Re: Oracle 8.1.6 on Linux 2.2.12-20smp - memory problems (leak?)

From: Rolf Czedzak <520040398192-0001_at_t-online.de>
Date: Sat, 21 Jul 2001 21:23:47 GMT
Message-ID: <9hs9ei$pkv$03$3@news.t-online.com>

Prasanth Duvvur <pduvvur_at_weaveit.net> schrieb in im Newsbeitrag: a2251aba.0107020828.7a80c940_at_posting.google.com...
> Hi,

Hi,
please allow some questions:
Is memory declare to the OS correctly (append "mem=512M" in lilo conf file)? It looks like, but please confirm.
Is system slowing down while aging? (Using swapping more often) Does starting some memory hungry jobs change life time of Your system? How much swap space do You provide?

Well, first of all low free memory is nothing to worry about, it just tells about efficient memory usage (basically). Your using of vmstat does not provide much information, because runnig vmstat without 'delay' parameter gives
You average values _since last boot time_ on most of the items. Try logging system status by

#! /bin/bash
# Parameter
# $1   delay    ! set to get reasonable resolution

if [ $# = 0 ]
  DELAY=60;
else
 DELAY=$1;
fi
vmstat -n ${DELAY} > $HOME/vmstat.out &
# exit

watching output by
tail -f $HOME/vmstat.out

> I have been running oracle Release 8.1.6.0.0 on Linux 2.2.12-20smp.
> There are 512Megs of memory on the box. There is nothing else running
> on the box.
>
> SGA:
>
> SVRMGR> show sga
> Total System Global Area 85712880 bytes
> Fixed Size 69616 bytes
> Variable Size 44605440 bytes
> Database Buffers 40960000 bytes
> Redo Buffers 77824 bytes
>
> After the server/database have been up for a few days the free OS
> memory comes down to 1-2 megs, and from then on ranges between 1-30
> Megs. Before ultimately crashing the server and the database by
> running out of memory.
>
> I get free memory stats using vmstat or top e.g.:
>
> vmstat
> procs memory swap io system
> cpu
> r b w swpd free buff cache si so bi bo in cs
> us sy id
> 1 0 0 280 11204 87740 342920 0 1 75 10 78 69
> 2 1 97

vmstat output should become more informative with 'delay'. The output above shows an almost idle (both cpu as memory usage) system, being able to provide
430M as buffers and cache? _average_ swap space used is next to nothing, but being an average ´since last boot' it's almost useless.

> The CPU usage stays pretty low.
 

> The only remedy I have been able to come up with so far is to
> periodically shutdown the database and start it up.
>
> Could someone help me in determining the source of the problem?
Received on Sat Jul 21 2001 - 16:23:47 CDT