MySQL 优化 01 之慢查询

最近线上生产环境中有些 MySQL 的语句特别慢,跑一个查询需要大概 5 分钟,已经到了不得不解决的地步了。

案例一:单表复杂语句的聚合计算

问题描述

单表数据超过 1500 万,写入不是特别频繁,查询很多,需要对表中订单的费用的进行实时计算,主要是基于时间以及订单状态进行运算。

1
2
3
SELECT count(order.id) AS order_count, sum(order.original_total_fee) AS original_total_fee, sum(order.settlement_fee) AS settlement_fee, sum(order.total_fee) AS total_fee, order.seller_id AS order_seller_id
FROM order
WHERE order.create_time < '2018-11-10 00:00:00' and create_time > '2016-10-10' AND order.order_status IN ('paid', 'refund_refused') AND order.partner_id IN (105203908, 105203916, 105203958, 105203910) AND order.seller_id = 96691 AND order.user_id NOT IN (105203908, 105203916, 105203958, 105203910) GROUP BY order.seller_id;

表中有以下索引:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246

{
"data":
[
{
"Table": "order",
"Non_unique": 0,
"Key_name": "PRIMARY",
"Seq_in_index": 1,
"Column_name": "id",
"Collation": "A",
"Cardinality": 16500090,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "order_order_id_device_id",
"Seq_in_index": 1,
"Column_name": "order_id",
"Collation": "A",
"Cardinality": 16500090,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "order_order_id_device_id",
"Seq_in_index": 2,
"Column_name": "device_id",
"Collation": "A",
"Cardinality": 16500090,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "order_device_id",
"Seq_in_index": 1,
"Column_name": "device_id",
"Collation": "A",
"Cardinality": 4125022,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "order_device_id",
"Seq_in_index": 2,
"Column_name": "order_status",
"Collation": "A",
"Cardinality": 4125022,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "order_device_id",
"Seq_in_index": 3,
"Column_name": "order_id",
"Collation": "A",
"Cardinality": 16500090,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "order_seller_id_order_name",
"Seq_in_index": 1,
"Column_name": "seller_id",
"Collation": "A",
"Cardinality": 18,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "order_seller_id_order_name",
"Seq_in_index": 2,
"Column_name": "order_name",
"Collation": "A",
"Cardinality": 5125,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "order_seller_id_create_time",
"Seq_in_index": 1,
"Column_name": "seller_id",
"Collation": "A",
"Cardinality": 5899,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "order_seller_id_create_time",
"Seq_in_index": 2,
"Column_name": "create_time",
"Collation": "A",
"Cardinality": 16500090,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "idx_order_status_seller_id_order_id",
"Seq_in_index": 1,
"Column_name": "order_status",
"Collation": "A",
"Cardinality": 18,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "idx_order_status_seller_id_order_id",
"Seq_in_index": 2,
"Column_name": "seller_id",
"Collation": "A",
"Cardinality": 31548,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "idx_order_status_seller_id_order_id",
"Seq_in_index": 3,
"Column_name": "order_id",
"Collation": "A",
"Cardinality": 16500090,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "idx_seller_id_user_id_create_time",
"Seq_in_index": 1,
"Column_name": "seller_id",
"Collation": "A",
"Cardinality": 18,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "idx_seller_id_user_id_create_time",
"Seq_in_index": 2,
"Column_name": "user_id",
"Collation": "A",
"Cardinality": 8250045,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order",
"Non_unique": 1,
"Key_name": "idx_seller_id_user_id_create_time",
"Seq_in_index": 3,
"Column_name": "create_time",
"Collation": "A",
"Cardinality": 16500090,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
}
]
}

这个 SQL 语句每次执行时间因需要计算的数据大小而异,平均时间再 1 分钟,更长会达到 3 分钟或更久。

问题分析

这个表已经有不少索引了,先不说这些索引的合理性,单从上面的语句来看,应该尽量使用现存的索引来优化查询,使用 explain 语句分析上面的查询输出如下:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

{
"data":
[
{
"id": 1,
"select_type": "SIMPLE",
"table": "order",
"type": "range",
"possible_keys": "order_seller_order_name,order_seller_id_create_time,idx_order_status_seller_id_order_id,idx_seller_id_user_id_create_time",
"key": "order_seller_id_create_time",
"key_len": "12",
"ref": null,
"rows": 22764,
"Extra": "Using where"
}
]
}

