已经运行很久的mysql，搭建主从

1、（主库）修改/etc/my.cnf配置文件
innodb_strict_mode=on
#innodb 共享表空间设置
innodb_file_per_table = 1
innodb_data_home_dir = /var/lib/mysql
innodb_data_file_path = ibdata1:6500M:autoextend
#innodb 日志相关
innodb_log_file_size = 500M
innodb_log_buffer_size = 8M
innodb_log_files_in_group = 4
innodb_log_group_home_dir=/var/lib/mysql
innodb_flush_log_at_trx_commit=1
#innodb 内存相关参数
innodb_buffer_pool_size=2048M
innodb_additional_mem_pool_size = 32M
innodb_thread_concurrency =0
innodb_flush_method = O_DIRECT
expire_logs_days    = 5
log_bin             = /var/lib/mysql/mysql-bin
log_bin_index       = /var/lib/mysql/mysql-bin.index




2、（从库）修改/etc/my.cnf配置文件
port          = 3306
socket        = /var/lib/mysql/mysql.sock
skip-external-locking
key_buffer_size = 16M
max_allowed_packet = 1M
table_open_cache = 64
sort_buffer_size = 512K
net_buffer_length = 8K
read_buffer_size = 256K
read_rnd_buffer_size = 512K
myisam_sort_buffer_size = 8M
binlog_format=mixed
server-id=50          //从库注意一定要修改此参数
lower_case_table_names=1
log-slow-queries = /tmp/mysql-slow.log
long_query_time = 2
max_connections=500


binlog-ignore-db=mysql #被忽略的数据库
#innodb 共享表空间设置
innodb_file_per_table = 1
innodb_data_home_dir = /var/lib/mysql
innodb_data_file_path = ibdata1:6500M:autoextend
#innodb 日志相关
innodb_log_file_size = 500M
innodb_log_buffer_size = 8M
innodb_log_files_in_group = 4
innodb_log_group_home_dir=/var/lib/mysql
innodb_flush_log_at_trx_commit=1
#innodb 内存相关参数
innodb_buffer_pool_size=2048M
innodb_additional_mem_pool_size = 32M
innodb_thread_concurrency =0
innodb_flush_method = O_DIRECT
log_slave_updates=1
expire_logs_days    = 5
log_bin             = /var/lib/mysql/mysql-bin
log_bin_index       = /var/lib/mysql/mysql-bin.index
relay_log           = /var/lib/mysql/replay-bin
relay_log_index     = /var/lib/mysql/replay-bin.index


3、修改主库innodb日志大小
[root@dsdb-1 mysql]# ls ib_logfile*
ib_logfile0  ib_logfile1
server mysql stop
mv ib_logfile*  /root
server mysql start


4、（主库）拷贝主库文件到备库

由于主库是已经运行很久的一个库，数据量在100G以上，在综合考虑mysqldump 备份恢复需要很久的时间，采用直接拷贝文件到备库的方式建立slave

由于主库包含InnoDB表的拷贝：
4.1在master主机上使用mysql命令行客户端，并阻塞所有的写入操作
mysql> FLUSH TABLES WITH READ LOCK;


4.2记录偏移量值


mysql> show master status;
+------------------+----------+--------------+------------------+
| File             | Position | Binlog_Do_DB | Binlog_Ignore_DB |
+------------------+----------+--------------+------------------+
| mysql-bin.000056 |      107 |              |                  |
+------------------+----------+--------------+------------------
4.3、在另外一个shell客户端关闭master数据库
mysqladmin shudown -p


 
4.4、拷贝并数据文件
mysqldump --all-databases --master-data > /tmp/datadump.db  
或者如果数据量比较大的话，可以用直接拷贝目录的方式，也省去了执行insert语句更新索引的开销。
tar czvf /tmp/databack.tar.gz /var/lib/mysql/


4.5 重启mysql master服务
 Service mysql start


4.6在master上释放读锁（如果前面执行过FLUSH TABLES WITH READ LOCK;）
mysql> UNLOCK TABLES;


4.7 主库传输文件到备库
Scp  /tmp/databack.tar.gz  10.10.12.50：/tmp


在master上设置复制用户及密码
GRANT REPLICATION SLAVE ON *.* TO 'repl'@'10.10.12.%' IDENTIFIED BY '123456';



