Workspace mode and incremental refresh
- ❌
Desktop Edition - ✔ Business Edition
- ✔ Enterprise Edition
- ✔Tabular Editor 2.X
- ✔Tabular Editor 3.4.2 and earlier
Important
This article only applies to versions 3.4.2 and earlier of Tabular Editor. Since the 3.5.0 update, Workspace Mode will not overwrite deployed Refresh Policy partitions from scheduled refreshes. Refresh policy partitions will also not be serialized in source control. You can change this setting in 'Tools > Preferences... > Save-to-Folder'.
Incremental Refresh creates new partitions upon the first scheduled refresh in a day. As a result, any local metadata (i.e. .bim
or database.json
) will be out-of-sync with the remote model metadata after the refresh. As a result, when working with a model that has tables configured with Incremental Refresh, Workspace Mode is not recommended.
Important
Setting up Incremental Refresh with Tabular Editor 3 is limited to dataset hosted in the Power BI Datasets service. For Analysis Services, custom partitioning is required.
Workspace Mode is not Recommended
The reason is because Workspace Mode will overwrite the remote model metadata with local metadata files; any out-of-sync changes (like to Policy Range partitions) will be lost. When working with Workspace Mode on these models, you would need to Apply refresh policy for tables using incremental refresh before saving changes every day.
Recommendation: Develop & Deploy from Local Metadata
Instead, it is recommended to develop the model from the local metadata files. Changes can be deployed excluding partitions governed by a Refresh Policy, so there is no risk of overwriting the policies created by Power BI. A second read/refresh instance of Tabular Editor can be connected to the remote model for testing purposes.
To deploy the model, go Model > Deploy... which opens the Deployment Wizard. Here you can select whether you want to include partitions governed by Incremental Refresh policies:
By deploying the model without these Policy Range partitions, you are mitigating any potential impact due to out-of-sync incremental refresh partitions between the metadata and remote model.