Multi-AZ web tier on AWS: Auto Scaling groups behind an Application Load Balancer

#Architecture

            ALB (subnets in AZ a, b, c)
     ┌─────────────┼─────────────┐
  ASG instance   ASG instance   ASG instance
  (ap-south-1a)  (ap-south-1b)  (ap-south-1c)

#1. Launch template

AMI built by a pipeline (EC2 Image Builder), IMDSv2 required, instance profile, user data that only pulls config and starts the app. No hand-built servers.

#2. Auto Scaling group

aws autoscaling create-auto-scaling-group --auto-scaling-group-name asg-web \
  --launch-template LaunchTemplateName=lt-web,Version='$Latest' \
  --min-size 3 --max-size 12 --desired-capacity 3 \
  --vpc-zone-identifier "subnet-a,subnet-b,subnet-c" \
  --target-group-arns arn:aws:elasticloadbalancing:ap-south-1:111122223333:targetgroup/tg-web/abc \
  --health-check-type ELB --health-check-grace-period 120 \
  --default-instance-warmup 120

• --health-check-type ELB: instances failing the load balancer health check get replaced, not only those failing EC2 status checks.
• Spread across AZs is automatic; ASG rebalances when an AZ comes back.

#3. Target group health check

aws elbv2 modify-target-group --target-group-arn <arn> \
  --health-check-path /healthz --health-check-interval-seconds 15 \
  --healthy-threshold-count 2 --unhealthy-threshold-count 3 --matcher HttpCode=200
aws elbv2 modify-target-group-attributes --target-group-arn <arn> \
  --attributes Key=deregistration_delay.timeout_seconds,Value=30

Keep /healthz fast and local. If it calls the database and the database blips, every instance fails at once and the ASG starts replacing healthy servers.

#4. Scaling policy

Target tracking on ALBRequestCountPerTarget or average CPU:

aws autoscaling put-scaling-policy --auto-scaling-group-name asg-web --policy-name cpu50 \
  --policy-type TargetTrackingScaling \
  --target-tracking-configuration '{"PredefinedMetricSpecification":{"PredefinedMetricType":"ASGAverageCPUUtilization"},"TargetValue":50}'

#5. Plan for losing an AZ

With 3 AZs, losing one removes ~33% of capacity. Either keep enough headroom (min 3, target ~50–60% CPU) or rely on scaling, and confirm the instance types have capacity in the remaining AZs (use multiple instance types in a mixed instances policy).

#6. Deployments

• Instance refresh for new AMIs/launch template versions, with minimum healthy percentage (for example 90%) and checkpoints.
• Or blue/green with two target groups and weighted ALB listener rules.

aws autoscaling start-instance-refresh --auto-scaling-group-name asg-web \
  --preferences '{"MinHealthyPercentage":90,"InstanceWarmup":120}'

#Stateless is the requirement

Sessions in ElastiCache/DynamoDB, uploads in S3, config in Parameter Store. If an instance holds state, the ASG replacing it causes data loss.

Tip

Test with AWS Fault Injection Service: stop instances in one AZ, or simulate an AZ power interruption, and watch the ALB HealthyHostCount and error rate.