on 02-04-2014 10:33 AM
Since performing a Hardware Migrations and unicode Conversion - certain tasks (mainly customer developed - but I have spotted a SAP transaction do it as well) remain in an 'On hold' Status in SM50 .
The specifics of our setup:-
SAP ECC 6.0 - SAP Basis 700 level 22
Oracle 11.2.0.3.0
Windows 2008 64-bit R2
SAP Kernel 7.20 UC patch 500
This problem is not being replicated on the development or test boxes which are running the same versions. To me it feels like a kernel issue. I was wondering if anyone else has experienced this. I'm considering upgrading the kernel to 7.21 - but I'm unsure if I should opt go standard or the ext. The trace file of the a lot of processes contains the entries like this.
A
A Tue Feb 04 09:28:51 2014
A ***SUBPOOL*** generating subroutine pool %_T000A0 for user PMW (05b10001).
A MainProg=SAPLSKBH, Incl=LSKBHF06, Line=549
A
A Tue Feb 04 09:29:32 2014
A ***SUBPOOL*** generating subroutine pool %_T000A1 for user PMW (05b10001).
A MainProg=SAPLSKBH, Incl=LSKBHF06, Line=549
Many Thanks
Liz
Just to completeness - the issue solved. In the end I deleted the instance file on the OS and recreated adding back each of the performance parameters in groups of ones and two's - checking at each stage the problem didn't reoccur. I can't pin point a particular parameter that caused the problem as in the end the parameters all went back in as is. I then re-adjusted the EM parameters to look more sensible and reimported the profile back into RZ10.
Thanks for everyones help with this matter.
Liz
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Liz,
Screenshot provided by you shows you've 27 WP's are active on your system,and all these wp's are in PRIV mode state for a Z report at your end.Are your users assigned wp's parking into PRIV state apart from running Z report my means with SAP standard Trnxs.(Are the processes hangs for Standard Tcodes as well.)
Also i suggest you to run same Z report on your test box & put same load on it to verify the issue is with memory or with the Z report your users running.
Hope it helps.
Thanks,
Gaurav
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Guarav
The main culprits are our own customer transaction agreeded - however there are a couple of SAP transactions that exhibit the same problem.- for instance ME2L list display by vendor. You get the the memory low. leave transaction before taking a break. Again I have tested this in the TEST box and no such messages appear. The Z reports and me2l all run fine in Test without entering PRIV mode even when under load. The memory in the user context is similar size to production (as shown in SM04).
Regards
Liz
Liz Weedon wrote:
The main culprits are our own customer transaction agreeded - however there are a couple of SAP transactions that exhibit the same problem.- for instance ME2L list display by vendor.
You need to decrease ztta/roll_extension value to avoid switching of other work processes with low memory consumption to PRIV mode. In other words, you need to balance em/initial_size_MB and ztta/roll_extension values. And, of course, take abaper help to optimize code of these "own customer transaction" with high memory requirements.
Hi Liz,
Could you please share how many work processes set on your system,as the processes entering into PRIV mode means your extended memory has been exhausted.
Do one trial by reducing no of dialog work processes if more as by this leads freeing up some memory for rest and check still rest DIA WP going into PRIV mode.
Thanks,
Gaurav
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Guarav
I have dropped the number of DIA's to 15 and the problem still existed then. Do you believe I should lower it still? I will try this as soon as I can. In the meantime I will add more dialogue proceeses to the test box to see if this causes this one to go into Priv mode.
Thanks for the suggestion.
Liz
Hi Liz,
Looking at the screenshot I can see your work processes are on hold due to status PRIV mode.
1. As suggested by Gaurav it is good idea to set the parameter rdisp/wppriv_max_no so that you limit
the amount of processes that can go into priv mode! You should determine value according
to your organizations internal standard.
http://scn.sap.com/thread/3450580
Private Memory (SAP Library - SAP Memory Management (BC-CST-MM))
2. Attach one of the logs for the affected work processes i.e. dev_w25 is one of them I can see in
screenshot.
3. Also we need to focus on why the processes are going to in priv mode A work process goes
into PRIV mode if it must allocate heap memory. Check for corresponding ST22 dumps this
should list the memory related dumps do you see any?
Kind Regards,
Johan
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Johan
The work process seems only to go into PRIV mode after the processing has completed and the results are displayed. If a user leaves the browser open the process remains open and assigned to that user. When they logout properley - the dialog process is released. However, those that close the browser without logging out - the process is still locked out to them and I have to manually the process.
Any ideas?
