线程池的说明

MySQL数据库默认是为每一个连接分配一个服务端的线程进行处理,在一个连接数少并且运行稳定的生产系统中,这通常不会有什么问题。但在下面两种情况下,会遇到性能问题
前端发起的连接数过多,造成服务端生成大量的线程,消耗了大量的CPU资源,而这些连接大部分时候处于sleep的状态,后端线程空转。
瞬间增加大量短连接的会话,造成服务端不断生成线程,高峰过后又有大量线程被销毁,这样会因为context-switch造成CPU性能大幅下降。
对于这种情况通常采用线程池的方法进行解决,线程池会预先创建一定数量的线程,当有连接的请求需要处理时,MySQL会从线程池中分配一条线程进行处理,到处理完成后,这条线程又会回到线程池中,下次它可能会被分配处理另外一个连接到请求。

单连接对单线程和线程池的对比

线程类型

适用场景

线程执行单位

one-thread-per-connection

连接数不多而且稳定

连接

pool-of-threads

高并发、短连接

SQL语句

需要说明的是:服务端的线程池和客户端的连接池不是一回事。

连接池和线程池的对比

池的类型

优化目标

功能

Connection Pool

客户端

减少建立连接的时间和连接占用的资源

Thread Pool

服务端

减少线程重复建立和context-switch的资源

线程池在MySQL的企业版才有,社区版没有,下面的图说明了MySQL企业版通过线程池有效地提高MySQL扩展性

(https://www.mysql.com/products/enterprise/scalability.html)

MySQL 线程池安装 mysql线程池优化_线程池


Oracle声称线程池让MySQL的扩展性提高了60倍。遗憾的是免费的社区版中并没有这个功能,但免费的Percona和MariaDB中都有这个功能。这里以MariaDB为例介绍线程池的作用。

安装MariaDB

在MariaDB的官方网站上可以下载安装包,在Linux上的安装方法如下:

groupadd mysql
useradd -g mysql mysql
cd /usr/local
tar -zxvpf /path-to/mariadb-VERSION-OS.tar.gz
ln -s mariadb-VERSION-OS mysql
cd mysql
./scripts/mysql_install_db --user=mysql
chown -R root .
chown -R mysql data

启动MariaDB:

root@yaoyuan ~# /usr/local/mysql/bin/mysqld_safe  --user=mysql &

登陆MariaDB,root或者mysql用户登陆不需要密码

[mysql@yuan ~]$ mysql
Welcome to the MariaDB monitor.  Commands end with ; or \g.
Your MariaDB connection id is 5
Server version: 10.6.11-MariaDB MariaDB Server


Copyright (c) 2000, 2018, Oracle, MariaDB Corporation Ab and others.


Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.


MariaDB [(none)]> \s
--------------
mysql  Ver 15.1 Distrib 10.6.11-MariaDB, for linux-systemd (x86_64) using readline 5.1


Connection id:    5
Current database:  
Current user:    mysql@localhost
SSL:      Not in use
Current pager:    stdout
Using outfile:    ''
Using delimiter:  ;
Server:      MariaDB
Server version:    10.6.11-MariaDB MariaDB Server
Protocol version:  10
Connection:    Localhost via UNIX socket
Server characterset:  latin1
Db     characterset:  latin1
Client characterset:  utf8mb3
Conn.  characterset:  utf8mb3
UNIX socket:    /tmp/mysql.sock
Uptime:      1 hour 34 min 51 sec


Threads: 1  Questions: 47  Slow queries: 0  Opens: 37  Open tables: 30  Queries per second avg: 0.008
--------------

配置线程池

控制线程池的主要参数是thread_handling,这个参数的默认值one-thread-per-connection,就是一个连接对应一个线程。

MariaDB [information_schema]> show variables like 'thread_h%';
+-----------------+---------------------------+
| Variable_name   | Value                     |
+-----------------+---------------------------+
| thread_handling | one-thread-per-connection |
+-----------------+---------------------------+
1 row in set (0.001 sec)

激活线程池的方法是在参数文件中把这个参数配置成pool-of-threads:

[mariadb]
...
thread_handling=pool-of-threads

启动后可以看到线程池的相关参数的默认值

MariaDB [(none)]> show variables like 'thread%';
+--------------------------------+-----------------+
| Variable_name                  | Value           |
+--------------------------------+-----------------+
| thread_cache_size              | 151             |
| thread_handling                | pool-of-threads |
| thread_pool_dedicated_listener | OFF             |
| thread_pool_exact_stats        | OFF             |
| thread_pool_idle_timeout       | 60              |
| thread_pool_max_threads        | 65536           |
| thread_pool_oversubscribe      | 3               |
| thread_pool_prio_kickup_timer  | 1000            |
| thread_pool_priority           | auto            |
| thread_pool_size               | 1               |
| thread_pool_stall_limit        | 500             |
| thread_stack                   | 299008          |
+--------------------------------+-----------------+
12 rows in set (0.075 sec)

hread_pool_size:线程池的大小,这个参数的默认值由cpu的个数决定,默认是CPU的个数,这里是虚拟机,只有一个虚拟CPU,因此是1。
thread_pool_max_threads:线程池中的最大线程,默认65536。

测试线程池

测试环境是只有一个虚拟CPU的虚拟机。

参数配置 /etc/my.cnf

[mariadb]
thread_handling=pool-of-threads
thread_pool_size=10
max_connections=1000
thread_pool_max_threads=50

