Forum Discussion
Mirror Between Dataverse and Fabric
Dears,
I have an application in dataverse. For the time being, lets call it NamedApp
Not sure exactly yet where it is inside the dataverse.
I would like to mirror the tables which belong to that app into my Fabric Bronze Lakehouse.
I was asking chatGPT and I got this answer about my questions :
Security – managed identity or service principal?
Use an Entra ID App Registration (service principal) configured as an Application User in the AGORA Dataverse environment with read access.
This is set/configured from the Dataverse / Synapse Link side.
Is the mirror created at workspace or lakehouse level?
For Dataverse, you don’t create a “mirror” item like SQL – you configure Synapse Link in Dataverse, pointing to a Lakehouse in a Fabric workspace (WS_PRD_BRZ_INGEST).
So: you choose the workspace + lakehouse as the target, but the setup is initiated from Dataverse, not from Fabric’s Mirror menu.
Is the mirror configured in Dataverse or Fabric?
It’s configured in Dataverse (AGORA) via Synapse Link.
You select the Fabric workspace & lakehouse as the destination.
What is created – Parquet files or tables, and where?
Synapse Link writes Delta/Parquet data into the Lakehouse you selected.
In Fabric, you’ll see:
Tables under the Lakehouse’s Tables tab (e.g. account, contact, etc.)
Backed by the corresponding Delta files under /Tables/<tableName>/... in the Lakehouse’s Files.
The questions I would be delighted to have validated here are:
1) Don't we really have the option to do mirroring with dataverse from Fabric side? as we have for example with SQL Server? The best option is really the one pointed above (synapse link)?
2) The mirror is self managed? meaning, we say what we want to mirror as tables and it will be near real time with no manual intervention?
3) This will land in my bronze lakehouse in the tables tab? not in the files tab?
4) Shall I use a service principal or a managed identity for security? or shall I use dedicated technical Entra user ?
Thank you very much,
Pedro
Hi fabricpribeiro ,
Great question — and yes, Direct Link is a different technology from Synapse Link, and because of that the behavior is also very different.
Here’s the practical way to look at it:
Direct Link
It doesn’t copy data. Everything stays in Dataverse.
What you get in Fabric are shortcuts, not real Delta tables.
Because of that, Fabric needs to create a brand-new Lakehouse with the correct storage bindings.
👉 That’s why you can’t choose an existing Lakehouse.All objects land in dbo and you can't change the schema, since they’re references, not physical tables.
It’s great when you want:
No duplication
Very low latency
A simple way to query Dataverse data in Fabric
But you lose:
Control over schemas
Ingestion metadata
Medallion patterns (Bronze/Silver/Gold)
The ability to treat the data like a traditional engineering pipeline
Synapse Link (Mirroring)
This one does copy the data into your Lakehouse as Delta tables.
You can organize your Lakehouse however you want and apply a normal engineering flow.This fits much better when you want:
A true Bronze layer
Transformations later on
Lineage + ingestion metadata
Separation by schemas
Full control over the data lifecycle
So, which one makes more sense for you?
Considering that you’re already splitting your Lakehouse into structured vs. unstructured sources and you want clean schemas, governance and metadata:
👉 Mirroring (Synapse Link) aligns better with what you’re trying to build.
Direct Link is excellent — but it behaves more like a “live window” into Dataverse, not like an ingestion pipeline.
11 Replies
There was a Private Preview of Fabric Dataverse mirroring, as discussed in this post:
Real-time data warehousing with Microsoft Dataverse and Fabric zero-copy integration - Microsoft Power Platform Blog
And here (see figure 1 in middle of page):
Mirroring in Microsoft Fabric explained: benefits, use cases, and pricing demystified | Microsoft Fabric Blog | Microsoft Fabric
But I notice the private preview signup page has disappeared, and no mention on roadmap of public preview release of the feature. I can't find it now, but I'm 95% sure I saw Q3 2025 release somewhere originally.
It would be great if we could get confirmation from Microsoft, but my account rep tried to help and go nowhere.I removed the response because it was reported by someone who might be envious. Sorry, Fabricio.
Thank you very much for the reply SavioFerraz , but shall it be a service princial ? or instead a dedicated technical Entra user ?
Hi fabricpribeiro I would like to start doing a VERY IMPORTANT clarification, as of now, Fabric Dataverse Mirroring, similar to other currently available like PostgreSQL, Oracle, etc. as shown below, IS IN PRIVATE PREVIEW, that is, just a few selected companies and/or individuals can use this, so, are you one of the luky ones?
If not, what we currently have is not mirroring, but LINKS from dataverse into Fabric and there're currently two types: Synapse Link and Fabric Link; I wrote a series on this topic Dataverse Meets Fabric (Series)
about "In this series, I’ll walk you through exactly how to get Dynamics 365 (Dataverse) data into Microsoft Fabric using the two officially supported options.
We’ll explore the pros and cons, compare them head-to-head, and I’ll also include step-by-step video walkthroughs and infrastructure-as-code (Bicep) templates to help you automate your setup. "
In this context, I'm answering your questions...
#1. Don't we really have the option to do mirroring with dataverse from Fabric side? - Not mirroring (will come, but not just yet), BUT Fabric LINKS either Fabric Link or Synapse Link,
#2. The best option is really the one pointed above (synapse link)? - Depends on several factors, read my article Dataverse Meets Fabric -Comparing Dataverse Sync Options to know the pros/cons
#3. This will land in my bronze lakehouse in the tables tab? not in the files tab? - Once you establish your link, there will be tables created in the Laksehouse like this, there will be delta parquet files receiving updates from Dataverse vie the link at intervals.
4) Shall I use a service principal or a managed identity for security? or shall I use dedicated technical Entra user ? - Depends on the Link, for Fabric link has to be the workspace managed identity, for Synapse linke, it would be preferable to use a Entra ID security group with the Service Principal acting as Synapse admin (ref. Dataverse Meets Fabric - Link via Azure Synapse Analytics) 😁
So, I suggest you to read the full series, you will understand all there's is not know on Links Fabric/Synpase to bring your data from Dataverse into Fabric.
Hope this information has been useful, if so, I would appreciate a kudos or mark this as solution to your question! Best of lucks!
