MySQL InnoDB事务的隔离级别有四级,默认是“可重复读”(REPEATABLE READ)。

  • 未提交读(READ UNCOMMITTED)。另一个事务修改了数据,但尚未提交,而本事务中的SELECT会读到这些未被提交的数据(脏读)。
  • 提交读(READ COMMITTED)。本事务读取到的是最新的数据(其他事务提交后的)。问题是,在同一个事务里,前后两次相同的SELECT会读到不同的结果(不重复读)。
  • 可重复读(REPEATABLE READ)。在同一个事务里,SELECT的结果是事务开始时时间点的状态,因此,同样的SELECT操作读到的结果会是一致的。但是,会有幻读现象(稍后解释)。
  • 串行化(SERIALIZABLE)。读操作会隐式获取共享锁,可以保证不同事务间的互斥。

四个级别逐渐增强,每个级别解决一个问题。

  • 脏读,最容易理解。另一个事务修改了数据,但尚未提交,而本事务中的SELECT会读到这些未被提交的数据。
  • 不重复读。解决了脏读后,会遇到,同一个事务执行过程中,另外一个事务提交了新数据,因此本事务先后两次读到的数据结果会不一致。
  • 幻读。解决了不重复读,保证了同一个事务里,查询的结果都是事务开始时的状态(一致性)。但是,如果另一个事务同时提交了新数据,本事务再更新时,就会“惊奇的”发现了这些新数据,貌似之前读到的数据是“鬼影”一样的幻觉。

借鉴并改造了一个搞笑的比喻:

  • 脏读。假如,中午去食堂打饭吃,看到一个座位被同学小Q占上了,就认为这个座位被占去了,就转身去找其他的座位。不料,这个同学小Q起身走了。事实:该同学小Q只是临时坐了一小下,并未“提交”。
  • 不重复读。假如,中午去食堂打饭吃,看到一个座位是空的,便屁颠屁颠的去打饭,回来后却发现这个座位却被同学小Q占去了。
  • 幻读。假如,中午去食堂打饭吃,看到一个座位是空的,便屁颠屁颠的去打饭,回来后,发现这些座位都还是空的(重复读),窃喜。走到跟前刚准备坐下时,却惊现一个恐龙妹,严重影响食欲。仿佛之前看到的空座位是“幻影”一样。

一些文章写到InnoDB的可重复读避免了“幻读”(phantom read),这个说法并不准确。

做个试验:(以下所有试验要注意存储引擎和隔离级别)

mysql> show create table t_bitfly\G;
CREATE TABLE t_bitfly (
id bigint(20) NOT NULL default '0',
value varchar(32) default NULL,
PRIMARY KEY (id)
) ENGINE=InnoDB DEFAULT CHARSET=gbk

mysql> select @@global.tx_isolation, @@tx_isolation;
+-----------------------+-----------------+
| @@global.tx_isolation | @@tx_isolation  |
+-----------------------+-----------------+
| REPEATABLE-READ       | REPEATABLE-READ |
+-----------------------+-----------------+

试验一:

t Session A                                   Session B
|
| START TRANSACTION;            START TRANSACTION;
|
| SELECT * FROM t_bitfly;
| empty set
|                                                             INSERT INTO t_bitfly
|                                                             VALUES (1, 'a');
|
| SELECT * FROM t_bitfly;
| empty set
|                                                             COMMIT;
|
| SELECT * FROM t_bitfly;
| empty set
|
| INSERT INTO t_bitfly VALUES (1, 'a');
| ERROR 1062 (23000):
| Duplicate entry '1' for key 1
v (shit, 刚刚明明告诉我没有这条记录的)

如此就出现了幻读,以为表里没有数据,其实数据已经存在了,傻乎乎的提交后,才发现数据冲突了。

试验二:

t Session A                                     Session B
|
| START TRANSACTION;              START TRANSACTION;
|
| SELECT * FROM t_bitfly;
| +------+-------+
| | id   | value |
| +------+-------+
| |    1 | a     |
| +------+-------+
|                                                            INSERT INTO t_bitfly
|                                                            VALUES (2, 'b');
|
| SELECT * FROM t_bitfly;
| +------+-------+
| | id   | value |
| +------+-------+
| |    1 | a     |
| +------+-------+
|                                                            COMMIT;
|
| SELECT * FROM t_bitfly;
| +------+-------+
| | id   | value |
| +------+-------+
| |    1 | a     |
| +------+-------+
|
| UPDATE t_bitfly SET value='z';
| Rows matched: 2  Changed: 2  Warnings: 0
| (怎么多出来一行)
|
| SELECT * FROM t_bitfly;
| +------+-------+
| | id   | value |
| +------+-------+
| |    1 | z     |
| |    2 | z     |
| +------+-------+
|
v

本事务中第一次读取出一行,做了一次更新后,另一个事务里提交的数据就出现了。也可以看做是一种幻读。


那么,InnoDB指出的可以避免幻读是怎么回事呢?

http://dev.mysql.com/doc/refman/5.0/en/innodb-record-level-locks.html

By default, InnoDB operates in REPEATABLE READ transaction isolation level and with the innodb_locks_unsafe_for_binlog system variable disabled. In this case, InnoDB uses next-key locks for searches and index scans, which prevents phantom rows (see Section 13.6.8.5, “Avoiding the Phantom Problem Using Next-Key Locking”).

准备的理解是,当隔离级别是可重复读,且禁用innodb_locks_unsafe_for_binlog的情况下,在搜索和扫描index的时候使用的next-key locks可以避免幻读。

关键点在于,是InnoDB默认对一个普通的查询也会加next-key locks,还是说需要应用自己来加锁呢?如果单看这一句,可能会以为InnoDB对普通的查询也加了锁,如果是,那和序列化(SERIALIZABLE)的区别又在哪里呢?

