DevOps
Ship a service the way production actually does, by merge, not by hand.
DevOps is the practice of shipping and operating software as one continuous, automated flow: build → test → deploy → observe → repeat. No snowflake servers, no manual steps, everything in code.
Every engineering team runs on it. The job is no longer "write the feature", it is "get the feature to users reliably and roll it back safely when it breaks." Companies pay for engineers who can make deploys boring.
Build a containerised web app and ship it to Kubernetes with a full GitOps pipeline and monitoring. A merge to main should deploy itself; a bad change should roll back on its own; and you should be able to see the system's health at a glance.
Milestones
Build it in order. Tick each one when its proof check passes.
1. Containerise the app
Package a small web service into a reproducible image.
- Write a small web service (any language) with a /health endpoint
- Write a multi-stage Dockerfile and build a slim image
- Run it locally with docker compose and a database
2. Build the CI pipeline
Every push builds, tests, and publishes an image automatically.
- Add a GitHub Actions workflow: lint → test → build → push to a registry
- Tag images by commit SHA
- Fail the build if tests fail
3. Provision infra as code
Stand up the cluster + dependencies from a single command.
- Define the cluster and a managed database in Terraform
- Use remote state and a module for the repeated parts
- Destroy and recreate to prove it is reproducible
4. Ship by merge with GitOps
The cluster converges to what is in Git, deploys and rollbacks included.
- Package the app as a Helm chart
- Have ArgoCD auto-sync the cluster to the repo
- Trigger a deploy with a PR, then introduce a failing change and watch it roll back
5. Make it observable
See latency, traffic, errors, and saturation, and alert on what matters.
- Export metrics and ship logs centrally
- Build a Grafana dashboard for the four golden signals
- Define an SLO and an alert tied to its error budget
Everything you need
Curated, mostly-free resources to actually complete the project.
Definition of done
- A public repo with the app, Dockerfile, Helm chart, Terraform, and CI workflow
- A merge-to-deploy GitOps pipeline with automatic rollback on failure
- A live Grafana dashboard (golden signals) and one SLO-based alert
- A README explaining the architecture and how to reproduce it from zero
Take it further, verified
This is the spine of the Platform Engineering track, built on real Hetzner sandboxes and machine-verified week by week.
Building is the proof. Get it verified.
Anyone can follow a tutorial. The Stratiflux Academy turns building into something employers trust: real sandboxes, an AI pair-engineer, and machine-verified missions you defend in a live viva. Seats are free or near-free for selected talent, selection is the gate, not money.