SQL Server 2025: Cloud and Hybrid Integration with Azure Arc and Microsoft Fabric

The Ground-to-Cloud-to-Fabric Model

SQL Server 2025 continues Microsoft’s push toward a unified data platform by strengthening cloud and hybrid integration. Rather than forcing migrations, the focus is on managing SQL Server consistently across environments and making operational data available for analytics without heavy ETL pipelines.

Two technologies drive this shift: Azure Arc and Microsoft Fabric. Together they implement what Microsoft calls the Ground-to-Cloud-to-Fabric model — SQL Server runs wherever you need it, Arc provides centralized management and governance, and Fabric receives near-real-time analytics data through mirroring.

Azure Arc: A Control Plane for SQL Server Anywhere

Azure Arc extends Azure management and governance to SQL Server instances running outside Azure — including on-premises servers, edge locations, and other clouds. Once connected through Azure Arc, SQL Server instances can be inventoried, governed, monitored, and secured centrally without moving data.

Arc-enabled SQL Server provides the following capabilities regardless of where the instance runs:

• Inventory — see all SQL Server instances, versions, editions, and OS in one place
• Licensing management — pay-as-you-go billing, Extended Security Updates (ESU) for older versions
• Best Practices Assessment — automated review of configuration and security posture
• Azure Policy integration — apply and audit governance rules consistently
• Azure Monitor integration — centralized health and telemetry
• Microsoft Defender for SQL — threat detection and vulnerability assessment
• Automated local backups — managed backup scheduling
• Always On management — monitor and manage AG health from Azure portal

What’s New in Arc for SQL Server 2025

The most significant Arc improvement in SQL Server 2025 is full Microsoft Entra managed identity support. SQL Server instances can now authenticate to Azure services — including Azure Storage for backups, Azure Key Vault for encryption keys, and Fabric for mirroring — without storing or managing any credentials. The system-assigned managed identity is provisioned through the Azure Arc agent and used automatically.

Microsoft Fabric: Analytics Without Heavy ETL

Microsoft Fabric is Microsoft’s unified analytics platform that combines data engineering, data warehousing, real-time analytics, and Power BI under a single SaaS experience. SQL Server 2025 builds a direct integration path to Fabric through mirroring — making operational data available for analytics with minimal infrastructure and no custom pipeline code.

Data mirrored from SQL Server lands in OneLake, Fabric’s unified storage layer, as Delta Parquet files. From there it is immediately accessible to Fabric Data Warehouse, Power BI, notebooks, and AI workloads — all without the operational overhead of maintaining ETL pipelines.

Fabric Mirroring: The SQL Server 2025 Difference

Fabric Mirroring was first released in preview in March 2024 using Change Data Capture (CDC). SQL Server 2025 introduces a fundamentally better mechanism: the Change Feed. Understanding the difference matters when planning your mirroring architecture.

Fabric Mirroring Prerequisites and Constraints

• SQL Server must be Arc-enabled with the Azure Extension for SQL Server installed
• On-premises Windows only — not Azure VMs, not Linux (as of April 2026)
• Source database must use the Full Recovery Model — Simple recovery is not supported
• Tables must have a Primary Key to be eligible for mirroring — tables without a PK are excluded
• Required ports for Azure Arc must be open for outbound data movement to OneLake
• Microsoft Fabric capacity must be available in your Azure subscription
• If network connectivity is interrupted, SQL Server holds transaction log records until connectivity restores — monitor log growth during extended outages

-- Verify recovery model before enabling mirroring
SELECT name, recovery_model_desc
FROM sys.databases
WHERE name = 'YourDatabase';
-- Must show FULL — change if needed:
-- ALTER DATABASE YourDatabase SET RECOVERY FULL;

-- Find tables without a Primary Key (excluded from mirroring)
SELECT t.name AS table_name
FROM sys.tables t
WHERE NOT EXISTS
(
    SELECT 1 FROM sys.key_constraints kc
    WHERE kc.parent_object_id = t.object_id
    AND   kc.type = 'PK'
)
ORDER BY t.name;

Workshop

Goal: Register a SQL Server instance with Azure so it can be centrally managed and prepared for Fabric Mirroring.

