I've started building the SisuOps course.
The first piece isn't a video or a slide deck. It's code. Two repos, built the GitOps way, and every module after this builds on them.
Repo 1: the infra template
This is the whole cluster, written as code.
OpenTofu creates the Kubernetes cluster on UpCloud. Prefer another cloud? The OpenTofu part can be rewritten for that provider, plus a couple of small provider-specific settings. The GitOps setup on top stays the same. FluxCD takes over from there and runs everything on it straight from Git. Ingress, certificates, every client site, and the automatic image updates all live in this one repo.
It's the same setup that runs sisuops.dev, netvoima.fi and getbusygenius.com today. I copied it out and stripped the client data.
Repo 2: a small Next.js app to test it for real
An infra repo with nothing running on it doesn't prove much. So the second repo is an example Next.js app you deploy onto your own cluster.
Here's the loop it shows you. You change a line of text and push to main. A minute or two later it's live on your own domain, with a real certificate.
Nobody ran kubectl. Nobody SSH'd anywhere.
That loop is what the whole course is about, and you see it working on day one.
The mistakes are already in there
A lot of the settings in this template exist because something broke on my production cluster first. Each one has a comment telling the story.
A few of the stories:
- A staging push went live in production within minutes. One small rule in how images get tagged makes that impossible now.
- One settings change took every site down for about 15 minutes. It looked like a harmless edit. There's a safe way to make that change. It's the less obvious one.
- A chart edit never deployed. No error, no warning. Flux just kept running the old version.
- A routine rollout starved the sites next to it on small nodes. One setting fixed it.
You'll hit every one of these yourself eventually. Or you can read why the setting is there before it bites you.
Testing it costs a few euros a day
You don't need a big cluster to learn this. Two small nodes on UpCloud are enough, and a test cluster costs a few euros per day.
Build it and run the whole flow on a real domain. Then tear it down when you're done.
(There's one teardown gotcha that keeps billing you if you miss it. It's covered.)
What's next
These two repos are the base. Next I'm writing the modules around them, going through why each piece is set up the way it is.
I'm building this solo, about five hours a week, and I'll keep posting progress here.
Want the full system? Join the list and you'll get updates like this one as the course comes together.
Already on the list? The first 50 refundable $20 deposits lock in the early-bird price, $148 instead of $297. You can put yours down on the pricing page.
