cancel
Showing results for 
Search instead for 
Did you mean: 

NW2004S BI 7.0 system freezes at times

Former Member
0 Kudos

Hello All,

Our BI 7.0 system running on oracle 10G, AIX 6.0 has total 24GB RAM allocated.

I have the following allocations for SAP memories.

ztta/roll_extension 2500000000 Byte Max. extended mem. per session (external mode)

abap/heap_area_dia 2000000000 Byte Max. heap memory for dialog workprocesses

abap/heap_area_nondia 2000000000 Byte Max. heap memory for non-dialog workprocesses

abap/heap_area_total 2000000000 Byte Max. usable heap memory

abap/heaplimit 40000000 Byte Workprocess restart limit of heap memory

extended memory allocated is:

EM/TOTAL_SIZE_MB =16GB

___________________________

History of Extended memory and heap memory usage on system:

__________________________

App.Server Buffer Name

Date EM allocat EM used EM avail. Heap used Heap tot. Analy.Time

tiger_BXP_00 Extended and heap memory (MEM)

Tue 03/08/2011 1,396,736 8,015,872 16,777,216 0 1,665,663 09:32:25

Mon 03/07/2011 7,208,960 7,995,392 16,777,216 0 1,656,772 23:32:43

Sun 03/06/2011 720,896 4,923,392 16,777,216 0 1,651,905 23:32:30

Sat 03/05/2011 3,473,408 4,026,368 16,777,216 0 0 23:32:30

Fri 03/04/2011 827,392 2,187,264 16,777,216 0 0 23:32:19

Thu 03/03/2011 4,612,096 5,804,032 16,777,216 0 0 23:32:33

Wed 03/02/2011 5,373,952 8,839,168 16,777,216 0 1,953,124 23:32:22

Tue 03/01/2011 5,296,128 7,729,152 16,777,216 0 1,953,124 23:32:19

Mon 02/28/2011 5,160,960 6,508,544 16,777,216 0 1,953,124 23:32:39

Sun 02/27/2011 1,015,808 4,759,552 16,777,216 0 1,953,124 23:32:32

Sat 02/26/2011 1,945,600 3,383,296 16,777,216 0 1,953,124 23:32:31

Fri 02/25/2011 6,938,624 9,080,832 16,777,216 0 1,953,103 23:36:02

Thu 02/24/2011 6,524,928 9,080,832 16,777,216 0 1,953,103 23:35:27

Wed 02/23/2011 8,077,312 9,007,104 16,777,216 0 1,953,103 23:34:58

Tue 02/22/2011 4,337,664 6,176,768 16,777,216 0 1,953,066 23:32:55

Mon 02/21/2011 659,456 4,517,888 16,777,216 0 1,953,066 23:33:09

Sun 02/20/2011 3,391,488 3,641,344 16,777,216 0 1,953,066 23:32:37

Sat 02/19/2011 749,568 1,953,792 16,777,216 0 0 23:32:44

Fri 02/18/2011 5,849,088 8,249,344 16,777,216 0 1,953,066 23:34:01

Thu 02/17/2011 6,107,136 8,249,344 16,777,216 0 1,953,066 23:37:11

Wed 02/16/2011 4,616,192 8,249,344 16,777,216 0 1,953,066 23:34:36

Average 4,013,495 6,303,744 16,777,216 0 1,538,936

__________________________________________________________

As I am using AIX with

ES/TABLE SHM_SEGS

the memory allocation would be same for dialog and non dialog work process.

As some of the work processes are using Heap memory and almost exhausting avaialbe heap memory,

after extended memory user limit of

ztta/roll_extension 2500000000 Byte Max. extended mem. per session (external mode)

is reached,

Is it good idea to increase ztta/roll_extension to 4GB to prevent non dialog work process using up all the

avaialble heap memory?

Can someone suggest what would be good approach here?

Regards,

Venkat

Accepted Solutions (1)

Accepted Solutions (1)

Former Member
0 Kudos

Also we have observed that when we killed a user's session from SM04 and also it sassociated process

in SM50, still the user's memory is not released in SM04 for long time.

The user's are coming through BI portal.

Is there some parameters that control the memory release to free pool when a session or work process is finished

or killed?

I suspect some memory wastage due this setting, but not sure what parameters control memory release?

Any idea on this?

Former Member
0 Kudos

Users come through BI Portal are Plug-IN Http type users. In SM04 if you want to clear a users session,

highligh the users session and in the menu --> log off user session wide (many times) until the users session is completely our of the system.

Killing a work process will not clear the user session as user context is loaded in the system buffer.

Also check icm/ keep alive, rdisp/ gui auto logout and icm/idle timeout parameters according to your system. This wil clear a users session who are being idle for long time.

Regards

SM.

Answers (8)

Answers (8)

Former Member
0 Kudos

upd...

I have processes= 320

sessions = 400

My oracle directory was not full at the time of issue

and CPU utilization was >30% idle and is dynamic due to LPAR.

Former Member
0 Kudos

The last thing that comes to my mind is measuing network performance between the CI and client.

Roundtrip time

You can use the SAP tool NIPING to measure the roundtrip time:

Start the NIPING server on <target_host> using:

niping -s

Start the NIPING client on <source_host> using:

niping -c -H <target_host> -B 20 -L 10000

For example, the output is as follows:

-


times -


avg 0.181 ms

max 30.678 ms

min 0.061 ms

tr 107.963 kB/s

excluding max and min:

av2 0.178 ms

tr2 109.808 kB/s

The approximate roundtrip time is displayed in the line before the last line - in this case, it is 178 microseconds.

It is useful to repeat this test several times to familiarize yourself with the typical values and differences.

For more information about NIPING, see Note 500235.

Former Member
0 Kudos

ok, I will increase ztta/roll_extension to 4GB and swap to 32GB

The user's are unable to login using sap gui and also already logged in users could not navigate to other tcode's.

system remaining in not usable state for 10 to 15 mins, (though not all BTC and Dia are not used up.)

Former Member
0 Kudos

I'm not convinced that this might be a memory issue,but can you perform the below:

1. Check the CPU Utilization at that time, make a note of TOP CPU consuming process

2. Also check the number of oracle proceses, sessions value configured in your Database. set it upto 2ABAP+J2EE AP THREADS AND SESSIONS TO 2 PROCESSES.

3. This is unusual, but check the archive log directory at that time.

4. Check SM21 and see what are the problems reported by SAP.

Regards

SM.

Former Member
0 Kudos

update:

rdisp/wp_no_dia 30

rdisp/wp_no_btc 20

I dont see any work processes showing private in sm50 when system is freezing,

either in dpmon,

I assume BTC processes using up all the heap and freezing the system? Does non dialog show in private mode? or only dialog processes?

Former Member
0 Kudos

I have total 25GB physical memory on this BI 7.0 system.

I have only 1 server node configured and Max heap size configured is 4GB

my oracle SGA is: ~4.7GB

SQL> show sga

Total System Global Area 4714397696 bytes

Fixed Size 2090344 bytes

Variable Size 2365590168 bytes

Database Buffers 2332033024 bytes

Redo Buffers 14684160 bytes

EM is 16GB

EM/TOTAL_SIZE_MB 16384

my os is AIX 6.0

current swap allocation is: 20GB

Former Member
0 Kudos

According to this stats:

tiger_BXP_00 Extended and heap memory (MEM)

Tue 03/08/2011 1,396,736 8,015,872 16,777,216 0 1,665,663 09:32:25

Mon 03/07/2011 7,208,960 7,995,392 16,777,216 0 1,656,772 23:32:43

Sun 03/06/2011 720,896 4,923,392 16,777,216 0 1,651,905 23:32:30

Sat 03/05/2011 3,473,408 4,026,368 16,777,216 0 0 23:32:30

Fri 03/04/2011 827,392 2,187,264 16,777,216 0 0 23:32:19

Thu 03/03/2011 4,612,096 5,804,032 16,777,216 0 0 23:32:33

The maximum amount of EM ever used was around 9GIG.

