Characteristics of MySQL InnoDB storage engine -- Analysis of row lock

First put a mind map , I can't help but use the free version of the template , After all, I still eat soil every day , No money for membership . Understand this picture , In fact, for innodb You have mastered the characteristics of the storage engine .

One 、mysql Common storage engines
myisam: Unsupported transaction , Table locks . Data files and index files are separated , Files can be compressed
innnodb： Support transactions , Row lock （ It can be upgraded to table lock ）, Support primary and foreign keys .
Today's main discussion innodb Storage engine for , Don't add more to others .

Two 、 The level of isolation of things
First, find a picture on the Internet , Come to a conclusion , stay mysql Of innodb Search engine , default RR Repeatable reading level , It can solve unreal reading . The picture is not accurate , Marked

In the figure, there are explanations for each case , Here I quickly skim over in a simple and clear way
Dirty reading ： Business A Read transaction B Uncommitted data A read B update Not submitted
It can't be read repeatedly ： Business A Read 2 Time ,2 The secondary data is inconsistent A read 2 Time ,B update Has been submitted
Fantasy reading ：A Range query read 2 Time ,B insert after （ Critical lock ： Opening left and closing right can solve , You can add a write lock to the current range search , Will result in the inability to insert data , Not twice
The range lookup result set is inconsistent ）

3、 ... and 、 Thinking based on isolation level
Ordinary scenes , Concurrency or low business requirements ： Our query statements are basically fast illumination MVCC
In this concurrent scenario , Ensure isolation , Then you can only lock it , Use the mutex of lock to control ,LBCC The current reading .
1. Use locks lbcc The current reading
insert
delect
update
// Row lock classification
select .....for update Write lock
select .....lock in share mode Read the lock
2. snapshot MVCC
ordinary select Statements are snapshots

Four 、innodb How to realize the next lock , Why can critical lock solve unreal reading ？
1.innnodb By default, an uplink lock is added to the index , If the query condition is not index , Sorry, it will be upgraded to table lock
2. Implementation and classification of row lock ：
Record locks ：
1. When sql Search the index data according to the method of finding equivalent matching
2. Hit data
Clearance lock ： Left open right open
1. When sql Search the index data according to the method of finding equivalent matching
2. When the data does not hit
Critical lock ： Left open right closed
1. When SQL When performing data retrieval on the index by search range
2. And when there is data hit

This is the end of the additional description of the figure , If you have any questions or different opinions, you are welcome to learn .

So far, we are talking about the extended locking mechanism in the isolation level of things
But I still want to share the communication behavior of things , Especially a very common scene , In the same class , Transaction method A Call non transactional methods B, The transaction is effective , And whether it's A still B If an exception occurs in the method, the transaction will be rolled back .

Have time to write code to verify , Here is a conclusion written by the author . You can reprint and view https://blog.csdn.net/qq_39355504/article/details/109893065