Microsoft Fabric Business Intelligence Data Engineering

Azure Data Factory to Fabric: Upgrade Guide for Reporting Pipelines

Azure Data Factory to Fabric: Upgrade Guide for Reporting Pipelines
Microsoft Fabric

Azure Data Factory to Fabric: Upgrade Guide for Reporting Pipelines

⏱️ 10 min read
Microsoft Fabric · Data Engineering · December 2026
Azure Data Factory to Fabric upgrade guide for reporting pipelines: mount, assess, map connections, upgrade and validate, by Numlytics

Mount first, upgrade in stages, validate before cutover.

Azure Data Factory to Fabric is an upgrade you can run in stages, and for reporting teams that matters. The pipelines that load your warehouse and refresh your Power BI semantic models can move one at a time, while everything else keeps running where it is.

This guide follows Microsoft's assessment-first path: mount your factory, assess readiness, map connections, upgrade, and validate before cutover. It focuses on the reporting pipelines that feed dashboards, and it flags the parts of Microsoft's tooling that are still in preview, so check the current documentation before you run them.

Why Upgrade Azure Data Factory to Fabric

Microsoft's position is clear. Azure Data Factory stays fully supported, your pipelines keep running on the same pricing, and there is no forced upgrade or deadline. The reason to move is where new features land. Fabric Data Factory is built on the same engine and connector library as ADF, so pipelines feel familiar, and the surrounding platform is what changes.

For reporting, three additions stand out. Pipelines work directly with OneLake, warehouses and lakehouses. A semantic model refresh activity lets a pipeline refresh a Power BI model as soon as its load finishes. Outlook and Teams activities can alert owners when a reporting pipeline fails. Fabric also adds built-in CI/CD through deployment pipelines and Copilot help for building pipelines.

Azure Data Factory to Fabric: Differences That Shape the Upgrade

Most pipeline logic carries across. The differences below are the ones that change the plan, so read them before you assess anything.

AreaAzure Data FactoryFabric Data Factory
DatasetsSeparate reusable objectsProperties defined inline in each activity
Linked servicesLinked servicesConnections, with no dynamic properties; parameterize the connection object instead
Global parametersGlobal parametersFabric variable libraries
Integration runtimesSelf-hosted IR, VNet IR, SSIS IROn-premises data gateway, VNet data gateway; run SSIS packages from a Fabric pipeline
TriggersOne trigger can serve many pipelinesPer-pipeline schedules, no central trigger hub; tumbling window becomes interval schedule
Execute pipelineExecute pipeline activityInvoke pipeline activity with a Fabric data pipeline connection
IdentityManaged identityFabric workspace identity
Pricing modelAzure pay-as-you-go, per activity run and data movementCapacity-based (Fabric F SKU)

The most disruptive items for a metadata-driven framework are the connection and scheduling changes. If your reporting loads rely on one parameterized linked service or one trigger serving dozens of pipelines, budget design time, not just conversion time.

Choose an Upgrade Path

Microsoft describes four routes, and you can combine them across waves.

RouteWhat it doesBest for
Mount in Fabric (ADF item)Live view, side-by-side testing, ownership mappingRuns and bills in ADF until you upgrade
Built-in upgrade experienceAssessment first, then guided upgrade of ready pipelinesMost estates; low-risk, incremental
PowerShell upgrade moduleScripted, bulk or CI/CD-driven upgradesLarge estates and repeatable waves
Manual rebuildRecreate pipelines, connections and schedules in FabricLow-parity or custom logic; a chance to modernise
The five stages of an Azure Data Factory to Fabric upgrade for reporting pipelines: mount, assess, map connections, upgrade and validate

The five stages of the upgrade.

Mount Your Factory in Fabric

Mounting is the safest first step. In a Fabric workspace, choose New item, then Azure Data Factory, and pick your factory. You can also start from the Manage tab in ADF. You need contributor access to the factory and member or admin rights in the workspace, and both must sit in the same Microsoft Entra tenant. Keeping the workspace in the same region as the factory is recommended.

You can then view, edit, validate and trigger pipelines from Fabric. Execution does not move. Pipelines, triggers, schedules and integration runtimes keep running in ADF, and Azure still bills you for them. Some items, such as global parameters, are not visible in the Fabric view. If you change your mind, select Unmount.

