Memory Problems

[Richard Kail]

Hello !

On Wed, 1 Dec 1999, Chris Tooley wrote:

> I have a Samba server running on RedHat 6.0 (Samba 2.0.5a).  The server is a
> dual 233 with 192 meg of RAM and it is constantly choking on memory.  It is
> running several other servers (sendmail, apache, mysql, openldap, and Knox's
> Arkeia Backup Server) but if samba isn't running I max out at about 50-58
> meg of memory used, but if smbd and nmbd are running it hits about 185-188
> meg of memory used, with about 45 meg of that being cached memory.  

As far as I understand, this is quite normal in this situation. Linux
tries to use all the memory available. If there are not enought processes
to fill up the memory, it will use the memory as disc and page cache. This
is what you see als 45MB cached memory.

Even if your box is swapping some pages out, this shouldn't worry you.
Cache Pages have not the lowest priority. If there are some totaly unused
pages lurking somewhere, Linux will swap them out and use the so freed
memory as cache.

Try to analyse the "ps axm" output and look at the output from "vmstat 1".
You have a memory problem, if you get substantial si/so values from vmstat
when running under full load. You have also a memory problem if your box
can't use a substantial part of the memory as cache/buffer, lets say 20%
for your case.

You have too much memory, if you get a substancial free-value, as
seen with top or "free". This means that your system isn't able to use all
the memory you plugged in.

You have not too much, but also not too less memory, if you get
cached/buffered values in a good range (this is very dependend what you
are running on the machine), only a small percentage of the whole used
virtual memory is on disk (swap) and you have no substancial exchange
between RAM and Swap, as seen with vmstat.

You have too less memory, if you get small cached/buffered values, much
swap space is used /and/ you see substancial exchange between RAM and
Swap, seen with vmstat.

Kind regards,
	Richard

-- 
"Either gravity is different than we think it is or time is messed up
somehow" -- Michael Nieto, about the unexpected slowdown of space probes.