Si vous utilisez MySQL de façon sérieuse, je ne peux vous dire à quel point vous devriez suivre le blogue MySQL Entomologist religieusement! L’auteur y expose les bogues, les problèmes, les particularités de MySQL dans un style efficace, sans biais ni embellissement : que la vérité crue!
Comme par exemple les nombreux problèmes avec les indexes fulltext ou les statistiques persistentes de InnoDB.
Leave a Comment » | 
bases de données, database, InnoDb, MySQL | Tagué: base de données, bd, blogue, bogues, database, db, entomologist, fulltext, InnoDb, MySQL, statistics | 
Permalien
Publié par endormitoire
Une page où vous pourrez avoir réponse à presque toutes vos questions concernant MySQL!
Leave a Comment » | bases de données, MySQL | Tagué: InnoDb, MyISAM, MySQL, snippets, tips | Permalien
Publié par endormitoire
Vous cherchez des solutions à des problèmes 1000 fois rencontrés? Il y a de grandes chances que vous trouviez ce qu’il vous faut ici!
Un excellent blogue consacré à MySQL sur lequel je suis tombé par hasard, lefred.be.
Un article portant sur l’erreur 1215 (« Cannot add foreign key constraint« ).
Les InnoDB Page merge & split expliqués en détail dans ce billet.
Comment choisir l’index approprié pour les requêtes avec égalité ou LIKE ?
Les attaques de réseau et MySQL : une série d’articles sur MySQL et la sécurité.
Configuration de base de MySQL 5.7 : quelques trucs et suggestions.
Leave a Comment » | bases de données, database, MySQL | Tagué: 1215, 5.7, attacks, attaque, configuration, index, InnoDb, LIKE, merge, network, page, réseau, split | Permalien
Publié par endormitoire
Sure, there are tools and script to help you fine tune your MySQL server but there’s nothing like using your brain.
Pythian has done a wonderful job of explaining what all those InnoDb-related variables in a series of articles…
Exposing Innodb Internals via System Variables: Part 1, Memory
Exposing Innodb Internals via System Variables: Part 2, I/O (structure and logs)
Exposing Innodb Internals via System Variables: Part 3, I/O (Table data)
Exposing Innodb Internals via System Variables: Part 4, Concurrency
Exposing Innodb Internals via System Variables: Part 5, Consistency / Statistics handling
Leave a Comment » | InnoDb, MySQL | Tagué: InnoDb, MySQL, performance, server, tuning, variables | Permalien
Publié par endormitoire
Note: this article was originally posted in French here.
You! Yes! Yep, you! I’m talking to you my friend! You, the one that I help out in forums, on IRC, on mailing lists!
Each and every day, I answer questions related to MySQL, mostly about SQL queries or optimization problems. Whether it’s on dBForums, Stack Overflow, MySQL.com, developpez.net or on IRC (irc.freenode.net, channel #mysql), it’s always the same thing! If you want me to help you, make it easy for me!
So, here’s a quick guide explaining what you have to do if you want us/me to help you solve your problem as quickly as possible.
Explain what your problem is as simply and clearly as possible
If you’re unable to explain, in one sentence, what you are trying to do, we’re off to a very bad start. You have to know what you’re talking about first! You have to know what you’re trying to do in the first place!
A good example: « I’m trying to find the best salesman in every store located in a district that sells less than the national district average« . This is clear and simple to understand!
A bad example: « My SQL query is slow, there it is » :
SELECT
COUNT(*) AS x1k1, w.nndt, w.n01_d
FROM
p_usr_htgg w
WHERE
EXISTS (SELECT
w2.id
FROM
puw_user w2
LEFT JOIN
temp_eoms te ON te.mt = w2.mt
WHERE
w2.k1 = w.k1 AND w.sales = MAX(w2.sales)
)
GROUP BY
w.nndt, w.n01_d
HAVING
COUNT(*) > 3
ORDER BY
x1k1 DESC
LIMIT
0, 100
Can you see the difference between these two examples? In one case, the problem is clear. In the second case, just seeing the SQL without any further explanation makes me want to puke! How can I help with all that SQL garbage?
Start small! If you have no clue about how to use JOINs, don’t focus on the the ROLLUP or aggregate stuff!!! Learn the basics while you’re at it! Learn the lingo as well. If you don’t know what something means, ask it! Don’t pretend you know when you don’t!
One more thing : pick names that are meaningful for tables and columns! Acronyms and cryptic abbreviations tell me nothing! Same thing for table and column names in Italian, Greek, Spanish or German! Use English !
Be precise
« I have a huge database with an enormous amount of concurrent users and my query is very slow and it is executed many many times« .
What is huge for you is probably a joke for Google, Amazon and GitHub or for someone working in bioinformatics: give me numbers (number of records, tablespace size, etc.), not your impressions!
« Enormous » is often relative! Fifty concurrent users and a table of 100 million rows on a server with 2G of RAM is an enormous amount of users! But on a server with 16 processors and 128G of RAM, 50 concurrent users is not a problem at all : once again, give me numbers, not adjectives!
« Many many times » doesn’t mean anything to me! What does that mean exactly? 200 times/minute? 15 times per minute? What? Give me precise numbers! What do you mean by slow? 3 seconds or 0.2 seconds? Be precise!
Don’t make me google for you
Please, make some minimal effort! Often times, I can give you an URL explaining your problem faster than the time it took you to write your question! You would have saved everyone’s precious time if you would have googled in the first place!
Give me some test data and expected results
« It doesn’t work! I should have six xyz records for every abcd record but I only get five ».
This tells me nothing useful! I don’t know your database, your tables, your domain, the content of the columns, etc. Even worse, I don’t know what’s in those tables, how the data is spread out, the selectivity of some columns, etc. Creating a test example with some data only takes two minutes on SQLFiddle!
Show me the EXPLAIN
Your query is slow? So what! Why? I have no way to know unless you give me the EXPLAIN of your query, I cannot help you at all nor even know why MySQL picks a particular access plan over a more promising one!
Show me the SHOW CREATE TABLE
Even if you provided me with the EXPLAIN of your SQL query, I still need to know what the table looks like. One is useless without the other! I also need the SHOW CREATE TABLE, it’s as simple as that!
Are both keys referenced in your JOIN of the same type and length? Is the LIKE in your query way too slow because we have to scan 4Mb of text in every record? Are we trying to match records by looking into a 1G BLOB column? Is it an InnoDB, a MyISAM or another table type? What are the character set and collation?
Learn to use SHOW VARIABLES
In some cases, your query is fine and the problems you encounter are caused by session/server/client parameters! You have to know how to use SHOW VARIABLES!
Learn to use SHOW STATUS
Sometimes, you have to be able to check the status of your server. This is essential! In that case, SHOW STATUS is another command you must know!
Use SHOW TABLE STATUS
The SHOW TABLE STATUS will allow you, in a glance, to have a precise picture of what’s in your database. You’ll be able to see, in one shot, table names, their type, their size, collation info, etc. Very useful command : for you and for me!
Give me some volumetrics
I just said it earlier and I’ll say it again : I need numbers, measures, to help you! Give me numbers, not fuzzy impressions!
Most of your queries execute in 0.1 second but your server cannot keep up? Why? Perhaps, it would have been useful to know that you have 3500 concurrent connections at any moment hitting this poor server! Or that one particular query is executed 900 times per minute!
Your query takes 27 seconds to complete? Perhaps you should have told me that your table has 728 million records!
You wonder why the EXPLAIN of your query shows a full table scan even though you’ve indexed just about everything on that table? Had I known your table only has 8 records, I would have told you that in that case a full table scan is perfectly normal!
If you want me to help you, I need a minimum of context ! Only numbers can tell me the real story! Not « enormous » nor « many » or « slow » !
Take a deep breath : do not panic!
Premature optimization is the root of all evil.
« I have this HYPER-HUGE-BIGGEST-OH-MY-GOD database and my query, with 2 INNER JOIN, is executed every 15 seconds and we add 2500 records to the main table every month… »
Take it easy! Relax! MySQL is not Excel : it can handle lots of data! For my own amusement, my server at home has 10000+ tables! Performance is mostly a matter of hardware and every database is different!
Unless you’re trying to compete with Google or you have a very complex query that causes real problems, don’t try to optimize every SQL query and make it execute in less that 100 milliseconds! In some cases, it’s just not possible possible. Focus on the real problems!
Don’t make me sweat over a query for 4 hours and end up telling me that this SQL runs once a year on January 1st at 2h15 AM and it takes 5 minutes to complete when only 2 users are connected! Pick your battles! I don’t mind helping you out but I don’t like to waste my time!
Don’t try to optimize everything! Work on what counts!
Use SQLFiddle
It’s easy to use and it works great when debugging SQL queries! And it’s free! And it’s here! And it allows us to work on exactly the same thing!
Use a pastebin
Please, don’t flood the IRC channel with a query of 35 lines! It’s annoying for everyone and it’s rude! Use a pastebin (or even better, SQLFiddle). There’s Pastie, gist.github et pastebin.com.
Use the MySQL client
Some problems are extremely hard to reproduce and difficult to understand. The last thing we need is some PHP code on top of your query! Another thing: please, don’t use crap like PHPMyAdmin or any other clumsy tool when trying to debug or fix things! All you need is the MySQL client ! It works just fine on every platform supported by MySQL, it’s rock solid and, honestly, it’s all you need! If you’re on Windows, you can always use HeidiSQL.
Use \G
Instead of forcing me to scroll to the right for eons just to see the complete result of your query, an EXPLAIN or a SHOW TABLE STATUS, add \G at the end of your SQL statement to force the client to display results vertically and limit them to 80 chars per line if necessary!
Format your code
SELECT count(*) as x1k1, w.nndt, w.n01_d from P_USR_HTGG w where exists ( SELECT w2.id FROM puw_user w2 left join temp_eoms te ON te.mt = w2.mt where w2.k1 = w.k1 AND w.sales = MAX(w2.sales)) group by w.nndt, w.n01_d having count(*) > 3 order BY x1k1 desc limit 0, 100
Can you quickly see what going on in that query? Not me! And it’s the exact same query that was at the beginning of this post! Which one is easier to read?
Write the reserved words and functions in uppercase! SELECT, FROM, WHERE, INNER JOIN, SUBSTRING, HAVING, GROUP BY, ORDER BY, IN, EXISTS, COUNT, etc. Indent your code and regroup « logical » parts (for instance, sub-selects) of the work done by your query. If I have to read your query with a magnifier to identify what I’m looking for, it doesn’t help!
Don’t be overexcited
We’ve been working on your query for 30 minutes and now you’re satisfied with the speed and results? Comes out you only needed an extra index?
Don’t precipitate yourself to add the index in production! What sometimes takes 3 seconds on test data can take hours on a production database! Besides, when applying changes to a production database, you should always make sure you have a recent backup first!
This advice can sound somewhat useless and stupid but not so long ago, a poor guy I was helping had the very bright idea to add the missing index to his production database right away… only to realize that his production table had 94 million records! The index creation took a few hours… during the peak hours!
Be open to criticism
Lots of times, a poor design or a bad choice of data types is the problem, not the circumvoluted query you’re trying to optimize. Your query is unnecessary complex because of this! Be open to suggestions : we’re just trying to help you!
Like here.
Say thanks
I’m helping people with their MySQL/SQL problems because I like it and the diverse nature of problems out there helps me stay mentally sharp and up-to-date with MySQL. If I spend 2 hours of my time helping you, a « thank you » is not much to ask in return!
Save
Leave a Comment » | MySQL | Tagué: dBForums.com, developpez.net, EXPLAIN, forums.MySQL.com, gist.github, HeidiSQL, InnoDb, IRC, irc.freenode.net, MyISAM, MySQL, optimisation, pastebin, pastebin.com, Pastie, PHPMyAdmin, requêtes, SHOW CREATE TABLE, SHOW STATUS, show table status, SHOW VARIABLES, SQL, SQLFiddle, SQLFiddle.com, StackOverflow.com | Permalien
Publié par endormitoire
Si vous constatez une dégradation de certaines requêtes SQL sur votre serveur MySQL 5.6, c’est probablement que vos statistiques ne sont pas à jour…
Un excellent article sur un des nouveaux paramètres qui pourrait vous aider, ici.
Leave a Comment » | MySQL | Tagué: indexes, InnoDb, MySQL, persistent statistics, statistiques, statistiques persistentes, STATS_SAMPLE_PAGES | Permalien
Publié par endormitoire
L'endormitoire 