有這樣一條查詢語句:mysql
select sum(index_count) as count, theday ,appVersion, channel, type
from tableA
where project_id='qjp' and theday>='2019-05-17' and theday<='2019-05-23'
group by theday
複製代碼
執行用時21s. 數據量55萬左右,實在太慢了。sql
首先我須要肯定這21s 都用在什麼地方了,即性能瓶頸在哪裏。須要用到Msql的 Query Profiler 診斷分析工具。使用方法以下:bash
root@localhost : (none) 10:53:11> set profiling=1;
Query OK, 0 rows affected (0.00 sec)
複製代碼
mysql> select sum(index_count) as count, theday from tableA where project_id='qjp' and theday>='2019-05-17' and theday<='2019-05-23' group by theday
+-----------+-----------+------------+------------+------
| count | theday | appVersion | channel | type |
+-----------+-----------+------------+------------+------
| 180205137 | 2019-05-17 | 1.2.6 | 55550006 | |
| 168597045 | 2019-05-18 | 1.2.7 | 55550337 | |
| 153154098 | 2019-05-19 | 1.2.7 | 55550006 | |
+-----------+------------+------------+----------+------+
7 rows in set (21.03 sec)
複製代碼
在開啓 Query Profiler 功能以後,MySQL 就會自動記錄全部執行的 Query 的 profile 信息了。app
mysql> show profiles;
+----------+-------------+----------------------------------------------------------------------------------------------------------------------------------------------+
| Query_ID | Duration | Query |
+----------+-------------+----------------------------------------------------------------------------------------------------------------------------------------------+
| 1 | 21.02640425 | select sum(index_count) as count, theday from tableA where project_id='qjp' and theday>='2019-05-17' and theday<='2019-05-23' group by theday|
+----------+-------------+----------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set, 1 warning (0.00 sec)
複製代碼
經過執行 「show profiles 」 命令獲取當前系統中保存的多個 Query 的 profile 的概要信息。工具
mysql> show profile for query 1;
+----------------------+-----------+
| Status | Duration |
+----------------------+-----------+
| starting | 0.000117 |
| checking permissions | 0.000020 |
| Opening tables | 0.000032 |
| init | 0.000039 |
| System lock | 0.000021 |
| optimizing | 0.000023 |
| statistics | 0.000124 |
| preparing | 0.000032 |
| Creating tmp table | 0.000042 |
| Sorting result | 0.000018 |
| executing | 0.000016 |
| Sending data | 21.021533 |
| Creating sort index | 0.003272 |
| end | 0.000019 |
| query end | 0.000022 |
| removing tmp table | 0.000728 |
| query end | 0.000019 |
| closing tables | 0.000023 |
| freeing items | 0.000064 |
| logging slow query | 0.000020 |
| Opening tables | 0.000025 |
| System lock | 0.000146 |
| cleaning up | 0.000051 |
+----------------------+-----------+
23 rows in set, 1 warning (0.00 sec)
複製代碼
從上面的可看出時間主要用在 Sending data ,所謂的「Sending data」 並非單純的發送數據,而是包括「收集 + 發送 數據」。後來我又嘗試將查詢的列減小,可是沒有什麼效果。後來無心間檢查下表結構發現其中 有兩列projet_id和 theday的列類型是text,好像知道什麼了(不知道是誰設計的,可是其實沒有必要),改這兩列爲varchar類型,再執行一次查詢,只用了0.5秒。性能
自此問題解決。總結一下,設計表的時候,列類型必定要考慮好,text類型儘可能少用,設置爲varchar的時候,長度夠用就好,越短性能越好。ui
=============我是分割線======下面是追加的內容===================spa
查看錶結構:.net
mysql> desc tmp_table;
+-------------+--------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+-------------+--------------+------+-----+---------+-------+
| project_id | text | YES | MUL | NULL | |
| index_count | varchar(100) | YES | | NULL | |
| theday | text | YES | MUL | NULL | |
| app_version | varchar(32) | NO | | NULL | |
| channel | varchar(32) | YES | | NULL | |
+-------------+--------------+------+-----+---------+-------+
8 rows in set (0.00 sec)
複製代碼
修改列類型:設計
alter table 表名 MODIFY COLUMN 列名 VARCHAR(16)
複製代碼
對了,在處理的過程當中,我還嘗試了加索引,可是發現加索引後,查詢時間不快反而變的更慢。這裏有一篇關於索引的文章 blog.csdn.net/u014470581/…