Description

      When the controller can not provide services when the problem occurs, it reflects the importance of the cluster, of course, reliability is one of the advantages of the cluster.

      This test is designed to test when a node in the cluster fails, how quickly it will take for other nodes to take over the switch belong to the fail node ago.

      This case is tested by one switch connected to a cluster controller, this cluster may be 3、5、7 ... ...

      During the test, we continue to send packet_in packets from the switch to the cluster. And then shutdown down the controller which is the master role of this switch. After the new role_request message is delivered, stop the capture and caculate the time by the last responded packet_in message and the first role_request message.

Suggestions

1. The transmission frequency of packet_in
   We suggest the packet_in interval is 50ms, that is 20 copies of packet_in messages are send in one second, and this can be configured in IxNetwork.
2. How to shutdown the controller(ONOS)

   1. Active approach

      onos> system:shutdown

   2. Passive approach

      $ sudo killall -9 java