I've setup a 3 node cluster, and in the client application its using the bolt+routing connection. I've also setup some Gatling stress tests, but it seems that all the queries always end up on just one of the cluster nodes. What is cluster balancing based on? Is it one using a single node because its not being stressed enough?
check out this link
You'll have to explicit create read or write transactions via the driver. Write transactions (or default, if not a read transaction) will always go to the leader node in the cluster. Read transactions will round-robin between the follower and read replica nodes.