Context

• Source RFE: https://issues.redhat.com/browse/RFE-8313
• Summary: OCP support for GCP placement policies (spread and compact) to control relative placement of control-plane and worker nodes within a zone.

Problem / Ask

• Add native OpenShift support to define and manage GCP placement policies for control plane and worker nodes to achieve (a) high availability via spread across failure domains and (b) low latency via compact placement.

Goals

• Provide installer-time configuration for GCP placement policies for control plane and (optionally) default worker pools.
• Provide day-2 configuration.
• Ensure conformance with GCP’s placement policies and failure domain model.
• Deliver documentation and guidance for using placement policies.

Non-Goals

• Implementing new GCP placement primitives beyond what Google Cloud provides.
• Automatic cross-region or cross-zone failover orchestration beyond standard OCP/GCP capabilities.

Personas

• Cluster administrators/SREs operating OCP on GCP.
• Application teams needing deterministic HA or low-latency node placement.

Scope & Requirements

• Installer
• Cluster API
• Day-2 updates
• Validation: Validate requested policy availability in the chosen zone(s); provide clear errors and guidance if constraints are unmet.
• Drift detection: Detect and report if actual GCP instance placement drifts from requested policy
• Test coverage

Dependencies

• Cluster API provider for GCP (capg) support for placement policies.
• OpenShift installer (GCP platform) changes.
• GCP APIs/quotas/constraints for placement policies per zone.

Risks & Mitigations

• Zone capacity/quotas might block spread/compact requests -> add preflight validation and actionable errors.
• Conflicts between user-specified zones and placement policies -> document precedence and validation.
• Latency/HA assumptions dependent on GCP guarantees -> clearly document expected behavior and caveats.

Acceptance Criteria

• Users can set placement policy for control plane in install-config and see it applied.
• Users can set placement policy per MachineSet/NodePool for workers.
• E2E tests 
• Documentation published with examples and operational guidance.

Notes from RFE

• Spread placement policies: distribute instances across availability domains within a zone to maximize resilience.
• Compact placement policies: place instances close together (e.g., same rack) to minimize network latency.
• Business need: HA compliance, low-latency workloads, align OCP with core GCP features, reduce operational risk and drift.

Links

• RFE: https://issues.redhat.com/browse/RFE-8313
• Components referenced in RFE: Installer (GCP), Machine API/Cluster API provider for GCP, MachineSets/NodePool, Docs