Myslq ten kinds of locks, an article will take you to fully analyze

MYSLQ Ten locks , An article will give you a full analysis

MySQL There are two core knowledge points , Index and lock . The previous articles have explained in detail MySQL Index implementation mechanism , Today, let's study together MySQL Lock of .

1 、 Why lock up ？

When multiple transactions operate on the same batch of data concurrently , If not locked , There is no guarantee of transaction isolation , Finally, the data is disordered .

Locking is to ensure the correctness of data under concurrent operations .

2 、 What are the categories of locks ？

According to the granularity of the lock, it can be divided into ： Table locks 、 Page lock 、 Row lock 、 Record locks 、 Clearance lock 、 Temporary key lock

According to the attributes of the lock, it can be divided into ： Shared lock 、 Exclusive lock

According to the locking mechanism, it can be divided into ： Optimism lock 、 Pessimistic locking

Let's introduce these locks in turn ：

Table locks ：

MyISAM and InnoDB All engines support table locking .

advantage ： Low overhead , Locked fast , A deadlock will not occur .

shortcoming ： Strong locking , High probability of lock conflict , Lowest degree of concurrency .

Lock mode ：

#  Yes user Read lock on the meter 
lock table user read;
#  At the same time user Read lock on the meter , Yes order Table write lock 
lock tables user read, order write;

When do I need a table lock ？

1. When you need to update most of the data in the table
2. Transactions involve multiple tables , Complex business logic , Adding table lock can avoid deadlock .

Page lock ：

advantage ： Cost and lock speed are between table lock and row lock .

shortcoming ： A deadlock occurs , Lock granularity is between table lock and row lock , The concurrency is average .

At present, only BDB The engine supports page locking , Few application scenarios .

Row lock ：

Only InnoDB The engine supports row locking , In addition, the lock is added to the index .

advantage ： Spending big , Lock the slow ; A deadlock occurs .

shortcoming ： The locking granularity is small , The probability of lock conflict is low , High concurrency .

In addition, record the lock 、 Clearance lock 、 Key lock belongs to row lock .

Record locks （Record Locks）：

That is, lock a record .

#  Yes id=1 The user of is locked 
update user set age=age+1 where id=1;

Clearance lock （Gap Locks）：

That is, lock a certain range , But the critical data that does not include the range .

#  Yes id Greater than 1 And less than 10 The user of is locked 
update user set age=age+1 where id>1 and id<10;

above SQL The lock range of is (1,10).

Temporary key lock （Next-Key Locks）：

It consists of record lock and clearance lock , Include both the record itself and the scope , Left open right closed interval .

#  Yes id Greater than 1 And less than or equal to 10 The user of is locked 
update user set age=age+1 where id>1 and id<=10;

Shared lock （ Also called read lock 、S lock ）：

effect ： Prevent other transactions from modifying the current data .

Lock mode ：

stay select Add... At the end of the statement lock in share mode keyword .

#  Yes id=1 The user adds a read lock 
select * from user where id=1 lock in share mode;

Exclusive lock （ Also known as write lock 、X lock ）：

effect ： Prevent other transactions from reading or updating the current data .

Lock mode ：

stay select Add... At the end of the statement for update keyword .

#  Yes id=1 The user of adds a write lock 
select * from user where id=1 for update;

Optimism lock ：

Always assume that others will not modify the current data , So every time the data is read, it will not be locked , Only when updating data version Judge whether others have modified the data ,Java Of atomic The classes under the package use optimistic locks （CAS） Realized .

It is suitable for reading more and writing less .

Lock mode ：

1. Read version

   select id,name,age,version from user id=1;
   

2. Update data , Judge version Have you ever modified .

   update user set age=age+1 where id=1 and version=1;
   

Pessimistic locking ：

Always assume that others will modify the current data , So every time you read , Always lock .

It is suitable for writing more and reading less .

Lock mode ：

#  Add read lock 
select * from user where id=1 lock in share mode;
#  Add write lock 
select * from user where id=1 for update;

This paper summarizes the knowledge points ：

Lock mode ：

#  Add read lock 
select * from user where id=1 lock in share mode;
#  Add write lock 
select * from user where id=1 for update;

This paper summarizes the knowledge points ：