MySQL manual里还有一段:

13.2.8.5. Avoiding the Phantom Problem Using Next-Key Locking (http://dev.mysql.com/doc/refman/5.0/en/innodb-next-key-locking.html)

To prevent phantoms, InnoDB uses an algorithm called next-key locking that combines index-row locking with gap locking.

You can use next-key locking to implement a uniqueness check in your application: If you read your data in share mode and do not see a duplicate for a row you are going to insert, then you can safely insert your row and know that the next-key lock set on the successor of your row during the read prevents anyone meanwhile inserting a duplicate for your row. Thus, the next-key locking enables you to “<span style="padding-top: 0px; padding-right: 0px; padding-bottom: 0px; padding-left: 0px; margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px;">lock</span>” the nonexistence of something in your table.

我的理解是说,InnoDB提供了next-key locks,但需要应用程序自己去加锁。manual里提供一个例子:

SELECT * FROM child WHERE id > 100 FOR UPDATE;

这样,InnoDB会给id大于100的行(假如child表里有一行id为102),以及100-102,102+的gap都加上锁。

可以使用show innodb status来查看是否给表加上了锁。

再看一个实验,要注意,表t_bitfly里的id为主键字段。实验三:

t Session A                                       Session B
|
| START TRANSACTION;                START TRANSACTION;
|
| SELECT * FROM t_bitfly
| WHERE id<=1
| FOR UPDATE;
| +------+-------+
| | id   | value |
| +------+-------+
| |    1 | a     |
| +------+-------+
|                                                           INSERT INTO t_bitfly
|                                                           VALUES (2, 'b');
|                                                           Query OK, 1 row affected
|
| SELECT * FROM t_bitfly;
| +------+-------+
| | id   | value |
| +------+-------+
| |    1 | a     |
| +------+-------+
|                                                           INSERT INTO t_bitfly
|                                                           VALUES (0, '0');
|                                                           (waiting for lock ...
|                                                           then timeout)
|                                                           ERROR 1205 (HY000):
|                                                           Lock wait timeout exceeded;
|                                                           try restarting transaction
|
| SELECT * FROM t_bitfly;
| +------+-------+
| | id   | value |
| +------+-------+
| |    1 | a     |
| +------+-------+
|                                                           COMMIT;
|
| SELECT * FROM t_bitfly;
| +------+-------+
| | id   | value |
| +------+-------+
| |    1 | a     |
| +------+-------+
v

可以看到,用id<=1加的锁,只锁住了id<=1的范围,可以成功添加id为2的记录,添加id为0的记录时就会等待锁的释放。

MySQL manual里对可重复读里的锁的详细解释:

http://dev.mysql.com/doc/refman/5.0/en/set-transaction.html#isolevel_repeatable-read

For locking reads (SELECT with FOR UPDATE or LOCK IN SHARE MODE),UPDATE, and DELETE statements, locking depends on whether the statement uses a unique index with a unique search condition, or a range-type search condition. For a unique index with a unique search condition, InnoDB locks only the index record found, not the gap before it. For other search conditions, InnoDB locks the index range scanned, using gap locks or next-key (gap plus index-record) locks to block insertions by other sessions into the gaps covered by the range.


一致性读和提交读,先看实验,实验四:

t Session A                                                    Session B
|
| START TRANSACTION;                             START TRANSACTION;
|
| SELECT * FROM t_bitfly;
| +----+-------+
| | id | value |
| +----+-------+
| |  1 | a     |
| +----+-------+
|                                                                        INSERT INTO t_bitfly
|                                                                                VALUES (2, 'b');
|                                                                        COMMIT;
|
| SELECT * FROM t_bitfly;
| +----+-------+
| | id | value |
| +----+-------+
| |  1 | a     |
| +----+-------+
|
| SELECT * FROM t_bitfly LOCK IN SHARE MODE;
| +----+-------+
| | id | value |
| +----+-------+
| |  1 | a     |
| |  2 | b     |
| +----+-------+
|
| SELECT * FROM t_bitfly FOR UPDATE;
| +----+-------+
| | id | value |
| +----+-------+
| |  1 | a     |
| |  2 | b     |
| +----+-------+
|
| SELECT * FROM t_bitfly;
| +----+-------+
| | id | value |
| +----+-------+
| |  1 | a     |
| +----+-------+
v

如果使用普通的读,会得到一致性的结果,如果使用了加锁的读,就会读到“最新的”“提交”读的结果。

本身,可重复读和提交读是矛盾的。在同一个事务里,如果保证了可重复读,就会看不到其他事务的提交,违背了提交读;如果保证了提交读,就会导致前后两次读到的结果不一致,违背了可重复读。

可以这么讲,InnoDB提供了这样的机制,在默认的可重复读的隔离级别里,可以使用加锁读去查询最新的数据。

http://dev.mysql.com/doc/refman/5.0/en/innodb-consistent-read.html

If you want to see the “freshest” state of the database, you should use either the READ COMMITTED isolation level or a locking read:
SELECT * FROM t_bitfly LOCK IN SHARE MODE;


结论:MySQL InnoDB的可重复读并不保证避免幻读,需要应用使用加锁读来保证。而这个加锁度使用到的机制就是next-key locks。

==================== 结尾 ====================

作者: bitfly. 转载请注明来源或包含本信息. 谢谢
链接: http://blog.bitfly.cn/post/mysql-innodb-phantom-read/