From owner-freebsd-stable@FreeBSD.ORG Wed Apr 27 14:21:26 2005 Return-Path: Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 2236916A506 for ; Wed, 27 Apr 2005 14:21:26 +0000 (GMT) Received: from www.provisio.net (paco.to [209.31.146.75]) by mx1.FreeBSD.org (Postfix) with ESMTP id 272F343D31 for ; Wed, 27 Apr 2005 14:21:25 +0000 (GMT) (envelope-from paco@paco.to) Received: from www.provisio.net (localhost [127.0.0.1]) by www.provisio.net id j3RELMgs061073 envelope-from paco@paco.to for ; Wed, 27 Apr 2005 10:21:22 -0400 (EDT) (envelope-from paco@paco.to) Received: from localhost (paco@localhost)j3RELMaB061070 for ; Wed, 27 Apr 2005 10:21:22 -0400 (EDT) (envelope-from paco@paco.to) X-Authentication-Warning: www.provisio.net: paco owned process doing -bs Date: Wed, 27 Apr 2005 10:21:22 -0400 (EDT) From: Paco Hope X-X-Sender: paco@www.provisio.net To: freebsd-stable@freebsd.org In-Reply-To: <20050425211207.B43358@carver.gumbysoft.com> Message-ID: <20050427094252.M52102@www.provisio.net> References: <20050421021940.X10760@www.provisio.net> <20050425211207.B43358@carver.gumbysoft.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Provisio-MailScanner-Information: Contact Paco Hope phone: 703-606-1905 X-Provisio-MailScanner: Judged to be clean X-MailScanner-From: paco@paco.to Subject: Re: [SOLVED] installkernel succeeds, but old kernel boots X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 27 Apr 2005 14:21:26 -0000 In case anyone is curious, here's the answer. The information needed to solve this wasn't in my email. Look in the archives for the original problem description. The short and amusing answer: it was booting off a stale vinum drive, and then running off the current drive once vinum got loaded. Here's the long answer. I'm one of the unlucky ones who was using vinum in 5.1 and 5.2 only to discover that it broke in 5.3. I am booting off a vinum root, too. I have the a partition adjusted on both root drives so that it appears to be a normal root partition, but then the d partition is the real vinum partition. Well, here's what was going on. At boot time I would see: kernel: vinum: incompatible sector sizes. mroot.p1 has 0, mroot has 512. Ignored. My root vinum volume ("mroot") would come up OK, but one of its plexes (mroot.p1) would be down because its only subdisk was missing. For some reason, it would never see the config in /dev/ad1s1, it would never find the drive on /dev/ad1s1d and thus never start up the mroot.p1 plex. If I interactively ran vinum and typed "read /dev/ad1s1" it would find all the drives that are on ad1s1. I could individually start subdisks, which would refresh and life would be happy for a while. I was seeing some kernel panics, though, so I eventually quit manually starting up these drives. If I didn't actively mirror, I didn't have panics. Well, my 'make installkernel' was installing onto the live, active mirror of my root filesystem. For some reason, though, the boot selector on the console was trying (and succeeding) booting the second disk. Since it wasn't up-to-date, it still had a legit filesystem on it, but not one that had the right kernel. My 5.3 kernel would boot, load the vinum module, switch the root partition to the vinum root, and then run. I was dumbfounded at how the file I saw as /boot/kernel/kernel was not the kernel I booted. So I ran vinum and manually started up the subdisk that corresponded to my root partition. It synchronized. Then I rebooted. Two things are true now that make me happy: 1. I'm booting my 5.4-STABLE kernel. 2. vinum seems to be working. No panics yet and I no longer have this problem where the ad1s1 vinum partitions are lost at boot time. If I get time, I will probably switch to gmirror for mirroring these partitions, but I feel less pressure to do so, now that this seems to be working again. I hope this explanation helps someone. Paco