mysql倒排字段类型_mysql倒排的优化

今天数据库负载就直线上升，数据库连接数撑爆。把语句抓出来一看，罪魁祸首是一条很简单的语句：SELECT * FROM eload_promotion_code WHERE 1 AND exp_time<1478782591 AND cishu=0 order by id desc limit 454660,20; 二话不说先把这个语句kill了，然后慢慢看怎么优化。

先看一下这个表的索引：

>show index from eload_promotion_code\G

*************************** 1. row ***************************

Table: eload_promotion_code

Non_unique: 0

Key_name: PRIMARY

Seq_in_index: 1

Column_name: id

Collation: A

Cardinality: 921642

Sub_part: NULL

Packed: NULL

Null:

Index_type: BTREE

Comment:

Index_comment:

*************************** 2. row ***************************

Table: eload_promotion_code

Non_unique: 1

Key_name: idx_cishu_exp

Seq_in_index: 1

Column_name: cishu

Collation: A

Cardinality: 15

Sub_part: NULL

Packed: NULL

Null:

Index_type: BTREE

Comment:

Index_comment:

*************************** 3. row ***************************

Table: eload_promotion_code

Non_unique: 1

Key_name: idx_cishu_exp

Seq_in_index: 2

Column_name: exp_time

Collation: A

Cardinality: 921642

Sub_part: NULL

Packed: NULL

Null:

Index_type: BTREE

Comment:

Index_comment:

可以看到id为主键，idx_cishu_exp为(cishu,exp_time)的唯一索引

看一下这个语句的执行计划，可以看到排序没有用到索引

explain SELECT * FROM eload_promotion_code WHERE 1 AND exp_time<1478782591 AND cishu=0 order by id desc limit 454660,20\G

*************************** 1. row ***************************

id: 1

select_type: SIMPLE

table: eload_promotion_code

type: ref

possible_keys: idx_cishu_exp

key: idx_cishu_exp

key_len: 4

ref: const

rows: 460854

Extra: Using where; Using filesort

1 row in set (0.00 sec)

将select * 换成select id后再看执行计划，可以用索引覆盖

>explain select id FROM eload_promotion_code WHERE 1 AND exp_time<1478782591 AND cishu=0 order by id desc limit 454660,20 \G

*************************** 1. row ***************************

id: 1

select_type: SIMPLE

table: eload_promotion_code

type: range

possible_keys: idx_cishu_exp

key: idx_cishu_exp

key_len: 8

ref: NULL

rows: 460862

Extra: Using where; Using index; Using filesort

1 row in set (0.00 sec)

好吧，这个语句有救了，采用延时关联先取出id,然后根据id获取原表所需要的行，改写后的语句来了：select * from eload_promotion_code inner join (select id from eload_promotion_code where exp_time<1478782591 AND cishu=0 order by id desc limit 454660,20) as x on eload_promotion_code.id=x.id;

执行一下，0.3s出结果。

这样就算完了。