创建1万条记录的表

[mysql@yuan ~]$ mysqlslap --create="create table table_a(col1 int primary key auto_increment, col2 varchar(255))" --query="insert into table_a(col2) values(md5(rand()))" --concurrency=10 --number-of-queries=10000 --no-drop
Benchmark
  Average number of seconds to run all queries: 1.982 seconds
  Minimum number of seconds to run all queries: 1.982 seconds
  Maximum number of seconds to run all queries: 1.982 seconds
  Number of clients running queries: 10
  Average number of queries per client: 1000

并发500个连接进行测试

[mysql@yuan ~]$  mysqlslap --query='select count(*) from table_a where col2<>md5(rand())' --concurrency=500 --iterations=10
Benchmark
  Average number of seconds to run all queries: 7.809 seconds
  Minimum number of seconds to run all queries: 7.762 seconds
  Maximum number of seconds to run all queries: 8.033 seconds
  Number of clients running queries: 500
  Average number of queries per client: 1

每个SQL语句运行的时间大约是7秒多,实际上这是因为并发造成context-switch而影响了执行的效率,如果是单线程会快很多,例如:

[mysql@yuan ~]$  mysqlslap --query='select count(*) from table_a where col2<>md5(rand())' --concurrency=1 --iterations=10
Benchmark
  Average number of seconds to run all queries: 0.015 seconds
  Minimum number of seconds to run all queries: 0.015 seconds
  Maximum number of seconds to run all queries: 0.016 seconds
  Number of clients running queries: 1
  Average number of queries per client: 1

观察500个连接运行时的状态参数

[mysql@yuan ~]$  mysqladmin extended-status -i5 -c100|grep Thread
| Threadpool_idle_threads                                | 24                                               |
| Threadpool_threads                                     | 50                                               |
| Threads_cached                                         | 0                                                |
| Threads_connected                                      | 371                                              |
| Threads_created                                        | 113                                              |
| Threads_running                                        | 26                                               |

| Threadpool_idle_threads                                | 32                                               |
| Threadpool_threads                                     | 50                                               |
| Threads_cached                                         | 0                                                |
| Threads_connected                                      | 195                                              |
| Threads_created                                        | 113                                              |
| Threads_running                                        | 18                                               |

| Threadpool_idle_threads                                | 20                                               |
| Threadpool_threads                                     | 50                                               |
| Threads_cached                                         | 0                                                |
| Threads_connected                                      | 465                                              |
| Threads_created                                        | 113                                              |
| Threads_running                                        | 30                                               |

| Threadpool_idle_threads                                | 24                                               |
| Threadpool_threads                                     | 50                                               |
| Threads_cached                                         | 0                                                |
| Threads_connected                                      | 287                                              |
| Threads_created                                        | 113                                              |
| Threads_running                                        | 25                                               |

| Threadpool_idle_threads                                | 48                                               |
| Threadpool_threads                                     | 50                                               |
| Threads_cached                                         | 0                                                |
| Threads_connected                                      | 47                                               |
| Threads_created                                        | 113                                              |
| Threads_running                                        | 2                                                |

分析MariaDB的状态参数可以得到以下结论:

  • Threadpool_idle_threads+Threads_running=Threadpool_threads
  • Threads_created值一直没有变,表示一直没有创建新的线程处理连接。
  • Threads_cached是在one-thread-per-connection时缓存的线程,在线程池使用时是零。

实际互联网的应用中,网民登录到网站后大部分时间是浏览网页,执行操作的时间只占很少一部分,可以使用下面的命令模拟网民的每一分钟执行一个SQL:

[mysql@yuan ~]$  mysqlslap --query='select count(*) from table_a where col2<>md5(rand());select sleep(6);' --concurrency=1 --iterations=20
Benchmark
  Average number of seconds to run all queries: 6.015 seconds
  Minimum number of seconds to run all queries: 6.015 seconds
  Maximum number of seconds to run all queries: 6.017 seconds
  Number of clients running queries: 1
  Average number of queries per client: 1

再次观察MariaDB的状态参数

| Threadpool_idle_threads                                | 19                                               |
| Threadpool_threads                                     | 20                                               |
| Threads_cached                                         | 0                                                |
| Threads_connected                                      | 5                                                |
| Threads_created                                        | 133                                              |
| Threads_running                                        | 2    
                                            |
| Threadpool_idle_threads                                | 19                                               |
| Threadpool_threads                                     | 20                                               |
| Threads_cached                                         | 0                                                |
| Threads_connected                                      | 5                                                |
| Threads_created                                        | 133                                              |
| Threads_running                                        | 2                                                |

| Threadpool_idle_threads                                | 19                                               |
| Threadpool_threads                                     | 20                                               |
| Threads_cached                                         | 0                                                |
| Threads_connected                                      | 5                                                |
| Threads_created                                        | 133                                              |
| Threads_running                                        | 2                                                |

发现这时线程池里面的线程数下降到20个,大部分时间线程池中只有一个线程在提供服务。

总结

经过前面的测试,我们可以看到,线程池对减少MySQL服务端线程数量的作用非常明显,通过包括一定数量的线程的线程池可以向大量的连接提供服务。
如果连接中执行的SQL运行时间较长,例如数据仓库类的业务,就不适合用线程池。另外非常简单的SQL(例如select 1)也会因为启动了线程池而比直接连接时运行得稍微慢一点。