有一种提高Replication性能的方法，就是增加mysql Replication结构的深度，就是一个master 只Replication给一个slave，在由这个slave 复制给其他的slave。
结构如下：
这张图中 master2 就是master1 的slave，而slave1，slave2，slave3 是master2 的slave
以下是大致的配置过程：
环境如下：
Maste1 ：
Hostname：media
Ip：172.72.15.41
Master2：
Hostname:rac1
Ip：172.72.15.151
Slave1：
Hostname:rac2
Ip：172.72.15.152
Mysql的版本都是5.6.10
先配置master1端的/etc/my.cnf文件，加入下述配置，然后重启master1的MySQL服务：
binlog-format=ROW
log-bin=master-bin.log
log-bin-index=master-bin.index
log-slave-updates=true
gtid-mode=on
enforce-gtid-consistency=true
master-info-repository=TABLE
relay-log-info-repository=TABLE
sync-master-info=1
slave-parallel-workers=2
binlog-checksum=CRC32
master-verify-checksum=1
slave-sql-verify-checksum=1
binlog-rows-query-log-events=1
server-id=1
sync_binlog=1
report-port=3306
port=3306
report-host=hostname
innodb_flush_log_at_trx_commit=1
再修改master2、slave1、slave2、slave3端的my.cnf文件，加入下述配置，然后重启slave的MySQL服务：
binlog-format=ROW
log-bin=slave-bin.log
log-bin-index=slave-bin.index
log-slave-updates=true
gtid-mode=on
enforce-gtid-consistency=true
master-info-repository=TABLE
relay-log-info-repository=TABLE
sync-master-info=1
slave-parallel-workers=2
binlog-checksum=CRC32
master-verify-checksum=1
slave-sql-verify-checksum=1
binlog-rows-query-log-events=1
server-id=2
sync_binlog=1
report-port=3306
port=3306
report-host=hostname
innodb_flush_log_at_trx_commit=1
skip_parallel_works=4
注意：在master2端启用binlog，就是为了salve1 slave2 slave3 同步master2数据，所以采用和master1主服务器类似的配置。
环境中只复制new1这个数据库
一：导出数据：
在导出数据之前把数据库的表都锁上，并置于read only状态
|
1
2
3
4
5
6
7
|
mysql> flush tables with
read
lock;
Query OK, 0 rows affected (0.00 sec)
mysql>
set
global read_only=on;
Query OK, 0 rows affected (0.00 sec)
mysql>
|
再开始maste1上面把数据导出来，由于5.6版本增加了gtid的功能，所以在用mysqldump导出的时候，可以通过 --set-gtid-purged 这个选项控制是否要用gtids来恢复，默认是开启的！
|
1
2
3
4
5
6
7
8
|
[root@media ~]
# mysqldump -u root -psbcenter -B new1 > /tmp/new1.sql
Warning: Using a password on the
command
line interface can be insecure.
Warning: A partial dump from a server that has GTIDs will by default include the GTIDs of all transactions, even those that changed suppressed parts of the database. If you don't want to restore GTIDs, pass --
set
-gtid-purged=OFF. To
make
a complete dump, pass --all-databases --triggers --routines --events.
[root@media ~]
#
[root@media ~]
# scp /tmp/new1.sql 172.72.15.151:/root/
root@172.72.15.151's password:
new1.sql 100% 8346 8.2KB
/s
00:00
[root@media ~]
#
|
二：到maste2,slave1上导入数据
如果mysqldump 导出数据的时候 --set-gtid-purged 没有设置成off，那么master2启动mysqld 的时候需要启用gtid_mode=on
Master2上导入数据
[root@rac1 ~]# mysql < new1.sql
Slave1 上导入数据
[root@rac2 ~]# mysql < /root/new1.sql
三：创建复制用户
在master 1 上创建master2的复制用户
mysql> GRANT REPLICATION SLAVE ON *.* TO 'rep1'@'172.72.15.151' IDENTIFIED BY '123456';
mysql> flush privileges ;
Query OK, 0 rows affected (0.00 sec)
在master2上创建slave1 的复制用户
mysql> GRANT REPLICATION SLAVE ON *.* TO 'rep1'@'172.72.15.152' IDENTIFIED BY '123456';
mysql> flush privileges;
Query OK, 0 rows affected (0.00 sec)
四：在master2和slave1 上配置master属性
在master2上配置master属性
mysql> change master to
-> master_host='172.72.15.41',
-> master_user='rep1',
-> master_password='123456',
-> master_auto_position=1;
Query OK, 0 rows affected, 2 warnings (0.17 sec)
mysql>start slave；
在slave1 上配置master属性
mysql> change master to
-> master_host='172.72.15.152',
-> master_user='rep1',
-> master_password='123456',
-> master_auto_position=1;
Query OK, 0 rows affected, 2 warnings (0.17 sec)
mysql>start slave；
－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－
－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－
五：验证！
在master1 的new1上创建一张表
mysql> use new1
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
mysql>
mysql>
mysql> show tables;
+----------------+
| Tables_in_new1 |
+----------------+
| a |
+----------------+
1 row in set (0.00 sec)
mysql> create table b like a;
Query OK, 0 rows affected (0.24 sec)
mysql> insert into b select * from a;
Query OK, 1031 rows affected (0.13 sec)
Records: 1031 Duplicates: 0 Warnings: 0
mysql>
到master2查看是否已经复制成功
mysql> use new1;
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
mysql> show tables;
+----------------+
| Tables_in_new1 |
+----------------+
| a |
| b |
+----------------+
2 rows in set (0.00 sec)
mysql> select count(*) from b;
+----------+
| count(*) |
+----------+
| 1031 |
+----------+
1 row in set (0.00 sec)
mysql>
再到slave1 上查看
mysql> use new1
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
mysql> show tables;
+----------------+
| Tables_in_new1 |
+----------------+
| a |
| b |
+----------------+
2 rows in set (0.00 sec)
mysql> select count(*) from bl
-> ;
ERROR 1146 (42S02): Table 'new1.bl' doesn't exist
mysql> select count(*) from b;
+----------+
| count(*) |
+----------+
| 1031 |
+----------+
1 row in set (0.00 sec)
mysql>
整个过程就是这样，另外要提及的一点是如果mysql 版本是5.6以下的或者是没开启gtids
的，需要在master2上配置log-slave-updates 参数，这个参数的作用是master2接受到master1的binlog到relay日志中去时，也会更新到自己的binlog中去，以便给它的slave使用！