Data Integration analyses all the transformations from the source to the target. If all the transformations are compatible in the target, it pushes the entire mapping logic to the target. If it cannot push the entire mapping logic to the target, Data Integration first pushes as much transformation logic to the source database and then pushes as much transformation logic as possible to the target database.

When a transformation is not supported in the mapping, the task partially pushes down the mapping logic to the point where the transformation is supported for pushdown optimization. However, this applies only to Databricks SQL endpoints.

When you enable full pushdown optimization, you can determine how Data Integration handles the job when pushdown optimization does not work. You can use the If pushdown mode is not possible, cancel the task option to set the task to fail or run without pushdown optimization.