作者:刘晨

网名 bisal ,具有十年以上的应用运维工作经验,目前主要从事数据库应用研发能力提升方面的工作,Oracle ACE ,拥有 Oracle OCM & OCP、EXIN DevOps Master 、SCJP 等国际认证,国内首批 Oracle YEP 成员,OCMU 成员,《DevOps 最佳实践》中文译者之一,CSDN & ITPub 专家博主,公众号”bisal的个人杂货铺”,长期坚持分享技术文章,多次在线上和线下分享技术主题。

本文来源:原创投稿

*爱可生开源社区出品,原创内容未经授权不得随意使用,转载请联系小编并注明来源。


我们知道,Oracle 中除了使用 select … for update ,其他查询语句不会出现锁,即没有读锁,读一致性通过多版本解决的,可以保证在不加锁的情况下,读到同一时间的数据。

前两天同事在微信群推了一篇文章,大概意思就是通过使用 insert into select 做了数据的备份,导致了 select 的表锁住,进而影响了正常的使用。

问题来了,Oracle 中执行的 insert into select 很正常,不会出现锁表,难道相同的语句用在了 MySQL ,就会锁住整张表?

我们能进行验证,MySQL 5.7 中执行如下语句,会出现什么现象?

insert into test_1 select * from test_2;

test_1 和 test_2 定义如下,test_1 存在五条记录,

​mysql> show create table test_1\G;
*************************** 1. row ***************************
       Table: test_1
Create Table: CREATE TABLE `test_1` (
  `id` int(11) NOT NULL,
  `name` varchar(10) NOT NULL,
  PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4
1 row in set (0.04 sec)
​
mysql> show create table test_2\G;
*************************** 1. row ***************************
       Table: test_2
Create Table: CREATE TABLE `test_2` (
  `id` int(11) NOT NULL,
  `name` varchar(10) NOT NULL
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4
1 row in set (0.01 sec)
​
mysql> select * from test_1;
+----+--------+
| id | name   |
+----+--------+
|  1 | test_1 |
|  2 | test_2 |
|  3 | test_3 |
|  4 | test_4 |
|  5 | test_5 |
+----+--------+
5 rows in set (0.01 sec)

默认情况下,show engine innodb status 显示的锁信息很有限,可以开启锁监控,如果仅需要在 show engine innodb status 显示具体的锁,可以仅打开 innodb_status_output_locks,

该参数的默认值 OFF ,而且只可以在全局层面打开,

​mysql> show variables like 'innodb_status_output_locks';
+----------------------------+-------+
| Variable_name              | Value |
+----------------------------+-------+
| innodb_status_output_locks | OFF   |
+----------------------------+-------+
1 row in set (0.44 sec)
​
mysql> set global innodb_status_output_locks=on;
Query OK, 0 rows affected (0.02 sec)
​
mysql> show variables like 'innodb_status_output_locks';
+----------------------------+-------+
| Variable_name              | Value |
+----------------------------+-------+
| innodb_status_output_locks | ON    |
+----------------------------+-------+
1 row in set (0.01 sec)

在会话 1 中,开启一个事务,将 test_1 的 name=’test_1′ 这行记录导入 test_2 ,

mysql> begin;
Query OK, 0 rows affected (0.00 sec)
​
mysql> insert into test_2 select * from test_1 where name = 'test_1';
Query OK, 1 row affected (0.00 sec)
Records: 1  Duplicates: 0  Warnings: 0

查看锁的信息,可以看到,有五个 record lock ,虽然我只从 test_1 读取一行数据,但实际上对 test_1 的所有记录都加了锁,而且显式对 test_1 加了一个 IS 的意向锁,因此这种操作,确实影响了 select 表的并发执行,

