We've discussed the three main observability techniques of modern applications: logging, metrics collection, and distributed tracing. All of them have their advantages, but logging is definitely the most important way of collecting information from your application. That's because it is simple, does not require any special infrastructure (although it is good to have one), and is least likely to fail you.
But logging has some limitations. Unstructured logging messages make it harder to extract insights from logs. Logging is also not suitable for the periodic probing of information about resource usage and performance. It is good for auditing purposes and post-mortem analysis after major failures but is rarely helpful in tracking current information and reacting to sudden events.
That's why metrics collection systems are a natural and valuable extension of logging infrastructures. They allow you to collect information in real time, create custom metrics...