To access AWS resources, we have to use a custom endpoint, pointing to an NLB and use the IP/Port of the resource as the target group of the NLB.For example take Redis:
Redis -> Target Group -> NLB -> Endpoint -> DC(internal)
For simple Redis i.e. cluster mode disabled, we have no issues as we can put the IP/Port of the Primary endpoint in the target group but with Cluster mode enabled, we will get a configuration endpoint and to handle the Ask/Move redirects we need Redis to advertise our endpoint, which clients can access from our internal network, as of now Redis will advertise the internal IP/Port which we can't access internally.
To modify this behaviour, we came across the below options:
cluster-announce-ip
cluster-announce-port
cluster-announce-bus-port
replica-announce-ip
replica-announce-port
Since we can't modify the parameters of the AWS managed ElastiCache, we are looking into deploying Redis Cluster on EC2 instances.Will be helpful if somebody can help us with the below questions:
1.We were able to set our endpoint IP/Port using cluster announce parameters but as soon as we add them, the node gets disconnected from the cluster because that address is not resolvable internally.Can't we advertise multiple addresses, one for internal and the other for external clients?
2.What should be the value of cluster-announce-bus-port ?
3.How will replication work if our endpoint is not accessible in the AWS internal network?
4.Where should we use replica-announce-ip and replica-announce-port , only on the Replicas? Since after failover or some other issue a master can be a replica also.What should be the value of these?
So, to summarize we need Redis cluster deployed in EC2 instances to advertise the IP/port given by us and Replication should work properly gievn that the IP(of our endpoint) is not resolvable internally to the EC2 instances. Let me know if you need any further info.