From owner-freebsd-stable@FreeBSD.ORG Thu Nov 26 13:48:47 2009 Return-Path: Delivered-To: freebsd-stable@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id CBDB9106568F for ; Thu, 26 Nov 2009 13:48:47 +0000 (UTC) (envelope-from jdc@koitsu.dyndns.org) Received: from QMTA03.westchester.pa.mail.comcast.net (qmta03.westchester.pa.mail.comcast.net [76.96.62.32]) by mx1.freebsd.org (Postfix) with ESMTP id 5EDB68FC27 for ; Thu, 26 Nov 2009 13:48:47 +0000 (UTC) Received: from OMTA01.westchester.pa.mail.comcast.net ([76.96.62.11]) by QMTA03.westchester.pa.mail.comcast.net with comcast id 9pW41d0050EZKEL53pon09; Thu, 26 Nov 2009 13:48:47 +0000 Received: from koitsu.dyndns.org ([98.248.46.159]) by OMTA01.westchester.pa.mail.comcast.net with comcast id 9pom1d0073S48mS3MponcD; Thu, 26 Nov 2009 13:48:47 +0000 Received: by icarus.home.lan (Postfix, from userid 1000) id 94F811E3035; Thu, 26 Nov 2009 05:48:45 -0800 (PST) Date: Thu, 26 Nov 2009 05:48:45 -0800 From: Jeremy Chadwick To: freebsd-stable@freebsd.org Message-ID: <20091126134845.GA57738@icarus.home.lan> References: <4B0E3ABA.3030606@tzim.net> <790a9fff0911260040i1456d7c0j4f8327d24d2966cf@mail.gmail.com> <20091126085034.GA51998@icarus.home.lan> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.20 (2009-06-14) Subject: Re: Can't use gpt labels re-importing pool X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 26 Nov 2009 13:48:48 -0000 On Thu, Nov 26, 2009 at 12:56:50PM +0100, Ivan Voras wrote: > Jeremy Chadwick wrote: > > >Why are people bothering with GPT labels (or in some cases, glabels) > >when AHCI (whether it be ataahci.ko or ahci.ko) is in use? Under what > >circumstance would the device name change dynamically in this situation? > > > >I've never witnessed this happening with AHCI, at least on Intel > >systems, and I've hot-swapped hard disks many times over. > > It isn't specific to AHCI, but this is where most people encountered > it first. The issue in question is "how to make ZFS survive drive > renaming" for any cause - driver change, controller change, drive > shuffling, etc. > > In general, ZFS will taste individual drives on "zpool import" so > will try to do the right thing, but it might pick up different > labels for different drives, etc. Using glabel tricks simply makes > the naming a bit more consistent. What I'm having trouble understanding is where the "more consistent" concept comes from. I guess it ultimately depends on one's system configuration. For example: Tom Evans' situation greatly benefits from use of labels, since his configuration consists of 1) multiple SATA controllers, and 2) moving drives around between different controllers. This isn't going to happen in most production server environments, where the there is a static number of disks and (usually) a single controller. This is the "situation" I was referring to; "environment" or "configuration" would have been a more accurate choice of word. On 4-disk Supermicro systems (Intel ICHx + AHCI based, with hot-swap drive bays, with FreeBSD 7.x/8.x and ataahci), depending on what ICHx ports the drives are plugged into, your drive bays/disks are going to be static/consistent. SATA port #0 = ad4, SATA port #1 = ad6, and so on. These won't change unless you do something like, say, disable a SATA port or disable AHCI mode in the BIOS. Regardless, I'm almost certain I've made the mistake on a 4-disk system of doing (physical) system cleaning, which involves dusting out all of the bays and so on -- and ended up inserting a drive/carrier/disk into a bay which it wasn't part of prior to the shutdown. "zpool import" took care of things for me. If someone wants me to validate my own memory (the more I work Grave shift the more I find my memory failing me...), I'd be more than happy to spend a weekend messing around moving disks + reporting back what the behaviour is on RELENG_8. I have a "spare" 5-disk (6 ports) system which can be used for this purpose (Supermicro X7SBA + 5 disks). I spend most of my time messing around with disks at my night job as is... :-) -- | Jeremy Chadwick jdc@parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP: 4BD6C0CB |