Cache architecture scheme of ten million level shopping cart system

​ Author's brief introduction ： Xiaoming Java Asking the way , Senior R & D Engineer of a large financial Internet company , Good at orders / High security in the trading field / You can use / Concurrent / Performance architecture design and implementation , Focus on the research of computer and financial technology
If this article is good , Please pay attention. 、 give the thumbs-up 、 Collect three companies to support bloggers

Introduction to this article ： This paper mainly introduces redis Design architecture scheme in ten million level system , Including the main architecture design 、 Cache consistency scheme 、 Big value Treatment plan and redis Current limiting and fault recovery degradation scheme

Cache architecture main frame composition ：

The main architecture includes cache clusters 、 Cache current limiting 、 Cache size value Handle 、HotKey Detection system 、redis Failback degradation and redis-mysql Data synchronization

Cache consistency solution ：

One 、 For all DB Operations do not add specific operations to delete the cache , But through canal monitor binlog The way
Two 、 After the data is confirmed to have been submitted to the database , Change through monitoring , After parsing the corresponding data , Filter out those that are not added, deleted or modified
binlog
3、 ... and 、 Then, the related tables, key fields and cache prefixes that need to handle data consistency are configured through constant classes key, Assemble the cache to be deleted key. And through mq Of ack Mechanism to ensure that the cache will be deleted .

Cache throttling solution ：

prevent redis After the crash , The system is not working properly , So we need to do a downgrade .
system operation redis In general, all the methods of RedisCache and Redislock Two classes to handle , So we went through AOP Cut the way , Make an aspect of all the methods in these two classes , If you execute here redis During operation ,redis Hang up , Connection failed , Then the method will throw exceptions , We are on the method of section processing , The catching , Then record it as redis Hang up is still a temporary fluctuation of the network

  Big Value Treatment scheme ：

One 、 Built redis colony
Two 、 stay canal Monitoring has been created in , Store in mysql in redis_large_key_log Tabular canal instance
3、 ... and 、 In the early hours of every day 3 spot , Will Redis The big of the middle key（value） data , adopt canal instance Send to MQ
Four 、 consumption redis_large_key_log Tabular binlog data , The data contains Redis The big key（value） Information
5、 ... and 、 Will be big value Cut into several key, When returning, splice and assemble

​