Keepalived实现Nginx负载均衡机高可用
keepalived简介
Keepalived 软件起初是专为LVS负载均衡软件设计的,用来管理并监控LVS集群系统中各个服务节点的状态,后来又加入了可以实现高可用的VRRP功能。因此,Keepalived除了能够管理LVS软件外,还可以作为其他服务(例如:Nginx、Haproxy、MySQL等)的高可用解决方案软件。
Keepalived软件主要是通过VRRP协议实现高可用功能的。VRRP是Virtual Router RedundancyProtocol(虚拟路由器冗余协议)的缩写,VRRP出现的目的就是为了解决静态路由单点故障问题的,它能够保证当个别节点宕机时,整个网络可以不间断地运行。
所以,Keepalived 一方面具有配置管理LVS的功能,同时还具有对LVS下面节点进行健康检查的功能,另一方面也可实现系统网络服务的高可用功能。
keepalived官网
keepalived的重要功能
keepalived 有三个重要的功能,分别是:
- 管理LVS负载均衡软件
- 实现LVS集群节点的健康检查
- 作为系统网络服务的高可用性(failover)
keepalived高可用故障转移的原理
Keepalived 高可用服务之间的故障切换转移,是通过 VRRP (Virtual Router Redundancy Protocol ,虚拟路由器冗余协议)来实现的。
在 Keepalived 服务正常工作时,主 Master 节点会不断地向备节点发送(多播的方式)心跳消息,用以告诉备 Backup 节点自己还活看,当主 Master 节点发生故障时,就无法发送心跳消息,备节点也就因此无法继续检测到来自主 Master 节点的心跳了,于是调用自身的接管程序,接管主 Master 节点的 IP 资源及服务。而当主 Master 节点恢复时,备 Backup 节点又会释放主节点故障时自身接管的IP资源及服务,恢复到原来的备用角色。
那么,什么是VRRP呢?
VRRP ,全 称 Virtual Router Redundancy Protocol ,中文名为虚拟路由冗余协议 ,VRRP的出现就是为了解决静态踣甶的单点故障问题,VRRP是通过一种竞选机制来将路由的任务交给某台VRRP路由器的。
Keepalived安装
配置主keepalived
//关闭防火墙 [root@master ~]# systemctl disable --now firewalld [root@master ~]# sed -i "s/SELINUX=enforcing/SELINUX=disabled/g" /etc/selinux/config [root@master ~]# setenforce 0 //安装keepalived [root@master ~]# yum -y install keepalived //查看安装生成的文件 [root@master ~]# rpm -ql keepalived /etc/keepalived /etc/keepalived/keepalived.conf /etc/sysconfig/keepalived /usr/bin/genhash /usr/lib/.build-id /usr/lib/.build-id/6c /usr/lib/systemd/system/keepalived.service /usr/libexec/keepalived /usr/sbin/keepalived //备服务器上安装keepalived [root@slave ~]# yum -y install keepalived //在master上安装nginx [root@master ~]# yum -y install nginx [root@master ~]# cd /usr/share/nginx/html/ [root@master html]# echo 'master' > index.html [root@master html]# systemctl enable --now nginx [root@master html]# ss -antl State Recv-Q Send-Q Local Address:Port Peer Address:Port LISTEN 0 128 0.0.0.0:80 0.0.0.0:* LISTEN 0 128 0.0.0.0:22 0.0.0.0:* LISTEN 0 128 [::]:80 [::]:* LISTEN 0 128 [::]:22 [::]:* //在slave上安装nginx [root@slave ~]# yum -y install nginx [root@slave ~]# cd /usr/share/nginx/html/ [root@slave html]# echo 'slave' > index.html [root@slave html]# systemctl enable --now nginx [root@slave html]# ss -antl State Recv-Q Send-Q Local Address:Port Peer Address:Port LISTEN 0 128 0.0.0.0:80 0.0.0.0:* LISTEN 0 128 0.0.0.0:22 0.0.0.0:* LISTEN 0 128 [::]:80 [::]:* LISTEN 0 128 [::]:22 [::]:*
Keepalived配置
配置主Keepalived
