Forum Discussion

fazithzakirll's avatar
fazithzakirll
Regular Visitor
6 months ago
Solved

Delta Lake Refresh Issue

I’m running 4 Power BI reports in Direct Lake mode from a Microsoft Fabric Lakehouse, each connected to different semantic models.

Three reports work as expected, but one report (let’s call it “Ops Dashboard”) is not updating correctly.

Setup

  • All reports have “Keep your Direct Lake data up to date” = ON

  • Two reports (including Ops Dashboard) use a manual calendar table

Expected Behavior

When I click “Refresh visuals”, the data should reflect the latest data from the Lakehouse, and the semantic model refresh time updates accordingly.

Actual Behavior (Issue)

For Ops Dashboard:

  • The refresh time updates after clicking Refresh visuals

  • But the data itself does NOT update

  • To get the latest data, I must manually refresh the semantic model in the workspace

For the other 3 reports:

  • Refresh visuals correctly updates both timestamp and data

Question

Has anyone faced this issue where:

  • Direct Lake is enabled

  • “Keep your Direct Lake data up to date” is ON

  • Refresh visuals updates the refresh time but not the data

Any ideas what could cause this behavior in just one semantic model/report?

Thanks in advance!

  • Hi fazithzakirll ,

    Since the other three reports are working correctly with the same setup, it seems the issue is specific to this semantic model.

    The behavior you mentioned refresh time updating but data remaining unchanged typically indicates that the model is displaying cached data and not retrieving the latest information from the Lakehouse until a full model refresh occurs.

    As you’ve confirmed it’s connected directly to the Lakehouse and not the SQL endpoint, here are a few things to check.

    1. Compare this model to the others for differences, such as calculated columns, complex measures, or larger tables.

    2. Because this model uses a calendar table in Import mode, you might try recreating or temporarily removing it to see if  Refresh visuals starts updating the data.

    3. If possible, make a small test semantic model using the same tables to see if it updates as expected. If the test model works, the issue is likely within this specific model.

     

    These steps should help determine if it’s a model level issue rather than a Direct Lake problem.

     

    Thank You.

10 Replies

  • Hi fazithzakirll 

     

    That’s a strange one, especially when your other reports are behaving fine.

    It sounds like your "Ops Dashboard" semantic model might be falling back to DirectQuery mode. When this happens, "Refresh visuals" can sometimes behave inconsistently because it's no longer using the high-speed Direct Lake path.

    This fallback usually happens for a few specific reasons:

    • Memory Pressure: If the columns needed for that specific report are too large, the capacity might refuse to load them into memory and switch to DirectQuery instead.

    • Unsupported Features: Direct Lake has some strict requirements. If your "Ops Dashboard" uses certain features (like complex DAX calculated columns or RLS in a specific way), it might be forcing a fallback.

    • Lakehouse Schema Changes: If the underlying Delta tables for that specific model were updated in a way that altered the schema, the link can sometimes "glitch" until a full manual refresh is triggered.

    I’d suggest using the Performance Analyzer in Power BI Desktop (connected to the service) or looking at the DAX Studio metrics to see the "Direct Lake Readiness" of your tables. Microsoft has a great guide on how to check if your queries are actually using Direct Lake or falling back to DirectQuery: Analyze Direct Lake query processing.

     

    If you find it is falling back, you might need to investigate your dax measures and perhaps use the vertipaq analyzer to see if you have some columns in your tables that are too large to be loaded into memory (eg. high cardinality columns like date/time columns etc.).

    Br
    Asger

     

  • Hi fazithzakirll,

     

    Since 3 reports work under the same setup, most probably the issue lies within that semantic model. Additionally, refresh timestamp updates but the data does not, it suggests the model is not re-reading the latest Lakehouse data.

     

    Could you confirm whether the semantic model is connected to the SQL Endpoint (check in settings, if model is showing DirectQuery or referencing SQL Endpoint instead of Direct Lake or Lakehouse)? It is a known observation that the SQL Endpoint can lag behind Delta commits and takes longer to sync. If yes, you should switch to Lakehouse, or give a long delay between tables update and report refresh.

     

    Please share whatever works for you eventually, so that it helps us all in such scenarios.

    • fazithzakirll's avatar
      fazithzakirll
      Regular Visitor

      The Semantic Models are all connected direclty to the Lakehouse. I am looking into other Solutions

      • V-yubandi-msft's avatar
        V-yubandi-msft
        Community Support

        Hi fazithzakirll ,

        Since the other three reports are working correctly with the same setup, it seems the issue is specific to this semantic model.

        The behavior you mentioned refresh time updating but data remaining unchanged typically indicates that the model is displaying cached data and not retrieving the latest information from the Lakehouse until a full model refresh occurs.

        As you’ve confirmed it’s connected directly to the Lakehouse and not the SQL endpoint, here are a few things to check.

        1. Compare this model to the others for differences, such as calculated columns, complex measures, or larger tables.

        2. Because this model uses a calendar table in Import mode, you might try recreating or temporarily removing it to see if  Refresh visuals starts updating the data.

        3. If possible, make a small test semantic model using the same tables to see if it updates as expected. If the test model works, the issue is likely within this specific model.

         

        These steps should help determine if it’s a model level issue rather than a Direct Lake problem.

         

        Thank You.

  • Hi fazithzakirll 

     

    Can you add more details on what you mean by: " a manual calendar table" ??

    How are the calendar tables for other reports when compared to the one with the issue?

     

  • I Created a New Table in the Semantic Model, 
    "Calendar =

    ADDCOLUMNS (
        CALENDAR ( DATE(2020,1,1), TODAY()),  
        "Year", YEAR([Date]),
        "Moanth", MONTH([Date]),
        "Day", DAY([Date])
    )
    • Pragati11's avatar
      Pragati11
      Super User

      HI fazithzakirll 
      I have a feeling this manual calendar table is creating issues as sometimes data is not refreshed properly because of various reasons.

      In your semantic model, what is the storage mode for this manual calendar table?
      Can you share picture of your semantic model please?

       

      • fazithzakirll's avatar
        fazithzakirll
        Regular Visitor

        The Calendar table is in Import Mode, However, I have another report which has a similar Calendar table which is working as expected. This is where I am not sure if it is because of the calendar table.

  • Hi fazithzakirll ,

    We haven’t received a response from your end yet. Please let us know whether the issue has been resolved or if you’re still facing any difficulties. Feel free to reach out if you need further assistance.

     

    Thank you.

  • Hi fazithzakirll ,

    Could you let us know if your issue has been resolved or if you are still experiencing difficulties? Your feedback is valuable to the community and can help others facing similar problems.