> Well, they got the server back up. Said /etc/rc.d/rc.local was "complete
> gobledegook".
Define gobbledegook i.e. what was in there, was it all zeros? Also, is root on
xfs or not.
The unfortunate thing is that so far the only way to upgrade from the rpms
is to build your own kernel. The cvs tree is currently at 2.4.6-pre2, there
is a patch against 2.4.5 on the ftp site in the patches directory. There are
a lot of xfs fixes in this since the 1.0 release.
I cannot remember exactly what compiler was used to build the kernel rpms,
it should have been the egcs-2.91.66 version, but I was not directly involved
so I cannot be sure.
>
> I'm back in and everything else seems to be in OK shape.
>
> Is there a fairly straightforward way to upgrade the kernel and XFS? I'm a
> little wary of using 2.4.5 because of the swapping issues. But it might be
> worth a shot.
What swapping issues - swapping in 2.4.5 is no different than previous 2.4
versions, it is just people started commenting on it. The 2.4.6-pre2 kernel
does have changes in it to make swapoff go faster if that is your concern.
>
> > Well, check the Linux mailing lists -- reading those, and having battled
> > with a Tyan Tiger 133 for ages now, I would be extremely hesistant depend
> > on anything with a VIA chip set onboard. I'm sure that they'll get it
> > right at some point, but at the moment, stability and reliability still
> > seems to equal Intel, and then preferably a 440BX chip set.
>
> well I certainly can't replace the motherboard. Unless things *really* get
> bad and that's the only possible solution.
There is nothing about XFS which should be directly affected by the particular
type of motherboard you have. XFS may stress the system in different manners,
and expose kernel/bios/hardware bugs that ext2 would not. However, if a board
works
without problems on Linux normally, but is crashing with XFS then I would say
the most likely explaination is that you have an I/O load which is triggering a
bug in XFS. Without some form of oops message (decoded) or a stack backtrace
from kdb, there is not much we can do to diagnose the problem. There is a good
chance that running the latest kernel will make the problem 'go away' since
we have fixed a lot of problems. However, your situation with being so far from
the machine does not make it easy to do experiments with different kernels.
>
> If this happens again I'll probably just have them install Debian + ext2.
Steve
|