Couchbase in AWS with replicas and server groups per AZ


We have a “Best Practices” question around how the latest version of Couchbase behaves with respect to AWS availability zones.

If we deploy a single couchbase cluster (9 nodes) across 3 availability zones in AWS where each AZ has it’s own Couchbase server group defined. What happens if we lose an entire availability zone in AWS (~ 3 nodes) assuming we have just 1 replica?

Are we safe to perform a rebalance of the last remaining 6 nodes because of the “Server Group” functionality which should guarantee that all the replicas exist in one of the other availability zones?



Yes, that is my understanding of how rack zone awareness (i.e. Couchbase node groups) works -