Original question:
> This system is running 4.0G and very sluggish
>
> Virtual Memory Statistics: (pagesize = 8192)
> 112285 active pages
> 261990 inactive pages
> 229166 free pages
> 94206 wired pages
> 102243726720 virtual memory page faults
> 15282601 copy-on-write page faults
> 85238040 zero fill page faults
> 3954465 reattaches from reclaim list
> 122684781 pages paged in
> 523299 pages paged out
> 11671637208 task and thread context switches
> 2018357838 device interrupts
> 85583682813 system calls
>
> swapon -s
> Swap partition /dev/re0b:
> Allocated space: 256000 pages (2000MB)
> In-use space: 256000 pages (100%)
> Free space: 0 pages ( 0%)
>
> Swap partition /dev/rzd0c:
> Allocated space: 2221084 pages (17352MB)
> In-use space: 478894 pages ( 21%)
> Free space: 1742190 pages ( 78%)
>
> Total swap allocation:
> Allocated space: 2477084 pages (19352MB)
> In-use space: 734894 pages ( 29%)
> Available space: 1742190 pages ( 70%)
>
> 'top' shows half of 8GB main memory is being used. I cannot find any
> culprits in top or 'ps aux'. How do I go about diagnosing this problem?
===========================================================================
"Serbia, Jorge" <Jorge.Serbia_at_compaq.com> a compaq engineer sent
me the following URL on "Improving Large-Memory System Performance"
http://www.tru64unix.compaq.com/faqs/publications/best_practices/VLM_BP/TITLE.HTM
Here are some of the replies I received:
=======================================
From:
<alan_at_nabeth.cxo.dec.com>
% vmstat 2
It will calculate the rates for you, and more reliably
than you could over a time difference of 5 - 8 minutes;
fractions of seconds. You can use different sample rates,
but 2 seconds is a nice round number. Ignore the first
line of this since it is the total since the syste booted,
which is basically what you've already posted. A decent
sample will be about a dozen lines. Most people that
look at vmstat(1) listings are used to seeing samples
collected over a few seconds.
========================================================================
From:
"Nemholt, Jesper Frank" <JesperFrank.Nemholt_at_compaq.com>
What is the output of swapon -sv ???
The output of swapon -s only show "in use" swap, but doesn't show if
it's
actively in use or just reserved (eager swap mode).
If you have a link called /sbin/swapdefault, you run in eager mode. If
not,
you run in lazy mode.
The output of uptime is also interesting, to see if it's a high runqueue
causing the problem.
Additionally, the output of /sbin/sysconfig -q vm and /sbin/sysconfig -q
proc
And how many files you have in /tmp
At last, som data from /usr/sbin/collect running it over a period of
time
when the machine is at its worst.
"/usr/sbin/collect -F" will give you data every 10 seconds. The -F
option
will make all values of the same scale. Warning : There's a lot of data
!
"/usr/sbin/collect -F -f filename.cgz" will save the data for later
playback.
==========================================================================
From:
Udo de Boer <Udo.de.boer_at_ubero.nl>
Start with "vmstat 1 100" to look at what the cpu and memory is doing.
Ignore the first line from vmstat. Also look at the disks. Are they very
busy. You could have a disk throughput problem. You can try iostat. Next
look if you have a network problem. See netstat -i. (or was it netstat
-a).
You have to do this twice to get an interval. Look at the errors and
collisions.
===========================================================================
Things are back to normal after a reboot. I will use "collect" to gather
performance statistics in a file from now on. I will also run "syscheck"
and see its reccommendations. I think the URL Jorge Serbia sent is very
useful. Thanks everyone who responded.
Sridhar
Received on Tue Jul 17 2001 - 21:20:34 NZST