[MySQL] lock

lock

A lock is a mechanism by which a computer coordinates multiple processes or threads to access a resource concurrently . In the database , Data is also a shared resource . How to ensure the consistency of data concurrent access 、 Validity is a problem that all databases must solve , Lock conflicts are also an important factor affecting the performance of concurrent database access .

• Global lock ： Lock all tables in the database .
• Table lock ： Lock the whole table with each operation .
• Row-level locks ： Each operation locks the corresponding row data .

Global lock

Global lock is to lock the whole database instance , After locking, the entire instance is in a read-only state , Subsequent inclusion DML sentence 、DDL The transaction commit statements of the statement will be blocked . Its typical use scenario is to make a logical backup of the whole database , Lock all tables , To get a consistent view , Ensure data integrity .

#  Add global lock 
flush tables with read lock;

#  The data backup 
# Windows Command line 
mysqldump -h39.105.19.217 -uroot –p123456 dataBaseName > scriptName.sql

#  Release the lock 
unlock tables;

characteristic ：

• If it is backed up on the primary database , The update cannot be performed during the backup , Business basically stopped .
• If you are backing up from a library , During the backup, the slave database cannot synchronize the binary logs from the master database , Will cause master-slave delay .

# InnoDB Storage engine , Consistent data backup without lock 
mysqldump --single-transaction -uroot –p123456 itcast > itcast.sql

Table lock

Table lock , Lock the whole table with each operation . Large locking size , The probability of lock conflict is high , Low concurrency . Apply to MyISAM、InnoDB Wait for the storage engine . For table lock , It is mainly divided into the following three categories ：

• Table locks
• Metadata lock （meta data lock, MDL）
• Intent locks

Table locks

• Read the lock （ Shared lock ）
  
  

• Write lock （ Exclusive lock ）
  
  

#  Add read lock 
lock tables tableName read;
#  Add write lock 
lock tables tableName write;
#  Release the lock 
unlock tables;

Conclusion ： Read locks do not block reads from other clients , But it will block writing . The write lock will block the reading of other clients , It will block the writing of other clients .

Metadata lock

MySQL5.5 Introduce metadata lock , When accessing a table, it will be locked automatically . Metadata can be simply understood as the table structure of a table , When a table involves uncommitted transactions , You cannot modify the table structure of this table . Metadata lock is to avoid DML And DDL Conflict , Ensure the correctness of reading and writing .

When adding to a table 、 Delete 、 Change 、 When checking the operation , Shared lock of Canadian dollar data lock . When changing the table structure , Exclusive lock of Canadian dollar data lock .

#  Check the metadata lock in the database 
select object_type,object_schema,object_name,lock_type,lock_duration from performance_schema.metadata_locks;

Intent locks

for fear of DML When executed , Conflict between row lock and table lock , stay InnoDB Intention lock is introduced into the storage engine , So that the table lock doesn't have to check every row
Whether the data is locked , Use intent lock to reduce the checking of table lock .

• Intention sharing lock （IS）： By statement select … lock in share mode add to . Compatible with shared locks of table locks , Exclusive locks with table locks are mutually exclusive .
• Intention exclusive lock （IX）： from insert、update、delete、select…for update add to . Shared locks and exclusive locks with table locks are mutually exclusive .
• Intent locks are not mutually exclusive , Once the transaction is committed , Intention sharing lock 、 The intentional exclusive lock will be automatically released .
  
  

#  Check the locking of intention lock and row lock 
select object_schema,object_name,index_name,lock_type,lock_mode,lock_data from performance_schema.data_locks;

Row-level locks

Row-level locks , Each operation locks the corresponding row data . The locking granularity is small , The probability of lock conflict is low , High concurrency . Apply to InnoDB In the storage engine .InnoDB The data is organized based on indexes , Row locking is achieved by locking the index items on the index , Instead of adding to the record
lock . For row level locks , It is mainly divided into the following three categories ：

• Row lock ： Lock a single line record , Prevent other transactions from... On this line update and delete. stay Read Commit、Repeatable Read Both isolation levels support .
  

