Frequently asked questions for Cosmos DB in Fabric

Microsoft Fabric is an enterprise-ready, end-to-end data platform. Fabric unifies data movement, databases, data engineering, data science, real-time intelligence, BI with copilot, and application development. You no longer need to put together these services individually from multiple vendors.

Cosmos DB in Fabric is an AI-optimized NoSQL database, automatically configured to meet your application needs through a simplified experience. Developers can use Cosmos DB in Fabric to build AI applications with ease, without managing complex database settings. Cosmos DB in Microsoft Fabric is based on Azure Cosmos DB, which provides dynamic scaling, high availability, and reliability for the database.

Cosmos DB is a distributed NoSQL database. You can store semi-structured or unstructured data in Cosmos DB in Fabric. Cosmos DB in Fabric can be used alongside your relational data and your data in OneLake in Fabric, enabling a unified data platform for your applications.

Cosmos DB data is made available to Fabric OneLake automatically. This integration provides deep integration of Cosmos DB with the rest of Fabric platform, enabling seamless analytics, Real-time intelligence, User Data Functions (UDFs), GraphQL, Data Science and BI with Copilot, and data agents all in one place.

For more information, see Cosmos DB in Fabric.

Cosmos DB in Fabric uses the same underlying engine and infrastructure as Azure Cosmos DB, providing the same performance, reliability, and availability guarantees. However, there are key differences:

• Integration: Cosmos DB in Fabric is tightly integrated with Microsoft Fabric and OneLake, providing automatic data mirroring for analytics without ETL pipelines.
• Management: Cosmos DB in Fabric offers a simplified management experience with optimized defaults, reducing database management complexity.
• Billing: Usage is measured in Fabric capacity units (CUs) rather than Azure request units (RUs), and is billed through your Fabric capacity.
• Authentication: Cosmos DB in Fabric uses Microsoft Entra authentication exclusively, with no primary/secondary keys.
• Licensing: Requires a Power BI Premium, Fabric Capacity, or Trial Capacity license.

For more information, see Cosmos DB in Fabric overview and billing and utilization.

Yes. Cosmos DB in Fabric allows applications to store arbitrary JSON documents without schema definitions or hints. The flexible, schemaless data model is ideal for semi-structured or unstructured data and makes it easy to evolve your data model over time. Data is immediately available for query using the NoSQL query language.

For more information, see Cosmos DB in Fabric overview.

To get started with Cosmos DB in Fabric:

1. Ensure you have a Power BI Premium, Fabric Capacity, or Trial Capacity license.
2. Navigate to the Fabric portal and create a new Cosmos DB database in your workspace.
3. Create containers to store your data.
4. Connect to your database using the Cosmos DB SDKs with Microsoft Entra authentication.

For a step-by-step guide, see Quickstart: Create a Cosmos DB database in Microsoft Fabric.

Microsoft Fabric exposes an endpoint that is compatible with the Cosmos DB software development kits (SDKs). These SDKs along with the corresponding Azure Identity library should be used to connect to the database directly using Microsoft Entra authentication. For more information, see connect to Cosmos DB in Microsoft Fabric using Microsoft Entra ID.

Cosmos DB in Fabric supports the Cosmos DB SDKs, including:

• .NET SDK
• Python SDK
• Java SDK
• JavaScript/Node.js SDK
• GO SDK
• Rust SDK
• Apache Spark SDK

Use these SDKs along with the Azure Identity library for Microsoft Entra authentication. For more information, see connect to Cosmos DB in Fabric.

No, primary and secondary keys are not supported. You must use Microsoft Entra identities (user identities, service principals, or managed identities) to authenticate.

For more information, see authentication for Cosmos DB in Fabric.

Cosmos DB in Fabric primarily supports the NoSQL query language for querying data.

The NoSQL query language provides a powerful, American National Standards Institute (ANSI) Structured Query Language (SQL)-like syntax for working with JSON data. This language is designed to be familiar to users with SQL experience, while also supporting the flexibility and hierarchical nature of JSON documents.

The built-in mirroring feature for Cosmos DB in Fabric also supports the use of T-SQL to query data. Mirroring and the SQL analytics endpoint allow you to use familiar T-SQL syntax to work with your Cosmos DB data, making it easier to integrate with existing SQL-based workflows and tools.

For more information, see use the NoSQL query language.

Yes. Cosmos DB in Fabric supports aggregation via aggregate functions in the NoSQL query language, including COUNT, MAX, MIN, AVG, and SUM. These functions can be used in queries to perform analytics on your data.

For more information, see NoSQL query language documentation.

Yes. Cosmos DB in Fabric supports cross-document transactions within a single partition. Transactions are scoped to a single logical partition and executed with ACID semantics (atomicity, consistency, isolation, durability) as "all or nothing," isolated from other concurrently executing operations. If exceptions occur, the entire transaction is rolled back.

Transactions can be executed using transactional batch operations in the SDKs.

For more information, see Cosmos DB transactions.

Cosmos DB in Fabric supports optimistic concurrency control (OCC) through HTTP entity tags (ETags). Every resource has an ETag that is set on the server every time a document is updated. ETags can be used with the If-Match header to allow the server to decide whether a resource should be updated. If the ETag is no longer current, the server rejects the operation with an "HTTP 412 Precondition failure" response code, and the client must refetch the resource to acquire the current ETag value.

Most of the Cosmos DB SDKs include classes to manage optimistic concurrency control. For more information, see optimistic concurrency control in Cosmos DB.

Yes. Cosmos DB in Fabric supports cross-database queries, allowing you to query data across multiple Cosmos DB databases and even SQL databases within the same Fabric workspace. This unified query experience enables powerful analytics across your entire data estate.

For more information, see cross-database queries in Cosmos DB in Fabric.

Every Cosmos DB in Fabric database automatically mirrors data to OneLake in the Delta Parquet format. This mirroring happens in near real-time without any additional configuration or setup. The mirrored data is immediately available for analytics, data science, Power BI reporting, and other Fabric workloads.

For more information, see mirror OneLake in Cosmos DB in Fabric.

Data replication from Cosmos DB in Fabric to OneLake occurs in near real-time. Inserts, updates, and deletes are replicated with minimal latency, typically within seconds depending on the volume of changes.

For more information, see mirror OneLake in Cosmos DB in Fabric.

No. Data replication to OneLake is a core feature of Cosmos DB in Fabric and cannot be disabled. All data in your Cosmos DB containers is automatically mirrored to OneLake in the Delta Parquet format.

For more information, see mirror OneLake in Cosmos DB in Fabric.

Yes. In OneLake, Cosmos DB tables are stored as v-ordered Delta tables, which support Direct Lake mode in Power BI. This enables high-performance, low-latency reporting directly over your Cosmos DB data without data duplication.

For more information, see create reports with Cosmos DB in Fabric.

You can check the replication status by navigating to the replication section for your database in the Fabric portal. This section displays metadata about replication, including the status of the last sync and any errors that may have occurred.

For more information, see mirror OneLake in Cosmos DB in Fabric.

Request units (RUs) are a performance currency that abstracts the system resources (CPU, IOPS, and memory) required to perform database operations. All database operations, including reads, writes, queries, and updates, are measured in RUs. For example, a point read for a 1-KB item consumes one request unit.

In Cosmos DB in Fabric, request units are converted to Fabric capacity units (CUs) for billing and usage reporting purposes.

For more information, see request units in Cosmos DB in Fabric and billing and utilization.

All containers in Cosmos DB in Fabric use autoscale provisioned throughput. With autoscale, containers automatically scale throughput based on workload demands, scaling between 10% and 100% of the maximum configured throughput (RU/s). When your workload is idle, it scales down to 10% of the maximum to minimize costs. When demand increases, it scales up instantly without any warm-up period.

