Date: Tue, 18 Oct 2011 18:21:01 +0200 From: Alfred Bartsch <bartsch@dssgmbh.de> To: freebsd-geom@freebsd.org Subject: Re: disk partitioning with gmirror + gpt + gjournal (RFC) Message-ID: <4E9DA76D.9080605@dssgmbh.de> In-Reply-To: <20111018100827.GA91400@mid.pc5.i.0x5.de> References: <4E69A152.6090408@rdtc.ru> <4E69EB15.50808@rdtc.ru> <4E9D2117.4090203@dssgmbh.de> <20111018081920.GA97840@mid.pc5.i.0x5.de> <4E9D49D2.8020801@dssgmbh.de> <20111018100827.GA91400@mid.pc5.i.0x5.de>
next in thread | previous in thread | raw e-mail | index | archive | help
Am 18.10.2011 12:08, schrieb Nicolas Rachinsky: > * Alfred Bartsch <bartsch@dssgmbh.de> [2011-10-18 11:41 +0200]: >> Thanks for pointing this out. I'm using filesystem labels >> (tunefs), no GEOM labels other than gmirror. > > ufs labels are handled by glabel. > >> After executing "gmirror remove gm0 <disk1>, the partitions on >> this disk show up as <disk1>p#, not as mirror/gm0p#, so there is >> IMHO no ambiguity. If you have experienced problems with gmirror >> - glabel configurations, I'm interested in more details. > > But <disk1>p# contains the same filesystem with the same label as > mirror/gm0p#. If glabel sees <disk1>p# on boot before mirror/gm0p# > it may associate ufs/<name> with <disk1>p# instead of > mirror/gm0p#. > > This can happen with "glabel label"-labels, ufs-labels and > gpt-labels. > Thanks for clarifying this. As I'm looking for a robust configuration, I will refrain from applying any of these labels until a consistent usage will be possible. So with some minor changes the configuration looks like ... create the UFS file systems (without anylabels): # newfs -J mirror/gm0p7.journal # newfs -J mirror/gm0p8.journal # newfs -J mirror/gm0p9.journal # newfs -J mirror/gm0p10.journal ... /etc/fstab could then look like # Device Mountpoint FStype Options Dump Pass# /dev/mirror/gm0p2 none swap sw 0 0 /dev/mirror/gm0p7.journal / ufs rw,noatime,async 1 1 /dev/mirror/gm0p10.journal /home ufs rw,noatime,async 2 2 /dev/mirror/gm0p9.journal /usr ufs rw,noatime,async 2 2 /dev/mirror/gm0p8.journal /var ufs rw,noatime,async 2 2 > > If you want verbose names, you might be able to use gconcat (with > one component) since gconcat can use hardcoded provider names, > which should avoid this problem. > > It would be nice to add the possibility to hardcode providers to > glabel. It's IMHO not the only weakness concerning glabel. Are there any other objections to this method of disk partitioning? -- Alfred Bartsch Data-Service GmbH
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4E9DA76D.9080605>