AZ-305 Practice Question 27: Messaging Regional Failover
Microsoft Azure Solutions Architect Expert (AZ-305) · Test 3 · question 7 of 10
A payments processor runs its settlement pipeline on an Azure Service Bus Premium namespace in North Europe. After a regional incident at a competitor, the risk team writes these requirements into the design:
- If North Europe is lost, no message that the broker acknowledged may be lost; the RPO must be zero. - Messages already queued, and their state changes, must exist in the second region, not only the queue and topic definitions. - Client applications must keep using the same namespace hostname, with no configuration change at failover. - Only the processor's own incident commander may trigger a regional switch; Azure must never do it automatically.
What should you include in the design?
Show answer and explanation
Answer:
Geo-Replication continuously replicates both metadata and data (queues, topics, subscriptions, filters, the messages inside those entities, and every state and property change) from a primary region to a secondary. In synchronous replication mode the system commits a request in both regions before acknowledging it to the client, which Microsoft's own comparison table lists as RPO 0, no data loss on promotion. The namespace keeps a single hostname that always points at the current primary, so client applications need no change, and all promotions are customer initiated: Azure never triggers one.
Option A is the closest wrong answer and fails the second requirement explicitly. Geo-Disaster Recovery replicates metadata only, the entities and their configuration, so after a failover the secondary namespace has the right queues but none of the messages that were sitting in them. Microsoft now names Geo-Replication as the recommended choice for most disaster recovery scenarios for exactly this reason.
Option B pushes the platform's job into the application and still misses the target. Dual publishing gives no atomicity across two namespaces, so a failure between the two sends leaves the regions divergent; consumers must build their own de-duplication; and completion state in one namespace is invisible to the other, so failing over replays work that was already done.
Option D protects against the wrong failure. Availability zones are separate facilities inside North Europe and give excellent resilience to a zonal outage, but the requirement is the loss of the region, which zone redundancy does not address.
Reference: Azure Service Bus Geo-Replication