We have an environment where mainframe invokes web services in webMethods using XML RPC server. Considering that there is one EntireX broker, is it technically correct to have two XML RPC servers with the same service names(RPC/XXXX/YYYY) installed in two machines to connect to the EntireX broker to achieve high availability for XML RPC server?
You only have restricted high availability with this setup.
This only works if an RPC Server can notify the Broker before it is going down (deregister from the Broker).
But if the machine where the RPC Server is running is crashing or the network connection to the Broker gets lost the Broker will not know that the RPC Server is no longer active and continue to assign client requests to this server.
You can minimize the disruption from a crashed XML RPC Server by setting the SERVER-NONACT timeout to a smaller value. The longest the server would be unavailable should be no more than twice the SERVER-NONACT value. A small SERVER-NONACT will generate more “noise” between the server and EntireX Broker, but in a busy server it won’t matter much - the busier the server is, the shorter the SERVER-NONACT can be with less overhead and shorter time before the inactive server is noted and removed.
Rolf- when we say that HA is getting restricted, I understand the issues by the scenarios stated. Considering that I do not control the broker, is there any other ideal way to set up the high availability with respect to XML RPC server?