Observability with Prometheus and Grafana

Rates and SRE practices

In SRE, many objectives and alerts are framed as rates (errors per request, failed checks per probe, etc.). An SLO might state, for example, that the error rate must not exceed 1% over 30 days. Practices such as error budgets and SLO burn-rate alerting rely on rate-based measurements. Some useful references include:

• Google Cloud: SRE error budgets and maintenance windows
• Google SRE Workbook: Alerting on SLOs