* BUG in the kernels 2.4.x
@ 2001-08-10 8:04 Eduard Phetisov
2001-08-10 6:46 ` Marcelo Tosatti
0 siblings, 1 reply; 2+ messages in thread
From: Eduard Phetisov @ 2001-08-10 8:04 UTC (permalink / raw)
To: linux-kernel, haible, osrc, quintela, dwmw2
I think what I find a bug in the kernels 2.4.x.
Bug apears then several processes allocates some memory. If size of the
free memory less then size of claimed memory by processes, system went
down. Such phenomenon apears on the several Linux systems.
Affected systems
Servers:
Compaq Proliant 8500 (CPU 8xXeon 600, 8Gb RAM, Raid Compaq 5300, OS
RedHat 7.0, 7.1, kernels 2.4.2 2.4.3 2.4.4 2.4.5 2.4.7), start.sh
parameters 200 50 10
WIT (CPU 8xXeon 700, 8Gb RAM, Raid Mylex DAC960 OS RedHat 6.2, 7.0, 7.1,
SuSe 7.2, Mandrake 8.0, kernels 2.4.2 2.4.4 2.4.7 2.4.8) start.sh
parameters 200 50 10
IBM (CPU 2xPentium III 500, RAM 1Gb, OS RedHat 6.2, kernels 2.4.2 2.4.3
2.4.4 2.4.5) start.sh parameters 200 10 10
Workstantions:
Cpu Pentium III 600 , RAM 128Mb, OS RedHat 7.0, kernels 2.4.1 2.4.2
2.4.3 2.4.4 2.4.5 2.4.7, start.sh parameters 70 10 2
Cpu Celeron (Copermine) 600, RAM 64Mb, OS RedHat 7.1, kernels 2.4.1
2.4.2 2.4.3 2.4.4 2.4.5 2.4.7, start.sh parameters 70 10 2
Cpu Celeron (Copermine) 900, RAM 128Mb, OS RedHat 7.1, kernels 2.4.1
2.4.2 2.4.3 2.4.4 2.4.5 2.4.7, start.sh parameters 70 10 2
Cpu Celeron (Copermine) 912, RAM 128Mb, OS Mandrake 8.0, kernels 2.4.3
2.4.4, start.sh parameters 70 10 2
with following programs
// Memuse.c is a very funny programm
#include <malloc.h>
#include <stdio.h>
main(int argc, char **argv)
{
int ms, rss, i, sum;
char * arr;
if (argc < 2)
{
printf("usage %s MS(MB) RSS(MB)\n", argv[0]);
exit(255);
}
ms = atoi(argv[1])*1024*1024;
rss = atoi(argv[2])*1024*1024;
printf("MS=%d bytes, RSS=%d bytes\n", ms, rss);
printf("Allocating memory ...\n");
arr = malloc(ms);
if (arr==NULL)
exit(printf("Malloc returns NULL\n"));
for(i=0;i<ms;arr[i]=(i++ & 255));
printf("Memory Allocated\n");
while(1)
{
for(sum=i=0;i<rss;arr[i] = (i++) & 255);
sleep(3);
}
}
and with shell script
#!/bin/bash
i=0
if [ $# -ne 3 ]
then
echo Usage $0 Count Use RSS
exit -1
fi
while [ $i -ne $1 ]
do
memuse $2 $3 &
echo $i
i=`expr $i + 1`
done;
Analogous situation must appear every morning, because we have a big
database in our busines (Oarcle, database size is about 100Gb and
awerage transaction activity about 1000 transactions per second), which
used by over then 400 people. Kernels 2.2.x (we use 2.2.19 with several
cenges, which allow to use 1.6Gb in the shared memory) works properly in
this situation, but it can't use more then 4Gb of memory, so please help
us and solve this problem.
^ permalink raw reply [flat|nested] 2+ messages in thread* Re: BUG in the kernels 2.4.x
2001-08-10 8:04 BUG in the kernels 2.4.x Eduard Phetisov
@ 2001-08-10 6:46 ` Marcelo Tosatti
0 siblings, 0 replies; 2+ messages in thread
From: Marcelo Tosatti @ 2001-08-10 6:46 UTC (permalink / raw)
To: Eduard Phetisov; +Cc: linux-kernel, haible, osrc, quintela, dwmw2
On Fri, 10 Aug 2001, Eduard Phetisov wrote:
> I think what I find a bug in the kernels 2.4.x.
> Bug apears then several processes allocates some memory. If size of the
> free memory less then size of claimed memory by processes, system went
> down.
What do you mean by system went down ?
Complete lockup or slowndown?
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2001-08-10 8:16 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2001-08-10 8:04 BUG in the kernels 2.4.x Eduard Phetisov
2001-08-10 6:46 ` Marcelo Tosatti
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®