关于基础介绍及原理略过，直接上操作步骤吧。
mysql：5.7
操作系统：centos7.4
地址规划：
MHA manage：10.21.20.89
mysql主：10.21.20.71
mysql从1：10.21.20.131
mysql从2：10.21.20.132
vip：10.21.20.77
- 安装好mysql及主从同步建立以后，先在主库上创建监听用户。
在各个主机上建立互信，为了方便直接一路回车到底。然后将私钥文件及authorized_keys文件复制给其他主机。四台服务器都要执行。create user 'admin'@'%' identified by 'admin'; grant all on *.* to 'admin'@'10.21.20.%' identified by 'admin';# ssh-keygen -t rsa Generating public/private rsa key pair. Enter file in which to save the key (/root/.ssh/id_rsa): Enter passphrase (empty for no passphrase): Enter same passphrase again: Your identification has been saved in /root/.ssh/id_rsa. Your public key has been saved in /root/.ssh/id_rsa.pub. The key fingerprint is: SHA256:wyMb8WZ6yJ0jxXN1FgWOwbVAPwW/GJRaKHfaefvXak0 The key's randomart image is: +---[RSA 2048]----+ | o=o*+o| | . ooX = | | . o O.O .| | = + =oo.| | o S . ....| | . @ * .E| | * = oo| | o . . =| | ....| +----[SHA256]-----+ # ssh-copy-id -i .ssh/id_rsa.pub migumysql@10.21.20.71 # ssh-copy-id -i .ssh/id_rsa.pub migumysql@10.21.20.131 # ssh-copy-id -i .ssh/id_rsa.pub migumysql@10.21.20.132 - 安装MHA
首先下载mha4mysql-manager、mha4mysql-node。
由于我的安装环境没有公网，因此我将下载好的MHA安装包及依赖包做成了本地yum源。具体方法如下：# wget https://github.com/yoshinorim/mha4mysql-manager/releases/download/v0.58/mha4mysql-manager-0.58-0.el7.centos.noarch.rpm # wget https://github.com/yoshinorim/mha4mysql-node/releases/download/v0.58/mha4mysql-node-0.58-0.el7.centos.noarch.rpm
（1）找到一台可以连接公网的服务器，利用yum工具将依赖包下载到服务器。做成yum源以后，打包上传到安装的服务器。
(2) 将mha4mysql-manager、mha4mysql-node复制到/home/mysql/MHA/mharepo目录下，然后创建yum源。# yum install yum-utils -y # yumdownloader mha4mysql-manager mha4mysql-node --resolve --destdir=/home/mysql/MHA/mharepo # yum install -y createrepo
# createrepo -pdo /home/mysql/MHA/mharepo
# cd /home/mysql/MHA/
# tar -czvf mharepo.tar.gz mharepo
（3）将mharepo.tar.gz上传四台需要安装的服务器解压。然后新建一个yum文件，指向解压的目录。
# cat /etc/yum.repos.d/abc.repo
[abc]
name=test
baseurl=file:///root/mharepo
gpgcheck=0
enabled=1
(4) 安装MHA，在3台mysql上安装node，在管理节点安装manager。
# yum install mha4mysql-manager -y
# yum install mha4mysql-node -y
- 在管理节点编辑MHA的配置文件，并启动MHA manager
测试ssh及集群状态# cat conf/mha.cnf [server default] manager_workdir=/home/mysql/MHA manager_log=/home/mysql/MHA/log/manager.log user=admin password=admin ssh_user=mysql ssh_port=22 repl_user=slave repl_password=abc123 ping_interval=1 shutdown_script="" master_ip_online_change_script="" report_script="" master_ip_failover_script="/home/mysql/MHA/scripts/master_ip_failover" [server1] hostname=10.21.20.71 port=13306 candidate_master=1 master_binlog_dir="/home/mysql/mysql/logs/" [server2] hostname=10.21.20.131 port=13306 candidate_master=1 master_binlog_dir="/home/mysql/mysql/logs/" [server3] hostname=10.21.20.132 port=13306# masterha_check_ssh --conf=/home/mysql/MHA/conf/mha.cnf # masterha_check_repl --conf=/home/mysql/MHA/conf/mha.cnf 忽略部分内容 10.21.20.71(10.21.20.71:13306) (current master) +--10.21.20.131(10.21.20.131:13306) +--10.21.20.132(10.21.20.132:13306) Sat Apr 11 11:45:12 2020 - [info] Checking replication health on 10.21.20.131.. Sat Apr 11 11:45:12 2020 - [info] ok. Sat Apr 11 11:45:12 2020 - [info] Checking replication health on 10.21.20.132.. Sat Apr 11 11:45:12 2020 - [info] ok. Sat Apr 11 11:45:12 2020 - [warning] master_ip_failover_script is not defined. Sat Apr 11 11:45:12 2020 - [warning] shutdown_script is not defined. Sat Apr 11 11:45:12 2020 - [info] Got exit code 0 (Not master dead). MySQL Replication Health is OK.
