Retail Edge Backbone (GCP & Tailscale)

This project demonstrates a professional, hybrid Cloud-Edge architecture designed for the retail industry. It focuses on high-performance data synchronization, local store autonomy, and a Zero Trust security model.

---

Core Pillars

• Cloud Backbone (GCP): Centralized control plane and data aggregation using GKE and Artifact Registry.
• Retail Edge (Store): Lightweight, autonomous nodes running K3s for local resilience and low-latency processing.
• Zero Trust Mesh: Secure, identity-based communication via Tailscale (WireGuard), eliminating public IP exposure.
• SRE Focused: Built with the 4 Golden Signals (Latency, Traffic, Errors, Saturation) in mind using Go-native instrumentation.

Architecture Overview

The system bridges the gap between decentralized retail locations and a centralized Google Cloud environment:

1. Infrastructure as Code: Fully managed via Terraform for reproducible environments.
2. Networking: A private overlay mesh connects all nodes. Edge nodes join the network automatically using OAuth2 authentication.
3. Security: Automated TLS (Let's Encrypt) for all internal services via Tailscale, enabling secure browser-native APIs (e.g., Camera/Scanner) in-store.

Architecture Decisions (ADR)

ADR 001: Switching from GKE to k3s for Edge Nodes

• Context: Local simulation of edge nodes using GKE resulted in ~80% CPU saturation on development hardware.
• Decision: Replaced GKE with k3s for edge environments.
• Consequence: Reduced resource footprint significantly while maintaining Kubernetes API compatibility, better mirroring actual retail hardware constraints.

ADR 002: Multi-Stage Distroless Builds

• Context: Retail store connectivity can be unstable or bandwidth-limited.
• Decision: Implemented multi-stage Docker builds using gcr.io/distroless/static.
• Consequence: Reduced image size from ~800MB to ~20MB, improving deployment speed and security (reduced attack surface).

Observability & SRE

This project is instrumented to monitor the 4 Golden Signals:

• Latency: Tracking sync duration between Edge and GCP.
• Traffic: Measuring the number of processed retail transactions.
• Errors: Monitoring 5xx rates and failed sync attempts.
• Saturation: Observing CPU/Memory pressure on k3s nodes.

Quick Start (Automated via Makefile)

The project uses a Makefile to abstract complex workflows. Ensure you have the GCP SDK and Tailscale installed.

1. Initialize Infrastructure

make infra-init
make infra-up

Note on Security: This repository uses Gitleaks in the CI pipeline. All previously exposed test keys have been revoked, rotated, and purged from the git history using git filter-repo.