0%

体系结构

连接层 -> 服务层 -> 引擎层 —> 存储层

存储引擎

存储引擎指存储数据,建立索引,更新/查询数据等技术的实现方式。
存储引擎是基于表的,不是基于库的。

查某个表的存储引擎

1
show create table 表名;

查数据库支持的存储引擎

1
show engines;

创建表时指定引擎

1
2
3
4
5
create table a(
)engine = MyISAM;

create table b(
)engine = InnoDB;

InnoDB

  • 支持事务,行级锁,外键
  • 每张表对应一个表空间文件XXX.ibd,路径:/var/lib/mysql/*.ibd
  • 逻辑存储结构: TableSpace(表空间), Segment(段), Extent(区), Page(页), Row(行)

MyISAM

MySQL早期的默认存储引擎

  • 不支持事务,不支持外键
  • 支持表锁,不支持行锁
  • XX.sdi 存储表结构信息, XX.MYD 存储数据 XX.MYI存储索引

Memory

表数据存储在内存中,只能将这些表作为临时表或缓存使用

  • 内存存放
  • hash索引(默认)
  • XX.sdi 存储表结构

索引

索引(index)是帮助MySQL高效查询数据的数据结构。
优缺点:提高查询效率,但索引本身占用空间,且降低更新表(INSERT,UPDATE,DELETE)的速度。

索引结构

索引结构 描述
B+Tree索引 最常见索引, 大部分引擎都支持B+树索引
Hash索引 底层用哈希表实现,只有精确匹配索引列的查询才有效,不支持范围查询和排序
R-tree(空间索引) MyISAM的一个特殊索引类型,主要用于地理空间数据类型,通常使用较少
Full-text(全文索引) 一种通过建立倒排索引,快速匹配文档的方式,类似与Lucene,Solr,ES

为什么InnoDB选择B+树,不用二叉树,B树

  • 二叉树:顺序插入时,退化成链表,查询性能低;层级较深,查询速度慢。
  • B-Tree: 叶子节点和非叶子节点都保存数据,查询不稳定;相比B+树IO消耗大
  • B+Tree: 只有叶子节点存储数据,使得树更矮,减少IO操作次数;所有叶子节点构成一个有序链表
    MySQL在原B+Tree基础上,增加一个指向相邻叶子节点的链表指针,提高区间访问性能。

索引分类

分类 含义 特点 关键字
主键索引 针对于表中主键创建的索引 默认自动创建,只有一个 PRIMARY
唯一索引 避免同一个表中某数据列的重复 可以有多个 UNIQUE
常规索引 快速定位特定数据 可以有多个
全文索引 查找文本中关键词 可以有多个 FULLTEXT

在InnoDB中,根据索引的存储形式,又可以分如下两种:

分类 含义 特点
聚集索引(Clustered Index) 将数据和索引一起存储,叶子节点保存了行数据 有且只有一个
二级索引(Secondary Index) 将数据和索引分开存储,叶子节点关联对应主键 可存在多个

聚集索引选取规则:

  • 如果存在主键,主键索引就是聚集索引
  • 如果不存在主键,使用第一个唯一索引作为聚集索引
  • 如没有主键和合适的唯一索引,InnoDB会自动生成一个rowid作为隐藏的聚集索引

索引语法

创建索引

1
create [unique|fulltext] index index_name on table_name (index_col_name,...);

查看索引

1
show index from table_name;

删除索引

1
drop index index_name on table_name;

例:按如下要求完成索引的创建

1
2
3
4
5
6
7
8
9
10
11
create table tb_user(
id bigint auto_increment primary key,
name varchar(10),
phone char(11),
email varchar(36),
profession varchar(36),
age tinyint unsigned,
gender char(1),
status int,
createtime datetime
)engine=InnoDB;

要求:

  • name设置索引
  • phone非空创建唯一索引
  • profession,age,status设置联合索引
  • email建立索引

创建索引

1
2
3
4
create index idx_user_name on tb_user(name);
create unique index idx_user_phone on tb_user(phone);
create index idx_user_pro_age_sta on tb_user(profession,age,status);
create index idx_user_email on tb_user(email);

查看索引

1
2
3
4
5
6
7
8
9
10
11
12
mysql> show index from tb_user;
+---------+------------+----------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+---------+------------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment | Visible | Expression |
+---------+------------+----------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+---------+------------+
| tb_user | 0 | PRIMARY | 1 | id | A | 0 | NULL | NULL | | BTREE | | | YES | NULL |
| tb_user | 0 | idx_user_phone | 1 | phone | A | 0 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_name | 1 | name | A | 0 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_pro_age_sta | 1 | profession | A | 0 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_pro_age_sta | 2 | age | A | 0 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_pro_age_sta | 3 | status | A | 0 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_email | 1 | email | A | 0 | NULL | NULL | YES | BTREE | | | YES | NULL |
+---------+------------+----------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+---------+------------+

SQL性能分析

查询SQL执行频率

通过如下命令,查看SQL中增删改查的频率

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
mysql> show global status like 'Com_______';
+---------------+-------+
| Variable_name | Value |
+---------------+-------+
| Com_binlog | 0 |
| Com_commit | 0 |
| Com_delete | 0 |
| Com_import | 0 |
| Com_insert | 0 |
| Com_repair | 0 |
| Com_revoke | 0 |
| Com_select | 4 |
| Com_signal | 0 |
| Com_update | 0 |
| Com_xa_end | 0 |
+---------------+-------+

慢查询日志

记录所有执行时间超过指定参数long_query_time(默认:10秒)的所有SQL语句日志

MySQL的慢查询日志默认没有开,需要手动配置

1
2
3
4
5
6
7
mysql> show variables like 'slow_query%';
+---------------------+-----------------------------------+
| Variable_name | Value |
+---------------------+-----------------------------------+
| slow_query_log | OFF |
| slow_query_log_file | /var/lib/mysql/localhost-slow.log |
+---------------------+-----------------------------------+

修改/etc/my.cnf.d/mysql-server.cnf, 在[mysqld]最后添加两行,开启MySQL慢查询日志,设置慢查询超时时间2秒

1
2
slow-query-log=on
long_query_time=2

重启mysqld, 执行systemctl restart mysqld,再确认修改后的配置已生效

1
2
3
4
5
6
7
8
9
10
11
12
mysql> show variables like 'slow_query%';
+---------------------+-----------------------------------+
| Variable_name | Value |
+---------------------+-----------------------------------+
| slow_query_log | ON |
+---------------------+-----------------------------------+
mysql> show variables like 'long_query_time';
+-----------------+----------+
| Variable_name | Value |
+-----------------+----------+
| long_query_time | 2.000000 |
+-----------------+----------+

profile详情

先打开profile开关

1
2
3
4
5
6
7
8
9
mysql> select @@profiling;
+-------------+
| @@profiling |
+-------------+
| 0 |
+-------------+
1 row in set, 1 warning (0.00 sec)

mysql> set profiling = 1;

执行SQL操作后,通过show profiles查看指令的执行耗时

1
2
3
4
5
6
7
8
9
mysql> show profiles;
+----------+------------+------------------------------+
| Query_ID | Duration | Query |
+----------+------------+------------------------------+
| 1 | 0.00027175 | select * from tb_user |
| 2 | 0.00008450 | show profiles for query 1 |
+----------+------------+------------------------------+
mysql> show profile for query 1;
mysql> show profile cpu for query 1;

explain

1
explain select * from tb_user where id = 1;