Containers created in the Fabric portal have a default autoscale throughput of 5,000 RU/s. This can be adjusted between 1,000 and 50,000 RU/s using the Cosmos DB SDK.

For more information, see autoscale throughput in Cosmos DB in Fabric.

No. All containers in Cosmos DB in Fabric must use autoscale provisioned throughput. Serverless and manual (standard) provisioned throughput is not supported. Containers created through the SDK must have throughput set to autoscale during container creation, or an error will be thrown stating "Offer Type is restricted to Autoscale for your account.".

For more information, see limitations for Cosmos DB in Fabric.

• Containers support a maximum autoscale throughput of 50,000 request units per second (RU/s) by default.
• Containers created in the Fabric portal are automatically allocated 5,000 RU/s maximum autoscale throughput.
• Containers created using an SDK can be set with a minimum of 1,000 RU/s up to the maximum allowed autoscale throughput.

Maximum throughput beyond 50,000 RU/s can be increased by submitting a support ticket.

For more information, see limitations for Cosmos DB in Fabric.

You can read and update autoscale throughput on a container using the Cosmos DB SDK. Use the SDK's throughput management methods to get the current throughput and replace it with a new value.

For code examples, see autoscale throughput in Cosmos DB in Fabric.

Cosmos DB in Fabric supports automatic indexing by default. All properties in your JSON documents are automatically indexed by default. Users can also define custom indexing policies to include or exclude specific paths, configure index types, and optimize for your query patterns.

Cosmos DB in Fabric supports several index types:

• Range index: Support for range queries on numeric, string, and date types.
• Spatial index: Support for geospatial queries using point, line, and polygon data types.
• Composite index: Support for optimizing queries that filter or sort on multiple properties simultaneously.
• Vector index: Support for indexing and searching vector embeddings for AI applications using DiskANN or quantized flat vector indexes.
• Full-text index: Support for full-text indexing and search on your documents with language-specific support.

For more information, see indexing in Cosmos DB in Fabric and vector indexing.

You can customize indexing policies when creating a container or update them later using the Cosmos DB SDK. Indexing policies allow you to specify which paths to include or exclude from indexing, configure index types (range, spatial, composite), and optimize query performance.

For more information, see customize indexing policies in Cosmos DB in Fabric.

Yes. Cosmos DB in Fabric supports vector indexing and search, enabling AI-powered applications with similarity search capabilities. You can store and index vector embeddings alongside your JSON documents and perform efficient vector searches using DiskANN or quantized flat vector indexes.

For more information, see vector indexing in Cosmos DB in Fabric and hybrid search.

Cosmos DB in Fabric provides several security features to help protect your data by default. These features include, but aren't limited to:

• Microsoft Entra authentication for secure access
• Data encryption at rest and in transit
• Workspace-based access control through Fabric permissions

For more information, see security for Cosmos DB in Fabric.

Cosmos DB in Fabric inherits user Fabric workspace permissions. For example, if a user has workspace viewer permissions, they have read-only access to the Cosmos DB artifact. Currently, you can set item-level permissions. However, they will be applied to all Cosmos DB artifacts within the workspace.

For more information, see limitations for Cosmos DB in Fabric.

No. Customer-managed key (CMK) encryption is not currently available for Cosmos DB in Fabric.

Private Link is not currently supported at the Cosmos DB artifact level. However, Private Links are available at the Fabric tenant level to secure connectivity to the Fabric service.

For more information, see Private links in Fabric.

No. All data in Cosmos DB in Fabric remains within your Fabric tenant and region. Data replication to OneLake occurs within the customer's environment and does not leave the tenant boundaries.

Cosmos DB in Fabric compute and storage usage is billed through your Fabric capacity using capacity units (CUs). Request units (RUs) consumed by Cosmos DB operations are automatically converted to capacity units for billing purposes. The conversion formula is: 100 RU/s = 0.067 CUs/hr.

