Operating System - HP-UX
1834925 Members
2580 Online
110071 Solutions
New Discussion

CPU Bottleneck and /tmp questions

 
SOLVED
Go to solution
Angela L. Shepherd_1
Occasional Advisor

CPU Bottleneck and /tmp questions

I have two questions:

1. My /tmp files system filled up. I cleared out the file that caused the problem but my /tmp file system still shows 100%. Why is my /tmp file still showing 100%.

2. I'm having CPU bottleneck problems, End of global swap space is nearly full. I obtained this information in Glance. The system table report shows shared memory available 1024gb and requested 1.66gb. I have 3.00gb of physical memory. Can I add more swap space without causing more problems.
6 REPLIES 6
Tracey
Trusted Contributor
Solution

Re: CPU Bottleneck and /tmp questions

Angela,

I'll answer the first question:

You may have removed the file, but the process that created it still thinks it is open, so the OS will not release that file space until the process dies, or you reboot.
Darrel Louis
Honored Contributor

Re: CPU Bottleneck and /tmp questions

For your filesystem full problem:
du -s /tmp
Devide the output by two, should give size in mb.

fuser -c /tmp, check which processes are having files open on /tmp

Swap:
You can add extra swap to the system, what you should check, what is causing the high swap usage???

Can you please show the output of
swapinfo -t

Do you have a databse running on the server?

check recommended kernel parameters.

http://www.docs.hp.com//hpux/onlinedocs/os/KCparams.OverviewAll.html


CHRIS_ANORUO
Honored Contributor

Re: CPU Bottleneck and /tmp questions

Hi Angela,

The files in /tmp should not bother once cleared and on reboot the will clean out if you have set CLEAR_TMP=1 in /etc/rc.config.d/clean_tmps.

With RAM of 3Gb, increase your swap device upto 2Gb, enable swapmem_on=1, this will give you pseudo memory of 75%RAM. Set nbuf and bufpages to zero for dynamic buffer caching.
When We Seek To Discover The Best In Others, We Somehow Bring Out The Best In Ourselves.
Kofi ARTHIABAH
Honored Contributor

Re: CPU Bottleneck and /tmp questions

Angela:

The file you removed may still have been opened by a running process - if that is the case, "it is still there". Find out what process created that file and shut it down (and restart it) to reclaim your disk space - usually, you should not remove a file without first verifying that it is not "in use" by a running process.

you can use lsof
http://eigen.ee.ualberta.ca/hppd/hpux/Sysadmin/lsof-4.51/
to see what files are owned by what processes.

good luck.
nothing wrong with me that a few lines of code cannot fix!
James R. Ferguson
Acclaimed Contributor

Re: CPU Bottleneck and /tmp questions

Hi:

You can safely and easily add more device swap.

Enable the kernel parameter 'swapmem_on' by setting it to one (1). This is called "pseudo swap" and allows for the use of system memory as additional swap space.

You may need to increase the kernel too, in order to create additional swap environments.

You can setup a new logical volume as an additional swap space. For best performance, make the size of this secondary swap the same as your primary, and set the swap priority to be the same too.

For additional procedural guidelines for managing swap space, see the System Administration Tasks manual:

http://docs.hp.com/hpux/onlinedocs/B2355-90672/B2355-90672.html

...JRF...
Darrel Louis
Honored Contributor

Re: CPU Bottleneck and /tmp questions

Dynamic Buffer cache.

When the parameters bufpages and nbuf are both set to their default value of 0, the size of the buffer cache grows or shrinks dynamically, depending on competing requests for system memory.

The value of dbc_max_pct sets the maximum percentage of physical memory that can be allocated to the dynamic buffer cache.

It is possible to set both dbc_max_pct and dbc_min_pct to the same value, 12 for example, and create a kernel that uses exactly that percentage of physical memory for the buffer cache, regardless of the size of physical memory.

If your dbc_max_pct is set to 50 %, that means that 1.5 Gb is allocated for Dynamic buffer cache.
As Chris stated set nbuf and bufpages to zero.
If dbc_max_pct is 50, bring it close to dbc_min_pct or equalize those two.

See the url in the previous answer given.