The brief
I wanted the homelab to tell me when something broke, instead of me finding out the slow way. The answer was the TICK stack plus Grafana, with alerts going into Discord because that is where I already am.
What I built
- Telegraf agents on every host pushing metrics into InfluxDB.
- Chronograf for quick data exploration. Kapacitor for alert logic.
- Grafana for the dashboards I actually come back to.
- Discord webhooks receiving alerts in a dedicated channel.
- Ansible and GitLab CI handling deployment and config drift.
Why it mattered
The TICK stack is a real production tool, the same stack Orange runs as its centralised observability platform for managed customers. Building and operating it at home meant I understood how Telegraf agent config, InfluxDB retention policy and Kapacitor alert logic connect before I was tuning them in customer environments. The TICK era has since been replaced by Prometheus, Loki and OpenTelemetry, but the mental model for how metrics flow from host to alert carried across stacks.