JosiahSiegel/fabric-onelake-2025

Microsoft Fabric Integration with Azure Data Factory (2025)

Overview

Microsoft Fabric is a unified SaaS analytics platform combining Power BI, Azure Synapse Analytics, and Azure Data Factory capabilities. ADF provides native connectors for Fabric Lakehouse and Fabric Warehouse, enabling seamless data movement between ADF and Fabric workspaces.

Microsoft Fabric Lakehouse Connector

The Fabric Lakehouse connector enables read and write operations to Microsoft Fabric Lakehouse for tables and files.

Supported Activities

| Activity | Supported | |----------|-----------| | Copy Activity (source and sink) | Yes | | Lookup Activity | Yes | | Get Metadata Activity | Yes | | Delete Activity | Yes |

Quick Reference

• Linked service type: Lakehouse
• Authentication: Managed Identity (preferred) or Service Principal
• Required IDs: workspaceId and artifactId
• Sink types: LakehouseTableSink (tables), LakehouseFileSink (files)
• Source type: LakehouseTableSource
• Table action options: append or overwrite

Finding Workspace and Artifact IDs:

1. Navigate to Fabric workspace in browser
2. Copy workspace ID from URL: https://app.powerbi.com/groups/<workspaceId>/...
3. Open Lakehouse settings to find artifact ID
4. Or use Fabric REST API to enumerate workspace items

For complete linked service, dataset, and copy activity JSON examples, see references/lakehouse-examples.md.

Microsoft Fabric Warehouse Connector

The Fabric Warehouse connector provides T-SQL based data warehousing capabilities within the Fabric ecosystem.

Supported Activities

| Activity | Supported | |----------|-----------| | Copy Activity (source and sink) | Yes | | Lookup Activity | Yes | | Get Metadata Activity | Yes | | Script Activity | Yes | | Stored Procedure Activity | Yes |

Quick Reference

• Linked service type: Warehouse
• Authentication: System-Assigned Managed Identity (recommended), User-Assigned MI, or Service Principal
• Required properties: endpoint, warehouse
• Sink type: WarehouseSink
• Write behaviors: insert or upsert
• Table option: autoCreate (creates table if missing)

For complete linked service, copy activity, stored procedure, and script activity JSON examples, see references/warehouse-examples.md.

OneLake Integration Patterns

ADF supports three integration patterns with OneLake:

| Pattern | Description | Key Benefit | |---------|-------------|-------------| | ADLS Gen2 Shortcuts | Reference ADLS data via OneLake shortcuts (zero-copy) | No data duplication | | Incremental Load | Watermark-based incremental copy to Lakehouse | Efficient updates | | Cross-Platform Invoke | Use InvokePipeline activity to call Fabric pipelines | Hybrid orchestration |

OneLake Shortcuts are the preferred approach when data already exists in ADLS Gen2 -- they provide instant zero-copy access without data movement. Use ADF Copy Activity only when data transformation or format conversion is needed.

For complete pipeline JSON examples for all three patterns, see references/onelake-patterns.md.

Permission Configuration

Azure Data Factory Managed Identity Permissions in Fabric

For Fabric Lakehouse:

1. Open Fabric workspace
2. Go to Workspace settings -> Manage access
3. Add ADF managed identity with Contributor role
4. Or assign Workspace Admin for full access

For Fabric Warehouse:

1. Navigate to Warehouse SQL endpoint
2. Execute SQL to create user:

CREATE USER [your-adf-name] FROM EXTERNAL PROVIDER;
ALTER ROLE db_datareader ADD MEMBER [your-adf-name];
ALTER ROLE db_datawriter ADD MEMBER [your-adf-name];

Service Principal Permissions

App Registration Setup:

1. Register app in Microsoft Entra ID
2. Create client secret (store in Key Vault)
3. Add app to Fabric workspace with Contributor role
4. For Warehouse, create SQL user as shown above

Best Practices (2025)

1. Use Managed Identity -- System-assigned for single ADF, user-assigned for multiple. Avoid service principal keys when possible. Store any secrets in Key Vault.

2. Enable Staging for Large Loads -- Use staging with compression for data volumes > 1 GB, complex transformations, or Fabric Warehouse loads.

3. Leverage OneLake Shortcuts -- Use ADLS Gen2 -> OneLake Shortcut -> Direct Access instead of ADLS Gen2 -> Copy Activity -> Lakehouse. No data movement, instant availability, reduced costs.

4. Monitor Fabric Capacity Units (CU) -- Track CU consumption per pipeline run, peak usage, and throttling. Optimize with incremental loads, off-peak scheduling, and right-sized parallelism.

5. Use Table Option AutoCreate -- Set tableOption: "autoCreate" on WarehouseSink for automatic schema management and faster development.

6. Implement Error Handling -- Configure retry policies on Copy activities and add WebActivity-based failure logging with dependencyConditions: ["Failed"].

Common Issues and Solutions

| Issue | Error Message | Solution | |-------|--------------|----------| | Permission Denied | "User does not have permission to access Fabric workspace" | Add ADF managed identity as Contributor; for Warehouse, create SQL user; allow 5 min propagation | | Endpoint Not Found | "Unable to connect to endpoint" | Verify workspaceId/artifactId; check workspace URL; ensure Lakehouse/Warehouse is not paused | | Schema Mismatch | "Column types do not match" | Use tableOption: "autoCreate" or explicit column mappings in translator | | Performance Degradation | Slow copy performance | Enable staging, increase parallelCopies (4-8), increase DIUs (8-32), check CU throttling |

Resources

• Fabric Lakehouse Connector
• Fabric Warehouse Connector
• OneLake Documentation
• Fabric Capacity Management
• ADF to Fabric Integration Guide

Progressive Disclosure References

• Lakehouse Examples: references/lakehouse-examples.md - Complete linked service, dataset, copy activity, and lookup JSON examples
• Warehouse Examples: references/warehouse-examples.md - Complete linked service, copy activity, stored procedure, and script activity JSON examples
• OneLake Patterns: references/onelake-patterns.md - Pipeline patterns for shortcuts, incremental loads, and cross-platform Invoke Pipeline