Greywolf wibbled one day
> On Wed, 28 Mar 2001, Tim Preston wrote:

> # I've been seeing file system corruption on a regular basis since I build
> # my -current playbox using 1.5S a couple of months ago.
> #
> # This is on a RAID1 file system running of IDE disks, with softdep on.
> #
> # Interestingly I only see this problem on one of the file systems, and
> # it's the one I ran fsirand on as it is likely to be NFS exported at some
> # point, however it contains all my source and and obj trees and so it's
> # by far the most heavily used filesystem on the box so this is likely
> # coincidence...

> Do you s'pose that fsirand might be causing the problems?
> (shooting at straws, grasping at the dark).

I have no idea, but it's the only difference (besides the relative load)
between the file systems. I am not seeing these problems to any degree
with the others.


> You might try it on a smaller filesystem and see what happens.

Hmmm. I'll do some file shuffling and newfs one of the other filsystems,
run fsirand on it before I repopulate then run some create/delete type
scripts over it for a few days and see what happens.


> Incidentally, do you have a reason you need to run fsirand?  (Typically
> it's only necessary in a potentially sorta-hostile environment (full-on
> hostile environments will probably not want to use NFS at all)).

Unfortunatley that box is sort of in one of those environments at the
moment. Where I really don't want to run NFS (and I'm not at the moment)
but it might turn out to be stupidly useful in the near future, hence
the (unusual for me) paranoia =)


> I somehow don't think it's the problem, but I've been right once or twice
> before on hunches, so I figure it's worth a shot.

Well I'll report back after my testing. Although I'm still interested in
other people's  ideas =)


-- 
I am become Reginald, purveyor of fitted kitchens...