Microsoft Fabric is a unified, SaaS‑based analytics platform that consolidates data engineering, data science, real‑time analytics, warehousing, and BI into a single architecture. For organisations currently operating Azure Synapse, migration requires analysing how notebooks, pipelines, and SQL assets map into the Fabric model.
Traditionally, data platform migrations can take weeks or months, are manually intensive and prone to errors. This blog walks through the migration process and how we accelerate this using our team of experienced consultants and AI-powered tools.
The table below summarises the core challenges and transformations involved and forms the basis for the high‑level migration plan.
The Migration Challenge
Not all Synapse objects have a direct Fabric equivalent. Understanding what can be automated and what requires manual work is critical to planning your migration:
Synapse Object |
Fabric Equivalent |
Automated? |
Notes |
|---|---|---|---|
| Notebooks (.json) | .Notebook folders | ✅ Fully | Structural conversion – cell format, metadata, and code transformations. |
| Pipelines (.json) | Data Factory pipelines | ⚠️ Partial | The activity structure is similar, but the linked service and runtime references must be remapped. |
| Linked Services | Fabric Connections | ❌ Manual | No Git-based format – connections are created in the Fabric UI. |
| Datasets | N/A | ❌ N/A | Datasets don’t exist in Fabric – data access uses Lakehouses and Shortcuts. |
| Integration Runtimes | N/A | ❌ N/A | Fabric manages compute internally – no equivalent concept. |
| SQL Scripts | SQL endpoint / Warehouse | ⚠️ Partial | SQL is portable, but the execution context changes. |
| Spark Configurations | Environment items | ❌ Manual | Different configuration model. |
| Triggers | Schedules | ⚠️ Partial | Concept maps, but the format differs. |
| Dataflows | Dataflow Gen2 | ❌ Manual | Different engine (Power Query M) – must be recreated. |
| Databases | Lakehouse / Warehouse | ⚠️ Partial | Fabric Migration Assistant supports DACPAC-based schema and data migration. |
The pain points are connections and compute references. Every Synapse-linked service (Key Vault, SQL databases, storage accounts) must become a Fabric Connection created manually in the UI. Notebook code can be converted, but it won’t run until the backing connections exist.
Migration Tasks
Notebook Migration
There are no automated tools, notebooks need to be exported and imported into Fabric manually. Notebooks provide the biggest opportunity for cost and time savings because Coeo can automate the entire transformation reliably.
Pipeline Migration
Pipelines require a combination of automated transformation and manual remediation.
Tools
- FabricPipelineUpgrade PowerShell module
- Converts activities, remaps supported items, and imports into Fabric.
- Fabric Data Factory Migration Assistant (web-based)
- Profiles ARM templates, maps connectors, embeds datasets inline, deploys connections, pipelines, and schedules.
Effort
- Supported activities (such as Copy, Lookup, ForEach, etc.) migrate cleanly, but areas like linked services, Mapping Data Flows, OAuth connectors, global parameters, and triggers still require manual effort.
Dedicated SQL Pool Migration
Fabric’s built‑in Warehouse Migration Assistant automates schema translation using DACPAC import and provides AI‑assisted error correction.
Tools
- DACPAC export tools (SQLPackage / Visual Studio)
- Fabric Warehouse Migration Assistant
- Copilot for T‑SQL fixes
- Data Factory copy jobs for data movement
Schema translation is mostly automated, with manual work required for UDF rewrites, identity columns, and external tables.
Connections, Security, and Gateways
The migration still requires manual work to:
- Recreate Fabric Connections (no Git‑based format)
- Map Linked Services → Fabric
- Configure On-Premises Data Gateways VNet Gateway
- Replace SQL auth with Entra ID
These foundational components gate notebook and pipeline execution and cannot yet be automated.
Which Gateway Should You Choose?
We have a choice of Gateway types in Fabric: The On-Premises Data Gateway (OPDG) or the VNet Data Gateway. An OPDG is like a Self-Hosted Integration Runtime (SHIR) in Synapse. A VNet Data Gateway is a fully managed SaaS service that connects Azure data sources, other clouds, and on‑prem data sources to Microsoft Fabric. Fabric is a SaaS product with no customer‑managed networking layer, and the VNet Data Gateway exists to bridge network‑isolated data sources into Fabric without exposing public endpoints. Synapse integrates with VNets natively and uses Managed Private Endpoints and Self-Hosted Integration Runtimes (SHIR).
VNet Gateways have fixed hardware limits (2 cores 8GB RAM) and support 6 concurrent queries/operations, which can cause performance and scalability issues. To increase throughput, you can scale-out to additional nodes, up to a cluster maximum of 7 nodes (42 concurrent operations). Each node consumes a fixed 4 Capacity Units (CUs) per hour in your Fabric capacity, which can easily saturate smaller capacities.
We have hit these scalability limits in projects and recommended deploying On-Premises Data Gateways in these situations instead, which support running on larger hardware and up to 10 nodes per cluster. This means you can scale far beyond the current VNet Data Gateway limits. The downside is you need to install and manage the VMs yourself, but you can then scale to hundreds of concurrent queries.
The Coeo AI-Driven Approach
Our approach uses AI agents utilising MCP servers to create a predictable, repeatable path for organisations moving from Synapse to Fabric while maximising automation and minimising risk. By automating manual tasks like notebook and pipeline conversions, Coeo’s engineers can focus on what really matters: testing, code optimisation, security and empowering our clients to extract the maximum benefit from their data.
If you’d like to find out more about how we can slash your migration time by weeks and months and help you benefit from the latest Fabric features, then please contact us via our website or LinkedIn page.
You can also catch my upcoming session at FABCON in Atlanta in March: Peachy Migrations: Sweet Moves from Synapse to Fabric