[KLUG Members] Memory dump?
Peter Buxton
members@kalamazoolinux.org
Thu, 11 Dec 2003 01:03:14 -0500
On Wed, Dec 10, 2003 at 07:46:20PM -0500, copycon was only escaped
alone to tell thee:
> I am running RH9 with the KDE desktop I have 368mg of RAM and I run KDE
> system gaurd to monitor the system and the memory just keeps increasing
> like thier is a leak or that it is not being released back to the system
> after a program is closed it gets to the point were programs wont even
> open anymore and I have to reboot this usually takes about 8 days
I'm assuming you've already used `ps afx` to look for zombie processes.
And from what you wrote, this isn't a simple "disk cache buffers
mistaken for program-assigned memory" error. And I take it top(1),
sorted by memory, isn't showing you enough history to spot the offenders
ps axv | sort -n -k 8
Merely set up a cron job to run that every 4 hours or so and save the
results to files named like this:
FILE_NAME=`date +%F_%X`
Don't save to /tmp, it gets erased every reboot.
ps(1) has very messy output, though. Another method:
#! /bin/sh
PDIR=/root/`date +%F_%H.%M`
mkdir $PDIR
cd $PDIR
for i in $( find /proc/ -maxdepth 1 -name [1-9]\* )
do
STATFILE="${i}/status"
if [ -a $STATFILE ] && [ ! /proc/self -ef $i ]
then
# Note that /Pid: / contains a TAB, not a SPACE (Ctrl-v, TAB)
OUR_OUT=`grep ^Pid: $STATFILE | sed 's/Pid: //'`
grep -E '(^Name|^Vm)' $STATFILE > $OUR_OUT
fi
done
exit
You're going to record a lot of transient PIDs, of course, but they
should not be the memory leak offenders.
--
[The Basement Tapes were] like the Watergate tapes... Bob
would say, 'We should destroy this.' -- Robbie Robertson