It was fixed my increasing the socket settings to the approximate time taken for cluster to come up again or reelect a leader. Also increase maxSize to like 10, spring could not handle 100 connections adequately. Now process is a bit slow but still more reliable.