[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Serious filesystem problems with k7 kernels



Hello,
  I'm really not sure where to go for this, so i thought I'd try here.
Please CC me as I'm not on the list.

I've got two servers that are physically in different places but have
the same hardware. Same disk types, same board but video and ethernet
cards are different.

Recently I have noticed severe filesystem corruption.  It appears to
be with the reading of files.  Scary things like dpkg segfaulting
at random times, daemons dying or going into T state and all sorts
of other excitement have been happening on a regular basis.  I first
thought it was software RAID1 so I got rid of that, then I thought
it was XFS so I got rid of that too and went to ext2 on a normal IDE
disk.  Still the problems occurred. The problem appears to be random,
to get something to work you just keep running it until it stays
running, you get several different errors in the meantime.
/var/lib/dpkg/available was one file that particularly got corrupted a
lot.

I checked the memory with the memtester program and it was fine.

The servers were running different kernels. One of them was running
kernel-image-2.4.24-1-k7 and the other was running a custom 2.4.24
kernel.  I did notice going back to kernel-image-2.4.22-1-386 on the
first server fixed the problems, which got me thinking along those
lines.

So, just to see how it went I installed and booted into 
kernel-image-2.4.24-1-386. The server works without a problem.  I then
booted the second one into the same kernel image, again the problems
went away.  Rebooting between the -386 and -k7 kernels turns the
problems off and on.  It is definitely to do with the differences in
the kernel.

The hardware is a MSI KT266 Pro2 board with a 1GHz Althon processor. The
drives are IBM IDE drives (which I hate, but thats another story).

Any idea where to look first? Could it be ACPI?

 - Craig
-- 
Craig Small      GnuPG:1C1B D893 1418 2AF4 45EE  95CB C76C E5AC 12CA DFA5
Eye-Net Consulting http://www.enc.com.au/   MIEE         Debian developer
csmall at : enc.com.au                      ieee.org           debian.org 



Reply to: