9. Long Running/Stuck/Performance Issues on ACT_UPG Phase?

ACT_UPG - SUM Shadow Instance Process - Applies to all kind of upgrades. A nightmare for
basis peoples. Sometimes, It will complete within a coffee break, Sometimes, It won't complete even after having breakfast next day.

This is regarding the performance improvement - memory parameters to speedup the process of ACT_UPG long running/stuck shadow instance process.

Note 1387739 - OOM during shadow upgrades vers 10

Above note shares steps to stop and start shadow instance in OS level via SAPup stopshd and SAPup startshd post increasing memory parameters on Shadow Instance Profile ( <DIR_PUT>/system/<SID>/SYS/profile). However, these kind of cases usually helps during Out of memory Errors as mentioned in note description itself or if we already decided to increase before ACT_UPG.

 Scenario 1:

✨However, There might be an scenario which we need to increase during ACT_UPG phase itself and shadow operations is in progress (Not yet failed with memory errors) since we would have missed to check this parameter prior (before starting ACT_UPG). 

 We can able to increase memory parameter without cancelling running transaction/ restarting shadow instance.

Login Shadow Instance via SAPGUI and use SE38 - RSMEMORY (dynamic parameter) Report to change the same. It will reflect in runtime itself. 

Note: Parameter change will revert automatically if Shadow Instance Restarted (which will be usually post ACT_UPG phase).

Scenario 2:

In case you missed to check but ACT_UPG process not started yet, then we can create a SUM Checkpoint before ACT_UPG phase prior/OR do this action once shadow instance created. 

Login Shadow Instance to set this memory parameter permanently - Import Shadow Instance Number Profile - Edit and Add/Change memory parameters - Restart Shadow Instance.

Now, What will be the recommended value?

Note 3517086 - ECS Non Security parameters 

As per Note 1387739, Below are the recommended values.

abap/heap_area_total = 12000000000
abap/heap_area_nondia = 10000000000
ztta/roll_extension_nondia = 6000000000
ztta/roll_area = 20000000 (the parameter is deprecated with Kernel 741)

For Rise/Private Cloud, Ensure to check note 3517086 and set memory values as per "Allowed Ranges" 

99% of memory failures (TSV_TNEW_PAGE_ALLOC_FAILED) would be due to Background job failures during ACT_UPG - Non Dialog.

Hence, No need to touch Dialog parameters - heap_area_dia or ztta/roll_extension_dia

Thanks for visiting !!!