MHA：
MHA工作原理总结为以下几条：
- 从宕机崩溃的master保存二进制日志事件（binlog events）；
- 识别含有最新更新的slave；
- 应用差异的中继日志（relay log）到其他slave；
- 应用从master保存的二进制日志事件（binlog events）；
- 提升一个slave为新master；
- 使用其他的slave连接新的master进行复制。
MHA特点:
- 故障切换时间 10-30秒
- 自动监控Master以及故障转移
- 在线切换Master到异机
1.环境介绍
|
服务器角色 |
IP |
节点 |
端口 |
职责 |
|
manager、master |
192.168.163.3 |
mha1 |
3310 |
监控、主master |
|
slave1 |
192.168.163.16 |
mha2 |
3310 |
从slave |
|
slave2 |
192.168.163.20 |
mha3 |
3310 |
从slave |
OS：CentOS7.8
MySQL：Mysql-5.7.23
MHA：mha4mysql-manager-0.58-0.el7.centos.noarch.rpm
mha4mysql-node-0.58-0.el7.centos.noarch.rpm
Keepalived：keepalived-1.3.5-19.el7.x86_64
关闭防火墙
systemctl stop firewalld systemctl disable firewalld
1.1配置yum源
mha需要很多依赖包，很过国内的网络yum源没有一些依赖，可直接使用下面的yum源，会省去大部分时间
cat > /etc/yum.repos.d/epel.repo << EOF [epel] name=Extra Packages for Enterprise Linux 7 - $basearch failovermethod=priority enabled=1 gpgcheck=0 mirrorlist=http://mirrors.fedoraproject.org/mirrorlist?repo=epel-7&arch=$basearch EOF
1.2安装依赖包
三台主机都需要安装依赖包
yum install perl-DBI perl-DBD-MySQL perl-Config-Tiny perl-Log-Dispatch perl-Parallel-ForkManager perl-Time-HiRes perl-Params-Validate perl-DateTime -y yum install cpan perl-ExtUtils-Embed perl-ExtUtils-CBuilder perl-ExtUtils-MakeMaker -y
1.3安装软件
mha1:
三台主机都选择安装manager，node节点都安装
rpm -ivh mha4mysql-manager-0.58-0.el7.centos.noarch.rpm rpm -ivh mha4mysql-node-0.58-0.el7.centos.noarch.rpm
mha2:
rpm -ivh mha4mysql-node-0.58-0.el7.centos.noarch.rpm
mha3:
rpm -ivh mha4mysql-node-0.58-0.el7.centos.noarch.rpm
1.4修改主机名和hosts文件
三台主机都修改
[root@mha1 ~]# cat /etc/hosts 127.0.0.1 localhost localhost.localdomain localhost4 localhost4.localdomain4 ::1 localhost localhost.localdomain localhost6 localhost6.localdomain6 192.168.163.3 mha1 192.168.163.16 mha2 192.168.163.20 mha3
1.5 ssh 互通环境
mha1： ssh-keygen ssh-copy-id -i ~/.ssh/id_rsa.pub root@192.168.163.3 ssh-copy-id -i ~/.ssh/id_rsa.pub root@192.168.163.16 ssh-copy-id -i ~/.ssh/id_rsa.pub root@192.168.163.20
三台主机依次登录测试，直到可以直接登录其他主机
ssh 192.168.163.3 ssh 192.168.16316 ssh 192.168.163.20
2. mysql安装部署
2.1mysql配置文件
注意更改server_id，在同一个集群中，更改为不同的server_id
[root@mha1 ~]# vim /data/3310/my.cnf [mysqld] basedir=/application/mysql datadir=/data/3310/data socket=/data/3310/mysql.sock log_error=/data/3310/mysql.log port=3310 server_id=331003 log_bin=/data/3310/mysql-bin gtid-mode=on enforce-gtid-consistency=true slow_query_log=ON skip-name-resolve # 建议加上 非必须项 relay_log_purge = 0 # 关闭自动清理中继日志 log_slave_updates = 1 # 从库通过binlog更新的数据写进从库二进制日志中，必加，否则切换后可能丢失数据 relay-log = relay-log
2.2加入启动服务
[root@mha1 ~]# vim /etc/systemd/system/mysqld3310.service [Unit] Description=MySQL Server Documentation=man:mysqld(8) Documentation=http://dev.mysql.com/doc/refman/en/using-systemd.html After=network.target After=syslog.target [Install] WantedBy=multi-user.target [Service] User=mysql Group=mysql ExecStart=/application/mysql/bin/mysqld --defaults-file=/data/3310/my.cnf LimitNOFILE = 5000
2.3初始化mysql
/application/mysql/bin/mysqld --initialize-insecure --user=mysql --basedir=/application/mysql --datadir=/data/3310/data