Use the mounted view to list what exists, assign owners per domain and agree naming and folder conventions. Our Microsoft Fabric migration checklist gives a wave structure that fits this stage.

Assess Pipeline Readiness Before You Upgrade

From your factory in Fabric, select Assess and Upgrade. The readiness assessment is read-only: it changes nothing and does not start an upgrade. It covers the whole estate, can be run on demand and can be downloaded as a report.

Microsoft's planning guide sorts every pipeline and activity into four categories.

  • Ready: can move to Fabric today, for example Copy activities and basic orchestration.
  • Needs review: requires manual changes first, such as converting Synapse notebooks or Spark job definitions.
  • Coming soon: depends on a capability Fabric does not have yet.
  • Not compatible: uses a feature that needs redesign.

Start the triage with reporting. Tag each pipeline by the report or semantic model it feeds, then rank by business impact, complexity and ease of upgrade. Ready pipelines that feed low-risk reports make good pilots. Clean up unused and duplicate pipelines at this stage too, so you do not carry them across.

Map Connections and Start the Upgrade

Select Review connections to map ADF linked services to Fabric connections. For common sources, such as Blob Storage, ADLS Gen2, SQL Server, Azure SQL Database and Azure Data Explorer, the experience can create connections for you using authentication it can map safely, including workspace identity. For the rest, pick an existing Fabric connection or create one in the workspace settings.

Do not skip this. If you map nothing, pipelines still upgrade, but activities that depend on unmapped connections are deactivated, and you must configure and re-enable them before running anything.

Then select Start upgrade, which Microsoft labels as preview. Selected pipelines land in a Fabric workspace folder prefixed with the source factory name, named in the form factory name, underscore, pipeline name. If a pipeline with the same name already exists, the tool skips it. Datasets disappear in the process: their settings are applied inline to each activity.

Scripted alternative: the Microsoft.FabricPipelineUpgrade PowerShell module imports your factory, converts it, takes a resolutions file that maps linked services to connection IDs, and exports the result. Microsoft recommends the built-in experience for most upgrades and the module for scripted, bulk or CI/CD-driven work.

"A partial upgrade is a supported outcome, not a failure."

What the Azure Data Factory to Fabric Upgrade Does Not Do

The built-in experience does not upgrade everything. Microsoft lists the items below as out of scope today, and your assessment shows which of your pipelines touch them.

ItemWhat to do
Self-hosted integration runtimeReplace with an on-premises data gateway
Managed virtual network IRUse the Fabric VNet data gateway
Change data captureRebuild as Copy job items
Tumbling window triggersRedesign as interval schedules; watermark and backfill logic needs rework
Custom event triggers and chained triggersNot upgraded; rebuild by hand
Dynamic (parameterized) linked servicesEach permutation becomes its own connection, or parameterize the connection object
SAP ECC, SAP BW and other long-tail connectorsNo equivalent connector; redesign
User-assigned managed identityUse workspace identity as a workaround
Copy upsert into Lakehouse tablesCopy to staging, then run a notebook MERGE
Global parametersRecreate as a Fabric variable library

Global parameters deserve a closer look because they often hold environment names and paths used by reporting loads. The conversion is manual: export your parameters from the Manage hub, create a variable library, add each parameter as a variable and rewrite the expressions.

-- Azure Data Factory expression
@globalParameters('Environment')

-- Fabric variable library expression
@pipeline.libraryVariables.Environment

Variable libraries can be versioned and deployed through deployment pipelines, so build one library per environment if you promote Dev, Test and Prod.

Mapping Data Flows in an Azure Data Factory to Fabric Upgrade

Mapping Data Flows used to be the sticking point. Eligible flows can now be upgraded to MDF transforms in Dataflow Gen2, which keep the Spark-based execution model, and Microsoft lists the transforms and the upgrade experience as generally available.

The limits still matter. Flowlets, the Data Flow Library and user-defined functions are not supported. MDF transforms run only through the pipeline Dataflow activity, managed virtual network connectivity is not supported, and on-premises and VNet data gateways cannot be used with them. Not every Mapping Data Flow capability is available, so test each flow's output against the original.

