[Mimedefang] memory leak?

[Stephen Smoogen]

On Mon, 2004-02-09 at 14:45, Ron Peterson wrote:
> On Mon, 9 Feb 2004, David F. Skoll wrote:
> > On Mon, 9 Feb 2004, Ron Peterson wrote:
> > 
> > > This system has 6GB RAM.  On boot, after everything gets going, the system
> > > consumes just under 1GB.  Over time, this useage is slowly creeping
> > > up.
> > 
> > Which process is growing?  (Use "ps" to find the process with the large
> > VSZ entry.)
> 
> No single process is getting too large.  Sorting 'top' by memory useage
> shows the mimedefang processes at the top, all using the expected 20-30MB.
> 
> > When you say the usage is creeping up, how do you know?  Linux, by default,
> > will use up all free memory for buffer cache purposes.  It is normal
> > for a Linux system to show practically 100% memory usage after it has been
> > running for a while.
> >
> > MIMEDefang is supposed to be highly resistant to memory leaks, but it's not
> > impossible that there are some.
> 
> What we have been seeing is that the system's memory useage slowly (over a
> day or two) creeps up to full utilization.  Then the swap space starts
> being used.  I don't know how hard it's getting hit, just that it goes
> from 0% to something slightly more than 0%.
> 
> That would all be o.k., but for the fact that the load average also
> spikes.  We're running a load average just over 1% right now.  The CPUs
> are over 80% idle.  Everything is going full bore.  By tomorrow morning, I
> expect the load average to be over 50, and the CPU idle to be way
> down.  This machine is also serving IMP, and those users start dropping
> off and complaining.  That function has existed for quite some time now
> without any problems.

Ok this sounds more like a problem with items going into Device Wait...
and hitting the load average up. The usual way I used to debug these
problems (when I did support for RH) would be to get a listing of
everything in the machine, and what services you are running.

CPU:
Memory:
Disk Drives:
Controller(s):
  Disk:
  Network:
  Other:

Services (versions):
  sendmail
  spamassassin
  mimedefang
  imp
  ????

netstat -natp
ps -auxww | grep D

also realize that x86 hardware has to do various internal hacks to deal
with memory above 2-4 Gigs. [2 gigs for some types of hardware and 4
gigs for other types.] Also check for things in your /proc/slabinfo to
see if anything is full.. or if dmesg is reporting some sort of network
issue.

Some weird problems like you have listed have gone away with a  mem=2g
and seeing how the box performs. If that were to occur.. then it would
be motherboard/cpu/pci card related and the only real fix is to go with
a 64 bit MB/CPU (Opteron/Itanium/sparc64 type stuff).



-- 
Stephen John Smoogen		smoogen at lanl.gov
Los Alamos National Lab  CCN-5 Sched 5/40  PH: 4-0645
Ta-03 SM-1498 MailStop B255 DP 10S  Los Alamos, NM 87545
-- So shines a good deed in a weary world. = Willy Wonka --