直接nohup masterha_manager -conf=conf/mha.cnf > log/mha-0318.log &启动，当crt关闭时MHA manager可能会自动停止，网上找了下资料，可能是crt的bug，将启动命令写到shell里可以避免此情况。有待验证。# ln -s /home/migumysql/mysql/bin/mysql /usr/local/bin/mysql # ln -s /home/migumysql/mysql/bin/mysqlbinlog /usr/local/bin/mysqlbinlog# cat start_mah.sh #!/bin/bash cd `dirname $0` nohup masterha_manager -conf=conf/mha.cnf > log/mha-0410.log & # ./start_mah.sh - 在数据库节点安装keepalived（本例中为10.21.20.71、10.21.20.131）。主库的priority要比从库的高。
# yum install keepalived -y # cat /etc/keepalived/keepalived.conf global_defs { notification_email { 123456.qq.com } notification_email_from 123456.qq.com smtp_server 127.0.0.1 smtp_connect_timeout 30 router_id MYSQL-HA script_user root enable_script_security } vrrp_script check_mysql_alived { script "/etc/keepalived/mysql_monitor.sh" interval 5 } vrrp_instance VI_1 { state BACKUP interface eth0 virtual_router_id 191 priority 120 nopreempt advert_int 2 authentication { auth_type PASS auth_pass Cmcc } track_script { check_mysql_alived } virtual_ipaddress { 10.21.20.77/24 } } # # cat /etc/keepalived/mysql_monitor.sh #!/bin/bash check(){ /home/migumysql/mysql/bin/mysql -u'root' -p'Migu123!' -S /home/migumysql/mysql/logs/mysql.sock -e "select user()" > /dev/null 2>&1 r1=$? r2=$(/usr/sbin/ss -nltp | grep 3306 | wc -l) if [[ $r1 -ne 0 ]] || [[ $r2 -eq 0 ]];then echo "error" else echo "ok" fi } if [ $(check) = "ok" ];then echo "MySQL OK!" exit 0 else sleep 2 if [ $(check) = "error" ];then systemctl stop keepalived pkill -9 keepalived echo "MySQL Error! Keepalived Stop." fi fi # # chmod a+x /etc/keepalived/mysql_monitor.sh # systemctl start keepalived - 此时查看vip是否在主库上，从库没有vip。验证正确后，停止主库查看主从是否自动切换，vip是否会漂移到从库。
# tailf log/manager.log 忽略部分内容 ----- Failover Report ----- mha: MySQL Master failover 10.21.20.71(10.21.20.71:13306) to 10.21.20.131(10.21.20.131:13306) succeeded Master 10.21.20.71(10.21.20.71:13306) is down! Check MHA Manager logs at localhost:/home/migumysql/MHA/log/manager.log for details. Started automated(non-interactive) failover. The latest slave 10.21.20.131(10.21.20.131:13306) has all relay logs for recovery. Selected 10.21.20.131(10.21.20.131:13306) as a new master. 10.21.20.131(10.21.20.131:13306): OK: Applying all logs succeeded. 10.21.20.132(10.21.20.132:13306): This host has the latest relay log events. Generating relay diff files from the latest slave succeeded. 10.21.20.132(10.21.20.132:13306): OK: Applying all logs succeeded. Slave started, replicating from 10.21.20.131(10.21.20.131:13306) 10.21.20.131(10.21.20.131:13306): Resetting slave info succeeded. Master failover to 10.21.20.131(10.21.20.131:13306) completed successfully.
在主库和从库上观察ip发现，vip已经自动漂移到从库上。至此高可用完成。
