Reprocessing the SSAS cube fails - then it succeeds if I try again

Question

So, I am fundamentally new to the BI concept, and I have inherited the existing ETL process, which is a two-step process:

• Loads data into a database that is used only when processing a cube

• Starts processing the SSAS cube with the specified database

It looks pretty well insulated, but sometimes (once a week, sometimes twice) it does not work with the following exception:

"OLAP Storage Module Errors: Attribute Key Not Found"

Now interesting is that:

• A dimension that has a problem is usually not the same thing (i.e. there is no single dimension that consistently has this failure )

• The original table , when I check it, really contains an attribute key , which, according to him, was not found

• If I immediately reprogram sizes and cubes manually through SSMS, they successfully process without incident.

Both in the aforementioned task and when it is processed through SSMS, I use "ProcessFull", so it must completely recycle them.

Does anyone encounter such a problem? I scratch my head about it ... because if it was a real data integrity problem, reprocessing the cube again will not fix it. What can happen? I was instructed to find out why this happens, but I cannot reproduce it sequentially and cannot point to the problem of data integrity as the main reason.

Thanks for any input you can provide!

Answer 1

I ran into a simulation problem and I found that data processing before cube processing worked for me.

Answer 2

You said that the package starts processing, what are the settings for processing?

I am wondering if the latest measurements are being processed, which might explain this. This will probably only happen in full if the number of transactions and parallelism are unusual. Check the settings in your manual mode in full accordance with the settings of the automated process.

Answer 3

Measurements must be processed first.

How else does a cube know how to display things?

Perhaps there was a sale of a new item that appears in the fact table, but which, for example, was not added to the product size. It will not be able to make the connection and will work.

Answer 4

I saw something similar to both AS2005 and AS2008. We get the error of measuring the error not found, despite the processed dimensions.

We load and process sizes daily. In the first month of the month, new sections are created. In the first weekend of the month, after daily processing of sizes, fact tables are loaded and sections are processed. All this is done through SSIS. Partition processing fails for several months. If we wait one day, let the daily processing of measurements be done once, the sections then process OK. Testing showed that starting ProcessUpdate in one key dimension AFTER loading the fact table allows you to process the associated partition. Note that the interaction depends on each section and depends on the fact table loaded. If we load fact table A, process the measurement, then section A will be processed. If we are late with loading fact table B, we will have to wait until dimension processing starts again before section B is processed.

It is well known that a ProcessUpdate measurement will invalidate the associated indexes and partition aggregates, so there are links between the measurement and processing of partitions. I suspect that we may see some additional undocumented interaction.