MySQL transactions and locks (V)

Database notes

• Mysql Index bottom layer （ One ）
• Explain Tools and index optimization （ Two ）
• SQL stay MySQL Execution process in （ 3、 ... and ）
• MySQL Index optimization （ Four ）
• MySQL Transactions and locks （ 5、 ... and ）
• MVCC and BufferPool（ 6、 ... and ）

Transactions and locks

Business

Definition

The business is made up of a group of SQL A logical processing unit made up of statements , have ACID Four properties

attribute

• A Atomicity ： In business all operation Agreement

• C Uniformity ： Operations within a transaction == As a result of Data modification Should agree ==

• I Isolation, ： During transaction processing ** The intermediate state is invisible to the outside ** Of ; External state changes are also invisible to transactions .

• D persistence ： After the transaction completes , Yes == The modification of data is permanent .==

Problems caused by concurrent transaction processing

• Dirty write —— Last update covers Updates made by other firms

• Dirty reading —— Business A Read the business B Modified but not submitted The data of 【 Does not satisfy the isolation 】

• Don't reread —— Business A Internal same query SQL Inconsistent results 【 Does not satisfy the isolation 】

• Fantasy reading —— Business A Read transaction B New data submitted 【 Does not satisfy the isolation 】

Transaction isolation level

Definition

Databases have different isolation levels , The default isolation level is Repeatable

• View the transaction isolation level of the current database :

  show variables like 'tx_isolation';
  

• Set the transaction isolation level ：

  set tx_isolation='REPEATABLE-READ';
  

classification

• Read uncommitted
• Read submitted
• Repeatable
• Serializable

Read uncommitted

set tx_isolation='read-uncommitted';

technological process ：

1. client A With client B Each setting is Read uncommitted
2. client B** Update table account,** but Do not commit transaction
3. client A Query table account, Read B Modified data
4. If at this time ,B Roll back ,A Read dirty data

Possible problems ：

Dirty reading 、 Don't reread 、 Fantasy reading

Read submitted

set tx_isolation = 'read-committed';

technological process ：

1. client A With client B Each setting is Read submitted
2. client B** Update table account,** Commit transaction
3. client A Query table account, Read B Modified data

Possible problems ：

Don't reread 、 Fantasy reading

Repeatable

set tx_isolation = 'repeatable-read';

adopt undo Log files for version control , Achieve isolation

technological process ：

1. client A With client B Each setting is Repeatable
2. client B** To the table account Add a new one id by 100 The data of ,** Commit transaction
3. client A Query table account, You can go to id by 100 The data of

Possible problems ：

Fantasy reading

Serialization

set tx_isolation = 'serializable';

adopt Lock To achieve serialization

technological process ：

• Case one

  1. client A With client B Each setting is Serializable
  2. client A perform Inquire about id by 1 The sentence of 【 add Row lock 、 Read the lock 、 Clearance lock 】
  3. client B here to update id by 1 The statement of will Block waiting , to update id by 2 The statement can be executed normally

• The second case

  1. client A With client B Each setting is Serializable
  2. client B here Insert id by 1 The sentence of 【 add Row lock 、 Write lock 】
  3. client A perform Inquire about id by 1 The statement of will Block waiting

Lock details

Definition ：

Locks are used to control Critical section resources For who

The following lock categories , The database implements Row lock and table lock

Lock classification

• performance —— Optimism lock and Pessimistic locking

• Operation type —— Read the lock and Write lock

• Lock granularity —— Row lock and Table locks and Clearance lock

Various types of locks can be crossed , Not necessarily mutually exclusive

Optimism lock

Definition ：

Each thread uses the version number comparison to achieve the goal Critical section resources

Pessimistic locking

Definition ：

Each thread vies for the lock Critical section resources

Read the lock

Definition ：

Shared lock S

For the same data , If ** Threads A Get the read lock **

Multiple thread read operations can be performed simultaneously , Except for threads A Other thread writes will be blocked .

It's a pessimistic lock

Write lock

Definition ：

Exclusive lock X

For the same data , If ** Threads A Get the write lock **

Only the thread A Be able to read and write data , The read and write operations of other threads will be blocked .

Row lock

Definition ：

Lock one row of data per operation , Smaller particle size .

classification ：

• Row lock + Read the lock —— Lock the data in this row , Other threads' writes to this row are blocked

  select * from test_innodb_lock where a = 2 lock in share mode;
  

• Row lock + Write lock —— Lock the data in this row , Other threads' read and write operations to this row are blocked

  select * from test_innodb_lock where a = 2 for update;
  

expenses ：

because The row lock should be positioned to a certain row , So it locks slowly , The cost will increase , And there will be a deadlock .

Lock escalation ：

InnoDB The row lock of is the lock added for the index , Instead of recording the lock added . If Update the non indexed field or the index becomes invalid , Row locks may be upgraded to table locks .

Line lock analysis ：

#  Inquire about InnoDB Row lock contention 
show status like 'innodb_row_lock%';

• Innodb_row_lock_time_avg （ Average waiting time ）

• Innodb_row_lock_waits （ Total waiting times ）

• Innodb_row_lock_time（ The total waiting time ）

Optimize according to the situation

Table locks

Definition ：

Lock one row of data per operation , The particle size is large , Applicable to whole table data migration .

classification ：

• Table locks + Read the lock —— Lock the table data , Other threads' writes to the table are blocked

  lock table  The name of the table  read;
  

• Table locks + Write lock —— Lock the table data , Other threads' read and write operations to the table are blocked

  lock table  The name of the table  write;
  

Clearance lock

Definition ：

Clearance lock , Is to lock the gap between two lines 【 Section 】 Only in Repeatable To take effect

Example ：

Existing interval id by (3,10) (10,20) (20,+∞)

• stay session1 In the implementation of

  #  UPDATE statement 
  update account set name = 'zhuge' where id > 8 and id <18;
  ## MySQL A clearance lock will be added   Lock the  (8,18)
  ##  because 8 Fall to the original (3,10) The range of ,18 Fall in the (10,20) The range of , So the locked interval is  (3,20]
  

• Lock according to clearance , Lock the (3,20]

Add a write lock to the interval , other session This interval cannot be inserted 、 Delete 、 Modify the operating

View the related data table of system library lock

##  see INFORMATION_SCHEMA Data table related to system library lock 
#  View transactions 
select * from INFORMATION_SCHEMA.INNODB_TRX;
#  Check the lock 
select * from INFORMATION_SCHEMA.INNODB_LOCKS;
#  View lock wait 
select * from INFORMATION_SCHEMA.INNODB_LOCK_WAITS;
#  Release the lock ,trx_mysql_thread_id It can be downloaded from INNODB_TRX See in the table 
kill trx_mysql_thread_id
#  View lock wait details 
show engine innodb status\G;

summary

• MyISAM stay DQL front , Will give == The tables involved are read locked .== stay DML It will be given before ** The tables involved are locked .** 【MyISAM Table locks 】
• InnoDB stay DQL front ,（ Non serial isolation level ） No locks . stay DML It will be given before == The rows involved are read locked .==【InnoDB Row lock 】
• When performing updates, try to operate with original values update Table name set Field 1 = Field 1 - 50 where Conditions
• Data is retrieved by index , Avoid row lock escalation
• Minimize the range of search conditions , Avoid gap locks
• ** Control transaction size ,** It may be locked SQL Execute after the transaction