Skip to content

Latest commit

 

History

History
252 lines (216 loc) · 19.6 KB

Explain执行计划详解.md

File metadata and controls

252 lines (216 loc) · 19.6 KB
SQL优化到底流程
  • explain 查看目前的索引信息
  • 查看目标sql数据分布
  • 将线上数据同步到本地库
  • 将你认为能提个速度的索引在内部环境实验，看是不是你想要的效果
  • 主要关注explain的rows、filtered、Extra字段，因为需要你优化的sql一般都是有了索引的，你要做的是怎么创建更高效的索引，比如减少回表次数、去掉filesort、减少零时表的生成等等 image
id

id： ：表示查询中执行select子句或者操作表的顺序，id的值越大，代表优先级越高，越先执行。

select_type

select_type：表示 select 查询的类型，主要是用于区分各种复杂的查询，例如：普通查询、联合查询、子查询等。

table

查询的表名，并不一定是真实存在的表，有别名显示别名，也可能为临时表，例如上边的DERIVED、 <union1,4>等。

partitions

查询时匹配到的分区信息，对于非分区表值为NULL，当查询的是分区表时，partitions显示分区表命中的分区情况。

+----+-------------+----------------+---------------------------------+-------+---------------+---------+---------+------+------+----------+-------------+
| id | select_type | table          | partitions                      | type  | possible_keys | key     | key_len | ref  | rows | filtered | Extra       |
+----+-------------+----------------+---------------------------------+-------+---------------+---------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | one            | p201801,p201802,p201803,p300012 | index | NULL          | PRIMARY | 9       | NULL |    3 |      100 | Using index |
+----+-------------+----------------+---------------------------------+-------+---------------+---------+---------+------+------+----------+-------------+
type

type：查询使用了何种类型，它在SQL优化中是一个非常重要的指标，以下性能从好到坏依次是：system > const > eq_ref > ref > ref_or_null > index_merge > unique_subquery > index_subquery > range > index > ALL

