Observing applications in Dapr

In a monolithic application made up of a few components running on a limited number of nodes, understanding how the application is behaving is a relatively simple task. In such a context, there is an expectation that by monitoring the activity of some processes on one or maybe two nodes to get higher availability, and their usage of the nodes' central processing unit (CPU) and memory over time, we can get a good perspective on the application's behavior. Log files would be available on the nodes, and the collection of those files could be arranged with any classic monitoring solution.

Once the number of working components increases dramatically, however, when we leverage the advantages coming from building microservices around business capabilities, the level of complexity grows along with it. We have more processes to monitor, and it is also likely that these are related to each other and might even have dependencies to external components...