Update a Characteristic in an Infocube.

[former_member713182]

Hi All:

Is there a way to update a characteristic in an Infocube (for example, due to change of a business rule) without doing a full reload of the whole cube. (It has too much data to do a deletion and a full reload.). This change affects only a certain sets of rows in the cube, for example, all records with Plant "PP01". As additional info, the infocube has aggregates (compressed/rollout).

Thanks for your advice.

Best regards.

[former_member713182]

Hi Suman & Martin:

First of all thanks a lot for your replies.

@ Suman.

What if I do the following.

- Delete aggregates.

- Do the change of data, using remodelling as Martin says.

- Build aggregates again.

Do you think it will work?.

Thanks a lot.

[RamanKorrapati]

Hi Raj,

No and never.

Remodeling is just like add/delete infoobject of char or key fig.

At cube level you can't change any data.

You need to changes by reloading from source or by  using routines.

Thanks

[Former Member]

Hi Raj,

It will work as Suman & Martin suggested. Just follow the steps as suman mentioned and drop the data from cube too rather than changing it. Also, make sure Aggregates got deleted from cube.

Thanks,

Karan Lokesh.

[former_member182470]

Hope you have not read my last reply. I have clearly explained that remodeling will not help you. it is to add,delete or replace a char/KF but not to modify the data which has already loaded.

If I am in your position, I will take some down time. I will drop all the aggregates/Cube data. I will make changes in the transformation rules as per new logic. I will reload data to Cube and rebuild the aggregates. This will be more optimistic. I will simulate this whole process in Dev and then do it in Production eventually.

[MGrob]

Agreed he can't modify the existing records  in th cube. But he doesn't need to do that  he can remodel in the way of negating the current records and insert the new records. This is the only option without deleting the whole cube as all suggested.

An other option you might want consider is loading the data from your current cube into a new cube applying your change logic in transformation to the new cube. Then you can swap the cubes in multiprovider without downtime.

Sorry if I wasn't clear.

Martin

[Former Member]

Hi Raj,

Finally how did you load the data to cube, Even I want to update historical data with enhanced infoobject in the cube?

[former_member182196]

Dear Raj,

You can add new fields with the help of remodeling.

Steps..

1. Tack a backup of cube where u wants to add a new field.

2. Stop the process Chains Which will feeding the data for cube.

3. Without deleted any data in cube add new field in cube.

4. Create a DSO with same Key Fields and  as a Cube with new field which u want to add in cube and get the full load.

please see the attached update.

regards,

Satendra

8099805591

[Sergiu]

1. Transfer data from cube_original to cube_transform with filters for rows you need to update.

2. Transfer data from cube_transform to cube_original all but with key figures values multiplied by minus one. This will reduce key figures' values to zero. Rows remain but with no impact on key figures.

3. Transfer data from cube_transform to cube_original with a new characteristic - it can be a constant or code with specific rules.

When you have a lot of daily updates for instance a sales order I use a characteristic (VBELLN+POSNR) and attributes for updates.

[Former Member]

Hi,

If data reloading is an issue, I would suggest an alternative. Create another data target and map the required fields only i.e considering Plant PP01 and load the data. Join both the data targets in a multiprovider and use for reporting.

Do this if reloading of data is a real pain, deletion of data and full reload is always recommended.

Br.,

VRYA

[former_member182470]

This is not possible, as you have to change the business rule(transformation). You have to drop all the cube data and reload again. You cannot update/modify the cube data just like that. Remodeling will not help you.

[MGrob]

Why should remodelling not work in this case?  Update the current existing data with a customer exit

Have a look here http://scn.sap.com/docs/DOC-41711

Martin

[former_member182470]

Hi Martin/Raj,

If you update through customer exit to both F-Fact anf E-Fact tables, SID inconsistency issues will be raised. It will be tedious to create SIDs for these modified records. Hope you know that Remodeling is to Add, Delete or Replace Char/KFs, but not to Modify.

Let's say he updates the data to cube through cust exit, what is he going to do with Aggregates?

To avoid all this mess, it's advisable to drop the Aggegates and Cube. make the change in the transformation rule and load all the data freshly to cube. There will not be any inconsistencies in the future. Hope you understand , Why did I say  like that.

Regards,

Suman

[MGrob]

Hi

You can go through remodelling that is the only option without rebuild the cube.. Search on scn on remodelling you will find the answer

Martin

[RamanKorrapati]

Hi Raj,

Seems like not clear what you need. pls explain with example data.

Updating single column/object not possible. but updating new records can be possible.

Am not sure. But try with self loop transformation and better to use some routine to update records where plant = PP01.  My guess may leads to duplicate.

if your cube don't have data for PP01 till now then no problem.

Thanks