MySQL一个异常查询问题追查_MySQL
问题
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">线上碰到的问题：相同的语句，只是最后的limit行数不同。奇怪的是，limit 10 的性能比limit 100的语句还慢约10倍。隐藏用户表信息，语句及结果如下SELECTf1 , SUM(`f2`) `CNT` FROM TWHERE f1 IS NOT NULL AND f3 = '2014-05-12' GROUP BY f1 ORDER BY `CNT` DESC LIMIT 10;执行时间3 min 3.65 secSELECTf1 , SUM(`f2`) `CNT` FROM TWHERE f1 IS NOT NULL AND f3 = '2014-05-12' GROUP BY f1 ORDER BY `CNT` DESC LIMIT 100;执行时间1.24Sec.性能差距非常大！</code>
分析
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">MySQL Tips：追查语句执行时最常用的方法，是通过explain来看语句的执行计划。 </code>
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">更有冲击性的效果是通过缩小范围后，在这个数据下，limit 67和limit 68的执行计划相差很大。两个执行计划:LIMIT 67 id: 1select_type: SIMPLEtable: atype: rangepossible_keys: A,B,Ckey: Bkey_len: 387ref: NULLrows: 2555192Extra: Using where; Using temporary; Using filesort1 row in set (0.00 sec) LIMIT 68id: 1select_type: SIMPLEtable: atype: refpossible_keys: A,B,Ckey: Akey_len: 3ref: constrows: 67586Extra: Using where; Using temporary; Using filesort1 row in set (0.00 sec)可以看到，两个语句的执行计划不同：使用的索引不同。</code>
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">MySQL Tips：explain的结果中，key表示最终使用的索引，rows表示使用这个索引需要扫描的行数，这是个估计值。表中 索引A定义为 (f3, f4, f1, f2, f5); 索引B定义为(f1, f2, f3);</code>
一个确认
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">虽然rows是估计值，但是指导索引使用的依据。既然limit 68能达到rows 67586, 说明在第一个语句优化器可选结果中，也应该有此值，为什么不会选择索引A？先确认一下我们上面的这个结论。</code>
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">MySQL Tips：MySQL语法中能够用force index 来强行要求优化器使用某一个索引。</code>
Explain SELECT f1 , SUM(f2) CNT FROM t force index(A) WHERE f1 IS NOT NULL AND f3 = ‘2014-05-12’ GROUP BY P ORDER BY CNT DESC LIMIT 67/G
id: 1
select_type: SIMPLE
table: a
type: ref
possible_keys:A
key: A
key_len: 3
ref: const
rows: 67586
Extra: Using where; Using temporary; Using filesort
1 row in set (0.00 sec)
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">顺便说明，由于我们指定了force index，因此优化器不会考虑其他索引，possible_keys里只会显示A。我们关注的是rows:67586。这说明在limit 67语句里，使用索引A也能够减少行扫描。</code>
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">MySQL Tips：MySQL优化器会对possiable_key中的每个可能索引都计算查询代价，选择最小代价的查询计划。</code>
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">至此我们大概可以猜测，这个应该是MySQL实现上的bug：没有选择合适的索引，导致使用了明显错误的执行计划。</code>
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">MySQL Tips：MySQL的优化器执行期间需要依赖于表的统计信息，而统计信息是估算值，因此有可能导致得到的执行计划非最优。</code>
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">但要说明的是，上述Tip是客观情况造成（可接受），但本例却是例外，因此优化器实际上可以拿到能够作出选择正确结果的数据(rows值)，但是最终选择错误。</code>
