Resources

Catalogued

MICROSOFT PURVIEW MIGRATION ARCHITECTURE: SANDBOX TO PRODUCTION

Technical Evaluation, Architecture Patterns, Constraints, and Recommended Approach

Guide · 5 min read · By Saurabh Nikumbha ·

ARCHITECTURAL CONTEXT

Microsoft Purview operates as a global tenant-level governance service with several critical architectural characteristics that fundamentally shape migration planning. The service maintains a single governance account per Azure Entra ID tenant, accessed through global endpoints at purview.azure.com.

The architecture relies heavily on Managed Identities that are tightly bound to specific subscriptions and resource groups. Network isolation is achieved through VNets, private endpoints, and DNS zones, whilst Key Vault provides secure credential storage for scanning operations.

CORE DEPENDENCIES

  • Global governance endpoints
  • Subscription-bound Managed Identities
  • VNet and private endpoint topology
  • Private DNS zone configuration
  • Key Vault credential integration
  • Data source connectivity patterns

CORE CONSTRAINT: TENANT-LEVEL SINGLETON

SINGLE INSTANCE LIMITATION

Microsoft Entra ID tenant allows exactly one Purview governance account. No parallel instances permitted.

This architectural constraint represents the most significant factor in migration planning. The singleton model means organisations cannot create a second Purview instance for parallel operations, testing, or gradual cutover scenarios. Resource Mover cannot relocate Purview across subscriptions, and the Purview resource ID remains static and non-portable throughout its lifecycle.

Managed Identities tied to the original subscription must be completely recreated during any migration, which cascades into numerous downstream dependencies. This constraint fundamentally eliminates traditional blue-green deployment patterns and requires careful orchestration of the migration sequence.

TECHNICAL IMPLICATIONS

  • No dual-instance testing environment
  • Complete MSI recreation required
  • All RBAC bindings must be reestablished
  • Data source re-registration mandatory

ARCHITECTURAL IMPLICATIONS

The singleton constraint creates a ripple effect across multiple architectural layers. Each component tightly coupled to the Purview instance requires careful consideration during migration planning.

ComponentImpactMigration Requirement
Global URIsFixed to tenantNo action required
Atlas Metadata BackendInstance-specificExport and reimport via API
Network EndpointsResource ID boundComplete recreation with new IDs
Managed IdentitiesSubscription lockedRegenerate and rebind all permissions
RBAC AssignmentsMSI dependentReapply all role assignments
Data Source ScansMSI authenticatedRe-register and reconfigure
SHIR IntegrationIdentity bindingRe-register with new credentials
Key Vault SecretsMSI access policiesRemap to new identity

OPTION 1: RESOURCE MOVER (INFRASTRUCTURE ONLY)

TECHNICAL SUMMARY

Azure Resource Mover supports migration of supporting infrastructure components but cannot relocate the Purview governance account itself. This approach allows preservation of network topology whilst requiring Purview reconstruction.

MOVEABLE RESOURCES

  • Virtual Networks and subnets
  • Private endpoints
  • Key Vault instances
  • Storage accounts (scan logs, classification rules)
  • Log Analytics and Application Insights
  • Private DNS zones

TECHNICAL ADVANTAGES

  • Preserves existing IP address ranges
  • Reduces DNS and private endpoint rebuild complexity
  • Minimal impact to SHIR networking topology
  • Maintains storage account continuity

TECHNICAL DISADVANTAGES

  • Purview instance rebuild still mandatory
  • Complete MSI recreation required
  • Key Vault secrets must be remapped
  • Risk of broken data lineage continuity
  • No reduction in cutover complexity

OPTION 2: FULL REBUILD IN PRODUCTION

DEPLOY NEW PURVIEW INSTANCE

Create fresh governance account in Production subscription with clean resource ID

RECREATE NETWORKING

Build VNets, private endpoints, and DNS private zones aligned to Production topology

ESTABLISH MANAGED IDENTITIES

Generate new MSI and configure RBAC assignments across all dependent resources

CONFIGURE KEY VAULT

Re-link scanning credentials and establish access policies for new identity

REGISTER DATA SOURCES

Reconnect all data assets with appropriate authentication configurations

REBUILD SCANNING CONFIGURATION

Recreate scan rules, schedules, and classification rulesets

RECONSTRUCT COLLECTIONS

Rebuild collection hierarchy and domain structures according to governance model

IMPORT METADATA

Utilise Atlas API to reimport glossary terms, classifications, and custom attributes

OPTION 2: TECHNICAL TRADE-OFFS

TECHNICAL ADVANTAGES

This approach delivers the cleanest architectural outcome, establishing Production governance with proper separation from Sandbox environments.