实例:给tb_user表生成100w条数据,用于测试查询性能

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
#!/usr/bin/env python3
import random
from datetime import datetime, timedelta
import string

#insert into tb_user(name,phone,email,profession,age,gender,status,createtime) values('Jony','12345678111', 'example@163.com', 'RD',18,'M',1,NOW());

def generate_phone():
return '1' + ''.join([str(random.randint(0, 9)) for _ in range(10)])

def generate_random_date():
start_date = datetime(2024, 1, 1)
end_date = datetime(2024, 12, 31)
time_between_dates = end_date - start_date
days_between_dates = time_between_dates.days
random_number_of_days = random.randint(0, days_between_dates)
random_date = start_date + timedelta(days=random_number_of_days)
return random_date.strftime('%Y-%m-%d')

def generate_random_name(length=5):
letters = string.ascii_letters
return ''.join(random.choice(letters) for _ in range(length))

def insert_data():
with open('./test.sql', 'w') as f:
for i in range(1000000):
name = generate_random_name()
phone = generate_phone()
email = generate_random_name() + '@163.com'
profession = random.choice(['Sales', 'R&D', 'Marketing', 'OPS'])
age = random.randint(0,130)
gender = random.choice(['F','M'])
status = random.choice([0, 1])
date = generate_random_date()
output = "insert into tb_user(name,phone,email,profession,age,gender,status,createtime) values('{name}','{phone}', '{email}', '{profession}',{age},'{gender}', {status}, '{date}');".format(name=name, phone=phone, email=email, profession=profession, age=age, gender=gender, status=status, date=date)
f.write(output + '\n')

insert_data()

再从SQL文件中批量导入tb_user表的数据到数据库pc

1
mysql -u root -p -D pc < test.sql 

explain执行计划中各字段含义

1
2
3
4
5
6
mysql> explain select name from tb_user where id = 1;
+----+-------------+---------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | const | PRIMARY | PRIMARY | 8 | const | 1 | 100.00 | NULL |
+----+-------------+---------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
  • id
    表示查询中执行select子句或操作表的顺序,id相同执行顺序从上到下,id不同,值越大越先执行
  • select_type
    表示select类型,常见取值有simple, primary, union, subquery(有子查询)
  • type
    表示连接类型,性能由好到差为NULL, system, const, eq_ref, ref, range, index, all
  • possible_keys
    显示可能应用在这张表的索引,一个或多个
  • key
    表示实际用到的索引,如果为NULL说明没使用索引
  • key_len
    表示索引中字节数,该值为索引字段最大可能长度,并非实际使用长度,长度越短越好
  • rows
    MySQL认为必须要执行查询的行数,在InnoDB表中,是一个估计值,可能并不总是准确的
  • filtered
    表示返回结果的行数占需读取行数百分比,filtered值越大越好

例: explain type = NULL (没有查表)

1
2
3
4
5
6
mysql> explain select 'A';
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------+
| 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 |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+----------------+

例: explain type = const (走主键索引/唯一索引)

1
2
3
4
5
6
mysql> explain select name from tb_user where id = 1;
+----+-------------+---------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+-------+---------------+---------+---------+-------+------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | const | PRIMARY | PRIMARY | 8 | const | 1 | 100.00 | NULL |
+----+-------------+---------+------------+-------+---------------+---------+---------+-------+------+----------+-------+

例: explain type = ref (走普通的非唯一索引)

1
2
3
4
5
6
mysql> explain select name from tb_user where name = 'uXrBh';
+----+-------------+---------+------------+------+---------------+---------------+---------+-------+------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+---------------+---------------+---------+-------+------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_name | idx_user_name | 43 | const | 1 | 100.00 | Using index |
+----+-------------+---------+------------+------+---------------+---------------+---------+-------+------+----------+-------------+

索引应用

最左前缀法则

  • 对于索引了多列的场景,最左前缀法则指查询从索引的最左列开始,且不跳过索引中的列
  • 如果跳跃某一列,索引将部分失效(后面字段索引失效)

例:先查看tb_user表的索引

1
2
3
4
5
6
7
8
9
10
11
12
mysql> show index from tb_user;
+---------+------------+----------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+---------+------------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment | Visible | Expression |
+---------+------------+----------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+---------+------------+
| tb_user | 0 | PRIMARY | 1 | id | A | 245768 | NULL | NULL | | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_name | 1 | name | A | 269997 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_pro_age_sta | 1 | profession | A | 3 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_pro_age_sta | 2 | age | A | 535 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_pro_age_sta | 3 | status | A | 1051 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_phone | 1 | phone | A | 271498 | NULL | NULL | YES | BTREE | | | YES | NULL |
| tb_user | 1 | idx_user_email | 1 | email | A | 269341 | NULL | NULL | YES | BTREE | | | YES | NULL |
+---------+------------+----------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+---------+------------+

例: 符合最左前缀法则,走联合索引(profession,age,status)

1
2
3
4
5
6
mysql> explain select * from tb_user where profession = 'R&D' and age = 31 and status = 1;
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro_age_sta | idx_user_pro_age_sta | 154 | const,const,const | 261 | 100.00 | NULL |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------+

例:查询条件改成只有profession,仍然走联合索引(profession,age,status)

1
2
3
4
5
6
mysql> explain select * from tb_user where profession = 'R&D';
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------+--------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------+--------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro_age_sta | idx_user_pro_age_sta | 147 | const | 127438 | 100.00 | NULL |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------+--------+----------+-------+

例:查询条件改成age, status,索引失效,走了全表扫描

1
2
3
4
5
6
mysql> explain select * from tb_user where age = 31 and status = 1;
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | ALL | NULL | NULL | NULL | NULL | 271498 | 1.00 | Using where |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-------------+

例:查询条件改成profession, status,索引部分失效,只有profession走了索引

1
2
3
4
5
6
mysql> explain select * from tb_user where profession = 'R&D' and status = 1;
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------+--------+----------+-----------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------+--------+----------+-----------------------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro_age_sta | idx_user_pro_age_sta | 147 | const | 127438 | 10.00 | Using index condition |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------+--------+----------+-----------------------+

例:查询条件改成age, status, profession, 走联合索引(profession,age,status) (只要存在即可,和语句中的先后位置无关)

1
2
3
4
5
6
explain select * from tb_user where age = 31 and status = 1 and profession = 'R&D';
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro_age_sta | idx_user_pro_age_sta | 154 | const,const,const | 261 | 100.00 | NULL |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------+

范围查询

联合索引中,出现范围查询,则范围查询右侧的列索引失效

例: age使用了范围查询,status没有走索引

1
2
3
4
5
6
mysql> explain select * from tb_user where profession = '软件工程' and age > 30 and status = '0';
+----+-------------+---------+------------+-------+----------------------+----------------------+---------+------+------+----------+-----------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+-------+----------------------+----------------------+---------+------+------+----------+-----------------------+
| 1 | SIMPLE | tb_user | NULL | range | idx_user_pro_age_sta | idx_user_pro_age_sta | 149 | NULL | 1 | 10.00 | Using index condition |
+----+-------------+---------+------------+-------+----------------------+----------------------+---------+------+------+----------+-----------------------+

