Login Logout

Howto MySQL : optimisation avancée.

Pour le guide d’installation et d’usage courant, consultez /HowtoMySQL.


Après quelque temps d’utilisation, il est intéressant de regarder les résultats du script mysqltuner.pl téléchargeable via http://mysqltuner.pl/

# perl mysqltuner.pl

 >>  MySQLTuner 1.0.1 - Major Hayden <major@mhtx.net>
 >>  Bug reports, feature requests, and downloads at http://mysqltuner.com/
 >>  Run with '--help' for additional options and output filtering

-------- General Statistics --------------------------------------------------
[--] Skipped version check for MySQLTuner script
[OK] Currently running supported MySQL version 5.0.51a-24+lenny3
[OK] Operating on 64-bit architecture

-------- Storage Engine Statistics -------------------------------------------
[--] Status: +Archive -BDB -Federated +InnoDB -ISAM -NDBCluster

[--] Data in MyISAM tables: 556M (Tables: 3831)
[--] Data in InnoDB tables: 34M (Tables: 968)
[--] Data in MEMORY tables: 3K (Tables: 9)
[!!] Total fragmented tables: 335

-------- Performance Metrics -------------------------------------------------
[--] Up for: 20h 59m 23s (2M q [29.301 qps], 31K conn, TX: 5B, RX: 300M)
[--] Reads / Writes: 23% / 77%
[--] Total buffers: 58.0M global + 2.6M per thread (100 max threads)
[OK] Maximum possible memory usage: 320.5M (10% of installed RAM)
[OK] Slow queries: 0% (1/2M)
[OK] Highest usage of available connections: 11% (11/100)
[OK] Key buffer size / total MyISAM indexes: 16.0M/216.0M
[OK] Key buffer hit rate: 97.0% (31M cached / 942K reads)
[OK] Query cache efficiency: 77.2% (921K cached / 1M selects)
[!!] Query cache prunes per day: 78490
[OK] Sorts requiring temporary tables: 0% (0 temp sorts / 59K sorts)
[!!] Joins performed without indexes: 29333
[OK] Temporary tables created on disk: 25% (22K on disk / 88K total)
[OK] Thread cache hit rate: 99% (26 created / 31K connections)
[!!] Table cache hit rate: 0% (64 open / 82K opened)
[OK] Open file limit used: 12% (126/1K)
[OK] Table locks acquired immediately: 99% (1M immediate / 1M locks)
[!!] InnoDB data size / buffer pool: 34.4M/8.0M

-------- Recommendations -----------------------------------------------------
General recommendations:
    Run OPTIMIZE TABLE to defragment tables for better performance
    MySQL started within last 24 hours - recommendations may be inaccurate
    Enable the slow query log to troubleshoot bad queries
    Adjust your join queries to always utilize indexes
    Increase table_cache gradually to avoid file descriptor limits
Variables to adjust:
    query_cache_size (> 16M)
    join_buffer_size (> 128.0K, or always use indexes with joins)
    table_cache (> 64)
    innodb_buffer_pool_size (>= 34M)


Il y a aussi tunning-prime qui est plus verbeux, il complète bien mysqltuner.

         - By: Matthew Montgomery -

MySQL Version 5.1.49-3-log x86_64

Uptime = 102 days 22 hrs 6 min 25 sec
Avg. qps = 196
Total Questions = 1747455439
Threads Connected = 29

Server has been running for over 48hrs.
It should be safe to follow these recommendations

To find out more information on how each of these
runtime variables effects performance visit:
Visit http://www.mysql.com/products/enterprise/advisors.html
for info about MySQL's Enterprise Monitoring and Advisory Service

The slow query log is enabled.
Current long_query_time = 2.000000 sec.
You have 533228 out of 1747455460 that take longer than 2.000000 sec. to complete
Your long_query_time seems to be fine

The binary update log is NOT enabled.
You will not be able to do point in time recovery
See http://dev.mysql.com/doc/refman/5.1/en/point-in-time-recovery.html

Current thread_cache_size = 32
Current threads_cached = 12
Current threads_per_sec = 0
Historic threads_per_sec = 0
Your thread_cache_size is fine

Current max_connections = 250
Current threads_connected = 28
Historic max_used_connections = 250
The number of used connections is 100% of the configured maximum.
You should raise max_connections

