I'd suggest you use the Dataflow Graphs view to drive your check:
- Click the Dataflows icon .
- To inspect the number of rows at an individual node, click the node and select show schema. The number of rows is at the top of the schema modal window.
- Keep working your way left, until you find the leftmost node in the graph whose row count is lower than you expect. Click on that node and select either Create erroneous rows table or Create complement table, whichever option can be selected:
- This creates a new Active Table whose contents are the content removed from your dataflow at that step. Analysis of the contents should point out where your data disappeared from the dataflow.
- Once you know where your modeling steps removed the data, click Edit Mode and modify the operation to correct any mistake.
To help narrow down the location of your lost rows, I would recommend you start by checking significant table joins. This can easily be caused by running into integrity constraint violations (ICVs). For additional information on ICVs, see https://discourse.xcalar.com/t/white-check-mark-i-want-to-use-the-change-data-type-but-my-column-has-invalid-values/328