例:把 age > 30 改成 age >= 30,都可以走索引 (业务允许情况下,尽量用>=代替>)

1
2
3
4
5
6
mysql> explain select * from tb_user where profession = '软件工程' and age >= 30 and status = '0';
+----+-------------+---------+------------+-------+----------------------+----------------------+---------+------+------+----------+-----------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+-------+----------------------+----------------------+---------+------+------+----------+-----------------------+
| 1 | SIMPLE | tb_user | NULL | range | idx_user_pro_age_sta | idx_user_pro_age_sta | 154 | NULL | 1 | 10.00 | Using index condition |
+----+-------------+---------+------------+-------+----------------------+----------------------+---------+------+------+----------+-----------------------+

在索引列上进行运算,索引会失效

例: 对phone做substring运算,发现索引失效, type = ALL

1
2
3
4
5
6
7
8
9
10
11
12
13
14
mysql> explain select * from tb_user where phone = '10931949622';
+----+-------------+---------+------------+------+----------------+----------------+---------+-------+------+----------+-----------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------+----------------+---------+-------+------+----------+-----------------------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_phone | idx_user_phone | 45 | const | 1 | 100.00 | Using index condition |
+----+-------------+---------+------------+------+----------------+----------------+---------+-------+------+----------+-----------------------+
1 row in set, 1 warning (0.00 sec)

mysql> explain select * from tb_user where substring(phone,10,2) = '22';
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | ALL | NULL | NULL | NULL | NULL | 271498 | 100.00 | Using where |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-------------+

字符串不加引号

字符串类型字段使用时,不加引号,索引将失效。

例:查询手机号时,忘了加引号,也能查出结果,但索引失效

1
2
3
4
5
6
mysql> explain select * from tb_user where phone = 10931949622;
+----+-------------+---------+------------+------+----------------+------+---------+------+--------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------+------+---------+------+--------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | ALL | idx_user_phone | NULL | NULL | NULL | 271498 | 10.00 | Using where |
+----+-------------+---------+------------+------+----------------+------+---------+------+--------+----------+-------------+

头部模糊匹配,索引失效

例: like查询手机号。使用头部模糊匹配’%262’,索引失效;尾部模糊匹配’109%’,索引有效

1
2
3
4
5
6
7
8
9
10
11
12
13
14
mysql> explain select * from tb_user where phone like '109%';
+----+-------------+---------+------------+-------+----------------+----------------+---------+------+------+----------+-----------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+-------+----------------+----------------+---------+------+------+----------+-----------------------+
| 1 | SIMPLE | tb_user | NULL | range | idx_user_phone | idx_user_phone | 45 | NULL | 2713 | 100.00 | Using index condition |
+----+-------------+---------+------------+-------+----------------+----------------+---------+------+------+----------+-----------------------+
1 row in set, 1 warning (0.00 sec)

mysql> explain select * from tb_user where phone like '%622';
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | ALL | NULL | NULL | NULL | NULL | 271498 | 11.11 | Using where |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-------------+

or连接的条件

用or分割开的条件,如果or前条件中列有索引,or后条件没有索引,查询不会走索引

1
2
3
4
5
6
mysql> explain select * from tb_user where phone = '10931949622' or age = 23;
+----+-------------+---------+------------+------+----------------+------+---------+------+--------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------+------+---------+------+--------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | ALL | idx_user_phone | NULL | NULL | NULL | 271498 | 10.00 | Using where |
+----+-------------+---------+------------+------+----------------+------+---------+------+--------+----------+-------------+

数据分布情况也会影响是否使用索引

如果MySQL评估使用索引比全表更慢,则不使用索引

例: tb_user表数据profression都是not null的,使用where is not null查询没有走索引

1
2
3
4
5
6
mysql> explain select * from tb_user where profession is not null ;
+----+-------------+---------+------------+------+----------------------+------+---------+------+--------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------------+------+---------+------+--------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | ALL | idx_user_pro_age_sta | NULL | NULL | NULL | 271498 | 50.00 | Using where |
+----+-------------+---------+------------+------+----------------------+------+---------+------+--------+----------+-------------+

索引提示

SQL提示,是优化数据库的重要手段。指在SQL语句中加入一些认为提示来达到优化操作的目的。

例:先创建一个profession单列索引

1
mysql> create index idx_user_pro on tb_user(profession);

再查询profession为’R&D’的员工。MySQL可以用普通索引idx_user_pro,也可以用联合索引idx_user_pro_age_sta

1
2
3
4
5
6
mysql> explain select * from tb_user where profession = 'R&D';
+----+-------------+---------+------------+------+-----------------------------------+----------------------+---------+-------+--------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+-----------------------------------+----------------------+---------+-------+--------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro_age_sta,idx_user_pro | idx_user_pro_age_sta | 147 | const | 127438 | 100.00 | NULL |
+----+-------------+---------+------------+------+-----------------------------------+----------------------+---------+-------+--------+----------+-------+

use index (建议使用索引)

1
2
3
4
5
6
mysql> explain select * from tb_user use index(idx_user_pro) where profession = 'R&D';
+----+-------------+---------+------------+------+---------------+--------------+---------+-------+--------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+---------------+--------------+---------+-------+--------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro | idx_user_pro | 147 | const | 127772 | 100.00 | NULL |
+----+-------------+---------+------------+------+---------------+--------------+---------+-------+--------+----------+-------+

ignore index

1
2
3
4
5
6
mysql> explain select * from tb_user ignore index(idx_user_pro) where profession = 'R&D';
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------+--------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------+--------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro_age_sta | idx_user_pro_age_sta | 147 | const | 127438 | 100.00 | NULL |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------+--------+----------+-------+

force index (强制使用索引)

1
2
3
4
5
6
mysql> explain select * from tb_user force index(idx_user_pro) where profession = 'R&D';
+----+-------------+---------+------------+------+---------------+--------------+---------+-------+-------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+---------------+--------------+---------+-------+-------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro | idx_user_pro | 147 | const | 90499 | 100.00 | NULL |
+----+-------------+---------+------------+------+---------------+--------------+---------+-------+-------+----------+-------+

覆盖索引

查询使用到了索引,并且查询的所有列,在该索引中已经全部能够找到(不需要回表)

例: tb_user表已有联合索引(profession,age,status), 查询所有列在该索引中能找到时, Extra=’Using index’,当查询字段新增name时,Extra=’NULL’

1
2
3
4
5
6
7
8
9
10
11
12
13
14
mysql> explain select id,profession,age,status from tb_user where profession = 'R&D'  and age = 31 and status = '0';
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro_age_sta | idx_user_pro_age_sta | 154 | const,const,const | 249 | 100.00 | Using index |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------------+
1 row in set, 1 warning (0.00 sec)

mysql> explain select id,profession,age,status,name from tb_user where profession = 'R&D' and age = 31 and status = '0';
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------+
| 1 | SIMPLE | tb_user | NULL | ref | idx_user_pro_age_sta | idx_user_pro_age_sta | 154 | const,const,const | 249 | 100.00 | NULL |
+----+-------------+---------+------------+------+----------------------+----------------------+---------+-------------------+------+----------+-------+

