Building Basic Dimensions and Cubes - Expert Cube Development with SSAS Multidimensional Models - page 59

Processing errors are unfortunately as common as deployment errors, and they are
caused by four basic types of problems:

• Changes in the underlying relational schema mean that Analysis Services

contains invalid references. For example, you might build a measure from
a column in a fact table that is subsequently renamed. You'd only find this
out if you refreshed your DSV or at processing time when Analysis Services
generated SQL that used the old name.

• Key errors : You can think of Analysis Services as performing an inner join

between tables in a snowflaked dimension, or between dimension tables
and fact tables, although it very rarely does so in the SQL it generates. For
example, if it finds a dimension key value in a fact table that doesn't exist
in the dimension table, by default, it will raise an error and the processing
will fail. As mentioned in Chapter 1 , Designing the Data Warehouse for Analysis
Services , you should try to ensure this never happens in your ETL, even if
some accidents are inevitable. You can configure processing so that key
errors are ignored, or unknown keys are assigned to a special unknown
member on a dimension by clicking on the Change Settings button in
the Process dialog to show the Change Settings dialog and going to the
Dimension Key Errors tab. However, we do not recommend you do this
except as insurance against these accidents. You'll get a lot more flexibility
with a custom solution. That said, at this stage of the cube development
process, it can be useful to ignore errors just to get the cube to process so you
can show it to your users.