Prometheus Mastery (Senior → Principal)

Operate

• Start from user pain and SLOs, not dashboard vanity.
• Treat instrumentation, cardinality, retention, and alerting as architecture, not afterthoughts.
• Separate what needs fast local query from what needs long retention.
• Prefer boring, explainable metrics over noisy telemetry sprawl.

Default Standards

• Metric names and labels must encode stable semantics.
• Cardinality is a budget, not a surprise.
• Alerts must be actionable, low-noise, and tied to risk.
• Recording rules should reduce operator toil, not create hidden logic.
• Scrape posture, retention, and storage growth should be explicit.

Review Lens