Does EKS Auto Mode automatically revert modifications to built-in general-purpose NodePool?

Yes, EKS Auto Mode does automatically revert modifications to built-in NodePools, and this behavior is intentional by design.

Understanding the Revert Behavior:

The built-in NodePools (general-purpose and system) created by EKS Auto Mode are managed resources with the managed-by: eks label. Documentation states that "You cannot modify the built-in node pools," this means AWS EKS actively reconciles these resources back to their intended configuration. When you patch these NodePools, EKS's control plane detects the drift from the expected state and automatically reverts the changes to maintain the default configuration. something what you have experienced.

The rapid revert cycle you experienced is expected reconciliation behavior - each time you or any automation attempted to modify the NodePool, EKS reverts it, incrementing the generation counter.

Recommended Approach:

Yes, using a custom NodePool with the built-in NodePools disabled is the correct and recommended approach when you need to control configuration like instance-category selection. When creating your cluster with Auto Mode, you can disable the default NodePool creation by setting nodePools: [] in your EKS Cluster configuration via Console, CLI or API, then create your own custom NodePools with your desired specifications.

This approach gives you control over:

• Instance types and categories
• Disruption budgets and maintenance windows
• Node lifecycle policies
• Other NodePool-specific configurations

Regarding Version Differences:

Reconciliation interval differences between EKS versions 1.33 and 1.34 is not documented, but the fundamental principle remains: built-in NodePools are managed resources that should not be modified.

To avoid production issues like the one you experienced, always use custom NodePools for workload-specific requirements rather than attempting to modify the built-in ones.

Sources
EKS Auto Mode - Eksctl User Guide