GlobeBounds - Continuous Deployment Pipeline

GlobeBounds is an interactive geospatial visualisation tool that allows users to explore country boundaries using GeoJSON data. Users can select a country and view its precise polygon shape, with the map dynamically adjusting to focus on that region.

Overview

GlobeBounds is deployed using a fully automated continuous deployment pipeline built with GitHub Actions, Amazon S3, and Amazon CloudFront.

The pipeline ensures that every code change is automatically built, validated, and deployed to production without any manual steps.

What This Project Demonstrates

• End to end continuous deployment pipeline setup
• Secure AWS authentication using least privilege IAM
• Automated frontend build and deployment
• Debugging real deployment issues related to CDN caching

Tech Stack

• GitHub Actions
• AWS S3
• AWS CloudFront
• Vite and TypeScript
• ESLint

Deployment Workflow

Every push to the repository triggers the deployment pipeline.

Pipeline Steps

1. Checkout the latest code from the repository
2. Install project dependencies
3. Build the application and run lint checks
4. Authenticate with AWS using a least privilege IAM user
5. Sync the /dist folder to S3 and remove outdated files

This ensures the production site always reflects the latest successful build.

Secure AWS Integration

The pipeline uses a least privilege IAM user rather than full access credentials. Permissions are restricted to:

• Uploading files to the S3 bucket
• Triggering CloudFront cache invalidation

This follows standard security practices for production deployments.

Challenge: Deployment Not Updating

During development, deployments were completing successfully but changes were not visible on the live site.

Root Cause

CloudFront was serving cached content from edge locations, resulting in stale files being delivered to users.

Solution

A cache invalidation step was added to the pipeline:

aws cloudfront create-invalidation --distribution-id <ID> --paths "/*"

This forces CloudFront to fetch updated files from S3 after each deployment.

Results

• Fully automated deployment pipeline
• No manual uploads required after setup
• Reliable production updates on every push
• Improved understanding of CDN caching behaviour

Evidence

Pipeline Execution

S3 Deployment

Live Site

Key Learnings

• Deployment pipelines must account for caching layers such as CDNs
• CloudFront invalidation is required for consistent updates
• Build output paths must align with deployment commands
• IAM permissions should be scoped to the minimum required access