After meeting a full stack developer from Tencent, I saw what it means to be proficient in MySQL tuning

One 、 Preface

MySQL Tuning for many programmers , It's a very difficult problem , Most of the cases are due to the database problems and processing ideas are not clear . It's going on MySQL Before optimization, we must know that MySQL Query process of , A lot of query optimization work actually follows some principles MySQL The optimizer of can run as expected .
Just yesterday, I took time out of my busy schedule to interview a full stack developer from Tencent , I said it ：MYSQL How to start with performance tuning ？ His answer is ： Based optimization 、 Philosophy of optimization 、 Optimize requirements 、 Optimization ideas 、 Storage engine layer 、 Database optimization 、 Wait for the details , Well, I admit I lost .
But I seriously doubt that he came prepared , Otherwise, no one can remember so clearly and orderly , Sure enough , He told the truth after he took office ;
There is a book recommended by many bosses in their team 【MySQL Performance optimization and high availability architecture practices 】 file

Well, don't say more , Xiaobian shares it directly , After learning, you can also write on your resume “ Master MySQL performance optimization ”, At the same time, I hope you can hang and beat other interviewers ;
Catalog

• InnoDB Storage cited the entire architecture
• MySQL Transactions and locks
• SLQ Language performance optimization

 

 

• MySQL Server optimization
• MySQL Performance monitoring
• MySQL Master slave copy
• PXC Highly available solutions

 

 

• be based on MHA Realized MySQL. Auto fail over cluster
• MySQL Group Replication
• Keepalived+ High availability architecture of dual master replication
• Database sub database sub table and middleware introduction
• Mycat Middleware details

 

MySQL Architecture introduction
 

The current industry MySQL Mainstream branch versions have Oracle The official version of MySQL、 Percona Server、MariaDB. Next, let's look at the characteristics of each branch .
 

　InnoDB Storage engine architecture
 

from MySQL 5.5 Version start ,InnoDB Is the default table storage engine , It is characterized by supporting transactions 、 Support data row lock 、 the Multi version concurrency MVCC、 Support foreign keys .InnoDB The architecture of the storage engine is shown in the figure 2-1 Shown , Including memory pool 、 Background thread and underlying data file
 

 

MySQL Transactions and locks
 

MySQL Transaction isolation level
Transactions also satisfy isolation through the locking mechanism . stay InnoDB In the storage engine , There are different levels of isolation , They have different isolation .
 

Lock wait
Lock waiting refers to the lock generated during a transaction , Other transactions need to wait for the previous transaction to release its lock to occupy the resource . If the transaction has not been released , You need to keep waiting , Until the lock waiting time is exceeded , An error of waiting timeout will be reported . stay MySQL Pass through innodb_lock_wait_timeout Parameter to control the lock waiting time , The unit is seconds . Pictured 3-9 Shown , You can do this by saying show variables like '%innodb_lock_wait%' To see the lock wait timeout .
 

SQL Sentence performance optimization
 

Speaking of SQL Sentence performance optimization , I believe everyone knows some simple skills ： Don't use SELECT *、 Don't use NULL Field 、 Using indexes properly 、 Select the appropriate data type for the field, etc . Do you really understand these optimization techniques ？ Do you understand how it works ？ This chapter starts from the perspective of theory and practice , Explain the principles behind these optimization suggestions .
 

Create high-performance indexes
The index is to improve MySQL An important way to query performance . Try to avoid thinking about adding indexes afterwards , Because you may need to monitor a lot of SQL To locate the problem , And the time needed to increase the index must be much longer than the time needed to increase the index initially .
 

MySQL Server comprehensive optimization
 

Most use Linux Large and medium-sized Internet websites with operating systems are using MySQL As a back-end database service , So how to optimize MySQL Server is what we want to study . Now? MySQL 5.7 Version for multi-core CPU、 Solid state disk 、 The lock mechanism has better optimization . in addition ,MySQL 5.7 The version has improved the optimizer a lot , such as MySQL 5.7 Of in Sentence query can use index range scan The way ,Union all No more temporary tables , Sorting efficiency has also been improved . We from MySQL 5.7 Version of the storage engine enhancements , Hardware 、 operating system 、 Configuration parameter optimization 、 The design specification is optimized in several aspects MySQL The server
 

MySQL Performance monitoring
 

MySQL Accepted by more and more enterprises . With the development of the enterprise ,MySQL Storage data is expanding , Performance analysis 、 Monitoring and early warning is very important . In some cases , Usually a set of MySQL monitor / Graphical tools , And then according to MySQL Monitor the information provided by the panel to perform further tuning .
 

 

MySQL Master slave copy
 

MySQL The master-slave replication function of is built on MySQL High availability of database 、 High performance application foundation , It can be used to share the read load of the main database , It is also highly available HA And so on . Master-slave replication means that data can be copied from one MySQL The database server master node replicates to another one or more MySQL Database server slave node . Master slave replication can be used for real-time data backup 、 Read / write separation 、 High availability HA And so on .
 

 

PXC Highly available solutions
 

 

 

be based on MHA Realized MySQL Auto fail over cluster
 

MHA It's an excellent set of actions MySQL High availability software for failover and master-slave promotion in high availability environment . It's made up of two parts ：MHAManager（ The management node ） and MHA Node（ Data nodes ）, Pictured 9-1 Shown .MHA Manager Can be independently deployed on a separate machine to manage multiple master-slave Primary and secondary replication clusters , It can also be deployed in one slave From the node .
 

MySQL Group Replication
 

For a long time MySQL Officials lack native MySQL Clusters are flexible and can provide solutions with strong consistency , So third party companies are based on Galera Agreed Percona XtraDB Cluster（PXC） Accumulated many customer cases . 2016 year 12 month 12 Japan Oracle Released MySQL Group Replication The first of GA edition , And provide their own comparison Galera Performance testing .MGR As a star product promoted by the government , Indeed, it has the same old brand products as the market （ Such as PXC） Eligibility for competition .MGR Is based on Paxos High availability solution of distributed consistency protocol , Perfectly solved MySQL Defects in consistency and high availability . future MGR The probability of the plan will become the bank 、 insurance 、 High availability and disaster recovery solutions for financial services such as securities .
 

 

Keepalived+ High availability architecture of dual master replication
 

 

All right, that's it , I hope the little partner who gets this document can cut down his favorite offer