[Bug] `node-lifecycle` label not present on AmazonLinux2023 nodegroups

[ebr]

What were you trying to accomplish?

Created new nodegroups and was expecting to see the node-lifecycle labels on the nodes, with spot or on-demand as potential values.

What happened?

The label isn't present on new nodes.

How to reproduce it?

eksctl create nodegroup -f nodegroups.yaml

nodeGroups:
  - name: ng-al2023
    amiFamily: AmazonLinux2023
    instancePrefix: worker-basic
    instancesDistribution:
      instanceTypes:
        - t3.xlarge
        - m4.xlarge
      onDemandBaseCapacity: 1
      onDemandPercentageAboveBaseCapacity: 0
      spotAllocationStrategy: 'capacity-optimized'
    minSize: 1
    maxSize: 5
    desiredCapacity: 1
    labels:
      my-own-label: value
    taints:
      - effect: NoSchedule

Logs

2025-07-08 15:41:00 [ℹ] will use version 1.31 for new nodegroup(s) based on control plane version
2025-07-08 15:41:03 [ℹ] nodegroup "ng-al2023" will use "ami-07947e1daef0a863a" [AmazonLinux2023/1.31]
2025-07-08 15:41:05 [ℹ] 1 nodegroup (ng-al2023) was included (based on the include/exclude rules)
2025-07-08 15:41:05 [ℹ] will create a CloudFormation stack for each of 1 nodegroups in cluster "..."
...
2025-07-08 15:44:10 [ℹ] waiting for at least 1 node(s) to become ready in "ng-al2023"
2025-07-08 15:44:10 [ℹ] nodegroup "ng-al2023" has 1 node(s)
2025-07-08 15:44:10 [ℹ] node "..." is ready
2025-07-08 15:44:10 [✔] created 1 nodegroup(s) in cluster "..."
2025-07-08 15:44:10 [✔] created 0 managed nodegroup(s) in cluster "..."
2025-07-08 15:44:13 [ℹ] checking security group configuration for all nodegroups

Anything else we need to know?

The only two changes from the working configuration were:

• upgrade eksctl from 0.201.0 to 0.210.0
• switch to Amazon Linux 2023

Versions

$ eksctl info
eksctl version: 0.210.0
kubectl version: v1.30.0
OS: linux

[ebr]

The labels are supposed to be set in the bootstrap helper script:

eksctl/pkg/nodebootstrap/assets/scripts/bootstrap.helper.sh

Line 22 in 9de2b87

NODE_LABELS="${NODE_LABELS},node-lifecycle=${INSTANCE_LIFECYCLE},alpha.eksctl.io/instance-id=${INSTANCE_ID}"

But neither node-lifecycle nor alpha.eksctl.io/instance-id labels are set on the node

I verified that the values are present in the metadata, so the way the script is trying to retrieve them should continue to work.

[ebr]

Upon further investigation, AL2023 no longer supports the bootstrap script: https://github.com/eksctl-io/eksctl/blob/main/pkg/nodebootstrap/README.md?plain=1#L1-L6.

This makes it unclear how to label nodes based on EC2 metadata, such as node-lifecycle, going forward

Also, how to execute commands on bootstrap. For example, mounting the instance store disks, or anything else that was previously done via preBootstrapCommands

[sapphirew]

I'm able to reproduce this using AL2023 node. As a workaround, I think you can use managedNodeGroup by specifying spot: true or not specifying this value. The provisioned node would have label eks.amazonaws.com/capacityType: SPOT|ON_DEMAND. I will need to take a deeper look on how to label nodes with node-lifecycle etc for AL2023.