Compare instance families and watch an Auto Scaling group replace an instance

Steps

1. Compare one size from each family. This only reads AWS's catalogue; nothing is launched:

aws ec2 describe-instance-types --instance-types m5.large c5.large r5.large i3.large g4dn.xlarge --query 'InstanceTypes[].[InstanceType,VCpuInfo.DefaultVCpus,MemoryInfo.SizeInMiB,InstanceStorageSupported]' --output table
aws ec2 describe-instance-types --instance-types g4dn.xlarge --query 'InstanceTypes[0].GpuInfo.Gpus[0].Name' --output text

m5 is general purpose, c5 compute optimized, r5 memory optimized (16384 MiB against c5's 4096 for the same 2 vCPUs), i3 storage optimized with local disks, and g4dn accelerated computing (the second command prints T4, its GPU). If a type is not offered in your Region, the command names it; remove it and run again. 2. Find a free-tier-eligible type in this Region:

aws ec2 describe-instance-types --filters Name=free-tier-eligible,Values=true --query 'InstanceTypes[].InstanceType' --output text

If the list includes t3.micro, set TYPE=t3.micro; otherwise set TYPE=t2.micro. Do not pick anything bigger. 3. Write a launch template: the latest Amazon Linux AMI, your free-tier type, and a network interface with no public IP address. The template itself costs nothing:

TYPE=t3.micro
cat > clf-lab14-lt.json <<EOF
{
  "ImageId": "resolve:ssm:/aws/service/ami-amazon-linux-latest/al2023-ami-kernel-default-x86_64",
  "InstanceType": "$TYPE",
  "NetworkInterfaces": [{"DeviceIndex": 0, "AssociatePublicIpAddress": false}],
  "TagSpecifications": [{"ResourceType": "instance", "Tags": [{"Key": "Name", "Value": "clf-lab14-instance"}]}]
}
EOF
aws ec2 create-launch-template --launch-template-name clf-lab14-lt --launch-template-data file://clf-lab14-lt.json

Change the first line if step 2 told you to use t2.micro. 4. Create the Auto Scaling group across two default subnets, so it spans two Availability Zones. Minimum, maximum and desired capacity are all 1:

SUBNETS=$(aws ec2 describe-subnets --filters Name=default-for-az,Values=true --query "join(',', Subnets[0:2].SubnetId)" --output text)
echo "$SUBNETS"
aws autoscaling create-auto-scaling-group --auto-scaling-group-name clf-lab14-asg --launch-template 'LaunchTemplateName=clf-lab14-lt,Version=$Latest' --min-size 1 --max-size 1 --desired-capacity 1 --vpc-zone-identifier "$SUBNETS"

The echo must show two subnet IDs separated by a comma; if it is empty, see the Notes. Then check every minute or so until the instance is InService:

aws autoscaling describe-auto-scaling-groups --auto-scaling-group-names clf-lab14-asg --query 'AutoScalingGroups[0].Instances[].[InstanceId,AvailabilityZone,LifecycleState]' --output text

5. Confirm the instance has no public IP address:

aws ec2 describe-instances --filters Name=tag:aws:autoscaling:groupName,Values=clf-lab14-asg Name=instance-state-name,Values=pending,running --query 'Reservations[].Instances[].PublicIpAddress'

It prints []. If it prints an address, go straight to the clean-up. 6. Break it and watch the group heal itself. Copy the instance ID from step 4 and terminate that instance while telling the group to keep its desired capacity:

aws autoscaling terminate-instance-in-auto-scaling-group --instance-id i-0123456789abcdef0 --no-should-decrement-desired-capacity

Put your own instance ID in place of the example. Run the describe-auto-scaling-groups command from step 4 again every minute: within a few minutes a new instance ID appears, possibly in the other Availability Zone, because the desired capacity is still 1. Read the group's own account of what happened:

aws autoscaling describe-scaling-activities --auto-scaling-group-name clf-lab14-asg --query 'Activities[].Description' --output text

7. Think about what is not here. In a real web application this group would sit behind an Application Load Balancer, which would give users one address, run health checks and send each request only to a healthy instance; the group would register each new instance with it. This lab leaves that piece out because Elastic Load Balancing is charged for every hour a load balancer exists. In your own notes, write which job is auto scaling's (how many instances) and which is the load balancer's (which instance gets each request).