with Automated Failover, Data Replication, and Business Continuity for Critical Systems
This project implements a production-grade cloud-native disaster recovery platform on AWS. The platform automatically detects primary region failure and switches traffic to a standby DR region without any human intervention. It combines Infrastructure as Code, GitOps, AI-powered security monitoring, and automated backup to achieve business continuity for critical systems.
Primary Region: AWS us-east-1 DR Region: AWS us-west-2 RTO Target: 15 minutes RPO Target: 5 minutes
A multi-region active / warm-standby topology on AWS. Route 53 health checks gate traffic between us-east-1 (primary) and us-west-2 (DR). GitOps via ArgoCD keeps both clusters in lockstep with the repository, while Velero replicates cluster state to S3 with cross-region replication. An AI layer — Llama 3.3 70B via Groq — handles anomaly detection on the OpenSearch logging pipeline and ISO 22301-aware drift analysis on driftctl reports.
Failover sequence (drill-measured: 6m 42s end-to-end)
1. Route 53 health check fails 3× consecutive (~60s detection)
2. CloudWatch alarm → SNS → EventBridge
3. Lambda promotes RDS read replica (~5-8 min)
4. Lambda scales DR EKS nodegroup 1 → 3 nodes
5. Route 53 DNS shifts to DR load balancer
6. ArgoCD reconciles workloads to Git HEAD
7. SNS email to operator — service restored
- Two VPCs across us-east-1 and us-west-2 with public and private subnets
- Two EKS clusters (Kubernetes 1.30) with managed node groups
- RDS PostgreSQL 16 primary with cross-region read replica
- S3 buckets with cross-region replication for Velero backups
- Lambda function for automated failover orchestration
- Route53 health checks with automatic DNS failover
- 18 IAM roles using IRSA (IAM Roles for Service Accounts)
- DynamoDB table for Terraform state locking
- AWS Config for resource configuration recording
- ArgoCD — app-of-apps GitOps pattern syncing all components from GitHub
- OPA Gatekeeper — admission controller enforcing ISO 22301 policies
- Prometheus + Grafana — 7 dashboards, 149 alert rules
- OpenSearch + Fluent Bit — centralised SIEM with 113,000+ log entries
- Velero — daily automated backups with 72-hour retention
- AI Anomaly Detector — Kubernetes CronJob running every 5 minutes. Queries OpenSearch for security-relevant logs and sends them to Groq llama-3.3-70b-versatile for threat analysis. Fires SNS alert for MEDIUM and above threats. Results indexed to OpenSearch security-analysis index.
- AI Drift Analyzer — Always-running Deployment. Reads driftctl infrastructure scan reports every 6 hours and sends to Groq AI for risk assessment. Returns risk level, affected resources, remediation steps, and ISO 22301 compliance impact.
- driftctl CronJob — Compares live AWS infrastructure against Terraform state every 6 hours.
- Route53 health check detects primary failure in ~60 seconds
- CloudWatch alarm fires
- EventBridge triggers Lambda
- Lambda promotes RDS read replica to standalone primary (~5-8 minutes)
- Lambda scales DR EKS nodegroup from 1 to 3 nodes
- Route53 DNS switches to DR load balancer
- SNS email alert sent to operator
| Component | Technology |
|---|---|
| Infrastructure as Code | Terraform 1.9 with reusable modules |
| Container Orchestration | AWS EKS 1.30 |
| Database | RDS PostgreSQL 16 with cross-region replica |
| Backup | Velero 1.14 + S3 Cross-Region Replication |
| GitOps | ArgoCD with app-of-apps pattern |
| Monitoring | Prometheus + Grafana (kube-prometheus-stack) |
| Logging / SIEM | OpenSearch + Fluent Bit |
| Policy Enforcement | OPA Gatekeeper |
| Drift Detection | driftctl + Groq llama-3.3-70b-versatile |
| AI Threat Detection | Groq llama-3.3-70b-versatile |
| Failover Orchestration | AWS Lambda (Python 3.12) |
| DNS Failover | Route53 health-check-based routing |
| Alerts | AWS SNS |
| Container Registry | AWS ECR |
| Standard | Implementation |
|---|---|
| ISO 22301 §8.4 — Business Impact Analysis | RTO/RPO targets defined and validated through automated health checks |
| ISO 22301 §8.5 — Business Continuity Strategy | Active/Passive multi-region architecture with automated promotion |
| NIST SP 800-34 §3.4 — Alternate Site | Independent us-west-2 DR region with separate VPC and RDS instance |
| NIST SP 800-34 §3.5 — Data Backup | Velero daily backups + RDS automated backups + S3 CRR |
| OPA Gatekeeper | All Deployments must have memory limits and liveness probes |
| Encryption at rest | EBS volumes encrypted, RDS storage encrypted, S3 SSE-AES256 |
| Secrets management | IRSA for pod-level IAM, no hardcoded credentials anywhere |
| Audit trail | OpenSearch logs, AWS CloudTrail, ArgoCD audit log |
cloud-native-dr-platform/
├── terraform/
│ ├── primary/ # us-east-1 infrastructure
│ ├── dr/ # us-west-2 infrastructure
│ └── modules/ # vpc, eks, rds, s3, route53
├── kubernetes/
│ ├── app/ # dr-status-monitor application
│ ├── argocd/ # app-of-apps GitOps configuration
│ ├── monitoring/ # Prometheus and Grafana values
│ ├── logging/ # OpenSearch and Fluent Bit values
│ ├── velero/ # Backup schedules
│ ├── opa-gatekeeper/ # ISO 22301 admission policies
│ ├── ai-detection/ # AI anomaly detector CronJob
│ └── drift-detection/ # driftctl + AI analyzer
├── lambda/
│ └── failover_handler.py
├── docker/
│ └── app/ # Flask application and Dockerfile
├── ansible/
│ └── playbooks/ # Tool installation and deployment
└── docs/
└── architecture.png # Reference architecture diagram
# 1. Clone the repository
git clone https://github.com/ZaheerrAhmed/cloud-native-dr-platform.git
cd cloud-native-dr-platform
# 2. Configure AWS credentials
aws configure
# 3. Deploy primary region (us-east-1)
cd terraform/primary
terraform init
terraform apply
# 4. Deploy DR region (us-west-2)
cd ../dr
terraform init
terraform apply
# 5. Install Kubernetes tools via Ansible
cd ../../ansible
ansible-playbook playbooks/install-tools.yaml
ansible-playbook playbooks/deploy-k8s-tools.yamlIn October 2025, AWS us-east-1 suffered a 15-hour DNS outage affecting 2,500+ companies including Netflix, Reddit, and McDonald's with $581 million in insurance losses. This platform addresses exactly that scenario — automated cross-region failover that detects failure within 60 seconds and completes recovery in 15 minutes with zero human intervention.
Author: Zaheer Ahmad Programme: Diploma in Artificial Intelligence Operations — EduQual Level 6 Institution: Al-Nafi International College