For every layer of the monitoring stack, we can always find a counterpart collector. For instance, at the application level, we can dump metrics manually; at the host level, we would install a metrics collector on every box; as for Kubernetes, there are APIs for exporting the metrics that we are interested in, and, at the very least, we have kubectl at hand.
When it comes to the container level collector, what options do we have? Perhaps installing the host metrics collector inside the image of our application does the job, but we'll soon realize that it could make our container way too clumsy in terms of size as well as resource utilizations. Fortunately, there's already a solution for such needs, ...
