From owner-freebsd-current@FreeBSD.ORG Sat Jul 12 21:04:44 2003 Return-Path: Delivered-To: freebsd-current@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id E44D437B401 for ; Sat, 12 Jul 2003 21:04:44 -0700 (PDT) Received: from c001.snv.cp.net (h001.c001.snv.cp.net [209.228.32.115]) by mx1.FreeBSD.org (Postfix) with SMTP id 524FB43FBF for ; Sat, 12 Jul 2003 21:04:43 -0700 (PDT) (envelope-from jeff@walters.name) Received: (cpmta 6580 invoked from network); 12 Jul 2003 21:04:40 -0700 Received: from 68.59.45.70 (HELO 10.0.1.51) by smtp.register-admin.com (209.228.32.115) with SMTP; 12 Jul 2003 21:04:40 -0700 X-Sent: 13 Jul 2003 04:04:40 GMT From: Jeff Walters To: Sean Kelly , Don Lewis Date: Sun, 13 Jul 2003 00:04:41 -0400 User-Agent: KMail/1.5.2 References: <200307130136.h6D1arM7023826@gw.catspoiler.org> <20030713032425.GA77413@edgemaster.zombie.org> In-Reply-To: <20030713032425.GA77413@edgemaster.zombie.org> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200307130004.41123.jeff@walters.name> cc: current@FreeBSD.org Subject: Re: sporadic disk syncing failures when shutting down X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list Reply-To: jeff@walters.name List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 13 Jul 2003 04:04:45 -0000 On Saturday 12 July 2003 11:24 pm, Sean Kelly wrote: > > syncing disks, buffers remaining... 54 54 54 54 54 54 54 54 54 54 54 54 > > 54 54 54 54 54 54 54 54 giving up on 54 buffers > > Uptime: 6m42s > > Terminate ACPI > > Rebooting... > > > > Each time this has happened, fsck finds and nukes a bunch of empty > > directories. The last time this happened, the /etc/rc.d/yp* files that > > mergmaster updated were missing after the reboot and fsck had done its > > work. Nothing has ever shown up in lost+found. > > > > Has anyone else seen this? > > I have seen this a lot, but not as much lately. Sadly, I don't have any > more data than you on why it happens. I've seen it give up on a rather > frighteningly large number of buffers before, though... I hate to even mentioned such an unscientific observation where I made multiple changes at once, but I'll provide a data point. I also saw this problem crop up at the same time as I tried the SCHED_ULE scheduler a couple of months ago. I re-cvsup'ed CURRENT and switched back to SCHED_4BSD and it went away.