We have several cubes with non-linear aggregates in them. (DISTINCT COUNT to be specific). These cubes have partitions in each measure group. Each partition also has an aggregation design associated to it.
When you Process Full, there aren't any issues. However, when you process individual partitions, the non-linear aggregates no longer compute correctly. Sometimes they double or triple what the value should be. Sometimes they are missing
values and don't produce correct calculations. For example, the distinct count of days is currently producing 30 for August 2010 even though we have 31 days worth of data loaded into the cube.
No, there is nothing wrong with the cube definition. No, there is nothing wrong with the database. No, there is nothing wrong with the data.
We fixed this problem several months ago in 2 of our cubes based on a suggestion on this forum to process indexes on the entire cube after we finished processing the dimensions. That fixed the problem for 2 of our cubes and those cubes continue to
We just finished switching 2 other cubes to partition processing from Process Full. We added a job step to Process Indexes on the entire cube once the partitions were processed. However, this does NOT fix the problem and the non-linear aggregates
are NOT computing correctly, eve
View Complete Post