小树不倒我不倒 2019-05-13
MHA:Master High Availability,对主节点进行监控,可实现自动故障转移至其他从节点;通过提升某一从节点为新的主节点,基于主从复制实现,还需要客户端配合实现,目前MHA主要支持一主二从,即一台充当master,一台充当备用master,另外一台充当从数据库,出于机器成本的考虑,淘宝进行了改造,目前淘宝TMHA已经一主一从。
MHA是由一台manager服务器远程监控主服务器,当主服务器挂了提升一台从服务器作为主服务器。
当主节点挂了,manager首先要查看哪台从节点,同步的数据最多,然后提升同步最多的从节点为主节点,再将其余的MySQL服务器对他做从节点。
如果原主节点没彻底死透,manager会让新的主机通过ssh协议远程连接到原先的主节点,拉取二进制日志进行同步。如果主节死透了那就放弃。
主机 | IP |
---|---|
Manager | 192.168.73.111 |
Master | 192.168.73.110 |
Slave1 | 192.168.73.112 |
Slave2 | 192.168.73.113 |
1.安装chrony服务
[root@Manager ~]# yum install -y chrony
2.修改chrony配置文件
[root@Manager ~]# vim /etc/chrony.conf server 172.22.0.1 iburst allow 192.168.0.0/16 local stratum 10
3.启动chrony服务
[root@Manager ~]# systemctl start chronyd
4.将MySQL服务器与Manager服务器进行时间同步
4.1在所有MySQL主机上修改配置文件并启动,并启动服务
[root@Master ~]# sed -i '/^server 0/i server 192.168.73.111 iburst' /etc/chrony.conf [root@Master ~]# systemctl start chronyd
4.2确认时间同步
[root@Master ~]# chronyc sources -v 210 Number of sources = 1 .-- Source mode '^' = server, '=' = peer, '#' = local clock. / .- Source state '*' = current synced, '+' = combined , '-' = not combined, | / '?' = unreachable, 'x' = time may be in error, '~' = time too variable. || .- xxxx [ yyyy ] +/- zzzz || Reachability register (octal) -. | xxxx = adjusted offset, || Log2(Polling interval) --. | | yyyy = measured offset, || \ | | zzzz = estimated error. || | | \ MS Name/IP address Stratum Poll Reach LastRx Last sample =============================================================================== ^* 192.168.73.111 4 6 377 54 +25us[ +41us] +/- 105ms
当主节点宕机,manager会让从节点通过ssh协议去尝试连接主节点,并拉取二进制日志,所以要时用密钥的认证方式让从节点登陆到主节点拉取数据。
1.在manager服务器上生成私钥文件
[root@Manager ~]# ssh-keygen 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:yAvC2PJUlRyAf1udlrVXzmIsUljTdUdW6X6FVpQ3Ajo root@Manager The key's randomart image is: +---[RSA 2048]----+ | ..ooo ++. +%| | . .o o oo.=*| | .. E = oo*o| | + ...... B o B.+| |o = ..ooS. . =...| | + . ... ..| | . . .| | | | | +----[SHA256]-----+
2.将公钥文件复制给自己
[root@Manager ~]# ssh-copy-id 127.0.0.1
3.将整个~/.ssh目录复制给所有的MySQL主机
[root@Manager ~]# scp -r ~/.ssh 192.168.73.110:/root
至此所有环境准备完毕
1.修改配置文件
[root@Master ~]# vim /etc/my.cnf [mysqld] server-id=1 log-bin binlog-format=row skip_name_resolve
2.启动数据库服务
[root@Master ~]# systemctl start mariadb
3.创建主从复制账号
[root@Master ~]# mysql -e "GRANT REPLICATION SLAVE ON *.* TO 'repluser'@'192.168.73.%' IDENTIFIED BY 'CentOS';"
4.添加mha的管理账号,让管理节点远程连接到主机用来设置主从调整
[root@Master ~]# mysql -e "GRANT ALL ON *.* TO 'mhauser'@'192.168.73.%' IDENTIFIED BY 'centos';"
1.修改配置文件
[root@Slave1 ~]# vim /etc/my.cnf [mysqld] server-id=2 read-only log-bin relay_log_purge=0 skip_name_resolve
2.启动服务
[root@Slave1 ~]# systemctl start mariadb
3.配置CHANGE MASTER TO
MariaDB [(none)]> CHANGE MASTER TO MASTER_HOST='192.168.73.110', MASTER_USER='repluser',MASTER_PASSWORD='centos',MASTER_PORT=3306,MASTER_LOG_FILE='mariadb-bin.000001',MASTER_LOG_POS=245; Query OK, 0 rows affected (0.00 sec)
4.启动线程
MariaDB [(none)]> START SLAVE; Query OK, 0 rows affected (0.00 sec)
在Slave2节点上也执行相同的操作,此处步骤省略,需要注意server-id需要修改为和其他主从节点不同
5.测试
主节点导入hellodb库
[root@Master ~]# mysql < hellodb_innodb.sql
从节点查看是否同步
slave1
[root@Slave1 ~]# mysql -e "SHOW DATABASES;" +--------------------+ | Database | +--------------------+ | information_schema | | hellodb | | mysql | | performance_schema | | test | +--------------------+
