MySQL toolset: the official performance testing tool mysqlslap

brief introduction

MySQL As the most popular open source database , It is widely used in various fields , As a MySQL DBA, We often perform some performance tests on the database to take the initiative （ Or passive ） Make an assessment of business stress , To determine the current load and maximum performance capacity of the database .

Common performance testing tools are sysbench and tpcc, Both of these are excellent pressure measuring tools , But they all need special compilation or installation , And you need some development ability to modify the specific test statements .

mysqlslap It's with MySQL It will be installed automatically during installation , and mysqlslap The functions of many custom tests are encapsulated to the outside , Users only need to provide... Externally SQL Statement can customize the test statement , It will be easier to use .

Introduction

mysqlslap Provides a number of parameters to configure the type of project being tested , Here, only some common parameters are extracted for description , For details, please refer to mysqlslap Own help information .

Parameter name	explain
login-path=#	The new version MySQL Login method provided
-a, --auto-generate-sql	Automatic generation SQL sentence
--auto-generate-sql-add-autoincrement	Add auto incrementing columns to the automatically generated table
--auto-generate-sql-execute-number=#	In the test , perform SQL The total number of times
--auto-generate-sql-guid-primary	Generation based GUID Primary key of
--auto-generate-sql-load-type=name	Load model tested , Include mixed, update, write, key,read, The default is mix
--auto-generate-sql-secondary-indexes=#	Automatically generated tables , Number of secondary indexes
--auto-generate-sql-unique-query-number=#	In the test , Number of query statements that use unique indexes
--auto-generate-sql-unique-write-number=#	In the test , Using a unique index DML Number of statements
--auto-generate-sql-write-number=#	In the test , Each thread executes insert Number of statements , The default is 100
--commit=#	In the test , How many statements are executed once commit
-c, --concurrency=name	In the test , Number of concurrent threads / Number of clients
--create=name	Custom table creation statement , Or is it SQL File address
--create-schema=name	In the test , Database name used
--detach=#	In the test , Reconnection occurs after a certain number of statements are executed
-e, --engine=name	Specify the storage engine when creating tables
-h, --host=name	Specify the... Of the test instance host Address
-u, --user=name	Specify the user name of the test instance
-p, --password=name	Specify the password for the test instance
-P, --port=#	Specify the port of the test instance
-i, --iterations=#	Specify the number of times the test is repeated
--no-drop	Specify that the library table used for the test will not be deleted after the test is completed
-x, --number-char-cols=name	Specify the... In the test table varchar The number of columns
-y, --number-int-cols=name	Specify the... In the test table int The number of columns
--number-of-queries=#	Specifies the execution of each thread SQL Maximum number of statements （ Imprecise ）
--only-print	Be similar to dry run, Output what will happen , But it will not really be implemented
-F, --delimiter=name	Use the SQL When the sentence is , Explicitly specify the separator between statements
--post-query=name	After the specified test is completed , Executed query statements , Or is it SQL Statement file
--pre-query=name	Specify before the test starts , Executed query statements , Or is it SQL Statement file
-q, --query=name	When specifying tests , Executed query statements , Or is it SQL Statement file

Practical experience

Here to mysqlslap Make a simple trial , The target instance uses Tencent cloud database MySQL, The simplest test example is as follows ：

[email protected]:~# mysqlslap --concurrency=100  --iterations 10 -a  --auto-generate-sql-add-autoincrement --engine=innodb --number-of-queries=1000 -h172.1.100.10 -uroot -p
Enter password:
Benchmark
	Running for engine innodb
	Average number of seconds to run all queries: 0.046 seconds
	Minimum number of seconds to run all queries: 0.042 seconds
	Maximum number of seconds to run all queries: 0.049 seconds
	Number of clients running queries: 100
	Average number of queries per client: 10

[email protected]:~#

The simplest test model used in this example , from mysqlslap Generate all test statements and create table statements by yourself .

If you want to compare the performance under different concurrency , Can be tested in this way ：

[email protected]:~# mysqlslap --concurrency=100,200 --iterations 10 -a --auto-generate-sql-add-autoincrement --engine=innodb --number-of-queries=1000 -h172.16.0.40 -uroot -p
Enter password:
Benchmark
	Running for engine innodb
	Average number of seconds to run all queries: 0.046 seconds
	Minimum number of seconds to run all queries: 0.045 seconds
	Maximum number of seconds to run all queries: 0.048 seconds
	Number of clients running queries: 100
	Average number of queries per client: 10

Benchmark
	Running for engine innodb
	Average number of seconds to run all queries: 0.037 seconds
	Minimum number of seconds to run all queries: 0.034 seconds
	Maximum number of seconds to run all queries: 0.052 seconds
	Number of clients running queries: 200
	Average number of queries per client: 5

[email protected]:~#

The output result is time , Therefore, the more time it takes, the better the overall performance .

If you want to customize the test statement , Then you can put SQL The statement is written in the file , for example ：

[email protected]:~# cat create.sql
CREATE TABLE a (b int);INSERT INTO a VALUES (23);
[email protected]:~# cat query.sql
SELECT * FROM a;
SELECT b FROM a;
[email protected]:~#
[email protected]:~# mysqlslap --concurrency=100,200 --query=query.sql --create=create.sql --delimiter=";" --engine=innodb --number-of-queries=1000 -h172.16.0.40 -uroot -p
Enter password:
Benchmark
	Running for engine innodb
	Average number of seconds to run all queries: 0.045 seconds
	Minimum number of seconds to run all queries: 0.045 seconds
	Maximum number of seconds to run all queries: 0.045 seconds
	Number of clients running queries: 100
	Average number of queries per client: 10

Benchmark
	Running for engine innodb
	Average number of seconds to run all queries: 0.023 seconds
	Minimum number of seconds to run all queries: 0.023 seconds
	Maximum number of seconds to run all queries: 0.023 seconds
	Number of clients running queries: 200
	Average number of queries per client: 5

[email protected]:~#

Summary

Available on the market MySQL There are many pressure measuring tools , Each tool has its pros and cons , Learn more about tools , Then select the most appropriate tool according to the actual needs to complete the required stress test in the most efficient way .