← Azure projects

Azure · DR Design

Azure SQL Database Geo-Resilience

Design Azure SQL regional resilience around the required failover scope, stable application endpoints, data-loss tolerance, and secondary-region dependencies.

PlatformAzure
DomainResilience
LevelAdvanced
Last reviewed2026-09-19
Use whenProtect a managed SQL data tier from regional failure while preserving application connectivity, authentication, and network access after failover.
Key decisionChoose replication scope deliberately
Primary servicesAzure SQL Database · Active Geo-Replication · Failover Groups
Cost focusDatabase service tier and vCores/DTUs
On this page

Building blocks

Azure SQL DatabaseActive Geo-ReplicationFailover GroupsPrivate LinkAzure Monitor

Design goal

Keep the database recoverable across a regional event without assuming database replication alone will make the application usable after failover.

Architecture

Azure SQL provides service-managed local high availability. Regional recovery can use active geo-replication for individual databases or failover groups for grouped databases and stable listener endpoints.

Primary AppPrimary SQLGeo SecondaryDR App
Primary App
Primary SQL
Geo Secondary
DR App

Architecture decisions

Decision

Choose replication scope deliberately

Why

Use active geo-replication for database-level control and failover groups when grouped failover and stable endpoints are required.

Trade-off

Database-level replication gives precise control, while grouped failover simplifies application endpoints; the choice changes failover coordination, cost, and operational flexibility.

Decision

Design the whole dependency chain

Why

Secondary-region networking, identity, secrets, and application configuration must work after database failover.

Trade-off

Preparing the full secondary environment improves recovery confidence, but means paying for and maintaining more than the database replica alone.

Security

Protect the control and data paths deliberately. Use contained users or appropriately synchronized identity configuration, private connectivity where required, and equivalent network controls in both regions.

Availability

Design for the failure domain that must be survived. Use service-tier capabilities and zone redundancy where available to reduce local failure impact before invoking regional DR.

Disaster recovery

Treat regional recovery as a separate operating state. Define planned and forced failover procedures, expected data-loss behavior, failback, and application connection-string strategy.

Cost drivers

  • Database service tier and vCores/DTUs
  • geo-secondary compute
  • backup retention
  • data transfer
  • monitoring

Design assumptions

  • Application supports retry/failover behavior
  • Secondary region is approved
  • RPO/RTO are defined

Implementation plan

  1. Define the acceptable data loss, recovery time, failover scope, and application connection behavior before selecting a replication pattern.
  2. Choose the database/service tier and configure the secondary database or failover group in a region that supports the required features.
  3. Prepare Private Endpoints, DNS, firewall rules, identities, keys, and dependent services in both regions so failover does not stop at the database.
  4. Point applications at the stable listener/connection pattern appropriate to the chosen design and define read-only routing only if it is needed.
  5. Configure replication/failover monitoring, then execute planned failover and failback tests while measuring application reconnection and data-loss behavior.

Validate the design

  • Perform a planned failover and confirm the application reconnects through the intended stable endpoint.
  • Verify authentication, Private DNS, firewall/private-endpoint paths, and dependent services work in the secondary region.
  • Measure application recovery time and observed data-loss behavior against the agreed RTO/RPO.
  • Fail back during a controlled test and confirm monitoring reports replication health before returning to normal operations.

Architecture basis

Continue a learning path