A poll application deployed on Kubernetes, from source code to the cluster
- context
- EPITECH · team of 2
- period
- Oct 2026
- platforms
- Kubernetes · Docker · Azure · Bash

> cat context.md
An EPITECH DevOps project, built with Quentin Dutournier-Ottenin: take the poll application from source code to a Kubernetes cluster. My part: the Kustomize base and overlays, PostgreSQL and Redis, the Azure setup, the build, push and deploy scripts, the PodDisruptionBudgets and the documentation. Quentin wrote the poll, worker and result Deployments and Services, the autoscalers, the rolling updates and the spread of replicas across nodes.
> cat how-it-works.md
- 01Kustomize manifests describe every service, with a shared base and one overlay per target: Minikube for local work, AKS on Azure.
- 02PostgreSQL runs as a StatefulSet with its own persistent volume and loads its schema on first start. Redis requires a password, and both passwords come from a Kubernetes Secret built from a local .env file.
- 03poll and result scale with HorizontalPodAutoscalers, spread their replicas across nodes and keep a PodDisruptionBudget, so rolling updates and node drains cause no downtime.
- 04deploy.sh checks the tools and the kubectl context, builds the images tagged with the Git commit, pushes them to Azure Container Registry and applies the overlay with that tag. The published images can be pulled without an account.