Kubernetes and AI Meetup Lecture Notes

Introduction

• Speaker: Siva Ram
• Community: Active Kubernetes and AI Meetup
• Followers: 14,000+ on Meetup, 9,000+ on LinkedIn
• Mission: Accelerate Kubernetes adoption through shared learning.

Community Growth

• Global footprint:
  • Australia (Melbourne, Sydney)
  • Canada
  • UK
  • Singapore
  • Philippines
• Upcoming events:
  • August 15 in Australia
  • August 29 in UK
  • September 4 in Australia
• Open invitation for volunteers and speakers
• Follow us on LinkedIn, Twitter, and Telegram.

Certifications and Promotions

• Cub Strut Program:
  • Complete five certifications with Linux Foundation: CKA, CKAD, CKS, KCNA, KCA.
  • Promo codes available for bundles and individual certifications.
  • 25% off total certifications, and up to 48.5% off on bundles.

Today's Topic: Scaling with Carpenter on EKS (AWS Cloud)

• Speaker: Gladwin Neo
• Introduction to Carpenter:
  • Carpenter as a "Cluster Autoscaler with steroids".
  • Purpose: Efficiently manage resources in Kubernetes clusters.

Scaling Concepts

• Cluster Autoscaler:
  • Scales worker nodes based on resource requests.
  • Responds to thresholds (e.g., CPU utilization).
• Horizontal Pod Autoscaler (HPA):
  • Scales Pods based on metrics like CPU or memory usage.

Carpenter Features

1. What is Carpenter?
   • Flexible, efficient scaling tool for Kubernetes on AWS.
2. How Carpenter Works:
   • Bypasses traditional node group management.
   • Uses EC2 Fleet API to provision new nodes directly.
3. Benefits of Carpenter:
   • Removes complexity of managing multiple node groups.
   • Supports diverse workloads and instance types (e.g., CPU architectures).
   • Consolidation: Reduces overall cost by optimizing node usage.
4. Best Practices:
   • Avoid running Carpenter Pods on nodes managed by Carpenter.
   • Deploy on EC2 instances for greater control.
   • Use diverse instance types for better cost optimization.
   • Enable consolidation feature for underutilized nodes.
   • Utilize custom resource definitions for fine-tuning Node Pools.

Demonstration

• Comparison between Carpenter and Cluster Autoscaler in scaling workloads.
• Carpenter optimally spins up fewer, larger instances compared to Cluster Autoscaler.

Questions and Answers

• Encouragement for questions, fostering a collaborative learning environment.
• Discussion about scaling best practices, resource management, and community involvement.

Closing

• Reminder that Carpenter is currently open-source and available for AWS EKS.
• Community involvement and future collaboration opportunities.