- Hi, I'm bex. I work at Microsoft on Linux things, but this talk has nothing to do with that. - 139 euros is not a typo. That's my entire monitoring budget. It's really less than that and you'll see why. - I run a significant load bearing piece of internet infrastructure ...

- I am of course talking about puzzlesecretary.com - This is a production web service. People use it every day. - In full disclosure, runs on one Azure VM. No cluster. No fleet. Just one server, bravely existing without a service mesh. - But it still needs monitoring.

- You all know the proper way. This is your day job. - Metrics, logs, traces, dashboards, alerts, a small galaxy of YAML. - Step 1: install and wire up 17 things. Step 2: your soul leaves your body. - I love all of you for building this. But .. do not want. - My side project is playing games and improving puzzlesecretary - NOT monitoring puzzle secretary

- I have some more confessions to make. - I'm not going to get up every morning and check traffic graphs. - I'm not keeping a monitor in the corner with pretty lines. - I need a system that yells at me. Not one that waits politely to be consulted.

- My monitoring stack is Home Assistant on a Home Assistant Green. - It costs 139 euros. It lives in my data center.

- My monitoring stack is Home Assistant on a Home Assistant Green. - It costs 139 euros. It lives in my data center. - Like PagerDuty, it can blow up your phone.

- My monitoring stack is Home Assistant on a Home Assistant Green. - It costs 139 euros. It lives in my data center. - Like PagerDuty, it can blow up your phone. - Unlike PagerDuty, home assistant can haunt you

- My monitoring stack is Home Assistant on a Home Assistant Green. - It costs 139 euros. It lives in my data center. - Like PagerDuty, it can blow up your phone. - Unlike PagerDuty, home assistant can haunt you - Nothing says SEV-1 like your hallway being disappointed in you.

- Here's why this isn't as ridiculous as it sounds. - HA already stored time-series data: batteries, temperatures, water meter readings. - What is CPU load but a temperature sensor for nerds? - It already ran MQTT because one of my lights needs it. - It was already on Tailscale because I wanted remote access. - It was already always-on because it if it isn't I have to turn on the lights with a switch ... like an animal. - I didn't install a monitoring stack. I noticed I already had one.

- There is a dashboard that I don't look at - But when something breaks, the history is there. - I can see the curve that led to the problem.

- Telegraf runs as a container on the VM. Gathers stats about both the server and the game system itself. - Publishes MQTT messages over Tailscale to Home Assistant. - Home assistant makes decisions about alerts and tracks stats - That's it. That's the whole thing.

- The take away is this - I didn't set out to build something weird. - I asked: what do I already have, already know, and already pay for? And then I looked to see if those things would just solve my problem. - This lets me stay focused on the fun part — adding games, improving UX. Delivering my product. - Monitoring: Home Assistant. Transport: Tailscale. Collection: Telegraf from InfluxDB - The point isn't "use Home Assistant." - Before you build a platform for your platform, ask what you already have, know or pay for. - Thank you. I'm bex. The slides are at the QR code.

Mermaid rendering: use `marp --html` to enable