How would you implement incremental loads with CDC in Fabric?

• A. Use change data capture on source systems but do not propagate changes to lakehouse tables.
• B. Use change data capture on source systems but only for full reloads.
• C. Use change data capture on source systems, propagate incremental changes to lakehouse tables via pipelines or Spark jobs, and maintain proper sink upserts.
• D. CDC is not supported in Fabric.

Answer: (C)

Incremental loads with CDC in Fabric rely on capturing changes at the source and applying them to the lakehouse as they happen. Start by enabling CDC on the source systems to capture inserts, updates, and deletes. Then propagate those change events into Fabric’s lakehouse using pipelines or Spark jobs that process the CDC stream. The key is applying those changes to the target tables with sink upserts (MERGE behavior) so inserts create new rows, updates modify existing ones, and deletes remove rows as needed. This keeps the lakehouse in sync with the source with minimal data movement.

This approach is preferred because it delivers true incremental loading and maintains accurate state over time, rather than reloading entire datasets. The other options miss the essential propagation of changes or rely on full reloads, and CDC is indeed supported in Fabric.