RDS for MySQL must automatically fail over to another AZ after a primary-AZ failure is detected. Brief connection interruption is acceptable, but manually changing the endpoint name is not.
The application uses the RDS DB endpoint and implements reconnection and appropriate DNS refresh. Read-load distribution is not the primary requirement.
Which configuration is most appropriate?
Use a single-standby Multi-AZ DB instance for automatic failover and reconnect through the same DB endpoint.
Detailed explanation
RDS changes the endpoint's DNS record during failover; existing connections must be re-established.
RDS changes the endpoint's DNS record during failover; existing connections must be re-established.
Manual promotion can be used operationally but does not meet this automatic-failover requirement by itself.
Manual promotion can be used operationally but does not meet this automatic-failover requirement by itself.
Restoration is possible, but it is not automatic standby failover without manual endpoint changes.
Restoration is possible, but it is not automatic standby failover without manual endpoint changes.
The database may fail over while an IP-pinned application fails to follow the endpoint's DNS update.
The database may fail over while an IP-pinned application fails to follow the endpoint's DNS update.
Check official documentation
Compare the following points with the official documentation.
Review single-standby Multi-AZ behavior, failover/DNS updates, and reconnection requirements.
Official references
What to confirm
Explain why a stable endpoint name does not guarantee uninterrupted connections and why this standby is not a read endpoint.
Key points
- RDS Multi-AZ DB instance
- Standby
- RDS endpoint
Notes
- Environment: AWS SOA-C03 exam guide and the linked AWS documentation
- Reviewed against the linked AWS documentation on 2026-10-04.
Foundation review
Identify the single-standby deployment
This DB instance deployment uses a synchronous standby in another AZ for high availability. The standby cannot serve reads. Distinguish this from a Multi-AZ DB cluster with readable replicas and from an asynchronous read replica.
A stable name is not an uninterrupted connection
RDS updates endpoint DNS during failover, but existing connections must be re-established. Typical 60–120-second failover is not a fixed guarantee; workload and recovery conditions can extend it. Test DNS caching, retries, and transaction replay, avoiding repeated side effects while verifying application recovery.