Going in Production - Expert Cube Development with SSAS Multidimensional Models - page 306

We recommend handling processing errors by setting up the ErrorConfiguration
property on objects rather than having to remember to set them every time we do
processing. Discarding rows when they contain errors is probably a bad thing to do
since it means our cube could potentially contain incorrect values; it is usually better
to assign these values to a visible UnknownMember . This way, even if we can't see
these values assigned correctly for a particular hierarchy, the totals seen at the All
Member will still be correct, and when browsing the cube, we'll be able to see where
exactly these errors are occurring. Remember, though, that this use of Unknown
Members is purely an insurance policy in case errors get past the checks we have
implemented in our ETL.

We probably don't want processing to completely fail either if an error is
encountered; it's better to let processing continue and then reprocess objects later
once we have fixed the data problems. Therefore, we should ignore the number of
errors generated, use the Report and Continue option for each type of error, and
then log these errors to a text file.

Managing processing with Integration
Services
Just as we used Integration Services to manage our partitions, we can also use it
to manage our dimension and partition processing very easily. Using Integration
Services gives us the following benefits:

• Complete control over the order in which objects are processed and the type

of processing used. It's very easy to implement even very complex processing
strategies using the Integration Services Control Flow.

• Logging Using Integration Services' own logging features (which we should

already be using in our ETL), we will be able to record exactly what gets
processed, how long each step takes, and any errors that are raised.

• If processing fails for whatever reason, Integration Services makes it very

easy to do things like send an e-mail to the cube administrator to tell them
that this has happened.

• Creation of new partitions. As we have shown, using Integration Services, we

can use a script task to automatically manage our partitions using AMO. As a
result, it makes sense to manage partition processing in the same place.