Skippr Docs

Appearance

DynamoDB CDC to MotherDuck Final-State Guide ​

June 2026

DynamoDB to MotherDuck CDC stays manageable when the integration is explicit about item keys, stream ordering, and final-state merges.

Short Answer ​

DynamoDB CDC to MotherDuck works by turning DynamoDB Streams item events into warehouse row changes. The source emits INSERT, MODIFY, and REMOVE records, and MotherDuck applies them with DuckDB MERGE semantics plus order-token guards so the target table stays aligned to the latest valid item state.

What changes for MotherDuck is the operating style, not the correctness rules. You still need a stable business key, tombstone protection for deletes, and a newer-wins rule for retries. The benefit is that the destination path stays lightweight while still giving you transactional final-state behavior.

Why Teams Struggle with This ​

Small-team destinations can create false confidence. The setup feels simpler, so teams sometimes skip the exact rules that make CDC trustworthy.

  • DynamoDB Streams must still be enabled correctly with NEW_AND_OLD_IMAGES.
  • Partition key and sort key combinations need to be carried forward as the warehouse identity when the table uses both.
  • REMOVE events need delete handling, not just a note that a delete happened.
  • A small destination does not eliminate replay or stale-write problems.

How Skippr Handles It ​

Skippr keeps the source details visible by using DynamoDB Streams directly and storing the last processed sequence number per shard for resume. That makes the setup easier to explain than a pile of custom Lambda glue.

For MotherDuck, Skippr adds _skippr_order_token and tombstone tables automatically and applies final-state reconciliation with DuckDB semantics. The integration stays compact, but the warehouse contract is still deliberate.

  • Native DynamoDB Streams capture for item-level mutations.
  • Durable per-shard resume positions for restarts.
  • MotherDuck final-state reconciliation with order-token guards.
  • Delete protection through companion tombstone tables.

What the First Useful Version Looks Like ​

This setup makes sense when the team wants DynamoDB CDC without adopting a heavier warehouse stack right away.

The important habit is to stay just as strict about keys, delete semantics, and newer-wins ordering as you would in a larger platform.