At times an endpoint may not reply when it is invoked in a BPEL process. In normal circumstances, the invoke faults. To avoid this, you can attach a policy to the endpoint reference in the PDD file to indicate that additional retry attempts are allowed at a specified interval for the endpoint. This technique is described in
Process Developer > Part IV: Testing and Deployment > Process Deployment > Endpoint Reference Addressing Considerations
.
Alternately, you can name a service in the policy section of the PDD file whose job it is to provide a retry interval, that you may want to calculate dynamically, and do additional work programmatically. The service could also specify an alternate endpoint to try when the invoke is faulting or suspended.
This service, which is a retry-policy service, is a process you can create and that the Process Server instantiates when an invoke activity is faulting (or suspended). The retry-policy process must be based on a WSDL named
retryCheck.wsdl
.
In essence, the WSDL defines the
retryCheck
operation that receives information from the server regarding the fault name, process Id, and other details of the faulting invoke activity. The process you write is instantiated when the invoke activity faults (or can be suspended), and it receives an input message identifying details of the fault.