← Nannix

Home Lab

Learning to listen to my server.

Graphs, temperatures and Telegram messages: the tools I use to understand how my everyday machine is doing.

Learning to listen to my server.
From my lab.

Powered on is no longer enough

At first I wanted to see applications running. Then I started asking how much they ask of the machine: CPU, memory, temperatures, storage. On a 2016 computer these questions are practical. I like being able to observe the system and connect the numbers to what I am doing with it.

I changed my mind about monitoring

I installed Netdata and liked its wealth of information and alerts. On my MacBook, though, it used too much CPU. I moved to Beszel for lighter monitoring. That choice came from use: I wanted resources left for the applications I had started the server to host.

I still find Netdata interesting. This experience taught me to choose according to my machine and what I need to see. I can really like a dashboard and still decide it asks for more than I want to give it.

Different questions, different tools

With Scrutiny I observe disk health information and temperature. With Uptime Kuma I check whether websites and services are reachable. I like how each tool helps me look at a specific aspect: machine condition, disk condition and availability of what I use.

I am also learning what those signals cannot tell me. A disk with reassuring readings today can still fail; a reachable service can have other problems. Graphs help me ask questions and act, while backups remain a separate responsibility.

The automations I enjoy building

A Telegram bot alerts me when CPU stays at maximum for around ten minutes and sends temperature warnings. It reports connection interruptions and recovery with the elapsed time. I can also ask for server status, temperatures and warnings. I like getting information without always sitting in front of a dashboard.

These are the automations I find most satisfying: small, built around something I actually need. I use an external service like Telegram because its convenience makes sense to me here. I am looking for a balance I can live with every day.

Learning when to look and when to let it run

I could always add another graph. I prefer looking for alerts that tell me when something needs attention and information that helps me understand it. I want the lab to leave me room to do other things as well as look after the lab itself.

Read further

Documentation and references

Explore the library ↗︎