Mini Edge Agent

A minimal Go-based Kubernetes controller simulator that mimics basic edge agent behavior.

Overview

This project demonstrates how a lightweight edge-like controller can:

• Watch Kubernetes resources
• React to cluster state changes
• Enable or disable workloads based on connectivity conditions

Inspired by Kubernetes controller patterns and edge computing concepts (e.g., KubeEdge).

Features

• Written in Go
• Uses client-go informers
• Watches Deployments
• Simulates edge online/offline behavior
• Scales workloads automatically

How It Works

1. The agent connects to the Kubernetes API.
2. It watches Deployment resources using informers.
3. Based on a flag (--edge-online), it:
   • Scales deployments to zero when the edge is offline.
   • Leaves them running when online.

Usage

Start a local cluster

minikube start

Create a test deployment

kubectl create deployment demo --image=nginx

Run the Mini Edge Agent

go run main.go --edge-online=false

Verify behavior

kubectl get deployments

You should see the deployment scaled to 0 replicas when the edge is offline.

Example Scenario

This simulates a situation where an edge node becomes unavailable. The agent reacts by disabling workloads to prevent inconsistent or unnecessary execution while the edge is offline.

Project Structure mini-edge-agent/ ├── main.go ├── go.mod ├── go.sum ├── config/ │ └── config.yaml

inspiration

This project was created to understand Kubernetes controller mechanics and explore how edge agent behavior can be simulated using core Kubernetes primitives without building a full edge platform.

Future Improvements

1. Restore original replica counts when the edge comes back online
2. Filter deployments using labels
3. Read configuration from a YAML file
4. Introduce a CRD-based control mechanism
5. Simulate node-level edge conditions