AZ-305 Practice Question 37: NoSQL Point in Time Recovery
Microsoft Azure Solutions Architect Expert (AZ-305) · Test 4 · question 7 of 10
A gaming company keeps player profiles in an Azure Cosmos DB for NoSQL account with a write region in West Europe and a read region in East US. After a faulty deployment silently overwrote thousands of profiles and went unnoticed for four days, the company sets these requirements:
- recovery to any chosen second within the previous 30 days, - a restore the team can perform itself from the portal or CLI, without opening a support request, - the recovered data must land in a separate account so it can be inspected before anything is promoted, - taking the backups must not consume the container's provisioned request units.
What should you include in the design?
Show answer and explanation
Answer:
Continuous backup with point-in-time restore is built for recovering from an accidental write or delete within a container. The restore point can be any timestamp inside the retention window, which is 7, 30, or 35 days depending on the tier chosen, so the 30-day tier covers a four-day-old mistake comfortably. The restore is self-service from the Azure portal, CLI, PowerShell, or an ARM template, and it restores into a new account, which is precisely the "inspect before promoting" behavior the company wants. Microsoft also notes that the backup runs in the background without consuming extra provisioned throughput or affecting performance.
Option B leaves a recovery window two orders of magnitude too small. With a four-hour interval and two retained copies, the oldest recoverable state is roughly eight hours old, so a corruption noticed on day four is unrecoverable no matter who is called.
Option C is a bespoke rebuild of the platform feature with worse guarantees. The change feed does not surface deletes by default, the replay logic is the company's to write and test, and the Function, its storage account, and its throughput all become new components to operate and pay for.
Option D confuses replication with backup. Every region holds the *current* state; the faulty overwrite replicated to East US within seconds of being written. Multi-region writes improve availability and write latency, and provide no historical point to return to.
Reference: Continuous backup with point-in-time restore in Azure Cosmos DB