99.95% Uptime: Multi-Region Kubernetes on AWS EKS

The Challenge

• ✗Frequent outages: 3-4 unplanned downtime incidents per month, costing $50K per incident in lost revenue
• ✗Slow deployments: 45-minute release cycles prevented rapid feature delivery
• ✗High infrastructure costs: Inefficient resource utilization cost $15,000/month without high availability
• ✗No disaster recovery: Single region deployment meant total loss on regional failure
• ✗Manual scaling: Ops team manually scaled infrastructure during traffic spikes

The Solution

We architected a multi-region Kubernetes infrastructure on AWS EKS with automatic failover, GitOps deployment, and advanced observability. Here's the complete technical breakdown:

1. Multi-Region EKS Architecture

Deployed 3 EKS clusters across us-east-1 (primary), us-west-2, and eu-west-1 with active-active configuration:

# Terraform configuration for multi-region EKS

# Primary Cluster (us-east-1)
resource "aws_eks_cluster" "primary" {
  name            = "production-primary"
  version         = "1.29"
  role_arn        = aws_iam_role.eks_role.arn

  vpc_config {
    subnet_ids            = aws_subnet.primary_public[*].id
    endpoint_private_access = true
    endpoint_public_access  = true
  }

  depends_on = [aws_iam_role_policy_attachment.eks_cluster_policy]
}

# Node Group (3 AZs for high availability)
resource "aws_eks_node_group" "primary" {
  cluster_name    = aws_eks_cluster.primary.name
  node_group_name = "production-primary-nodes"
  node_role_arn   = aws_iam_role.node_role.arn
  subnet_ids      = aws_subnet.primary_public[*].id

  scaling_config {
    desired_size = 9    # 3 nodes per AZ
    max_size     = 20
    min_size     = 9
  }

  instance_types = ["t3.xlarge"]

  tags = {
    Name = "Production Primary"
    Environment = "production"
    Region = "us-east-1"
  }
}

# Secondary Clusters (us-west-2, eu-west-1) - Same configuration

# Route53 Health Check for failover
resource "aws_route53_health_check" "primary" {
  fqdn              = aws_lb.primary.dns_name
  port              = 443
  type              = "HTTPS"
  failure_threshold = 3
}

# Route53 Weighted Routing (Active-Active)
resource "aws_route53_record" "weighted" {
  zone_id = aws_route53_zone.main.zone_id
  name    = "api.example.com"
  type    = "A"

  alias {
    name                   = aws_lb.primary.dns_name
    zone_id                = aws_lb.primary.zone_id
    evaluate_target_health = true
  }

  set_identifier = "primary"
  weighted_routing_policy {
    weight = 100
  }
}

2. GitOps with ArgoCD

Implemented ArgoCD for declarative, Git-driven deployments across all clusters:

• ✓Automated sync: Every commit to main branch auto-deploys within 5 minutes
• ✓Blue-green deployments: Zero-downtime releases with automatic rollback on failure
• ✓Progressive rollouts: 10% → 25% → 50% → 100% deployment strategy
• ✓Cross-cluster deployment: Single Git push deploys to all 3 regions simultaneously

3. Advanced Observability Stack

4. Disaster Recovery & Auto-Scaling

# Horizontal Pod Autoscaler (HPA)
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: api-server-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: api-server
  minReplicas: 9
  maxReplicas: 50
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 70
  - type: Resource
    resource:
      name: memory
      target:
        type: Utilization
        averageUtilization: 80
  behavior:
    scaleDown:
      stabilizationWindowSeconds: 300
      policies:
      - type: Percent
        value: 50
        periodSeconds: 15
    scaleUp:
      stabilizationWindowSeconds: 0
      policies:
      - type: Percent
        value: 100
        periodSeconds: 15

# Velero for cluster-level backups
velero backup-location create aws \
  --provider aws \
  --bucket $BUCKET \
  --secret-file ./credentials-velero

# Automated backup schedule
velero schedule create daily-backup --schedule="0 2 * * *"

The Results

Technical Stack Used

Related Services We Provided