注:

  • using index condition (查找使用了索引,但需要回表)
  • using where; using index; (查找使用了索引,且需要数据都能在索引列中找到,无需回表)

聚集索引和辅助索引

注:

  • 聚集索引
  • 辅助索引 叶子节点存主键ID (聚集索引叶子节点已经存储行数据,没必要每个辅助索引也存一份行数据,只存主键ID节省空间)

前缀索引

当字段类型为字符串(varchar,text)时,可以只将字符串的一部分前缀,建立索引,这样可以大大节约索引空间,从而提高索引效率

语法:

1
create index idx_XXX on table_name(column(n));

前缀长度
可以根据索引选择性决定,选择性指不重复索引值/数据表记录总数,选择性越高说明查询效率越好
唯一索引选择性为1,是最好的索引选择性

1
select count(distinct substring(email, 1, 5)/count(*) from tb_user;

例: 给email建立前缀索引,取长度为5

1
create index idx_email_5 on tb_user(email(5));

单列索引与联合索引

联合索引指一个索引包含了多个列。
业务场景中,如果存在多个查询条件,建议建立联合索引,而非单列索引。

SQL优化

插入数据

  • 批量插入,不要一条一条插入

    1
    insert into test values(1),(2),(3);
  • 手动提交事务

    1
    2
    3
    4
    5
    start transaction;
    insert into tb_test values(1),(2),(3);
    insert into tb_test values(4),(5),(6);
    insert into tb_test values(7),(8),(9);
    commit;
  • 主键顺序插入

  • 大批量插入数据,使用load指令加载本地文件到数据库

    1
    2
    3
    mysql --local-infile -u root -p
    set global local_infile = 1;
    load data local infile '*.log' into table `tb_user` fields terminated by ',' terminated by '\n';

主键优化

  • 页分裂

  • 页合并
    删除一行记录时,实际上记录没有被物理删除,只是被标记为删除,当页中删除记录达到阈值(MERGE_THERSHOLD),InnoDB会试图合并两个页以节省空间

  • 主键设计原则
    满足业务需求情况下,尽量降低主键长度
    插入数据时,尽量选择顺序插入,选择使用AUTO_INCREMENT自增主键
    尽量不使用UUID或其他自然主键(比如身份证号)

order by优化

using filesort: 在排序缓冲区sort buffer中完成排序操作,未通过索引直接返回排序结果
using index: 通过有序索引顺序扫描直接返回的,这种无需额外排序,效率较高

例: 没有建立索引,排序查询走Using filesort.

1
2
3
4
5
6
mysql> explain select id, age, phone from tb_user order by age;
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+----------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+----------------+
| 1 | SIMPLE | tb_user | NULL | ALL | NULL | NULL | NULL | NULL | 271498 | 100.00 | Using filesort |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+----------------+

例:建立索引,排序查询走Using index

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
mysql> create index idx_user_age_phone on tb_user(age,phone);
mysql> explain select id, age, phone from tb_user order by age,phone;
+----+-------------+---------+------------+-------+---------------+--------------------+---------+------+--------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+-------+---------------+--------------------+---------+------+--------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | index | NULL | idx_user_age_phone | 47 | NULL | 271498 | 100.00 | Using index |
+----+-------------+---------+------------+-------+---------------+--------------------+---------+------+--------+----------+-------------+

mysql> explain select id, age, phone from tb_user order by age desc ,phone desc;
+----+-------------+---------+------------+-------+---------------+--------------------+---------+------+--------+----------+----------------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+-------+---------------+--------------------+---------+------+--------+----------+----------------------------------+
| 1 | SIMPLE | tb_user | NULL | index | NULL | idx_user_age_phone | 47 | NULL | 271498 | 100.00 | Backward index scan; Using index |
+----+-------------+---------+------------+-------+---------------+--------------------+---------+------+--------+----------+----------------------------------+

mysql> explain select id, age, phone from tb_user order by age asc ,phone desc;
+----+-------------+---------+------------+-------+---------------+--------------------+---------+------+--------+----------+-----------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+-------+---------------+--------------------+---------+------+--------+----------+-----------------------------+
| 1 | SIMPLE | tb_user | NULL | index | NULL | idx_user_age_phone | 47 | NULL | 271498 | 100.00 | Using index; Using filesort |
+----+-------------+---------+------------+-------+---------------+--------------------+---------+------+--------+----------+-----------------------------+

注:

  • 使用排序字段建立合适索引,多字段排序时也遵循最左前缀法则
  • 尽量使用覆盖索引
  • 多字段排序,一个升序一个降序,需指定联合索引创建的规则(ASC/DESC)
  • 如果不可避免出现filesort,大数量排序时可适当增加排序缓冲区大小sort_buffer_size(默认256k)
    1
    2
    3
    4
    5
    6
    mysql> show variables like 'sort_buffer_size';
    +------------------+--------+
    | Variable_name | Value |
    +------------------+--------+
    | sort_buffer_size | 262144 |
    +------------------+--------+

group by优化

先删除所有索引,测试group by查询结果为Using temporary(效率很低)

1
2
3
4
5
6
mysql> explain select profession,count(*) from tb_user group by profession;
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-----------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-----------------+
| 1 | SIMPLE | tb_user | NULL | ALL | NULL | NULL | NULL | NULL | 271498 | 100.00 | Using temporary |
+----+-------------+---------+------------+------+---------------+------+---------+------+--------+----------+-----------------+

创建索引后,测试group by查询结果为Using index

1
2
3
4
5
6
7
mysql> create index idx_pro_age_sta on tb_user(profession,age,status);
mysql> explain select profession,count(*) from tb_user group by profession;
+----+-------------+---------+------------+-------+-----------------+-----------------+---------+------+--------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+---------+------------+-------+-----------------+-----------------+---------+------+--------+----------+-------------+
| 1 | SIMPLE | tb_user | NULL | index | idx_pro_age_sta | idx_pro_age_sta | 154 | NULL | 271498 | 100.00 | Using index |
+----+-------------+---------+------------+-------+-----------------+-----------------+---------+------+--------+----------+-------------+

limit优化

分页的记录越靠后,查询耗时越大。
优化思路:一般分页查询时,通过创建覆盖索引+子查询形式优化。

例:

1
2
select * from tb_sku order by id limit 9000000,10;
select s.* from tb_sku s, (select id from tb_sku order by id limit 900000,10) a where s.id = a.id;

count优化

表数据量大时, count(*)执行耗时较多

  • count(主键)
    遍历整张表,把每一行主键id取出来返回给服务层。服务层直接按行累加(无需判断null)。
  • count(字段)
    如果没有not null约束,需要把每一行字段值取出来返回给服务层,服务层需判断是否为null,不为null才累加。
  • count(1)
    遍历整张表,但不取值。服务层对于返回的每一行,放一个数字’1’进去,直接按行进行累加。
  • count(*)
    不会把全部字段取出来,而是专门做了优化,不取值,服务层直接按行累加。

总结: count(*)效率最高

update优化

更新数据时,一定要根据索引更新,否则会出现锁表的现象。

例: 没有对name建立索引的条件下,以下update语句会锁住整张表

1
update student set no = '123' where name = 'haha';