mysql> show engine innodb status \G;
...
------------
TRANSACTIONS
------------
Trx id counter 3255
Purge done for trx's n:o < 3254 undo n:o < 0 state: running but idle
History list length 3
LIST OF TRANSACTIONS FOR EACH SESSION:
---TRANSACTION 422059634232944, not started
0 lock struct(s), heap size 1136, 0 row lock(s)
---TRANSACTION 422059634231120, not started
0 lock struct(s), heap size 1136, 0 row lock(s)
---TRANSACTION 3254, ACTIVE 4 sec
3 lock struct(s), heap size 1136, 6 row lock(s), undo log entries 1
MySQL thread id 23, OS thread handle 140584218986240, query id 16201659 localhost root
TABLE LOCK table `bisal`.`test_1` trx id 3254 lock mode IS
RECORD LOCKS space id 44 page no 3 n bits 72 index PRIMARY of table `bisal`.`test_1` trx id 3254 lock mode S
Record lock, heap no 1 PHYSICAL RECORD: n_fields 1; compact format; info bits 0
 0: len 8; hex 73757072656d756d; asc supremum;;
​
Record lock, heap no 2 PHYSICAL RECORD: n_fields 4; compact format; info bits 0
 0: len 4; hex 80000001; asc     ;;
 1: len 6; hex 000000000ca3; asc       ;;
 2: len 7; hex a80000011c0110; asc        ;;
 3: len 6; hex 746573745f31; asc test_1;;
​
Record lock, heap no 3 PHYSICAL RECORD: n_fields 4; compact format; info bits 0
 0: len 4; hex 80000002; asc     ;;
 1: len 6; hex 000000000ca3; asc       ;;
 2: len 7; hex a80000011c011c; asc        ;;
 3: len 6; hex 746573745f32; asc test_2;;
​
Record lock, heap no 4 PHYSICAL RECORD: n_fields 4; compact format; info bits 0
 0: len 4; hex 80000003; asc     ;;
 1: len 6; hex 000000000ca3; asc       ;;
 2: len 7; hex a80000011c0128; asc       (;;
 3: len 6; hex 746573745f33; asc test_3;;
​
Record lock, heap no 5 PHYSICAL RECORD: n_fields 4; compact format; info bits 0
 0: len 4; hex 80000004; asc     ;;
 1: len 6; hex 000000000ca3; asc       ;;
 2: len 7; hex a80000011c0134; asc       4;;
 3: len 6; hex 746573745f34; asc test_4;;
​
Record lock, heap no 6 PHYSICAL RECORD: n_fields 4; compact format; info bits 0
 0: len 4; hex 80000005; asc     ;;
 1: len 6; hex 000000000ca3; asc       ;;
 2: len 7; hex a80000011c0140; asc       @;;
 3: len 6; hex 746573745f35; asc test_5;;
​
TABLE LOCK table `bisal`.`test_2` trx id 3254 lock mode IX
​
...

解决方案 1 ,创建索引

我们为列 name 创建一个索引,

mysql> alter table test_1 add index idx_test_1_01 (name);
Query OK, 0 rows affected (0.18 sec)
Records: 0  Duplicates: 0  Warnings: 0

再次开启事务,

mysql> begin;
Query OK, 0 rows affected (0.00 sec)
​
mysql> insert into test_2 select * from test_1 where name = 'test_1';
Query OK, 1 row affected (0.00 sec)
Records: 1  Duplicates: 0  Warnings: 0

此时看下锁,这次没对 test_1 加任何的锁,只是对'test_1'这行记录加了共享锁(lock mode S locks gap before rec),其实是加到了索引上,


​mysql> show engine innodb status \G;
...
------------
TRANSACTIONS
------------
Trx id counter 3268
Purge done for trx's n:o < 3268 undo n:o < 0 state: running but idle
History list length 4
LIST OF TRANSACTIONS FOR EACH SESSION:
---TRANSACTION 422059634232944, not started
0 lock struct(s), heap size 1136, 0 row lock(s)
---TRANSACTION 422059634231120, not started
0 lock struct(s), heap size 1136, 0 row lock(s)
---TRANSACTION 3263, ACTIVE 3 sec
4 lock struct(s), heap size 1136, 2 row lock(s), undo log entries 1
MySQL thread id 23, OS thread handle 140584218986240, query id 16201664 localhost root
TABLE LOCK table `bisal`.`test_1` trx id 3263 lock mode IS
RECORD LOCKS space id 44 page no 4 n bits 72 index idx_test_1_01 of table `bisal`.`test_1` trx id 3263 lock mode S
Record lock, heap no 2 PHYSICAL RECORD: n_fields 2; compact format; info bits 0
 0: len 6; hex 746573745f31; asc test_1;;
 1: len 4; hex 80000001; asc     ;;
​
TABLE LOCK table `bisal`.`test_2` trx id 3263 lock mode IX
RECORD LOCKS space id 44 page no 4 n bits 72 index idx_test_1_01 of table `bisal`.`test_1` trx id 3263 lock mode S locks gap before rec
Record lock, heap no 3 PHYSICAL RECORD: n_fields 2; compact format; info bits 0
 0: len 6; hex 746573745f32; asc test_2;;
 1: len 4; hex 80000002; asc     ;;
 ...

解决方案2:更改隔离级别

在创建索引前,之所以会出现锁表的情况,和隔离级别是相关的,首先看下数据库的隔离级别。ISO 和 ANSI SQL 标准制定了 4 种事务隔离级别的标准,包括如下,
Read Uncommitted
Read Committed
Repeatable Read
Serializable
然而不是所有的数据库厂商都遵循这些标准,例如 Oracle 不支持 RU 和 RR ,MySQL 则支持所有级别。Oracle 默认隔离级别是 RC ,MySQL 默认隔离级别是 RR 。
P.S. 
https://dev.mysql.com/doc/refman/5.7/en/innodb-transaction-isolation-levels.html

RR 和 RC 下,InnoDB 引擎都提供了一致性的非锁定读,即通过多版本控制的方式来读取当前时刻的行数据,从技术实现上,MySQL 和 Oracle 是很相像的,都是通过回滚段来实现的 MVCC (Multi Version Concurrency Control),每行都可能有多个版本,即多个快照数据,避免对读加锁,提高读的并发。

