rpl_semi_sync_master_wait_no_slave 参数研究实验

最近在研究MySQL，刚学到半同步。

半同步的配置中，关于这两个参数：

　　rpl_semi_sync_master_wait_no_slave

　　rpl_semi_sync_master_wait_for_slave_count

发现很不好搞懂，请教了一些老师，也做了一些资料搜索，每个人给我的答案都不同：

　　

中文网络上一些典型的说法，这些都是错误的！！！：
rpl_semi_sync_master_wait_for_slave_count：
　　控制slave应答的数量，默认是1，表示master接收到几个slave应答后才commit。
　　
rpl_semi_sync_master_wait_no_slave ：
　　1.需要等待几个slave节点的ACK，否则一直waiting。
　　2.当一个事务被提交，但是Master没有Slave连接，这时M不可能收到任何确认信息，但M会在时间限制范围内继续等待。如果没有Slave链接，会切换到异步复制。是否允许master每个事务提交后都要等待slave的接收确认信号。默认为on，每一个事务都会等待。如果为off，则slave追赶上后，也不会开启半同步复制模式，需要手工开启。

好奇心驱使我通过实验来验证大家的说法，可惜的是——都是错的。通过官方文档，看的云里雾里的，只好自己探索。

 

探索内容

　　半同步参数

• rpl_semi_sync_master_wait_no_slave（重点）
• rpl_semi_sync_master_wait_for_slave_count

环境信息

role	ip	port	hostname	etc
master	192.168.188.101	4306	mysqlvm1	提示符为mysql>
slave	192.168.188.201	4306	mysqlvm1-1	提示符为mysql4306>
		5306		提示符为mysql5306>
		6306		提示符为mysql6306>
		7306		提示符为mysql7306>

 

MySQL版本

　　5.7.26

 

前置条件

　　已配置好主从复制。

 

配置增强半同步

　　1.加载lib，所有主从节点都要配置。

　　　　主库&从库：

install plugin rpl_semi_sync_master soname 'semisync_master.so';

install plugin rpl_semi_sync_slave soname 'semisync_slave.so';   

      

　　2.查看，确保所有节点都成功加载。

　　　　mysql> show plugins;

　　　　　　| rpl_semi_sync_master       | ACTIVE   | REPLICATION        | semisync_master.so | GPL     |

　　　　　　| rpl_semi_sync_slave        | ACTIVE   | REPLICATION        | semisync_slave.so  | GPL     |

 

　　3.启用半同步

　　　　1.先启用从库上的参数，最后启用主库的参数。

　　　　从库：

set global rpl_semi_sync_slave_enabled = 1;   # 1：启用，0：禁止

　　　　主库：

　　　　　　　　

set global rpl_semi_sync_master_enabled = 1;   # 1：启用，0：禁止

set global rpl_semi_sync_master_timeout = 60000;       # 60秒，时间长些便于实验

　　　　2.从库重启io_thread

　　　　stop slave io_thread;

　　　　start slave io_thread;

　　　　查看主库参数

　　　　　　

mysql> show global variables like "%sync%"; show global status like "%sync%";

+-------------------------------------------+------------+
| Variable_name                             | Value      |
+-------------------------------------------+------------+
| binlog_group_commit_sync_delay            | 100        |
| binlog_group_commit_sync_no_delay_count   | 10         |
| innodb_flush_sync                         | ON         |
| innodb_sync_array_size                    | 1          |
| innodb_sync_spin_loops                    | 30         |
| rpl_semi_sync_master_enabled              | ON         |
| rpl_semi_sync_master_timeout              | 60000      |
| rpl_semi_sync_master_trace_level          | 32         |
| rpl_semi_sync_master_wait_for_slave_count | 3          |
| rpl_semi_sync_master_wait_no_slave        | ON         |
| rpl_semi_sync_master_wait_point           | AFTER_SYNC |
| rpl_semi_sync_slave_enabled               | ON         |
| rpl_semi_sync_slave_trace_level           | 32         |
| sync_binlog                               | 1          |
| sync_frm                                  | ON         |
| sync_master_info                          | 10000      |
| sync_relay_log                            | 10000      |
| sync_relay_log_info                       | 10000      |
+-------------------------------------------+------------+
18 rows in set (0.00 sec)

