Need Help Understanding deltas

[Former Member]

We have recently had to reload our Material Movements Cube. Everything worked out well until we ran the nightly delta updates. The delta INIT for movements for example had 22,000+ records..... all the data was confirmed to be PERFECT.... then we immediately ran a delta update and there were zero records which was PERFECT.... then came the second night and the second set of delta updates and there were nearly 22,000 records and the data was definitely wrong.

When doing analysis on the delta INIT in the PSA I can see all the posting dates are from 4/3-4/5 which was our selection criteria upon setting up. When doing analysis on our delta UPDATES in the PSA I can clearly see it has documents with posting dates from 3/20-4/5... not sure why it would pick up those old dates. Also when the delta update was running we were concurrently running a FULL UPDATE of records from prior to 4/3 to capture the historical data. Our understanding is that this should not have had any affect on the delta updates so we are certainly puzzled as to why the deltas are wrong.

Any clues or suggestions?

Thanks!

[Former Member]

Hi Patrick,

Though you may have done the right thing I guess you should ideally done 2 init loads. One for prior to 4/3 and the other for 4/3-4/5. Then follow it up with the delta load for 4/3-4/5. Also check the selection are not overlapping.

Bye

Dinesh

[Former Member]

We're beginning to think it's because we deleted the setup tables... here's what we did;

1) Ran stock opening balance

2) Deleted and recreated setup tables with date range of Apr-3 to Apr-5

3) Ran delta init (Pulled all data from Apr-3 to Apr-5) Data looked good

4) Ran delta update (Pulled zero records - data still looked good)

5) Deleted and recreated setup tables with date range of PRIOR to Apr-3 (ie: No data after Apr-2)

6) Ran delta update while setup tables were still busy loading (Pulled way too many records - data looked bad)

7) Ran full update which loaded all records PRIOR to Apr-3 from setup tables - Data still looks bad

We deleted all the requests and reloaded from PSA one by one and have determined it's definitely the second Delta Update in step 6 that has the bad data.

[Former Member]

Hi Patrick,

Number of records selected in a delta depends upon the criteria for delta upload.

In a generic datasource for e.g. , this is indicated with the using a delta specific field, which could be a date, timestamp or a numeric pointer. Records for a delta upload are selected only if they have a higher value for this field, than the value in the last upload.

You can check this 'Largest value read by the delta upload' for a specific datasource in the table ROOSGENDLM,

field DELTAID.

So in your case if the extraction is using a generic datasource, you need to check if the posting date is really used to indicate the delta, or most probably some other field. In case you use a function module or infoset for the extraction, you may need to check the logic for the delta in there.

Best Regards,

Dabre