Prometheus

An open-source monitoring system that collects and stores time-series metrics and supports alerting.

2 min read 4 lesson sections
  • #prometheus
  • #monitoring
  • #metrics
  • #promql
  • #alerting

The quiet dashboard before the noisy incident

An application can look fine from the outside while response times creep upward and errors quietly multiply. Prometheus helps a team notice those changes by collecting numeric measurements—called metrics—over time. Instead of asking only “is it down?”, you can ask “when did the error rate start rising?”

Metrics have a shape and a story

Prometheus stores time-series data: a measurement associated with a timestamp and labels. Teams expose metrics from applications or infrastructure, and Prometheus commonly scrapes those endpoints on a schedule. You can query the collected data with PromQL and define alerting rules; Alertmanager can then handle how alerts are grouped and routed.

Metrics are not the whole story. They tell you that something changed and help show its scale, but they may not reveal the exact request or code path behind the change. Logs and traces can provide complementary evidence.

Start with a useful signal

Track a small set of signals that match the service: request rate, error rate, response duration, and resource saturation are common starting points. Label carefully—labels with unbounded values such as unique user IDs can create too many time series and strain storage.

Mini mission

Imagine a page becomes slow after a release. Pick one metric that could confirm the slowdown, one that could show its impact, and one question you would ask next. Then picture opening Grafana to explore the trend.

Prometheus helps you move from “something feels wrong” to evidence you can investigate.