Problem with memory

[alymarcor]

Hello everybody,

I have a problem with some data from memory. I don't understand why if I have 4GB memory installed when I see monitor performance tool I only see 2GB of real memory. Can somebody explain me what happen with memory and what is the meaning of memory active, virtual memory active, etc? Aditional in the event log of Oracle application appears the following message: HPUX Error: 233: No buffer space available. I atached a document with print screen of problem and Oracle log. Please help me,

Thanks,
Alymarcor

[Kent Ostby]

Here are some definitions

Real Memory: Amount of real memory in the machine thatis available to user processes and the virtual paging system
(not held by the kernel)

Active: Real memory currently being used.
Total: Total real memory in the system.

Virtual Memory:
Amount of virtual memory (also called pseudo-swap) in the system (as defined by pstat (see pstat(2)):
Active: Virtual memory currently in use.
Total: Total virtual memory in the system.

[RAC_1]

Do you have glance installed? Then post following glance -m

You may have default buffer cache set-i.e
dbc_max_pct at 50%.

kmtune -lq dbc_max_pct
kmtune -lq dbc_min_pct
kmtune -lq nbuf
kmtune -lq bufpages

No Buffer space available error can be caused by number of reasons.
Post following

swapinfo -mta
glance -m
ndd -get /dev/tcp tcp_conn_request_max

Anil

[alymarcor]

I don't have glance installed. I want to somebody tell me if the server is utilized the total of 4GB of memory or only 2GB.

[John Payne_2]

Alymarcor,

If you have Online diagnostics installed, you can go look for yourself.

for cstm:

cstm
map
sel dev nn [where nn is the device number for memory]
info
il

The information log will tell you what slots you have memory in, the sizes, total amount of memory, status, any errors, etc.

There is also an X related command, xstm. Select memory, run infomation. The report will come up.

Hope it helps

John

[A. Clay Stephenson]

You are being confused by the output of 'top'. Top is a meaningless tool for analyzing memory usage because you are reading something into it that is not there. Top only knows about memory related to processes; it reports nothing about kernel data structures such as buffer cache.

[Sathish C]

Hi
try using ipcs -mobt to see how may shared segments have also been occupied by other processes .

[Isralyn Manalac_1]

Another way to look how much physical memory you have is to check it in /var/adm/syslog/syslog.log. Search for "Physical".

Regards,

Isralyn

[alymarcor]

Thanks for your comments... But I don't understand why if I have 4GB of memory installed in the server, it uses 3.5GB of swap and only 2GB of memory active. You can see that in the document I have attached...

Thanks,
Alymarcor

[Steve Steel]

Hi


To be sure look at

ftp://eh:spear9@hprc.external.hp.com/memory.htm

It is a good site with great memory tools


Steve Steel

[Steve Steel]

Hi


This is probably an oracle issue with
memory leaks

#define ENOBUFS 233 /* No buffer space available */



What ENOBUFS means:
===================
In many cases, ENOBUFS just means there's no sufficient memory available and the system(kernel) can not allocate more


Steve Steel

[Ted Buis]

I'm with RAC. You likely have the maximum buffer cache set in the kernel at 50%. Typically this is too much. Also, you if you are using Oracle DB, then unless you have OnlineJFS you are double buffering and really making poor use of RAM. See the attached performance cookbook.