视图

视图(View)是一种虚拟存在的表,视图仅保存查询的SQL逻辑,不保存查询结果。

创建视图

1
create view tb_user_v1 as select id, name from tb_user where id <= 10;

查询视图

1
select * from tb_user_v1;

修改视图

1
alter view tb_user_v1 as select id from tb_user where id <= 10;

删除

1
drop view tb_user_v1;

检查选项cascade

视图可以插入数据,通过视图插入的数据不一定能在视图中查询到(比如插入id>10的记录)

例: 创建视图时指定with cascaded check option,可以阻止通过视图插入查不到的数据

1
2
3
4
mysql> create or replace view tb_user_v1 as select id, name from tb_user where id <= 10 with cascaded check option;

mysql> insert into tb_user_v1(id,name) values(100,'hello');
ERROR 1369 (HY000): CHECK OPTION failed 'pc.tb_user_v1'

存储过程

创建过程

1
2
3
4
5
6
7
8
9
10
11
12
13
14
delimiter $$
create procedure p1()
begin
select count(*) from tb_user;
end$$

delimiter ;
mysql> call p1();
+----------+
| count(*) |
+----------+
| 272633 |
+----------+
1 row in set (0.02 sec)

查看过程

1
2
select * from information_schema.ROUTINES where ROUTINE_SCHEMA = 'itcast';
show create procedure 存储过程名称;

删除过程

1
drop procedure if exists p1;

变量

  • 全局变量(GLOBAL)
  • 会话变量(SESSION)

查看系统变量

1
2
3
show [session|global] variables;
show [session|global] variables like '...';
select @@[session|global].变量;

设置系统变量

1
set [session|global] 系统变量名=值;

注:

  • 未指定session或global时,默认为session,会话级变量
  • mysql重启后,所设置全局参数会失效,要想不失效,可以在/etc/my.cnf中配置。

用户定义和使用变量

1
2
set @var_name := expr [, @var_name = expr]; -- 定义变量
select @var_name; -- 使用变量

注:用户定义变量无需对其进行声明或初始化,只不过获取到的值为NULL

if语句

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
create procedure p3()
begin
declare score int default 58;
declare result varchar(10);

if score >= 85 then
set result := '优秀';
elseif score >= 60 then
set result := '及格';
else
set result := '不及格';
end if
select result;
end;

call p3();

参数

例:入参为分数,出参为(优秀,及格,不及格)

1
2
3
4
5
6
7
8
9
10
11
12
13
create procedure p4(in score int, out result varchar(10))
begin
if score >= 85 then
set result := '优秀';
elseif score >= 60 then
set result := '及格';
else
set result := '不及格';
end if
select result;
end;

call p4(68, @result);

case

例:求月份所属季度

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
create procedure p6(in month int)
begin
declare result varchar(10);

case
when month >= 1 and month <= 3 then
set result := '第一季度';
when month >= 4 and month <= 6 then
set result := '第二季度';
when month >= 7 and month <= 9 then
set result := '第三季度';
when month >= 10 and month <= 12 then
set result := '第四季度';
else
set result := '非法参数';
end case;

select concat('输入月份为: ', month, ', 所属季度为: ', result);
end;

call p6(4);

while

例: 计算从1累加到n的和

1
2
3
4
5
6
7
8
9
10
11
create procedure p7(in n int)
begin
declare total int default 0;
while n>0 do
set total := total + n;
set n := n + 1;
end while;
select total;
end;

call p7(10);

repeat

例:计算从1累加到n的和

1
2
3
4
5
6
7
8
9
10
11
create procedure p8(in n int)
begin
declare total int default 0;
repeat
set total := total + n;
set n := n - 1;
until n <= 0
end repeat;

select total;
end;

游标(cursor)

用来存储查询结果集的数据类型

例:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
create procedure p11(in uage int)
begin
declare u_cursor cursor for select name,profession from tb_user where age <= uage;
declare uname varchar(100);
declare upro varchar(100);
declare exit handler for SQLSTATE '02000';

drop table if exists tb_user_pro;
create table if not exists tb_user_pro(
id int primary key auto_increment,
name varchar(100),
profession varchar(100)
);
open u_cursor;
while true do
fetch u_cursor into uname,upro;
insert into tb_user_pro values(null,uname,upro);
end while;
close u_cursor;
end;

call p11(40);

存储函数

存储函数是有返回值的存储过程,参数只能是IN类型。

例:从1到n累加

1
2
3
4
5
6
7
8
9
10
11
create function fun1(n int)
returns int
begin
declare total int default 0;
while n>0 do
set total := total + n;
set n := n - 1;
end while;

return total;
end;

锁(重点)

锁是计算机协调多个进程或线程并发访问某一资源的机制

  • 全局锁: 锁定数据库中的所有表
  • 表级锁: 每次操作锁住整张表
  • 行级锁: 每次操作锁住对应的行数据

全局锁

对整个数据库实例加锁,加锁后整个实例处于只读状态,后续的DML写语句,DDL语句都被阻塞。
典型的使用场景是做全库的逻辑备份,对所有的表进行锁定,从而获取一致性视图,保证数据的完整性。

全局锁实例

1
2
3
4
5
6
7
8
9
10
11
12
13
-- 加全局锁
flush table with read lock;

-- 加了全局锁后,可以读
update tb_user set name = 'XXX' where id = 1;
ERROR 1223 (HY000): Can't execute the query because you have a conflicting read lock

-- 但是不能写
update tb_user set name = 'XXX' where id = 1;
ERROR 1223 (HY000): Can't execute the query because you have a conflicting read lock

-- 释放锁
unlock tables;

注:数据库中加全局锁,是一个比较重的操作,存在以下问题:

  • 如果在主库上备份,那么在备份期间都不能更新
  • 如果在从库上备份,那么备份期间从库不能执行主库同步来的二进制日志(binlog),会导致主从延迟
  • 备份数据库的工具是 mysqldump,加上 –single-transaction 参数,完成不加锁的备份。
    https://www.cnblogs.com/xiaolincoding/p/16820726.html

表级锁

表级锁分为如下三类

  • 表锁
  • 元数据锁 (meta data lock, MDL)
  • 意向锁

表锁

  • 表共享读锁(read lock)
  • 表独占写锁(write lock)

语法
加锁: lock tables 表名… read/write
释放锁: unlock tables (或客户端断开连接时释放锁)

读锁的特点
a) 客户端c1加了读锁后,c1和c2都能读,但c1和c2都不能写。

1
2
3
4
5
6
7
8
9
10
11
12
mysql> lock tables tb_user read;
mysql> select * from tb_user where id = 1;
+----+-------+-------------+---------------+
| id | name | phone | email |
+----+-------+-------------+---------------+
| 1 | uXrBh | 15847705582 | YYdRR@163.com |
+----+-------+-------------+---------------+

mysql> update tb_user set name = 'XXX' where id = 1;
ERROR 1099 (HY000): Table 'tb_user' was locked with a READ lock and can't be updated

mysql> unlock tables

b) 客户端c1加了写锁后, c1能读也能写,但c2不能读也不能写。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
-- 客户端c1
mysql> lock table tb_user write;

mysql> update tb_user set name = 'XXXXX' where id = 1;
Query OK, 1 row affected (0.01 sec)
Rows matched: 1 Changed: 1 Warnings: 0

mysql> select * from tb_user where id = 1;
+----+-------+-------------+---------------+------------+------+--------+--------+---------------------+
| id | name | phone | email | profession | age | gender | status | createtime |
+----+-------+-------------+---------------+------------+------+--------+--------+---------------------+
| 1 | XXXXX | 15847705582 | YYdRR@163.com | OPS | 81 | F | 0 | 2024-08-22 00:00:00 |
+----+-------+-------------+---------------+------------+------+--------+--------+---------------------+

