undefined problems on hp/ux machine

[Christian Marquardt_1]

Hello,
this morning one of our hp/ux 11.11 machines was in an undefined state. I've tried to use top (top -s2) and got the error: "top: can't allocate sufficient memory".
I've tried the oracle "opmnctl status" and got the error "Memory fault (coredump)".
I think there was a memory problem. I've tried to reboot the machine, but the shell was hanging. Then I've tocced the machine and after reboot everything works fine. Right now my shell windows are hanging temporarily and doesn't answer. Minutes later the shell came back. After reading some forum entries I've tried the swapinfo -mat and got this output:
iukweb01:/#swapinfo -mat
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4096 2042 2054 50% 0 - 1 /dev/vg00/lvol2
reserve - 2054 -2054
memory 3059 3040 19 99%
total 7155 7136 19 100% - 0 -

Can anyone help me where to look and find out the problem???

regards
Christian

[TommyT]

If you do a bdf is any file system full?

/Tommy

[Christian Marquardt_1]

Hi Tommi,
the file systems are ok!

Christian

[Pete Randall · Accepted solution]

Christian,

It sounds to me like you're swapping. That would account for your shell windows hanging temporarily.

Run "vmstat 1 5" and look at the "po" column. Numbers greater than 10 in this column indicate memory pressure (though I prefer to see nothing but zeros!). If you do have insufficient memory the only answer is to either reduce the requirements for memory by eliminating some processes or to increase the amount of memory available by purchasing some more.


Pete

[Christian Marquardt_1]

Hello Pete,
the vmstat 1 5 shows in the first line a value of 29 for the "po". In the other four lines there are nulls.
The machine has 4 GB of ram for one 9iAS Webserver.

regards
Christian

[Amit Agarwal_1]

sorry, but I am not sure if this will help or not. You can try to increase the MAXSSIZ tunable, and then run top.

Just two days back there was post in forum where the author was able to solve the memory fault problem by increasing the MAXSSIZ. I think it might help you too.

[KVK]

hi

Can u check the #ll /var/adm/crash output
Is there any directory called crash.X with latest timestamp .
If so at ur machine OS core dump has been occured call it as crash .It will happen on abnormal reboot of the system.Due to a H/W prob or OS panic or TOC . You have to contact the HP response center for analyzing that Core Dump .

[Ermin Borovac]

To find out top memory-hungry processes run

$ UNIX95= ps -e -o vsz,user,pid,args | sort -n

Also you may want to revise memory buffer cache size. By default it is set to take 50% of total RAM size, which is way too much.

$ kmtune -q dbc_max_pct

You are advised to lower dbc_max_pct to 5-10%, which should relieve memory pressure.

[CAS_2]

Run the following commands in order to show how memory is used:

/usr/contrib/Q4/bin/kmeminfo
ipcs -mob

[Christian Marquardt_1]

Hello CAS,
if I use the /usr/contrib/Q4/bin/kmeminfo in the row "Available physical memory" appears the value 1048064.
Does this mean that I can use only 1 GB of memory? The machine has 4 GB of ram installed???
I'm a little bit confused right now?!?!?!?

regards
Christian

[CAS_2]

No.

Check the output of kmeminfo command. If it contains

Physical memory usage summary (in pages):

then the values are shown in memory pages. Each page is 4 KB, therefore, your box has 4 GB.

Please, post the output of required commands.