I have 6 nodes in a cluster performing reads and writes. When I stop the network interface on a one of the nodes the other 5 nodes block for 60 seconds and then resume normal operations. I did not observe the pause when using a replicated-cache configuration. The distributed-cache configuration is my preferred deployment for production. My configuration is using a TCP JGroups stack and synchronous replication. Note that when I stop a node gracefully or with kill -9 the other 5 nodes continue reading and writing data without pause as expected. I only observe the blocking when I pull the network cable on a single node. Any help would be appreciated.
topology
rack="even": node1, node3, node5
rack="odd": node2, node4, node6
version: 6.0.2.Final