TR1901 - parent/child limit exceeded in Transformer

[doddy]

G'day,

I have a cube model that builds on the desktop OK, but fails on our production server (UNIX) with the error message in the log file ...

(TR1901) The number of children (or parents) of a category cannot exceed 65,536. (Drill) in (followed by the .mdc name)

Checking the category counts in the model with the Show Counts facility, I have two dimensions sourced from the same query with their lowest level at 36,322 categories, and 3 other dimensions with around 300 categories. I can't see how I can have 65K+ child categories anywhere if my greatest number is 36K.

Does anyone have experience with this sort of error?

Am I misunderstanding the concept of the category count in relation to parent/child?

regards,
Doddy

 

[DoubleD]

My first suggestion would be to reconsider your model design. 36K categories is usually entirely to many to effectively trend or analyze. Remember, Cubes are for multidimensional analysis, when you want detail, you should be drilling through to an IWR report.

Pain is stress leaving the body.

DoubleD

 

[doddy]

Thanks for your reply DoubleD.

I would have agreed in the past, but we have recently rearranged our data warehouse into star-schema tables, and this is the modelling that Decision Stream recommends, with the Transformer output showing each dimension fed by a single query with its surrogate_key at the lowest level of the dimension (our 6K agents correspond to the 36K keys, a new key with each change of team/function of the agent).

Fact data for the measures then has just a surrogate_key relating to each dimension and a couple of measures.

This form of modelling works great in other cubes we have. The surrogate_keys are all suppressed as the lowest level of detail required is the agent, and this is the same across the dimensions.

Would cloaking work better than suppressing? I think I read that cloaking stops the categories being written to the MDC, whereas suppressing writes them to the MDC but hides them. I've never cloaked before.

regards,
Doddy

 

[sys2674]

Level in dimension doesnot appear in cube.But i can see in categories.urgent

Hi
I got stuck in some situtation,
I have 3 data source all are flat( .txt) files.
All have a common column called Part. which i have made unique.
I want to build cube:-
I have Dimenison
Part_status
Levels Part_Status
Part
Part_Desc
When i build a cube
I can see Part_status--->Part--->Part_desc in cateogories but i cannot see this in Cube.In cube. i can just see Part_Status-->part-->

Pls let me know How to proceed with this.I am struglling for that.
Thanks
SYS2674

 

[drlex]

Sys2674,
You need to start a new thread when you have a new Question.
However, on your Q, I suspect that there is no scope to Part_Desc and so it is not displayed.
To rememdy this, make your dimension
Part_Status
Part
Part
and make the lowest level unique. Then change the label of this level to Part_Desc. Part then 'drives' the model, but has a 1:1 relationship at the lowest 2 levels.

lex

soi la, soi carré