Principle of Redis Sentinel

First of all, the sentinel mode is a special mode, which is an implementation of Redis high availability.First of all, the sentinel is an independent process that can monitor, notify, and automatically failover the Redis instance.

Actually, each sentinel node performs heartbeat monitoring (including all redis instances and sentinel peers) through ping every second, and judges whether the node is online based on the reply.

If a sentinel thread finds that the main library does not respond to this PING within the given time ( down-after-milliseconds), then this sentinelThe code> thread thinks that the main library is unavailable, this situation is called "subjective failure" (ie SDOWN); this situation generally does not cause immediate failover automatically, but when multiple sentinel threadsIt is indeed found that the main library is unavailable and exceeds the in the configuration item sentinel monitor mymaster {#ip} {#port} {#number} in sentinel.conf#number (the gossip protocol is actually used here), generally the remaining sentinel threads will be responsible for the objective offline of the main library and automatic failover through the RAFT algorithm., this situation is called "objective failure" (ie ODOWN).

The specific process is shown in the figure below:

SentinelMode configuration items