A way of keeping a copy of an outside database, for example Azure SQL Database, Cosmos DB or Snowflake, continuously up to date in OneLake as Delta tables inside a mirrored database. You query it in T-SQL via its SQL analytics endpoint, and the replication compute costs nothing.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains Mirroring in context, with comparison tables and the common traps.
Terms in this definition
- Schema
The middle part of a Unity Catalog name (
catalog.schema.table), grouping tables, views, volumes, functions and models inside a catalog. A grant on it covers everything in it now and later, and nothing inside can be reached withoutUSE SCHEMA. - Azure SQL Database
Platform-as-a-service database offered as a single database or in an elastic pool, sized up to 4 TB or 128 TB on Hyperscale. SQL Agent, CLR and queries across databases aren't available.
- Cosmos DB
NoSQL database distributed globally, offering writes in multiple regions, automatic indexing and latency below 10 ms.
- OneLake
Built on Azure Data Lake Storage Gen2, it is the one logical data lake for an entire Microsoft Fabric tenant, provisioned automatically, and the place where every Fabric workload keeps its data.
- Mirrored database
What mirroring creates in Fabric: an almost real-time replica of a database held elsewhere, stored in OneLake as Delta tables. Users can query it with T-SQL through the read-only SQL analytics endpoint that comes with it.
- T-SQL
The dialect of SQL that Microsoft uses for Azure SQL and SQL Server. Azure Monitor logs are queried with KQL instead.
- SQL analytics endpoint
Read-only SQL access in Fabric to the data in a lakehouse. Power BI using DirectQuery through it performs more slowly than Direct Lake.
Related terms
- ALTER ANY EXTERNAL MIRROR
A permission that the account used for mirroring must hold in an Azure SQL Database, Azure SQL Managed Instance or SQL Server source before Fabric can mirror it. Anyone with CONTROL, or in the db_owner role, already has it.
- Azure Synapse Link for Cosmos DB
Feature for querying Cosmos DB's analytical column store without ETL and without using request units from the transactional store. New projects can no longer adopt it and should use Cosmos DB mirroring in Microsoft Fabric.
- Fabric Data Factory
Presented as the next step on from Azure Data Factory, this Fabric workload moves and reshapes data using copy jobs, mirroring, Dataflow Gen2 and pipelines.
- FTT
Failures to tolerate: how many disks or hosts can fail without losing a VM object, met by mirroring (RAID-1) or erasure coding (RAID-5/6). vSAN's default policy uses one.
- Metadata mirroring
Rather than copying data, this kind of mirroring brings only catalog, schema and table definitions into Fabric from places like Snowflake or Azure Databricks Unity Catalog. Queries then reach the source data where it sits, via OneLake shortcuts.
- NOLOCK
Using this hint on a table means no shared locks are requested, mirroring READ UNCOMMITTED; as a result a query can pick up dirty, non-repeatable or phantom data.
- Open mirroring
A mirroring approach where an application or third-party tool of any kind uploads its changes as Parquet or delimited text files, plus a _metadata.json file that identifies the keys, to the landing zone of a mirrored database in OneLake. Fabric applies those inserts, updates and deletes to Delta tables on its side.
- RAID
Grouping drives for speed or resilience, by mirroring (level 1) or parity (level 5), say. Software-defined versions come from Storage Spaces, and Storage Spaces Direct expects pass-through controllers.