Question 320
The previous outage involved this problem: A Cloud SQL account used by one microservice can read a customer payments table unrelated to its workload. What long-term control addresses the risk?
Explanation
Correct Answer
Review effective database permissions against application data-flow maps. This is preventive rather than a one-time repair; it addresses the underlying failure: A shared database principal or broad grant defeats service-level least privilege.
Incorrect Answers
This control would test Google Cloud IAM and database SQL privileges as distinct authorization layers; it responds to the risk in which a new Cloud SQL administrator cannot inspect an instance in the console even though their PostgreSQL database user has superuser permissions. It would not address the current root cause: a shared database principal or broad grant defeats service-level least privilege.
The failure in the question arises because a shared database principal or broad grant defeats service-level least privilege. A safeguard to examine skew and hot key ranges in multitenant Spanner workloads serves situations where Spanner latency rises for one tenant while aggregate database CPU appears healthy rather than this one.
This would recheck identity permissions during database project or organization changes. It is useful when a Spanner service account loses access after its database is moved to a different project, but it does not prevent the issue here: a shared database principal or broad grant defeats service-level least privilege.