• Clearance lock ： Lock index record gap （ This record is not included ）, Ensure that the index record gap remains unchanged , Prevent other transactions from going on in this gap insert, And then there's magic reading . stay Repeatable Read Support at isolation level .
  

• Temporary key lock ： Combination of row lock and clearance lock , Lock the data at the same time , And lock the gap in front of the data . stay Repeatable Read Support at isolation level .
  

Row lock

InnoDB The storage engine implements the following two types of row locks ：

• Shared lock （S）： Allow a transaction to read a line , Prevent other transactions from obtaining exclusive locks on the same dataset .
• Exclusive lock （X）： Allow to get transaction update data of exclusive lock , Prevent other transactions from obtaining shared locks and exclusions for the same dataset
  lock .
  
  
  When searching for a unique index , When performing equivalence matching on existing records , It will be automatically optimized as row lock .InnoDB The row lock of the storage engine is the lock added to the index , When retrieving without indexing , All records in the table will be locked , At this point, it will be upgraded to table lock .
  
  
  As soon as the client gets id by 1 This line of shared locks , Second, the client can obtain id by 3 The exclusive lock of this line , Because it's not the same row of data . And if client 2 wants to get id by 1 Exclusive locks in this line , It's going to be blocked , Think that shared locks and exclusive locks are mutually exclusive .
  
  When a client , perform update sentence , Would be id by 1 We'll lock our records . Client 2 , If so update Statement update id by 1 The data of , For id by 1 Data with exclusive lock , But client 2 will be blocked , Because exclusive locks are mutually exclusive . Until the client commits the transaction , Will release the row lock of this row , At this time, the client 2 is unblocked .
  

As soon as the client starts the transaction , And implement update sentence , to update name by Lily The data of , That is to say id by 19 The record of . Then update... In client 2 id by 3 The record of , But not directly , It's going to be blocked , Why? ？ Because at this time, the client is based on name When the field is updated name The field has no index , If there is no index , At this point, the row lock will be upgraded to a table lock （ Because row locks are locks on index entries , and name No index ）.

#  Check the locking of intention lock and row lock 
select object_schema,object_name,index_name,lock_type,lock_mode,lock_data from performance_schema.data_locks;

Clearance lock and temporary key lock

By default ,InnoDB Storage engine in Repeatable Read Transaction isolation level running , Search and index scan with keylock , To prevent unreal reading .

• Equivalent query on Index （ unique index ）, When locking records that do not exist , Optimized for clearance lock .
  

• Equivalent query on Index （ Non unique index ）, When traversing to the right, the last value does not meet the query requirements , Key lock degenerates into clearance lock .

InnoDB The storage engine uses B+ Tree index , Leaf nodes are ordered two-way linked lists . If , According to this secondary index, the query value is 18 The data of , And add a shared lock , We're just locking 18 Is this all right ？ Not at all , Because it's a non unique index , There may be more than one... In this structure 18 The existence of , therefore , Will continue to look back when locking , Find a value that does not meet the condition （ In the current case, that is 29）. It will be right at this time 18 Add temporary key lock , Also on 29 Lock the previous gap .

• Range query on Index （ unique index ）, Access to the first value that does not satisfy the condition .

The query criteria are id>=19, And add a shared lock . At this point, we can use the existing data in the database table , Divide the data into three parts ：[19]、(19,25]、(25,+∞]. So the database locking situation is ,19 With a row lock ,25 The key lock of （ contain 25 And 25 The gap before ）, Positive infinite presence
Key lock （ Positive infinity and the gap before ）.

Be careful ： The only purpose of a gap lock is to prevent other transactions from inserting gaps . Gap locks can coexist , A gap lock adopted by one transaction will not prevent another transaction from adopting a gap lock on the same gap .

Reference resources ：https://www.bilibili.com/video/BV1Kr4y1i7ru