Key Takeaways
  • Mounting your factory changes nothing: pipelines run and bill in ADF.
  • Assess first; the readiness report is read-only and downloadable.
  • Map connections before upgrading, or activities arrive deactivated.
  • Triggers arrive disabled and global parameters are not upgraded.
  • Validate reporting refreshes in non-production before cutover.

Validate Reporting Pipelines Before Cutover

After the upgrade, Microsoft's own list is short: validate connections and credentials, re-enable and configure triggers, run end-to-end tests, and test in a non-production environment before you upgrade production. For reporting pipelines, turn that into a checklist.

  1. Connections: every mapped connection authenticates, and secrets work. Key Vault integration in Fabric is more limited than in ADF, so compare supported sources and authentication types first.
  2. Data: row counts, schemas and checksums match between the old and new run. Watch for type mismatches and locale-specific date parsing.
  3. Business measures: key report figures agree within a tolerance you set before testing.
  4. Refresh chain: the load, the semantic model refresh and any downstream alert fire in the right order, using the Fabric semantic model refresh activity where it suits.
  5. Schedules: triggers are re-enabled one pipeline at a time, and the old ADF schedule is turned off on cutover so nothing runs twice.
  6. Performance and cost: run durations are inside your reporting window, and capacity use is checked against your SKU. Our Microsoft Fabric capacity estimator helps size it.

Run both estates in parallel for an agreed period, and keep the ADF version until reporting owners sign off. If you are weighing the cost side, see our guide to Azure Data Factory cost optimisation for what you are paying today.

Upgrade Now or Wait? A Decision Table

SituationSuggested moveWhy
Pipelines are mostly Copy and basic orchestrationUpgrade nowThese land in the Ready category and carry the least risk.
Estate depends on SHIR, SSIS IR or a managed VNetPlan firstGateways and network design must be in place before pipelines move.
Heavy use of Mapping Data FlowsUpgrade in a pilotEligible flows can move to Dataflow Gen2, but check the MDF limitations first.
Parameterized linked services drive a metadata frameworkWait or redesignConnections do not support dynamic properties, so the framework needs rework.
Fabric capacity not yet in placeMount and assess onlyViewing the factory uses no capacity, so you can plan without committing.
Reporting is already on Fabric or moving soonUpgrade the refresh chain earlyFabric pipelines can refresh semantic models directly and sit next to the reports they feed.

Pipelines that come from other sources follow different rules. For SSIS packages, see our SSIS to Fabric Data Factory guide. For Synapse pipelines and the Spark notebooks they call, read the Synapse to Fabric Migration Assistant guide, and for how a combined estate can be planned, see the Synapse to Fabric case study scenario.

Azure Data Factory to Fabric Upgrade FAQ

Is Azure Data Factory being retired?

Microsoft says no. Azure Data Factory remains fully supported, pipelines keep running on the same pricing, and there is no forced upgrade or deadline. New innovation is being built into Fabric Data Factory, which is the reason to plan a move rather than a reason to rush one.

Does mounting my factory in Fabric change anything?

No. A mounted factory is a live view. Pipelines, triggers and integration runtimes keep running in Azure Data Factory, and you are billed by Azure for ADF, not by Fabric capacity. You can unmount at any time.

Can I upgrade only some pipelines?

Yes. A partial upgrade is a supported outcome. Upgrade the ready pipelines, keep the rest running in Azure Data Factory, and redesign the ones that depend on unsupported features.

What happens if I do not map connections?

Pipelines still upgrade, but activities that depend on an unmapped connection are deactivated. Create or map the Fabric connections, then re-enable those activities before you run anything.

Do triggers and global parameters upgrade automatically?

Schedule and storage event triggers upgrade but arrive disabled, so you enable them after validation. Other trigger types are rebuilt by hand. Global parameters are not upgraded by either the built-in experience or the PowerShell tool; you recreate them as Fabric variable libraries.

Are Mapping Data Flows supported?

Eligible Mapping Data Flows can be upgraded to MDF transforms in Dataflow Gen2, and Microsoft lists both the transforms and the built-in upgrade experience as generally available. Flowlets, the Data Flow Library and user-defined functions are not supported, so review the limitations first.

If you want help assessing your own estate, see our Microsoft Fabric migration service or request a free pipeline readiness assessment.