Back to archive

Stop Paging on CPU

Page on user symptoms, not causes. Burn-rate alerting turns error budgets into sane on-call.

Production idea

Cause alerts, symptom alerts, burn rate, multi-window paging, and runbooks.

job:slo_errors:ratio_rate1hjob:slo_errors:ratio_rate5mrunbook_url
  • Page only on symptoms.
  • Use multi-window burn-rate alerts.
  • Attach a runbook URL to every page.
  • Review every page weekly.