Mysql database master-slave synchronization, consistency solution

---

Method 1 Semi-synchronous replication

Between asynchronous replication and synchronous replication , The master database will not return to the client immediately after executing the transaction submitted by the client ,
Instead, wait until at least one receives from the library and writes to redo log in , Will return to the client , As opposed to asynchronous replication , Semi-synchronous replication improves data security

• The principle of semi synchronous replication
  The transaction is finished in the main database binlog after , You need to return a received... From the library , To return to the client
  Make sure that after the transaction is committed binlog Transfer to at least one slave Library
  There is no guarantee that this transaction will be applied from the database binlog
  advantage ： Make use of database native function , Relatively simple
  shortcoming ： The write request latency of the main library will increase , Throughput will decrease

Method 2 database middleware

• All read / write requests go through the database middleware , Usually, write requests go to the main database , Read request from library
  Record all routes to the write library key, If in the window of master-slave synchronization （ hypothesis 500ms）, There's a read request coming ,
  At this time, the slave database may be old data , You need to key Read requests on are routed to the main database access
  After the master-slave synchronization time is over , Corresponding key Read requests continue to be routed to the slave library
  advantage ： Can guarantee absolute consistency ,
  shortcoming ： The cost of database middleware is high

Method 3 Cache record write key Law

• Write requests
  If key To write , It was recorded that cache in , Set the cache timeout for master-slave synchronization , such as 500ms
  Then modify the master database
  Read request
  Check the cache first , Corresponding key There's no data
  If there is , Description cache hits , This key A write operation has just occurred , At this point, you need to route the request to the main database to read the latest data
  without , Description cache misses , Illustrate this key There has been no write operation recently , At this point, the request is routed to the slave Library , Continue read write separation
  advantage ： Relative to database middleware , Lower cost
  shortcoming ： To ensure consistency , Introduced cache Components , And when reading and writing the database, there are more cache operations

Please correct me if there is any mistake