Epic Goal

Enable NodePool CRs to support setting minimum replicas to 0, allowing node pools to scale down completely when using autoscaling (scale-from-zero functionality).

Why is this important?

To support autoscaling from/to zero for Hypershift NodePools (OCPSTRAT-1525), the NodePool API must allow min-replicas=0 to be set alongside autoscaling configuration. This enables:

• Cost optimization by scaling down to zero nodes during off-hours or low-demand periods
• On-demand scaling that automatically provisions nodes when workloads require them
• Efficient resource management for development and testing environments

Implementation Details

The implementation includes the following changes (per PR #6975):

API Validation Changes

• Change NodePoolAutoScaling.Min validation from minimum=1 to minimum=0
• Update CRD manifests across all feature gates to allow min: 0
• Add min to the required fields list in CRD specifications
• Update API documentation to reflect 0 <= Min <= Max constraint

Controller Logic Updates

• Modify replica defaulting logic in hypershift-operator/controllers/nodepool/capi.go
• Change boundary checks from < Min to <= Min to properly handle Min=0
• Apply changes to both MachineDeployment and MachineSet replica management

RBAC Updates

• Add watch verb to cluster-autoscaler RBAC for infrastructure.cluster.x-k8s.io API group
• Allows informer to watch rather than poll for better performance

Scenarios

• Users configure a NodePool with autoscaling enabled and min: 0, max: N
• When no workloads are present, the node pool scales down to 0 nodes
• When workloads are scheduled, cluster-autoscaler scales up the node pool
• After workload removal, the node pool scales back down to 0 nodes
• API validation accepts this configuration without errors

Acceptance Criteria

• NodePool API accepts min: 0 configuration with autoscaling enabled
• API validation allows the combination of autoScaling.min: 0 and autoScaling.max: N
• NodePool controller properly handles zero-replica state
• Replica defaulting logic correctly processes min: 0
• RBAC permissions updated for cluster-autoscaler
• CI tests pass with the new functionality

Dependencies (internal and external)

• PR #6975: https://github.com/openshift/hypershift/pull/6975 (implementation PR)
• CNTRLPLANE-1479: Cluster API AWS provider scale from zero support (required for AWS)
• OCPSTRAT-1525: Parent feature for autoscaling from/to zero on HyperShift
• Cluster API: Opt-in Autoscaling from Zero proposal support
• Kubernetes Cluster Autoscaler: Must support scale-from-zero with provider metadata

Done Checklist

• CI - CI is running, tests are automated and merged
• Release Enablement - Release enablement details provided
• DEV - Upstream code and tests merged (PR #6975)
• DEV - Upstream documentation merged
• DEV - Downstream build attached to advisory
• QE - Test plans in Polarion
• QE - Automated tests merged (E2E test included in PR #6975)
• DOC - Downstream documentation merged