Nice read. You are well ahead if you are thinking about security compliance, data integrity issues, etc etc that such a model would bring. And the pressure is sure on from the beginning to design such a system that would
- Work as normal for all the production traffic and continue to route to the old service
- Identify & handle the test traffic in production and route to the new service under test
- The monitoring and metrics know to disregard such test traffic
- And that all this is a for each service
Not to say this is not possible but it has to be in the ‘engineering’ culture.
For me..I am still battling the ‘this is frontend’ / ‘backend’ issue