WebLogic Server (WLS) Managed Servers Losing Connectivity to Admin Server
Your managed servers are losing connection with one another and with the cluster in which they belong. The logs indicate that the servers are running correctly, but unicast messages are being lost. When enough unicast messages are lost, the server is removed from the cluster. However, the server is still there, so it is returned to the cluster again after a few seconds or minutes.
This sort of situation is often explained by network latency caused by heavy use.
If it isn’t possible to reduce the network latency, you can adjust the Idle Periods Until Timeout parameter in order to allow the system enough time for the messages to arrive.