可以看到上面的语句使用了 order_seller_id_create_time 这个索引,这个不难理解,因为 where 子句中用了 seller_id 的等值以及 create_time 的范围,符合最左前缀的要求,按理来说查询应该不是很慢,rows 显示也就只有几万条数据,不算太大,尝试尽量减少数据范围,继续观察对计算时间的影响,做了以下几种修改:

缩小 create_time 时间范围

只有当数据明显减少之后计算时间才会缩短

简化 where 子句的条件

去掉 user_id 和 partner_id 的删选,尽量简化 where 子句的删选,索引选择依然没有变,但是如果不缩小数据范围,依然很慢。

在做了以上几种尝试之后,初步能得出结论:慢不在于查询,而在于聚合运算

既然索引本身无助于聚合运算,尝试对数据库硬件进行升级,来保证 MySQL 有充分的资源进行计算,虽然查了一圈没有找到特别详细的描述 MySQL 做聚合运算的原理,猜测肯定会用到不少资源。

方案尝试

方案一

原有 MySQL 实例用的是云,6G 内存,ssd 的实例,直接增加内存到 12G,然后直行上面的运算,并没有显著改善

方案二

修改 innodb_buffer_pool_size 大小从原来的 4G 到 8G,并且更改 innodb_buffer_pool_instances 为 2,让每个 buffer_pool 分配 4G 内存,然后直行上面的计算,速度得到显著改善。

方案总结

提升硬件资源有利于 MySQL 执行聚合运算,但是最终是有利于 MySQL 把查询索引都缓存了改善了查询,还是改善了聚合计算所需要的资源在这个场景之下都不太好能得出结论,猜测二者都有可能。

innodb_buffer_pool_size 是 MySQL 为执行 MySQL 查询计算所预留的一个缓存区,本质上是越大越好,因为这样的话,MySQL 会把大部分运算直接在内存中执行,innodb_buffer_pool_instances 是为了在并发大的时候提升并发读和并发写的,单个 innodb_buffer_pool_size 的大小是 innodb_buffer_pool_size/innodb_buffer_pool_instances,官方建议 innodb_buffer_pool_size 最小不要小于 1G。

案例二:大表 join 查询

问题描述

两个大表 join 查询,大表 2000 万以上,小表 1500 万以上

1
2
3
4
SELECT order_base.id AS order_base_id, order_base.user_id AS order_base_user_id
FROM order_base INNER JOIN order ON order_base.id = order.order_id
WHERE order.seller_id = 96691 AND order_base.paid_time >= '2015-10-03T16:43:07' AND order_base.paid_time < '2018-11-12T16:43:07' AND order_base.paid = 1 AND order.partner_id NOT IN (105203908, 105203916, 105203958, 105203910) ORDER BY order_base.order_id DESC
LIMIT 10;

order_base 索引如下:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
{
"data":
[
{
"Table": "order_base",
"Non_unique": 0,
"Key_name": "PRIMARY",
"Seq_in_index": 1,
"Column_name": "id",
"Collation": "A",
"Cardinality": 20935372,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order_base",
"Non_unique": 0,
"Key_name": "order_id",
"Seq_in_index": 1,
"Column_name": "order_id",
"Collation": "A",
"Cardinality": 20935372,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order_base",
"Non_unique": 1,
"Key_name": "uuid",
"Seq_in_index": 1,
"Column_name": "uuid",
"Collation": "A",
"Cardinality": 20935372,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order_base",
"Non_unique": 1,
"Key_name": "idx_order_base_user_id_id",
"Seq_in_index": 1,
"Column_name": "user_id",
"Collation": "A",
"Cardinality": 6978457,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order_base",
"Non_unique": 1,
"Key_name": "idx_order_base_user_id_id",
"Seq_in_index": 2,
"Column_name": "id",
"Collation": "A",
"Cardinality": 20935372,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order_base",
"Non_unique": 1,
"Key_name": "idx_order_base_user_id_order_id",
"Seq_in_index": 1,
"Column_name": "user_id",
"Collation": "A",
"Cardinality": 20935372,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order_base",
"Non_unique": 1,
"Key_name": "idx_order_base_user_id_order_id",
"Seq_in_index": 2,
"Column_name": "order_id",
"Collation": "A",
"Cardinality": 20935372,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order_base",
"Non_unique": 1,
"Key_name": "idx_order_base_user_id_create_time",
"Seq_in_index": 1,
"Column_name": "user_id",
"Collation": "A",
"Cardinality": 10467686,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order_base",
"Non_unique": 1,
"Key_name": "idx_order_base_user_id_create_time",
"Seq_in_index": 2,
"Column_name": "create_time",
"Collation": "A",
"Cardinality": 20935372,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
},
{
"Table": "order_base",
"Non_unique": 1,
"Key_name": "idx_order_base_paid_time",
"Seq_in_index": 1,
"Column_name": "paid_time",
"Collation": "A",
"Cardinality": 20935372,
"Sub_part": null,
"Packed": null,
"Null": "",
"Index_type": "BTREE",
"Comment": "",
"Index_comment": ""
}
]
}

