Client applications should communicate with the high availability services only through the logical hostname of the logical host of the high availability service. You should ensure that client applications are written to accept a communications error and possibly retry after a few minutes.
Consider a typical client connection. The client is connected to machineA through the logical host called snap. If machineA fails, then snap fails over to machineB. According to machineB, the client connection does not exist, and will send the client a connect reset message, which will appear to the client as a communication error. The client must reconnect to the server to obtain a new connection from machineB when DB2 starts.