Mysql调优之profile详解

前言

  1. 在我们做mysql性能分析的时候,最常用的有三种方式:

    (1)慢查询 (分析出现出问题的sql)
    (2)Explain (显示了mysql如何使用索引来处理select语句以及连接表。可以帮助选择更好的索引和写出更优化的查询语句)
    (3)Profile(查询到 SQL 会执行多少时间, 并看出 CPU/Memory 使用量, 执行过程中 Systemlock, Table lock 花多少时间等等.)
  2. 本章主要是对profile做简单的概述,用来对某一条sql语句进行性能分析。
  3. Profiling是从 mysql5.0.3版本以后才开放的。但是在mysql5.7之后,profile信息将逐渐被废弃,mysql推荐使用performance schema。
  4. profile此工具可用来查询SQL执行状态,System lock和Table lock 花多少时间等等,对定位一条语句的I/O消耗和CPU消耗 非常重要。(SQL 语句执行所消耗的最大两部分资源就是IO和CPU)

profile工具使用

  1. 查看自己的mysql版本:

    mysql> select version(); 
    +------------+
    | version()  |
    +------------+
    | 5.6.35-log |
    +------------+ 
  2. 查看是否开启profile功能(profiling=on代表开启):

    mysql> show variables like '%profil%';
    +------------------------+-------+
    | Variable_name          | Value |
    +------------------------+-------+
    | have_profiling        | YES  |
    | profiling              | OFF    |
    | profiling_history_size | 15    |
    +------------------------+-------+ 
  3. 开启profile:

    mysql> set profile=1; 
  4. 开启profile之后,执行要分析的sql语句:

    mysql> select t1.*,t2.action from pre_forum_thread as t1
    left join 
    (select a.* from pre_forum_threadmod as a,(select tid,max(dateline) as dateline from pre_forum_threadmod group by tid) as b
    where a.tid=b.tid and a.dateline=b.dateline) as t2
    on t1.tid=t2.tid
    where t1.displayorder>=0 and t1.fid in (47,49) and t1.tid > 100318 
    and (t1.authorid =7683017 or t2.action<>'DWN' or t2.action is null )
    order by t1.dateline desc limit 20;
  5. 查看生成的profile信息:

    mysql> show profiles;
    +----------+------------+--------------------------------------------------------------------------------------------------------+
    | Query_ID | Duration  | Query                                                                                                  |
    +----------+------------+--------------------------------------------------------------------------------------------------------+
    |        1 | 1.37183777 | select t1.*,t2.action from pre_forum_thread as t1  |
    |        2 | 0.00078796 | show columns from `bbs`.`t2` |
    |        3 | 0.00150425 | show columns from `bbs`.`pre_forum_thread` |
    +----------+------------+--------------------------------------------------------------------------------------------------------+
  6. 获取指定的query语句的开销:

    mysql> show profile for query 2; 
    +----------------------+----------+
    | Status              | Duration |
    +----------------------+----------+
    | starting            | 0.000147 |
    | checking permissions | 0.000023 |
    | Opening tables      | 0.000047 |
    | init                | 0.000081 |
    | System lock          | 0.000031 |
    | optimizing          | 0.000034 |
    | statistics          | 0.001650 |
    | preparing            | 0.000046 |
    | executing            | 0.000018 |
    | Sending data        | 2.460588 |
    | end                  | 0.000041 |
    | query end            | 0.000019 |
    | closing tables      | 0.000022 |
    | freeing items        | 0.000055 |
    | cleaning up          | 0.000085 |
    +----------------------+----------+ 
    
  7. 关闭profile:

    mysql> set profiling=0;
  8. 相关具体的参数:

    type: 
       ALL                --显示所有的开销信息 
     | BLOCK IO          --显示块IO相关开销 
     | CONTEXT SWITCHES  --上下文切换相关开销 
     | CPU                --显示CPU相关开销信息 
     | IPC                --显示发送和接收相关开销信息 
     | MEMORY            --显示内存相关开销信息 
     | PAGE FAULTS        --显示页面错误相关开销信息 
     | SOURCE            --显示和Source_function,Source_file,Source_line相关的开销信息 
     | SWAPS              --显示交换次数相关开销的信息 
    
    例如,想要查看cpu和io开销可以执行命令:
    mysql> SHOW profile CPU,BLOCK IO  FOR query 2;

总结

  1. 一般简易的流程:

    (1)set profiling=1; //打开profile分析
    (2)run your sql1;
    (3)run your sql2;
    (4)show profiles;    //查看sql1,sql2的语句分析
    (5)SHOW profile CPU,BLOCK IO io FOR query 1; //查看CPU、IO消耗
    (6)set profiling=0; //关闭profile分析

参考链接

https://blog.csdn.net/ty_hf/a…
https://www.linuxidc.com/Linu…

    原文作者:繁星落眼眶
    原文地址: https://segmentfault.com/a/1190000017099375
    本文转自网络文章,转载此文章仅为分享知识,如有侵权,请联系博主进行删除。
点赞