Message-ID: <20050828144634.43837.qmail@web34110.mail.mud.yahoo.com>
Date: Sun, 28 Aug 2005 07:46:34 -0700 (PDT)
From: "Bryan J. Smith" <b.j.smith@ieee.org>
Reply-To: b.j.smith@ieee.org
Subject: Re: Will we ever see XFS supported in Red Hat Enterprise Linux?
To: Joshua Baker-LePain <jlb17@duke.edu>
Cc: linux-xfs@oss.sgi.com
In-Reply-To: <Pine.LNX.4.58.0508280729370.30086@chaos.egr.duke.edu>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Sender: linux-xfs-bounce@oss.sgi.com
Errors-to: linux-xfs-bounce@oss.sgi.com
Precedence: bulk
Content-Length: 3036
Lines: 75

Joshua Baker-LePain <jlb17@duke.edu> wrote:
> See this thread on the nahant mailing list for their most
> recent thoughts on this:
> Starts here:
>
https://www.redhat.com/archives/nahant-list/2005-June/msg00280.html
> Most relevant message (IMO):
>
https://www.redhat.com/archives/nahant-list/2005-June/msg00304.html
> Basically, they *really* like 4K stacks,

That's right.  It totally slipped my mind on that with 2.6.

To be honest, I haven't been deploying XFS as much since the
official XFS 1.2 on Red Hat Linux 7.x (plus one 1.3 on Red
Hat Linux 9).  The Red Hat Linux 7.x installs definitely have
the most time, and I very much trust and appreciate its
availability.

But my Fedora Core 3 tests have been positive, even if they
aren't official.  I have _avoided_ anything but the official
XFS releases prior, including the 2.4 backport.  I don't use
the 3rd party RPM sets and I don't use the CentOS Plus kernel
(although it probably is no less reliable than FC3's).

> XFS doesn't, and they won't go back to 8K or spend the time
> making XFS work with 4K.  

I was just happy with (so far) the capabilities of my few
Fedora Core 3 installs with XFS.  Seems to be handling
Quotas, XATTRs for ACLs/SELinux and NFS services without
issue.  So they are using 4K stacks, and it seems to work for
me on FC3's 2.6 implementation, but I haven't hammered it yet
like I have RHL7.x in the past.

BTW, I can understand Red Hat's insistance on sticking with
4K stacks because of the reduced issues for x86.  I remember
reading a bit awhile back when I had to deal with early 2.6
workstations (using nVidia's driver).

So, what are the "real issues" of using 4K stacks for XFS
(among others)?  Is it just a shift in the trust from the
proven Irix/MIPS history?  Or is it more than that?

> Oh, and, again, they don't see that XFS buys you over ext3.

That's gotta stop.  No offense, but that seems to be 100%
political statement.  I can't be the only consultant
deploying RHEL who constantly wishes he could offer clients a
way to backup all XATTRs on files without some added
procedures/hacks.  I mean, if Red Hat is pushing for SELinux
so damn hard, WTF can't they include a solid backup mechanism
for it on Ext3?

I just want to smack Red Hat silly on the existance of
xfsdump (and xfs_copy) and xfs_fsr among others.  Where are
the freak'n tools and support?  It's almost ironic that all
the "lack of kernel-interface/user-space support" Red Hat
claims about JFS and ReiserFS (which is true) is not applied
equally in reverse on Ext3!

And then there's the big damn size issue.  That's the kicker.
 I'm trying put to in >1.1TB (>1TiB) filesystems and I have
to start explaining the limitations to clients.  That just
eats right into the MS FUD / Windows pundit non-sense.

I'm not some ReiserFS puke or JFS enthusiast, I've got XFS in
production.  But it's all old systems.


-- 
Bryan J. Smith                | Sent from Yahoo Mail
mailto:b.j.smith@ieee.org     |  (please excuse any
http://thebs413.blogspot.com/ |   missing headers)