During APD execution, the job fails with error and throws a dump. The dump can be seen in transaction ST22 or double clicking on job log itself.
This dump is seen when the work processess are busy and there are not enough system resources to handle the APD execution
Mostly the APD is processing lakhs or records at a time.
- Make sure that work processess are free while running the APD, go to SM66 and analyze the status of your BW system. If possible try to cancel the jobs that are not required
- Run the below programs to clear the temporary tables before repeating the failed APD-
Go to transaction RSANWB, search for your APD technical name, double click on name to open the APD in display mode.
Go to top menu. Choose Environment -> Clean up temporary tables
Press “Delete” button
3. Check the performance setting of the APD, uncheck the option “Execute in memory”
Press F3, F3
4. If none of the above actions work out, the only solution is to split the APD or give filters so that data is lesser at one chunk of execution
e.g. If APD is fetching data from cube, consider building a new cube with only relevant data e.g. Plant IN01
If the APD is fetching information from cubes or DSO, try to reduce the number of requests in cube/DSO.
You can build a new cube for the APD which is daily delete and full load from old cube so that it has only one request per day based on APD selection only.
Then use this new cube in the APD. In this way, APD will not fail.