Forum Discussion
Deployment Pipeline
Hi otalog,
Thanks for the additional details. and thanks to SamInogic for sharing valuable insights.
According to the Microsoft documentation, when deploying content to an existing stage, relationships between items aren't kept if you deploy only one item. For example, if you deploy a report that's bound to a semantic model, only the report is deployed. If you want to deploy everything connected to the report, Microsoft recommends using Select related so that the report and its connected semantic model are deployed together.
The documentation also notes that:
• If a dependent item doesn't exist in the target stage, the deployment can fail.
• Deploying an item without its dependent item can lead to unexpected results if the dependent item in the target stage has changed and is no longer identical to the one in the source stage.
Could you confirm whether, when deploying from Test to Production, you deployed both the report and the semantic model together (using Select related if applicable), or only the semantic model?
For more details, please refer to the below official documentation:
Fabric Application lifecycle management (ALM) deployment pipelines deploy content - Microsoft Fabric | Microsoft LearnI hope this helps. Please feel free to reach out if you have any further questions.
Thank you.Hi,
Yes, that could be the reason.
If the Test workspace has a separate semantic model and report, but the Production workspace has a report with an embedded semantic model, then the two environments are set up differently. In that case, deploying only the semantic model may not update the report in Production as expected.
I'd first check whether the Production report is actually connected to the deployed semantic model or if it's still using its own embedded model. If it has an embedded model, you'll likely need to republish it as a thin report that connects to the shared semantic model.
Keeping the same setup in both Test and Production is the best approach, as it helps avoid issues like this during deployments.
Hope this helps.
Thanks!
Thank you for the reply. After deploying to production, I always refresh the semantic model. However, I noticed a discrepancy between our environments when downloading the files. In the test environment, the report and the semantic model (dataset) are correctly separated, with the report simply pointing to the model. In production, downloading the semantic model results in a file that has the report embedded directly within it. Could this architectural difference be the root cause of the issue, If so how do I resolve it?
Hi,
Yes, that could be the reason.
If the Test workspace has a separate semantic model and report, but the Production workspace has a report with an embedded semantic model, then the two environments are set up differently. In that case, deploying only the semantic model may not update the report in Production as expected.
I'd first check whether the Production report is actually connected to the deployed semantic model or if it's still using its own embedded model. If it has an embedded model, you'll likely need to republish it as a thin report that connects to the shared semantic model.
Keeping the same setup in both Test and Production is the best approach, as it helps avoid issues like this during deployments.
Hope this helps.
Thanks!