I have a cluster of 5 memebers. Three of them are masters and all of them are data nodes. What is the best way of configuring a round robin? Only including the masters and then they will take care of redistributing the requests?
Describe the issue:
Relevant Logs or Screenshots:
By default, every node is coordinate node, and coordinate code can redistribute the requests, so you can include all of the nodes in client’s configuration, most clients can round robin the configured nodes properly. However, when one of the nodes fails, your client will encounter connection errors, so it’s recommended that using a proxy or load balancer before the cluster, and setting up a health check strategy for the backend nodes, when one of the backend nodes fails, it will remove the node from the backend node list as soon as possible.