Hi these are the memory parameters we have in place, I have been adjusting these in the hope to solve this problem so they may not be optimum.
abap/heap_area_dia = 10000000000
abap/heap_area_nondia = 0
abap/heap_area_total = 34358689792
abap/heaplimit = 40000000
abap/buffersize = 1200000
em/initial_size_MB = 32767
em/stat_log_size_MB = 20
em/stat_log_timeout = 0
rdisp/PG_SHM = 16384
rdisp/ROLL_SHM = 32768
rdisp/ROLL_MAXFS = 32768
rdisp/PG_MAXFS = 32768
rtbb/buffer_length = 200000
rtbb/max_tables = 500
sap/bufdir_entries = 2000
zcsa/db_max_buftab = 8000
zcsa/presentation_buffer_area = 19601408
zcsa/table_buffer_area = 100000512
ztta/roll_first = 1
ztta/roll_area = 3000000
ztta/roll_extension = 10000000000
ztta/parameter_area = 32000
Hi Liz,
Is your parameter rdisp/gui_auto_logout set on the system? If it is to what value?
http://help.sap.com/saphelp_nw73ehp1/helpdata/en/bb/1bcf2122fd4a76948816b1342f20d7/content.htm
For the HTTP sessions I would perhaps look at the aforementioned article you must then decide on how handle it as per your requirements.
What I would also suggest is temporarily increasing the ICM trace level logging on with particular user, kill browser and then review the relevant dev_w* file for the process and the ICM logs for more details.
Regards,
Johan
Hi
Still having problems and after much investigation this is the current situation:-
Many processes go into priv mode - this seems to be happening on both the ITS side and normal gui side.
It looks like the as soon as the user context uses above 30 Mb of memory (as shown in SM04) the process goes into PRIV mode. My understanding of memory is as follows:-
ztta/roll_first is used first this is set to 1Mb
If this is not sufficient then is hits the extended memory limited by which ever is the lower value ztta/roll_extension = (10GB) set in my system) or em/initial_size_MB = (32GB)
This should be more than enough to stop it going into PRIV mode - or am I missing something obvious.
There are no errors being logged and when the user leaves the screen the memory if freed. However the exception to this is the ITS - if the user fails to logout and closes the browser the process hangs and if eventually disconnected by the timeout parameter. The strange thing is on the test box the same about of memory is used bythe user context but the processes do not go into PRIV mode. This box has less memory and page file.
Many Thanks
Liz
Hi
abap/heap_area_dia = 10000000000
abap/heap_area_nondia = 0
abap/heap_area_total = 34358689792
abap/heaplimit = 40000000
abap/buffersize = 1200000
em/initial_size_MB = 32767
em/stat_log_size_MB = 20
em/stat_log_timeout = 0
rdisp/PG_SHM = 16384
rdisp/ROLL_SHM = 32768
rdisp/ROLL_MAXFS = 32768
rdisp/PG_MAXFS = 32768
rtbb/buffer_length = 200000
rtbb/max_tables = 500
sap/bufdir_entries = 2000
zcsa/db_max_buftab = 8000
zcsa/presentation_buffer_area = 19601408
zcsa/table_buffer_area = 100000512
ztta/roll_first = 1
ztta/roll_area = 3000000
ztta/roll_extension = 10000000000
ztta/parameter_area = 32000
These are the main ones - let me know if there are any specific ones you need.
Thanks Liz
Hi Liz,
As per SAP Note 1659422
You must apply a kernel with at least the patch level specified on the 'SP Patch Level'.So i also suggest you to update kernel.
Thanks,
Gaurav
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi,
Please refer
SAP Note 1584498 - Work processes in "On Hold" status.Also share value for rdisp/wppriv_max_no parameter.
Thanks,
Gaurav
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Liz
Kindly check the following things in your system
1. As per you screen shot all the program are custom development (Z), Enable the ST12 trace in your QAS(test) system and then check the over all performance of transaction with help of ABAPer
2. Check the memory & buffer settings as per SAP Notes
1280759 SAP recommends only 64-Bit Windows for SAP Appl. Servers
88416 Zero administration memory management for the ABAP server
Regards
Sriram
Hi Liz,
IT looks like your system needs tuning of SAP buffers. Processes are not getting enough memory for execution and hence it goes to PRIV mode.
Could you share the following
1) Server configuration in terms of CPU and RAM
2) Server type Windows or Unix which version
3) DAtabase version
4) No of Work processes configured ( DIA, BGD, UPD, SPO)
5) No of concurrent users on the system.
Regards,
Deepak Kori
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Deepak
I'm not convinced it is a memory issue mainly because the processes go into priv mode almost instantly after the results are returned onto the screen after the processing is complete.
However more details of the setup:-
32GB or RAM CPU - Intel Xeon CPU ES-2420 0 @ 1.90 Ghz (2 Processors)
Windows Server 2008 R2 -Service Pack 1
Oracle 11.2.0.3.0 64-bit
At the moment I have over configured the DIA to compensate for the processes not dying. Looking at the CPU time in SM50 none of the processes are heavily used they just don't die.
ST02 reports no memory issues for example extended memory is only 18.65% used.
Any more info let me know.
Thanks for taking the time to reply
Liz
Liz Weedon wrote:
I'm not convinced it is a memory issue mainly because the processes go into priv mode almost instantly after the results are returned onto the screen after the processing is complete.
The programs that lead to high memory consumption is your custom developments. You need to consult with your abap team for further analysis. Can you provide more information about these developments: is these WDA applications or classis ABAP reports? What type of output they produce?
User | Count |
---|---|
80 | |
24 | |
12 | |
9 | |
7 | |
6 | |
5 | |
5 | |
4 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.