Ensuring sustainability of integration

Errors in processing operations occur when they were not expected. The reason may be an error in the code, or an abnormal termination due to large operation flow, transaction deadlock or a memory lack. The classic architecture ensured consistency due to transaction of user request operation, but the distributed architecture demand check stability during the testing.

How to test the stability of the integration and the application as a whole? To come up with test scenarios based on architecture descriptions? What to do in legacy projects when the architecture descriptions are inaccurate? How does understanding the business architecture help with this? And how to fix the situation if the interaction of systems is unstable? It is clear that it is impossible to give comprehensive answers to them within the framework of one talk, the topic is too broad. But I will share the practices that help me get answers, and I hope this will be useful.