I have a cluster consisting of 3 master nodes, 4 client nodes, and 16 data nodes. I utilize AWS ELB for request balancing. Within the ELB setup, I allocate 4 client nodes and employ a Round Robin algorithm to distribute the load. This algorithm checks the health status of cluster, and if it's green, assigns the request accordingly. However, if a client node is removed from the cluster, the ELB remains unaware and continues to assign load to that node, resulting in unresponsive requests.
How can this issue be mitigated? Is there a way to inform the ELB about nodes being kicked out?