In page 212 is detailed the implementation of message gateway, in order to divide the message in 2. One to process the payment and another one to continue the message avoiding to charge the credit card twice.

But what happens if the message fails after accepting the payment but before sending the message to the bus (like the bus endpoint went down) ? It will change the credit card but will be retried again after the bus goes up again.

Is this scenario possible ?

Hello and thankyou for your question.

Firstly, my apologies for the delay. There is an issue with the Wrox notification system and I was not being notified of new messages.

As for your question. In short yes, there is always the possibility that there could be a failure at any point. By adding a message after each operation, you reduce the potential for a failure to cause harm.

But, there's always that small chance it will happen at an awkward moment. The strategy here is to minimise the possibility as much as possible.

As we talked about in the book, though. If you can make your operations idempotent the problem goes away. If you REALLY REALLY REALLY cannot tolerate failure, you may want to investigate distributed locking as described by Martin Kleppmann http://martin.kleppmann.com/2016/02/...d-locking.html.

Hope that helps and I hope the late response does not cause offence.