Context

• Source RFE: https://issues.redhat.com/browse/RFE-8314
• Summary: OCP support for GCP N4A machine types (Axion/ARM) and custom machine types on GCP.

Problem / Ask

• Provide OpenShift support for the GCP N4A (Axion/ARM) family and related custom machine types to enable cluster provisioning and operations on these shapes.

Goals

• Qualify and support `n4a-*` machine types for control plane and worker nodes where applicable.
• Expose N4A/custom machine types via installer configuration (platform.gcp) and Machine API (MachineSets/NodePools providerSpec) for day-1/day-2.
• Ensure payload and images support ARM64/aarch64 where required (e.g., RHCOS aarch64 on GCP; multi-arch payload readiness).
• Validate Cluster Autoscaler compatibility with N4A/custom shapes.
• Publish documentation for install/operations on N4A and custom shapes.

Non-Goals

• Introducing new instance classes beyond GCP’s offerings.
• Guaranteeing cross-architecture workload portability beyond documented multi-arch support.

Personas

• Platform/SRE teams operating OCP on GCP.
• Application teams requiring cost/performance from ARM-based instances and custom VM shapes.

Scope & Requirements

• Installer: Add install-config options to select N4A/custom shapes; validation for region/zone availability.
• Release/payload: Build and publish required aarch64 images/artifacts for GCP if applicable.
• Autoscaler
• CI/Testing: Add coverage for N4A provisioning and operations; upgrade/replace scenarios.
• Documentation: Examples and guidance for common topologies and constraints.

Dependencies

• Cluster API provider for GCP (capg) support for N4A/custom machine types.
• OpenShift installer (GCP platform) exposure and validation.
• Release engineering for aarch64 payloads/images on GCP.
• Cluster Autoscaler integration updates as needed.

Risks & Mitigations

• Quotas/availability per region/zone -> preflight checks and actionable error messages.

Acceptance Criteria

• Users can select N4A/custom shapes in install-config and via MachineSets/NodePools.
• Clusters successfully provision and operate on N4A-based nodes; autoscaling and MHC function correctly.
• Required aarch64 images/payloads available and validated where applicable.
• Docs published with configuration examples and limitations.

Notes from RFE

• Business driver: very-large customers blocked until N4A supported.
• Affected components: capg provider, installer, RHCOS aarch64 on GCP, autoscaler, docs.

Links

• RFE: https://issues.redhat.com/browse/RFE-8314