What the node termination handler does

The AWS node termination handler runs on each Spot instance and watches for termination notifications. When it detects a signal that an instance is going to be reclaimed, it helps your workloads exit cleanly. This improves application stability and reduces interruptions.

The handler helps the Kubernetes control plane respond correctly to events that make a node unavailable, such as:

• EC2 maintenance events
• Spot interruptions
• Auto Scaling group scale‑in
• Auto Scaling group Availability Zone rebalancing
• Manual EC2 instance termination

Without a handler, pods may not shut down gracefully, and your applications may take longer to recover.

How the node termination handler works

When the DaemonSet detects a termination event, it:

1. Identifies that the Spot instance is being reclaimed.

2. Uses the 2‑minute Spot interruption notice to prepare the node.

3. Taints and cordons the node so new pods aren’t scheduled.

4. Drains existing connections from running pods.

5. Lets Kubernetes reschedule replacement pods on healthy nodes.

Using the node termination handler with Ocean

Ocean doesn’t conflict with the AWS node termination handler. You can install the handler on your nodes if you want to manage graceful shutdown logic. However, it’s not required.

Ocean already analyzes how your containers use compute resources and automatically launches or removes nodes as needed. Ocean ensures that your workloads stay available while also scaling down underutilized nodes to reduce costs.