A production-grade Change Data Capture (CDC) system that streams PostgreSQL WAL (Write-Ahead Log) changes from Happy Robot's AWS RDS instance to Snowflake. Built with Python, Snowflake Stored Procedures, DBT, and Dagster, this solution replicates 215 tables across 6 schemas (~8 GB) using the same proven architecture as our MySQL Binlog CDC pipeline.
Key Metrics:
- Tables: 215 across 6 schemas (cash_reconciliation, fin_approval_log, member, origination, program_access, transactions)
- Initial Load: 193 tables, 28.7M rows, ~25 minutes
- Latency: Hourly CDC cycles (configurable)