Prerequisites:

• A running SQL Server 2025 instance (on-premises Windows)
• An Azure subscription with Arc permissions
• Local admin rights on the SQL Server host
• Outbound HTTPS access from the SQL Server host to Azure endpoints
• .NET Framework 4.7.2 or later on the host

-- Confirm version and edition on the SQL Server machine
SELECT @@VERSION;

-- Confirm the instance is not in a container
-- Containers are not supported for Arc-enabled SQL Server

# Run as Administrator on the SQL Server host
# The generated script installs the Arc agent and registers with Azure
# Script is unique to your subscription and instance — do not share it

.\ArcOnboardingScript.ps1

-- Verify managed identity is configured from SQL Server side
-- Run after Arc agent is installed
SELECT name, type_desc
FROM sys.server_principals
WHERE type_desc = 'EXTERNAL_LOGIN'
AND name LIKE '%managed%';

Goal: Mirror SQL Server 2025 data to Microsoft Fabric OneLake for near-real-time analytics without impacting OLTP workloads.

Prerequisites:

• Arc-enabled SQL Server 2025 (on-premises Windows) — completed in Part 1
• System-assigned managed identity enabled
• Source database on Full Recovery Model
• All tables to be mirrored have Primary Keys
• Microsoft Fabric workspace with available capacity

-- Confirm Full Recovery Model
ALTER DATABASE YourDatabase SET RECOVERY FULL;

-- Create a dedicated mirroring login with minimal permissions
-- sysadmin is NOT required for SQL Server 2025 Change Feed mirroring
CREATE LOGIN fabric_mirror_login WITH PASSWORD = 'StrongPasswordHere!',
    CHECK_POLICY = ON, CHECK_EXPIRATION = ON;

USE YourDatabase;
CREATE USER fabric_mirror_user FOR LOGIN fabric_mirror_login;

-- Grant only the minimum permissions needed
GRANT SELECT ON SCHEMA::dbo TO fabric_mirror_user;
GRANT VIEW DATABASE STATE TO fabric_mirror_user;
GRANT VIEW SERVER STATE TO fabric_mirror_login;

-- Monitor mirroring activity from SQL Server side (DMVs)
-- Check for active replication processes
SELECT
    session_id,
    status,
    command,
    wait_type,
    wait_time,
    LEFT(text, 200) AS query_text
FROM sys.dm_exec_requests r
OUTER APPLY sys.dm_exec_sql_text(r.sql_handle) t
WHERE command LIKE '%MIRROR%'
   OR command LIKE '%FABRIC%';

-- Monitor transaction log usage during mirroring
SELECT
    name,
    log_size_mb = size * 8.0 / 1024,
    log_used_pct = FILEPROPERTY(name, 'SpaceUsed') * 100.0 / size
FROM sys.databases
WHERE name = 'YourDatabase';

Validating Workload Isolation

One of the key benefits of this architecture is that analytics workloads run entirely in Fabric — they never query SQL Server directly. Validate this by confirming SQL Server is not serving heavy read queries while mirroring is active:

-- While analytics are running in Fabric, confirm no heavy read queries on SQL Server
SELECT
    session_id,
    status,
    command,
    wait_type,
    cpu_time,
    logical_reads,
    LEFT(t.text, 200) AS query_text
FROM sys.dm_exec_requests r
OUTER APPLY sys.dm_exec_sql_text(r.sql_handle) t
WHERE r.logical_reads > 100000  -- Flag high-read queries
ORDER BY r.logical_reads DESC;

-- If this returns no rows while Fabric reports are running,
-- your workload isolation is working correctly.

Final Thoughts

SQL Server 2025 does not force a cloud migration. It provides a realistic hybrid path that meets most organizations where they are. Azure Arc delivers centralized control and modern identity across on-premises instances. Fabric Mirroring using the new Change Feed delivers near-real-time analytics data to OneLake without the overhead of CDC change tables, without sysadmin access, and without custom ETL pipelines.

References

• Microsoft Learn – SQL Server Enabled by Azure Arc
• Microsoft Learn – Azure Arc for SQL Server Prerequisites
• Microsoft Learn – Mirroring SQL Server in Microsoft Fabric
• Microsoft Learn – Microsoft Fabric Overview
• Microsoft Learn – Azure Policy Overview
• Data Savvy – How Fabric Mirroring Transformed with SQL Server 2025
• Reitse’s Blog – Mirroring SQL Server 2025 to Microsoft Fabric: A Level Deeper