[root@master ~]# cd /etc/keepalived/ [root@master keepalived]# ls keepalived.conf [root@master keepalived]# mv keepalived.conf{,.bak} [root@master keepalived]# vim /etc/keepalived/keepalived.conf ! Configuration File for keepalived global_defs { router_id lb01 } vrrp_instance VI_1 { state MASTER interface ens160 //这里是修改成本机的网卡名称 virtual_router_id 51 priority 100 advert_int 1 authentication { auth_type PASS auth_pass baozi } virtual_ipaddress { 192.168.30.200 } } virtual_server 192.168.30.200 80 { delay_loop 6 lb_algo rr lb_kind DR persistence_timeout 50 protocol TCP real_server 192.168.30.131 80 { weight 1 TCP_CHECK { connect_port 80 connect_timeout 3 nb_get_retry 3 delay_before_retry 3 } } real_server 192.168.30.132 80 { weight 1 TCP_CHECK { connect_port 80 connect_timeout 3 nb_get_retry 3 delay_before_retry 3 } } } [root@master ~]# systemctl enable --now keepalived
配置备Keepalived
[root@slave ~]# cd /etc/keepalived/ [root@slave keepalived]# ls keepalived.conf [root@slave keepalived]# mv keepalived.conf{,.bak} [root@slave keepalived]# vim /etc/keepalived/keepalived.conf ! Configuration File for keepalived global_defs { router_id lb02 } vrrp_instance VI_1 { state BACKUP interface ens160 virtual_router_id 51 priority 90 advert_int 1 authentication { auth_type PASS auth_pass baozi } virtual_ipaddress { 192.168.30.200 } } virtual_server 192.168.30.200 80 { delay_loop 6 lb_algo rr lb_kind DR persistence_timeout 50 protocol TCP real_server 192.168.30.131 80 { weight 1 TCP_CHECK { connect_port 80 connect_timeout 3 nb_get_retry 3 delay_before_retry 3 } } real_server 192.168.30.132 80 { weight 1 TCP_CHECK { connect_port 80 connect_timeout 3 nb_get_retry 3 delay_before_retry 3 } } } [root@slave ~]# systemctl enable --now keepalived
让keepalived监控nginx负载均衡机
keepalived通过脚本来监控nginx负载均衡机的状态
在master上编写脚本
[root@master ~]# mkdir /scripts [root@master ~]# cd /scripts [root@master scripts]# vim check_n.sh #!/bin/bash nginx_status=$(ps -ef|grep -Ev "grep|$0"|grep '\bnginx\b'|wc -l) if [ $nginx_status -lt 1 ];then systemctl stop keepalived fi [root@master scripts]# chmod +x check_n.sh [root@master scripts]# vim notify.sh #!/bin/bash VIP=$2 sendmail (){ subject="${VIP}'s server keepalived state is translate" content="`date +'%F %T'`: `hostname`'s state change to master" echo $content | mail -s "$subject" 763342488@qq.com } case "$1" in master) nginx_status=$(ps -ef|grep -Ev "grep|$0"|grep '\bnginx\b'|wc -l) if [ $nginx_status -lt 1 ];then systemctl start nginx fi sendmail ;; backup) nginx_status=$(ps -ef|grep -Ev "grep|$0"|grep '\bnginx\b'|wc -l) if [ $nginx_status -gt 0 ];then systemctl stop nginx fi ;; *) echo "Usage:$0 master|backup VIP" ;; esac [root@master ~]# chmod +x /scripts/notify.sh [root@master ~]# ll 总用量 8 -rwxr-xr-x 1 root root 127 5月 21 12:56 check_n.sh -rwxr-xr-x 1 root root 661 5月 21 12:57 notify.sh
在slave上编写脚本
