现在生产环境MySQL数据库是一主一从,由于业务量访问不断增大,故再增加一台从库。前提是不能影响线上业务使用,也就是说不能重启MySQL服务,为了避免出现其他情况,选择在网站访问量低峰期时间段操作。
一般在线增加从库有两种方式,一种是通过mysqldump备份主库,恢复到从库,mysqldump是逻辑备份,数据量大时,备份速度会很慢,锁表的时间也会很长。另一种是通过xtrabackup工具备份主库,恢复到从库,xtrabackup是物理备份,备份速度快,不锁表。为什么不锁表?因为自身会监控主库日志,如果有更新的数据,就会先写到一个文件中,然后再回归到备份文件中,从而保持数据一致性。
服务器信息:
MySQL主从是基于binlog日志,所以在安装好数据库后就要开启binlog。这样好处是,一方面可以用binlog恢复数据库,另一方面可以为主从做准备。
原有主库配置参数如下:
# vi my.cnf
server-id = 1 #id要唯一
log-bin = mysql-bin #开启binlog日志
auto-increment-increment = 1 #在Ubuntu系统中MySQL5.5以后已经默认是1
auto-increment-offset = 1
slave-skip-errors = all #跳过主从复制出现的错误
mysql> grant all on *.* to 'sync'@'192.168.18.%' identified by 'sync';
# vi my.cnf
server-id = 3 #这个设置3
log-bin = mysql-bin #开启binlog日志
auto-increment-increment = 1 #这两个参数在Ubuntu系统中MySQL5.5以后都已经默认是1
auto-increment-offset = 1
slave-skip-errors = all #跳过主从复制出现的错误
# mysqldump -uroot -p123 --routines --single_transaction --master-data=2 --databases weibo > weibo.sql
参数说明:
--routines:导出存储过程和函数
--single_transaction:导出开始时设置事务隔离状态,并使用一致性快照开始事务,然后unlock tables;而lock-tables是锁住一张表不能写操作,直到dump完毕。
--master-data:默认等于1,将dump起始(change master to)binlog点和pos值写到结果中,等于2是将change master to写到结果中并注释。
# scp weibo.sql root@192.168.18.214:/home/root
mysql> create table test_tb(id int,name varchar(30));
# mysql -uroot -p123 -e 'create database weibo;'
# mysql -uroot -p123 weibo < weibo.sql
# head -25 weibo.sql
-- CHANGE MASTER TO MASTER_LOG_FILE='mysql-bin.000001', MASTER_LOG_POS=107; #大概22行
mysql> change master to master_host='192.168.18.212',
-> master_user='sync',
-> master_password='sync',
-> master_log_file='mysql-bin.000001',
-> master_log_pos=107;
mysql> start slave;
mysql> show slave status\G;
ERROR 2006 (HY000): MySQL server has gone away
No connection. Trying to reconnect...
Connection id: 90
Current database: *** NONE ***
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.18.212
Master_User: sync
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mysql-bin.000001
Read_Master_Log_Pos: 358
Relay_Log_File: mysqld-relay-bin.000003
Relay_Log_Pos: 504
Relay_Master_Log_File: mysql-bin.000001
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
......
可以看到IO和SQL线程均为YES,说明主从配置成功。
mysql> show tables;
+---------------------------+
| Tables_in_weibo |
+---------------------------+
| test_tb |
发现刚才模拟创建的test_tb表已经同步过来!
在上面配置基础上做实验,先删除掉从库配置:
mysql> stop slave; #停止同步
mysql> reset slave; #清除从连接信息
mysql> show slave status\G; #再查看从状态,可以看到IO和SQL线程都为NO
mysql> drop database weibo; #删除weibo库
此时,从库现在和新装的一样,继续前进!
# innobackupex --user=root --password=123 ./
生成一个以时间为命名的备份目录:2015-07-01_16-49-43
# ll 2015-07-01_16-49-43/
total 18480
drwxr-xr-x 5 root root 4096 Jul 1 16:49 ./
drwx------ 4 root root 4096 Jul 1 16:49 ../
-rw-r--r-- 1 root root 188 Jul 1 16:49 backup-my.cnf
-rw-r----- 1 root root 18874368 Jul 1 16:49 ibdata1
drwxr-xr-x 2 root root 4096 Jul 1 16:49 mysql/
drwxr-xr-x 2 root root 4096 Jul 1 16:49 performance_schema/
drwxr-xr-x 2 root root 12288 Jul 1 16:49 weibo/
-rw-r--r-- 1 root root 21 Jul 1 16:49 xtrabackup_binlog_info
-rw-r----- 1 root root 89 Jul 1 16:49 xtrabackup_checkpoints
-rw-r--r-- 1 root root 563 Jul 1 16:49 xtrabackup_info
-rw-r----- 1 root root 2560 Jul 1 16:49 xtrabackup_logfile
# scp -r 2015-07-01_16-49-43 root@192.168.18.214:/home/root
# sudo rm -rf /var/lib/mysql/
# sudo mv 2015-07-01_16-49-43/ /var/lib/mysql
# sudo chown mysql.mysql -R /var/lib/mysql
# sudo /etc/init.d/mysql start
# ps -ef |grep mysql #查看已经正常启动
mysql 8832 1 0 16:55 ? 00:00:00 /usr/sbin/mysqld
mysql> create table test_tb2(id int,name varchar(30));
# cat /var/lib/mysql/xtrabackup_info
uuid = 201af9db-1fce-11e5-96b0-525400e4239d
name =
tool_name = innobackupex
tool_command = --user=root --password=... ./
tool_version = 1.5.1-xtrabackup
ibbackup_version = xtrabackup version 2.2.11 based on MySQL server 5.6.24 Linux (x86_64) (revision id: )
server_version = 5.5.43-0ubuntu0.12.04.1-log
start_time = 2015-07-01 16:49:43
end_time = 2015-07-01 16:49:46
lock_time = 1
binlog_pos = filename 'mysql-bin.000001', position 429 #这个位置
innodb_from_lsn = 0
innodb_to_lsn = 1598188
partial = N
incremental = N
format = file
compact = N
compressed = N
mysql> change master to master_host='192.168.18.212',
-> master_user='sync',
-> master_password='sync',
-> master_log_file='mysql-bin.000001',
-> master_log_pos=429;
mysql> start slave;
mysql> show slave status\G;
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.18.212
Master_User: sync
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mysql-bin.000001
Read_Master_Log_Pos: 539
Relay_Log_File: mysqld-relay-bin.000002
Relay_Log_Pos: 363
Relay_Master_Log_File: mysql-bin.000001
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
......
可以看到IO和SQL线程均为YES,说明主从配置成功。
mysql> show tables;
+---------------------------+
| Tables_in_weibo |
+---------------------------+
| test_tb |
| test_tb2 |
发现刚才模拟创建的test_tb2表已经同步过来。