Question 850
A database team reports a successful correction for this problem: Database replicas are created successfully but the team never tests whether a failed primary can be recovered into service. What evidence supports closing the incident?
Explanation
Correct Answer
The designated replacement accepts required writes and application reconnection meets RTO. This is the intended effect of: Perform a controlled role transition and full application transaction test. The original defect was deployment verification stops at replica existence rather than failover and client recovery.
Incorrect Answers
Even if this observation were recorded, it would support a case where an API needs to scale a high volume of eventually consistent catalog reads but all requests target the single Cloud SQL writer. It would not verify that the designated replacement accepts required writes and application reconnection meets RTO.
That result would matter when an AlloyDB cluster's recovery runbook has never been tested after switching to new application connection pools. It does not establish that the designated replacement accepts required writes and application reconnection meets RTO, which is the evidence needed here.
Even if this observation were recorded, it would support a case where a Spanner multi-region deployment has low read latency in Europe but US-based writes suffer after a planned shift in user population. It would not verify that the designated replacement accepts required writes and application reconnection meets RTO.