Forum Discussion

jaryszek's avatar
jaryszek
Icon for Super User rankSuper User
1 year ago
Solved

How to create deployment pipeline for different customers?

Hello,  how to create a deployment pipelines in 2 kind of architectures for different customers?:   1) data is on customer tenant , they are using one lake shortcut to connect to OneLake. How to...
Need Help
  • v-sgandrathi's avatar
    v-sgandrathi
    1 year ago

    HI jaryszek,

    Thank you AmiraBedh for your response in this thread.

     

    For production environments, please consider the following additional points:

    Verify that shared data has the appropriate access rights in both tenants, as incorrect permissions often cause shortcuts or tables to be missing.

    When making changes in DEV, such as adding tables or columns, remember to refresh metadata or apply rebind rules in your pipelines to ensure TEST and PROD remain aligned.

    If you support multiple customers, automating deployments using options like REST API or Git integration can help manage not only semantic models but also pipelines, notebooks, and dataflows.

    It is also advisable to set up monitoring for SQL endpoints and Lakehouse, and to prepare a rollback strategy in case any issues arise during deployment.

    These steps are not mandatory, but they will contribute to maintaining stability and consistency as you scale your solution across customers.

     

    Thank you.

AmiraBedh's avatar
AmiraBedh
Icon for Super User rankSuper User
1 year ago

Hello !

Thank you for posting on Fabric community.

If you host the analytics in your Fabric tenant, but read the customer data in their tenant without copying it.

The customer needs to enable external data sharing in their Fabric tenant and create a data share for the folders or tables they want you to use. They invite your tenant.

https://learn.microsoft.com/en-us/fabric/governance/external-data-sharing-overview

https://blog.fabric.microsoft.com/en-us/blog/whats-new-with-onelake-shortcuts

You accept the share in your tenant so this creates a read-only OneLake shortcut in a Lakehouse you choose (no B2B guest needed for the data path). 

https://learn.microsoft.com/en-us/fabric/governance/external-data-sharing-overview

Then create 3 workspaces: Customer DEV / TEST / PROD and each workspace, you create a Lakehouse. Inside each Lakehouse, you need to create a shortcut to the correct path the customer shared 

After that you create a deployment pipeline and attach the DEV/TEST/PROD workspaces.

https://learn.microsoft.com/en-us/fabric/cicd/deployment-pipelines/intro-to-deployment-pipelines

and add the deployment rules so that when you deploy, the semantic model in each stage rebinds to that stage Lakehouse (this is key for Direct Lake).

https://learn.microsoft.com/en-us/fabric/cicd/deployment-pipelines/create-rules

https://blog.crossjoin.co.uk/2025/01/12/changing-the-source-lakehouse-of-power-bi-direct-lake-models-in-deployment-pipelines/

 

Pipelines don’t deploy across tenants so all 3 stages must be in your tenant. Cross-tenant comes from the shared shortcut, not from the pipeline. If the customer wants everything in their tenant, you need to package or automate with REST APIs/Git and let them run their own pipeline there. 

https://community.fabric.microsoft.com/t5/Fabric-platform/FABRIC-deployment-pipeline-deployment-across-tenant/m-p/4414525

The 2nd case is when your data in your tenant and you want to separate workspaces or Lakehouses per customer.

You need one pipeline per customer, each with its own DEV/TEST/PROD workspaces and a Lakehouse per stage.

You need to create customer DEV/TEST/PROD workspaces and Lakehouse DEV/TEST/PROD (and/or Warehouses), plus any Data Factory items, notebooks...

Then build your semantic model in DEV (Direct Lake or Import) and reports.

The rest is almost the same as your 1st situation.