Synapse to Fabric Migration Assistant: A Hands-On Reporting Guide and Its Limits
Four steps, then test reporting before you reroute.
Synapse to Fabric migration no longer has to start from a blank page. Microsoft's Fabric Migration Assistant for Data Warehouse copies a dedicated SQL pool into a Fabric warehouse in four steps, and the last step is the one reporting teams care about: rerouting connections so reports read from the new warehouse.
This hands-on reporting guide walks through each step, shows where the assistant stops, and covers the separate tools for Spark notebooks and pipelines. The steps follow Microsoft Learn at the time of writing, and two of the tools are still in preview, so check the current documentation before you run them.
Synapse to Fabric Migration Assistant: What It Covers
The Fabric Migration Assistant is built into Fabric. It copies metadata and data from the source database and automatically converts the schema to Fabric Data Warehouse, with AI help for incompatibilities. Besides Synapse dedicated SQL pools, it supports SQL Server and, in preview, Teradata metadata.
It captures tables, views, functions, stored procedures and security objects such as roles, permissions and dynamic data masking. It works from an uploaded DACPAC file or a direct connection to the source, and it always creates a new warehouse item. Note that the new warehouse uses case-insensitive collation, whatever the default warehouse setting is.
The four steps of the warehouse Migration Assistant.
When to Use the Assistant and When to Go Manual
Use the assistant when the warehouse is the main prize and you want a guided, repeatable path with a clear list of what failed. Go manual, or mix methods, when you want fine control. Microsoft also documents stored procedures that convert the schema, extract data to ADLS and flag unsupported T-SQL, for teams that prefer an IDE-driven approach.
Either way, decide the scope first. Our Azure Synapse vs Microsoft Fabric guide helps decide which workloads should move, and the Microsoft Fabric migration checklist keeps each wave consistent.
Before You Start: Prerequisites for Synapse to Fabric Migration
- A Fabric workspace with an active or trial capacity. The assistant creates the new warehouse inside it.
- Either a DACPAC extracted from the dedicated SQL pool, or the server name, database name and authentication details for a direct connection.
- For AI-assisted fixes, Copilot enabled by your tenant admin on a paid F2 or P1 capacity or higher. Trial capacities do not support Copilot, and tenants outside the US and France need an extra cross-region setting.
You can extract a DACPAC from a dedicated SQL pool with SQL Server Object Explorer in Visual Studio 2022, or with the SqlPackage command-line utility. Inventory the sessions that connect to the pool now, because you need them for step 4.
Step 1: Copy the Metadata
In the Fabric workspace, select Migrate, choose the Azure Synapse Analytics dedicated SQL pool tile, and pick either the file upload or the direct connection. Select the objects to migrate, name the destination warehouse and review.
The assistant translates the T-SQL metadata to syntax Fabric supports and opens a summary with migrated objects and objects that need attention. The State and Details columns show where it adjusted something, for example a data type. Expect type changes such as nvarchar to varchar, and datetimeoffset is not supported, so time zone offsets need a separate column.
You can add objects later with Get more objects. If a name already exists, you choose Keep existing, Keep incoming or Rename incoming. Be careful with Keep incoming on a table, because it drops and re-creates it, and data in it is lost.
Step 2: Fix Problems
Scripts that could not be translated land in the Fix problems list. The assistant splits them into primary objects, which depend on nothing, and dependent objects, which wait until the objects they rely on are fixed. Primary objects are ordered by how many objects depend on them, so you clear the biggest blockers first.
Each failed object opens as a query with comments explaining what was changed. Fix it by hand or use Fix query errors for a Copilot suggestion, then run it. Copilot makes mistakes, so review every suggestion. You can skip objects you do not need.
Handle security here, before any data moves. Roles, permissions and dynamic data masking usually migrate, but SQL authenticated users should be replaced with Microsoft Entra users, and column-level encryption needs another approach. IDENTITY columns also behave differently in Fabric.
Step 3: Copy the Data
Select Copy data, choose Use a copy job and name it. Connect to the source, pick the tables, choose the destination warehouse, map columns and choose a mode. A one-time full copy is recommended for migration, and continuous incremental copy is available if you need the two sides in step for a while.
This is the right moment to plan the parallel run. Keep the old pool available, and agree how long both sides run before reporting switches.
Step 4: Reroute Reporting Connections
This is where reporting migrates. Identify what connects to the old pool: on a dedicated SQL pool, the session information in the pool shows application name, login and IP address. Then work in this order.
- Repoint reporting connections, such as Power BI semantic models and reports, to the new warehouse.
- Test the Fabric warehouse with some reports before rerouting, and compare the figures against the old pool.
- Update data loading connections. For Fabric and Power BI pipelines, use the List Connections REST API to find connections to the old pool, then update them in Manage connections and gateways.
- Mark Reroute connections complete in the assistant.
Validate at three levels: table row counts, key business measures with an agreed tolerance, and finished reports side by side. Our Synapse to Fabric case study scenario shows how that validation fits a wider consolidation, and the Snowflake to Fabric case study shows the same checks from a different source.
