topic Re: swapinfo information in Operating System - HP-UX

swapinfo information

mehul_3 — Mon, 05 Dec 2005 01:24:56 GMT

Hi all
Following result is shown after executing swapinfo
# swapinfo -atm
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 6144 277 5867 5% 0 - 1 /dev/vg00/lvol2
reserve - 5867 -5867
memory 2336 2209 127 95%
total 8480 8353 127 99% - 0 -

Could anybody tell what does reserve column indicate? and 99% utilization leads to a problem "Out of memeory" while application is running.
Note: if physical memory is 3GB then in above swapinfo what does memory column indicate?

Rgds,
Mehul

Re: swapinfo information

Joseph Loo — Mon, 05 Dec 2005 01:35:22 GMT

hi,

From "man swapinfo", the reserve column means the paging space on reserve which is needed by processes that r currently running, but that has not yet been allocated from one of the paging areas.

the memory seems quite low, do a vmstat to verify how bad swapping is occurring:

# vmstat 2 15
15 2-seconds interval

regards.

Re: swapinfo information

Steve Steel — Mon, 05 Dec 2005 02:50:30 GMT

Hi

Best explanation I know is on

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

I would advise anybody to bookmark this site

Steve Steel

Re: swapinfo information

Geoff Wild — Mon, 05 Dec 2005 02:55:04 GMT

What is dbc_max_pct set to? You may be able to free up some ram for your applications...

With 3GB of ram, dbc_max_pct of 20 or 25% should be fine.

Post a vmstat -S 2 10

-S Report the number of processes swapped in and out (si and so)

You may need to add more memory to the system - or add more swap space...

Rgds...Geoff

Re: swapinfo information

Thomas J. Harrold — Tue, 06 Dec 2005 09:06:32 GMT

Mehul,

The "memory" row indicates that you have pseudo-swap or "lazy" swap mode enabled. (default on HP-UX). Up to 75% of physical memory can be dedicated to use as paging space. This is key on large memory systems. It's tough to set aside 2x physical memory (traditional recommendation) for paging space on a system with 128GB of RAM!

You can disable this with a kernel parameter change, but unless your app routinely chews up a ton of memory, I'd leave it set at the default.

-tjh

Re: swapinfo information

Ted Buis — Tue, 13 Dec 2005 21:51:40 GMT

Still if reducing buffer cache doesn't work, then you need more virtual address space on your system. You can buy more RAM, which is your best option, or you can create more device swap on a disk, or both. You are currently using 99% of your virtual address space of 8480MBbyes. Reducing buffer cache can't give you that much more space. Once you reach 100% on that bottom line of swapinfo -tam, you can no longer exec any new processes, so in won't be just poor performance, but processes not running or even dying.

Re: swapinfo information

mehul_3 — Tue, 13 Dec 2005 22:36:28 GMT

Thanks for explanation.
We are still facing "Out of memory" problem from last 10 days.Either we have to abort the instance or kill user session which is not ideal solution.
Pl. suggest everlasting solution?

Waiting for immediate response.

Regards,
Mehul

Re: swapinfo information

Patrick Wallek — Tue, 13 Dec 2005 23:31:08 GMT

If you still see swap usage and vmstat shows a value consistently greater than 0 for the po column, then you choices are:

1) Reduce the memory usage on the system

2) Upgrade RAM

Re: swapinfo information

RAC_1 — Tue, 13 Dec 2005 23:33:35 GMT

What is the OS version?? What is oracle version??
How many instances on this system??

what are the following kernel settings??
maxssiz
maxssiz_64bit
maxdsiz
maxdsiz_64bit
maxtsiz
maxtsiz_64bit
shmmax

Re: swapinfo information

Yogeeraj_1 — Wed, 14 Dec 2005 00:05:46 GMT

hi Mehul,

Are you running an Oracle Database?

maybe you can also perform some tuning at this level also..

kind regards
yogeeraj

Re: swapinfo information

mehul_3 — Wed, 14 Dec 2005 02:40:35 GMT

OS version ::HP-UX11.11
Oracle version :: 9.2.0.1.0
How many instances on this system :: One Instance

Following are kernel settings
maxssiz 134217728
maxssiz_64bit 1073741824
maxdsiz 1073741824
maxdsiz_64bit 2147483648
maxtsiz 67108864
maxtsiz_64bit 1073741824
shmmax 2147483648

We have increased shmmax from 536870912 to 2147483648 to fit entire SGA and(Which is 1500MB,PGA 600MB) but problem still persists.

Note:Physical RAM is 3GB.
Concurrent user 25

Regards,
Mehul

Re: swapinfo information

RAC_1 — Wed, 14 Dec 2005 02:46:26 GMT

You did not provide this.

What is oracle version??
How many instances on this system??

Re: swapinfo information

mehul_3 — Wed, 14 Dec 2005 03:56:00 GMT

already provided pl.check it.

Regards,
Mehul

Re: swapinfo information

RAC_1 — Wed, 14 Dec 2005 04:21:28 GMT

Get shminfo tool and check shared memory usage. I also inclined towards reducing the SGA. try it.

Re: swapinfo information

Yogeeraj_1 — Wed, 14 Dec 2005 06:45:16 GMT

hi again,

can you post the output of "show sga"

login to the database as '/ as sysdba'

regards
yogeeraj

Re: swapinfo information

Rita C Workman — Wed, 14 Dec 2005 10:39:02 GMT

Set up some disk for swap and set it up so it will be there when you reboot too. That will keep you from totally choking out for the moment. I notice you only have lvol2.

[Memory total is the total of all available and would include the lvol2 disk]

3Gb is not alot of memory running anything Oracle over 8i. You need to look at increasing this, cause the newer versions of Oracle just keep putting more of a strain on memory, so you also need to look at all your kernel parms.
Post all your parms...and let folks look them over, there may be a few others that could be suggested for changes to help.

Just a thought,
Rgrds,
Rita