[NTLUG:Discuss] Linux crashes

[MadHat]

S Jackson wrote:
> 
> I'm surprised to here these problems being blamed on SMP.  I have 5 SMP
> machines (ranging from a dual PPro to a dual P3-500) and I have never
> had any problems.  I never had problems with 2.2.5 but I use 2.2.9+ on
> all my machines now, so maybe you could try a kernel update?
> 

I agree.  I know that SMP is still fairly new with x86 based machines
(as compared to commercial *nix run on say SPARC or others), but I run
several machines with Dual CPUs and have yet to see a problem (2.2.9+ as
well).  And these machines are hit pretty hard.  Now I could beleive
that something in the SMP may not be compatible with say the Team
Fortress server (or how it is trying to utilize the SMP), but I wouldn't
just say it is the SMP itself.  Remember how SMP works.  It works on
threading or on processes.  If an application only does threading
(meaning that it runs everything in one process, but has multiple
instances of the same commands going on in the one process), it has to
be written to work with SMP using the SMP threading libraries, if it
spawns children (like httpd, where one master is started and it starts
new processes to handle new requests), then the kernel will handle the
SMP by sending new processes to the CPU that is not being used as much
(the one with the lowest load).  There may be something in a server, or
daemon, that is trying to share memory with another process, or
something of that nature and it just isn't working right, messages get
crossed and WHAM!  Once again, this is just me thinking out loud.  If
any of this is wrong, I am sure someone will let us know.  I think I
have made sence...

> Jackson
> 
> alan at tieless.com wrote:
> >
> > I've seen the same problem on my dual Pentium Pro.  I've managed to narrow it
> > to a problem with the SMP code. I'm not a kernel  hacker,  so that's as far as
> > I can go.
> >
> > -Alan
> >
> > > I've been having a problem with my Linux server crashing recently...
> > > Here's my config:
> > > -dual PII 350, SuperMicro MB, AMI bios
> > > -128 MB RAM
> > > -3Com 3C509b NIC connected to Georgia Tech's network
> > > -Sound Blaster 16
> > > -Kernel 2.2.5 SMP
> > > -Running: Netscape, Maple, DrScheme, KDE, Half-Life TFC dedicated server
> > > with 20+ players, mp3 player, ProFTPD (fairly recent to fix a hole, but
> > > there may be other holes)
> > >
> > > The CPU usage is about 50%, which is mainly the Team Fortress server.
> > > Memory usage is pretty high, but I don't think it's running out.
> > > Everything's going peachy, and all of a sudden the machine hardlocks.
> > > Under Windows I'd just reboot and not think twice, but Linux doesn't
> > > usually do this... I could understand the server crashing, but hardlocks
> > > are disconcerting. Any suggestions that don't involve removing the
> > > TFC server?
> > >
> > > Thanks!
> > > -Reeves

-- 
MadHat