Prometheus
Exporters & Instrumentation
Get metrics out of apps and third-party systems into the Prometheus exposition format.
Prometheus only scrapes HTTP endpoints in its exposition format, so two paths exist to produce them. Direct instrumentation uses a client library (Go, Java, Python, etc.) inside your code to define counters, gauges, and histograms and expose them on /metrics. This is the richest approach because the metrics reflect your domain — orders processed, queue depth, cache hits.
A universal travel adapter: your appliance (the database) has its own plug shape, and the exporter is the adapter that converts it to the standard socket Prometheus expects.
Key Concepts
1
For systems you cannot modify — databases, message brokers, the host OS, hardware — exporters act as adapters. An exporter (node_exporter for host metrics, plus exporters for MySQL, Redis, Kafka, etc.) talks to the target in its native protocol and re-exposes the data as Prometheus metrics. The four golden signals and the RED/USE methods guide which metrics to add: instrument latency, traffic, errors, and saturation rather than everything.