Direct Lake fallback reasons

Fabric

PowerBI semantic models utilizing the Direct Lake storage mode read data directly from tables stored in OneLake. However, if a DAX query reaches the limitations of the Direct Lake mode, it may switch to DirectQuery mode (this behavior can be controlled, details follow). These limitations are as follows:

As of the writing of this article, the following features were unsupported:

  • Single lakehouse/warehouse
  • Composite models
  • T-SQL based views
  • Calculated columns and calculated tables
  • String length > 4000 characters
  • Complex/No structured delta table column types (Binary, GUID)
  • Relationships based on DateTime types
  • Hierarchies / Excel drillthrough
  • RLS/OLS in Warehouse
  • Direct Lake model resource limits

Switching from Direct Lake mode to DirectQuery mode results in reduced performance and, in some cases, query failure, as not all DAX features are supported in DirectQuery mode.

In the following series of articles, we will take a closer look at these limitations.

Environment:

  • MS Fabric Capacity: Trial (F64)
  • Spark runtime: 1.2 (Spark 3.4, Delta 2.4)
  • Database: AdventureWorks