Re: Fun with OOM on 2.4.18-pre9

Re: Fun with OOM on 2.4.18-pre9

[Jesse Pollard]

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).

-------------------------------------------------------------------------
Jesse I Pollard, II
Email: pollard@navo.hpc.mil

Any opinions expressed are solely my own.

Re: Fun with OOM on 2.4.18-pre9

[toad]

[-- Attachment #1: Type: text/plain, Size: 1731 bytes --]

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

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: Fun with OOM on 2.4.18-pre9

[Pavel Machek]

Hi!

> > 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.

No, it should *kill* kaffe, not only say it is doing that.
									Pavel

-- 
(about SSSCA) "I don't say this lightly.  However, I really think that the U.S.
no longer is classifiable as a democracy, but rather as a plutocracy." --hpa

Fun with OOM on 2.4.18-pre9

[toad]

[-- Attachment #1: Type: text/plain, Size: 899 bytes --]

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?

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]