Edge redirect with Lambda@Edge + DynamoDB Global Tables

Replicates the Split Cloudflare Workers template using Lambda@Edge and DynamoDB Global Tables instead of Cloudflare Workers and Durable Objects.

• Redirect at the edge: CloudFront viewer-request invokes Lambda@Edge, which runs the full Split/FME SDK in consumer_partial mode, reads the rollout plan from the nearest DynamoDB Global Table replica, evaluates getTreatment(key, featureFlag), and returns an HTML page showing the redirect target (or you can change it to a 302).
  • Treatment on → https://google.com
  • Treatment off (or control) → https://apple.com
• Sync Lambda: Runs on a schedule and populates DynamoDB with the rollout plan via the Split Synchronizer (DynamoDB storage wrapper).

Architecture

• CloudFront – Viewer-request → Lambda@Edge.
• Lambda@Edge – Full Split/FME SDK + DynamoDB wrapper; uses AWS_REGION so the DynamoDB client hits the nearest Global Table replica (lower latency).
• DynamoDB Global Table – One table with replicas in multiple regions (US, EU, APAC, SA). Sync writes in the primary region; edge reads from the nearest replica, or a fallback region when the edge runs in a region without a replica.
• Sync Lambda – EventBridge schedule → Synchronizer only (writes rollout plan to DynamoDB).

---

How to run the demo

1. Prerequisites

• AWS CLI configured (aws configure or env vars).
• Terraform >= 1.0.
• Node.js 20 (Terraform will run npm ci and zip the Lambdas).
• Harness account: Create a feature flag in FME and get a server-side SDK key (same environment as the flag).

2. Create the feature flag in Split

1. In Harness FME (e.g. Harness Feature Flags) create a feature flag (e.g. my_feature).
2. Define at least two treatments: on and off.
3. Set the default rule or targeting rules as you like (e.g. 100% on, or by key).
4. Copy the server-side SDK key for that environment (FME Settings → SDK keys → Server-side key).

3. Deploy with Terraform

Terraform packages the Lambdas by running npm ci in edge/ and sync/, which requires a package-lock.json in each directory. Generate the lockfiles once from the repo root:

cd edge && npm install && cd ..
cd sync && npm install && cd ..

Then deploy. From the repo root:

cd terraform
cp terraform.tfvars.example terraform.tfvars

Edit terraform.tfvars and set:

split_sdk_key = "YOUR_SPLIT_SERVER_SIDE_SDK_KEY"
# Optional: feature_flag_name = "my_feature"   # must match the flag name in Split

Deploy (Terraform will run npm ci and zip for both Lambdas, then create DynamoDB, Lambdas, CloudFront, EventBridge):

terraform init
terraform apply -var-file=terraform.tfvars

Confirm with yes. Note the outputs: cloudfront_domain, cloudfront_url, sync_lambda_name.

4. Run the first sync

The sync Lambda runs on a schedule (e.g. every 5 minutes). To have data immediately, invoke it once:

aws lambda invoke --function-name $(terraform output -raw sync_lambda_name) out.json
cat out.json

You should see {"ok":true,"message":"Synchronization finished"}. If you see an error, check that SPLIT_SDK_KEY is correct and the flag name exists.

5. Test the redirect

Use the CloudFront URL from the Terraform output:

# Replace with your CloudFront domain from: terraform output cloudfront_domain
curl -I "https://YOUR_DISTRIBUTION.cloudfront.net/"

You should get 200 with an HTML page showing the redirect target (and treatment/key). Depending on the treatment for the default key, it will show google.com or apple.com.

Per-user key (optional): the edge reads the key from the query string. If your flag is set up to segment by key:

curl -I "https://YOUR_DISTRIBUTION.cloudfront.net/?key=user123"
curl -I "https://YOUR_DISTRIBUTION.cloudfront.net/?key=user456"

Different keys can get different treatments (and thus different redirects) based on your Harness FME targeting rules.

6. Toggle the flag in Harness

• In Harness FME, change the flag to off (or adjust targeting).
• Wait for the next scheduled sync (e.g. 5 minutes), or invoke the sync Lambda again.
• Hit the same CloudFront URL again; the redirect should reflect the new treatment.

---

Project layout

• terraform/ – DynamoDB Global Table, sync Lambda, Lambda@Edge, CloudFront, EventBridge.
• sync/ – Sync Lambda: DynamoDB storage wrapper + Split Synchronizer only.
• edge/ – Lambda@Edge: full Split/FME SDK + DynamoDB wrapper, evaluates getTreatment at the edge.