What this covers
Architecture, deployment, scaling, failure drills, and security controls for a dual-AZ edge footprint with serverless ingestion and S3 origin failover.
Implementation trail
- Foundation and networking prerequisites
- Template deployment and DNS cutover
- Ingestion routing and fan-out patterns
- Scaling, cache, and AZ-failure handling
- Operations, observability, and security
Lay down VPC and edge foundations
Allocate two public subnets across AZs, attach an internet gateway, and ensure ALB and API Gateway endpoints are reachable through CloudFront.
- Parameterize VPC and subnet CIDRs in the template to match your landing zone conventions.
- Keep ALB and instance security groups minimal-HTTP/HTTPS from the world and SSH only from trusted CIDRs.
- Enable CloudFront origin failover so static assets can flow from S3 if the ALB origin degrades.
Deploy and cut over traffic
Deploy cf_templates/multi-az-edge-workload.yaml, then point DNS at the CloudFront distribution or attach a custom domain with ACM.
- Seed health checks for ALB targets before sending production traffic.
- Validate CloudFront behaviors for static assets versus API paths backed by API Gateway and Lambda.
- Warm caches for high-traffic paths to reduce cold-start latency during launch.
Keep ingestion online across AZs
Route ingestion to API Gateway (HTTP API) that triggers Lambda to fan out to Kinesis or S3; keep Lambda outside subnets unless VPC access is required.
- Enable throttling and JWT/authorizer controls on API Gateway to manage bursty producers.
- Use DLQs for Lambda and monitor function errors alongside API Gateway 5xx metrics.
- Keep stateful dependencies (RDS/DynamoDB) in Multi-AZ or global table mode for consistent ingest.
Scale and recover from AZ loss
Let the Auto Scaling Group span both AZs with cross-zone load balancing on the ALB; temporarily raise desired/max capacity when one AZ is impaired.
- Tune scaling policies on CPU or request counts and ensure instance profiles allow S3/CloudWatch access for bootstrap scripts.
- Document runbooks for promoting capacity in the healthy AZ and draining impaired nodes.
- Confirm CloudFront health checks continue to pass when only one AZ serves traffic.
Operate with observability and guardrails
- Create CloudWatch dashboards for ALB 5xx, Lambda errors, API Gateway latency, and ASG capacity.
- Mirror TLS enforcement end-to-end and attach AWS WAF to CloudFront for L7 protections.
- Schedule game days that rehearse CloudFront origin failover and AZ-isolation scenarios.
