Forum Discussion

Rdomart's avatar
Rdomart
New Member
6 months ago

Direct Lake - Delta Protocol Violation

Hello everyone, 

 

I am facing a recurrent issue with my Fabric / Power BI Data Architecture. 

We have a Lakehouse with several Delta Tables that are partionned on one or more fields. (example : ref.dim_Customers)

On top of these tables we have created a series of shortcuts that are used in different schema (example : market.dim_Customers and sales.dim_Customers). For readibility and usability by our Data Analysts.  This architecture works well at the Lakehouse, Sql Endpoint Level. 

On top of these shortcuts we have created some Direct Lake Model. (Direct Lake Only, using the Direct Lake connection string with a Service Principal). 

 

But we are facing a recurrent issue with theses models. Once in a while (often after a redeployment) we get the following error : 

"Premium_ASWL_Error_Details_Label","detail":{"type":1,"value":"Delta protocol violation: partition values cannot be found for the table" market Dim_client.

 

The only way to solve this issue is to manually delete the source table (ref DimClient) and rebuild it. We do not re-generate the shortcuts. This issues occurs for every shorcuts in our DiretLake model referencing a partitionned table. 

 

I've looked on the Forum but haven't found any solution, or any hint on how to solve thi issue. 

 

Thank you for your help. 

 

 

Need Help

5 Replies

  • GilbertQ's avatar
    GilbertQ
    Super User
    6 months ago

    Hi Rdomart 

     

    When you are getting this issue, could you have a look at your source data to make sure that there are values that are used in the partitioning? To enable the inserts to partition the data correctly into your delta tables.

  • Rdomart's avatar
    Rdomart
    New Member
    6 months ago

    Hello GilbertQ , 

    I will look into it next I get this error. 

    But I'm pretty sure these tables are correct : I've tried to query them using SparkSQl & Sql Endpoind and there are working well (both shorcuts and tables) 

     

  • Rdomart's avatar
    Rdomart
    New Member
    6 months ago

    Update  : I got the error and had a look on source tables. Partitions, Files and Data are ok. It is really only related to the DIrect Lake mode

     

  • v-achippa's avatar
    v-achippa
    Community Support
    6 months ago

    Hi Rdomart,

     

    Thank you for reaching out to Microsoft Fabric Community.

     

    Thank you GilbertQ for the prompt response.

     

    Thank you for the response. Since the tables and partitions work correctly in Spark sql and the sql endpoint, this is not a data issue.
    This is a Direct Lake limitation when using shortcuts over partitioned delta tables, and can occur after redeployments due to metadata desynchronization. Rebuilding the table works because it refreshes the metadata.

    For Direct Lake models, avoid shortcuts and prefer the base delta tables directly. If shortcuts are required, please raise a Microsoft support ticket, as this is a direct lake engine issue.

    Create a Fabric and Power BI Support Ticket - Power BI | Microsoft Learn

     

    Thanks and regards,

    Anjan Kumar Chippa

    • v-achippa's avatar
      v-achippa
      Community Support
      6 months ago

      Hi Rdomart,

       

      As we haven’t heard back from you, we wanted to kindly follow up to check if your issue is resolved? or you have raised the support ticket?

       

      Thanks and regards,

      Anjan Kumar Chippa