 

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Innodb_data_fsyncs                         | 664      |
| Innodb_data_pending_fsyncs                 | 0        |
| Innodb_os_log_fsyncs                       | 413      |
| Innodb_os_log_pending_fsyncs               | 0        |
| Rpl_semi_sync_master_clients               | 4        |
| Rpl_semi_sync_master_net_avg_wait_time     | 0        |
| Rpl_semi_sync_master_net_wait_time         | 0        |
| Rpl_semi_sync_master_net_waits             | 199      |
| Rpl_semi_sync_master_no_times              | 21       |
| Rpl_semi_sync_master_no_tx                 | 48       |
| Rpl_semi_sync_master_status                | ON       |
| Rpl_semi_sync_master_timefunc_failures     | 0        |
| Rpl_semi_sync_master_tx_avg_wait_time      | 3280008  |
| Rpl_semi_sync_master_tx_wait_time          | 72160195 |
| Rpl_semi_sync_master_tx_waits              | 22       |
| Rpl_semi_sync_master_wait_pos_backtraverse | 0        |
| Rpl_semi_sync_master_wait_sessions         | 0        |
| Rpl_semi_sync_master_yes_tx                | 20       |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+

19 rows in set (0.00 sec)

 

============================我是分割线=====================================

 

 

实验1：

场景：

　　rpl_semi_sync_master_wait_no_slave=ON

　　rpl_semi_sync_master_wait_for_slave_count=3

　　slave存活1，其他slave停止，会发生什么？

步骤：

　　1.只保留一个slave，停止其他3个slave

mysql5306> stop slave;
mysql6306> stop slave;
mysql7306> stop slave;

　　2.立即查询master

　　　　为减少信息干扰，只截取了需要关注的数据

mysql> show global variables like "%sync%"; show global status like "%sync%";

+-------------------------------------------+------------+
| Variable_name                             | Value      |
+-------------------------------------------+------------+
| rpl_semi_sync_master_enabled              | ON         |
| rpl_semi_sync_master_timeout              | 60000      |
| rpl_semi_sync_master_wait_for_slave_count | 3          |
| rpl_semi_sync_master_wait_no_slave        | ON         |
| rpl_semi_sync_master_wait_point           | AFTER_SYNC |
| rpl_semi_sync_slave_enabled               | ON         |
+-------------------------------------------+------------+
18 rows in set (0.00 sec)

 

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 4        |
| Rpl_semi_sync_master_status                | ON       |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

　　3.等待一分钟，再查询master

mysql> show global variables like "%sync%"; show global status like "%sync%";

+-------------------------------------------+------------+
| Variable_name                             | Value      |
+-------------------------------------------+------------+
| rpl_semi_sync_master_enabled              | ON         |
| rpl_semi_sync_master_timeout              | 60000      |
| rpl_semi_sync_master_wait_for_slave_count | 3          |
| rpl_semi_sync_master_wait_no_slave        | ON         |
| rpl_semi_sync_master_wait_point           | AFTER_SYNC |
| rpl_semi_sync_slave_enabled               | ON         |
+-------------------------------------------+------------+
18 rows in set (0.00 sec)

 

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 1        |
| Rpl_semi_sync_master_status                | ON       |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

　　4.另外开启一个master会话，开始一个事务并提交

mysql> insert into new values(4);
[挂起……]

　　5.查看master

　　　　发现无变化

mysql> show global variables like "%sync%"; show global status like "%sync%";

+-------------------------------------------+------------+
| Variable_name                             | Value      |
+-------------------------------------------+------------+
| rpl_semi_sync_master_enabled              | ON         |
| rpl_semi_sync_master_timeout              | 60000      |
| rpl_semi_sync_master_wait_for_slave_count | 3          |
| rpl_semi_sync_master_wait_no_slave        | ON         |
| rpl_semi_sync_master_wait_point           | AFTER_SYNC |
| rpl_semi_sync_slave_enabled               | ON         |
+-------------------------------------------+------------+
18 rows in set (0.00 sec)

 

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 1        |
| Rpl_semi_sync_master_status                | ON       |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

　　6.查看存活的slave

　　　　发现事务已经被应用

mysql> select * from new;