Based on these stats you can increase ZTTA/roll_extenstion to 4GB and

What do you mean by system freeze, Is your Internet Explorer freezes while pulling a BI Report?

Regards

SM.

Former Member
0 Kudos

If ZTTA/ROLL_EXTENSION is set to 4G, it takes 4G from em/initial_size_mb. After this value reached then sap gets into abap/heap_area_dia which is 2g so total 6GB is allocated to per user process.

Increasing ZTTA/ROLL_EXTENSION will prevent wp getting into heap too oftenly and this would enable other users to use this memory at the same time which is not the case with heap memory.

Regards

SM.

Edited by: Sri on Mar 15, 2011 4:51 PM

Former Member
0 Kudos

Is raising: ZTTA/ROLL_EXTENSION to 3GB from current 2.5GB on my system would be helpful, so that

the work process doesn't use up all the heap?

my current parameters:

ztta/roll_extension 2500000000

abap/heap_area_dia 2000000000

abap/heap_area_nondia 2000000000

abap/heap_area_total 2000000000

My system is using up all the heap sometimes and system becoming very slow,

So I assume bumping up the ZTTA/ROLL_EXTENSION to 3GB will prevent this behaviour provided

I have enough memory.

I have total EM allocated is 16GB of which 13 to 14GB max is being used at times.

I have 21GB swap allocation, So will bump up swap to 30GB and ZTTA/ROLL_EXTENSION to 3GB

to prevent BXP hang situation.

I think these changes ar ein line with your answer.

Please let me know if this would be good idea or you have other suggestion.

Former Member
0 Kudos

I forgot, you said BI 7.0 so you must be having J2EE on this host as well, so the new question is How many J2EE server nodes did you configured? and what are their heap settings?

How many work process did you configured, my question is when you are limiting heap at 2GIG I want to know how many work process are in the state of PRIV mode.

Also

swap space = at least 2X memory so which should be >= 32 GIG.

Regards

SM.

Former Member
0 Kudos

Hi,

Total memory available is 25 GB, then 25% of this is utilized by Database by database I don't know how you have changed in your systems and the remaining would be around 18GB. You have given extended memory size as 16 GB so the available memory for the OS is just 2 GB. Also can you check the value for parameter em/initial_size_MB to what value it is set to. I also thin the value for ztta/roll_extension should be bit lower.

Regards,

Vamshi.

Former Member
0 Kudos

Hello All,

Any update to fix my issue.

Can you suggest.

Regards,

Venkat

Former Member
0 Kudos

Hello Venkat,

If you have enough free physical memory in your server, you should not have any problems raising the ztta/roll_extension in order to the user to use more EM memory.

Cheers,

Maurício

Former Member
0 Kudos

Hello Venkat,

The thing here is:

- Do you really need to set up such large selection criteria in order to fully use this 4,5gb of memory? (EM+heap).

Technically there is no problem in raising the memory quotas to 4gb or even larger values, as long as you have enough memory available for all the work process to use such quota. Remeber that when you raise the quotas, you are not raising for only one transaction that has a big memory consumption, but for all of them. After raising the quotas, you can face a situation where several big reports are running in parallel and you run out of memory available and can cause a production down situation.

You can raise if you size the solution in order to have enough free memory available to your system to run fine, but always do this with caution. Check on ST06 if you have enough physical resources and also take a look on ST02 to be sure if you have enough free memory.

Cheers,

Maurício

Former Member
0 Kudos

Maurício, Thanks for your suggestion.

Per my understanding 4,5gb(heap+extended) of max memory which is allotted currently per user, is not not sufficient on the system, which is indiacted by all the heap memory being used up by some users.

Do you suggest bumping up either of the below parameters for our situation, to prevent all the heap memory

being used up?

Parameter: current value: change to:

ztta/roll_extension 2500000000 3GB

or

abap/heap_area_dia 2000000000 3GB

abap/heap_area_nondia 2000000000 3GB

abap/heap_area_total 2000000000 3GB

Else do you suggest a different solution for all the heap memory being used up currently?

Can you please suggest.

Regards,

Venkat