Re: Processing of complex AS 2005 DBs - only reliable from BI Dev

What's non-standard in our cubes is the extended use of Fact Dimensions
needed to establish many-to-many relationships.

Various dimensions are built from fact table records, i.e. one dimension
member corresponds to one fact table record.

It's the processing of such dimensions that seems to be the problem. When
run on the development machine, everything works fine, doing the same thing
on a busier server machine results in key attribute not found errors.

When the fact dimension is processed, not all of the records are read in
some cases. For example I have a fact table containing 3,780,017 records of
which only 3,538,945 were read during processing of the key attribute of the
fact dimension (success was still reported). Later, when processing the
measure group, the key attribute errors occured.

There seems to be a timeout on processing queries that interferes here. Is
there a way to control this timeout or make it unlimited while processing?

The source database is SQL Server 2005 db.

Stan

"Vidas Matelis" wrote:

Stan,

From Management studio if you select Process Full, that should process
dimensions at first and then process all cubes.
Could you run SQL Profiler and make sure that this is happening? You would
see at the beginning statements to process all dimensions.
Then you would see something like "Processing of "x" cube has started".
After processing of cube started, you should not see any messages about
processing dimensions. Could you confirm that this is the case for your
processing?

With AMO, when processing sometimes does not work, do you just re-process
and it works? Can you use SQL Profiler same way to check.

Also, is there anything special to cubes that fail? Like parent/child
dimensions, relationship between fact table and dimension not regular,
custom rollups, distinct count measures, etc? Anything that would make your
cube not standard.

Vidas Matelis
http://www.ssas-info.com



"Stan Kondrat" <StanKondrat@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:C3BF9FB7-96D1-4487-B600-EE62A0611527@xxxxxxxxxxxxxxxx
We only do full process. From Management Studio, select the database,
choose
Process from the context menu, choose "Process Full" in Process Options
column. That should process both the dimensions and the cubes.

Do you think there could be a timing issue here?

With AMO we tried processing all dimensions first (Full Process) and then
all cubes (Full Process). Sometimes it works, sometimes it doesn't.

Stan

"Vidas Matelis" wrote:

Stan,

There is no known problem regarding this problem. Actually we also do
processing like that all the time and have not encountered any problems.
I did see similar problems, but they were allways my mistakes - example I
forget to update dimensions before starting processing cubes.

I use AMO scripts to do processing and sometimes do manuall processing
from
SQL Server Management studio.

For full reprocessing:
1. I have SSIS package that does full reprocess of all dimensions.
2. Then I process all cubes/measuregroups/partitions.

For incremental processing:
1. I process each dimension with ProcessUpdate option.
2. I reprocess all partitions that were updated with processFull option.

Could you please specify how do you do full/incremental processing?

Vidas Matelis
http://www.ssas-info.com



"Stan Kondrat" <StanKondrat@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:C3AB290C-D1C0-40DE-A359-B9847A95F415@xxxxxxxxxxxxxxxx
Hi,

I have seen various people posting similar problems, but there were no
clear
answers.

We have a quite complex AS 2005 database, with many-to-many
relationships,
parent-child hierarchies, calculation scripts, etc.

When fully processed from BI Development Studio, everything works like
a
charm. Doing the same thing (i.e. full process of the whole database)
from
SQL Server Management Studio or programmatically using AMO results in
processing errors (dimension keys not matching fact table entries -
which
is
not true).

Each time the customer updates the data (which usually only means new
records in the fact tables) the same horror happens over and over.
Ignoring
the processing errors is not an option - the affected records really
don't
make it to the cubes. The scheduled processing tasks are useless then.
The
only way to fix the customer problem is to open the affected AS 2005 db
in
BI
Development Studio and fully process it from there.

Is this a known bug? Are there any other work-arounds?

Thanks in advance for any help on this.

Stan




.

Relevant Pages
Loading