OneLake Shortcut
A OneLake shortcut is an object in Microsoft Fabric’s OneLake that points to data stored somewhere else — another Fabric item, Azure Data Lake Storage, Amazon S3, Google Cloud Storage, and other sources. Microsoft’s documentation describes shortcuts as appearing as folders and behaving like symbolic links: Spark, SQL, and Power BI read the data through the shortcut as if it were local.
What it gives you
- No separate copy to reconcile. The authoritative data stays at the source, so there is no second copy to keep in sync by hand. Caching and metadata propagation, described below, can still affect what a reader sees at a given moment.
- Tables, not just files. A shortcut in a lakehouse’s Tables folder that points to Delta data is recognized as a table, and its schema is synchronized automatically.
- One namespace. Data from several clouds and accounts can be addressed through the same OneLake paths and permissions.
What to know before relying on one
| Behavior | Behavior to account for | Why it matters |
|---|---|---|
| Deletion | Deleting the shortcut leaves the target untouched, but deleting a file inside it deletes the file at the target if the user has write permission there | A cleanup in Fabric can remove source data |
| Identity | Internal shortcuts check the calling user’s permissions at the target; external ADLS and S3 shortcuts use a cloud connection bound when the shortcut is created. Direct Lake over SQL and delegated-identity T-SQL pass the item owner’s identity instead of the caller’s | Who is actually reading the source depends on the path, not only on the user |
| Caching | Shortcut caching, currently supported for Google Cloud Storage, S3, S3-compatible, and on-premises data gateway shortcuts, can keep external files in a workspace cache for 1 to 28 days to reduce cross-cloud egress. If the remote provider has a more recent version, the request is served from the remote provider and the cache is updated. Files over 1 GB are not cached | Can reduce repeated cross-cloud reads; cache retention is not a freshness guarantee. Check the version and freshness actually seen by consumers |
| Propagation | Schema and metadata updates apply automatically, but propagation time can vary | A schema change at the source may not appear immediately |
| Fragility | Moving, renaming, or deleting the target path can break the shortcut | The source owner can break Fabric consumers without touching Fabric |
Shortcut or copy?
A shortcut suits data that is already in an open format, owned and maintained elsewhere, and read rather than transformed in Fabric. Where data lives in an operational database rather than in files, mirroring is the replicating alternative. Either way, the source owner becomes a dependency: the shortcut removes the copy, not the need to agree on paths, formats, and permissions with whoever controls the target. How shortcuts fit into Fabric’s architecture is worked through in Microsoft Fabric Architecture: OneLake, Capacity, and the Move from Synapse.
References
Discover more from Insightful Data Lab
Subscribe to get the latest posts sent to your email.