Redis cerebral fissure

summary

Cleft brain problem , It sounds very high-end . In fact, it is the master-slave topology , The master node is mistakenly switched , As a result, the original master node is still processing requests , A phenomenon that results in the loss of data in the new master node . Generally speaking , The master-slave structure has the problem of brain crack .

Cleft brain problem

explain :redis The master-slave structure is normal , Normal synchronization , Client requests are processed normally .

explain : master Due to the large amount of data processed , Or network congestion , No, ack Heartbeat bag , The sentry goes offline with the main node by mistake , This is the time
Elected a new master, And this time is old master Still dealing with set a 10 This request

explain : old master The network has been restored , It is found that new people have been elected master, Then the local data is cleared , Start full synchronization of new master data
, Lead to set a 10 This data is missing , new master I didn't know that this request had happened

Solution

1. To configure min-slaves-to