+------+
| name |
+------+
|    2 |
+------+
1 rows in set (0.00 sec)

 

　　7.等待master事务超时后完成。

　　 　　这里master提交到收到结果，用了1分钟0.01秒,该事务因slave的ack应答不够，发生了等待。

mysql> insert into new values(4);
Query OK, 1 row affected (1 min 0.01 sec)

 

　　8.查看master

　　　　可以发现，事务超时后，master已经转为异步复制（Rpl_semi_sync_master_status=OFF）

 

mysql> show global variables like "%sync%"; show global status like "%sync%";

+-------------------------------------------+------------+
| Variable_name                             | Value      |
+-------------------------------------------+------------+
| rpl_semi_sync_master_enabled              | ON         |
| rpl_semi_sync_master_timeout              | 60000      |
| rpl_semi_sync_master_wait_for_slave_count | 3          |
| rpl_semi_sync_master_wait_no_slave        | ON         |
| rpl_semi_sync_master_wait_point           | AFTER_SYNC |
| rpl_semi_sync_slave_enabled               | ON         |
+-------------------------------------------+------------+
18 rows in set (0.00 sec)

 

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 1        |
| Rpl_semi_sync_master_status                | OFF      |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

　　9.依次启动已经停止的slave，每启动一个，便立即查看一次master状态

　　　　可以看到，启动slave的动作会立即被master接收到，并且master会自动切换回半同步模式（Rpl_semi_sync_master_status=ON）

 

mysql5306> start slave;

mysql> show global status like "%sync%";

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 2        |
| Rpl_semi_sync_master_status                | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

mysql6306> start slave;

mysql> show global status like "%sync%";

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 3        |
| Rpl_semi_sync_master_status                | ON       |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

mysql7306> start slave;

mysql> show global status like "%sync%";

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 4        |
| Rpl_semi_sync_master_status                | ON       |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

============================我是分割线=====================================

 

实验2：

场景：

　　rpl_semi_sync_master_wait_no_slave=OFF

　　rpl_semi_sync_master_wait_for_slave_count=3

　　slave存活1，其他slave停止，会发生什么？

步骤：

　　1.继续实验1的环境，首先设置参数并查看master状态

　　　　为减少信息干扰，只截取了需要关注的数据 

mysql> set global rpl_semi_sync_master_wait_no_slave=OFF;
Query OK, 0 rows affected (0.00 sec)

mysql> show global variables like "%sync%"; show global status like "%sync%";

+-------------------------------------------+------------+
| Variable_name                             | Value      |
+-------------------------------------------+------------+
| rpl_semi_sync_master_enabled              | ON         |
| rpl_semi_sync_master_timeout              | 60000      |
| rpl_semi_sync_master_wait_for_slave_count | 3          |
| rpl_semi_sync_master_wait_no_slave        | OFF        |
| rpl_semi_sync_master_wait_point           | AFTER_SYNC |
| rpl_semi_sync_slave_enabled               | ON         |
+-------------------------------------------+------------+
18 rows in set (0.00 sec)

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 4        |
| Rpl_semi_sync_master_status                | ON       |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

　　2.只保留一个slave，依次停止其他3个slave

mysql5306> stop slave;
mysql6306> stop slave;
mysql7306> stop slave;

 

　　3.等待一分钟，查看master状态。（在等待期间可以通过反复查看master状态，来观察master数据的变化）

　　　　master已经转为异步模式复制。 

mysql> show global status like "%sync%";

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 1        |
| Rpl_semi_sync_master_status                | OFF      |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

　　4.使用另一个master的会话，开始一个事务并提交。 这里master提交到收到结果，该事务没有发生等待。

mysql> insert into new values(5);
Query OK, 1 row affected (0.02 sec)

 

　　5.立即查看存活的slave

　　　　　事务已经被应用。

mysql> select * from new;

+------+
| name |
+------+
|    4 |
|    5 |
+------+
2 rows in set (0.00 sec)

 

　　6.依次启动其他slave，每启动一个便立即查看master状态

　　　　可以看到，启动slave的动作会立即被master接收到，并且master会自动切换回半同步模式（Rpl_semi_sync_master_status=ON）

mysql5306> start slave;

mysql> show global status like "%sync%";

