AWS re:Invent 2024 - Automate your entire Kubernetes cluster with Amazon EKS Auto Mode

Have you ever wished setting up and managing Kubernetes could be easier? At AWS re:Invent 2024, Alex Kestner (Senior Product Manager, AWS) and Todd Neal (Senior Software Engineer, AWS) introduced a new feature for Amazon EKS: Auto Mode. This blog post will explain the main ideas from their talk, showing you what Amazon EKS Auto Mode does and how it can help manage Kubernetes clusters.

The Evolution of Kubernetes Management

Before diving into EKS Auto Mode, let's take a quick look at how we got here:

• Cloud Computing Revolution: Alex reminded how AWS fundamentally changed application development with services like Amazon S3 and Amazon EC2 in 2006 [12].
• Rise of Kubernetes: As cloud systems became more complicated, Kubernetes became the most popular way to manage container-based apps. Today, 84% of organizations surveyed by the Cloud Native Computing Foundation are using Kubernetes or trying it out.
• Amazon EKS Launch (2017): AWS introduced Amazon EKS to simplify Kubernetes management, handling the control plane while customers managed the worker nodes.
• Continuous Improvement: Over seven years, AWS made over 250 enhancements to EKS, covering areas like compute management, security, scalability, and networking.

The Challenge: Data Plane Management

Even with these improvements, customers still had to do some work to manage the part of the cluster where their apps run. This usually meant:

1. Choosing cluster architecture
2. Selecting and configuring essential Kubernetes plugins
3. Provisioning and configuring compute resources
4. Continuous monitoring and maintenance
5. Cost optimization

Introducing Amazon EKS Auto Mode

Amazon EKS Auto Mode makes these tasks much easier. Here's what it does:

1. One-Click Cluster Creation: Auto Mode provides pre-configured clusters based on best practices derived from AWS's experience running millions of clusters. This allows you to quickly set up a production-ready Kubernetes environment with minimal effort.
2. Managed Compute: Auto Mode introduces a new level of compute management. It automatically selects and scales Amazon EC2 instances based on your application's needs, using new EC2 managed instances that delegate lifecycle management to AWS. These instances run on Bottlerocket OS, a container-optimized operating system that improves both security and performance.
3. Integrated Networking: Networking in Auto Mode is streamlined and fully integrated. It includes an improved VPC CNI for efficient pod networking, CoreDNS running locally on each node for fast service discovery, and managed load balancer integration for easy external access to your services.
4. Simplified Storage: Storage management is made easier with Auto Mode's built-in EBS CSI driver, allowing for straightforward provisioning of block storage. It also introduces a new storage class specifically designed for Auto Mode, further simplifying persistent storage setup for your applications.
5. Automatic Optimization: Auto Mode continuously analyzes and adjusts compute resources to ensure cost efficiency. It supports various Amazon EC2 purchase options, including Spot instances and Reserved Instances, allowing you to leverage cost-saving strategies while maintaining performance.
6. Health Monitoring and Auto-Repair: Keeping your cluster healthy is a key feature of Auto Mode. It actively monitors the health of your nodes and automatically replaces any that are found to be faulty, ensuring high availability for the applications.
7. Easy Upgrades: Cluster version upgrades, which can be complex and time-consuming, are simplified with Auto Mode. It handles the upgrade process with minimal user intervention, reducing the operational burden of keeping your Kubernetes environment up to date.

You can see all the features in details on the Amazon EKS Auto Mode documentation.

EKS Auto Mode in Action

Todd Neal provided a live demonstration of EKS Auto Mode's capabilities:

• Cluster Creation: With just a few clicks, a fully functional EKS Auto Mode cluster was created, initially scaled to zero (no nodes running).
• Application Deployment: Todd deployed a sample retail application, showcasing how Auto Mode automatically provisioned and scaled nodes as needed.
• Cluster Upgrade: Upgrading from Kubernetes 1.30 to 1.31 was demonstrated with a single click. Auto Mode handled the complex process of migrating workloads to new nodes running the updated version.
• Health Monitoring: Todd induced a failure on a node to show how Auto Mode detected the issue, reported it via Kubernetes-native constructs, and automatically replaced the faulty node.

Kubernetes Conformance and Ecosystem

Alex made it clear that EKS Auto Mode still works with all standard Kubernetes tools. This means you can use the same tools you're already familiar with. Many partner companies have already tested their products with EKS Auto Mode, and lots of businesses are now using it in their daily operations.

Key Takeaways

Amazon EKS Auto Mode helps Kubernetes users on AWS in several ways. It makes it faster to start new apps and update old ones. It also makes daily cluster management and upgrades easier. Auto Mode saves money by automatically choosing the right size and type of computers for your apps. It also improves security by using a special operating system (Bottlerocket) and updating it automatically. Overall, EKS Auto Mode makes Kubernetes easier to use and manage. By doing many hard tasks automatically, it lets teams spend more time building apps instead of managing the system.

To learn more about EKS Auto Mode, Alex Kestner shared the ongoing Event series where you can go deep into the Auto Mode capabilities. He also shared the EKS Auto Mode guide and the launch blog. These will help you understand more about how this new way of running Kubernetes on AWS works and how it can help.

For those who want to watch the full presentation, you can find the recording on the AWS YouTube channel. The video provides a more detailed explanation and demonstration of EKS Auto Mode's features.