[root@slave ~]# mkdir /scripts [root@slave ~]# vim /scripts/check_n.sh #!/bin/bash nginx_status=$(ps -ef|grep -Ev "grep|$0"|grep '\bnginx\b'|wc -l) if [ $nginx_status -lt 1 ];then systemctl stop keepalived fi [root@slave ~]# vim /scripts/notify.sh #!/bin/bash VIP=$2 sendmail (){ subject="${VIP}'s server keepalived state is translate" content="`date +'%F %T'`: `hostname`'s state change to master" echo $content | mail -s "$subject" qinghao_yu@163.com } case "$1" in master) nginx_status=$(ps -ef|grep -Ev "grep|$0"|grep '\bnginx\b'|wc -l) if [ $nginx_status -lt 1 ];then systemctl start nginx fi sendmail ;; backup) nginx_status=$(ps -ef|grep -Ev "grep|$0"|grep '\bnginx\b'|wc -l) if [ $nginx_status -gt 0 ];then systemctl stop nginx fi ;; *) echo "Usage:$0 master|backup VIP" ;; esac [root@master ~]# chmod +x /scripts/check_n.sh [root@master ~]# chmod +x /scripts/notify.sh
配置Keepalived加入监控脚本的配置
配置主Keepalived
[root@master ~]# vim /etc/keepalived/keepalived.conf ! Configuration File for keepalived global_defs { router_id lb01 } vrrp_script nginx_check { script "/scripts/check_n.sh" interval 10 weight -20 } vrrp_instance VI_1 { state MASTER interface ens160 virtual_router_id 51 priority 100 advert_int 1 authentication { auth_type PASS auth_pass baozi } virtual_ipaddress { 192.168.30.200 } track_script { nginx_check } notify_master "/scripts/notify.sh master 192.168.100.200" notify_backup "/scripts/notify.sh backup 192.168.100.200" } virtual_server 192.168.30.200 80 { delay_loop 6 lb_algo rr lb_kind DR persistence_timeout 50 protocol TCP real_server 192.168.30.131 80 { weight 1 TCP_CHECK { connect_port 80 connect_timeout 3 nb_get_retry 3 delay_before_retry 3 } } real_server 192.168.30.132 80 { weight 1 TCP_CHECK { connect_port 80 connect_timeout 3 nb_get_retry 3 delay_before_retry 3 } } } [root@master ~]# systemctl restart keepalived
配置备Keepalived
backup无需检测nginx是否正常,当升级为MASTER时启动nginx,当降级为BACKUP时关闭
[root@slave ~]# vim /etc/keepalived/keepalived.conf ! Configuration File for keepalived global_defs { router_id lb02 } vrrp_instance VI_1 { state BACKUP interface ens160 virtual_router_id 51 priority 90 nopreempt advert_int 1 authentication { auth_type PASS auth_pass baozi } virtual_ipaddress { 192.168.30.200 } notify_master "/scripts/notify.sh master 192.168.100.200" notify_backup "/scripts/notify.sh backup 192.168.100.200" } virtual_server 192.168.30.200 80 { delay_loop 6 lb_algo rr lb_kind DR persistence_timeout 50 protocol TCP real_server 192.168.30.131 80 { weight 1 TCP_CHECK { connect_port 80 connect_timeout 3 nb_get_retry 3 delay_before_retry 3 } } real_server 192.168.30.132 80 { weight 1 TCP_CHECK { connect_port 80 connect_timeout 3 nb_get_retry 3 delay_before_retry 3 } } } [root@slave ~]# systemctl restart keepalived
测试验证
模拟master挂掉,slave继承.
