On Wed, 24 Jul 2002, Daniel Mose wrote:
> Cheer up people, The guy that you just lost was working with XFS in
> a production environment. He simply couldn't take the responsability
> towards his boss for having a production server that broke down every
> three days or so. It was quite brave of him to have XFS running at his
> company in the first place.
I had a test box in my department december 2000 that did all the stuff the
stuff that the rest of my machines also did including testing with all the
different clients. It wasn't untill May 2001 (Official 1.0 release) that I
started using it on production machines (one by one).
Today it is on all production machines and we have had a few very small
XFS problems yes but most were not caused directly by XFS. Our current
internet gateway/web/squid/mail server is XFS and is running a 2.4.18 CVS
XFS for the last 130 days or so.
The box ran a 1.0.2 release before that which was the only version that
acted funny together with a squid cache. After moving up the problem never
came back. Previous releases <1.0.2 never gave problems either.
It just shows you that you need to test in your environment and stay off
when it works flawless. I am not inclined to upgrade the kernel of this
box for a long time since everything is working dandy now.
> This kind of service would probably not be very hard to set up. I believe
> it could be done at some university by some professor, or student, where
> the XFS is already running.
> (I have the feeling that a lot of XFS testing and development is done by
> academy people now as we speak)
the oss.sgi.com server is a server with XFS filesystems and is used no
only by the XFS folks but also for the MIPS folks for instance. It does
get a beating and maybe Russel, Eric or Steve can quote some numbers of
GB/day it dishes out.
Cheers
Seth
|