Guide
SSIS to Microsoft Fabric migration guide
To migrate SSIS to Microsoft Fabric, first decide for each package whether to keep it running or convert it. Microsoft offers two ways to keep packages running and one way to rebuild them. Converting turns each package into native Fabric pipelines and notebooks. This guide covers the routes, eight steps from inventory to cutover, what breaks, and how Novuz proves each converted package.
The routes
What are the ways to move SSIS to Microsoft Fabric?
| Route | What you end up with | Best for | Limits to know |
|---|---|---|---|
| Invoke SSIS Package activity in Fabric | Your existing packages, run from a Fabric pipeline | Running packages with minimal change | In preview. Microsoft lists no on-premises or private-network sources and no custom or third-party components |
| Azure-SSIS integration runtime | Your existing packages, run by Azure Data Factory and called from Fabric | Keeping packages running in Azure while the rest moves | The SSIS logic stays as it is |
| Rebuild in Fabric Data Factory | Fabric pipelines and Dataflow Gen2, built by your team | A small number of simple packages | Each package is redesigned by hand |
| Convert with Novuz | Native Fabric pipelines, notebooks and warehouse objects | Estates where the packages carry the business rules | Scope is fixed at assessment; Script Tasks are reviewed by an engineer |
The first three rows repeat Microsoft’s documentation as cited on the SSIS to Fabric options page. Last checked: 2026-11-17.
Eight steps
How do you migrate SSIS packages to Microsoft Fabric, step by step?
- Inventory the estate. Count every package, data flow, Script Task, connection, package configuration and SQL Agent job. The count sets the cost, not the data volume.
- Find the dead packages. Packages no job calls and tables nothing reads leave the scope, with the evidence recorded.
- Choose a route for each package. Keep simple, stable packages running if you must; convert the ones that carry business rules.
- Choose the Fabric target for each object. Set-based T-SQL usually goes to the warehouse; heavy transforms go to Spark notebooks on the lakehouse.
- Convert the control flow and data flows. Precedence constraints become pipeline dependencies, and lookups and slowly changing dimensions become MERGE patterns.
- Rebuild the schedules. SQL Agent jobs, steps and dependencies become Fabric pipeline schedules and triggers, in the same run order.
- Reconcile every package. Run the original and the converted logic on the same inputs and compare schemas, row counts and column values.
- Cut over by domain. Run old and new side by side for one domain, switch it, then retire that domain’s packages.
Mapping
What does each SSIS object become in Fabric?
| SSIS or SQL Server object | Fabric target |
|---|---|
| Control flow and precedence constraints | Data Factory pipeline activities and dependencies |
| Data flow tasks | Spark notebooks or Dataflow Gen2 |
| Execute SQL tasks | Warehouse stored procedures or notebook cells |
| Script Tasks and Script Components | Notebook code, reviewed by an engineer |
| Stored procedures and functions | Fabric warehouse T-SQL or PySpark |
| Parameters and configurations | Pipeline parameters and variable libraries |
| SQL Agent jobs and schedules | Pipeline schedules and triggers |
Risks
What breaks in an SSIS to Fabric migration?
- Script Tasks and Script Components hold C# or VB.NET code with no direct Fabric equivalent.
- Package configurations and environment variables are spread across files, the catalog and SQL Agent jobs.
- Lookup, Slowly Changing Dimension and Merge Join behave differently when rebuilt as Spark or T-SQL.
- Event handlers and precedence constraints carry error handling that is easy to lose.
- The Fabric warehouse supports a subset of T-SQL, so some procedures need a different pattern.
Proof
How does Novuz prove a converted SSIS package matches?
Novuz runs the original package and the converted pipeline on the same input data. It compares schemas, row counts and column-level values, and marks each object pass, fail or review, with every miss listed.
The assessment costs $3,000, takes three business days and fixes the price. A First Domain of up to 1,500 components takes 5 to 6 weeks. The software runs inside your own Azure tenant.
FAQ
Frequently asked questions
How do I migrate SSIS packages to Microsoft Fabric?
Inventory every package, Script Task and SQL Agent job. Decide per package whether to keep it running, through the Invoke SSIS Package activity or the Azure-SSIS runtime, or convert it to native Fabric pipelines and notebooks. Rebuild the schedules, reconcile each package on the same inputs, and cut over by domain. Novuz converts and reconciles the packages by fixed rules.
Can SSIS packages run in Microsoft Fabric without changes?
Some can. Microsoft’s Invoke SSIS Package activity runs existing packages from a Fabric pipeline and is in preview, with no on-premises or private-network sources and no custom components. The Azure-SSIS integration runtime also keeps packages running. Both keep the SSIS logic as it is.
How long does an SSIS to Fabric migration take?
It depends on the package count and complexity. With Novuz, the assessment takes three business days and a First Domain of up to 1,500 components takes 5 to 6 weeks. The assessment gives a dated plan for the full estate.
Should SSIS logic go to the Fabric warehouse or the lakehouse?
Often both. Set-based T-SQL logic usually fits the warehouse, and heavy transforms fit Spark notebooks on the lakehouse. Decide per object, by complexity and by how the object is used downstream.
What happens to SQL Agent jobs in Fabric?
They are rebuilt as Fabric pipeline schedules and triggers. Keep the original dependencies and run order, so the old and new systems can run in parallel on the same schedule before cutover.
- 1 Microsoft facts as cited on the SSIS to Fabric options page, from Microsoft Learn. Last checked: 2026-11-17.
Sources and further reading
Written by Ted Ghose, Co-founder and CTO, Novuz
Last reviewed:
Find out what your logic debt is costing you.
Every object counted, lineage, dead code and a fixed price in three business days. $3,000, credited.
Not sure which door? The scorecard takes ten minutes and tells you.