> Yes - I was thinking of using xfs on some of our computers, but because of 
> time pressures and so on, we've switched our xfs computers back to ext3 
> :-(. Also I was starting to worry that some of the millions of patches 
> redhat that apply to the kernel might stuff up xfs, as they don't appear 
> to test xfs. There is a system that might require xfs soon, however.
> 
> What would be great would be to make a separate kernel-xfs package which 
> people could optionally install. Maybe it's worth asking on the xfs 
> mailing lists...

Reading through the xfs mailing lists it looks like RH disabled the non-4K
stacks option (8k stacks), and xfs _really_ wants 8k stacks or the whole
thing becomes rather unstable.  A bunch of RH vm patches assume 4k stacks
so turning 8k stacks on would require backing out (at least some of) those
patches...

We have machines running SGI's version of RHEL (SGI ProPack-3) so we can 
probably work out what they think is sufficient to enable xfs (at least on 
ia64) but only for 2.4 kernels...

Of course SGI have decided to switch horses and ProPack-4 is based on 
Suse (who don't disable xfs in their kernels)...

> >>> Jeremy Sanders <[log in to unmask]>   http://www-xray.ast.cam.ac.uk/~jss/
> >>> X-Ray Group, Institute of Astronomy, University of Cambridge, UK.
> >
> > Hmm, the UCS Unix Support keep trying to convince me that we are the only
> > SL users in Cambridge (but they possibly just don't know any better)...
> 
> I don't think we've told them since we've updated. We've got ~25 desktop 
> computers running SL, with another 30 odd systems to do in the immediate 
> future. We did try using SuSE Enterprise to replace Fedora, but I really 
> could not stand yast, and the horrible things it does behind your back.

I could ignore yast, but autoyask (the automated installer) drove me mad.

With suse 9.2 running the *same* install on the same machine would
sometimes give different results, and the installer would occasionally
fail to find some devices (I thought it *must* be a memory fault but
running memtest for 3 days showed no errors, and we never had problem on
that hardware with FC/SL -- before or since).

 -- Jon