一、概述
本文将介绍mysql的MM+Keepalived方案。该方案由两个mysql服务器组成，这两个mysql互为主备。其中一台主作为写服务器，另一台主作为读服务器。通过keepalived软件管理写vip，当承担写服务器的mysql出现故障时，将写vip漂移到读服务器上，实现高可用。
二、节点介绍
本次实验采用2台虚拟机，操作系统版本Centos6.10，mysql版本5.7.25
node1 10.40.16.61 主库 提供写服务
node2 10.40.16.62 主库 提供读服务
还须预留1个vip，现在不用配置，这里先提一下，后面的安装步骤用得到
10.40.16.71 写vip
三、安装
1. 配置双主架构
安利一个自己写的mysql一键安装脚本https://www.cnblogs.com/ddzj01/p/10678296.html
mysql搭建完成后，就可以配置互为主备的架构了。
这样node1和node2就互为主备了
在node2上将数据库设置为只读模式
(root@localhost)[(none)]> set global read_only = 1;
2. 安装keepalive软件
node1&node2:
yum install -y keepalived
四、修改配置文件
1. node1
编辑配置文件/etc/keepalived/keepalived.conf
! Configuration File for keepalived vrrp_script chk_mysql { script "/etc/keepalived/check_mysql.sh" # 自定义检查脚本 interval 30 # 设置检查间隔时长，可自行设定 } vrrp_instance VI_1 { state BACKUP # BACKUP状态，具体意思后面介绍 interface eth0 virtual_router_id 51 priority 100 advert_int 1 nopreempt # 防止主库切换到从库后，主库恢复后自动切换回主库 authentication { auth_type PASS auth_pass 1111 } track_script { chk_mysql } virtual_ipaddress { 10.40.16.71/24 # vip } }
编辑检查mysql主库的脚本文件/etc/keepalived/check_mysql.sh
#!/bin/bash source /root/.bash_profile ###填数据库相关信息### DB_USER='root' DB_PASSWD='root' U_EMAIL='[email protected]' ###################### ###判断如果上次检查的脚本还没执行完，则退出此次执行 if [ `ps -ef | grep -w "$0" | grep -v "grep" | wc -l` -gt 2 ]; then exit 0 fi mysql_con="mysql -u$DB_USER -p$DB_PASSWD" error_log="/etc/keepalived/logs/check_mysql.err" ###如果error_log目录不存在则创建目录 if [ -d /etc/keepalived/logs ]; then usleep else mkdir -p /etc/keepalived/logs fi ###定义一个简单判断mysql是否可用的函数 function execute_query { $mysql_con -e "select 1;" 2>> $error_log } ###定义无法执行查询，且mysql服务异常时的处理函数 function service_error { echo -e "`date "+%F %H:%M:%S"` ----mysql service error, now stop keepalived----" >> $error_log service keepalived stop >> $error_log 2>&1 echo "master1 keepalived stopped" | mail -s "master1 keepalived stopped, please take notice!" $U_EMAIL 2>> $error_log echo -e "\n---------------------------------------------------------\n" >> $error_log } ###定义无法执行查询，但mysql服务正常的处理函数 function query_error { echo -e "`date "+%F %H:%M:%S"` ----query error, but mysql service ok, retry after 30s----" >> $error_log sleep 30 execute_query if [ $? -ne 0 ]; then echo -e "`date "+%F %H:%M:%S"` ----still can't execute query----" >> $error_log ###关闭本机mysql echo -e "`date "+%F %H:%M:%S"` ----stop mysql service----" >> $error_log service mysql stop &>> $error_log sleep 2 ###给执行和缓冲时间 ###关闭本机keepalived echo -e "`date "+%F %H:%M:%S"` ----stop keepalived----" >> $error_log service keepalived stop &>> $error_log echo "master1 keepalived stopped" | mail -s "master1 keepalived stopped, please take notice!" $U_EMAIL 2>> $error_log echo -e "\n---------------------------------------------------------\n" >> $error_log else echo -e "`date "+%F %H:%M:%S"` ----query ok after 30s----" >> $error_log echo -e "\n---------------------------------------------------------\n" >> $error_log fi } ###检查开始: 执行查询 execute_query if [ $? -ne 0 ]; then service mysql status &> /dev/null if [ $? -ne 0 ]; then service_error else query_error fi fi
chmod +x /etc/keepalived/check_mysql.sh
该脚本的作用有两个
a. mysql进程挂掉了，直接关闭keepalived服务
b. mysql进程正常，但是无法执行查询，比如进程卡死了等等原因。则同时关闭mysql和keepalived
2. node2
编辑配置文件/etc/keepalived/keepalived.conf
! Configuration File for keepalived
vrrp_instance VI_1 {
state BACKUP # BACKUP状态，具体意思后面介绍
interface eth0
virtual_router_id 51
priority 90 # 优先级设置为90，这个值设置比节点1低
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
notify_master /etc/keepalived/notify_master_mysql.sh
virtual_ipaddress {
10.40.16.71/24 # vip
}
}
编辑检查mysql从库的脚本文件/etc/keepalived/notify_master_mysql.sh
#!/bin/bash source /root/.bash_profile ###当keepalived监测到本机转为MASTER状态时，执行该脚本 ###填数据库相关信息### DB_USER='root' DB_PASSWD='root' U_EMAIL='[email protected]' ###################### change_log='/etc/keepalived/logs/state_change.log' mysql_con="mysql -u$DB_USER -p$DB_PASSWD" echo -e "`date "+%F %H:%M:%S"` ----master2 keepalived change to MASTER----" >> $change_log ###如果error_log目录不存在则创建目录 if [ -d /etc/keepalived/logs ]; then usleep else mkdir -p /etc/keepalived/logs fi slave_info() { ###检查从库状态 slave_stat=`$mysql_con -e "show slave status\G"` Slave_IO_Running=`echo $slave_stat | egrep -w "Slave_IO_Running" | awk '{print $2}'` Slave_SQL_Running=`echo $slave_stat | egrep -w "Slave_SQL_Running" | awk '{print $2}'` Master_Log_File=`echo $slave_stat | egrep -w "Master_Log_File" | awk '{print $2}'` Read_Master_Log_Pos=`echo $slave_stat | egrep -w "Read_Master_Log_Pos" | awk '{print $2}'` Relay_Master_Log_File=`echo $slave_stat | egrep -w "Relay_Master_Log_File" | awk '{print $2}'` Exec_Master_Log_Pos=`echo $slave_stat | egrep -w "Exec_Master_Log_Pos" | awk '{print $2}'` } action() { ###解除read_only属性 echo -e "`date "+%F %H:%M:%S"` ----set read_only = 0 on master2----" >> $change_log $mysql_con -e "set global read_only = 0;" 2>> $change_log echo "master2 keepalived change to MASTER，线上数据库切换至master2" | mail -s "master2 keepalived change to MASTER" $U_EMAIL 2>> $change_log echo -e "---------------------------------------------------------\n" >> $change_log } slave_info if [ $Slave_SQL_Running == 'Yes' ]; then i=0 #一个计数器 ###判断从master接收到的binlog是否全部在本地执行(这样仍无法完全确定从库已追上主库，因为无法完全保证io_thread没有延时(但由网络传输问题导致的从库落后的概率很小) until [ $Master_Log_File == $Relay_Master_Log_File -a $Read_Master_Log_Pos == $Exec_Master_Log_Pos ] do if [ $i -lt 10 ]; then #将等待exec_pos追上read_pos的时间限制为20s echo -e "`date "+%F %H:%M:%S"` ----Relay_Master_Log_File=$Relay_Master_Log_File, Exec_Master_Log_Pos=$Exec_Master_Log_Pos is behind Master_Log_File=$Master_Log_File, Read_Master_Log_Pos=$Read_Master_Log_Pos, wait......" >> $change_log #输出消息到日志，等待exec_pos=read_pos i=$(($i+1)) sleep 2 slave_info else echo -e "The waits time is more than 20s,now force change. Master_Log_File=$Master_Log_File Read_Master_Log_Pos=$Read_Master_Log_Pos Relay_Master_Log_File=$Relay_Master_Log_File Exec_Master_Log_Pos=$Exec_Master_Log_Pos" >> $change_log action exit 0 fi done action else echo -e "master2's slave status is not running,now force change. Master_Log_File=$Master_Log_File Read_Master_Log_Pos=$Read_Master_Log_Pos Relay_Master_Log_File=$Relay_Master_Log_File Exec_Master_Log_Pos=$Exec_Master_Log_Pos" >> $change_log action fi
chmod +x /etc/keepalived/notify_master_mysql.sh
该脚本的作用有三个
a. slave sql线程没有运行，直接将从库只读关闭
c. slave sql线程正在运行，如果从库没有延迟，直接将从库只读关闭
b. slave sql线程正在运行，如果从库有延迟，等待一段时间(这个自己设置)再将从库只读关闭
这里解释下keepalived.conf的"state BACKUP"的意思，在Keepalived中有两种模式，分别是master->backup模式和backup->backup模式，这两种模式有什么区别呢？
在master->backup模式下，一旦主库宕掉，虚拟IP会自动漂移到从库，当主库修复后，keepalived启动后，还会把虚拟IP抢过来，即使你设置nopreempt（不抢占）的方式抢占IP的动作也会发生。
在backup->backup模式下，当主库宕掉后虚拟IP会自动漂移到从库上，当原主恢复之后重启keepalived服务，并不会抢占新主的虚拟IP，即使是原主优先级高于从库的优先级别，也不会抢占虚拟IP。
所以，为了减少虚拟IP的漂移次数，生产中我们通常是把修复好的主库当做新主库的备库。因而采用backup->backup模式居多。
五、启动Keepailived
node1&node2:
service keepalived start
注意因为我们使用的是backup->backup模式，所以启动keepalived的顺序需要先启动node1，再启动node2，这样vip才会在node1上。如果先启动node2，再启动node1，node1并不会把虚拟IP抢过来。
六、测试
1. 模拟主库宕机的情况
关闭node1的mysql数据库
service mysql stop
查看node1的keepalived日志/etc/keepalived/logs/check_mysql.err
查看node1和node2的vip，发现vip已经转移到了node2
node2的read_only参数也变成了off
node2的切换日志可以查看/etc/keepalived/logs/state_change.log
看看有没有收到邮件，呵呵:-)
failover正常！
2. 模拟主库正常，但是无法查询
重新打开node1的mysql数据库和keepalived服务
service mysql start
service keepalived start
重启node2的keepalived服务
service keepalived restart
在node2上将数据库设置为只读模式
(root@localhost)[(none)]> set global read_only = 1;
这样就跟最初的状态一致了，主库是node1，从库是node2，vip在node1上。
在node1上将参数max_connections设置得足够小
(root@localhost)[(none)]> set global max_connections = 2;
在node1上然后多开几个连接，直到出现无法连接的情况。以此来模拟无法查询的情况。
查看node1的keepalived日志/etc/keepalived/logs/check_mysql.err
查看node1，发现keepalived和mysql服务都已经停止
[root@mysqla ~]# service mysql status
MySQL is not running
[root@mysqla ~]# service keepalived status
keepalived is stopped
node2的read_only参数也变成了off
vip也已经漂移过来了
[root@mysqlb ~]# ip a
failover正常！
七、手动切换
还有一个场景就是，如何手工切换。举个例子，node2目前是主库，但是node1通过各种办法修复好了，我想让node1当主库。
重启node1的mysql服务
service mysql restart
这样node2是主库，node1是备库，vip在node2上
在node2上创建手工切换的脚本vi /etc/keepalived/manual_switch_to_master
