virtual memory management

[Kris Kennaway]

On Sat, Jan 20, 2007 at 03:51:38AM -0500, Kris Kennaway wrote:
> On Sat, Jan 20, 2007 at 09:13:48AM +0100, Zbigniew Szalbot wrote:
> > Hello again,
> > 
> > >> The swap size usage grow so big probably because I started wget to
> > >> download an iso image and then WinSCP to grab it from the FBSD machine
> > >>  to my laptop. When I started wget, the swap usage was around 19% and
> > >> had been like that for many days.
> > >
> > > That should not cause such a thing (wget does not try to fit the whole
> > > file in memory, so it won't be pushing stuff out to swap).  Look at the
> > > process sizes in top to see what is using the swap space - something(s)
> > > that is still running is using that 482MB.
> > 
> > I do not see such a process:
> > 
> >   PID USERNAME   THR PRI NICE   SIZE    RES STATE    TIME   WCPU COMMAND
> > 21442 root         2  20    0   224M 26128K kserel   0:06  0.00% java
> >   896 mysql       16  20    0 70756K 14764K kserel 255:35  0.00% mysqld
> > 98693 bind         1  96    0 32812K 32172K select   0:22  0.00% dnscache
> >  5035 www          1   4    0 28372K 15660K accept   5:05  1.86% httpd
> >  5026 www          1   4    0 28240K 15104K accept   4:46  0.00% httpd
> >  5065 www          1   4    0 28128K 15196K accept   4:29  0.00% httpd
> >  5030 www          1   4    0 27892K 15144K accept   4:21  0.00% httpd
> >  5126 www          1   4    0 27784K 14864K accept   4:20  0.00% httpd
> >  5029 www          1   4    0 27760K 14644K accept   4:22  0.00% httpd
> >  5027 www          1   4    0 27740K 15140K accept   4:30  0.00% httpd
> >  5028 www          1   4    0 27516K 14812K accept   4:03  0.00% httpd
> > 95977 www          1   4    0 27216K 14532K accept   2:21  0.00% httpd
> >   703 root         1  96    0 16412K  2296K select   4:35  0.00% httpd
> > 91014 mailman      1   8    0 11492K  1600K nanslp   6:00  0.00% python
> > 91012 mailman      1   8    0 11024K  1560K nanslp   5:32  0.00% python
> > 91010 mailman      1   8    0 11008K  1568K nanslp   5:23  0.00% python
> > 91009 mailman      1   8    0 11008K  1552K nanslp   5:20  0.00% python
> 
> I see lots of them; every one in that list is contributinig.  If you
> add up all those process sizes you'll see where the space is going.

By which I mean the difference between size and res, which indicates
the amount of process memory allocated but not currently resident in
RAM.  This isn't a foolproof method (see e.g. the FAQ entry on
rpc.statd), but it's true in your case.

> Basically you are just overloading your system by trying to run too
> much at once.  Reduce the load or add more RAM.
> 
> Kris


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20070120/74392df1/attachment.pgp