2.4启动mysql
systemctl start mysqld3310 systemctl enable mysqld3310
3.一主两从搭建
3.1 mha1
mysql> create user 'slave'@'192.168.163.%' identified by '123'; mysql> create user mha@'192.168.163.%' identified by '123'; mysql> grant replication slave on *.* to slave@'192.168.163.%' identified by '123'; mysql> grant all on *.* to mha@'192.168.163.%'; mysql> flush privileges;
也可以选择用mysqldump的方式来恢复从库，此时是新建的库，并没有数据。
mysqldump -uroot -p123 -S /data/3310/mysql.sock -A --master-data=2 --single-transaction -R -E --triggers >/tmp/full.sql
从库source /tmp/full.sql
查看主库master状态
3.2 mha2
mysql> change master to master_host='192.168.163.3',master_user='slave',master_port=3310,master_password='123',master_log_file='master-log.000003',master_log_pos=234;
Last_IO_Error: Got fatal error 1236 from master when reading data from binary log: 'Could not find first log file name in binary log index file'
主库添加log-bin-index 参数后，从库报这个错误：Got fatal error 1236 from master when reading data from binary log: 'Could not find first log file name in binary log index file'
mysql> stop slave; Query OK, 0 rows affected (0.00 sec) mysql> reset slave; Query OK, 0 rows affected (0.01 sec) mysql> start slave; Query OK, 0 rows affected (0.00 sec) mysql> show slave statusG;
3.3 mha3
mysql> change master to master_host='192.168.163.3',master_user='slave',master_port=3310,master_password='123',master_log_file='master-log.000003',master_log_pos=234; mysql> stop slave; Query OK, 0 rows affected (0.00 sec) mysql> reset slave; Query OK, 0 rows affected (0.01 sec) mysql> start slave; Query OK, 0 rows affected (0.00 sec)
4.mha搭建
Manager工具
masterha_check_ssh 检查MHA的SSH配置状况
masterha_check_repl 检查MySQL复制状况
masterha_manger 启动MHA
masterha_check_status 检测当前MHA运行状态
masterha_master_monitor 检测master是否宕机
masterha_master_switch 控制故障转移（自动或者手动）
masterha_conf_host 添加或删除配置的server信息
Node工具
save_binary_logs 保存和复制master的二进制日志
apply_diff_relay_logs 识别差异的中继日志事件并应用于其它slave
filter_mysqlbinlog 去除不必要的ROLLBACK事件(MHA已不再使用这个工具)
purge_relay_logs 清除中继日志(不会阻塞SQL线程)
启动mha
nohup masterha_manager --conf=/etc/mha_master/mha.cnf --remove_dead_master --ignore_last_failover < /dev/null > /etc/mha_master/manager.log 2>&1 &
关闭mha
masterha_stop --conf=/etc/mha_master/mha.cnf
集群状态检查
masterha_check_status --conf=/etc/mha_master/mha.cnf
集群连接检查
masterha_check_ssh--conf=/etc/mha_master/mha.cnf
集群复制检查
masterha_check_repl --conf=/etc/mha_master/mha.cnf
集群切换
masterha_master_switch --conf=/etc/mha_master/mha.cnf --master_state=alive --new_master_host=192.168.163.16 --new_master_port=3310 --orig_master_is_new_slave --running_updates_limit=10000
4.1安装软件
见1.3 已安装
4.2配置文件
mha1: mkdir -p /etc/mha_master/app1 touch /etc/mha_master/manager.log cat > /etc/mha_master/mha.cnf <<EOF [server default] manager_log=/etc/mha_master/manager.log manager_workdir=/etc/mha_master/app1 password=123 ping_interval=1 remote_workdir=/tmp repl_password=123 repl_user=slave ssh_user=root user=mha [server1] candidate_master=1 hostname=192.168.163.3 port=3310 ssh_port=22 [server2] candidate_master=1 hostname=192.168.163.16 port=3310 ssh_port=22 [server3] candidate_master=1 hostname=192.168.163.20 port=3310 ssh_port=22 EOF
