If you haven’t already, check out Arun Ulag’s hero blog “Microsoft Build 2026: Building Agentic Apps with Microsoft Fabric and Microsoft Databases” for a complete look at all of our Microsoft Build announcements across our Fabric and database offerings.
__________________________________________
From replication to real-time: The next step for Mirroring users
If you are already using Mirroring in Microsoft Fabric, you know the value of having your operational data continuously replicated into Fabric as OneLake Delta tables — ready for analytics, AI, and reporting across Fabric workloads. And if you have enabled Delta Change Data Feed (CDF) through Extended Capabilities, you are already capturing row-level inserts, updates, and deletes incrementally, eliminating the need for full table reloads.
But what if you don't just want to process those changes in batch? Instead, you want to react to the changes as they happen.
That's exactly what the Mirrored Database Change Feed connector (Preview) enables. It lets you stream Delta CDF updates from your Mirrored Databases directly into Fabric Eventstreams, unlocking low-latency, event-driven applications and real-time intelligence — all without writing custom Spark jobs or building your own change-consumption pipelines.
Figure: Setup and configuration of Mirrored Database Change Feed Connector for Eventstream.
Why this matters
Many teams that adopt Mirroring follow a natural progression:
- Replicate: Core Mirroring brings operational data into OneLake continuously and for free.
- Track changes: Extended Capabilities like Delta CDF add row-level change visibility for incremental processing.
- React in real time: The Mirrored Database Change Feed connector streams those changes into Eventstreams for immediate, event-driven action.
Previously, consuming CDF changes typically meant writing Spark notebooks to poll for incremental updates. This is effective for batch analytics, but not ideal when you need very low (minutes if not seconds) latency or want to trigger actions the moment something changes. The Mirrored Database Change Feed connector removes that friction by providing a fully managed, streaming path from your mirrored change feeds into Eventstreams.
This works with all sources supported by Fabric Mirrored Databases: Azure SQL, Snowflake, Cosmos DB, Oracle, and those from Open Mirroring partners. So, wherever your operational data lives, the path to real-time intelligence is the same.
Easily discover and consume Mirrored Database Change Feeds
The connector integrates directly into the Fabric experience. You can discover your mirrored databases in the Real-Time Hub, select a database with CDF enabled, and configure an Eventstream destination — no code required.
Example scenario: A fintech company mirrors its Azure SQL Database — which powers its loan-processing application — into OneLake using Fabric Mirroring. With CDF already enabled, a data engineer opens the Real-Time Hub and, in a few clicks, connects the mirrored change feed to a new Eventstream. Within minutes, loan-status updates are streaming into Fabric — no Spark job, no custom connector, no infrastructure to manage.
Stream change events with full row-level fidelity
Once connected, the connector continuously publishes change events into your Eventstream. Each event reflects the row-level operation — insert, update, or delete — along with metadata such as change type and timestamps, preserving the full fidelity of what happened in the source database and mirrored to the OneLake Delta tables. The streaming change events mirror the structure of your source tables, so there's no need to parse low-level change logs or reason about internal Delta formats.
Example scenario: A healthcare platform mirrors its PostgreSQL patient-records database into Fabric. When a clinician updates a patient's medication or discharge status, the change event arrives in the Eventstream with the exact operation type (update), the affected columns, and a precise timestamp — giving downstream consumers the context they need to distinguish a new admission from a dosage adjustment, without querying the full table.
Process, route, and act on changes with Eventstreams
With changes flowing into Eventstreams, you can apply the full range of Eventstream processing capabilities — SQL operators, no-code transformations, filtering, and aggregation — and route outputs to multiple destinations simultaneously:
- Eventhouse for real-time dashboards and KQL-powered analytics.
- Activator to trigger alerts and automated actions when conditions are met.
- Lakehouse or other destinations for enriched, processed change data.
Example scenario: An e-commerce company mirrors its Azure Cosmos DB order database into Fabric. Using Eventstream operators, the team filters for high-value order cancellations, enriches them with customer-tier data, routes the processed stream to Eventhouse for a live revenue-impact dashboard, and configures Activator to alert the customer-success team in Teams whenever a premium customer cancels an order above a threshold.
End-to-end example: From Mirrored Azure Cosmos DB to real-time order intelligence
Let's walk through a scenario that many modern application teams will recognize.
A food-delivery startup runs its order-management system on Azure Cosmos DB. Every order placement, status change (accepted, preparing, out-for-delivery, delivered), and cancellation is written to Cosmos DB as the app's primary data store. The data team has already set up Mirrored Azure Cosmos DB to continuously replicate order data into OneLake, giving analysts access to order history in Fabric without building ETL pipelines. They've also enabled Delta CDF through Extended Capabilities so their Spark notebooks can process only incremental changes for daily reporting.
But the operations team needs more: they want a live command-center dashboard that shows order volume, average delivery times, and cancellation rates as they happen — not a report that refreshes every 30 minutes. They also want instant alerts when cancellation rates spike in a specific city, so they can investigate issues (driver shortages, restaurant delays) before they escalate.
This is how the Mirrored Database Change Feed connector makes this possible:
- Enable the connector: In the Real-Time Hub, the team selects their Mirrored Cosmos DB database (which already has CDF enabled) and creates an Eventstream with the change feed as the source.
- Process in Eventstreams: They use Eventstream operators to categorize events by order status (new, in-progress, delivered, cancelled), compute rolling metrics like cancellation rate per city over a 10-minute window, and enrich events with delivery-zone metadata.
- Route to Eventhouse: The processed stream feeds an Eventhouse table, powering a KQL real-time dashboard that displays live order volume by city, average time-to-delivery, and a cancellation heatmap — all updating within seconds.
- Trigger alerts with Activator: A parallel stream branch monitors cancellation-rate spikes. When any city exceeds a 15% cancellation rate in a rolling window, Activator sends an alert to the operations channel in Teams with the city name, current rate, and top cancellation reasons.
The result: the same mirrored Cosmos DB data that was already powering batch reporting now also drives a real-time operations dashboard and automated alerting — without duplicating infrastructure or writing custom streaming code. The team simply extended their existing Mirroring investment with a few clicks.
Get started today
To use the Mirrored Database Change Feed connector:
- Set up Mirroring: If you haven't already, create a Mirrored Database for your operational data source.
- Enable Delta CDF: Turn on Delta Change Data Feed through Extended Capabilities in your mirrored database configuration dashboard or via APIs.
- Connect to Eventstreams: Discover your CDF-enabled mirrored database in the Real-Time Hub and create an Eventstream with the change feed connector.
- Build your application: Add processing logic, route to destinations like Eventhouse, and set up Activator alerts as needed.
Explore the following resources
- Microsoft Fabric Eventstreams Overview
- Extended Capabilities in Mirroring in Microsoft Fabric: Optional Enhancements to Core Mirroring
- Real-Time Intelligence in Microsoft Fabric documentation
We’d love your feedback!
If you find this blog helpful, please give it a thumbs up!
Have ideas for what you'd like to see next? Drop us a comment or reach out to [email protected] — we’d love to hear what real-time scenarios you’re exploring and what topics you'd like us to cover in future posts.
Have new feature ideas? Submit them at Fabric Ideas - Microsoft Fabric Community
Microsoft Employee