Support for Reader endpoint in AWS Elasticache cluster mode disabled? about redis-rb

[supercaracal]

There is only the latter way unless using sentinel or cluster mode.

w = Redis.new(url: 'redis://your-primary-endpoint')
r = Redis.new(url: 'redis://your-reader-endpoint')

w.set('foo', 1)
r.get('foo')

[supercaracal]

You can specify a reader endpoint like this:

r = Redis.new(url: 'rediss://your-reader-endpoint', reconnect_attempts: 10, reconnect_delay: 1.5, reconnect_delay_max: 10.0)

r.get('foo')

https://aws.amazon.com/about-aws/whats-new/2019/06/amazon-elasticache-launches-reader-endpoint-for-redis/

[waitstime]

Thank you, @supercaracal, for your response, I'm aware of the reader endpoint. Is there a way to have a single Redis client connection that manages splitting read only calls between the primary endpoint and the reader endpoint and directs write calls to the primary endpoint? Or is it necessary to write special logic myself to handle multiple client connections (primary and reader) and directing calls to an endpoint based on the redis call?

[waitstime]

Thanks for the info! Closing this.

[dominicfollett]

@supercaracal Apologies for commenting on this closed ticket. But I'd like to ask for clarification.

I have an AWS Elasticache (cluster mode disabled) replication group with a write primary endpoint, and a readonly endpoint. In my scenario I am configuring Rails' cache store:

  config.cache_store = :redis_cache_store, {
    url: [
      ENV['REDIS_WRITE_HOST'],
      ENV['REDIS_READ_HOST'],
    ],
    ...

As it stands, I think I might run into problems with write operations being made on the readonly endpoint. You mention above that using cluster mode is a potential option. How would I leverage redis-clustering with ActiveSupport::Cache for my use-case? Thank you.

[supercaracal]

rcc = Redis::Cluster.new(nodes: "ENV['ELASTICACHE_ENDPOINT_URL']", replica: true, replica_affinity: :latency)
config.cache_store = :redis_cache_store, { redis: rcc }

The above setting might work for you if you use the redis gem version 5.* or later. Also, Redis cluster (cluster mode is enabled in ElastiCache) has some constraints because it does sharding. If bulk processing commands such as MSET, MGET, pipelining, and transaction features are used in redis_cache_store internally, some errors may occur. And clients need consideration of the replication delay. Sorry, I'm not familiar with redis_cache_store codes.

https://github.com/rails/rails/blob/main/activesupport/lib/active_support/cache/redis_cache_store.rb