How We Improved Rails Response Times by 87% — Fast Retro Blog

The whole observability setup lives on a single server, locked behind Tailscale so nothing is exposed to the public internet. I deploy it with Kamal 2 — the same tool I use for Fast Retro itself. The stack has four pieces: Prometheus — scrapes metrics every 5 seconds, stores 30 days of data Grafana — dashboards and exploration UI Loki + Promtail — log aggregation (auto-discovers container logs via Docker socket) Node Exporter + cAdvisor — server and container resource metrics Everything is defined in config files and auto-provisioned. Grafana boots with datasources and dashboards already configured — no manual clicking around.