通过唯一索引S锁与X锁来了解MySQL死锁套路

在初学者从源码理解MySQL死锁问题中介绍了使用调试 MySQL  源码的方式来查看死锁的过程，这篇文章来讲讲一个常见的案例。
这次我们讲一段唯一索引 S 锁与 X 锁的爱恨情仇

我们来看一个简化过的例子

# 构造数据
CREATE TABLE `t1` (
 `id` int(11) NOT NULL AUTO_INCREMENT,
 `name` varchar(10),
 `level` int(11),
 PRIMARY KEY (`id`),
 UNIQUE KEY `uk_name` (`name`)
);
INSERT INTO `t1` (`name`, `level`) VALUES ('A',0);

# 出现问题的sql语句如下，并发情况下就会出现死锁
INSERT ignore INTO `t1` (`name`, `level`) VALUES ('A',0);
update t1 set level = 1 where name = "A";

我们用之前介绍过的源码分析方式，先来看下这两条语句分别加什么锁，然后分析死锁形成的过程。

第一条语句

INSERT ignore INTO t1 (name, level) VALUES ('A',0);

在调试中得到的结果如下

可以看到这条语句对唯一键 uk_name 加共享锁（S锁），而且成功。

第二条语句

update t1 set level = 1 where name = "A";

 通过唯一键更新数据库字段。

这种情况在之前的文章已经介绍过，会对唯一索引加 X 锁，然后对主键索引加 X 锁

这样就可以非常轻松的复现死锁的问题了，步骤如下

1.开启两个 session，分别 begin
2.session1 执行INSERT ignore INTO t1 (name, level) VALUES ('A',0);
3.session2 执行INSERT ignore INTO t1 (name, level) VALUES ('A',0);
4.session1 执行update t1 set level = 1 where name = "A"; 进入等待状态
5.session2 执行update t1 set level = 1 where name = "A";，死锁产生，被回滚，同时事务 1 执行成功

详细的锁状态变化如下

死锁日志如下：

LATEST DETECTED DEADLOCK
------------------------
181208 23:00:52
*** (1) TRANSACTION:
TRANSACTION 53A7, ACTIVE 162 sec starting index read
mysql tables in use 1, locked 1
LOCK WAIT 3 lock struct(s), heap size 376, 2 row lock(s)
MySQL thread id 12, OS thread handle 0x700010522000, query id 1424 localhost root Updating
update t1 set level = 1 where name = "A"
*** (1) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 89 page no 4 n bits 72 index `uk_name` of table `lock_demo2`.`t1` trx id 53A7 lock_mode X locks rec but not gap waiting
Record lock, heap no 2 PHYSICAL RECORD: n_fields 2; compact format; info bits 0
 0: len 1; hex 41; asc A;;
 1: len 4; hex 80000001; asc ;;

*** (2) TRANSACTION:
TRANSACTION 53A8, ACTIVE 8 sec starting index read
mysql tables in use 1, locked 1
3 lock struct(s), heap size 376, 2 row lock(s)
MySQL thread id 96, OS thread handle 0x70001062e000, query id 1425 localhost root Updating
update t1 set level = 1 where name = "A"
*** (2) HOLDS THE LOCK(S):
RECORD LOCKS space id 89 page no 4 n bits 72 index `uk_name` of table `lock_demo2`.`t1` trx id 53A8 lock mode S
Record lock, heap no 2 PHYSICAL RECORD: n_fields 2; compact format; info bits 0
 0: len 1; hex 41; asc A;;
 1: len 4; hex 80000001; asc ;;

*** (2) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 89 page no 4 n bits 72 index `uk_name` of table `lock_demo2`.`t1` trx id 53A8 lock_mode X locks rec but not gap waiting
Record lock, heap no 2 PHYSICAL RECORD: n_fields 2; compact format; info bits 0
 0: len 1; hex 41; asc A;;
 1: len 4; hex 80000001; asc ;;

*** WE ROLL BACK TRANSACTION (2)

来详细看一下这个死锁日志

*** (1) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 89 page no 4 n bits 72 index uk_name of table lock_demo2.t1 trx id 53A7 lock_mode X locks rec but not gap waiting

事务 1 想获取 uk_name 唯一索引上的 X 锁 （非 gap 锁的记录锁）

*** (2) HOLDS THE LOCK(S):
RECORD LOCKS space id 89 page no 4 n bits 72 index uk_name of table lock_demo2.t1 trx id 53A8 lock mode S

事务 2 持有uk_name 唯一索引上的 S 锁（共享锁）

*** (2) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 89 page no 4 n bits 72 index uk_name of table lock_demo2.t1 trx id 53A8 lock_mode X locks rec but not gap waiting

事务 2 想获得 uk_name 唯一索引上的 X 锁（非 gap 锁的记录锁）
跟之前理论上推断的结论是一致的