Howdy,
I agree with some of your points which is why I'd be more inclined to choose tRFC over aRFC - however there are still design questions about capturing the error messages, things like:
- How many error messages are you expecting, is it 1% or significantly less?
- Is the number of error messages small enough that generating a single email notification per batch run is acceptable?
- Who would be addressing the error messages, technical people or business users?
These are questions which need to be answered regardless of the solution.
If you are expecting many error messages and cannot fix the errors upstream then I would consider using the application log for recording errors. If you still want an email notification once all tRFC queues have been resolved then you can have a "master" job create the tRFC queues and then wait until all the queues have been processed, then pick up the application log and email it.
Agree it is not possible for all messages to be processed at once (the number of queues to be created should be configurable or use a variant). But also if you have multiple application servers then you can distribute the jobs across those servers.
And you're correct that the design decision should include a review of the process distribution among application servers.
Hope this is not too OT - interested to see what the OP's final solution is.
Cheers
Alex