分享一则sql优化案例：生产数据库从385s优化到16.8s

Posted 2021-06-01 IT大咖说

◆ 概述

这条sql是典型的在数据量增加的情况下，mysql数据库自动选择了另一个执行计划，这里只要通过改写sql来实现该sql的优化，仅供参考。

◆ 定位慢sql

至于怎么获取到该问题sql，实际上只需要跑一下慢查询查一下就可以看到了..

有兴趣的朋友也可以看下之前介绍的慢查询平台来获取慢sql...

pt-query-digest slow.log --since '2021-01-28 00:00:00' --until '2021-01-28 23:59:00'  > /tmp/tms-slow.log

◆ 分析问题sql

可以看出只是查询一条记录但耗时385秒

SELECT DISTINCT
	t1.id,
	t1.shipment_no,
	t1.vehicle_no,
	t1.driver1_name,
	DATE_FORMAT( t1.latest_pickup_time, '%Y-%m-%d' ) AS latest_pickup_date,
	DATE_FORMAT( t1.latest_pickup_time, '%H:%i' ) AS latest_pickup,
	t1.latest_pickup_time,
	t1.version,
	t1.domain_name,
	t1.insert_user 
FROM
	fsl_shipment t1
	LEFT JOIN fsl_order_movement_unit t2 ON t1.id = t2.shipment
	LEFT JOIN fsl_order_release t3 ON t2.order_release = t3.id 
WHERE
	t1.project_code = 'xx' 
	AND t1.shipment_no IS NOT NULL 
	AND t1.shipment_status IN ( 'xx', 'xx' ) 
	AND t1.is_a_shipment = 'N' 
	AND t1.sendncicflag IS NULL 
	AND t3.customer = '3xxx6' 
	AND t1.custom_type IN (	'xx','xx')

对应的执行计划如下：

对应的表数据量情况如下：

◆ 业测环境测试

这里要说一下为什么在业测环境之所以只需要0.7s，其实是因为生产环境的t3表customer结果集比较大，导致先筛选t1表，在筛选t2表，最后筛选t3表，导致耗时接近400s；而UAT环境的t3表customer结果集小时则先筛选t3表，最后再筛选t1表，速度在1秒内。

◆ 改写sql优化

这里耗时16s。

SELECT DISTINCT
	t1.id,
	t1.shipment_no,
	t1.vehicle_no,
	t1.driver1_name,
	DATE_FORMAT( t1.latest_pickup_time, '%Y-%m-%d' ) AS latest_pickup_date,
	DATE_FORMAT( t1.latest_pickup_time, '%H:%i' ) AS latest_pickup,
	t1.latest_pickup_time,
	t1.version,
	t1.domain_name,
	t1.insert_user 
FROM
	fsl_shipment t1
	LEFT JOIN fsl_order_movement_unit t2 ON t1.id = t2.shipment
	LEFT JOIN (select id from fsl_order_release  where  customer = 'xxx') t3   ON t2.order_release = t3.id 
WHERE
	t1.project_code = 'DD' 
	AND t1.shipment_no IS NOT NULL 
	AND t1.shipment_status IN ( '18', '20' ) 
	AND t1.is_a_shipment = 'N' 
	AND t1.sendncicflag IS NULL 
	AND t1.custom_type IN ('0','4');

对应的执行计划如下;

来源：

https://www.toutiao.com/i6923561903268643331/

“IT大咖说”欢迎广大技术人员投稿，投稿邮箱：[e-mail]

来都来了，走啥走，留个言呗~

 IT大咖说  |  关于版权 

感谢您对IT大咖说的热心支持！

相关推荐

推荐文章

• 
  

• 
  

• 
  

• 
  

•