Skip to main content

Knowledge Base

These are teaching notes. I mentor engineers at Developer Akademie, and a page lands here when I have had to explain something for the third time, or when I want to point at one canonical answer instead of retyping it in a code review.

That makes the coverage uneven on purpose. Docker gets several pages because I use it daily and because most of the Dockerfiles I review have the same three problems. There is nothing at all on Kubernetes, because I don't run it.

If you want the applied version — real numbers, real postmortems — that is in the blog and the project write-ups. This section is the reference layer underneath.

What's here​

Containers​

Concepts, building your first image, and Compose. The one I'd point a colleague at is First Docker Image — it goes past FROM python:3.9 into non-root users, multi-stage builds, minimal base images and where Trivy fits in the pipeline. Most Dockerfiles I review pick up at least three findings from that page.

DevOps​

Practice-level notes: where Docker actually sits in a delivery workflow, and implementing DevOps against the 12-Factor principles. The 12-Factor page is the one I keep returning to — most "we need DevOps" conversations turn out to be a config-in-the-code problem wearing a costume.

Git​

Branching and workflow. Deliberately short: Git basics are the best-documented topic on the internet and I have nothing to add to them. The security side — signed commits, scanning before the commit lands, what to do when a key is pushed — is in Git Security, which goes considerably deeper.

Environment variables​

How to load them, and more importantly what not to put in them. Pairs with Secrets Management, which is the longer argument for why .env files are a local-development convenience and not a secrets strategy.

Where I stand​

A few positions that run through all of this, so you know what you are reading:

  • A security check that only warns is not a check. If a Semgrep finding can't block a merge, it will be ignored by the third sprint. Pick fewer rules and make them blocking.
  • Golden paths beat guardrail documents. Nobody reads the wiki page about IAM. They do use the module that gets IAM right by default.
  • If you can't measure the rollback, you don't have one. An automated rollback with no SLO behind it is a button nobody trusts enough to press.
  • Agents propose, humans approve. I'll let a model read every log in the estate. It doesn't get to run terraform apply.

What's missing​

Honest gaps, so you don't go looking: Kubernetes, service meshes, and anything about scale beyond a few dozen services. I work on platforms in the tens-of-services range. I would rather have nothing here than have you find out the hard way that a page was written from someone else's blog post instead of from production.