-- 客户端c2
-- 不能读,被阻塞
mysql> select * from tb_user limit 1;
^C^C -- query aborted
ERROR 1317 (70100): Query execution was interrupted

-- 不能写,直接报错
mysql> mysql> update tb_user set name = 'XXX' where id = 1;
ERROR 1064 (42000): You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near 'mysql> update tb_user set name = 'XXX' where id = 1' at line 1

元数据锁(meta data lock, MDL)

  • MDL加锁是系统自动控制,无需显式使用,访问一张表时会自动加上。
  • MDL锁作用是维护表元数据的一致性,表上有活动事务时,不可以对元数据进行写入操作。
  • MySQL5.5中引入MDL,对一张表进行增删改查时,加MDL读锁(共享), 对表结构进行变更时,加MDL写锁(排他)。避免了DML与DDL的冲突,保证读写正确性。
对应SQL 锁类型 说明
lock tables xxx read/write SHARED_READ_ONLY/SHARED_NO_READ_WRITE
select, select … lock in share mode SHARED_READ 与SHARED_READ, SHARED_WRITE兼容, 与EXCLUSIVE互斥
insert, update, delete, select … for update SHARED_WRITE 与SHARD_READ, SHARED_WRITE兼容, 与EXCLUSIVE互斥
alter table … EXCLUSIVE 与其他MDL互斥

元数据锁的特点
客户端c1开启事务,执行select操作, 客户端c2开启事务,执行alter table操作会阻塞

1
2
3
4
5
6
7
8
9
10
11
12
13
-- 客户端c1手动开启事务,执行select操作, 不提交
mysql> begin;
Query OK, 0 rows affected (0.00 sec)

mysql> select * from demo;
Empty set (0.00 sec)

-- 客户端c2开启事务,执行alter table,发现被阻塞 (EXCLUSIVE与SHARED_READ互斥)
mysql> begin;
Query OK, 0 rows affected (0.00 sec)

mysql> alter table demo add column1 int;
b^

查看元数据锁的指令:

1
select object_type,object_schema,object_name,lock_type,lock_duration from performance_schema.metadata_locks;

意向锁

为了避免DML执行时,加的行锁与表锁冲突,InnoDB引入意向锁,使得表锁不用检查每行数据是否有锁,使用意向锁减少表锁的检查。

意向锁有两种:

  • 意向共享锁(IS): 由语句select … lock in share mode添加
  • 意向排他锁(IX): 由insert,update,delete,select … for update添加

意向锁兼容性:

  • 意向共享锁(IS): 与表锁共享锁兼容,与其他表锁互斥。
  • 意向排他锁(IX): 与表锁共享锁及排它锁都互斥,意向锁之间不会互斥。

可以通过以下SQL,查看意向锁及行锁的加锁情况:

1
select object_schema,object_name,index_name,lock_type,lock_mode,lock_data from performance_schema.data_locks;

意向锁1实例:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
-- 客户端c1开启事务,加IS锁+行锁
mysql>begin;

mysql>select * from demo where id = 1 lock in share mode;

mysql> select object_schema,object_name,index_name,lock_type,lock_mode,lock_data from performance_schema.data_locks;
+---------------+-------------+-----------------+-----------+-----------+------------------------+
| object_schema | object_name | index_name | lock_type | lock_mode | lock_data |
+---------------+-------------+-----------------+-----------+-----------+------------------------+
| pc | demo | NULL | TABLE | IS | NULL |
| pc | demo | GEN_CLUST_INDEX | RECORD | S | supremum pseudo-record |
| pc | demo | GEN_CLUST_INDEX | RECORD | S | 0x000000000300 |
+---------------+-------------+-----------------+-----------+-----------+------------------------+

-- 此时,客户端c2可以加表共享读锁,但不能加表排他写锁。
mysql> lock tables demo read;
Query OK, 0 rows affected (0.00 sec)

mysql> unlock tables;
Query OK, 0 rows affected (0.00 sec)

mysql> lock tables demo write;
^C^C -- query aborted
ERROR 1317 (70100): Query execution was interrupted

行级锁

行级锁,每次操作锁住对应的行数据,锁定粒度最小,发生锁冲突概率最低,并发度最高。应用在InnoDB存储引擎中。
行级锁是通过对索引加锁来实现的,不是对记录加锁。行级锁分如下三类:

  • 行锁(Record Lock)
  • 间隙锁(Gap Lock) RR隔离级别下支持
  • 临键锁(Next-Key Lock)
行锁
  • 共享锁(S):允许一个事务读一行,阻止其他事务获得排它锁。
  • 排它锁(X):允许获取排它锁事务更新数据,阻止其他事务获得相同数据集的共享锁和排他锁。
    SQL 行锁类型 说明
    insert 排他锁 自动加锁
    update 排他锁 自动加锁
    delete 排他锁 自动加锁
    select 不加锁
    select … lock in share mode 共享锁 需手动在select之后加lock in share mode
    select … for update 排它锁 需要手动在select之后加for update

行锁示例:
默认情况, InnoDB在Repeatable Read事务隔离级别运行,使用next-key lock进行搜索和索引扫描,防止幻读

  • 针对唯一索引进行检索时,对已存在记录进行等值匹配时,自动优化为行锁。
  • 不通过索引条件检索数据,InnoDB将对表中所有记录加锁,此时升级为表锁。

触发器

触发器是与表有关的数据库对象,指在insert/update/delete之前或之后,触发并执行触发器中定义的SQL语句集合。

触发器类型 NEW和OLD
INSERT型触发器 NEW表示将要或者已经新增的数据
UPDATE型触发器 OLD表示修改之前的数据,NEW表示将要或已经修改后的数据
DELETE型触发器 OLD表示将要或者已经删除的数据

– TODO –

InnoDB引擎

逻辑存储结构

Tablespace(表空间) -> Segment(段) -> Extent(区) -> Row(行) -> Page(页)

表空间(ibd文件), 一个MySQL示例可对应多个表空间,用于存储记录,索引等数据。
段, 分为数据段, 索引段, 回滚段, InnoDB是索引组织表,数据段就是B+树的叶子节点,索引段是B+树的非叶子节点,
区,表空间的单元结构,每个区大小为1M, 默认情况,InnoDB页大小为16K, 一个区中有64个连续的页
页, InnoDB磁盘管理的最小单元,每个页默认16KB,为保证页连续性,InnoDB每次从磁盘申请4到5个区
行, InnoDB数据按行存放。

表空间

1
2
3
4
# ls /var/lib/mysql/*.ibd
/var/lib/mysql/mysql.ibd
# ls /var/lib/mysql/pc/
demo.ibd tb_user.ibd

