Postgres high availability faces operational challenges when using Change Data Capture (CDC) due to logical replication slots being primary-local objects that can prevent failover. When CDC clients lag or poll infrequently, standby replicas may become ineligible for promotion, forcing administrators to choose between waiting for CDC advancement or breaking the CDC stream. MySQL's GTID-based approach avoids this coupling by embedding replication progress in the binary log itself, allowing seamless failover regardless of CDC client behavior.
Sort: