All stacks · Streaming & CDC
Change data capture from Postgres, MySQL, and SQL Server into Kafka.
DebeziumA production walkthrough of Postgres CDC with Debezium: how the replication slot fills your disk, how to snapshot a large table without locking it, and how to apply changes so replays are safe.
PostgreSQLA BI scan on the orders primary froze autovacuum, the replication slot kept WAL, and the disk hit 91%. Read the plan, manage bloat, and keep CDC from filling the volume.
Debezium is how most teams do CDC: log-based change events from Postgres, MySQL, and SQL Server into Kafka (or a compatible bus). It is not a sync tool and it is not a warehouse loader — it is a change log you must land idempotently.
The production work is snapshot vs streaming, heartbeats, TOAST columns, and the MERGE that makes “exactly-once” a key instead of a prayer. We already have a Postgres CDC guide in the catalog.
For databases you own and a Kafka platform you already run, Debezium is often cheaper and more controllable. Fivetran wins SaaS sources and teams that do not want to operate Connect. Many platforms run both.
A slot that was not consumed, a connector on a starved worker, or a snapshot that never finished. Monitor replication lag and slot size like a production SLO, not a dashboard curiosity.
Yes — vendor paths and Snowpipe Streaming exist. Kafka + Debezium is the open, inspectable path. Pick based on who pages when lag hits 45 minutes, not the architecture diagram.
One email a week with our latest tutorials. No spam.
Newsletter signup is not live yet. Use the contact form if you want to be notified.