Swap Memory 99%

[Prashantj]

Hi,

My client Devlopment server,HP-UX 11.11, oracle & Hp-DP cell manager loaded.
swapinfo reached 99%,i faced this problem 1 months back due to velocies daemon error in HP-DP,i had taken HP-Support & after adding some parameter in omnirc dir. problem has been solved.
But
Now the swapinfo between 94% to 95%
& at 12:25am to 12:35am (US TZ) it's reaching 99%
Swapinfo status.

Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 6652 411 6241 6% 0 - 1 /dev/vg00/lvol2
reserve - 6241 -6241
memory 3004 2467 537 82%
total 9656 9119 537 94% - 0 -
# kmtune | grep -i dbc

dbc_max_pct 50 - 50
dbc_min_pct 5 - 5

Please help me, waiting for your valueable advice.

Prashant.

[likid0]

How much physical memory do you have on your server?

You can all ways add some more swap, if you have disk space:



To create swap, you need a contiguous space on a disk/LUN.


Create vg01:

pvcreate /dev/rdsk/c25t0d0
pvcreate /dev/rdsk/c25t2d0
vgcreate -s 8 -p 128 /dev/vg01 /dev/dsk/c25t0d0 /dev/dsk/c25t2d0
vgchange -a y /dev/vg01

lvcreate -C y -L 16384 -n lvswap /dev/vg01

/usr/sbin/swapon -f -p 3 /dev/vg01/lvswap

add to /etc/fstab:

/dev/vg01/lvswap â ¦ swap pri=3 0 0

[Dennis Handly]

>dbc_max_pct 50 - 50

You probably want to reduce this to 5 or 10.

[Jeeshan]

For swap space creation , read the documents

http://docs.hp.com/en/B2355-90672/ch06.html

you can set the dbc_max_pct from 50 to 10 to increase your available physical memory space.

[Rasheed Tamton]

Hi,

As said above, reduce dbc_max_pct to 10 as first step.

>reserve - 6241 -6241
>memory 3004 2467 537 82%

If you have glance, just post the output of glance -m footer.


rgds.

[Prashantj]

Hi

Thanks for your valuable replay.
please tell me how to set the dbc_max_pct 50 to 10.

Please tell the concept of mention changes.

Prashant.

[Rasheed Tamton]

kctune dbc_max_pct=10

I do not have 11.11 systems now to check. I think on 11.11 it needs a reboot?. On the newer versions it is dynamic and no reboot is required.

dbc_max_pct value conrols the buffer cache.

[Rasheed Tamton]

If your Phys mem is 4GB, then I suggest you add another 2GB swap device as your OS is 11.11

[Fabien GUTIERREZ]

enlarging your swap space won t get you rid of your problem
your system memory shows still few but enough memory to be allocated in it much more then getting to swap
my opinion is that you may allocate a too much large sga from your oracle database (have a look at ipcs -ma to see values)
that directly get in your swap as it do not find a memory segment large enough to fill in your memory

[Tim Nelson · Accepted solution]

modify dbc_max_pct with kmtune -s dbc_max_pct=10
rebuild kernel with mk_kernel
prep new kernel with kmupate
reboot server

Then we can review your memory allocations.

-post your output of ipcs -ma
-post some output of vmstat 10 10

just because the used entry exists on your swap device with swapinfo does not mean you are swapping/deactivating. You will need to post some information as yes you are close but maybe not yet.

Either way you have a choice, either reduce your memory usage by tuning the kernel or adjusting application allocations or buy more memory.

4GB for an Oracle environment is not much but I guess it depends on the size of your application.