MySQL系列详解八：MySQL多线程复制演示-技术流ken

前言

Mysql 采用多线程进行复制是从 Mysql 5.6 开始支持的内容，但是 5.6 版本下有缺陷，虽然支持多线程，但是每个数据库只能一个线程，也就是说如果我们只有一个数据库，则主从复制时也只有一个线程在工作。相当于还是以前的单线程。 从 Mysql 5.7 开始支持同一数据库下并行主从复制。不过默认情况下，还是单数据库单个线程，如果需要使用多线程，需要在从节点进行配置。

Mysql 5.7 对主从复制增加了一种类型，共有两种类型，如下：

• DATABASE 基于库的并行复制 ， 每个数据库对应一个复制线程
• LOGICAL_CLOCK 基于组提交的并行复制方式，同一个数据库下可以有多个线程

下面的步骤，在从节点上进行配置。其实只需要修改两处变量即可。

首先需要搭建出来一个主从复制的架构出来，这里就不再演示，请参考上一篇博客《MySQL系列详解六：MySQL主从复制/慢同步演示-技术流ken》

演示：实现MySQL多线程复制

1.在从节点查看当前的主从复制的进程数

MySQL [(none)]> show processlist;
+----+-------------+--------------------+------+-------------+------+---------------------------------------------------------------+------------------+
| Id | User        | Host               | db   | Command     | Time | State                                                         | Info             |
+----+-------------+--------------------+------+-------------+------+---------------------------------------------------------------+------------------+
|   | ken         | 10.220.5.137: | NULL | Binlog Dump |  | Master has sent all binlog to slave; waiting for more updates | NULL             |
|   | system user |                    | NULL | Connect     |  | Waiting for master to send event                              | NULL             |
|   | system user |                    | NULL | Connect     |  | Slave has read all relay log; waiting for more updates        | NULL             |
|  | root        | localhost          | NULL | Query       |     | starting                                                      | show processlist |
+----+-------------+--------------------+------+-------------+------+---------------------------------------------------------------+------------------+

从上面看出只有一个主进程在等待同步。

2.下面查看复制类型和并行数量配置

MySQL [(none)]> show global variables like '%slave_para%';
+------------------------+----------+
| Variable_name          | Value    |
+------------------------+----------+
| slave_parallel_type    | DATABASE |
| slave_parallel_workers |         |
+------------------------+----------+
 rows in set (0.02 sec)

当前的复制类型是 DATABASE，也就是同一数据库下只有一个线程进行复制，不能并行复制。

3.停止从节点的复制

MySQL [(none)]> stop slave;
Query OK,  rows affected (0.00 sec)

4.修改复制类型为LOGICAL_CLOCK 

MySQL [(none)]> set global slave_parallel_type=LOGICAL_CLOCK;
Query OK,  rows affected (0.00 sec)

MySQL [(none)]> show global variables like '%slave_para%';
+------------------------+---------------+
| Variable_name          | Value         |
+------------------------+---------------+
| slave_parallel_type    | LOGICAL_CLOCK |
| slave_parallel_workers |              |
+------------------------+---------------+
 rows in set (0.00 sec)

5.修改并行数量为5

MySQL [(none)]> set global slave_parallel_workers=;
Query OK,  rows affected (0.00 sec)

MySQL [(none)]> show global variables like '%slave_para%';
+------------------------+---------------+
| Variable_name          | Value         |
+------------------------+---------------+
| slave_parallel_type    | LOGICAL_CLOCK |
| slave_parallel_workers |              |
+------------------------+---------------+
 rows in set (0.00 sec)

6.启动从节点

MySQL [(none)]> start slave;
Query OK,  rows affected (0.04 sec)

7.再次查看复制类型和并行数量配置

MySQL [(none)]> show processlist;
+----+-------------+--------------------+------+-------------+------+---------------------------------------------------------------+------------------+
| Id | User        | Host               | db   | Command     | Time | State                                                         | Info             |
+----+-------------+--------------------+------+-------------+------+---------------------------------------------------------------+------------------+
|   | ken         | 10.220.5.137: | NULL | Binlog Dump |  | Master has sent all binlog to slave; waiting for more updates | NULL             |
|  | root        | localhost          | NULL | Query       |     | starting                                                      | show processlist |
|  | system user |                    | NULL | Connect     |    | Waiting for master to send event                              | NULL             |
|  | system user |                    | NULL | Connect     |    | Slave has read all relay log; waiting for more updates        | NULL             |
|  | system user |                    | NULL | Connect     |    | Waiting for an event from Coordinator                         | NULL             |
|  | system user |                    | NULL | Connect     |    | Waiting for an event from Coordinator                         | NULL             |
|  | system user |                    | NULL | Connect     |    | Waiting for an event from Coordinator                         | NULL             |
|  | system user |                    | NULL | Connect     |    | Waiting for an event from Coordinator                         | NULL             |
|  | system user |                    | NULL | Connect     |    | Waiting for an event from Coordinator                         | NULL             |
+----+-------------+--------------------+------+-------------+------+---------------------------------------------------------------+------------------+
 rows in set (0.00 sec)

总结

最后说一下为什么需要多线程复制？因为主从之间的同步会有延时，多线程的目的是为了尽量减少这个延时时间。虽然如何优化主从是一个系统的功能，不同的场景需要不同的解决方案，但是多线程至少从基础上能减少延迟时间。另外根据数据库的实际情况，能否真正减少延时，以及配置多少线程，则需要反复的测试得出适合自己的数据。