Node Exporter
inari.node-exporter
A Docker Compose file for Node Exporter alone, generated from the catalog and checked by the safety inspector. No account needed.
Prometheus exporter for host-level metrics, reading CPU, memory, disk, filesystem, network and kernel statistics from the operating system and exposing them on a scrape endpoint.
Node Exporter reads the host's own instrumentation — the proc and sys filesystems on Linux — and translates it into Prometheus metrics, running one process per machine rather than one per service. Collectors are individually enableable and cover load average, filesystem fill, hardware sensors, systemd unit state and network interface counters among many others. A textfile collector reads metrics written to files by other scripts, which is the standard route for exposing batch job results. It reports on the host rather than on containers, so it is normally paired with a container-level exporter rather than replacing one.
You know it worked when
- The metrics endpoint returns Prometheus-formatted output.
- Reported memory and filesystem figures match the host rather than the container.
- A Prometheus instance scrapes the target and shows it as up.
- A metric such as load average tracks actual host activity.
Known sharp edges
- Running it in a container requires host paths mounted read-only and the host network and PID namespaces, or it reports the container's figures rather than the machine's.
- Filesystem metrics count every mount it can see, so container overlay mounts appear as real filesystems unless explicitly excluded.
- Enabling every collector produces a very large metric surface, most of which is never queried but all of which is stored.
- The endpoint is unauthenticated and describes the host's hardware, mounts and network configuration in detail.