Re: Hang after running out of memory

From: Craig Reyenga <creyenga_at_connectmail.carleton.ca>
Date: Mon, 21 Apr 2003 13:54:19 -0400
From: "Mike Hunter" <mhunter_at_ack.berkeley.edu>
> On Apr 20, "Craig Reyenga" wrote:
>
> > I have been following the thread here on current_at_ about problems
occuring
> > when running out of memory, so I wrote a C++ program that simply mallocs
> > about 144M of memory. My system has 256MB+491MB(swap). I can run the
system
> > out of memory just fine by running multiple copies of the program, and
the
> > kernel does kill one if I run too many, however shortly after, when I go
to
> > killall the programs the system hangs, (although pings, etc manage to
> > survive for a few mintues longer) and I have no choice but to press the
> > reset button. While it is hung, I _can_ press ^T and it shows the load,
> > which increases by 1 about every 45 seconds:
> >
> > load: 11.58  cmd: killall 20125 [running] 0.00u 0.00s 11% 292k
> > load: 12.27  cmd: killall 20125 [running] 0.00u 0.00s 12% 292k
> >
> > Also, the console is just as unusable as any other terminal. I'm not
sure
> > what is going on, but this is reproducible, and I can post more info
upon
> > request.
>
> I've experienced similar things after running out of memory.  It seemed
> like the scheduler went crazy:  The only way I could get my ssh session to
> keep talking was to start other ssh sessions, which would buy me about 20
> keystrokes.  shutdown -r got half done and failed into a weird state.  I
> think the kernel was from around April 8.
>

Was it SCHED_ULE? That's the one I've been using.

-Craig
Received on Mon Apr 21 2003 - 08:54:44 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:04 UTC