Current InnoDB index space = 34 M
Current InnoDB data space = 77 M
Current InnoDB buffer pool free = 98 %
Current innodb_buffer_pool_size = 6.00 G
Depending on how much space your innodb indexes take up it may be safe
to increase this value to up to 2 / 3 of total system memory

Max Memory Ever Allocated : 7.67 G
Configured Max Per-thread Buffers : 671 M
Configured Max Global Buffers : 7.01 G
Configured Max Memory Limit : 7.67 G
Physical Memory : 23.58 G
Max memory limit seem to be within acceptable norms

Current MyISAM index space = 4.13 G
Current key_buffer_size = 512 M
Key cache miss rate is 1 : 4965
Key buffer free ratio = 8 %
You could increase key_buffer_size
It is safe to raise this up to 1/4 of total system memory;
assuming this is a dedicated database server.

Query cache is enabled
Current query_cache_size = 512 M
Current query_cache_used = 52 M
Current query_cache_limit = 8 M
Current Query cache Memory fill ratio = 10.34 %
Current query_cache_min_res_unit = 4 K
Your query_cache_size seems to be too high.
Perhaps you can use these resources elsewhere
MySQL won't cache query results that are larger than query_cache_limit in size

Current sort_buffer_size = 2 M
Current read_rnd_buffer_size = 256 K
Sort buffer seems to be fine

Current join_buffer_size = 132.00 K
You have had 9527 queries where a join could not use an index properly
You should enable "log-queries-not-using-indexes"
Then look for non indexed joins in the slow query log.
If you are unable to optimize your queries you may want to increase your
join_buffer_size to accommodate larger joins in one pass.

Note! This script will still suggest raising the join_buffer_size when
ANY joins not using indexes are found.

Current open_files_limit = 5260 files
The open_files_limit should typically be set to at least 2x-3x
that of table_cache if you have heavy MyISAM usage.
You currently have open more than 75% of your open_files_limit
You should set a higher value for open_files_limit in my.cnf

Current table_open_cache = 2500 tables
Current table_definition_cache = 256 tables
You have a total of 5084 tables
You have 2500 open tables.
Current table_cache hit rate is 0%
, while 100% of your table cache is in use
You should probably increase your table_cache
You should probably increase your table_definition_cache value.

Current max_heap_table_size = 64 M
Current tmp_table_size = 256 M
Of 8549110 temp tables, 35% were created on disk
Effective in-memory tmp_table_size is limited to max_heap_table_size.
Perhaps you should increase your tmp_table_size and/or max_heap_table_size
to reduce the number of disk-based temporary tables
Note! BLOB and TEXT columns are not allow in memory tables.
If you are using these columns raising these values might not impact your
ratio of on disk temp tables.

Current read_buffer_size = 128 K
Current table scan ratio = 50592 : 1
You have a high ratio of sequential access requests to SELECTs
You may benefit from raising read_buffer_size and/or improving your use of indexes.

Current Lock Wait ratio = 1 : 115
You may benefit from selective use of InnoDB.
If you have long running SELECT's against MyISAM tables and perform
frequent updates consider setting 'low_priority_updates=1'
If you have a high concurrency of inserts on Dynamic row-length tables
consider setting 'concurrent_insert=2'.

Support des large pages

Le support des large pages permet d’avoir une gestion différente de la mémoire, notamment utile pour les requêtes avec des besoins en mémoire important. Ceci n’est utile dans la plupart des cas a priori.

Pour l’activer, il faudra vérifier que le kernel supporte les large pages :

#  cat /proc/meminfo | grep -i huge
HugePages_Total:       0
HugePages_Free:        0
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:       2048 kB

Si cela ne renvoie rien c’est que le support des large pages est désactivé. (options CONFIG_HUGETLBFS et CONFIG_HUGETLB_PAGE du noyau).

Pour l’activer, on fera les opérations suivantes :

# Set the number of pages to be used.
# Each page is normally 2MB, so a value of 20 = 40MB.
# This command actually allocates memory, so this much
# memory must be available.

# echo 20 > /proc/sys/vm/nr_hugepages

# Set the group number that is permitted to access this
# memory (110 in this case). The mysql user must be a
# member of this group.

