Fabric SQL Database: Auto-Provisioning Guide (Build 2026)

Fabric SQL Database went GA at Microsoft Build 2026 and quietly filled the biggest remaining gap in the Fabric platform: an OLTP-class transactional database that lives inside a Fabric workspace, shares OneLake for analytics, and provisions in seconds without leaving the Fabric portal. This guide covers what Fabric SQL Database is, when to choose it over Azure SQL Database or Fabric Warehouse, how to provision one, and the production readiness checklist for enterprise deployments.

What Is Fabric SQL Database?

Fabric SQL Database is a fully-managed Azure SQL Database engine hosted inside a Fabric workspace. It shares the security model, the governance surface, and the capacity billing of Fabric — but it exposes a transactional SQL surface identical to Azure SQL Database. That means:

• Full T-SQL surface, including transactions, stored procedures, triggers, and views.
• Millisecond-latency single-row lookups and updates (unlike Fabric Warehouse, which is optimized for analytical scans).
• Automatic replication of operational data to OneLake as Delta Parquet, so the same tables are queryable by Fabric Warehouse, Lakehouse, and Power BI Direct Lake+ semantic models without ETL.
• Provisioning in under 60 seconds from the Fabric portal.
• Billing consumed from your Fabric capacity's CU pool, not from a separate Azure subscription.

The critical architectural choice is the automatic OneLake replication. Every table you write to Fabric SQL Database is mirrored as Delta Parquet in OneLake within seconds. That means the analytics surface — Power BI, Fabric Warehouse queries, notebooks — always sees fresh transactional data without you writing a single ETL pipeline.

Fabric SQL Database vs Fabric Warehouse vs Azure SQL Database

The three products overlap enough to confuse architects. Here is when to pick each:

Choose Fabric SQL Database when: your team is building a Fabric-first application that has both transactional and analytical needs, and you want zero ETL between operational and analytical stores. Typical scenarios: internal product catalog with sales analytics, customer support case tracking with operational KPIs, IoT device registry with real-time telemetry queries.

Choose Fabric Warehouse when: your workload is 100% analytical. No transactions, no single-row lookups, high-throughput scans over large tables.

Choose Azure SQL Database when: your team is Azure-first, your app is not deeply integrated with Fabric analytics, or you need SQL Server-specific features not yet in Fabric SQL Database (SQL Server Agent, Service Broker, cross-database references outside a Fabric workspace).

Provisioning a Fabric SQL Database

The Build 2026 GA introduced automatic provisioning that eliminates the previous multi-step preview flow. The end-to-end steps:

1. Open your Fabric workspace.
2. + New Item → SQL Database.
3. Enter a database name. That is the only required field.
4. Fabric provisions the database in approximately 45-60 seconds and returns the connection string.

The connection string is a standard SQL Server connection string. Any client that speaks TDS (SQL Server Management Studio, Azure Data Studio, .NET SqlClient, JDBC, ODBC, Python pyodbc) connects immediately. Authentication is Entra ID by default; SQL authentication is disabled unless you explicitly enable it in the database settings.

Connection string format:

``` Server=tcp:{workspace-id}.database.fabric.microsoft.com,1433; Database={database-name}; Authentication=Active Directory Default; Encrypt=True; TrustServerCertificate=False; Connection Timeout=30; ```

Note that the server hostname now includes the workspace ID, which is different from the preview URL format. If you built any tooling against the preview endpoint, update it before you migrate.

Automatic OneLake Replication

Every table you create in Fabric SQL Database is automatically replicated as Delta Parquet to a hidden `_mirrored` folder in the workspace's OneLake. Replication latency is typically 5-15 seconds from commit to Delta write. The mirrored tables are queryable from:

• Fabric Warehouse (via a warehouse in the same workspace, or via a shortcut in a different workspace).
• Fabric Lakehouse SQL endpoint.
• Power BI semantic models in Direct Lake+ mode.
• Fabric notebooks (Spark SQL, PySpark).

You cannot write to the mirrored copies — writes must go through the SQL Database engine. This preserves ACID semantics on the operational store while giving the analytics surface real-time data.

The mirroring works with any table shape, including tables with computed columns, JSON columns, XML columns, and spatial types. Some SQL types (image, text, ntext, timestamp) are excluded from mirroring; use varbinary(max), varchar(max), nvarchar(max), and rowversion instead.

Performance Characteristics

Benchmarks against Azure SQL Database General Purpose S3 on equivalent workloads:

For OLTP workloads, Fabric SQL Database performs comparably to Azure SQL Database at the equivalent price point, with the significant advantage of faster cold-start recovery from auto-pause and CU-based concurrency scaling.

Cost Model

Fabric SQL Database consumes CU from your Fabric capacity based on:

• Compute active time — the CU consumed while the database is serving queries.
• Storage — billed at approximately $0.023/GB/month via OneLake.
• Backup — 7-day point-in-time restore included; extended retention available at additional CU cost.

For a typical operational workload of 1M queries/day at moderate complexity, expect approximately 40-80 CU-hours per month. That fits in an F32 as a standalone workload, or an F64 alongside typical Power BI reporting.

Fabric SQL Database auto-pauses after configurable idle time (default: 60 minutes) and consumes zero CU while paused. First query after pause takes 8-12 seconds to warm up.

Migrating from Azure SQL Database to Fabric SQL Database

If you have an existing Azure SQL Database you want to move to Fabric:

1. Schema migration: use SqlPackage (dacpac) or SSDT to extract and deploy the schema. Fabric SQL Database supports the full Azure SQL Database T-SQL surface, so schema portability is generally 1:1.
2. Data migration: use Azure Data Factory, Fabric Data Factory pipelines, or the built-in Import Data wizard for smaller databases.
3. Application changes: update the connection string to the new `{workspace-id}.database.fabric.microsoft.com` hostname. Everything else in your app stays the same.
4. Cutover: switch traffic during a low-usage window. Rollback plan: keep the Azure SQL Database up for 30 days after cutover.
5. Retire Azure SQL Database: after 30 days of stable Fabric operation.

Not every Azure SQL Database workload should move. If you have deep SQL Server Agent job dependencies, Service Broker messaging, or SQL Server Replication topologies, those features are not yet in Fabric SQL Database and staying on Azure SQL Database is the right call.

Production Readiness Checklist

• [ ] Database backup retention configured to match your RPO.
• [ ] Point-in-time restore tested against a real backup.
• [ ] Entra ID authentication configured for both users and any service principals.
• [ ] Row-level security policies deployed and tested against representative user identities.
• [ ] Auto-pause interval set appropriately (production apps typically use 4-8 hours, dev/test can auto-pause faster).
• [ ] Fabric capacity has 30% CU headroom above measured peak load.
• [ ] Monitoring configured — export query performance metrics to Log Analytics or a Fabric warehouse.
• [ ] Downstream OneLake replication verified — confirm mirrored Delta tables appear in the expected `_mirrored` folder within 15 seconds of writes.
• [ ] Semantic model or Fabric Warehouse queries against the mirrored data confirmed working under end-user identities.
• [ ] Disaster recovery plan documented — Fabric SQL Database supports geo-restore to a different Fabric region.

Related Guides

• Fabric Data Agent API: Public Preview Guide
• Microsoft Fabric vs Power BI Premium: Break-Even Analysis
• Direct Lake+ vs Import Mode: 2026 Decision Framework
• Microsoft Fabric Consulting Services

Ready to prototype Fabric SQL Database on your own workspace, or evaluate migrating from Azure SQL Database? Book a 30-minute strategy call.