Forum Discussion
help in refresh in Power BI service
- 10 months ago
Hi Anonymous ,
Thanks for the update. The issue might be due to the old partitions that weren’t fully cleared before the refresh. When you disable incremental refresh and republish, the Service still retains historical partitions unless you explicitly delete them. Try these below steps by using some sample data.
Disable incremental refresh in Desktop and refresh locally.
Publish to Service, overwrite dataset.
In Power BI Service > Dataset Settings > Incremental refresh history,
Delete all partitions (this step is critical).
Trigger a full refresh in Service, this rebuilds schema without incremental logic.
Re-enable incremental refresh in Desktop, set RangeStart/RangeEnd, and publish again.
Run the refresh again in Service.
Hope this helps.
Thank you.
Hi Anonymous
The error you’re encountering — “Expression.Error: The column 'UPD_DATE' of the table wasn’t found” — typically occurs when Power BI Service tries to apply query folding during an incremental refresh, but the column references in the partition queries don’t fully match the final query structure from Power BI Desktop. In your case, even though the column UPD_DATE was removed in Power Query, it still exists in the source database and may be referenced implicitly by the incremental refresh logic or cached metadata from an earlier version of the dataset. This mismatch often surfaces only in the Power BI Service (not in Desktop) because the Service re-generates SQL queries dynamically for each partition during refresh, and it may still be expecting that column due to cached schema or folding behavior.
To fix this, you should first ensure that every step in Power Query referencing UPD_DATE (even indirectly) is fully removed — including any applied filters or incremental refresh parameters that may depend on it. Then, refresh the preview in Power Query, save the PBIX, and republish the dataset as a new version (overwriting the existing one) so the Service rebuilds partitions from scratch with the updated schema. If possible, disable incremental refresh temporarily, refresh once to reset cached schema, then re-enable it and republish. This approach clears the old metadata references that often persist in the Service and ensures that incremental partitions are generated only using the currently valid columns.