架构

Buffer Pool

缓冲池: 缓存磁盘上经常操作的数据,执行增删改查操作时,先操作Buffer Pool数据(如Buffer Pool没有数据,就从磁盘加载并缓存),再以一定频率刷到磁盘,从而减少磁盘IO,加快处理速度
以页为单位, 底层采用链表管理Page, 根据状态,将Page分为三类:

  • free page 空闲页,未被使用
  • clean page 被使用页,数据没有被修改过
  • dirty page 被使用页,数据与磁盘中不一致

Change Buffer

更改缓冲区,执行DML语句时,如果这些数据Page没有在Buffer Pool中,不会直接操作磁盘,而是将数据变更存到Change Buffer,
未来数据被读取时,再将数据合并恢复到Buffer Pool中,再将合并后数据刷新到磁盘中。

Adaptive Hash Index (自适应哈希索引)

  • 自适应哈希索引,用于优化对Buffer Pool数据查询。InnoDB会监控对表上各索引页查询,如观察到hash索引可以提升速度,
    则建立hash索引,称之为自适应hash索引
  • 自适应哈希索引,无需人工干预,是系统根据情况自动完成的
    1
    2
    3
    4
    5
    6
    7
    mysql> show variables like '%hash_index%';
    +----------------------------------+-------+
    | Variable_name | Value |
    +----------------------------------+-------+
    | innodb_adaptive_hash_index | ON |
    | innodb_adaptive_hash_index_parts | 8 |
    +----------------------------------+-------+

Log Buffer

用于保存要写入到磁盘中的log(redo log, undo log), 默认大小16MB,参数:

  • innodb_log_buffer_size 缓冲区大小
  • innodb_flush_log_at_trx_commit 日志刷新到磁盘时机

磁盘结构

  • System Tablespace: 系统表空间是更改缓冲区的存储区域,如果表是在系统表空间而不是每个表文件或通用表空间中创建的,也可能含表和索引数据

    1
    2
    3
    4
    5
    6
    mysql> show variables like '%data_file_path%';
    +----------------------------+------------------------+
    | Variable_name | Value |
    +----------------------------+------------------------+
    | innodb_data_file_path | ibdata1:12M:autoextend |
    +----------------------------+------------------------+
  • File-Per-Table Tablespaces: 每个表的文件表空间包含单个InnoDB表数据和索引,并存储在文件系统的单个数据文件中

    1
    2
    3
    4
    5
    6
    mysql> show variables like '%innodb_file_per_table%';
    +-----------------------+-------+
    | Variable_name | Value |
    +-----------------------+-------+
    | innodb_file_per_table | ON |
    +-----------------------+-------+
  • Gerneral Tablespaces: 通用表空间,需通过craete tablespace语法创建通用表空间,创建表时可以指定该表空间

  • Undo Tablespaces: 撤销表空间,MySQL实例在初始化时会自动创建两个默认undo表空间,用于存储undo log

  • Doublewrite Buffer Files: 双写缓冲区,innoDB将数据页从Buffer Pool刷新到磁盘前,先将数据页写入双写缓冲区文件中,便于系统异常时恢复数据。
    (#ib_16384_0.dblwr, #ib_16384_1.dblwr)

  • Redo Log: 重做日志,两部分组成(redo log buffer, redo log),当事务提交后会把所有修改信息存到日志中,用于刷新脏页到磁盘发生错误时,进行数据恢复
    文件(ib_logfile0, ib_logfile1)

后台线程

  • Master Thread 核心线程,负责调度其他线程,将缓冲池中数据异步刷新到磁盘,包括脏页刷新,合并插入缓存,undo页回收
  • IO Thread InnoDB使用AIO处理IO请求
    线程类型 默认个数 职责
    Read Thread 4 负责读操作
    Write Thread 4 负责写操作
    Log Thread 1 负责将日志缓冲区刷新到磁盘
    Insert Buffer Thread 1 负责将缓冲区内容刷新到磁盘

事务原理

redo log

重做日志,记录事务提交时数据页的物理修改,用来实现事务的持久性,由两部分组成

  • redo log buffer(内存)
  • redo log file(磁盘)
    事务提交后会把所有修改信息都存到日志文件中,用于刷新脏页到磁盘发生错误时,进行数据恢复使用。

undo log

回滚日志,记录数据被修改前的信息,作用:提供回滚, MVCC
undo log 记录逻辑日志(01:05)
https://www.bilibili.com/video/BV1Kr4y1i7ru?spm_id_from=333.788.player.switch&vd_source=d8559c2d87607be86810cd806158bb86&p=140

MVCC

MVCC的几个重要概念:

当前读
读取记录的最新版本,读取时要保证其他并发事务不能修改当前记录,会对读取记录加锁。比如:
select … lock in share mode, select … for update, insert, delete都是一种当前读

例: 先在客户端c1开启事务, 再从客户端c2更新tb_user表id=1记录并提交,此时c1不能查到c2修改的记录

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
-- 客户端c1开启事务
mysql> begin;
mysql> select * from tb_user;
+----+-------+
| id | value |
+----+-------+
| 1 | 1 |
+----+-------+
1 row in set (0.00 sec)

-- 客户端c2更新tb_user表id=1记录,并提交事务
mysql> begin;
mysql> update tb_user set value = 2 where id = 1;
mysql> commit;

-- 切到客户端c1,此时c1中查不到c2修改的记录, 即可重复读
mysql> select * from tb_user;
+----+-------+
| id | value |
+----+-------+
| 1 | 1 |
+----+-------+

-- 如果通过select * lock in share mode 使用当前读,就可以查到c2修改的记录
mysql> select * from tb_user lock in share mode;
+----+-------+
| id | value |
+----+-------+
| 1 | 2 |
+----+-------+

快照读
简单的select就是快照读,读取的是记录数据的可见版本,有可能是历史数据,不加锁,是非阻塞读

  • read committed: 每次select都生成一个快照读
  • Repeatable Read: 开启事务后第一个select语句才是快照读
  • Serializable: 快照读退化为当前读

MVCC
全称 Multi-Version Concurrency Control 多版本并发控制。指维护一个数据的多个版本,使读写操作无冲突
快照读为MySQL实现MVCC提供了一个非阻塞读的功能,MVCC的实现依赖数据库记录中三个隐式字段、undo log日志、readView

MVCC实现原理

记录中的隐藏字段
隐式增加了两个字段DB_TRX_ID, DB_ROLL_PTR

隐藏字段 含义
DB_TRX_ID 最近修改事务ID,记录插入这条记录或最后一次修改改记录的事务ID
DB_ROLL_PTR 回滚指针, 指向这条记录的上一个版本,用于配合undo log, 指向上一个版本
DB_ROW_ID 隐藏主键, 如果表结构没有指定主键。将会生成该隐藏字段

