AWS Fargate for EKS
What you'll learn: How to run Kubernetes pods without managing any EC2 nodes using AWS Fargate profiles.
The Serverless Kubernetes Experience
Until now, your EKS cluster required node groups—either managed or self-managed EC2 instances that run your pods. AWS Fargate for EKS changes this entirely. With Fargate, you define which pods should run "serverlessly," and AWS provisions the exact compute capacity needed for each pod on-demand. You never see, patch, or pay for the underlying nodes.
Think of it like ordering food delivery: with node groups, you buy a whole restaurant kitchen (nodes) and manage it yourself. With Fargate, you simply order individual meals (pods) as needed—no kitchen maintenance required.
How Fargate Profiles Work
A Fargate profile tells EKS which pods to run on Fargate based on:
- Namespace
- Labels
When a pod matches a Fargate profile's selectors, EKS automatically:
- Provisions isolated compute capacity for that pod
- Launches the pod on Fargate infrastructure
- Tears down the compute when the pod terminates
You pay only for the vCPU and memory your pods actually consume, with per-second billing.
When to Use Fargate
Fargate excels for:
- Variable workloads where node capacity planning is difficult
- Security-sensitive applications requiring pod isolation
- Batch jobs and scheduled tasks
- Eliminating node patching and maintenance overhead
However, Fargate has limitations: no DaemonSets, no privileged containers, and no host networking. Certain workloads still need traditional node groups.
Key Takeaway: AWS Fargate for EKS lets you run pods without managing nodes by defining Fargate profiles that match specific namespaces and labels, giving you truly serverless Kubernetes compute.