ADR-013 Phasing and Ordering of OpenShift Implementation

Context and Problem Statement

The implementation needs a clear dependency order for MVP delivery. Storage, Advanced Cluster Management (ACM)/Red Hat Advanced Cluster Management, Automation, and hosted control planes all have prerequisites that affect sequencing across datacenters.

Decision Drivers

• Storage readiness is a prerequisite for stateful workloads and virtualization.
• Advanced Cluster Management (ACM) governance should align with the single-hub operating model using one Red Hat Advanced Cluster Management installation per hub cluster.
• Automation must manage day-2 operations across clusters.
• Clusters as a Service clusters replace the legacy Platform Services cluster concept.
• Cluster Management are preferred for managed clusters where feasible.

Considered Options

1. Build one datacenter at a time (Location A baseline, then repeat in Location B).
2. Build Advanced Cluster Management (ACM) first and deploy datacenters in parallel.
3. Prioritize shared clusters and defer Cluster-as-a-Service automation until after MVP.

Decision Outcome

Chosen option: One datacenter at a time with explicit readiness-gate enforcement and controlled cutover to Location B.

Implementation order:

1. Phase A - Foundation: deliver lane-scoped networking, trust/secrets, and security prerequisites.
2. Phase B - Shared services: deploy Clusters as a Service internal and external profiles in Location B, establish ACM governance plane, and enable GitOps control paths while keeping Location A scoped to the transitional OpenShift AI lane without Datacenter Services activation.
3. Phase C - Runtime services: onboard CaaS/VMaaS/DBaaS and tenant capabilities with policy baselines.
4. Phase D - Cutover: execute cutover only after readiness gates pass and rollback controls are validated.

Execution controls are delivered through wp-01-dc3-0-network-and-supporting-services, wp-08-platform-services-automation, and wp-16-business-continuity-and-disaster-recovery, with cutover release evidence approved under adr-004-single-active-phased-rollout.

Pros and Cons of the Options

Option 1 – One datacenter at a time (chosen)

Pros

• Aligns with single-active rollout and reduces operational risk.
• Enables iterative automation refinement before scaling.

Cons

• Slower overall delivery if parallel capacity is available.

Option 2 – Advanced Cluster Management first, parallel datacenters

Pros

• Faster multi-site readiness if hardware and teams are ready.

Cons

• Higher coordination complexity and larger initial scope.

Option 3 – Defer containers as a service automation

Pros

• Focuses on core platform readiness and shared clusters.

Cons

• Delays hosted control plane automation benefits.