The term infrastructure may be too broad here, but if we simply consider where our application runs and how it interacts with other components and users, it is obvious what we should monitor: the application hosts and the connecting network.

As collecting tasks at the host is a common practice for system monitoring, it is usually performed by agents provided by the monitoring framework. The agent extracts and sends out comprehensive metrics about a host, such as loads, disks, connections, or other process statistics that help us determine the health of a host.

For the network, these can be merely the web server software and the network interface on the same host, plus perhaps a load balancer, or even within a platform such as Istio. Although the way to collect telemetry data about the previously mentioned components depends on their actual setup, in general, the metrics we'd like to measure would be traffic, latency, and errors.