原因分析
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">MySQL优化器是按照查询代价的估算值，来确定要使用的索引。计算这个估算值的过程，基本是按照“估计需要扫描的行数”来确定的。</code>
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">MySQL Tips：MySQL在目前集团主流使用的5.1和5.5版本中只能使用前缀索引。</code>
因此，使用索引A只能用上字段f3，使用索引B只能用上字段f1。Rows即为使用了索引查到上下界，之后需要扫描的数据行数（估算值）。
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">上述的语句需要用到group和order by，因此执行计划中都有Using temporary; Using filesort。流程上按顺序先计算使用索引A的查询代价。之后依次计算其他possitabe_key的查询代价。由于过程中需要排序，在得到一个暂定结果后，需要判断是否有代价更低的排序方式(test_if_cheaper_ordering)。与之前的大同小异，也是依靠估计扫描行数来计算代价。在这个逻辑的实现过程中，存在一个bug：在估计当前索引的区分度的时候，没有考虑到前缀索引。即：假设表中有50w行数据，索引B(f1,f2,f3)，则计算索引区分度时，需要根据能够用上的前缀部分来确定。比如f1有1000个不同的值，则平均每个key值上的记录数为500.如(f1,f2)有10000个同的值，则平均每个组合key上的记录数为50，若(f1,f2,f3)有50w个不同的值，则平均每个组合key上的记录数为1。</code>
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">MySQL Tips：每个key上的记录数越少，说明使用该索引查询时效率最高。对应于show index from tbl 输出结果中的Cardinality值越大。</code>
在这个case下，索引A只能使用f1做前缀索引，但是在计算单key上的行平均值时用的是(f1,f2,f3)，这就导致估算用索引B估算的时候，得到的代价偏小。导致误选。
回到问题本身
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">1、为什么limit值大的时候反而选对了呢？</code>这是因为在计算B的查询代价时，查询需要返回的行数limit_rows也参与乘积，若limit值较大，则计算出来的B的代价就会更大，反而会由于代价。值超过A，而导致优化器最终选择A。
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">2、这个表有50w行数就，为什么limit相差为就差别这么大？这与语句本身有关。这个语句中有group by，这就意味着每多limit一个值，实际上需要扫描更多的行N。 这里N为“表的总行数”/“表中不同的f2值”。也就是说这个语句使得这个bug有放大作用。</code>
解决方案
<code style="font-family: inherit; font-size: 14px; padding: 0px; color: inherit; background-color: transparent; white-space: pre-wrap; border-top-left-radius: 0px; border-top-right-radius: 0px; border-bottom-right-radius: 0px; border-bottom-left-radius: 0px; border: 0px;">分析清楚后解决方法就比较简单了，修改代码逻辑，在执行test_if_cheaper_ordering过程中，改用字段f1的区分度来计算即可。</code>
Hot AI Tools
Undresser.AI Undress
AI-powered app for creating realistic nude photos
AI Clothes Remover
Online AI tool for removing clothes from photos.
Undress AI Tool
Undress images for free
Clothoff.io
AI clothes remover
AI Hentai Generator
Generate AI Hentai for free.
Hot Article
Hot Tools
Notepad++7.3.1
Easy-to-use and free code editor
SublimeText3 Chinese version
Chinese version, very easy to use
Zend Studio 13.0.1
Powerful PHP integrated development environment
Dreamweaver CS6
Visual web development tools
SublimeText3 Mac version
God-level code editing software (SublimeText3)
Hot Topics
How to use Xiaohongshu account to find users? Can I find my mobile phone number?