参数详情见网络
4.3集群状态检查
MHA连接检查
[root@mha1 ~]# masterha_check_ssh --conf=/etc/mha_master/mha.cnf
MHA复制检查
[root@mha1 ~]# masterha_check_ssh --conf=/etc/mha_master/mha.cnf
4.4开启集群
[root@mha1 ~]# nohup masterha_manager --conf=/etc/mha_master/mha.cnf --remove_dead_master --ignore_last_failover < /dev/null > /etc/mha_master/manager.log 2>&1 &
集群状态
[root@mha1 ~]# masterha_check_status --conf=/etc/mha_master/mha.cnf
5.手动故障切换
5.1停止mha监控
停止集群
[root@mha1 ~]# masterha_stop --conf=/etc/mha_master/mha.cnf
5.2切换主库为mha2
[root@mha1 ~]# masterha_master_switch --conf=/etc/mha_master/mha.cnf --master_state=alive --new_master_host=192.168.163.16 --new_master_port=3310 --orig_master_is_new_slave --running_updates_limit=10000 Sat Feb 20 09:44:43 2021 - [info] MHA::MasterRotate version 0.58. Sat Feb 20 09:44:43 2021 - [info] Starting online master switch.. Sat Feb 20 09:44:43 2021 - [info] Sat Feb 20 09:44:43 2021 - [info] * Phase 1: Configuration Check Phase.. Sat Feb 20 09:44:43 2021 - [info] Sat Feb 20 09:44:43 2021 - [warning] Global configuration file /etc/masterha_default.cnf not found. Skipping. Sat Feb 20 09:44:43 2021 - [info] Reading application default configuration from /etc/mha_master/mha.cnf.. Sat Feb 20 09:44:43 2021 - [info] Reading server configuration from /etc/mha_master/mha.cnf.. Sat Feb 20 09:44:44 2021 - [info] GTID failover mode = 1 Sat Feb 20 09:44:44 2021 - [info] Current Alive Master: 192.168.163.3(192.168.163.3:3310) Sat Feb 20 09:44:44 2021 - [info] Alive Slaves: Sat Feb 20 09:44:44 2021 - [info] 192.168.163.16(192.168.163.16:3310) Version=5.7.32-log (oldest major version between slaves) log-bin:enabled Sat Feb 20 09:44:44 2021 - [info] GTID ON Sat Feb 20 09:44:44 2021 - [info] Replicating from 192.168.163.3(192.168.163.3:3310) Sat Feb 20 09:44:44 2021 - [info] Primary candidate for the new Master (candidate_master is set) Sat Feb 20 09:44:44 2021 - [info] 192.168.163.20(192.168.163.20:3310) Version=5.7.32-log (oldest major version between slaves) log-bin:enabled Sat Feb 20 09:44:44 2021 - [info] GTID ON Sat Feb 20 09:44:44 2021 - [info] Replicating from 192.168.163.3(192.168.163.3:3310) Sat Feb 20 09:44:44 2021 - [info] Primary candidate for the new Master (candidate_master is set) It is better to execute FLUSH NO_WRITE_TO_BINLOG TABLES on the master before switching. Is it ok to execute on 192.168.163.3(192.168.163.3:3310)? (YES/no): yes Sat Feb 20 09:44:47 2021 - [info] Executing FLUSH NO_WRITE_TO_BINLOG TABLES. This may take long time.. Sat Feb 20 09:44:47 2021 - [info] ok. Sat Feb 20 09:44:47 2021 - [info] Checking MHA is not monitoring or doing failover.. Sat Feb 20 09:44:47 2021 - [info] Checking replication health on 192.168.163.16.. Sat Feb 20 09:44:47 2021 - [info] ok. Sat Feb 20 09:44:47 2021 - [info] Checking replication health on 192.168.163.20.. Sat Feb 20 09:44:47 2021 - [info] ok. Sat Feb 20 09:44:47 2021 - [info] 