Detailed explanation of MySQL and mvcc and the difference between RC and RR for snapshot reading

One 、 What is? MVCC

When we operate the database, we always these four categories  

Read

 

Reading and writing

 

Write to read

 

Write about

, Reading must have no problem with task data , But students who know something will know , Reading and writing 、 Write operations can easily lead to data inconsistencies .

Before that, the common way to solve such problems was

Lock

, You can tell by the name that this is a very complicated 、 Performance intensive operation , So the great gods are not satisfied with this operation , Thus in MySQL It's realized MVCC.

MVCC Not at all MySQL Unique , It's an idea , Baidu Encyclopedia explains as follows

Multi-Version Concurrency Control  Multi version concurrency control ,MVCC  It's a method of concurrency control , Generally in the database management system , Realize concurrent access to database ; Implement transaction memory in programming language .

MVCC There are some key words , Understand these keywords , You will understand what is MVCC.MVCC Is to solve the problem of reading and writing 、 Data inconsistency caused by writing and reading , The writing problem still needs to be locked to solve .

So we can use it  MVCC +  lock （ Optimism lock / Pessimistic locking ） To solve all the problems .

Two 、 The current reading 、 Read the snapshot

Current reading is to read the latest data , In order to ensure that the latest and accurate data is read , So it will lock when reading , Prevent other things from operating .

Snapshot reading is an unlocked way , When a thing wants to operate the database , Will form a snapshot on the basis of this thing , Other operations are to read this snapshot .

MVCC It is based on snapshot reading , stay MySQL The snapshot reading is based on these key points

• Three hidden parameters  （DB_ROW_ID  Hide primary key id、DB_ROLL_PTR  rollback pointer 、DB_TRX_ID  thing id）

• undo log  journal

• read view

3、 ... and 、 Hide fields

If we had a watch , There are two fields ,name、age, But in fact, the data in our table is like this

3-1、 Hide primary key

6byte, Implied self increasing ID（ Hide primary key ）, If the data table has no primary key ,InnoDB Will automatically use DB_ROW_ID Generate a clustered index

Cluster index

： Data storage and indexing exist together , Both logically and physically , A table can only have one clustered index .

notes ： Understanding clustering index can be a good understanding of MySQL Index rules for , Those who are interested can take a look at this  

MySQL Index details

3-2、 thing id

Record this record of the last operation id

3-3、 rollback pointer

rollback pointer , Point to the previous version of this record （ Store in rollback segment in ）, Used in conjunction with the following  undo log.

Four 、undo log

undo log  There are two kinds of logs

• insert undo log
    The database generates when inserting data , Only in
  Current things
  Only useful when rolling back , So it's useless when the current thing is over , It will be deleted .

• update undo log
    The database is updating 、 Generated when deleting , Except that the current thing will use , It will also be used when reading snapshots , So you can't delete , Only if a fast read or transaction rollback does not involve the log , The corresponding log will be
  purge
  Thread unified cleaning

Suppose our hidden primary key is from 1、2、3…,  Transaction primary key 、 The rollback pointer is also generated in this way  （ In fact, it's not such a rule ）, So let's operate on the above table , The following chain structure will be formed .

insert data

It is revised as Zhang San and Li Si

Change the age to 25

notes ： The pointer here points to this data , Not inside   Primary key .

5、 ... and 、Read View

What is read view ？ The operation of database is carried out by multiple things at the same time , Read and write . If there are two things at present ,A Things read ,B Things are updating data .

stay A When things start , A snapshot of the current database is formed , Record and maintain the current active transactions of the system ID.read view  It's mainly used for visibility judgment , It will judge the data of each record , This data may be real data , It could be undo log  Data in .

read view  With a visibility algorithm , To determine whether it is currently reading real data , still undo log The data of . It can be simply understood here read view  Internally maintains a thing id list , There are maximum and minimum values , Can judge other things id Whether it is within this visible range .

N、 Other

N-1、 The snapshot is read in RC and RR The difference between

• RC （read-committed） Read submitted ,  May lead to   It can't be read repeatedly 、 Fantasy reading

• RR （repeatable-read） Repeatable , May lead to   It can't be read repeatedly

Fantasy reading

 ：  thing A Query the database and find out 20 Data , Things then B Deleted 2 Data , At this time things A Then go to the query and find that only 18 The article , Which creates hallucinations .

We know that RR There will be no unreal reading below the level , The reason why there is no unreal reading , Is a snapshot read in RC and RR The generation strategies under are different .

RC Under isolation level , Every snapshot read generates and gets the latest Read View; And in the RR Under isolation level , The first snapshot read in the same transaction is created Read View,  The subsequent snapshot reads all capture the same Read View.