From owner-freebsd-geom@FreeBSD.ORG Mon Mar 24 03:39:58 2008 Return-Path: Delivered-To: freebsd-geom@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id BD807106566B for ; Mon, 24 Mar 2008 03:39:58 +0000 (UTC) (envelope-from Lee@dilkie.com) Received: from spock.dilkie.com (spock.dilkie.com [206.51.1.40]) by mx1.freebsd.org (Postfix) with ESMTP id 5A2FF8FC1C for ; Mon, 24 Mar 2008 03:39:58 +0000 (UTC) (envelope-from Lee@dilkie.com) Received: from [206.51.1.193] (borg.dilkie.com [206.51.1.193]) (authenticated bits=0) by spock.dilkie.com (8.14.2/8.14.2) with ESMTP id m2O3ccea079633 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Sun, 23 Mar 2008 23:38:40 -0400 (EDT) (envelope-from Lee@dilkie.com) Message-ID: <47E7223E.5070101@dilkie.com> Date: Sun, 23 Mar 2008 23:38:38 -0400 From: Lee Dilkie User-Agent: Thunderbird 2.0.0.12 (Windows/20080213) MIME-Version: 1.0 To: "Donald E. Holt" References: <1206284676.2073.28.camel@Hank.bullhouse.net> <47E6E34B.6000107@dilkie.com> <1206316676.33903.20.camel@Hank.bullhouse.net> In-Reply-To: <1206316676.33903.20.camel@Hank.bullhouse.net> X-Enigmail-Version: 0.95.6 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Scanned-By: MIMEDefang 2.63 on 206.51.1.40 Cc: GEOM Mailing List Subject: Re: Striped Drive for OS File System X-BeenThere: freebsd-geom@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: GEOM-specific discussions and implementations List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 24 Mar 2008 03:39:58 -0000 I think the gconcat vs gstripe decision boils down to your performance requirements. Gconcat lets you string together various size disks to make a really big disk, whereas gstripe requires them all the be the same size. Gstripe has better performance due to stripping, gconcat is only as fast as a drive. I'm using gconcat because this "big drive" holds a lot of video files and it's more important to just be able to go out and buy the lastest/cheapest drive and add to the end of the array when it gets full. I wouldn't put swap on a filesystem, it's better off in it's own partition/slice (I can never get that terminology straight). If the swap on your root drive is enough, you could go without swap on the /usr drive but it's up to you. oh, and after you've created your drive, don't forgot to add it to /etc/fstab (and remove the exiting /usr entry).. ie. /dev/concat/usr2_concat /usr2 ufs rw 2 2 Donald E. Holt wrote: > Lee, thanks so much for the great info. I may still bollix things up, > but at least I know that the operation itself isn't doomed from the > start... > > The man pages seem to be thorough, so I believe I can work out the > precise steps there. I do have two remaining questions, if I may beg > your patience. > > I am able to backup /usr to tape. In light of that fact, would you > recommend gstripe? I should do a bit more research on my own concerning > the differences between gstripe and gconcat prepared drives, I know, but > I'd value your advice. > > Second question. The drive that currently contains /usr also has a swap > partition (in addition to the swap partition on the first SCSI drive). I > assume the second swap isn't absolutely necessary, but would it be > worthwhile (or possible) to designate a portion of the concatenated > space to swap? > > Thanks again! > > On Sun, 2008-03-23 at 19:10 -0400, Lee Dilkie wrote: > > >> Donald, >> >> Yes, it can be done but as a gconcat rather than a gstripe if you need >> to expand in place. >> >> essentially, >> >> unmount /usr >> create gconcat from existing /usr and new drive (use raw drive.). >> then use growfs to expand gconcatted drive >> mount gconcatted version of /usr >> >> If you can do a backup and restore of /usr, you could create a stripped >> drive instead (if they are the same size or thereabouts). >> >> The only gotcha that I had was in FreeBSD 5.5. Growfs doesn't handle >> really big disks (which is what you get here). I found that the patch to >> fix growfs in 6.2 and manually did it and rebuilt growfs and it worked >> great. I've done it twice now, adding a third drive to my gconcat array. >> >> Also. I only used gconcat with the raw drives, ie "gconcat label >> big_drive /dev/ad2 /dev/ad3 /dev/ad4" but I presume it works the same >> with slices/partitions. >> >> -lee >> >> Donald E. Holt wrote: >> >>> I haven't seen this particular situation discussed, nor have I been able >>> to find any specific information on the web. I am not experienced in >>> GEOM or vinum, so I may be asking a dumb question. If so, I apologize. >>> >>> My system has FreeBSD installed across 2 SCSI drives. /dev/da2 >>> holds /, /tmp, /var, and a swap. /dev/da4 is devoted to /usr and another >>> swap. A third drive (dev/da5) is dedicated to a Win4BSD virtual machine, >>> and is mounted as /WinXP (there are still a few MS apps I can't get >>> around using). >>> >>> I'd like to expand /usr fs to an additional SCSI drive (physically >>> present now, but unused). Would it be possible to do that, perhaps with >>> gstripe, without distroying data on the current drive? >>> >>> Again, sorry if I've asked one of those eye-rolling, forehead-thumping >>> questions. >>> >>> Thanks! >>> _______________________________________________ >>> freebsd-geom@freebsd.org mailing list >>> http://lists.freebsd.org/mailman/listinfo/freebsd-geom >>> To unsubscribe, send any mail to "freebsd-geom-unsubscribe@freebsd.org" >>> >>> > _______________________________________________ > freebsd-geom@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-geom > To unsubscribe, send any mail to "freebsd-geom-unsubscribe@freebsd.org" >