Date: Mon, 29 Nov 2004 14:38:25 +0100 From: Pawel Jakub Dawidek <pjd@FreeBSD.org> To: Chris Hedley <cbh-freebsd-current@groups.chrishedley.com> Cc: marcel@freebsd.org Subject: Re: GEOM: gpt partitions on a gmirror array possible? Message-ID: <20041129133825.GL7232@darkness.comp.waw.pl> In-Reply-To: <20041127193532.X15946@teapot.cbhnet> References: <20041127193532.X15946@teapot.cbhnet>
next in thread | previous in thread | raw e-mail | index | archive | help
[-- Attachment #1 --] On Sat, Nov 27, 2004 at 07:58:35PM +0000, Chris Hedley wrote: +> Hi all, +> +> A hopefully quick question! I have looked for a solution for (or +> discussion about) this, if I've just failed to find it, please point me in +> the right direction. +> +> This is what I'm trying to do: I've created a RAID-1 array, m1, using +> gmirror on two scsi units, da0 and da1, and I'd like to chop it up into +> logical partitions using gpt, as I understand (correctly, hopefully) there +> are leanings toward gpt rather than bsdlabel as being the correct way to +> do things. Now when I initialise and create my gpt partition on +> mirror/m1, I get no new entries in /dev/mirror but I _do_ get separate +> /dev/da0p1 and /dev/da1p1 entries appearing. I've probably completely +> failed to understand the whole concept of GEOM tasting and unprejudiced +> hierarchy and so on as I decided I may as well try to create my UFS2 +> filesystem on /dev/da0p1 in /dev/mirror/m1p1's absence and see if +> /dev/da1p1 magically follows suit. Needless to say it doesn't, I just end +> up with a dirty RAID-1 disc that needs resynchronising, with da0p1 +> (un)mysteriously disappearing after the newfs. +> +> Is there a correct way to do this, or am I once again guilty of trying to +> use something that isn't quite ready yet? Perhaps I should do it the +> other way around and use gpt first and create one gmirror per partition; +> that solution didn't "feel" right for some reason, but I can't really +> quantify that particularly as, if I understand the GEOM documentation +> correctly, multiple gmirrors with their respective paritions on the same +> set of discs shouldn't compete with each for access. +> +> My system is FreeBSD/amd64, 6.0-current (Oct 25th vintage) using ahc for +> my test scsi discs and aac for my active discs. It is because GEOM_GPT class only allow to create GPT labels on rank#1 providers (i.e. disks). I'm not sure why we have this hack, maybe marcel@ knows something more (cc'ed). -- Pawel Jakub Dawidek http://www.FreeBSD.org pjd@FreeBSD.org http://garage.freebsd.pl FreeBSD committer Am I Evil? Yes, I Am! [-- Attachment #2 --] -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (FreeBSD) iD8DBQFBqyZRForvXbEpPzQRAhgSAKCPgwYs4PwT6KcaepgxWoRPQczJxgCcCVnt hLeVVz+ToYjPmCor5zR4b7M= =Eb4i -----END PGP SIGNATURE-----
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20041129133825.GL7232>