查看隐藏字段

1
ibd2sdi /var/lib/mysql/pc/tb_user.ibd

undo log
回滚日志, 在insert, update, delete时产生的,便于数据回滚的日志。

  • 当insert时候,产生的undo log只在回滚时需要,事务提交后可被立即删除。
  • 而update,delete时候,产生的undo log不仅在回滚时需要,在快照读时也需要,不会被立即删除。

undo log版本链
不同事务或相同事务对同一条记录进行修改,
会导致该记录undolog生成一条记录版本链表,链表头部是最新的旧记录,链表尾部是最早的旧记录

** readview **
readview (读视图)是快照读SQL执行时MVCC提取数据的依据,记录并维护系统当前活跃的事务id
readview 包括四个核心字段

字段 含义
m_ids 当前活跃的事务ID集合
min_trx_id 最小活跃事务ID
max_trx_id 预分配事务ID, 当前最大事务ID+1
creator_trx_id ReadView创建者的事务ID
  • trx_id == creator_trx_id 可以访问该版本, 说明数据是当前这个事务更改的
  • trx_id < min_trx_id 可以访问该版本, 说明数据已经提交了
  • trx_id > max_trx_id 不可以访问该版本, 说明该事务在ReadView生成后才开启
  • min_trx_id <= trx_id <= max_trx_id 如果trx_id不在m_ids中是可以访问该版本的

不同的隔离级别,生成readview的时机不同:

  • READ COMMITTED: 在事务中每一次执行快照读时生成Read View
  • REPEATABLE READ: 仅在事务中第一次执行快照读时生成ReadView, 后续复用该ReadView (可重复读)

MySQL管理

系统数据库

  • mysql 存储MySQL正常运行所需各种信息(时区、主从、用户、权限)
  • information_schema 提供了访问数据库元数据的各种表和视图,包含数据库,表,字段类型以及访问权限
  • performance_schema 为MySQL运行时状态提供了一个底层监控功能,主要用于收集数据库服务器性能参数
  • sys 包含一系列方便DBA和开发人员利用performance_schema性能数据库进行性能调优和诊断的视图

常用工具

mysql

例: 通过mysql -e直接执行语句,无需进入交互式命令行 (用户名root,密码123,数据库名称pc)

1
mysql -uroot -p123 pc -e "select * from tb_user"

mysqladmin

一个执行管理操作的客户端程序,可以用它检查服务器配置和当前状态,创建并删除数据库等。

mysqlbinlog

检查二进制日志

1
mysqlbinlog -s binlog.000006 /var/lib/mysql/binlog.000006

mysqlshow

mysqlshow客户端对象查找工具,用来很快查找存在哪些数据库,数据库中的表,表中的列或者索引。

1
2
3
4
5
6
7
8
9
10
11
12
13
# mysqlshow -uroot -pXXXXXX --count
mysqlshow: [Warning] Using a password on the command line interface can be insecure.
+--------------------+--------+--------------+
| Databases | Tables | Total Rows |
+--------------------+--------+--------------+
| gx_day15 | 12 | 70 |
| gx_day16 | 14 | 75 |
| information_schema | 79 | 29745 |
| mysql | 37 | 4090 |
| pc | 3 | 2 |
| performance_schema | 111 | 295786 |
| sys | 101 | 6398 |
+--------------------+--------+--------------+

mysqldump

mysqldump用于备份数据库和在不同数据库间作数据迁移。备份内容包含创建表,以及插入表的SQL语句

1
mysqldump -uroot -ppassword dbname > output.sql

查看MySQL版本

https://github.com/mysql/mysql-server/releases/tag/mysql-8.0.36

进程模型

1
2
systemctl cat mysqld | grep ExecStart=
ExecStart=/usr/libexec/mysqld --basedir=/usr

单进程多线程模型

1
2
3
4
5
6
7
8
ps axf | grep mysqld
16273 ? Ssl 0:33 /usr/libexec/mysqld --basedir=/usr
ps -T 16273
PID SPID TTY STAT TIME COMMAND
16273 16273 ? Ssl 0:00 /usr/libexec/mysqld --basedir=/usr
16273 16276 ? Ssl 0:00 /usr/libexec/mysqld --basedir=/usr
16273 16277 ? Ssl 0:00 /usr/libexec/mysqld --basedir=/usr
....

https://www.jcwlyf.com/newsContent-id-18272.html

mysqld进程

  • 监听线程,接受客户端连接请求;
  • 查询线程,处理客户端发来的SQL查询;
  • 复制线程,处理主从复制相关操作;
  • 后台线程,执行定期任务如日志清理、缓存刷新等。

例如:删除/data/log/目录下所有10天前的.log文件

1
find /data/log/ -type f -name "*.log" -mtime +10 -exec rm -f {} \;

只查看要删除的文件有哪些,不真正删除文件

1
2
logfiles=$(find /data/log/ -type f -name "*.log" -mtime +10)
echo $logfiles
阅读全文 »

数据库相关概念

  • 数据库(DataBase) 存储数据仓库
  • 数据库管理系统(DateBase Management System)(DBMS) 操纵和管理数据库的大型软件
  • SQL(Structured Query Language)(SQL) 操作关系型数据库的编程语言,定义了一套操作关系数据库的统一标准

关系模型

关系模型本质上就是若干个存储数据的二维表

  • 表的每一行称为记录(Record),记录是一个逻辑意义上的数据。
  • 表的每一列称为字段(Column),同一个表的每一行记录都拥有相同的若干字段。
  • 字段定义了数据类型(整型、浮点型、字符串、日期等),以及是否允许为NULL
阅读全文 »

问题描述

在RHEL中运行了一个自定义的systemd服务,启动报错 Start operation timed out, 在后台执行systemctl start也被阻塞, 不能自动退出。

service配置文件如下:

1
2
3
4
5
6
7
8
9
[Unit]
After=network.target

[Service]
Type=forking
ExecStart=/path/to/monitor_network.sh

[Install]
WantedBy=multi-user.target

解决方法

查阅资料,发现Type=forking有问题,这里应该改成Type=simple。

阅读全文 »

先查看当前网卡

1
2
3
4
nmcli con show
NAME UUID TYPE DEVICE
eth0 XX ethernet eth0
lo XX loopback lo

例如,配置eth0网卡的static ip为10.206.216.93, gateway 10.206.216.254, DNS 10.204.16.18

1
2
3
nmcli con del eth0
nmcli con add con-name eth0 autoconnect yes type ethernet ifname eth0 ip4 10.206.216.93/24 gw4 10.206.216.254 ipv4.dns "8.8.8.8 4.4.4.4"
nmcli con up eth0

使用ip命令配置临时路由

添加静态路由

1
ip route add <目的网络> via <下一跳IP> dev <网卡接口名称>

例: 给eth0网卡添加一个到达 192.168.2.0/24 网络,下一跳为 192.168.1.254 的路由

1
ip route add 192.168.2.0/24 via 192.168.1.254 dev eth0

删除静态路由

1
ip route del <目的网络> via <下一跳IP> dev <网卡接口名称>

例: 删除上述的静态路由

1
ip route del 192.168.2.0/24 via 192.168.1.254 dev eth0

使用nmcli配置永久路由

添加静态路由
例: 给eth0网卡添加一个到达 192.168.2.0/24 网络,下一跳为 192.168.1.254 的路由

1
nmcli connection modify eth0 +ipv4.routes "192.168.2.0/24 192.168.1.254"

如果需要添加多个路由,可以用逗号分隔的方式添加:

1
nmcli connection modify eth0 +ipv4.routes "192.168.2.0/24 192.168.1.254,192.168.3.0/24 192.168.1.254"

删除静态路由

1
nmcli connection modify eth0 -ipv4.routes "192.168.2.0/24 192.168.1.254"
阅读全文 »