MEMORY_PGFREE_FAILED?

[former_member211576]

Hi experts,

  I got an error MEMORY_PGFREE_FAILED when I ran a customized program which consumes a lot of memory. It usually showed TSV_TNEW_PAGE_ALLOC_FAILED before. Maybe after I adjust some parameters in profile, the error message is changed now. How do I solve this problem? Someone said I should update the kernel but other said I should increase paging file size.

--- profile ---

PHYS_MEMSIZE = 36851

paging file size = 108000MB

ztta/roll_first = 1

ztta/roll_area = 3000000

rdisp/PG_MAXFS = 262144

rdisp/PG_SHM = 32768

abap/heap_area_dia = 8000000000

abap/heap_area_nondia = 16000000000

abap/heaplimit = 40000000

--- TSV_TNEW_PAGE_ALLOC_FAILED is clear which indicated abap/heap_area_dia is reached

Memory consumption

Roll.... 2718272

EM...... 2002720576

Heap.... 8000015648

Page.... 40960

MM Used. 9604547088

MM Free. 373583888

--- MEMORY_PGFREE_FAILED I don't like this message, it does not indicate which type of memory overflows.

Memory consumption

Roll.... 16112

EM...... 8379584

Heap.... 0

Page.... 32768

MM Used. 2082944

MM Free. 2104176

--- environment: Windows 2012, SQL 2012

SAP kernel....... 721

created (date)... "Sep 16 2013 16:32:38"

create on........ "NT 6.0 6002 S x86 MS VC++ 16.00"

Database version. "SQL_Server_9.00 "

Patch level. 136

Patch text.. " "

292
>>>>>	SELECT MKPF~MBLNR MKPF~MJAHR MKPF~BKTXT MKPF~BUDAT MKPF~CPUTM MKPF~USNAM
294	MKPF~USNAM MSEG~MBLNR MSEG~MATNR MSEG~WERKS MSEG~LGORT
295	MSEG~ZEILE MSEG~BWART MSEG~MENGE MSEG~UMWRK MSEG~UMLGO MSEG~LIFNR
296	MSEG~AUFNR MSEG~EBELN MSEG~SHKZG MSEG~KOSTL MSEG~SGTXT MSEG~CHARG
297
298	INTO CORRESPONDING FIELDS OF TABLE IMSEG
299	FROM ( MKPF JOIN MSEG ON MKPF~MBLNR =  MSEG~MBLNR AND
300	MKPF~MJAHR =  MSEG~MJAHR )
301	WHERE  MSEG~MBLNR IN S_DOC
302	AND MSEG~MATNR IN S_MAT
303	AND MSEG~WERKS IN S_PLT
304	AND MSEG~LGORT IN S_STL
305	AND MSEG~BWART IN S_MTY
306	AND MSEG~SOBKZ IN S_SPE
307	AND MSEG~KOSTL IN S_KOSTL
308	AND MSEG~/BEV2/ED_AEDAT IN S_CDT
309	AND MKPF~USNAM IN S_PNM .

[Former Member · Accepted solution]

well, pagefile looks quiet big enough.

One important parameter is missing: abap/heap_area_total

In theory you can have multiple wps running into heap allocation at the same time,

Are there any errors, warnings in Windows System Eventlog indicating that the server is running out of memory?

Is the problem always happening when starting the customized report or just in some cases.

regards

Peter

[former_member211576 · Accepted solution]

Hi Peter,

  Thanks for your reply.

abap/heap_area_total uses default value: 38641074176

No, there is no errors in Event viewer -> system or application log(only from SAPTCP_22, same info as SM21).

It always happens when  this specific customized report is running. I know it uses too much memory but I would like the short dump can show the memory consumption more clear. Otherwise, some developers might challenge me why don't I update the kernel or increase the paging file size.

[Former Member · Accepted solution]

if the problem is reproduceable you can easily check whether the pagefile is sufficient:

just monitor the pagefile usage (and usage peak) during the runtime of the report using Windows perfmon.

if the pagefile size is not a problem I would first go for a current kernel.

Typically you can reproduce the problem in your test system - therefore you should not have a problem switching the kernel to a current version there.

regards

Peter