如果主库不包含InnoDB表的拷贝，拷贝方式可以考虑，不用停机：
    1 在master主机上使用mysql命令行客户端，将所有数据刷进磁盘，并阻塞所有的写入操作


mysql> FLUSH TABLES WITH READ LOCK;  
    2 拷贝并压缩数据文件


# tar czvf /tmp/databack.tar.gz /usr/local/mysql/data  
    3 释放读锁


mysql> UNLOCK TABLES;  


5、在从机上设置复制 
5.1 设置主库，从库的配置文件my.cnf 
(1) 需要设置解压出来文件的权限确保slave能够读取或更改它们


Mv mysql mysql_bak
Mv /tmp/databack.tar.gz  /var/lib/
tar xzvf databack.tar.gz  
mv   /var/lib/var/lib/mysql  /var/lib/


(2) 删除data目录下的auto.cnf文件及err日志
rm -rf  dsdb-1.pid
rm -rf  dsdb-1-relay-bin.000021
rm -rf  relay-log.info
rm -rf  dsdb-1-relay-bin.000022
rm -rf  dsdb-1-relay-bin.000023
rm -rf  dsdb-1-relay-bin.000024
rm -rf  dsdb-1-relay-bin.000025
rm -rf  dsdb-1-relay-bin.000026
rm -rf  dsdb-1-relay-bin.000027
rm -rf  dsdb-1-relay-bin.000028
rm -rf  dsdb-1-relay-bin.000029
rm -rf  dsdb-1-relay-bin.000030
rm -rf  dsdb-1-relay-bin.000031
rm -rf  dsdb-1-relay-bin.000032
rm -rf  dsdb-1-relay-bin.000033
rm -rf  dsdb-1-relay-bin.index
rm -rf  dsdb-1-relay-bin.000034
rm -rf  mysql-bin.000052
rm -rf  mysql-bin.index
rm -rf  master.info
rm -rf  dsdb-1.err
rm -rf  mysql-bin.000053


5.2 启动mysql服务
Service mysql start


如果在master上加读锁（如果前面执行过FLUSH TABLES WITH READ LOCK;）从库上执行
mysql> UNLOCK TABLES;
mysql> stop SLAVE; 


5.3 从库上执行change master操作


mysql> CHANGE MASTER TO master_host='10.10.12.49', master_port=3306, master_user='repl',master_password='123456', master_log_file='mysql-bin.000056', master_log_pos=107;
ERROR 1201 (HY000): Could not initialize master info structure; more error messages can be found in the MySQL error log
mysql> stop slave;
Query OK, 0 rows affected, 1 warning (0.00 sec)


mysql> reset slave;
Query OK, 0 rows affected (0.00 sec)


mysql> CHANGE MASTER TO master_host='10.10.12.49', master_port=3306, master_user='repl',master_password='123456', master_log_file='mysql-bin.000056', master_log_pos=107;
Query OK, 0 rows affected (0.02 sec)


8.4 从库启动复制进程
mysql> START SLAVE; 
9、（从库）检查同步情况
show slave status \G;


验证：
主库：
mysql> create database t;
Query OK, 1 row affected (0.03 sec)

mysql> show databases;
+--------------------+
| Database           |
+--------------------+
| information_schema |
| mysql              |
| performance_schema |
| t                  |
| test               |
+--------------------+
5 rows in set (0.00 sec)


mysql> use t;
Database changed

mysql> create table t(id int,a varchar(10)) ENGINE=InnoDB ROW_FORMAT=COMPRESSED 
    -> ;
Query OK, 0 rows affected (0.24 sec)

mysql> insert into t values(1,'a');
Query OK, 1 row affected (0.05 sec)

mysql> insert into t values(2,'a');
Query OK, 1 row affected (0.07 sec)

mysql> insert into t values(3,'a');
Query OK, 1 row affected (0.05 sec)


从库 查看：


mysql> select * From t;
+------+------+
| id   | a    |
+------+------+
|    1 | a    |
|    2 | a    |
|    3 | a    |
+------+------+
3 rows in set (0.00 sec)




至此主从master-slave已经建立起来了