It's highly recommended to use AWS managed or custom load balancer. This implementation relies on the Route53 MultiValueAnswer policy.

• Description
  • Best practice?
  • Requirements
  • What's being provisioned?
  • Todo

Description

This terraform template creates an AWS Auto Scaling Group with network based scaling policy and their CloudWatch alarms among with auto DNS updates on instance start and terminate by sending notifications from Auto Scaling Group lifecycle hooks to SNS. This SNS notifications trigger python based lambda functions to update your DNS entities hosted in Route53

Best practice?

This implementation may cover your needs when you cannot work with a load balancer due to specific network protocols or such. It applies DNS round-robin by using Route53 MultiValueAnswer policy.

Requirements

You have to edit variables.tfvars, especially the ones with your-### such as AMI name, VPC name, etc.

What's being provisioned?

• It deploys to your default VPC in eu-central-1 region which can be configured in variables.tfvars
• An Auto Scaling Group with 2 desired instances with the launch template defined below
• A launch template with the most recent Amazon Linux 2 AMI; amzn2-ami-hvm*
• 2 scaling policy and cloudwatch alarms that trigger based on networkOut metric of the ASG instances.
• 2 CloudWatch log groups for lambdas with 30 days of retention
• A security group with one TCP ingress, the ingress port defined as a terraform variable.
• 2 Lambda functions to register and deregister DNS
• 2 SNS topics to trigger register and deregister Lamdas
• An IAM role for lambda funciton to let them update the DNS (!! This is shitty, it uses AdministratorPolicy and needs to be updated by least privillaged scope !!)
• It's not using S3 or any other backend as a state management for the sake of the POC.

Todo

• Replace Admin level policy
• Estimated cost
• Add architecture diagram
• Lamda layers-
• Add license