AWS DMS uses exponential backoff for recoverable errors, and with default settings a CDC task can silently retry for up to 30 minutes before failing. This post explains how four key error-handling settings interact — RecoverableErrorCount, RecoverableErrorInterval, RecoverableErrorThrottling, and RecoverableErrorThrottlingMax — and how to tune them so CDC tasks surface failures within minutes. Three retry scenarios are compared, with the recommended configuration (Scenario C) producing task failure after ~225 seconds using bounded exponential backoff. The post also covers setting all apply error policies to STOP_TASK for migration cutover windows to prevent silent data drift, and pairs these settings with Amazon EventBridge rules and CloudWatch alarms on CDCLatencyTarget for complementary alerting. Step-by-step CLI instructions and validation guidance are included.
