Hi deepal,
>The problem is that pinging at the handler level basically amounts to >just asking "is there such a service?". Having it go to the message >receiver allows the ping to be responded to based on more useful data- >such as did the impl class or script or whatever load etc.. So its >really a question of fidelity- yes some level of liveness checking can >be done at a handler level but you get much improved info by getting the >ping to the message receiver. > > Service might have more than one MRs (most of the time it does) so checking only one MR wont give you the accurate results. The other problem I have is which MR you are going to call when you receive a ping request , are you going call all the MRs or only one.
There can be two types of ping requests. 1. Service level ping requests. 2. Operation level ping requests. In the first case, you are correct, we need to ping all the operations within the service. Then in the ping response simply we can say whether the service is up or not, or we can include all operations and their status. In the second case we can directly ping the MR assinged to the requested operation and the response accordingly. This implies that we need to develop a schema for the ping requests and ping responses. thanks Sameera.
