Re: kernel choices?

On Wed, Jul 16, 2003 at 11:19:46AM -0400, Kurt Lieber wrote:
> On Tue, Jul 15, 2003 at 08:36:00PM -0700 or thereabouts, adam morley wrote:
> > I've been vacilating on what kernel to pick for a production server.  
> 
> You were going to get a lengthier reply, but then I saw kerframil took care
> of that. :)
> 
> So, I'll just second everything that kerframil said and say that I use WOLK
> on all my home machines.  We're also in the process of converting all of
> Gentoo's infrastructure servers over to WOLK 4.3 as well.  We would have
> done this sooner but there was some uncertainty about the future of WOLK.
> 
> My second choice would be vanilla-sources, patched with the drivers you
> require for your machines.  I recommend against using gentoo-sources for
> server usage as it *tends* to be geared more towards desktop stuffs.

Does this mean the ebuild will get updated so I don't have to manually download it?  That would be beyond neat.  By everyone else's silence, am I to take it that you're all running either vanilla custom patched or WOLK?

Here's another interesting question: XFS, JFS or ReiserFS?  I was thinking about XFS for a while, but for the install I don't need XFS's features, and the potential loss of data on system crash (is this still true?) is a no-no for me.  I was thinking JFS just because its made by IBM, seems fairly stable, etc.  I've got reiserfs running on a machine, and its never given me problems, so its really a tossup.

Is there a whitepaper or anything comparing the different filesystems?  Ie: filesystem X does better when you have 5000 small files in one directory, filesystem Y does better for large dataset (ie: huge sequential reads/writes)?  Because if some do better on small file single directory setups, I might use them for maildir storage and something else for everything else.

I'd planned on just using ext3 for root, usr, opt, var because those aren't going to get hit hard (at all really) and I'd like to be able to screw with them off a default linux rescue cd w/o too much trouble.

I'd also like to hear if you folks think its a bad idea to make seperate root, usr, opt and var partitions.  That's what we do with our commercial unixes, in part for disaster recovery stuff (dumping partitions to offsite tapes, one of the reasons for the choice of ext3 for the boot filesystems since it supports dump).

I'd probaly keep the portage tree, /usr/src tree, etc. on an NFS mount or a seperate FS and just mount -o bind it to the right place. . .

thanks,

-- 
adam

--
[email protected] mailing list