  • system：当表仅有一行记录时(系统表)，数据量很少，往往不需要进行磁盘IO，速度非常快。
  • const：表示查询时命中 primary key 主键或者 unique 唯一索引，或者被连接的部分是一个常量(const)值。这类扫描效率极高，返回数据量少，速度非常快
mysql> EXPLAIN SELECT * from three where three_id=1;
+----+-------------+-------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
| id | select_type | table | partitions | type  | possible_keys | key     | key_len | ref   | rows | filtered | Extra |
+----+-------------+-------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
|  1 | SIMPLE      | three | NULL       | const | PRIMARY       | PRIMARY | 4       | const |    1 |      100 | NULL  |
+----+-------------+-------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
  • eq_ref：查询时命中主键primary key 或者 unique key索引， type 就是 eq_ref。
mysql> EXPLAIN select o.one_name from one o ,two t where o.one_id = t.two_id ; 
+----+-------------+-------+------------+--------+---------------+----------+---------+--------------------+------+----------+-------------+
| id | select_type | table | partitions | type   | possible_keys | key      | key_len | ref                | rows | filtered | Extra       |
+----+-------------+-------+------------+--------+---------------+----------+---------+--------------------+------+----------+-------------+
|  1 | SIMPLE      | o     | NULL       | index  | PRIMARY       | idx_name | 768     | NULL               |    2 |      100 | Using index |
|  1 | SIMPLE      | t     | NULL       | eq_ref | PRIMARY       | PRIMARY  | 4       | xin-slave.o.one_id |    1 |      100 | Using index |
+----+-------------+-------+------------+--------+---------------+----------+---------+--------------------+------+----------+-------------+
  • ref：区别于eq_ref ，ref表示使用非唯一性索引，会找到很多个符合条件的行
mysql> select o.one_id from one o where o.one_name = "xin" ; 
+--------+
| one_id |
+--------+
|      1 |
|      3 |
+--------+```

```sql
mysql> EXPLAIN select o.one_id from one o where o.one_name = "xin" ; 
+----+-------------+-------+------------+------+---------------+----------+---------+-------+------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key      | key_len | ref   | rows | filtered | Extra       |
+----+-------------+-------+------------+------+---------------+----------+---------+-------+------+----------+-------------+
|  1 | SIMPLE      | o     | NULL       | ref  | idx_name      | idx_name | 768     | const |    1 |      100 | Using index |
+----+-------------+-------+------------+------+---------------+----------+---------+-------+------+----------+-------------+
  • index_merge：使用了索引合并优化方法，查询使用了两个以上的索引。下边示例中同时使用到主键one_id 和 字段one_name的idx_name 索引
mysql> EXPLAIN select * from one o where o.one_id >1 and o.one_name ='xin'; 
+----+-------------+-------+------------+-------------+------------------+------------------+---------+------+------+----------+------------------------------------------------+
| id | select_type | table | partitions | type        | possible_keys    | key              | key_len | ref  | rows | filtered | Extra                                          |
+----+-------------+-------+------------+-------------+------------------+------------------+---------+------+------+----------+------------------------------------------------+
|  1 | SIMPLE      | o     | NULL       | index_merge | PRIMARY,idx_name | idx_name,PRIMARY | 772,4   | NULL |    1 |      100 | Using intersect(idx_name,PRIMARY); Using where |
+----+-------------+-------+------------+-------------+------------------+------------------+---------+------+------+----------+------------------------------------------------+
  • unique_subquery：替换下面的 IN子查询，子查询返回不重复的集合
value IN (SELECT primary_key FROM single_table WHERE some_expr)
  • index_subquery：区别于unique_subquery，用于非唯一索引，可以返回重复值
value IN (SELECT key_column FROM single_table WHERE some_expr)
  • range：使用索引选择行，仅检索给定范围内的行。简单点说就是针对一个有索引的字段，给定范围检索数据。在where语句中使用 bettween...and、<、>、<=、in 等条件查询 type 都是 range。 举个栗子：three表中three_id为唯一主键，user_id普通字段未建索引
mysql> EXPLAIN SELECT * from three where three_id BETWEEN 2 AND 3;
+----+-------------+-------+------------+-------+---------------+---------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type  | possible_keys | key     | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-------+------------+-------+---------------+---------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | three | NULL       | range | PRIMARY       | PRIMARY | 4       | NULL |    1 |      100 | Using where |
+----+-------------+-------+------------+-------+---------------+---------+---------+------+------+----------+-------------+

从结果中看到只有对设置了索引的字段，做范围检索 type 才是 range

mysql> EXPLAIN SELECT * from three where user_id BETWEEN 2 AND 3;
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | three | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    3 |    33.33 | Using where |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
  • index：Index 与ALL 其实都是读全表，区别在于index是遍历索引树读取，而ALL是从硬盘中读取。下边示例：three_id 为主键，不带 where 条件全表查询 ，type结果为index
mysql> EXPLAIN SELECT three_id from three ;
+----+-------------+-------+------------+-------+---------------+---------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type  | possible_keys | key     | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-------+------------+-------+---------------+---------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | three | NULL       | index | NULL          | PRIMARY | 4       | NULL |    1 |      100 | Using index |
+----+-------------+-------+------------+-------+---------------+---------+---------+------+------+----------+-------------+
  • ALL：将遍历全表以找到匹配的行，性能最差。
possible_keys

表示在MySQL中通过哪些索引，能让我们在表中找到想要的记录，一旦查询涉及到的某个字段上存在索引，则索引将被列出，但这个索引并不定一会是最终查询数据时所被用到的索引

key

区别于possible_keys，key是查询中实际使用到的索引，若没有使用索引，显示为NULL。具体请参考上边的例子

区别于possible_keys，key是查询中实际使用到的索引，若没有使用索引，显示为NULL。具体请参考上边的例子

key_len

表示查询用到的索引长度（字节数），原则上长度越短越好 。

  • 单列索引，那么需要将整个索引长度算进去；
