ReplicaSet self-healing lab
Kubernetes lab: a Node.js app with a deliberate failure endpoint, run as a three-pod ReplicaSet behind a NodePort Service, to watch the cluster restart failed containers.
- PERIOD
- "Oct 2026"
- STACK
- "Kubernetes, Docker, Node.js, Minikube"
- UPDATED
- "Oct 2026"
Overview
A short lab for watching self-healing happen. The app has a /failure endpoint that stops its own container. Run it as a ReplicaSet of three pods, call that endpoint, and watch a pod drop into Error and come back.
It also covers the plumbing around it: building the image, pushing it to a local registry, and configuring containerd on the node to pull from that registry.
What I built
- Node.js app with a deliberate failure point, and its Dockerfile.
- ReplicaSet and NodePort Service manifests.
- Step-by-step guide for Minikube on the kvm2 driver with a local image registry.
How it fits together
- curlfrom the host
- nodePort 30008
- ServiceNodePort, port 80
- targetPort 3000
- ReplicaSet3 pods
GET /failure stops one container. The pod shows Error, then restarts.
In the repository
$ git clone https://github.com/amga-d/node-app-failure-point.git
.
├── .gitignore
├── Dockerfile
├── package-lock.json
├── package.json
├── README.md
├── ReplicaSet.yaml
├── server.js
└── Service.yaml
- caff827inital commit: node-app-failure-point-v1
- JavaScript 73.2%
- Docker 26.8%