order 表索引同 案例一

如果最终没有数据满足条件,这个 SQL 语句非常慢,如果有数据满足条件,速度也不算特别理想。

问题分析

使用 explain 语句可以看到如下输出:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
{
"data":
[
{
"id": 1,
"select_type": "SIMPLE",
"table": "order",
"type": "index",
"possible_keys": "order_order_id_device_id,order_seller_id_order_name,seller_id_create_time,idx_seller_id_user_id_create_time",
"key": "order_order_id_device_id",
"key_len": "12",
"ref": null,
"rows": 473,
"Extra": "Using where"
},
{
"id": 1,
"select_type": "SIMPLE",
"table": "order_base",
"type": "eq_ref",
"possible_keys": "PRIMARY,idx_order_base_paid_time",
"key": "PRIMARY",
"key_len": "4",
"ref": "martin_order.order.order_id",
"rows": 1,
"Extra": "Using where"
}
]
}

关于 explain 语句的输出,详见 MySQL explain 详解,这里可以明显看到表 order 中的 ref 是 null,type 是 index,也就是是一个索引扫描,也就是说 order 表可能使用了一个不合理的索引 order_order_id_device_id 从而造成了 order 需要进行全索引扫描,这是非常糟糕的索引选择类型,这有可能是造成查询慢的原因之一

表 order_base 用的 type 是 eq_ref,意味着 order_base 使用的是唯一性索引或者 primary key,从上不能看出是 primary key,性能较好。

方案尝试

order 表有很多索引,在这个 join 语句的 where 子句中 order 表主要使用了 seller_id 和 partner_id,已经存在的索引中只有 order_seller_id_create_time 满足场景,因而可以在 join 的时候强制表 order 走这个索引:

1
2
3
4
SELECT order_base.id AS order_base_id, order_base.user_id AS order_base_user_id
FROM order_base INNER JOIN order force index(order_seller_id_create_time) ON order_base.id = order.order_id
WHERE order.seller_id = 96691 AND order_base.paid_time >= '2015-10-03T16:43:07' AND order_base.paid_time < '2018-11-12T16:43:07' AND order_base.paid = 1 AND order.partner_id NOT IN (105203908, 105203916, 105203958, 105203910) ORDER BY order_base.order_id DESC
LIMIT 10;

再使用 explain 输出如下:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
{
"data":
[
{
"id": 1,
"select_type": "SIMPLE",
"table": "order",
"type": "ref",
"possible_keys": "order_seller_id_create_time",
"key": "order_seller_id_create_time",
"key_len": "4",
"ref": "const",
"rows": 348810,
"Extra": "Using where; Using filesort"
},
{
"id": 1,
"select_type": "SIMPLE",
"table": "order_base",
"type": "eq_ref",
"possible_keys": "PRIMARY,idx_order_base_paid_time",
"key": "PRIMARY",
"key_len": "4",
"ref": "martin_order.order.order_id",
"rows": 1,
"Extra": "Using where"
}
]
}

可以看到表 order 的 type 已经是 ref,表示使用了一个非唯一性索引或者是满足最左前缀的索引,而且 ref 已经成了 const,是非常好的索引选择。

在 join 查询中,需要 match 的总行数是各个表 explain output 中 rows 的乘机,虽然说更改之后表 order 的 rows 明显增多,但是能够合理使用索引,还是明显改善查询速度,优化后的 SQL 语句基本是秒查。

update

上述优化对于 order_base.paid = 1 这个条件删选出的数据如果比较少时确实很快,10 万以下,但是如果非常多还是很慢,后来发现 order 表也有一个类似的 paid 状态,相对来说数据要少很多,换成 order.order_status='paid' 会有显著改善,也就是说上述 SQL 语句在不同的查询量级的情况下使用不同的方案更合适。

小结

MySQL 的慢查询大多都集中在索引使用不合理,对 MySQL 的索引原理要有深入理解才能准确定位问题,而且 MySQL 的性能问题是多方面的,有硬件也有数据的影响,要多方面综合考虑。

三月沙 wechat
扫描关注 wecatch 的公众号