ARA-C01 Question 250
Single answerDisaster recoveryA global retail company runs its production Snowflake account in AWS us-east-1 and must meet a disaster recovery objective of less than 15 minutes of data loss and less than 60 minutes to resume analytics in another region if the primary region becomes unavailable. The company has several large databases that are updated continuously by ELT pipelines and shared with downstream BI workloads. The architect wants to minimize operational overhead and avoid custom data export/import processes. Which approach should the architect recommend?
- A
Create database replication and failover groups to replicate the required databases and account objects to a secondary account in another supported region, and use account/database failover when a regional outage occurs.
- B
Configure Time Travel and Fail-safe on the primary account, and if us-east-1 fails, recover the databases into a new account in another region using historical data.
- C
Use Snowpipe to continuously unload changed data from the primary region to cloud storage in another region, then recreate databases manually in the DR account when needed.
- D
Create reader accounts in a secondary region and direct BI users there during an outage, because reader accounts automatically provide cross-region disaster recovery for shared databases.
Show answer and explanation
Correct answer: A
Explanation
For Snowflake disaster recovery, the recommended native approach is replication and failover. Database replication protects database objects and data, while failover groups extend replication to a broader set of supported account objects needed for application and analytics continuity. This is the most appropriate design when the business requires defined RPO/RTO targets and wants to avoid building custom data movement and reconstruction pipelines. Time Travel and Fail-safe are data protection features, not substitutes for cross-region DR. Reader accounts support data sharing use cases, not full production failover. In Snowflake documentation, architects should review replication, failover groups, supported objects, and scheduling/frequency considerations to ensure the configured replication cadence aligns with the required RPO and that cutover/failover procedures meet the desired RTO.
- A. Correct.
Correct. Snowflake's supported disaster recovery pattern is to use replication for databases and, where needed, failover groups to include account-level objects such as users, roles, warehouses, resource monitors, and other supported securable/account objects. This approach is designed for business continuity and disaster recovery across regions and/or clouds, with low operational overhead compared to building custom copy processes. For a target of less than 15 minutes of data loss, architects should configure an appropriate replication schedule/frequency that aligns with the RPO, and fail over to the secondary account when the primary region is unavailable. This best matches the requirement to minimize operational complexity while enabling relatively fast recovery.
- B. Incorrect.
Incorrect. Time Travel and Fail-safe protect against logical errors such as accidental data changes or object drops within the same account's data protection model; they are not a regional disaster recovery solution. Fail-safe is not customer-driven for immediate restore into another region and does not satisfy a controlled cross-region DR plan or a sub-hour recovery objective for analytics resumption.
- C. Incorrect.
Incorrect. Although exporting data to cloud storage and rebuilding in another region is technically possible, it introduces significant operational overhead, requires custom orchestration, and typically increases both RPO and RTO. It also does not preserve Snowflake metadata and account-level configuration as effectively as native replication/failover. This option conflicts with the requirement to avoid custom export/import processes.
- D. Incorrect.
Incorrect. Reader accounts are intended to let consumers access shared data without having their own Snowflake account, but they are not the mechanism for cross-region disaster recovery of a provider's production environment. They do not automatically replicate all required production databases and account objects for failover, nor do they provide a complete DR posture for the company's own analytics workloads.