Date: Mon, 22 Jan 2007 07:42:36 -0600 From: Doug Poland <doug@polands.org> To: Robert Huff <roberthuff@rcn.com> Cc: freebsd-questions@freebsd.org Subject: Re: more than 7 partitions on a SCSI-drive Message-ID: <20070122134235.GB1382@polands.org> In-Reply-To: <17844.85.335537.317957@jerusalem.litteratus.org> References: <ep0jcf$1meb$10@nermal.rz1.convenimus.net> <a969fbd10701211254ha01cb66q4ca4fe474c0dfdb@mail.gmail.com> <45B3E0D0.70005@u.washington.edu> <17844.85.335537.317957@jerusalem.litteratus.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Jan 21, 2007 at 07:07:49PM -0500, Robert Huff wrote: > Garrett Cooper writes: > > > One good reason I can think of is to partition (not the tech > > definition but the traditional definition, "to divide") filesystems > > such that if one person fills up "/", it won't cause a program that > > needs to write to "/var" or "/tmp" problems, which in the case of > > "/var" can bring down entire systems and infrastructures (happened > > before where I was working as IT when a CUPS server ran out of space > > on /var). Other than that.. not really sure. Maybe some of the > > older guard on the list know why. > I've had a /usr partition corrupt on a drive and I was able to boot into single user mode, mark the /usr partition read-only and mount it. I ran on a crippled but functional /usr partition until I could replace the drive. That would not have been possible with one large partition. > N) Dump - the preferred beckup method - works at the partition level. > Sure, you can flag files and directories "nodump" using chflags ... > but do you really want to manage that given modern disk sizes? > Also, file system quotas and snapshots work at the partition level. I manage an "all-in-one" server for a small non-profit. This is a rather "extreme" example of partitioning and, admittedly, I did it as an experiment, but you'll see my logic when you think about how dumps, quotas, snapshots work at the partition level. # Device Mountpoint FStype Options Dump Pass# /dev/da0s1b none swap sw 0 0 /dev/da0s1a / ufs rw 1 1 /dev/da0s2b /data ufs rw,groupquota 2 2 /dev/da0s2a /home ufs rw,groupquota 2 2 /dev/da0s1d /tmp ufs rw 2 2 /dev/da0s2h /usr ufs rw 2 2 /dev/da0s2d /usr/local ufs rw 2 2 /dev/da0s2g /usr/local/www ufs rw 2 2 /dev/da0s1h /usr/obj ufs noatime,async,rw 2 2 /dev/da0s2f /usr/ports ufs noatime,async,rw 2 2 /dev/da0s2e /usr/src ufs noatime,async,rw 2 2 /dev/da0s1e /var ufs rw 2 2 /dev/da0s1g /var/db ufs rw 2 2 /dev/da0s1f /var/mail ufs rw,groupquota 2 2 /dev/acd0 /cdrom cd9660 ro,noauto 0 0 snapshot_schedule="/,/usr:1:0:0 /usr/local,/usr/local/www:1:2:0 /var/db:2:1:0 /data,/var:0:1:0 /var/mail:1:2:0 /home:0:1:0" Using this partitioning scheme, I could also optimize /usr/obj /usr/src and /usr/ports for building. Since the underlying hardware is 3Ware RAID-10 on about 300GB of space, I had plenty of room with which to work. My only regret is making /usr/obj only 800MB in size. It gives me enough room for building world and ONE kernel, but not two kernels at one time. -- Regards, Doug
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070122134235.GB1382>