RWS Architecture article

ADR-012 SSO (Single Sign-On) for OpenShift and Hosted Services

A federated authentication solution is required across clusters and hosted services to simplify smart access operations. The platform baseline already assumes keycloak; this ADR de

  1. Typeadr
  2. Statusdraft
  3. Domainplatform
On this page
  1. Context and Problem Statement
  2. Decision Drivers
  3. Considered Options
  4. Decision Outcome
  5. Next Decision Trigger
  6. Pros and Cons of the Options
  7. Option 1 – Stand-alone clusters per datacenter
  8. Option 2 – Replicated clusters with single-active failover

Context and Problem Statement

A federated authentication solution is required across clusters and hosted services to simplify smart access operations. The platform baseline already assumes keycloak; this ADR decides the deployment architecture and resilience model.

Scope boundary: this ADR decides SSO service topology and failover behavior; broader IAM role/realm and integration architecture is handled in ADR-015.

Decision Drivers

  • Resilience to single-site disruptions.
  • Support for approximately 50,000 users (internal and external).
  • automation-friendly deployment and operations.
  • Compliance with MFA, audit logging, and cryptography requirements.

Considered Options

  1. Stand-alone Keycloak clusters per datacenter without cross-site replication.
  2. Replicated Keycloak clusters with single-active failover and database replication.

Decision Outcome

Decision remains draft. keycloak remains the baseline smart access component (see ADR-005), while deployment topology is under evaluation.

Next Decision Trigger

  • Decision owner: Platform Architecture with Security Architecture review.
  • Promote to accepted when failover test evidence, database replication posture, and operational runbooks are approved.
  • If replicated topology cannot meet service objectives, publish fallback constraints and revised option ranking before 2026-03-15.

Pros and Cons of the Options

Option 1 – Stand-alone clusters per datacenter

Pros

  • Simpler deployment and operational separation between sites.

Cons

  • No cross-site state replication; failover requires client reconfiguration.

Option 2 – Replicated clusters with single-active failover

Pros

  • Controlled failover with shared state across sites.
  • Aligns with resilience requirements for identity and access management.

Cons

  • More complex replication and load-balancing configuration.
  • Higher operational overhead for database and cache replication.