A company runs a workload in the AWS Cloud.

The company needs to upgrade DocumentDB to the latest version with no data loss while allowing continuous reads and writes. The company also must be able to test and verify the upgrade before switching production traffic. This is a classic requirement for performing an upgrade using a blue/green approach: build a new target environment on the new version, keep it in sync with the source, validate it, and then cut over by changing the endpoint (here, Route 53 DNS).

Option A implements this pattern using a new DocumentDB cluster running the latest version and AWS DMS to continuously migrate and replicate changes from the old cluster to the new cluster. Because the workload is continuously changing, a one-time export/import is insufficient; continuous replication is needed to keep the target cluster current during the test period. AWS DMS supports a “migrate and replicate” style of task that performs a full load and then applies ongoing changes (CDC) so the target stays synchronized. The question also states that change streams are enabled with a 24-hour retention period, which supports capturing and applying changes during migration/validation and helps ensure the replication stream can be maintained while testing.

Option A also addresses indexes by using the DocumentDB Index Tool to export and import indexes, which is important because indexes can affect query performance and behavior. After the company validates the new cluster, the cutover is done by updating the Route 53 record to point to the new cluster endpoint, switching all backend services without changing application configuration beyond DNS resolution.

Option B uses MongoDB CLI tools to export/import. This is not suitable for continuous write workloads because export/import is a point-in-time operation and would require downtime or risk data divergence during the test period. It also adds more operational overhead and does not provide continuous replication for the duration of validation.

Option C performs an in-place major version upgrade. That does not satisfy the requirement to test and verify the upgrade before backend services use the upgraded version because the upgrade happens directly on the production cluster. Even though a snapshot exists for rollback, production is still exposed to the upgrade immediately, which violates the requirement for pre-cutover verification.

Option D is incorrect because AWS DataSync transfers files between storage systems such as NFS/SMB and AWS storage services. It is not a database migration or replication service and cannot copy a DocumentDB database in a way that preserves database semantics and supports continuous replication.

Therefore, creating a new DocumentDB cluster, keeping it synchronized using AWS DMS (supported by change stream retention), validating it, and then cutting over via Route 53 DNS update (option A) meets all requirements.