sir,

We have Two server node,

1.server A who is responsible receiving and sending message from/to out side the world.

2.Server B who is receiving message from node A and process it and give response to node A

Our setup model is,

Server node A have "zero capacity distributed request cache" .whenever we receive request from outside ,we put request in that request cache .So,non-zero capacity of this request cache already definied in server node B and have listener for it. So always listener fires while  request are put in server node A.

above case happen vice versa like server node B give response for that request to zero-capacity response cache and corresponding non-zero capacity initialized already and have listeners for it.

We were testing in above setup, as per second 200 request from server A. but it was delayed to reach server node B.So,our server went unstable. After failed load, we again continued same load without restarting the server node.This time ,no problem. So we tried above algorithm like try 300 request/per second and it got failed .then restart the test again 300 and it was successful.

It may sound crazy.but ,I dont know,why it is happpening like that. is there any capacity in infinispan that have to increase or have to increase jgroups thread or anything have to be considered??

note :server node A have 2gb heap and server node B 3 gb heap. futureAsync is used to stopping thread get stuck.please comment if you dont understand sir.i can try to give better explaination .