Mysql：select ... for update

I met you these days select … for update Of sql sentence , Decided to sort it out mysql Two locking mechanisms of .
Mysql Database has two kinds of locks , One is shared lock , One is exclusive lock .

Shared lock （ Read the lock ,S lock ）

select … from … lock in share mode

Multiple transactions share a lock , however Only read , Do not modify . When a transaction gets the shared lock and the lock is not released , Another transaction cannot modify the locked data .

The biggest feature of shared locks is that they can be shared , Multiple transactions can acquire locks and read data at the same time , also , When the lock is not released , Data cannot be modified , This can avoid the problems of dirty reading and non repeatable reading of the database .

Exclusive lock ( The mutex , Write lock ,X lock )

select … for update;
Exclusive lock Cannot be shared by multiple transactions , If a transaction acquires an exclusive lock of a row of data , Then other transactions cannot obtain the lock of this row of data , Including shared lock and exclusive lock . Transactions that obtain exclusive locks can read and modify data , After the transaction is committed , Release the lock .

demonstration

Create a table for testing

CREATE TABLE `emipe_user_entity` (
  `id` varchar(50) CHARACTER SET utf8mb4 COLLATE utf8mb4_bin NOT NULL COMMENT ' Primary key ',
  `user_name` varchar(50) CHARACTER SET utf8mb4 COLLATE utf8mb4_bin DEFAULT NULL COMMENT ' user name ',
  `pass_word` varchar(50) CHARACTER SET utf8mb4 COLLATE utf8mb4_bin DEFAULT NULL COMMENT ' User password ',
  `status` int(11) DEFAULT NULL COMMENT ' User state ',
  PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin;

Use InnoDB Storage engine ,InnoDB Both row level locks are supported , It also supports table level locks , Row level locks are used by default
MySQL InnoDB The engine defaults to update,delete,insert The statement will automatically add an exclusive lock to the data involved .

Store data for the test form

INSERT INTO `emipe_user_entity` VALUES ('1', ' Little red riding hood ', '123456', 0);
INSERT INTO `emipe_user_entity` VALUES ('2', ' A passer-by ', '523456', 0);

Shared lock

Business A： Get shared lock , But don't commit the transaction

begin;
select * from emipe_user_entity where id = 1 lock in share mode;

Business B： Get shared lock , You can query the data

select * from emipe_user_entity where id = 1 lock in share mode;

Business C： Try modifying data with shared locks , Report errors

update emipe_user_entity set pass_word = '222222' where id = 1;

result ：

Business C First, we will wait for the shared lock to be released , After the lock is released , You can modify the modified data , Due to transaction A The lock has not been released , After a long wait , Business C Throw the wrong ：
Lock wait timeout exceeded;try restarting trasacrtion. Waiting for lock timeout ; Try to restart this transaction

Exclusive lock

Business A： Get the exclusive lock to query , The transaction does not commit

begin;
select * from emipe_user_entity where id = 1 for update;

Business B： Try to get an exclusive lock , Blocked

select * from emipe_user_entity where id = 1 for update;

Business C: Attempt to acquire the Shared lock , Blocked

select * from emipe_user_entity where id = 1 lock in share mode;

Business D： Without a lock , Can query data

select * from emipe_user_entity where id = 1;
--  The query is successful 

Ordinary query statements do not have any locking mechanism .