CODEWITHSUNDEEP
redisredis-sentinel
The crazy knight
The crazy knight

Reputation: 19

clarification on redis-sentinel read & write

I am trying to write a golang application and its read intensive so I am using redis for caching.

I am planning to use redis sentinel and thinking of making write(to redis) through master and read (from redis) using 3 slaves.

But I am not getting any configs to specifically mention read from slaves and write from master.

I read an article which says by default sentinel will use master for writing and slave for reading, is it true? Can someone confirm please

need to confirm if redis-sentinel by default redirects all the write to master and read to slave

Upvotes: 0

Views: 77

Answers (1)

Adam Marshall
Adam Marshall

Reputation: 7705

Welcome to Stack Overflow!

Redis Sentinel cannot be used in the manner you described. Sentinels are monitors, not proxies. They perform two general functions:

  1. They monitor the Redis nodes, and failover to a new primary when the current primary becomes unavailable
  2. Tells you which node is the current primary

In a proper Sentinel deployment, client applications would connect to a Sentinel and perform a query to identify the current primary node, then it's the client's responsibility to send writes to that node (and if desired, send reads to any of the replica nodes). However, since a failover could happen at any time, you'd have to re-query the Sentinel frequently (this is especially important if a split-brain scenario occurs).

Alternatively, you could have a separate process query the Sentinels that tracks the previous primary node, so that when a failover is detected, it updates a DNS record that points to the current primary node. The same can be done with a DNS record for a replica node.

Upvotes: 0

Related Questions