Slave2
[root@Slave2 ~]# mysql -e "SHOW DATABASES;" +--------------------+ | Database | +--------------------+ | information_schema | | hellodb | | mysql | | performance_schema | | test | +--------------------+
1.在管理节上安装mha4mysql-manager、mha4mysql-node,将两个包放在同一目录下
[root@Manager ~]# yum install *.rpm -y #这两个包有依赖管理需要一起安装
2.在所有被管理节点上安装mha4mysql-node
[root@Master ~]# yum install mha4mysql-node-0.56-0.el6.noarch.rpm -y
[root@Slave1 ~]# yum install mha4mysql-node-0.56-0.el6.noarch.rpm -y
[root@Slave2 ~]# yum install mha4mysql-node-0.56-0.el6.noarch.rpm -y
3.在管理节点上创建配置文件
[root@Manager ~]# vim /etc/mha/aap1.conf [server default] user=mhauser password=centos manager_workdir=/data/mastermha/app1/ manager_log=/data/mastermha/app1/manager.log remote_workdir=/data/mastermha/app1/ ssh_user=root repl_user=repluser repl_password=centos ping_interval=1 [server1] hostname=192.168.73.110 candidate_master=1 [server2] hostname=192.168.73.112 candidate_master=1 [server3] hostname=192.168.73.113 candidate_master=1
4.做检查
4.1检查ssh连接
[root@Manager ~]# masterha_check_ssh --conf=/etc/mha/aap1.conf
4.2检查主从复制
[root@Manager ~]# masterha_check_repl --conf=/etc/mha/aap1.conf
5.以上两项全部成功后启动程序
mha这个程序是跑在前台的,一次性的可以使用nohub或screen来解决跑在前台的问题
[root@Manager ~]# masterha_manager --conf=/etc/mha/aap1.conf
1.在master上跑个存储过程,导入存储过程
[root@Master ~]# mysql hellodb < testlog.sql
2.调用存储过程
MariaDB [(none)]> USE hellodb Reading table information for completion of table and column names You can turn off this feature to get a quicker startup with -A Database changed MariaDB [hellodb]> call pro_testlog;
3.另起一个主节点窗口将主节点断网
[root@Master ~]# ifdown ens33
4.manager端完成切换退出,查看日志,查看新的主节点是哪台slave
[root@Manager app1]# tail /data/mastermha/app1/manager.log Started automated(non-interactive) failover. The latest slave 192.168.73.112(192.168.73.112:3306) has all relay logs for recovery. Selected 192.168.73.112(192.168.73.112:3306) as a new master. 192.168.73.112(192.168.73.112:3306): OK: Applying all logs succeeded. 192.168.73.113(192.168.73.113:3306): This host has the latest relay log events. Generating relay diff files from the latest slave succeeded. 192.168.73.113(192.168.73.113:3306): OK: Applying all logs succeeded. Slave started, replicating from 192.168.73.112(192.168.73.112:3306) 192.168.73.112(192.168.73.112:3306): Resetting slave info succeeded. Master failover to 192.168.73.112(192.168.73.112:3306) completed successfully. #此处显示最新的主节点为192.168.73.112
由于从节点在配置文件中定义的为read-only,此时被提升为主能执行写操作时应为管理服务器上有管理账号,他将从节点的服务器全局变量read_only给关闭了
[root@Slave1 ~]# mysql -e "SELECT @@read_only;" +-------------+ | @@read_only | +-------------+ | 0 | +-------------+
为了防止服务服务重启再次变为read-only,此时需要对新主节点的配置文件进行修改将read-only行注释
[mysqld] server-id=2 #read-only log-bin relay_log_purge=0 skip_name_resolve
1.对hellodb.teachers表插入数据
[root@Slave1 ~]# mysql -e "INSERT hellodb.teachers VALUES(5,'Tang San',30,'M');"
2.Slave2主机上查看是否同步
[root@Slave2 ~]# mysql -e "SELECT * FROM hellodb.teachers;" +-----+---------------+-----+--------+ | TID | Name | Age | Gender | +-----+---------------+-----+--------+ | 1 | Song Jiang | 45 | M | | 2 | Zhang Sanfeng | 94 | M | | 3 | Miejue Shitai | 77 | F | | 4 | Lin Chaoying | 93 | F | | 5 | Tang San | 30 | M | #已经同步 +-----+---------------+-----+--------+
当原主节点被修复后,将其添加为从节点使用。