禁掉了VRRP协议,还是不能达到切换资源的目的的。
把VRRP恢复,规则清空即可
[root@centos7-01 111.com]# iptables -F [root@centos7-01 111.com]# iptables -nvL Chain INPUT (policy ACCEPT 19 packets, 1332 bytes) pkts bytes target prot opt in out source destination Chain FORWARD (policy ACCEPT 0 packets, 0 bytes) pkts bytes target prot opt in out source destination Chain OUTPUT (policy ACCEPT 17 packets, 1400 bytes) pkts bytes target prot opt in out source destination
3 测试3:关闭master上的keepalived服务
最暴力的测试方法就是,在master上把keepalived服务停止
[root@centos7-01 111.com]# systemctl stop keepalived
在从上ip add看看vip有没有被释放到从上
[root@centos7-02 conf]# ip add 1: lo:mtu 65536 qdisc noqueue state UNKNOWN qlen 1 link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 inet 127.0.0.1/8 scope host lo valid_lft forever preferred_lft forever inet6 ::1/128 scope host valid_lft forever preferred_lft forever 2: ens33: mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether 00:0c:29:73:7c:4c brd ff:ff:ff:ff:ff:ff inet 192.168.189.129/24 brd 192.168.189.255 scope global ens33 valid_lft forever preferred_lft forever inet 192.168.189.100/32 scope global ens33 valid_lft forever preferred_lft forever inet6 fe80::b485:96d0:c537:251e/64 scope link valid_lft forever preferred_lft forever 3: ens37: mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether 00:0c:29:73:7c:56 brd ff:ff:ff:ff:ff:ff
VIP 192.168.189.100 被释放到了在从上
查看从的日志,发现VIP被添加上
4 在浏览器访问VIP,会跳转到backup的默认主页上
由此证明,VIP已经到了backup机器上了。
5 再次在master上恢复开启keepalived服务
VIP已经释放回去了master上
[root@centos7-01 ~]# ip add 1: lo:mtu 65536 qdisc noqueue state UNKNOWN qlen 1 link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 inet 127.0.0.1/8 scope host lo valid_lft forever preferred_lft forever inet6 ::1/128 scope host valid_lft forever preferred_lft forever 2: ens33: mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether 00:0c:29:15:53:53 brd ff:ff:ff:ff:ff:ff inet 192.168.189.128/24 brd 192.168.189.255 scope global ens33 valid_lft forever preferred_lft forever inet 192.168.189.100/32 scope global ens33 valid_lft forever preferred_lft forever inet 192.168.189.150/24 brd 192.168.189.255 scope global secondary ens33:0 valid_lft forever preferred_lft forever inet6 fe80::243c:86d7:d85e:224d/64 scope link valid_lft forever preferred_lft forever 3: ens37: mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether 00:0c:29:15:53:5d brd ff:ff:ff:ff:ff:ff inet6 fe80::f3b5:ccf4:c98b:5e00/64 scope link valid_lft forever preferred_lft forever
6 测试成功
总结说明:
设置高可用集群,mater只需要一台机器,但是backup角色支持多台机器,backup角色的权重(priority)决定优先级(当mater崩溃的时候,会根据backup的权重从高到低进行优先级担任mater角色),backup的权重不能比mater高。
高可用集群除了可以利用Keepalived搭配Nginx之外,还可以利用Keepalived搭配MySQL,不过如果搭配MySQL的话,需要保持主从数据一致,不然的话,主宕机了,从的数据跟主的数据匹对不上,后果不堪设想。
上述实验,搭建了Keepalived+Nginx实现Web高可用集群
扩展
lvs三种模式详解
http://www.it165.net/admin/html/201401/2248.html
lvs几种算法
http://www.aminglinux.com/bbs/thread-7407-1-1.html
关于arp_ignore和 arp_announce
http://www.cnblogs.com/lgfeng/archive/2012/10/16/2726308.html
lvs原理相关的
http://blog.csdn.net/pi9nc/article/details/23380589