On Thu, Feb 14, 2002 at 08:56:10AM -0600, Jesse Pollard wrote:
> toad <mtoseland@cableinet.co.uk>:
> >
> > I do a make -j bzImage
> > I have 2 large processes (Kaffe) running in the background. They are
> > driven by scripts like this:
> >
> > while true;
> > do su freenet -c java freenet.node.Main;
> > done
> >
> > I have 512MB of RAM and no swap on 2.4.18-pre9.
> > Kernel eventually slows to a near complete halt, and starts killing
> > processes.
> > It kills Kaffe several times
> > Out of Memory: Killed process xyz (Kaffe)
> > (no I don't have logs, sorry)
> > Each time it's a different pid, having respawned from its parent
> > process. Then later, it apparently becomes unkillable - each time it
> > respawns it is *the same PID*. VT switching works, but otherwise the
> > system is unresponsive. It is not clear whether the make -j is still
> > running. Immediately before this, it did the same thing with dictd, but
> > eventually got around to Kaffe. After a fairly long wait I rebooted with
> > the reset switch.
> >
> > Any more information useful? Is this known behaviour?
>
> Known behaviour - ie. don't do that.
>
> On most systems a "make -j4", or 6, or even 8 will work. But SOMETHING has
> to give after memory fills up. The number of processes to use for -j
> depends on the system. On mine (256MB, dual pentium pro) I don't do more
> than 6, and 4 works best (lets me run solitare too).
Isn't that what the OOM killer is for?
>
> -------------------------------------------------------------------------
> Jesse I Pollard, II
> Email: pollard@navo.hpc.mil
>
> Any opinions expressed are solely my own.
-- The road to Tycho is paved with good intentions
This archive was generated by hypermail 2b29 : Fri Feb 15 2002 - 21:01:02 EST