192.168.163.16 can be new master. Sat Feb 20 09:44:47 2021 - [info] From: 192.168.163.3(192.168.163.3:3310) (current master) +--192.168.163.16(192.168.163.16:3310) +--192.168.163.20(192.168.163.20:3310) To: 192.168.163.16(192.168.163.16:3310) (new master) +--192.168.163.20(192.168.163.20:3310) +--192.168.163.3(192.168.163.3:3310) Starting master switch from 192.168.163.3(192.168.163.3:3310) to 192.168.163.16(192.168.163.16:3310)? (yes/NO): yes Sat Feb 20 09:44:49 2021 - [info] Checking whether 192.168.163.16(192.168.163.16:3310) is ok for the new master.. Sat Feb 20 09:44:49 2021 - [info] ok. Sat Feb 20 09:44:49 2021 - [info] 192.168.163.3(192.168.163.3:3310): SHOW SLAVE STATUS returned empty result. To check replication filtering rules, temporarily executing CHANGE MASTER to a dummy host. Sat Feb 20 09:44:49 2021 - [info] 192.168.163.3(192.168.163.3:3310): Resetting slave pointing to the dummy host. Sat Feb 20 09:44:49 2021 - [info] ** Phase 1: Configuration Check Phase completed. Sat Feb 20 09:44:49 2021 - [info] Sat Feb 20 09:44:49 2021 - [info] * Phase 2: Rejecting updates Phase.. Sat Feb 20 09:44:49 2021 - [info] master_ip_online_change_script is not defined. If you do not disable writes on the current master manually, applications keep writing on the current master. Is it ok to proceed? (yes/NO): yes Sat Feb 20 09:44:50 2021 - [info] Locking all tables on the orig master to reject updates from everybody (including root): Sat Feb 20 09:44:50 2021 - [info] Executing FLUSH TABLES WITH READ LOCK.. Sat Feb 20 09:44:50 2021 - [info] ok. Sat Feb 20 09:44:50 2021 - [info] Orig master binlog:pos is mysql-bin.000002:194. Sat Feb 20 09:44:50 2021 - [info] Waiting to execute all relay logs on 192.168.163.16(192.168.163.16:3310).. Sat Feb 20 09:44:50 2021 - [info] master_pos_wait(mysql-bin.000002:194) completed on 192.168.163.16(192.168.163.16:3310). Executed 0 events. Sat Feb 20 09:44:50 2021 - [info] done. Sat Feb 20 09:44:50 2021 - [info] Getting new master's binlog name and position.. Sat Feb 20 09:44:50 2021 - [info] mysql-bin.000001:2509 Sat Feb 20 09:44:50 2021 - [info] All other slaves should start replication from here. Statement should be: CHANGE MASTER TO MASTER_HOST='192.168.163.16', MASTER_PORT=3310, MASTER_AUTO_POSITION=1, MASTER_USER='slave', MASTER_PASSWORD='xxx'; Sat Feb 20 09:44:50 2021 - [info] Sat Feb 20 09:44:50 2021 - [info] * Switching slaves in parallel.. Sat Feb 20 09:44:50 2021 - [info] Sat Feb 20 09:44:50 2021 - [info] -- Slave switch on host 192.168.163.20(192.168.163.20:3310) started, pid: 1716 Sat Feb 20 09:44:50 2021 - [info] Sat Feb 20 09:44:52 2021 - [info] Log messages from 192.168.163.20 ... Sat Feb 20 09:44:52 2021 - [info] Sat Feb 20 09:44:50 2021 - [info] Waiting to execute all relay logs on 192.168.163.20(192.168.163.20:3310).. Sat Feb 20 09:44:50 2021 - [info] master_pos_wait(mysql-bin.000002:194) completed on 192.168.163.20(192.168.163.20:3310). Executed 0 events. Sat Feb 20 09:44:50 2021 - [info] done. Sat Feb 20 09:44:50 2021 - [info] Resetting slave 192.168.163.20(192.168.163.20:3310) and starting replication from the new master 192.168.163.16(192.168.163.16:3310).. Sat Feb 20 09:44:50 2021 - [info] Executed CHANGE MASTER. Sat Feb 20 09:44:51 2021 - [info] Slave started. Sat Feb 20 09:44:52 2021 - [info] End of log messages from 192.168.163.20 ... Sat Feb 20 09:44:52 2021 - [info] Sat Feb 20 09:44:52 2021 - [info] -- Slave switch on host 192.168.163.20(192.168.163.20:3310) succeeded. Sat Feb 20 09:44:52 2021 - [info] Unlocking all tables on the orig master: Sat Feb 20 09:44:52 2021 - [info] Executing UNLOCK TABLES.. Sat Feb 20 09:44:52 2021 - [info] ok. Sat Feb 20 09:44:52 2021 - [info] Starting orig master as a new slave.. Sat Feb 20 09:44:52 2021 - [info] Resetting slave 192.168.163.3(192.168.163.3:3310) and starting replication from the new master 192.168.163.16(192.168.163.16:3310).. Sat Feb 20 09:44:52 2021 - [info] Executed CHANGE MASTER. Sat Feb 20 09:44:53 2021 - [info] Slave started. Sat Feb 20 09:44:53 2021 - [info] All new slave servers switched successfully. Sat Feb 20 09:44:53 2021 - [info] Sat Feb 20 09:44:53 2021 - [info] * Phase 5: New master cleanup phase.. Sat Feb 20 09:44:53 2021 - [info] Sat Feb 20 09:44:53 2021 - [info] 192.168.163.16: Resetting slave info succeeded. Sat Feb 20 09:44:53 2021 - [info] Switching master to 192.168.163.16(192.168.163.16:3310) completed successfully.
如果失败：
5.3启动mha
[root@mha1 ~]# nohup masterha_manager --conf=/etc/mha_master/mha.cnf --remove_dead_master --ignore_last_failover < /dev/null > /etc/mha_master/manager.log 2>&1 & [1]1733
5.4查看目前集群状态
状态检查已经显示master节点切换到mha2上
[root@mha1 ~]# masterha_check_status --conf=/etc/mha_master/mha.cnf mha (pid:1733) is running(0:PING_OK), master:192.168.163.16
mha1:原master
mha2:现在的master主库 192.168.163.16
可以看到现在集群软件仍然部署在192.168.163.3上，只不过主节点现在切换到了mha2上
mha3: 192.168.163.20
6.自动故障切换
根据手动切换的方法可以将master从mha2切换为mha1
6.1安装keepalived
[root@mha3 ~ 10:21:01]# yum install keepalived -y
6.2配置文件
[root@mha1 ~]# vim /etc/keepalived/keepalived.conf
! Configuration File for keepalived
global_defs {
notification_email {
admin@test.com
}
notification_email_from admin@test.com
smtp_server 127.0.0.1
smtp_connect_timeout 30
router_id MYSQL_HA
}
vrrp_script chk_mysql {
script "killall -0 mysql" # 监控mysql健康性脚本
insterval 1
weight -10
}
vrrp_instance VI_1 {
state BACKUP
interface ens33 #根据实际网络接口进行更改
virtual_router_id 51
priority 100 #优先级，mha1设置100，mha2设置99，mha3设置98
advert_int 1
nopreempt #不主动抢占资源，只在master上设置
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
192.168.163.23
}
track_script {
chk_mysql
}
}
6.3传输配置文件
[root@mha1 ~]# scp /etc/keepalived/keepalived.conf root@192.168.163.16:/etc/keepalived/keepalived.conf keepalived.conf 100% 1290 836.2KB/s 00:00 [root@mha1 ~]# scp /etc/keepalived/keepalived.conf root@192.168.163.20:/etc/keepalived/keepalived.conf keepalived.conf
6.4修改keepalived.conf
mha2：
priority 99 #优先级，mha1设置100，mha2设置99，mha3设置98
mha3：
priority 98 #优先级，mha1设置100，mha2设置99，mha3设置98
6.5启动keepalived
