Mirroring vs. Iceberg interop: using Snowflake data in Microsoft Fabric (2026 guide)

The two routes in plain terms

Mirroring keeps a copy of Snowflake tables in Fabric. Microsoft's documentation does not mark it as a preview feature. It continuously replicates Snowflake tables into OneLake as Delta/Parquet, and each mirrored database gets a read-only SQL analytics endpoint. Replication is continuous, with no schedule option (Mirroring Snowflake (opens in a new tab)). When nothing changes, polling backs off to as slow as once an hour (mirroring tutorial (opens in a new tab)).

Apache Iceberg interop is the shared-table route, and it is generally available. OneLake and Snowflake read the same Iceberg tables. Snowflake can write Iceberg tables directly into OneLake, so there is a single copy (Jan 30, 2026 release note (opens in a new tab), Feb 3, 2026 Fabric blog (opens in a new tab), Snowflake database item (opens in a new tab)).

Optional extended mirroring capabilities, with Snowflake as the first source, are in preview (Mar 19, 2026 (opens in a new tab)).

What mirrors and what doesn't

Support depends on the table type (Mirroring Snowflake (opens in a new tab)).

Where costs come from and how to keep them down

Fabric does not charge for replication compute, and storage is free up to a capacity-based limit. Snowflake still bills warehouse compute and cloud-services credits when changes are read (Mirroring Snowflake (opens in a new tab)).

Microsoft suggests reusing an existing warehouse and mirroring only the tables you need. Schema changes can trigger a full reload, so watch dbt schema changes that can force that reload (Mirroring Snowflake (opens in a new tab)).

Security and networking caveats

Security set up in Snowflake must be recreated in Fabric. Private Link is not supported, so a private network setup needs a gateway (mirroring tutorial (opens in a new tab)).

Authentication can be username/password, Entra SSO, or a key pair. Workspace identity is not supported (Mirroring Snowflake (opens in a new tab)).

Data-infrastructure practice

Manage the Snowflake connection and the mirroring configuration as code with Terraform. Promote those changes through CI/CD, with Azure DevOps Pipelines or GitHub Actions. Coordinate dbt deploys with mirroring, because a schema change can trigger a full reload (Mirroring Snowflake (opens in a new tab)). That work still sits on sound data infrastructure: Terraform, CI/CD, dbt, agile delivery, Azure DevOps, GitHub Actions, and Airflow.

How MyOrgLab helps

MyOrgLab builds governed analytics in the customer's own account, including the connection and the configuration for the route you choose.