Amazon EKS — Node Management
EKS gives you three ways to run your workloads: managed node groups, self-managed nodes, and Fargate (serverless). Choosing the right compute option — and the right instance types — directly impacts cost, performance, and operational complexity.
Topics Covered
| # | Topic | Description |
|---|---|---|
| 1 | Managed Node Groups | AWS-managed EC2 nodes with automated updates |
| 2 | Self-Managed Nodes | Full control over node lifecycle with ASGs |
| 3 | Fargate Profiles | Serverless pods — no nodes to manage |
| 4 | Bottlerocket & AMI Types | Purpose-built OS and AMI options for EKS |
| 5 | Spot & Graviton Instances | Save up to 90% with Spot and ARM-based nodes |
| 6 | Hands-On Lab | Create and manage different node types |
| 7 | Practice Questions | Test your node management knowledge |
Compute Options Compared
| Feature | Managed Node Groups | Self-Managed Nodes | Fargate |
|---|---|---|---|
| Node management | AWS handles AMI updates, draining | You manage everything | No nodes — AWS manages |
| Scaling | ASG-backed, auto-scaling ready | ASG-backed, manual config | Per-pod, automatic |
| DaemonSets | Supported | Supported | Not supported |
| GPU workloads | Supported | Supported | Not supported |
| Cost | EC2 pricing | EC2 pricing | vCPU + memory per second |
| Customization | Launch template | Full control | Limited (resource requests only) |
| Best for | Most production workloads | Custom AMI, special kernels | Burst, batch, isolated workloads |
Recommendation: Start with managed node groups for most workloads. Add Fargate for batch jobs or dev namespaces. Use self-managed only when you need deep customization.
