Less memory available than installed?

[Guy Zelck_1]

We have an RP5470 hp-ux64bit 11.11 machine with 2cpus & 2GB total memory (verified!). Now when I install an Oracle db & iAS9 I run very quickly into memory shortage and the machine goes into a paging frenzy. So I have the impression I do not have 2GB at my disposal.
Using top (seems to be made by HP, so trustworthy) it shows me the following:
Memory: 921032K (778540K) real, 2075376K (1305548K) virtual, 126384K free
The 921MB used real memory + the 126K free real memory add up to a bit under 1GB. So where is my 2GB physical memory? Vmstat shows similar figures for free memory.
I searched the forum and saw a discussion about maxdsiz, I checked with SAM and I have the following:
maxdsiz=1GB (32 bit)
maxdsiz_64bit=2GB.
At 1st sight no problem for 64bit.
Is the fact I have 2 cpus limiting the memory to 1GB?
Can some clever admin clarify this?
Regards,
Guy (tiscalinet.be)

[T G Manikandan]

First check the amount of memory the system has identified using

echo phys_mem_pages/D | adb64 /stand/vmunix /dev/mem
multiply this by 4096
Also you can find it by doing a

cat /var/adm/syslog/syslog.log|grep Phy

It seems that only 1GB is being recognised.

If you have support tools Manager
use stm to diagonise
Start xstm, highlight 'Memory'. Tools->Exerciser

check for the logfiles for error messages.



Thanks

[T G Manikandan]

top does not show the real amount of memory installed on the machine.

Use the above commands to find out the amount of memory that is really configured on the machine.

Use Glanceplus to find out the memory usage.

You can also use
UNIX95= ps -e -o ruser,pid,vsz=Kbytes|more

to find out the proceses and their memory usage.

what are the values of dbc_max_pct.If it is 50% bring that down to 15%.
This makes the buffer cache to be around 300MB.
50% is 1Gb buffer cache.

Thanks

[Guy Zelck_1]

Hi T.G.,
Thanks for the valuable answers, at least I can check things out :
>echo phys_mem_pages/D | >adb64 /stand/vmunix /dev/mem
>multiply this by 4096
It shows 524.288, this x 4K = 2G.

>It seems that only 1GB is >being recognised.
So, no, it sees 2GB.

>If you have support tools >Manager
>use stm to diagonise
>Start xstm, >highlight 'Memory'. Tools->Exerciser
I did that and it ran successfully.

>top does not show the real >amount of memory installed >on the machine.
What does it show then, why doesn't it see more free memory?

>Use the above commands to >find out the amount of >memory that is really >configured on the machine.

>Use Glanceplus to find out >the memory usage.
Don't have this tool.

>You can also use
>UNIX95= ps -e -o >ruser,pid,vsz=Kbytes|more
I'll used this while installing again.

>what are the values of >dbc_max_pct.If it is 50% >bring that down to 15%.
>This makes the buffer cache >to be around 300MB.
>50% is 1Gb buffer cache.
Good point. It's 25% which is still 512MB. Is 10% too low?

Thanks for the help,
Guy.

[T G Manikandan]

The value can be brought down to 10% which is 200MB which is not very less.
This can be done when the machine is really lacking for memory.But right now,you can notice some unwanted processes using memory.

Find them using
UNIX95= ps -e -o ruser,pid,vsz=Kbytes
so that you can catch the process using huge amount of memory.
ALso make sure that mib2agt process is not using up more memory.

you can install glance from your application CD's and inspect in depth on the usage of resources.

Also check your memory related kernel parameters.Make sure that you do not have huge values for them.


Thanks

[Dietmar Konermann · Accepted solution]

To get a 1st class memory statistic:

/usr/contrib/Q4/bin/kmeminfo

With PHCO_25723 this you can do e.g.:

/usr/contrib/Q4/bin/kmeminfo -u

kmeminfo (3.42)
libp4 (7.57): Opening /stand/vmunix /dev/kmem

Loading symbols from /stand/vmunix

----------------------------------------------------------------------
Summary of user processes memory usage:

Process list sorted by resident set size ...

proc vas p_pid va_rss va_prss va_ucount command
0x044a1f040 0x0432156c0 15810 4004 3961 4504 libsel
0x042db7040 0x042573d00 3986 3910 3867
...


Or, to sort by VSS:

/usr/contrib/Q4/bin/kmeminfo -u vss
kmeminfo (3.42)
libp4 (7.57): Opening /stand/vmunix /dev/kmem

Loading symbols from /stand/vmunix

----------------------------------------------------------------------
Summary of user processes memory usage:

Process list sorted by virtual set size ...

proc vas p_pid va_rss va_prss va_ucount command
0x0425ce040 0x042721900 2132 1234 932 38292 prm3d
0x048038040 0x043247200


Regards...
Dietmar.

[Guy Zelck_1]

Thanks again Dietmar & TG,

With this I'm bound to find the cause.

Regards,
Guy.

[Jean-Louis Phelix]

Hello Guy,

Did you check the swapmem_on kernel parameter ? If it's set to 0 you won't be able to use more memory than your swap space.

Regards,

Jean-Louis.

[Guy Zelck_1]

Jean-Louis,

That is on. We have 4GB of swap space against 2GB om memory so it would have been OK even with swapmem_on=0.

[Guy Zelck_1]

All,
New information.
When I issue 'kmeminfo -V', I get these errors mentioned at the top, how serious is this?
atcoaicd:/usr/contrib# kmeminfo -V
kmeminfo (3.11)
libp4 (5.17): Opening /stand/vmunix /dev/kmem

Loading symbols from /stand/vmunix

======================================================================
Date: Mon Oct 14 09:29:56 2002

Processing pfdat table (496474 entries)...
Error: 0 sysmap pages but only 2 sysmap_32bit pages and 49637 sysmap_64bit pages
Error: found 131072 system pages but there is only 180711 P_SYS pages!!
Error: freemem is 254468 but there is only 249515 P_FREE pfdat entries!!

Thanks for your insights.
Guy.