Question 680
The previous outage involved this problem: A migration validation tool reports more orders at the destination because retried change events create duplicate inserts. What long-term control addresses the risk?
Explanation
Correct Answer
Inject retry and duplicate-event tests in migration pipelines. This is preventive rather than a one-time repair; it addresses the underlying failure: The apply process is not idempotent under at-least-once event delivery.
Incorrect Answers
The failure in the question arises because the apply process is not idempotent under at-least-once event delivery. A safeguard to record source and target checkpoint positions with validation reports serves situations where a parallel-run comparison flags mismatches because the same transaction is captured at different times in source and target snapshots rather than this one.
For the situation in which a MySQL source fills its disk because its binary logs accumulate during a slow long-running migration, it makes sense to monitor binary log storage and purge dependencies throughout continuous replication. Here, the apply process is not idempotent under at-least-once event delivery, so the safeguard misses the recurring fault.
This would set rollback feasibility limits based on verified write direction and divergence. It is useful when a team tries to roll back two days after cutover by pointing applications at the old source without copying intervening writes, but it does not prevent the issue here: the apply process is not idempotent under at-least-once event delivery.