Question 840
After an incident where a company claims two read replicas provide automatic HA for its Cloud SQL primary without enabling the approved failover configuration, which recurring control would best prevent the same defect?
Explanation
Correct Answer
Review HA and read-serving components as independent architecture layers. The original failure arose from: The deployment confuses read scaling with primary availability architecture. The control reduces that failure mode.
Incorrect Answers
For the situation in which Bigtable traffic is distributed among replicated clusters but a single unhealthy cluster continues receiving requests from some clients, it makes sense to exercise routing and latency impact after Bigtable cluster changes. Here, the deployment confuses read scaling with primary availability architecture, so the safeguard misses the recurring fault.
The stated control responds to a failure mode where an API needs to scale a high volume of eventually consistent catalog reads but all requests target the single Cloud SQL writer. It does not prevent this one: the deployment confuses read scaling with primary availability architecture.
For the situation in which a Terraform rollout attempts to recreate a production Cloud SQL instance because a mutable resource label was used as its identity, it makes sense to enforce database deletion protection and plan review in IaC pipelines. Here, the deployment confuses read scaling with primary availability architecture, so the safeguard misses the recurring fault.