Migration Assistant for Fabric Data Warehouse: what Synapse customers should do now

If you run an Azure Synapse Analytics dedicated SQL pool, this week's FabCon news is aimed at you. In the opening announcements from FabCon 2025 in Las Vegas, Microsoft announced the preview of a migration experience built into the Fabric UI, so Synapse data warehouse customers can move to Fabric Data Warehouse with guided steps instead of a hand-built project.

That matters because many Synapse customers have been asking the same question for a while: when, and how, do we move? Until now the answer was a mix of scripts, pipelines and patience. The Migration Assistant does not remove the work, but it gives the work a shape.

Here is what it does, based on the documentation published with the preview, and what I would do this week if I had a dedicated SQL pool in production.

What the Migration Assistant does

The Migration Assistant is a guided experience inside Fabric. It copies metadata and data from the source, converts the source schema to Fabric Data Warehouse, and uses AI assistance to help with objects that do not convert cleanly.

The process has four steps:

1. Migrate object schemas into a new Fabric warehouse, using a DACPAC file extracted from your dedicated SQL pool.
2. Fix problems, by updating T-SQL types and definitions for the objects that could not be migrated automatically.
3. Copy data, using a copy job in Fabric Data Factory.
4. Test and compare the old and new warehouse, then reroute connections from your applications.

The metadata it captures from the DACPAC covers tables, views, functions, stored procedures and security objects such as roles, permissions and dynamic data masking.

You start from the Migrate button in a Fabric workspace, upload the DACPAC and name the new warehouse. The assistant creates the warehouse and translates the T-SQL. It then shows a summary of what moved, what was adjusted and what failed.

The "Fix problems" step is the interesting part

Anyone who has done a warehouse migration knows that the schema conversion is the easy part. What follows is a long tail of objects that break for small reasons.

The assistant splits failed scripts into primary objects, which do not depend on anything else, and dependent objects. You are guided to fix primary objects first, sorted by how many other objects depend on them. That is a sensible order: fix the table that many views rely on before you touch the views.

Each failed object opens as a query with comments explaining what was adjusted. You can fix it by hand, or use Copilot through "Fix query errors". Copilot must be enabled for that, and the documentation is clear that its suggestions need to be verified before you run them. I agree. Treat it as a fast first draft, not a reviewer.

Where it stops

Fabric Data Warehouse is not a copy of a dedicated SQL pool, and the documentation says there is not full T-SQL compatibility between the two. Things to plan for:

• SQL authenticated users must be replaced with Microsoft Entra users.
• Column-level encryption needs another approach, such as encryption in the application layer and dynamic data masking.
• Identity columns need a different approach to generating unique keys.
• External tables and multi-statement table-valued functions are not supported.
• Indexes and transparent data encryption are not needed in Fabric, so expect them to disappear rather than migrate.

Data types also change. The planning guide maps money to decimal(19,4), datetime to datetime2, nvarchar to varchar, nchar to char and tinyint to smallint. The one to watch is datetimeoffset: it maps to datetime2, which does not store the time zone offset, so you need to move the offset into a separate column. If you have Unicode text or time zone logic in your model, look at those columns before anything else.

What Synapse customers should do now

This is a preview, so I would not schedule a production cutover on it yet. But there is a lot you can do without risk.

1. Extract a DACPAC. Visual Studio Code, SQL Server Data Tools or SqlPackage can all do it. This is your inventory, and it costs nothing.
2. Run a trial migration into a test workspace. You need a workspace with an active or trial capacity. The assistant creates a new warehouse, so your existing items are not touched.
3. Count the failures. The summary tells you how many objects migrated and how many need fixing. That number is the most honest estimate of your migration effort you will get this early.
4. Review security separately. Fix security objects that failed before you copy any data, so users do not get unintended access to sensitive data.
5. Find your connections. The guide includes a query against sys.dm_pdw_exec_sessions that lists applications, logins and IP addresses connected to the pool. Run it now. Every row is something you will have to reroute later.
6. Copy a few tables. The copy job offers a one-time full copy, which is recommended for migration, or continuous incremental copying. Copy a handful of fact tables and compare row counts and a few aggregates.

Microsoft's planning guide also offers a useful decision: lift and shift, or modernize in phases. Lift and shift fits a small number of data marts, a well-designed star schema, or time pressure. A warehouse that has grown over many years is more likely to need redesign. Be honest about which one you have.

Takeaway

After two decades in data, I have seen few migrations fail on the schema. They fail on the details: authentication, data types, forgotten connections and untested reports. The Migration Assistant puts those details in front of you early, and that is its real value.

My advice is simple: extract a DACPAC this month, run it through the assistant in a test workspace, and let the failure list tell you how big the job is. How many of your objects do you think will make it through on the first pass?

Sources

• FabCon 2025: Fueling tomorrow's AI with new agentic capabilities and security innovations in Fabric (Microsoft Fabric blog)
• Fabric Migration Assistant for Data Warehouse (Microsoft Learn)
• Migrate with the Fabric Migration Assistant for Data Warehouse (Microsoft Learn)
• Migration planning: Azure Synapse Analytics dedicated SQL pools to Fabric Data Warehouse (Microsoft Learn)