The Node Exporter is the most well-known Prometheus exporter, for good reason. It provides over 40 collectors for different areas of the OS, as well as a way of exposing local metrics for cron jobs and static information about the host. Like the rest of the Prometheus ecosystem, the Node Exporter comes with a sane default configuration and some smarts to identify what can be collected, so it's perfectly reasonable to run it without much tweaking.

Although this exporter was designed to be run as a non-privileged user, it does need to access kernel and process statistics, which aren't normally available when running inside a container. This is not to say that it doesn't work in containers—every Prometheus component can be run in containers—but that additional configuration is required for it to work. It is, therefore, recommended that the Node Exporter be run as a system daemon directly on the host whenever possible.

Node Exporter collectors might gather different metrics depending on the system being run, as OS kernels vary in the way they expose internal state and what details they make available. As an example, metrics exposed by node_exporter on macOS will be substantially different from the ones on Linux. This means that, even though the Node Exporter supports Linux, Darwin (macOS), FreeBSD, OpenBSD, NetBSD, DragonFly BSD, and Solaris, each collector within the Node Exporter will have their own compatibility matrix, with Linux being the kernel with the most support.

By design, this exporter only produces aggregated metrics about processes (such as how many are running, and so on) and not individual metrics per process. In the Prometheus model, each process of relevance needs to expose its own metrics, or have a companion exporter to do that job for it. This is one of the reasons why it is ill-advised in most cases to run a generic process exporter without an explicit whitelist.