SHADOW_IMPORT_INC issue during EHP 7 upgrade

[siva123]

Hi Everyone

I have somce serious issues upgrading EHP 7 with my present environment in the Production system. I need help from someone who has experience in EHP 7 upgrade in Windows 2012 with SQL DB, most importantly eperience with SQL in an Availability group.

When we reached the SHADOW_IMPORT_INC phase in pre-processing, I noticed the perfomrance of the Upgrade was seriously low and thats when i got to see a Snote "1945399 - performance analysis for SHADOW_IMPORT_INC and TABIM_UPG phase" which suggest us to look at few areas such as DB Performance, SQL query, File system I/O, et. when i checked with the respective teams , i got a reponse saying that everything was normal.

But there is a sugestion from a DB team to remove the database from the availability group (both from DB level from the SAP parameter level) will help improve the performance of the Pre-processing phase, as the synchronization happens at both the ends of DB it may have caused the poor performance of the Shadow Import in the preprocessing.

we did check the Package import rate/duration from the SUM/ABAP/log and found an average difference of 25% between our Quality and Production. this means, the Quality server wsa 25% faster in Package Import then the Production System. Please note that the Qualityand the Production servers have same Hardware configuration. There are 2 differences we can think of between Test and Production.

1.Quality system was in SIMPLE mode when we started the Preprocessing in Test but the Production was not.

2.Availability group was not configured for Quality but it is configured in Production

Present Environment

ECC 6 EHP 6

Windows 2012

SQL Server 2012 R2

thanks for the help in advance

regards,

ASK

[Former Member]

Hi Siva,

I have face similar issue during EHP7 upgrade and stop SUM before pre processing then stop sap again start sap with clean ipc then run the SUM tool again.I an not facing slowness issue .please try it and it will fix your issue.

Thanks

Manish

[Former Member]

Hi A SK,

I think I got the same situation as you on the Release Upgrade.

impacting both SHADOW_IMPORT and TABIM_UPG steps....

have you got a workaround for this? appreciate if you can share your experience.

System is full of free CPU and free memory and database is really not moving..

Thank you

[siva123]

HHi ferris,

can you give more detail ? pls share your environment detail such as Sap , OS , DB and hardware detail

thanks,

ASK

[Former Member]

Hi ASK,

It is a SAP ECC EHP upgrade from 3 to 7.

OS is Windows 2012 R2 and DB is MSSQL2012.

Hardware is HP blade but the production VMs are clustered (MSCS).

Two clustered VMs for SAP ASCS and two other clustered VMs for MSSQL2012.

CPU is 28 vCPU (14 core) on all four VMs, memory is 128GB on all four VMs....

I doubt it is a hardware resource constraint issue but currently there is no way out of this long running TABIM_UPG step. At different test runs, it can go down from just 3hrs. on the bad run, it can  go up to 35hrs....

Are you able to share what you experienced previously?

[siva123]

Hi,

My issue was due to DB mirroring setup. In our DB mirror setup, whenever a commit happens the data will firstly be written at the DR site then writes at the Primary Database instance. (This method of setup was to avoid data loss during a System/DBcrash). the same principles works when the SUM tool runs in to "pre-Processing". The data will be writen in the DR site first before the Primary DB instance (note: we switch the DB mode to simple only at the 'Execution' stage). This was the reason why it took so long time for each package to import during shadow phase. So we had to break the mirror setup before the Pre-processsing. Then we could see an improvement in the performance

If you have completed the Upgrade in QTY, you can compare the duration of each package import between QTY and PRD. That will help you understand how much more time PRD takes during the Pre-processing. There was a huge difference in our case between these systems and the only difference we had between QTY and PRD was DB mirror setup

EX log:

4 ETW000 SAPKW74003 touched.

4 ETW690 COMMIT "128" "769438150"

4 ETW000

24536164 bytes read.

4 ETW000 Transport overhead 0.0 %.

4 ETW000 Duration: 3:02 min (134814 bytes/sec).

3 ETW710 "1" "0"

THe below notes may be helpful to you

note #965908

1616401 - Understanding parallelism during the Upgrades, EhPs and

Support Packages implementations

1945399 - performance analysis for SHADOW_IMPORT_INC

and TABIM_UPG phase

Thanks,

ASK

[siva123]

any input on this ?

thanks,

ask

[siva123]

Hi Everyone,

nobody has experienced or encountered such things !!??

Thanks,

ASk

[Sriram2009]

Hi ASK

1. Just check the over all resource usage at OS level is this any bottle nick in memory usage?

2. During the upgrade process are you getting any error message?

3. If possible do the full system restart and then check the upgrade performance.

Regards

SS

[siva123]

Hi Sriram,

Point 1 - Memory usage was very less

Point 2 - No error

Point 3 - It is not ppossible in a Production environment. even if i want to do. can i stop the SUM tool in a middle?

it makes me think that, will i be able to stop SUM during Load Generation (SGEN) in shadow instance?

thanks,

ASK

[Sriram2009]

Hi ASK

You can wait I thing it's going to be over with slow process of SGEN

BR

SS