Question 120
After an incident where a Bigtable team claims replication ensures data protection but never measures how late changes reach the alternate cluster, which recurring control would best prevent the same defect?
Explanation
Correct Answer
Monitor cross-cluster replication freshness against recovery objectives. It targets the actual operating weakness: Replication delay is excluded from the recovery-point and consistency objectives. The immediate repair was: Measure replica lag and define a recovery decision based on observed data freshness.
Incorrect Answers
The stated control responds to a failure mode where an AlloyDB failover test reports success because the console shows a new primary, while users cannot complete transactions. It does not prevent this one: replication delay is excluded from the recovery-point and consistency objectives.
The failure in the question arises because replication delay is excluded from the recovery-point and consistency objectives. A safeguard to include database maintenance events in the application availability calendar serves situations where a trading application loses service during an unannounced maintenance period despite an available supported maintenance scheduling option rather than this one.
The failure in the question arises because replication delay is excluded from the recovery-point and consistency objectives. A safeguard to use service-account identity tests in release acceptance instead of relying on personal access serves situations where a Spanner application returns permission denied only when run under its deployment service account rather than this one.