Re: XFS

Hi,


On Mon, Oct 25, 2004 at 11:19:31PM +0200, Bogdan Costescu wrote:
> On Mon, 25 Oct 2004, Pasi Pirhonen wrote:
> 
> > I am actually _very happy_ if you 'start beating' the XFS. More people
> > willing to do so, more like it can be said to be stable. 
> 
> Well, I'll be beating XFS for this particular kernel. I have used XFS 
> for about 2 years now and I'm very happy with it, but not on any of 
> the "very important servers" that I'm taking care of. Especially, I 
> have not used it much with NFS.
> 
> > I have been very moderate with this and only XFS itself is enabled as in
> > 
> > CONFIG_XFS_FS=m
> > # CONFIG_XFS_QUOTA is not set
> 
> Arrgh, then I'll have to rebuild it. I would like to have QUOTA 
> enabled as well (and funny given today's post on nahant-beta list...)

I've been testing mostly 'the XFS part' so far. For my own home LAN,
the quota sin't needed. I am about to enable that myself too, but i
really would like to hear if that is actually working.

For my setup, there is mostly one user (builder) accessing the
NFS-server. Few others, but nothing compared to setup where there are
hundreds of users. 

But NFS-server being still critcal part a when mainframe emulator
(hercules) is runnig a month for 'a respin', one is not able to make
NFS unavailable :)

I saw the rant on nahant list too and was tempted to say 'We, Tao Linux
do support XFS', but it's not place for stupid things like that.

Mostly (as i think David too) i just like to 'b e quiet' and if someone
picks up our release, that'll be good, but not trying to promote it any
way - it's there and if someone finds it useable, it makes me (us)
happy.

Definetly i really don't want to start dealing with that 'clueless
mass' which seems to be bothering WhiteBox (judged by the traffic on
mailing lists).

> 
> > I am not in progress of converting the / or /usr or /var to XFS, but i
> > really like having my NFS-exports as XFS for just pure performance and
> > no problems with 'sudden I/O stalls'.
> 
> Same here.

The funny part being that those I/O stalls haven't been that much
there and one could not get one with UP machine (i tried very, very,
very hard, but wasn't able to trigger it with UP kernel).

My current NFS-server _is_ UP and most of my older machines are UP too. 
I just can't stand the sluggish behaviour of UP anymore while something
else is happening - SMP is way to go :)

> 
> > The problem being that some of these 'extra features' would likely to be
> > needed as in 'installation media'. That efectively makes it impossible
> > to keep those as in /contrib/
> 
> For XFS it might not be that stringent, but for the devices...
> 

Yeah. XFS isn't such issue as i don't have any plans to make that 'as
installable target' for now. For device drivers it is as the sole purpose
for device drivers merge is to make those available as in installation
media. Network components being less critical than disk subsystem.

For ny NFS-server, i still have a AMI megaraid w/ 2x18GB SCSI mirrored
as boot media. I just don't like the idea having OS on the same
contoller where the 1.6TB data-pool is. So for myself, the installation
support of 3w-9xxx have _never_ been a issue at all (i could have
installed to another disk and dump the OS later on to 3w-9xxx w/
support), but there are these people that insist of installing to same
ocntroller :P



-- 
Pasi Pirhonen - [email protected] - http://iki.fi/upi/