Migrating from Azure Synapse Analytics to Microsoft Fabric Data Warehouse

[Shubham_rai955]

If you have been running Azure Synapse Analytics dedicated SQL pools and wondering how to move to Microsoft Fabric Data Warehouse, you are not alone. Microsoft built a migration assistant directly into Fabric to make this transition straightforward. This guide walks through the entire process, from uploading your source database to rerouting your connections.

Introduction

Microsoft Fabric represents a shift in how organizations handle analytics workloads. The Fabric Migration Assistant automates much of the heavy lifting involved in moving metadata, schemas, and data from Synapse dedicated SQL pools to Fabric Data Warehouse. The migration process handles your tables, views, stored procedures, and functions, while providing AI-powered assistance through Copilot to fix compatibility issues.

Prerequisites

Before starting, you need a DACPAC file exported from your Azure Synapse dedicated SQL pool. This file captures all the metadata about your database objects. You also need contributor or admin permissions on the Fabric workspace where you will create the new data warehouse, plus read access to your source Synapse pool.

Step by Step Migration

Step 1: Copy Metadata

Navigate to your Fabric workspace and click the Migrate button. Select the Analytical T-SQL warehouse or database tile under Migrate to a warehouse. Review the overview information and click Next.

Upload your DACPAC file when prompted. Once the upload completes, click Next. On the Set the destination page, provide a name for your new Fabric warehouse and click Next. Review your inputs and click Migrate.

The Migration Assistant now translates your T-SQL metadata to supported Fabric syntax. When complete, you will see a summary showing migrated objects and any that need fixes. Click Show migrated objects to see what migrated successfully. The State column indicates if adjustments were made during translation, while the Details column explains those changes.

Step 2: Fix Problems

Select the Fix problems step to see scripts that failed to migrate. Click any failed object to open it in the query editor. The script shows your original code plus automatic comments explaining what adjustments were attempted.

Review the error information at the beginning of each script. For AI assistance, click Fix query errors in the Suggested action section. Copilot suggests fixes, but verify them carefully before proceeding. Click Run to validate and create the object. Continue fixing remaining scripts or skip ones you do not need. When finished, check the Fix problems step as complete.

Step 3: Copy Data

Select the Copy data step and click Use a copy job button. Name your job and click Create. On the Connect to data source page, provide credentials for your source Azure Synapse Analytics warehouse.

Choose the tables you want to migrate on the Choose data page. Select your Fabric warehouse as the destination from the OneLake catalog. Configure column mappings for each table, then choose your copy mode. For migration, select one-time full data copy rather than continuous incremental copying. Review the summary and click Save + Run. Once complete, check the Copy data step.

Step 4: Reroute Connections

Identify all systems currently connected to your Synapse pool. This includes Power BI reports, ETL pipelines, and custom applications. Update each connection string to point to your new Fabric warehouse instead of the old Synapse pool.

For Power BI and Fabric pipelines, use the List Connections REST API to find existing connections, then update them through Manage Connections and Gateways in Settings. Test each connection thoroughly before cutting over production workloads. Once complete, check the Reroute connections step.

Conclusion

The Fabric Migration Assistant removes traditional barriers to migrating data warehouses. The workflow breaks into four clear phases that you can execute methodically. Plan carefully, test thoroughly, but recognize that Microsoft has made this process significantly easier than legacy migration approaches.

[sampath1601]

That's simple than i thought