Seems like you can connect to any node and index and query data. If I have a cluster with 3 master nodes and 3+ data nodes, should I connect to the data nodes or master nodes to index or query documents? Should I round robin across the nodes i.e. if I have 6 app processes should those have connections to all elasticsearch nodes or just a single node? What is most efficient?
Leave dedicated master nodes to manage the cluster and send all requests to the data nodes.