# echo 110 > /proc/sys/vm/hugetlb_shm_group

# Increase the amount of shmem permitted per segment
# (900M in this case).

# echo 943718400 > /proc/sys/kernel/shmmax

Ajouter dans le /etc/security/limits.conf :

ulimit -l unlimited

Enfin dans la configuration MySQL :


Mais encore une fois, cela n’est pas nécessaire dans la plupart des cas.

À propos de shmmax

Since shm (shared memory) in Unix/Linux is only used to share memory between independent processes and mysql server is a single process (multi-threaded), I am convinced that shmmax value has no consequence on mysql. (OTOH Oracle has a multiprocess architecture and uses shared memory).

Il semblerait que contrairement à PostgreSQL, il ne soit pas nécessaire d’augmenter le kernel.shmmax dans la plupart des cas Source.

Benchmark avec sql-bench

Récupérer les fichiers de sql-bench avec bazar :

$ bzr branch lp:sql-bench

Installer quelques outils nécessaires pour le benchmark :

# apt install libclass-dbi-perl libdbd-mysql-perl mysql-client-5.1

Créer une base de données de test :

# mysqladmin create test
# mysql
mysql> GRANT ALL PRIVILEGES ON test.* TO 'test'@localhost IDENTIFIED BY 'test';

Résultat du benchmark :

# cd ~/mysql-5.5.12/sql-bench/sql-bench/
# perl run-all-tests --user='test' --password='test'
Benchmark DBD suite: 2.15
Date of test:        2011-05-19 14:01:18
Running tests on:    Linux 2.6.32-5-amd64 x86_64
Limits from:         
Server version:      MySQL 5.1.49 3
Optimization:        None

Operation                           seconds     usr     sys     cpu   tests
alter-table: Total time: 16 wallclock secs ( 0.08 usr  0.00 sys +  0.00 cusr  0.00 csys =  0.08 CPU)
ATIS: Total time:  4 wallclock secs ( 2.34 usr  0.77 sys +  0.00 cusr  0.00 csys =  3.11 CPU)
big-tables: Total time:  5 wallclock secs ( 0.92 usr  2.47 sys +  0.00 cusr  0.00 csys =  3.39 CPU)
connect: Total time: 18 wallclock secs ( 4.85 usr  2.88 sys +  0.00 cusr  0.00 csys =  7.73 CPU)
create: Total time: 53 wallclock secs ( 3.86 usr  1.75 sys +  0.00 cusr  0.00 csys =  5.61 CPU)
insert: Total time: 618 wallclock secs (221.17 usr 14.05 sys +  0.00 cusr  0.00 csys = 235.22 CPU)
select: Total time: 49 wallclock secs (11.28 usr  8.95 sys +  0.00 cusr  0.00 csys = 20.23 CPU)
transactions: Test skipped because the database doesn't support transactions
wisconsin: Total time:  5 wallclock secs ( 1.93 usr  0.47 sys +  0.00 cusr  0.00 csys =  2.40 CPU)
TOTALS                               788.00  244.02   30.97  274.99 2913950


La première fois que vous faites un SELECT, son résultat sera conservé dans le Query Cache de MySQL. Si ce cache n’a pas été invalidé, la même requête retournera directement son résultat à partir de cache, soit un résultat quasi-instantané.

On peut avoir des stats sur ce Query Cache :

mysql> SHOW STATUS LIKE 'Qcache%';
| Variable_name           | Value     |
| Qcache_free_blocks      | 28        |
| Qcache_free_memory      | 242412400 |
| Qcache_hits             | 226483735 |
| Qcache_inserts          | 22334124  |
| Qcache_lowmem_prunes    | 5337097   |
| Qcache_not_cached       | 3323695   |
| Qcache_queries_in_cache | 4229      |
| Qcache_total_blocks     | 8919      |
8 rows in set (0.00 sec)

On peut vider ce cache ainsi :

Query OK, 0 rows affected (0.00 sec)


Il peut être intéressant de changer le chemin du tmpdir pour le mettre sur un disque plus rapide ou plus gros par exemple.

# mkdir /ssd/mysqltmp
# chown mysql: /ssd/mysqltmp
# chmod 700 /ssd/mysqltmp

Puis dans la configuration MySQL :

tmpdir = /ssd/mysqltmp

OPTIMIZE TABLE avec des partitions

Quand on a des partitions il faut remplacer le OPTIMIZE TABLE par :