From owner-freebsd-hackers Sat May 29 5:55:56 1999 Delivered-To: freebsd-hackers@freebsd.org Received: from nothing-going-on.demon.co.uk (nothing-going-on.demon.co.uk [193.237.89.66]) by hub.freebsd.org (Postfix) with ESMTP id AB62A14FC5 for ; Sat, 29 May 1999 05:55:38 -0700 (PDT) (envelope-from nik@nothing-going-on.demon.co.uk) Received: (from nik@localhost) by nothing-going-on.demon.co.uk (8.9.2/8.9.2) id KAA80497; Sat, 29 May 1999 10:54:23 +0100 (BST) (envelope-from nik) Date: Sat, 29 May 1999 10:54:21 +0100 From: Nik Clayton To: Joel Ray Holveck Cc: Don Lewis , Graeme Tait , freebsd-hackers@FreeBSD.ORG, info@boatbooks.com Subject: Re: FS tuning (Was: File system gets too fragmented ???) Message-ID: <19990529105420.A78812@catkin.nothing-going-on.org> References: <199905271415.HAA10721@salsa.gv.tsc.tdk.com> <86lne8h3gj.fsf@detlev.UUCP> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Mailer: Mutt 0.95.4i In-Reply-To: <86lne8h3gj.fsf@detlev.UUCP>; from Joel Ray Holveck on Sat, May 29, 1999 at 12:03:54AM -0500 Organization: Nik at home, where there's nothing going on Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Sat, May 29, 1999 at 12:03:54AM -0500, Joel Ray Holveck wrote: > How do people like to set up their filesystems these days? I've heard > of people who like one big fs (not generally usable anymore because of > the 1024 cyl limit), others who like the small root fs and one big fs > for everything else, and some who like separate fs's for different > things. All other things (disk speed, etc) being equal, what's this > groups' opinion? Depends on the disk size. My FreeBSD box has 2 disks at the moment, 1 x 2GB and 1 x 4GB. They're set up as follows (da0 = 2GB, da1 = 4GB) da0s1a / 50MB da0s1e /var 50MB da0s1f /usr 200MB da0s1g /local/0 1700MB da1s1a /local/1 4000MB /local/* is my mount point for disks in general. The first disk is a special case, as it has three 'special' filesystems on it (/, /var, /usr). I sized these based on general expectations on what the box would be doing, and in the knowledge that they would be fairly static. A look in /var shows; tmp@ -> /local/1/var/tmp A look in /usr shows; X11R6@ -> /local/0/usr/X11R6 local -> /local/1/usr/local obj -> /local/1/usr/obj ports -> /local/0/usr/ports src -> /local/0/usr/src A look in /home shows; ncvs -> /local/1/home/ncvs nik -> /local/1/home/nik operator -> /local/1/home/operator These are all directories that are going to grow in size over the lifetime of the box. So making them filesystems in the BSD sense is foolish, as you end up having to either resize the filesystems (which is a pain) or start symlinking between two or more filesystems. This approach makes it very easy to put different directory trees on different disks -- when I had one disk, /usr/src and /usr/obj were on the same disk, putting one of them on to the second disk was a snap. It's also pretty easy to automate. A script in /usr/local/etc/rc.d runs This is a slightly simplified version of an approach used at my current employer, where they have systems that often have 15 or more disks attached. And instead of using symlinks they use AMD to NFS mount the directories in the right place. But the basic principle is the same, and it scales very well. N -- The trial continues tomorrow. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message