on 10-13-2009 9:53 AM
Hi community,
after a kenel upgrade on a SAP ECC 5.0 we have some memory related dumps
The errors in ST22 are system_no_roll, but in the trace of the work process we see the following logs, indicating a problem on heap memory instead:
A *** ERROR => User XXXYYY reached avail. heapsize = 1907 MB: see ST22 [abstor.c 1693]
A *** ERROR => Increase abap/heap_area_total [abstor.c 1694]
So I would like to know what parameters must be changed, roll memory or heap memory? All the parameters have the right values according to SAP note SAP Note 789477
Thanks,
Valerio
Is this 32 or 64bit?
Markus
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello again,
this is the extract of the trace log relevant to the problem.
A Mon Oct 12 06:28:41 2009
A *** ERROR => User XXXYYY reached avail. heapsize = 1907 MB: see ST22 [abstor.c 1693]
A *** ERROR => Increase abap/heap_area_total [abstor.c 1694]
A *** ERROR => heap memory WP0: 0 MB [abstor.c 1698]
A *** ERROR => heap memory WP1: 0 MB [abstor.c 1698]
A *** ERROR => heap memory WP2: 0 MB [abstor.c 1698]
A *** ERROR => heap memory WP3: 1907 MB [abstor.c 1698]
A *** ERROR => heap memory WP4: 0 MB [abstor.c 1698]
...
...
A *** ERROR => heap memory WP64: 0 MB [abstor.c 1698]
A *** ERROR => heap memory WP65: 0 MB [abstor.c 1698]
A *** ERROR => heap memory WP66: 0 MB [abstor.c 1698]
A *** ERROR => heap memory WP67: 0 MB [abstor.c 1698]
A *** ERROR => heap memory WP68: 0 MB [abstor.c 1698]
A
A Mon Oct 12 06:37:24 2009
A TH VERBOSE LEVEL FULL
A ** RABAX: level LEV_RX_PXA_RELEASE_MTX entered.
...
...
A ** RABAX: level LEV_RX_MEMO_CHECK entered.
A
A Mon Oct 12 06:37:25 2009
A ** RABAX: level LEV_RX_MEMO_CHECK completed.
...
...
A ** RABAX: level LEV_RX_STDERR entered.
A Mon Oct 12 06:37:25 2009
A
A ABAP Program ZPP34_R0027.
A Source ZPP34_R0027I_REPORT Line 145.
A Error Code SYSTEM_NO_ROLL.
A Module $Id: //bas/640_REL/src/krn/runt/abtsvhsh.c#9 $ SAP.
A Function HshMemGet Line 112.
A ** RABAX: level LEV_RX_STDERR completed.
...
...
A ** RABAX: end no http/smtp
A ** RABAX: end RX_BTCHLOG|RX_VBLOG
A Unable to fulfil request for 268435456 bytes of memory space1698]
Thanks, Valerio
Hello,
Parameter EM/TOTAL_SIZE_MB is 262144 (default value)
PSALLOC variable is not set for <sid>adm
We have the dump systematically for a daily job running early in the morning (when the load on the system is very low) and sporadically during the day for dialog sessions. The trace i posted refers to the job running in the morning.
I can add that today the job failed again, but with the error TSV_TNEW_BLOCKS_NO_ROLL_MEMORY and this time in the trace log there are no reference to heap memory anymore
A ABAP Program ZPP34_R0027 .
A Source ZPP34_R0027I_REPORT Line 145.
A Error Code TSV_TNEW_BLOCKS_NO_ROLL_MEMORY.
A Module $Id: //bas/640_REL/src/krn/runt/abtsv.c#29 $ SAP.
A Function tsvPageAdminResize Line 1550.
A ** RABAX: level LEV_RX_STDERR completed.
...
...
A ** RABAX: level LEV_RX_END entered.
A ** RABAX: level LEV_RX_END completed.
A ** RABAX: end no http/smtp
A ** RABAX: end RX_BTCHLOG RX_VBLOG
A No roll storage space of length 2775056 available for internal storage. table blocks..
The parameter for roll memory are correct and (as far as we know) the system should allocate extended memory and then heap memory when roll memory is no more available, but probably we miss something. These are the values for roll memory-related parameters
ztta/roll_area 6500000
ztta/roll_extension 2000683008
ztta/roll_extension_dia 2000683008
ztta/roll_extension_nondia 2000683008
Regards,
Valerio
Hi,
The dump TSV_TNEW_BLOCKS_NO_ROLL_MEMORY ususally means that the roll
buffer has been exhausted. Please check the usage of your roll buffer
with ST02->history->page near the bottom to see ROLL BUFFER history.
If the roll buffer has been exhausted, you can increase the size with
parameter rdisp/ROLL_MAXFS, but make sure you have sufficient disk
space to handle the additional size of this buffer if it writes to disk.
try increase the following parameters:
rdisp/PG_MAXFS gradually, i.e 81920 -> 160000 or is described in a step by
step way even more highly as in the SAP note #133909.
cheers,
Vincent
Hi,
You say's Parameter EM/TOTAL_SIZE_MB is 262144 (default value), but if you look in to the note 789477
For a server with at least 12 GB physical memory
EM/TOTAL_SIZE_MB = 16384 (1024x16=16384Mb)
; allows a maximum of 16 GB extended memory and
; can be larger, if required. This can be used
; to limit the total amount of Extended Memory,
; to prevent a swap bottleneck.
So in your case what is the size of physical memory?
You are saying it is 262144? can you check is it in Mb's or Kb's? and try to set this as per your Physical , virtual memory.
Regards,
Hari.
Hello community,
thanks for your replies to my problem, we solved the problem by setting these following parameter values:
abap/heap_area_dia 2000000000
abap/heap_area_nondia 4000000000 (4GB)
abap/heap_area_total 4000000000
We launched the job and monitored its execution, it allocates progressively nearly all the heap memory now available (it reaches about 3.8 GB of heap memory) but it completes successfully.
I will continue to investigate the issue with the hints you gave me in your last posts
Thanks for your support, the question in solved
Valerio
User | Count |
---|---|
93 | |
11 | |
10 | |
9 | |
9 | |
7 | |
6 | |
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.