Monitoring

Prometheus Exporters for Bare Metal

Instrument noisy hosts without melting CPUs, label metrics honestly, and keep dashboards readable.

Blended evenings 4 weeks · 22 lab hours Intermediate

980,000 KRW informational until enrollment confirms

Grafana tiles with muted blues on a widescreen

Program narrative

You install a curated set of exporters on lab metal, then intentionally spike cardinality to feel the pain. We close with a Grafana review session where mentors rip apart cute but useless panels.

What the labs include

  • node_exporter flag tour with safe defaults
  • Recording rules vs alerts philosophy
  • Label naming workshop
  • Cardinality disaster on purpose
  • Recording rule for SLI starter kit
  • Dashboard critique rubric
  • Handoff doc for overnight teams

Outcomes you can show a lead

  • Ship a dashboard a tired human can parse at 03:00
  • Explain cardinality tradeoffs to developers
  • Pick exporters that survive reboot storms

Straight FAQ

Kubernetes?

Mentioned for contrast; labs stay on bare metal hosts.

Long-term storage?

We discuss Thanos briefly; setup is out of scope.

What is excluded?

Tracing pipelines are not part of this syllabus.

Recent participant notes

“Cardinality disaster hour hurt, in the intended way. Recording rule starter kit is now our team default.”

Nari · Monitoring owner

“Mentor roast of my rainbow dashboard was kinder than production users would be.”

Iseul