"It took one tense meeting to realize we were grabbing knives before deciding what to cook. So we paused and asked: What journeys should a dev complete without opening a Slack thread? What needs to be self-served? What's our "platform MVP"? We weren't building a cathedral.We were starting with one dish: deploying a backend service safely. We chose principles first:✅ GitOps for consistency✅ IaC for transparency✅ Guardrails, not gates The tech came after."
"Remember that kitchen from Part 1? Loud, frantic. Devs yelling for infra help like chefs yelling for onions. Then came Part 2 - we laid out stations, built mise-en-place, and brought calm. But now, we stood in front of a blank floor plan.The owner (read: CTO) said: "Go build the kitchen." Where do you begin?"
"The dish we picked? "New backend service → CI → deploy → rollback" Sounds simple. Wasn't. Some devs were skeptical."Isn't this just another infra ticket wrapped in a CLI?" But we mapped every touchpoint: GitHub push triggers GitHub Actions → container build ArgoCD syncs the Helm chart to Kubernetes Sentry, Prometheus, Grafana wired in"
Chaos preceded the platform: developers faced frantic, ad-hoc infrastructure demands and a blank floor plan from leadership. The team prioritized principles before tools, defining a platform MVP focused on one end-to-end journey. The chosen path automated "New backend service → CI → deploy → rollback", enforced GitOps and Infrastructure-as-Code, and favoured guardrails over gates. Implementation mapped every touchpoint: GitHub Actions for builds, ArgoCD for syncing Helm charts to Kubernetes, and monitoring wired into Sentry, Prometheus, and Grafana. The approach emphasized starting small, iterating on a single golden path, and adding tooling after validating principles.
Read at Medium
Unable to calculate read time
Collection
[
|
...
]