Observability stack for Paycrest
Metrics, logs, traces and SLO-based alerts that page only when users are affected. A case study by Lukas Schmidt, Platform Engineer (Terraform & GCP).
Challenge
Paycrest, a payments startup, learned about outages from customers. The on-call engineer received hundreds of alerts a week, most of them noise.
Solution
I deployed Prometheus, Loki and Tempo with Grafana, instrumented services with OpenTelemetry, defined SLOs for the critical user journeys and replaced noisy alerts with burn-rate alerts linked to runbooks.