blog
49 posts in English and Spanish. Browse by tag or subscribe via RSS.
The closing post of the series. The thread that connects all six patterns, the evolution from glue spaghetti to platform control plane, and why having architectural vocabulary changes how platform teams make decisions.
Hardcoding deployment strategy into pipelines creates a hidden coupling that makes platforms less reliable. The Strategy pattern explains how to make canary, blue/green, and progressive delivery first-class platform concerns.
Every platform connects systems that weren't designed to talk to each other. That connective tissue is glue — and in most platforms, it's 60-80% of what the team actually builds. Here's what glue is, where it lives, and how to design it intentionally.
Pipelines that run kubectl apply directly are automated but invisible. The Command pattern explains why GitOps is architecturally sound — not just a convention — and what you gain from separating the definition of a change from its execution.
Should you expose Kubernetes directly to developers or abstract it? That's not a tooling debate — it's a design question. The Facade pattern and the Golden Path concept explain why your platform's interface is its most important architectural decision.
Manual reconciliation doesn't scale. The Kubernetes controller loop is the Observer pattern at cluster scale. Learn how operators, Argo Rollouts, and External Secrets Operator implement this pattern — and how to decide when to write your own.
Environment provisioning without a clear creation model is slow, inconsistent, and doesn't scale. The Factory and Builder patterns explain why — and what Crossplane, Backstage, and Terraform modules have in common.
Platform engineers apply design patterns every day without knowing it. This series bridges the gap between software architecture vocabulary and the real decisions you make in platform engineering.
bash ml-project/ ├── .git/ ├── src/ │ └── train.py ├── data/ │ └── training.csv 5GB, in Git └── models/ └── model.pkl 2GB, in Git
Terraform, Helm, GitOps... ¿para entrenar modelos? Sí.