  • Clean RBAC model aligned with Production IAM policies
  • VNet addressing and private endpoint layout designed for long-term governance scalability
  • Enables enforcement of Zero Trust network policies from inception
  • Full auditability with clean lineage tracking from Day 1
  • No legacy configuration baggage
  • Optimal for greenfield Production standards

TECHNICAL DISADVANTAGES

The comprehensive rebuild introduces significant implementation overhead and coordination requirements.

  • Complete reconstruction of scanning and classification rules
  • SHIR must be re-registered with new credential set
  • Custom migration scripts required for glossary and metadata
  • High dependency on client networking and IAM approval cycles
  • Extended timeline for data source re-registration
  • Potential metadata loss for non-exportable elements

OPTION 3: LEAVE PURVIEW IN SANDBOX

ARCHITECTURAL WEAKNESS

Maintaining Purview in Sandbox fundamentally contradicts enterprise governance principles and introduces significant technical debt.

TECHNICAL CONSTRAINTS

  • Sandbox VNets lack production-grade hardening
  • IAM roles not aligned to Production PIM policies
  • Insufficient private link configuration for enterprise requirements
  • Limited cross-subscription governance scalability
  • Non-compliant with separation of concerns principles

TECHNICAL RISKS

  • Misaligned DNS zones causing resolution failures
  • Legacy firewall rules introducing security gaps
  • No separation between test and production data lineage
  • Future architectural changes blocked by single-instance constraint
  • Audit trail confusion across environments

Architectural Recommendation: This option fails to meet enterprise governance standards and should be excluded from serious consideration. The technical debt accumulated will compound over time, creating increasingly severe operational constraints.

OPTION 4: STAGED CUTOVER (RECOMMENDED)

SANDBOX REMAINS ACTIVE

Existing Purview instance continues operations during preparation phase

BUILD PRODUCTION INFRASTRUCTURE

Deploy VNet, DNS zones, private endpoints, Key Vault, Log Analytics, and storage accounts

EXPORT ATLAS METADATA

Extract glossary, domains, custom attributes, and classifications via API

DELETE SANDBOX PURVIEW

Remove existing instance to satisfy tenant singleton constraint

DEPLOY PRODUCTION PURVIEW

Create new governance account in Production within minutes of deletion

APPLY RBAC CONFIGURATION

Establish role assignments for new Managed Identity

IMPORT METADATA

Rehydrate Atlas backend with exported governance content

REGISTER DATA SOURCES

Reconnect all data assets with Production authentication

VALIDATE FUNCTIONALITY

Test scans, lineage tracking, and governance workflows

Critical Note: Only one Purview instance exists at any point. "Hybrid" refers exclusively to preparation activities, not dual operation.

RECOMMENDATION AND EXECUTION PATH

RECOMMENDED APPROACH

Option 4: Staged Cutover — Balances risk mitigation with architectural quality whilst respecting tenant constraints

KEY ADVANTAGES

  • Respects singleton tenant constraint
  • Minimises production downtime
  • Enables Infrastructure-as-Code approach
  • Provides structured metadata migration
  • Low-risk networking and IAM implementation
  • Allows thorough testing of export scripts

EXECUTION PREREQUISITES

  • Production Landing Zone readiness confirmed
  • Network and DNS architecture approved
  • IAM and RBAC policies documented
  • Atlas export/import scripts tested
  • SHIR host infrastructure prepared
  • Cutover window scheduled (2–3 hours)

DISCOVERY

1–2 weeks

INFRASTRUCTURE BUILD

2–3 weeks

NETWORK CONFIG

1 week

IAM APPROVALS

1–2 weeks

CUTOVER WINDOW

2–3 hours

DATA SOURCE REGISTRATION

3–6 weeks

VALIDATION

1–2 weeks

Not sure where to start?

Book a 30-minute call and tell us what worries you about Microsoft 365: oversharing, an audit, or a Copilot rollout. Or email us and we'll reply within one working day.