SLOs, error budgets and burn-rate alerts

In build — target Aug 31, 2026. The plan below is complete; real configs, dashboards, alert rules and the postmortem land here as the build finishes.

Prometheus Grafana PromQL SLO methodology

What this project builds

Reliability engineering on top of the EKS cluster: define SLIs and a 99.9% SLO for the demo service — which makes the error budget concrete: about 43 minutes of allowed failure per 30 days — then build multi-window burn-rate alerts that page on budget spend, not on raw symptoms.

How it will be built

Prometheus + Grafana installed on the project-02 cluster by script.
SLO recording rules and multi-window, multi-burn-rate alert rules in PromQL (slo-rules.yaml).
A deliberate failure injected (inject-failure.sh), the alerts observed firing in the right order, and a written postmortem published from a proper template.

Why it matters

Every candidate shows Grafana dashboards; very few can explain an error budget, and almost none have published a postmortem. This maps directly onto the SLA work I already do in production — the project gives that experience its industry vocabulary.