1.6 — Deployment Pipelines and Impact Analysis
DP-600 Domain 1 · Skill 1.6
What You Need to Know
Changes have to reach production safely. Deployment pipelines move content between stages, and impact analysis tells you who and what breaks if you change an upstream item.
Deployment pipelines
| Fact | Detail |
|---|---|
| Stages | 2–10 stages (Development, Test and Production by default), each assigned to one workspace |
| What moves | Item definitions: semantic model metadata, report pages, notebook code, pipeline JSON. Data isn't copied, so refresh models after a first deployment |
| Not copied | Data, permissions, workspace settings, app content, and for semantic models: refresh schedule, credentials, role memberships, endorsement |
| Pairing | Items paired across stages are overwritten on deployment (their IDs and URLs stay the same). An unpaired item with the same name creates a duplicate |
| Autobinding | A report deployed to Test binds to the Test copy of its model. If that model isn't in Test, the deployment fails, so use Select related |
| Deployment rules | Change data sources or parameters per stage, for example a Test server or a production database name |
| Who can deploy | Pipeline admins who are at least Contributor in both the source and target workspaces |
| Automation | Deployment pipeline REST APIs from Azure DevOps or GitHub, authenticated with a service principal |
Incremental refresh policies are carried along when a semantic model is deployed. Its partitions and data in the target stay intact, so you don't reload history in production.
Impact analysis
Before you change or delete a lakehouse, warehouse, dataflow or semantic model, open its Impact analysis (or the workspace lineage view):
- It lists downstream items: models, reports, dashboards and dataflows, including those in other workspaces.
- It shows the number of views and viewers of the affected content, so you know how much a change matters.
- Notify contacts sends one message to the owners of all affected items, for example before you rename a column.
Use the lineage view for the end-to-end picture inside one workspace (sources → lakehouse → model → report). Use impact analysis for cross-workspace effects.
A typical release flow
- Develop in a feature workspace connected to a Git branch, and merge to main through a pull request.
- The Development workspace syncs from main.
- Deploy to Test with the deployment pipeline. Deployment rules point Test at test data. Refresh and validate.
- Deploy to Production, often with an approval step in Azure DevOps that calls the deployment pipeline API.
Key Concept: Deployment pipelines copy definitions and use rules for per-stage values. Impact analysis shows the downstream blast radius, across workspaces, before you make a change.
Exam Traps
- Expecting deployed semantic models to arrive with data, a refresh schedule or role members
- Deleting and recreating production items (which breaks their IDs, URLs and permissions) instead of deploying over paired items
- Checking only the current workspace's lineage when reports in other workspaces use the model
- Hard-coding production connection strings in development instead of using deployment rules or parameters
Check your understanding
Practice Scenario
You plan to rename a column in a lakehouse table. Several semantic models and reports in other workspaces may depend on it. Before you make the change, you want to see every affected item and warn its owners. What should you use?
Log in to save your progress.
Sources
- Introduction to deployment pipelines, Microsoft Learn
- The deployment pipelines process, Microsoft Learn
- Create deployment rules, Microsoft Learn
- Impact analysis, Microsoft Learn
- Lineage in Fabric, Microsoft Learn