[root@mha1 ~]# systemctl start keepalived [root@mha1 ~]# systemctl status keepalived [root@mha1 ~]# systemctl enable keepalived
6.6模拟故障
关闭mha1的主节点
mha1：
[root@mha1 ~]# systemctl stop mysqld3310 [1]+ Done nohup masterha_manager --conf=/etc/mha_master/mha.cnf --remove_dead_master --ignore_last_failover < /dev/null > /etc/mha_master/manager.log 2>&1
查看集群状态
[root@mha1 /etc/keepalived]# masterha_check_repl -conf=/etc/mha_master/mha.cnf
mha2：
mysql> show slave statusG; Empty set (0.00 sec) ERROR: No query specified mysql> show master status; +------------------+----------+--------------+------------------+------------------------------------------------------------------------------------+ | File | Position | Binlog_Do_DB | Binlog_Ignore_DB | Executed_Gtid_Set | +------------------+----------+--------------+------------------+------------------------------------------------------------------------------------+ | mysql-bin.000002 | 234 | | | b883ad36-731c-11eb-8f2b-000c297b5e9b:1-3, d85d6dce-7287-11eb-9701-000c2902c557:1-8 | +------------------+----------+--------------+------------------+------------------------------------------------------------------------------------+ 1 row in set (0.00 sec) mysql> show processlist; +----+-------+----------------------+------+------------------+------+---------------------------------------------------------------+------------------+ | Id | User | Host | db | Command | Time | State | Info | +----+-------+----------------------+------+------------------+------+---------------------------------------------------------------+------------------+ | 11 | slave | 192.168.163.20:58982 | NULL | Binlog Dump GTID | 598 | Master has sent all binlog to slave; waiting for more updates | NULL | | 15 | root | localhost | NULL | Query | 0 | starting | show processlist | +----+-------+----------------------+------+------------------+------+---------------------------------------------------------------+------------------+ 2 rows in set (0.00 sec)
mha3：
mysql> show slave statusG;
6.7恢复故障节点
要先恢复mha1和mha2的主从关系，此时mha2是主库，mha1作为一个备库来恢复
mha2：
mha1：
mysql> change master to master_host='192.168.163.16',master_user='slave',master_port=3310,master_password='123',master_log_file='master-log.000002',master_log_pos=234; Query OK, 0 rows affected, 2 warnings (0.01 sec) mysql> stop slave; Query OK, 0 rows affected (0.00 sec) mysql> reset slave; Query OK, 0 rows affected (0.01 sec) mysql> start slave; Query OK, 0 rows affected (0.01 sec) mysql> show slave statusG;
检查集群状态，发现活着的节点并没有mha1
[root@mha1 /etc/keepalived]# masterha_check_repl -conf=/etc/mha_master/mha.cnf
此时需要重新添加mha1进配置文件，mha1已经被集群剔除，需要重新添加。mha1打开的一个过程是需要读取配置文件的，现在mha1已经不在控制文件里，需要手动添加一下
[server1] candidate_master=1 hostname=192.168.163.3 port=3310 ssh_port=22
重新启动
[root@mha1 /etc/keepalived]# nohup masterha_manager --conf=/etc/mha_master/mha.cnf --remove_dead_master --ignore_last_failover < /dev/null > /etc/mha_master/manager.log 2>&1 & [1] 5023
6.8重新检查集群状态
[root@mha1 /etc/keepalived]# masterha_check_status --conf=/etc/mha_master/mha.cnf