  • 多列索引，不是所有列都能用到，需要计算查询中实际用到的列。

注意：key_len只计算where条件中用到的索引长度，而排序和分组即便是用到了索引，也不会计算到key_len中。

ref

常见的有：const，func，null，字段名。

  • 当使用常量等值查询，显示const，
  • 当关联查询时，会显示相应关联表的关联字段
  • 如果查询条件使用了表达式、函数，或者条件列发生内部隐式转换，可能显示为func
  • 其他情况null
rows

以表的统计信息和索引使用情况，估算要找到我们所需的记录，需要读取的行数。

==这是评估SQL 性能的一个比较重要的数据==，mysql需要扫描的行数，很直观的显示 SQL 性能的好坏，一般情况下 rows 值越小越好

mysql> EXPLAIN SELECT * from three;
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------+
|  1 | SIMPLE      | three | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    3 |      100 | NULL  |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------+
filtered

这个是一个百分比的值，表里符合条件的记录数的百分比。简单点说，这个字段表示存储引擎返回的数据在经过过滤后，剩下满足条件的记录数量的比例。

在MySQL.5.7版本以前想要显示filtered需要使用explain extended命令。MySQL.5.7后，默认explain直接显示partitions和filtered的信息。

Extra

不适合在其他列中显示的信息，Explain 中的很多额外的信息会在 Extra 字段显示

  • Using index：我们在相应的select操作中使用了覆盖索引，通俗一点讲就是查询的列被索引覆盖，使用到覆盖索引查询速度会非常快，SQl优化中理想的状态
mysql> EXPLAIN SELECT one_id from one ;
+----+-------------+-------+------------+-------+---------------+------------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type  | possible_keys | key        | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-------+------------+-------+---------------+------------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | one   | NULL       | index | NULL          | idx_two_id | 5       | NULL |    3 |      100 | Using index |
+----+-------------+-------+------------+-------+---------------+------------+---------+------+------+----------+-------------+

注意：想要使用到覆盖索引，我们在 select 时只取出需要的字段，不可select *，而且该字段建了索引。

  • Using where：查询时未找到可用的索引，进而通过where条件过滤获取所需数据，但要注意的是并不是所有带where语句的查询都会显示Using where。下边示例create_time 并未用到索引，type 为 ALL，即MySQL通过全表扫描后再按where条件筛选数据
mysql> EXPLAIN SELECT one_name from one where create_time ='2020-05-18';
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | one   | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    3 |    33.33 | Using where |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
  • Using temporary：表示查询后结果需要使用临时表来存储，一般在排序或者分组查询时用到。
mysql> EXPLAIN SELECT one_name from one where one_id in (1,2) group by one_name;
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | one   | NULL       | range| NULL          | NULL | NULL    | NULL |    3 |    33.33 | Using where; Using temporary; Using filesort |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
  • Using filesort：表示无法利用索引完成的排序操作，也就是ORDER BY的字段没有索引，通常这样的SQL都是需要优化的
mysql> EXPLAIN SELECT one_id from one  ORDER BY create_time;
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra          |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------+
|  1 | SIMPLE      | one   | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    3 |      100 | Using filesort |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------+

如果ORDER BY字段有索引就会用到覆盖索引，相比执行速度快很多

  • Using join buffer：在我们联表查询的时候，如果表的连接条件没有用到索引，需要有一个连接缓冲区来存储中间结果，先看一下有索引的情况：连接条件 one_name 、two_name 都用到索引
mysql> EXPLAIN SELECT one_name from one o,two t where o.one_name = t.two_name;
+----+-------------+-------+------------+-------+---------------+----------+---------+----------------------+------+----------+--------------------------+
| id | select_type | table | partitions | type  | possible_keys | key      | key_len | ref                  | rows | filtered | Extra                    |
+----+-------------+-------+------------+-------+---------------+----------+---------+----------------------+------+----------+--------------------------+
|  1 | SIMPLE      | o     | NULL       | index | idx_name      | idx_name | 768     | NULL                 |    3 |      100 | Using where; Using index |
|  1 | SIMPLE      | t     | NULL       | ref   | idx_name      | idx_name | 768     | xin-slave.o.one_name |    1 |      100 | Using index              |
+----+-------------+-------+------------+-------+---------------+----------+---------+----------------------+------+----------+--------------------------+

接下来删掉 连接条件 one_name 、two_name 的字段索引。发现Extra 列变成 Using join buffer，type均为全表扫描，这也是SQL优化中需要注意的地方

mysql> EXPLAIN SELECT one_name from one o,two t where o.one_name = t.two_name;
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------------------------------------------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra                                              |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------------------------------------------+
|  1 | SIMPLE      | t     | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    2 |      100 | NULL                                               |
|  1 | SIMPLE      | o     | NULL       | ALL  | NULL          | NULL | NULL    | NULL |    3 |    33.33 | Using where; Using join buffer (Block Nested Loop) |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------------------------------------------+
  • Impossible where：表示在我们用不太正确的where语句，导致没有符合条件的行。
mysql> EXPLAIN SELECT one_name from one WHERE 1=2;
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+------------------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra            |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+------------------+
|  1 | SIMPLE      | NULL  | NULL       | NULL | NULL          | NULL | NULL    | NULL | NULL | NULL     | Impossible WHERE |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+------------------+
  • No tables used：我们的查询语句中没有FROM子句，或者有 FROM DUAL子句
mysql> EXPLAIN select now();
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------+
| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref  | rows | filtered | Extra          |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------+
|  1 | SIMPLE      | NULL  | NULL       | NULL | NULL          | NULL | NULL    | NULL | NULL | NULL     | No tables used |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------+

详见 MySQL官方文档