innodb next-key lock引发的死锁

innodb的事务隔离级别是可重复读级别且innodb_locks_unsafe_for_binlog禁用，也就是说允许next-key lock

CREATE TABLE `LockTest` (
   `order_id` varchar(20) NOT NULL,
   `id` bigint(20) NOT NULL AUTO_INCREMENT,
   PRIMARY KEY (`id`),
   KEY `idx_order_id` (`order_id`)
 ) ENGINE=InnoDB AUTO_INCREMENT=16 DEFAULT CHARSET=utf8

事务1	事务2
begin delete from LockTest where order_id =  'D20'
	begin delete from LockTest where order_id =  'D19'
insert into LockTest (order_id) values ('D20')
	insert into LockTest (order_id) values ('D19')
commit	commit

事务1 执行到insert语句会block住，事务2执行insert语句会提示死锁错误

错误码: 1213
Deadlock found when trying to get lock; try restarting transaction

Execution Time : 00:00:00:000
Transfer Time : 00:00:00:000
Total Time : 00:00:00:000

show engine innodb status 显示死锁信息

------------------------
LATEST DETECTED DEADLOCK
------------------------
2014-04-30 15:01:55 a233b90
*** (1) TRANSACTION:
TRANSACTION 596042, ACTIVE 7 sec inserting
mysql tables in use 1, locked 1
LOCK WAIT 3 lock struct(s), heap size 320, 2 row lock(s), undo log entries 1
MySQL thread id 10851, OS thread handle 0x2abfb90, query id 251521 10.10.53.122 root update
insert into LockTest (order_id) values ('D20')
*** (1) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 502 page no 4 n bits 72 index `idx_order_id` of table `test`.`LockTest` trx id 596042 lock_mode X insert intention waiting
Record lock, heap no 1 PHYSICAL RECORD: n_fields 1; compact format; info bits 0
0: len 8; hex 73757072656d756d; asc supremum;;

*** (2) TRANSACTION:
TRANSACTION 596041, ACTIVE 19 sec inserting
mysql tables in use 1, locked 1
3 lock struct(s), heap size 320, 2 row lock(s), undo log entries 1
MySQL thread id 10848, OS thread handle 0xa233b90, query id 251522 10.10.53.122 root update
insert into LockTest (order_id) values ('D19')
*** (2) HOLDS THE LOCK(S):
RECORD LOCKS space id 502 page no 4 n bits 72 index `idx_order_id` of table `test`.`LockTest` trx id 596041 lock_mode X
Record lock, heap no 1 PHYSICAL RECORD: n_fields 1; compact format; info bits 0
0: len 8; hex 73757072656d756d; asc supremum;;

*** (2) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 502 page no 4 n bits 72 index `idx_order_id` of table `test`.`LockTest` trx id 596041 lock_mode X insert intention waiting
Record lock, heap no 1 PHYSICAL RECORD: n_fields 1; compact format; info bits 0
0: len 8; hex 73757072656d756d; asc supremum;;

*** WE ROLL BACK TRANSACTION (2)

简单分析上面的场景先删除再插入的sql是hibernage保存集合关联的处理方式。delete语句删除不存在且删除的order_id大于现有表中的所有order_id,所以delete语句会使用next-key锁住(当前最大-无穷大)

lock_id	lock_trx_id	lock_mode	lock_type	lock_table	lock_index	lock_space	lock_page	lock_rec	lock_data
596133:502:4:1	596133	X	RECORD	`test`.`LockTest`	idx_order_id	502	4	1	supremum pseudo-record
596134:502:4:1	596134	X	RECORD	`test`.`LockTest`	idx_order_id	502	4	1	supremum pseudo-record

比较奇怪的是为啥两个事务都拿到了相同区间的(当前最大-无穷大)的X锁。不过换成read-commited级别后就没死锁了。

终于在官方文档找到答案, 区间锁只是用来防止其他事务在区间中插入数据，区间x锁 与区间S锁效果是一样的。也就是说不会因为两个事务都用加相同区间锁而相互等待的

https://dev.mysql.com/doc/refman/5.1/en/innodb-record-level-locks.html

Gap locks in InnoDB are “purely inhibitive”, which means they only stop other transactions from inserting to the gap. Thus, a gap X-lock has the same effect as a gap S-lock.

当两个事务拿到相同区间锁后，就会阻止对方忘区间内做insert操作。所以第一个事务insert会阻塞,第二个事务会提示死锁