Hi @Debasis_Pal,

This is expected behaviour with the Databricks Power BI task. If you enable "Refresh after update," Databricks doesn’t just publish the semantic model metadata... it also triggers the Power BI data refresh and waits for that step to finish, which is why the workflow can sit there holding compute for a long-running Import model refresh. The product docs call out that this setting is optional and only applies when you're using Import mode, so for very large datasets, a better pattern is usually to leave that option off in the Databricks job and trigger the Power BI refresh asynchronously from something lightweight like ADF, Logic Apps, or an Azure Function instead. The Databricks docs for the Power BI task are here: Power BI task for jobs.

If the goal is to avoid long refresh windows altogether, it’s also worth checking whether the model can move away from full Import refresh behaviour. Databricks documents that with DirectQuery, the data is not stored in Power BI, and queries are pushed to the SQL warehouse at query time, which is often a better fit when freshness matters more than imported-cache performance. There’s also broader guidance here on Power BI with Azure Databricks: Power BI with Azure Databricks.

The optimal approach is usually to decouple "publish/update the model" from "refresh the dataset." Let Databricks finish the data prep and model update, then let an external async orchestrator kick off the Power BI refresh so your Databricks compute is not waiting around for an hour just to monitor Power BI.

If this answer resolves your question, could you mark it as “Accept as Solution”? That helps other users quickly find the correct fix.