From: Herbert Poetzl (herbert_at_13thfloor.at)
Date: Sun 29 Dec 2002 - 21:26:59 GMT
On Sun, Dec 29, 2002 at 04:10:27PM -0500, tedsuo wrote:
> > > >Using 2.4.20 kernel; what is the default file-nr value? And what is the
> > > >normal max. value for 1gb p1.5mhz machine?
> > > >
> > > >cat /proc/sys/fs/file-nr
> > > >7161 1452 10240
> > > >
> > > >When i compile new kernel, the max value is about 103xxx and it looks
> > > >very big for me, and i modify it as 10240.
> > > >
> > > >Now my uptime is about 5 days, and i hope the >crash problem is about
> > > >max value.
> > >
> > > Any news about this???
> > > or at least nobody reported back ...
> > I am not able to reproduce this issues, so how
> > should I get/provide more information ...
> > it seems that nobody had the time or was interested
> > to try the two patch sets I made for this purpose,
> i have tested your patch on another computer and all seem ok
> 2.4.20ctx-16 #1 SMP
> 16:03:14 up 1 day, 15:31, 4 users, load average: 0.03, 0.07, 0.02
> but the fact is on this computer no crash happen with ctx16
> ( 10 days of uptime before i test your patch)
in this case, I would suspect the following:
- your hardware of the crashing machine is faulty
- your kernel built on the crashing machine is faulty
- the combination of your kernel and the hardware
shows erroneous or at least unusual behaviour
which leads to races or failures ...
(ordered by probability)
> I cannot test your patch on the computer that crach with ctx16
> 10 reboots whith ctx16 in 3 days is enought for the clients...
you should never use bleeding edge technologies
for production systems ... (but I think you know that now)
> back on ctx14 and no crach happen...