比较一下 RR 和 RC ,最大的区别是两者对快照数据的定义不同,RR 模式下读取的是事务开始时的行快照数据,RC 模式下读取的则是该行最新的一份快照数据,我们通过实验,来看下这是什么意思。

如果是 RR 模式,模拟如下两个事务的操作。

T1 时刻,

会话 1 ,

mysql> begin;
Query OK, 0 rows affected (0.00 sec)
​
mysql> select * from test_1 where id=5;
+----+--------+
| id | name   |
+----+--------+
|  5 | test_5 |
+----+--------+
1 row in set (0.01 sec)

T2 时刻,

会话 2 ,

mysql> begin;
Query OK, 0 rows affected (0.00 sec)
​
mysql> update test_1 set name='test_6' where id=5;
Query OK, 1 row affected (0.00 sec)
Rows matched: 1  Changed: 1  Warnings: 0

T3 时刻,

会话 1 ,

mysql> select * from test_1 where id=5;
+----+--------+
| id | name   |
+----+--------+
|  5 | test_5 |
+----+--------+
1 row in set (0.01 sec)

T4 时刻,

会话 2 ,

mysql> commit;
Query OK, 0 rows affected (0.01 sec)

T5 时刻,

会话 1 ,

mysql> select * from test_1 where id=5;
+----+--------+
| id | name   |
+----+--------+
|  5 | test_5 |
+----+--------+
1 row in set (0.01 sec)

可以看到,无论在会话 2 的事务中 id=5 的记录如何改动,会话 1 的事务中,id=5 的记录值,都和事务开始时的值相同。

更改为 RC 模式,模拟如下两个事务的操作。

在两个会话中,都执行这个操作,

mysql> set session transaction_isolation='read-committed';
Query OK, 0 rows affected (0.00 sec)

T1 时刻,

会话 1 ,

mysql> begin;
Query OK, 0 rows affected (0.00 sec)
​
mysql> select * from test_1 where id=5;
+----+--------+
| id | name   |
+----+--------+
|  5 | test_5 |
+----+--------+
1 row in set (0.00 sec)

T2 时刻,

会话 2 ,

mysql> begin;
Query OK, 0 rows affected (0.00 sec)
​
mysql> update test_1 set name='test_6' where id=5;
Query OK, 1 row affected (0.00 sec)
Rows matched: 1  Changed: 1  Warnings: 0

T3 时刻,

会话 1 ,

mysql> select * from test_1 where id=5;
+----+--------+
| id | name   |
+----+--------+
|  5 | test_5 |
+----+--------+
1 row in set (0.00 sec)

T4 时刻,

会话 2 ,

mysql> commit;
Query OK, 0 rows affected (0.01 sec)

T5 时刻,

会话 1 ,

mysql> select * from test_1 where id=5;
+----+--------+
| id | name   |
+----+--------+
|  5 | test_6 |
+----+--------+
1 row in set (0.00 sec)

可以看到,在会话 2 的事务中改动 id=5 的值,在会话1的事务中得到了体现。

因此,RR 模式下读取的是事务开始时的行快照数据,RC 模式下读取的则是该行最新的一份快照数据。

如果隔离级别是 RC ,执行如上 insert into select 操作,

mysql> show variables like '%transaction_isolation%';
+-----------------------+----------------+
| Variable_name         | Value          |
+-----------------------+----------------+
| transaction_isolation | READ-COMMITTED |
+-----------------------+----------------+
1 row in set (0.00 sec)
​
mysql> begin;
Query OK, 0 rows affected (0.00 sec)
​
mysql> insert into test_2 select * from test_1 where name = 'test_1';
Query OK, 1 row affected (0.00 sec)
Records: 1  Duplicates: 0  Warnings: 0

此时看下锁信息,能看到 test_2 上是没有任何锁,因此不会出现 RR 会锁定 test_2 的情况,

mysql> show engine innodb status \G;
...
------------
TRANSACTIONS
------------
Trx id counter 3269
Purge done for trx's n:o < 3268 undo n:o < 0 state: running but idle
History list length 0
LIST OF TRANSACTIONS FOR EACH SESSION:
---TRANSACTION 422059634232944, not started
0 lock struct(s), heap size 1136, 0 row lock(s)
---TRANSACTION 422059634231120, not started
0 lock struct(s), heap size 1136, 0 row lock(s)
---TRANSACTION 3268, ACTIVE 108 sec
1 lock struct(s), heap size 1136, 0 row lock(s), undo log entries 1
MySQL thread id 23, OS thread handle 140584218986240, query id 16201671 localhost root
TABLE LOCK table `bisal`.`test_2` trx id 3268 lock mode IX
...

从语义上讲,RC 模式,其实破坏了 ACID 中的 I ,因为两个事务并未做到真正的隔离。而在 RR 模式,虽然两个事务做到了真正的隔离,但实际通过加锁,还是会产生一些问题的,因此隔离级别的选择,其实还是一种权衡的。