we are going to change our Elasticsearch cluster architecture.
currently we have 5 nodes per cluster, all are data + master eligible nodes.
- we are connecting all nodes from code via round robin method from code.
Here the problem is
-- when one node is unresponsive the request to that particular node becomes fails.
- Having independent master node to serve all requests
- Having independent client nodes to serve all requests
- Having haproxy to maintain high availability between master or client nodes.
give me the little bit elaborated way to form the cluster.