transient issue in directLake Semantic Model refresh

[stellahe102]

Hi,

I have a directLake semantic Model set to refresh daily at 7am, it occasionally yields this error in below. The semantic model is only set to refresh once a day at 7am, but as you can see in below screenshot, there are 2 different requests to refresh the same model: scheduled and directLake. The delta tables behind the directLake are completed usually before 3am, so there shouldn't be change at 7am to trigger any direcLake update. Any idea why the model is refreshed twice? we would want to avoid the transient error

"Failure details: The last refresh attempt failed because of an internal service error. This is usually a transient issue"

[stellahe102]

Hi @Anonymous 2 questions -

1. How/Where should I go to see if there is really a metadata update in the underlying delta tables? the pipeline generating the delta tables finishes a couple hours before the model refresh at 7am, so not sure why would the metadata still change at 7

2. there is only one scheduled refresh for the semantic model (see below), how should I stop the "additional" refresh? 

[Anonymous · Accepted solution]

Hi @stellahe102,

Here are the answers to your questions.

1. you can check recent changes by running the sys.delta_history  function in a Fabric Notebook (Spark) or SQL query editor.

2. Could you turn off the refresh that is highlighted in the screenshot below

Regards,

Vinay Pabbu

[stellahe102]

Hi @Anonymous I have an additional question - turning off the "Keep your directLake up to date" refresh button in the semantic model, will it affect the loading speed, or the CPU consumed when openning the dashboard built using this model?

[Anonymous]

Hi @stellahe102,

Glad that your issue is resolved.

Turning off the "Keep your DirectLake up to date" refresh button in the semantic model will not affect the loading speed or CPU usage when opening a dashboard.
CPU usage during dashboard opening depends on query complexity, not the DirectLake refresh setting.

If this post helps, then please consider Accepting as solution to help the other members find it more quickly, don't forget to give a "Kudos" – I’d truly appreciate it!

Regards,

Vinay Pabbu

[Anonymous]

Hi @stellahe102,

As we haven't heard back from you, At this time we are closing this thread. If you have any further issues, please start a new thread in the community forum, and we are here to assist you. Thankyou for your understanding and continuous support.

Thank you for being part of the Microsoft Fabric Community.

Regards,

Vinay Pabbu

[stellahe102]

1.  "sys.delta_history" is not working in fabric notebook, used "describe history table1" instead

2.  Already turned off the directlake sync, seems no transient error any more, thanks for the suggestion! 

[Anonymous]

Hi @stellahe102,

Thank you for reaching out to Microsoft Fabric Community Forum.

This unexpected DirectLake refresh at 7 AM is likely triggered by metadata changes in the underlying Delta tables. Even if no data changes occur, modifications such as schema updates, column reordering, or changes in Delta table properties can still prompt a refresh. Microsoft Fabric detects these changes and initiates an additional refresh.

If this additional refresh is unnecessary, consider adjusting the refresh schedule or disabling automatic DirectLake refresh to prevent transient errors.

If this post helps, then please consider Accepting as solution to help the other members find it more quickly, don't forget to give a "Kudos" – I’d truly appreciate it!

Regards,
Vinay Pabbu