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 2        |
| Rpl_semi_sync_master_status                | OFF      |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

mysql6306> start slave;

mysql> show global status like "%sync%";

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 3        |
| Rpl_semi_sync_master_status                | ON       |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

mysql7306> start slave;

mysql> show global status like "%sync%";

+--------------------------------------------+----------+
| Variable_name                              | Value    |
+--------------------------------------------+----------+
| Rpl_semi_sync_master_clients               | 4        |
| Rpl_semi_sync_master_status                | ON       |
| Rpl_semi_sync_slave_status                 | OFF      |
+--------------------------------------------+----------+
19 rows in set (0.00 sec)

 

============================我是分割线=====================================

结论：

使用直白简短好理解的语言：

 

【rpl_semi_sync_master_wait_for_slave_count】 ，

　　1.master提交后所需的应答数量！如果slave clients数量大于等于这个值，那么master会一路畅行无阻；如果低于这个值，master可能会在事务提交阶段发生一次超时等待，当等待超过参数（rpl_semi_sync_master_timeout）设定时，master就转为异步模式（原理见下一个参数）。

　　2.master将这个参数值作为标杆，用来和【Rpl_semi_sync_master_clients】参数做比较。

【rpl_semi_sync_master_wait_no_slave】

　　1.为OFF时，只要master发现【Rpl_semi_sync_master_clients】小于【rpl_semi_sync_master_wait_for_slave_count】，则master立即转为异步模式。

　　2.为ON时，空闲时间（无事务提交）里，即使master发现【Rpl_semi_sync_master_clients】小于【rpl_semi_sync_master_wait_for_slave_count】，也不会做任何调整。只要保证在事务超时之前，master收到大于等于【rpl_semi_sync_master_wait_for_slave_count】值的ACK应答数量，master就一直保持在半同步模式；如果在事务提交阶段（master等待ACK）超时，master才会转为异步模式。

 

无论【rpl_semi_sync_master_wait_no_slave】为ON还是OFF，当slave上线到【rpl_semi_sync_master_wait_for_slave_count】值时，master都会自动由异步模式转为半同步模式。

 

粗略看来，设置为OFF时，master不会因为slave的离线而造成事务等待，这似乎是一个更合适的选择，但是为什么5.7中默认参数为ON呢？目前我还不得而知。

另外，本次实验建立在空闲场景（或者说极微小负载场景）下，在高并发的场景下，这个参数又会导致怎样的结果，这块也尚需后续探索。

 

 

 ============================我是分割线=====================================

官方文档关于参数rpl_semi_sync_master_wait_no_slave的解释:

 

Controls whether the master waits for the timeout period configured by rpl_semi_sync_master_timeout to expire, even if the slave count drops to less than the number of slaves configured by pl_semi_sync_master_wait_for_slave_count during the timeout period.

 

When the value of rpl_semi_sync_master_wait_no_slave is ON (the default), it is permissible for the slave count to drop to less than rpl_semi_sync_master_wait_for_slave_count during the timeout period. As long as enough slaves acknowledge the transaction before the timeout period expires, semisynchronous replication continues.

 

When the value of rpl_semi_sync_master_wait_no_slave is OFF, if the slave count drops to less than the number configured in rpl_semi_sync_master_wait_for_slave_count at any time during the timeout period configured by rpl_semi_sync_master_timeout, the master reverts to normal replication.

 

This variable is available only if the master-side semisynchronous replication plugin is installed.

 

官方文档关于参数rpl_semi_sync_master_wait_for_slave_count的解释:

 

The number of slave acknowledgments the master must receive per transaction before proceeding. By default rpl_semi_sync_master_wait_for_slave_count is 1, meaning that semisynchronous replication proceeds after receiving a single slave acknowledgment. Performance is best for small values of this variable.

 

For example, if rpl_semi_sync_master_wait_for_slave_count is 2, then 2 slaves must acknowledge receipt of the transaction before the timeout period configured by rpl_semi_sync_master_timeout for semisynchronous replication to proceed. If less slaves acknowledge receipt of the transaction during the timeout period, the master reverts to normal replication.

 

 

============================我是结束线=====================================

EOF