Ref: https://learn.cantrill.io/courses/1820301/lectures/41301443

• Summary:
• Contents:

EC2 Auto Scaling Group (ASG) - Key Concepts

Diagram: https://github.com/acantril/aws-sa-associate-saac03/blob/main/1500-HA_and_SCALING/00_LEARNINGAIDS/ASG-1.png

• 🔧 Logical grouping of EC2 instances
  • Automates horizontal scaling and self-healing
  • Usually used together with LCs/LTs and ELBs to deliver elastic architectures
• Uses LCs/LTs to provision new instances when scaling out
  • ❗ An ASG has ONE single LC/LT associated to it at any given time
    • The LC/LT can be changed, but the previous one then gets disassociated
    • All launched instances launched follow the same LC/LT unless changed
• 💡 An ASG defines WHEN & WHERE, its LC/LT defines WHAT

EC2 ASG - X:Y:Z Definition

• ASG behavior can be defined with X:Y:Z → 3 super important values:
  1. X = Minimum size (MIN) = Minimum number of instances the ASG can have
  2. Y = Desired Capacity = Number of instances the ASG should have right now
  3. Z = Maximum size (MAX) = Maximum number of instances the ASG can have
  • 💡 e.g. 1:2:4 ASG → MIN=1; DESIRED=2; MAX=4
• 🔧 An ASG's job is to is to keep its running instances at the number specified in the DESIRED capacity → provisions/terminates instances to reach desired capacity
  • 💡 Examples:
    • If running instances=1 & DESIRED=2 → ASG provisions a new instance
    • If running instances=3 & DESIRED=2 → ASG terminates an instance
• DESIRED Capacity can be updated/adjusted:
  • Manually (no scaling policy, no intelligence, customer updates themselves)
    • ASG will still perform necessary actions when DESIRED is updated manually
  • Automatically via Scaling Policy (based on metrics)
    • e.g. increase desired capacity if CPU load is high
    • ❗ DESIRED Capacity always adjusted within the MIN-MAX range set by customer!

EC2 ASG - Architecture

Diagram: https://github.com/acantril/aws-sa-associate-saac03/blob/main/1500-HA_and_SCALING/00_LEARNINGAIDS/ASG-2.png

• ASGs are linked to a VPC → ASGs can span multiple subnets & AZs within that VPC
  • ASG instances launched in the configured subnets
  • By default, ASG attempts to keep the number of instances across subnets as even as possible
    • ❗ Not always the case though → see AZRebalance in ASG Scaling Processes
• Health monitoring
  • Default: EC2 health checks
    • If an EC2 instance fails, ASG will replace it → EC2 self-healing (don't confuse with EC2 Auto Recovery feature!)
  • Other health checks (e.g. ALB health checks) can be configured
• ❗ Easy to configure simple EC2 instance recovery
  • Configuration:
    1. Set a LT with desired configuration
    2. Set ASG to use multiple subnets in different AZs
    3. Set 1:1:1 to ASG
  • 👍 Cheap, simple, and effective HA for one instance (a nice trick)
    • Instance will self-recover if it fails
    • Instance will self-recover in a different AZ if an AZ fails