[lug] More Server Problems

[David L. Anselmi]

George Sexton wrote:
> My server tanked again (reference):
> 
> http://archive.lug.boulder.co.us/Week-of-Mon-20060123/031529.html
[...]
> I haven't had a chance to take the disks offline but there are no kernel
> messages normally associated with failing disks (write timeouts, dma
> timeouts, etc). I'm really suspecting a kernel issue in either the md or
> reiserfs subsystem. I suppose it could be a more generic issue relating to a
> Pentium D CPU but I think I would get some sort of message and the reiser
> file system wouldn't be hosed.

Really, you should run the disk diagnostics to rule them out.  If it 
isn't them I agree with you that it's a kernel (64bit probably) issue. 
But "i/o failure occurred trying to find stat data" sounds close enough 
to a failing disk to me.  And since the disks are taking themselves 
offline for you... ;-)

[...]
> The one thing I'd like to sort out how to do is to make SYSLOG-NG echo any
> kernel or other error messages to the console running on ttyS0.

You should be able to tell syslog to send things to the console (as you 
can tell it to send to a pipe or file or whatever).  I don't use NG but 
the man page should be obvious enough.  You could also tell it to log to 
the machine you're using for minicom.

The kernel will already log to the console though there's a setting 
somewhere to control how verbose it is.

> Does anyone know a simple way of doing this? I'd also be interested in ideas
> on how I can fix this issue. The only solution that I can think of would be
> to re-install linux, and choose a 32-bit installation. I have a machine that
> has the same disk sub-system but is running 32 bit kernel/linux and I'm not
> having any issues.

Really, check the disks before you reinstall (the machine will be down 
anyway).

Did you burn this in well before you sent it to the colo?  Maybe a 
lesson for next time...

Dave