ralph balazs
← Back to all projects
DevOps Infrastructure

Kubernetes GitOps Pipeline — Inception of Things

K3s clusters provisioned from scratch with Vagrant and Bash, Traefik ingress routing, and an Argo CD GitOps pipeline that auto-syncs deployments from Git.

Context
42 school project
Team
Team of 2
Status
My part: K3s provisioning and the validation suite — ingress part pending teammate availability
Stack
K3s/K3d · Argo CD · Vagrant · Docker · Bash · Traefik
Links

Goal

Inception of Things is a 42 team project about running a full GitOps loop end to end, locally: provision a Kubernetes cluster, expose applications through it, and let Git be the source of truth for what's deployed. This is a team-of-2 project; my part covers the cluster provisioning and its validation suite.

Provisioning that doesn't lie

I automated the provisioning of a 2-node K3s cluster (server + agent) with Vagrant/VirtualBox and Bash: network-interface detection, waiting on API readiness, and a token-based node join. Instead of relying on fragile fixed sleeps, I wrote a Bash validation suite that actually checks cluster health — nodes Ready, private IPs assigned, K3s services active — so provisioning regressions get caught instead of silently passing.

Ingress

Traefik ingress routes three applications by Host header with a default backend, with the provisioning waiting for controller readiness before applying manifests — this part of the project is still pending teammate availability.

GitOps

I contributed to the Argo CD GitOps pipeline, where Git is the single source of truth and Kubernetes manifests auto-sync on every commit — including a bonus setup running a self-hosted GitLab CE instance inside the cluster as the Git source itself.

What this taught me for MLOps

Treating infrastructure as something you validate, not just something you hope works, is directly transferable to ML pipelines — the same "don't trust it until you've checked it" instinct applies to model deployments as much as to cluster nodes.