Mar 22, 2024 am 08:40 AM
With the rapid development of social media, Xiaohongshu has become one of the most popular social platforms. Users can create a Xiaohongshu account to show their personal identity and communicate and interact with other users. If you need to find a user’s Xiaohongshu number, you can follow these simple steps. 1. How to use Xiaohongshu account to find users? 1. Open the Xiaohongshu APP, click the "Discover" button in the lower right corner, and then select the "Notes" option. 2. In the note list, find the note posted by the user you want to find. Click to enter the note details page. 3. On the note details page, click the "Follow" button below the user's avatar to enter the user's personal homepage. 4. In the upper right corner of the user's personal homepage, click the three-dot button and select "Personal Information"
Local users and groups are missing on Windows 11: How to add it
Sep 22, 2023 am 08:41 AM
The Local Users and Groups utility is built into Computer Management and can be accessed from the console or independently. However, some users find that local users and groups are missing in Windows 11. For some people who have access to it, the message suggests that this snap-in may not work with this version of Windows 10. To manage user accounts for this computer, use the User Accounts tool in Control Panel. The issue has been reported in previous iterations of Windows 10 and is usually caused by issues or oversights on the user's side. Why are local users and groups missing in Windows 11? You are running Windows Home edition, local users and groups are available on Professional edition and above. Activity
Log in to Ubuntu as superuser
Mar 20, 2024 am 10:55 AM
In Ubuntu systems, the root user is usually disabled. To activate the root user, you can use the passwd command to set a password and then use the su- command to log in as root. The root user is a user with unrestricted system administrative rights. He has permissions to access and modify files, user management, software installation and removal, and system configuration changes. There are obvious differences between the root user and ordinary users. The root user has the highest authority and broader control rights in the system. The root user can execute important system commands and edit system files, which ordinary users cannot do. In this guide, I'll explore the Ubuntu root user, how to log in as root, and how it differs from a normal user. Notice
Explore Windows 11 guide: How to access user folders on your old hard drive
Sep 27, 2023 am 10:17 AM
Certain folders are not always accessible due to permissions, and in today’s guide we will show you how to access user folders on your old hard drive on Windows 11. The process is simple but can take a while, sometimes even hours, depending on the size of the drive, so be extra patient and follow the instructions in this guide closely. Why can't I access my user folders on my old hard drive? User folders are owned by another computer, so you cannot modify them. You don't have any permissions on the folder other than ownership. How to open user files on old hard drive? 1. Take ownership of the folder and change permissions Find the old user directory, right-click on it and select Properties. Navigate to "An
What is sudo and why is it important?
Feb 21, 2024 pm 07:01 PM
sudo (superuser execution) is a key command in Linux and Unix systems that allows ordinary users to run specific commands with root privileges. The function of sudo is mainly reflected in the following aspects: Providing permission control: sudo achieves strict control over system resources and sensitive operations by authorizing users to temporarily obtain superuser permissions. Ordinary users can only obtain temporary privileges through sudo when needed, and do not need to log in as superuser all the time. Improved security: By using sudo, you can avoid using the root account during routine operations. Using the root account for all operations may lead to unexpected system damage, as any mistaken or careless operation will have full permissions. and
Tutorial: How to delete a normal user account in Ubuntu system?
Jan 02, 2024 pm 12:34 PM
Many users have been added to the Ubuntu system. I want to delete the users that are no longer in use. How to delete them? Let’s take a look at the detailed tutorial below. 1. Open the terminal command line and use the userdel command to delete the specified user. Be sure to add the sudo permission command, as shown in the figure below. 2. When deleting, be sure to be in the administrator directory. Ordinary users do not have this permission. , as shown in the figure below 3. After the delete command is executed, how to judge whether it has been truly deleted? Next we use the cat command to open the passwd file, as shown in the figure below 4. We see that the deleted user information is no longer in the passwd file, which proves that the user has been deleted, as shown in the figure below 5. Then we enter the home file
Windows 11 KB5031455 fails to install, causing other issues for some users
Nov 01, 2023 am 08:17 AM
Microsoft began rolling out KB2 to the public as an optional update for Windows 503145511H22 or later. This is the first update to enable Windows 11 Moment 4 features by default, including Windows Copilot in supported areas, preview support for items in the Start menu, ungrouping of the taskbar, and more. Additionally, it fixes several Windows 11 bugs, including potential performance issues that caused memory leaks. But ironically, the optional update for September 2023 will be a disaster for users trying to install the update, or even for those who have already installed it. Many users will not install this Wi
Analysis of user password storage mechanism in Linux system
Mar 20, 2024 pm 04:27 PM
Analysis of user password storage mechanism in Linux system In Linux system, the storage of user password is one of the very important security mechanisms. This article will analyze the storage mechanism of user passwords in Linux systems, including the encrypted storage of passwords, the password verification process, and how to securely manage user passwords. At the same time, specific code examples will be used to demonstrate the actual operation process of password storage. 1. Encrypted storage of passwords In Linux systems, user passwords are not stored in the system in plain text, but are encrypted and stored. L
