COF-C03 Snowflake Streams, Tasks, and Incremental Pipelines

About this video

Snowflake Streams and Tasks make it possible to build incremental transformation pipelines directly inside Snowflake, but understanding how they work together is more important than memorizing their definitions. In this video, we break down Snowflake-native pipeline design for the SnowPro Core COF-C03 exam and trace the complete flow from source changes to transformed target data. We’ll cover: Snowflake Streams Change tracking and stream offsets Stream metadata Stream consumption Stream staleness Snowflake Tasks Scheduled and triggered execution Serverless vs. warehouse-backed tasks Suspended task state Task privileges Task Graphs DAG dependencies Dynamic Tables Target lag MERGE and idempotency TASK_HISTORY TASK_DEPENDENTS Pipeline reconciliation Common COF-C03 pipeline mistakes One of the most important concepts is that a Stream does not physically store another copy of your source data. It tracks an offset against the source and exposes change records relative to that position. A diagnostic SELECT does not advance that offset. The offset moves when qualifying DML consumes the stream within a successfully committed transaction. Tasks provide the execution layer. They can run SQL on a schedule or through supported triggers, but newly created tasks begin suspended and must be resumed before they execute. We’ll also look at why task success alone is not enough. A successful task tells you that the SQL completed. It does not automatically prove that the correct number of records reached the target or that the resulting business state is accurate. For COF-C03, focus on the full reasoning chain: Source change → Stream offset → Task execution → Target state → Reconciliation That model will help you troubleshoot pipeline scenarios much more effectively than memorizing individual Snowflake objects. #Snowflake #COFC03 #SnowProCore