Hello,
In the last 48-hours we've had two machines hang with no messages on the
console and none in the logs, pings return, drive lights are all off and
don't ever flash. Both machines we running linux-2.4.24 with the
xfs-2.4.23-all-i386 patch as recommended by the XFS group. These
machines had been running fine on this kernel image since Jan 13,
2004. One machine was running XFS as a filesystem and the other was
running EXT2. These machines have both been in constant production for
over 4 years and there has not been any hardware changes, including
those to it's environment with is power and temperature controlled.
Both machines are Pentium-3 class, 256M of memory, both using SCSI disk
with different SCSI controllers. This is the the version string:
Linux version 2.4.24-xfs (j3gum@krypton) (gcc version 2.95.4 20011002
(Debian prerelease)) #1 SMP Tue Jan 13 22:05:12 CST 2004
If you need more specific info. I'll prob. keep these and the other
dozen on the same kernel image up for a few more days.
--
jeffrey hundstad
On Sat, 24 Jan 2004, Jeffrey E. Hundstad wrote:
> Hello,
>
> In the last 48-hours we've had two machines hang with no messages on the
> console and none in the logs, pings return, drive lights are all off and
> don't ever flash. Both machines we running linux-2.4.24 with the
> xfs-2.4.23-all-i386 patch as recommended by the XFS group. These
> machines had been running fine on this kernel image since Jan 13,
> 2004. One machine was running XFS as a filesystem and the other was
> running EXT2. These machines have both been in constant production for
> over 4 years and there has not been any hardware changes, including
> those to it's environment with is power and temperature controlled.
>
> Both machines are Pentium-3 class, 256M of memory, both using SCSI disk
> with different SCSI controllers. This is the the version string:
> Linux version 2.4.24-xfs (j3gum@krypton) (gcc version 2.95.4 20011002
> (Debian prerelease)) #1 SMP Tue Jan 13 22:05:12 CST 2004
>
> If you need more specific info. I'll prob. keep these and the other
> dozen on the same kernel image up for a few more days.
Hi Jeffrey,
There is a known VFS _SMP_ deadlock in 2.4.24. I'm not sure if that is
what you are hitting, but it is likely.
So try 2.4.25-pre7 (which contains an uptodated XFS tree) and check if the
problem goes away.
Please keep me informed.