#!/bin/bash source /root/.bash_profile ###在master2上手动执行将主库切换回master1的操作 ###填数据库相关信息### DB_USER='root' DB_PASSWD='root' U_EMAIL='[email protected]' MASTER1='10.40.16.61' MASTER2='10.40.16.62' REPL_USER='repl' REPL_PASSWD='123456' MASTER1_MYSQL_PATH='/usr/local/mysql/bin' ###################### ###如果error_log目录不存在则创建目录 if [ -d /etc/keepalived/logs ]; then usleep else mkdir -p /etc/keepalived/logs fi mysql_con="mysql -u$DB_USER -p$DB_PASSWD" echo -e "`date "+%F %H:%M:%S"` ----change to BACKUP manually----" >> /etc/keepalived/logs/state_change.log echo -e "`date "+%F %H:%M:%S"` ----set read_only = 1 on master2----" >> /etc/keepalived/logs/state_change.log $mysql_con -e "set global read_only = 1;" >> /etc/keepalived/logs/state_change.log ###kill掉当前客户端连接 echo -e "`date "+%F %H:%M:%S"` ----kill current client thread----" >> /etc/keepalived/logs/state_change.log if [ -e /tmp/kill.sql ]; then rm -f /tmp/kill.sql &> /dev/null fi ###这里其实是一个批量kill线程的小技巧 $mysql_con -e 'select concat("kill ",id,";") from information_schema.processlist where command="Query" or command="Execute" into outfile "/tmp/kill.sql";' $mysql_con -e "source /tmp/kill.sql" 2>> /etc/keepalived/logs/state_change.log sleep 2 ###给kill一个执行和缓冲时间 slave_info() { ###检查从库状态 slave_stat=`mysql -u$REPL_USER -p$REPL_PASSWD -h$MASTER1 -e "show slave status\G"` Master_Log_File=`echo $slave_stat | egrep -w "Master_Log_File" | awk '{print $2}'` Read_Master_Log_Pos=`echo $slave_stat | egrep -w "Read_Master_Log_Pos" | awk '{print $2}'` Relay_Master_Log_File=`echo $slave_stat | egrep -w "Relay_Master_Log_File" | awk '{print $2}'` Exec_Master_Log_Pos=`echo $slave_stat | egrep -w "Exec_Master_Log_Pos" | awk '{print $2}'` } slave_info until [ $Read_Master_Log_Pos = $Exec_Master_Log_Pos -a $Master_Log_File = $Relay_Master_Log_File ] do echo -e "`date "+%F %H:%M:%S"` ----Relay_Master_Log_File=$Relay_Master_Log_File, Exec_Master_Log_Pos=$Exec_Master_Log_Pos is behind Master_Log_File=$Master_Log_File, Read_Master_Log_Pos=$Read_Master_Log_Pos, wait......" >> /etc/keepalived/logs/state_change.log sleep 2 slave_info done ###然后解除master1的read_only属性并打开keepalived服务 echo -e "`date "+%F %H:%M:%S"` ----set read_only = 0 on master1----" >> /etc/keepalived/logs/state_change.log ssh $MASTER1 "$MASTER1_MYSQL_PATH/mysql -u$DB_USER -p$DB_PASSWD -e 'set global read_only = 0;' && /etc/init.d/keepalived start" 2>> /etc/keepalived/logs/state_change.log ###重启master2的keepalived服务，使VIP漂移到master1 echo -e "`date "+%F %H:%M:%S"` ----make VIP move to master1----" >> /etc/keepalived/logs/state_change.log /etc/init.d/keepalived restart &>> /etc/keepalived/logs/state_change.log echo "master2 keepalived restart，vip change to master1" | mail -s "master2 keepalived change to BACKUP" $U_EMAIL 2>> /etc/keepalived/logs/state_change.log echo -e "\n--------------------------------------------------\n" >> /etc/keepalived/logs/state_change.log
在node2上执行脚本
chmod +x /etc/keepalived/manual_switch_to_master
[root@mysqlb keepalived]# sh /etc/keepalived/manual_switch_to_master
在node2上查看ip，发现vip不见了
在node2上查看read_only参数
在node1上查看vip，发现vip已经漂移过来了
手工切换完成！
八、总结
mm+keepalive的配置简单，相对于传统的主从架构，能实现比较简单的写库故障转移。
文章参考：https://www.cnblogs.com/ivictor/p/5522383.html