A company uses AWS DMS to continuously capture changes from an on-premises PostgreSQL database and replicate them to Amazon Redshift. The DMS ongoing replication task reads PostgreSQL transaction logs, but the data engineer observes high latency in CDC and suspects the PostgreSQL source is the bottleneck. Which approach will verify that the PostgreSQL database is the source of the latency?

The CDCLatencySource metric in Amazon CloudWatch directly measures the latency between the source database and the DMS replication instance. This metric specifically indicates how far behind the DMS task is from the source database's transaction log, making it the most accurate way to identify if the PostgreSQL source is causing the bottleneck. Using CDCIncomingChanges would show the rate of changes received, but not the latency itself. Verifying postgresql.conf configuration is a troubleshooting step, not a direct measurement of current latency. Checking CloudWatch Logs for error messages might reveal issues, but it doesn't quantify the latency from the source.