I am evaluating couchbase community edition to deploy in AWS. But as rack awareness is not available in community edition , I am wondering how we can deploy production workload without compromising the data in case of whole AZ going down as both primary and replica docs can be in same AZ.
I am thinking to deploy based on following formula to sustain whole AZ failure.
Number of replica should be equal to … Number of nodes in one AZ + 1
This will ensure , that we have at least one replica is on a different AZ . But this is probably ok for small data sets, bigger dataset can require more RAM /nodes and cost.
Can anyone suggest anything better.