Traffic on the front door should be monitored carefully, and its metrics should be defined after giving proper thought to it. Front door implies any entry point exposed to the world. It should be HTTP API, or sockets or static web pages, and so on. Deploying public or partner-accessible microservices without protection is lunacy and a recipe for disaster, unless you like failing. Think about the scenario when someone decides to hit your endpoints 10x more than you planned for; sleepless nights for sure. So, monitor microservices that you publish and limit access to them.

The monitoring system should be capable of making you understand the things happening from outside the world, such as a sudden rise and fall in API calls while there are not many active users in the system. It is better if the monitoring system evolves by itself by learning the metrics it collects. One of the major factors is that the monitoring system should be able to show you the data flow and data about service performance. Developers need performance metrics to improve the code in a new version. Performance metrics includes CPU usage, memory usage, time taken to talk to a database or another service, the serving time of one request, response time versus number of request graphs, and so on.