For more information, see billing and utilization for Cosmos DB in Fabric.

You can monitor your Cosmos DB consumption using the Microsoft Fabric Capacity Metrics app. This app provides a centralized view of capacity consumption across all Fabric workloads, including Cosmos DB. You can filter the app to show only Cosmos DB-related activity and track usage trends.

For more information, see billing and utilization for Cosmos DB in Fabric and monitor Cosmos DB in Fabric.

A Power BI Premium, Fabric Capacity, or Trial Capacity is required to use Cosmos DB in Fabric. Your usage is measured against the capacity units (CUs) available in your Fabric capacity.

For more information on licensing, see Microsoft Fabric licenses.

Cosmos DB in Fabric is available in regions where Microsoft Fabric is supported. Your Cosmos DB database is located in the region of your Fabric workspace, which is based on the capacity region.

For the current list of supported regions, see Fabric regional availability.

Your Cosmos DB database is located in the region of your Fabric workspace. The workspace region is determined by the capacity assigned to it, which is displayed in Workspace settings under the License info page.

For more information, see Fabric regional availability.

Cosmos DB in Fabric databases are deployed in a single region (the region of your Fabric workspace). Multi-region deployments are not currently supported. However, the underlying infrastructure provides high availability within the region.

For more information, see limitations for Cosmos DB in Fabric.

Yes. Cosmos DB deploys all resources across availability zones, providing enhanced resilience and high availability within supported regions.

For more information, see Fabric availability zone support.

Databases support a maximum of 25 containers by default. This limit can be increased by submitting a support ticket.

For more information, see limitations for Cosmos DB in Fabric.

No. Cosmos DB stored procedures, triggers, and user-defined functions (UDFs) are not currently supported in Cosmos DB in Fabric.

For more information, see limitations for Cosmos DB in Fabric.

No. Artifact renaming is not currently supported for Cosmos DB in Fabric.

For more information, see limitations for Cosmos DB in Fabric.

Documents within Cosmos DB have a 2 MB limit.

JSON strings within a document greater than 8 KB are truncated when queried from the mirrored SQL analytics endpoint. The workaround is to create a shortcut of the mirrored database in Fabric Lakehouse or use Spark to query your data.

For more information, see limitations for Cosmos DB in Fabric.

Yes. You can use Fabric notebooks to interact with data directly in Cosmos DB in Fabric. Also, data in Cosmos DB in Fabric is automatically mirrored to OneLake in Delta Parquet format, making it accessible from Apache Spark notebooks. You can use Spark to perform analytics, data science workflows, and machine learning operations on your Cosmos DB data.

For more information, see mirror OneLake in Cosmos DB in Fabric or use Spark with Cosmos DB in Fabric or use Fabric notebooks with Cosmos DB in Fabric.

You can create Power BI reports directly over Cosmos DB in Fabric data using Direct Lake mode. The mirrored data in OneLake is stored as v-ordered Delta tables, enabling high-performance reporting without data duplication.

For more information, see create reports with Cosmos DB in Fabric.

Yes. You can implement CI/CD workflows for Cosmos DB in Fabric using Fabric's deployment pipelines and Git integration. This allows you to version control your database schemas, configurations, and deployment automation.

For more information, see continuous integration and deployment for Cosmos DB in Fabric.

Yes. You can migrate data from Azure Cosmos DB to Cosmos DB in Fabric using various methods, including:

• Azure Data Factory or Fabric Data Factory pipelines
• Azure Cosmos DB bulk import features in the SDKs
• Apache Spark for large-scale data migration
• Custom migration scripts using the SDKs
• Azure Cosmos DB Desktop Data Migration Tool

The migration process involves exporting data from Azure Cosmos DB and importing it into Cosmos DB in Fabric using compatible SDKs and connection strings with Microsoft Entra authentication.

Related content

• Learn about Cosmos DB in Fabric
• Review limitations of Cosmos DB in Fabric