- The warehouse assistant has four steps: copy metadata, fix problems, copy data, reroute.
- Fix security objects before you copy any data.
- Keep incoming on a table drops it and loses data.
- Reporting only migrates in step 4, so test reports before switching users.
- Spark and pipelines have separate preview tools and need manual work.
Synapse to Fabric Migration for Spark Notebooks
The Spark Migration Assistant, in public preview, migrates Spark pools, notebooks, Spark job definitions and lake databases. It copies and transforms items but does not move your data, and it does not finish the job: you still refactor code and validate results.
The most common change is replacing mssparkutils with notebookutils, and most file system calls are identical. Credential calls change more. Linked service based secret calls are not supported, so getSecretWithLS becomes notebookutils.credentials.getSecret with the Key Vault URL. Any remaining linked service, TokenLibrary or getSecretWithLS calls fail at runtime in Fabric.
Before you migrate, search the notebook repository for Synapse-specific patterns, including spark.synapse.linkedService, TokenLibrary, synapsesql and several spark.catalog methods. Notebooks with no matches are the safest to move first. If a pool has no custom configuration or libraries, you may not need to migrate it at all, because Fabric's Starter Pool starts sessions in seconds.
Microsoft's Fabric Assessment Tool, in the fabric-toolbox repository, scans a Synapse workspace and reports objects, linked services and likely blockers, which makes a good baseline. For the layer design behind the notebooks, see our guide to medallion architecture in Azure Synapse.
Synapse to Fabric Migration for Pipelines
The Synapse pipeline upgrade experience, also in preview, assesses readiness in your Synapse workspace, shows compatibility gaps at pipeline and activity level, upgrades supported pipelines into a Fabric workspace and exports results to CSV. Upgrade the Spark items first so that notebook and Spark job definition activities map to the right Fabric items.
If a Synapse pipeline only triggers a Spark job definition, consider Fabric's built-in scheduling instead of migrating the pipeline. Our guide to SSIS to Fabric Data Factory covers the pipeline patterns that carry across.
What the Synapse to Fabric Migration Assistant Does Not Cover
There is no full T-SQL compatibility between a dedicated SQL pool and a Fabric warehouse, so some code will need rewriting. Indexes and transparent data encryption are not needed in Fabric, but external tables and multi-statement table-valued functions are currently unsupported.
| Item | What to do | Where it applies |
|---|---|---|
| Spark configurations, custom libraries, executor settings | Recreate in a Fabric environment item | Spark Migration Assistant (preview) |
| Workspaces under a virtual network | Cannot be migrated by the assistant; plan manually | Spark Migration Assistant (preview) |
| Non-Delta tables and .NET for Spark | Convert to Delta; rewrite .NET code in Python or Scala | Spark Migration Assistant (preview) |
| GPU Spark pools | Not available in Fabric; use CPU or keep on Synapse | Spark Migration Assistant (preview) |
| SQL authentication users | Replace with Microsoft Entra users | Data Warehouse assistant |
| Column-level encryption | Use application-layer encryption and dynamic data masking | Data Warehouse assistant |
| External tables and multi-statement table-valued functions | Currently unsupported; redesign | Data Warehouse assistant |
| Report and semantic model connections | Repoint and validate by hand | Manual step 4 |
The assistants also do not do the work around the platform: report validation, capacity sizing, user communication and retirement of the old pool. Our guide to choosing a Microsoft Fabric migration partner lists what to ask for when you want help with that.
Synapse to Fabric Migration Assistant FAQ
What does the Fabric Migration Assistant for Data Warehouse migrate?
It copies metadata and data from a Synapse dedicated SQL pool and converts the schema to Fabric Data Warehouse. Captured objects include tables, views, functions, stored procedures and security objects such as roles, permissions and dynamic data masking. It creates a new warehouse item for the result.
Does the Migration Assistant move my Power BI reports?
No. It moves the warehouse. Reporting is the last step, where you repoint report and semantic model connections to the new warehouse, test them against the old one, and only then switch users over.
Do I need Copilot to use the Migration Assistant?
No, but the AI-assisted fixes need it. Copilot must be enabled by your tenant admin, and it requires a paid F2 or P1 capacity or higher, not a trial. You can still fix failed scripts by hand.
Is the Spark Migration Assistant generally available?
It is in public preview. It migrates Spark pools, notebooks, Spark job definitions and lake databases, but not Spark configurations, custom libraries or executor settings, and it cannot migrate workspaces under a virtual network.
Can I migrate Synapse pipelines with the assistant?
A separate upgrade experience, also in preview, assesses pipeline readiness, shows compatibility gaps and upgrades supported pipelines. Upgrade the Spark items first so notebook activities map correctly, and expect to rebuild anything unsupported.
What happens if I choose Keep incoming on a table?
It drops and re-creates the existing table. If the table has data, you lose it and must load it again, and the assistant does not recover dropped data. Back up or use Rename incoming when in doubt.
To plan your own move, see our Microsoft Fabric migration and Azure Synapse consulting services, or request a free migration assessment.