[root@master ~]# systemctl start keepalived [root@master ~]# systemctl start nginx [root@master ~]# ss -antl State Recv-Q Send-Q Local Address:Port Peer Address:Port LISTEN 0 128 0.0.0.0:80 0.0.0.0:* LISTEN 0 128 0.0.0.0:22 0.0.0.0:* LISTEN 0 128 [::]:80 [::]:* LISTEN 0 128 [::]:22 [::]:* [root@slave ~]# systemctl start keepalived [root@slave ~]# ss -antl State Recv-Q Send-Q Local Address:Port Peer Address:Port LISTEN 0 128 0.0.0.0:22 0.0.0.0:* LISTEN 0 128 [::]:22 [::]:* //此时客户端正常访问VIP到master [root@client ~]# curl 192.168.30.200 master //模拟master挂掉 [root@master ~]# systemctl stop nginx [root@master ~]# ss -antl State Recv-Q Send-Q Local Address:Port Peer Address:Port LISTEN 0 128 0.0.0.0:22 0.0.0.0:* LISTEN 0 128 [::]:22 [::]:* //此时再访问VIP会访问到slave [root@client ~]# curl 192.168.30.200 slave //此时的slave会因为master挂掉,自动上位,启动nginx [root@slave ~]# ss -antl State Recv-Q Send-Q Local Address:Port Peer Address:Port LISTEN 0 128 0.0.0.0:80 0.0.0.0:* LISTEN 0 128 0.0.0.0:22 0.0.0.0:* LISTEN 0 128 [::]:80 [::]:* LISTEN 0 128 [::]:22 [::]:*
脑裂
在高可用(HA)系统中,当联系2个节点的“心跳线”断开时,本来为一整体、动作协调的HA系统,就分裂成为2个独立的个体。由于相互失去了联系,都以为是对方出了故障。两个节点上的HA软件像“裂脑人”一样,争抢“共享资源”、争起“应用服务”,就会发生严重后果——或者共享资源被瓜分、2边“服务”都起不来了;或者2边“服务”都起来了,但同时读写“共享存储”,导致数据损坏(常见如数据库轮询着的联机日志出错)。
对付HA系统“裂脑”的对策,目前达成共识的的大概有以下几条:
- 添加冗余的心跳线,例如:双线条线(心跳线也HA),尽量减少“裂脑”发生几率;
- 启用磁盘锁。正在服务一方锁住共享磁盘,“裂脑”发生时,让对方完全“抢不走”共享磁盘资源。但使用锁磁盘也会有一个不小的问题,如果占用共享盘的一方不主动“解锁”,另一方就永远得不到共享磁盘。现实中假如服务节点突然死机或崩溃,就不可能执行解锁命令。后备节点也就接管不了共享资源和应用服务。于是有人在HA中设计了“智能”锁。即:正在服务的一方只在发现心跳线全部断开(察觉不到对端)时才启用磁盘锁。平时就不上锁了。
- 设置仲裁机制。例如设置参考IP(如网关IP),当心跳线完全断开时,2个节点都各自ping一下参考IP,不通则表明断点就出在本端。不仅“心跳”、还兼对外“服务”的本端网络链路断了,即使启动(或继续)应用服务也没有用了,那就主动放弃竞争,让能够ping通参考IP的一端去起服务。更保险一些,ping不通参考IP的一方干脆就自我重启,以彻底释放有可能还占用着的那些共享资源
脑裂产生的原因
一般来说,脑裂的发生,有以下几种原因:
- 高可用服务器对之间心跳线链路发生故障,导致无法正常通信
- 因心跳线坏了(包括断了,老化)
- 因网卡及相关驱动坏了,ip配置及冲突问题(网卡直连)
- 因心跳线间连接的设备故障(网卡及交换机)
- 因仲裁的机器出问题(采用仲裁的方案)
- 高可用服务器上开启了 iptables防火墙阻挡了心跳消息传输
- 高可用服务器上心跳网卡地址等信息配置不正确,导致发送心跳失败
- 其他服务配置不当等原因,如心跳方式不同,心跳广插冲突、软件Bug等
脑裂的常见解决方案
在实际生产环境中,我们可以从以下几个方面来防止裂脑问题的发生:
- 同时使用串行电缆和以太网电缆连接,同时用两条心跳线路,这样一条线路坏了,另一个还是好的,依然能传送心跳消息
- 当检测到裂脑时强行关闭一个心跳节点(这个功能需特殊设备支持,如Stonith、feyce)。相当于备节点接收不到心跳消患,通过单独的线路发送关机命令关闭主节点的电源
- 做好对裂脑的监控报警(如邮件及手机短信等或值班).在问题发生时人为第一时间介入仲裁,降低损失。例如,百度的监控报警短信就有上行和下行的区别。报警消息发送到管理员手机上,管理员可以通过手机回复对应数字或简单的字符串操作返回给服务器.让服务器根据指令自动处理相应故障,这样解决故障的时间更短.
当然,在实施高可用方案时,要根据业务实际需求确定是否能容忍这样的损失。对于一般的网站常规业务.这个损失是可容忍的
对脑裂进行监控
对脑裂的监控应在备用服务器上进行,通过添加zabbix自定义监控进行。
监控什么信息呢?监控备上有无VIP地址
备机上出现VIP有两种情况:
- 发生了脑裂
- 正常的主备切换
//从服务器上编写脚本 [root@node2 ~]# mkdir /script [root@node2 ~]# cd /script [root@node2 script]# vim chkvip.sh #!/bin/bash ip address show dev ens160 |grep 192.168.30.200/32 if [ $? -eq 0 ];then echo '1' else echo '0' fi [root@node2 script]# chmod +x chkvip.sh [root@node2 ~]#vim /usr/local/zabbix/etc/zabbix_agentd.conf ... //自定义监控 UnsafeUserParameters=1 UserParameter=check_slave,/script/chkvip.sh //zabbix监控机查看能否接受 [root@node3 ~]# /usr/local/zabbix/bin/zabbix_get -s 192.168.94.143 -k check_slave 0