If you are unable to create a new account, please email support@bspsoftware.com

 

Is it possibe to identify at which record the cube got failed

Started by Sarayucog, 12 Aug 2010 08:48:03 AM

Previous topic - Next topic

Sarayucog

Hi all,

Is it possibe to identify at which record the cube got failed.

Actually my problem is I have 1000 in my database. I am running a cube if it get failed after inserting the 500 records the cube got failed. Next time when I run the cube it should insert the records from 501 to 1000 into the cube not from scratch. IS IT POSSIBLE OR NOT.

Next time when I runt the cube it should not insert the records from scratch because those 500 records are already inserted into the cube. If it run from the scratch the data will get duplicate.

If its possible please tell me how it is. Tell me the detail process.....

If you tell me this you will be a life saver.


Sarayu  ???    :'(  


twlarsen

I wouldn't put a cube into production that failed...

If the cube fails, wouldn't there be a total of 0 records, so you could just re-run the cube and it wouldn't contain duplicates.

Why not do a full rebuild of the cube?

Sarayucog

Thank you very much...

but you didn't told is it possible or not.

Here my cube is taking 22 hrs to run the cube. So I am going to implement the incremental update.

IceTea


Sarayucog

Hi,

Is it possible to identify the timestamp of last record which is stored in the cube of last success of cube.



nmcdermaid

Anything is possible..

Transformer creates lots of temp files which you could maybe sift through but this is something Cognos would have to help you with.

If your data source keeps causing your cube build to fail then you really need to consistently fix your data source before the cube builds it.

Depending on what stage of the cube build fails, there might not be a simple way to identify which record is the problem. It might be loads of records. i.e. a uniqueness violation might be caused due to records 500, 579, 999, but by the time this is recognised in the cube commit stage, the record numbers bear no resemblance to database records.

You really need to identify and fix this completely in the data source.