Forum Discussion
Semantic Model Refresh Failure
My pipeline fails to refresh after some daysthen I go to semantic model settings and update the cloud connection each timethe cloud connection unlinks itselfonce the cloud conn is set, the refresh works fine for a day or two and after that same thinghappesrefresh successful, how can I fix the unlinking of cloud connection
My all pipelines fails at last step , where I refresh the semantic model .
Each time the problem remains same , ie the cloud connection unlinks.
Once I link again the cloud connection , the semantic model refresh activity works fine for day or two and again fails due to unlink of cloud connction.
Any permanent solution of this?
Hello IAMCS
We’re seeing this quite frequently: Fabric pipelines failing at the final semantic model refresh step. The native Semantic model refresh activity works in simple cases but is fragile in real‑world usage, particularly where there are concurrent refresh attempts, OAuth token instability with user accounts, or tight capacity constraints. As the activity wraps the Power BI REST API and polls for completion, any collision or transient issue often causes the pipeline to fail even though the model refreshes successfully when run manually.
For more reliable behaviour, many teams are calling the Power BI REST API directly via a Web activity, using a service principal and explicit polling logic. The commonly used endpoint is:
POST https://api.powerbi.com/v1.0/myorg/groups/{groupId}/datasets/{datasetId}/refreshes
This pattern allows better handling of “refresh already in progress” scenarios, clearer error reporting, and controlled retries. In enterprise environments with multiple pipelines or schedules, it has proven significantly more robust than relying solely on the built‑in activity.An earliest post on this issue can be found here -
Solved: Re: Semantic Model Refresh not working in Data Pip... - Microsoft Fabric Community
3 Replies
Hello IAMCS
We’re seeing this quite frequently: Fabric pipelines failing at the final semantic model refresh step. The native Semantic model refresh activity works in simple cases but is fragile in real‑world usage, particularly where there are concurrent refresh attempts, OAuth token instability with user accounts, or tight capacity constraints. As the activity wraps the Power BI REST API and polls for completion, any collision or transient issue often causes the pipeline to fail even though the model refreshes successfully when run manually.
For more reliable behaviour, many teams are calling the Power BI REST API directly via a Web activity, using a service principal and explicit polling logic. The commonly used endpoint is:
POST https://api.powerbi.com/v1.0/myorg/groups/{groupId}/datasets/{datasetId}/refreshes
This pattern allows better handling of “refresh already in progress” scenarios, clearer error reporting, and controlled retries. In enterprise environments with multiple pipelines or schedules, it has proven significantly more robust than relying solely on the built‑in activity.An earliest post on this issue can be found here -
Solved: Re: Semantic Model Refresh not working in Data Pip... - Microsoft Fabric Community
As of my experience this usually happens when someone re-publishes the semantic model ant that user does not have permission to that connection.
Hi IAMCS,
Thank you for reaching out to Microsoft Fabric Community.
Thank you deborshi_nag and cengizhanarslan for the prompt response.
As we haven’t heard back from you, we wanted to kindly follow up to check if the solution provided by the user's for the issue worked? or let us know if you need any further assistance.
Thanks and regards,
Anjan Kumar Chippa
