Global swap area full

[Dayanand Naik]

Hi,

Lately there are critical messages
popping up on the VPO screen as Global swap area is nearly full. Using glance, swapinfo the memory utilization showed 100% full
the server popped out throwing errors unable to fork process. As the process was not able run I had to reeboot the server. Can you tell me which process is utilizing more swap/memory or is there are reserve space allocated for
oracle/nnm process.

Regards.
Dayanand Naik

[Ravi_8]

Hi,
The oracle needs a min of 128 MB, may i know your current RAM and swap size.


later
ravi

[Alexander M. Ermes]

Hi there.
If you use telnet, you might need a patch.
We had the same kind of error.
After we patched the telnetd, it was ok.
I will try to find the patch.
Rgds
Alexander M. Ermes

[Dayanand Naik]

Hi,

here are the details
Physical mem 512 MB, 1024 swap space. on /vg00/lovl2 attached are the output of glance and swapinfo.

Regards
Dayanand Naik

[Mladen Despic]

Hi Dayanand,

A simple suggestion is to disable pseudo-swap
(change the kernel parameter "swapmemon" from 1 to 0). I would also try to add more physical memory if possible, for performance reasons.

The 'memory' line in your 'swapinfo' output is about pseudo-swap reservations. When memory is locked (by Oracle), it actually contributes to this percentage.

The amount of device swap is double the amount of physical memory, so if you disable pseudo-swap, that should be just fine.

On the other hand, the amount of free memory is low on your system, so you could investigate what is using all the memory. Chances are, however, that you will simply need more physical memory, for perfromance reasons.

For more details, see Bill Hassell's comment at this link:

http://forums.itrc.hp.com/cm/QuestionAnswer/1,1150,0x3994663ce855d511abcd0090277a778c,00.html