From owner-freebsd-performance@FreeBSD.ORG Mon Aug 25 21:49:16 2003 Return-Path: Delivered-To: freebsd-performance@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 0CA1316A4BF for ; Mon, 25 Aug 2003 21:49:16 -0700 (PDT) Received: from mail.lambertfam.org (www.lambertfam.org [216.223.208.55]) by mx1.FreeBSD.org (Postfix) with ESMTP id 449BF43FE3 for ; Mon, 25 Aug 2003 21:49:13 -0700 (PDT) (envelope-from lambert@lambertfam.org) Received: from laptop.lambertfam.org (laptop.int.lambertfam.org [10.1.0.2]) by mail.lambertfam.org (Postfix) with ESMTP id AC13F34D2D for ; Tue, 26 Aug 2003 00:48:00 -0400 (EDT) Received: by laptop.lambertfam.org (Postfix, from userid 1000) id 0B00C8879; Tue, 26 Aug 2003 00:48:56 -0400 (EDT) Date: Tue, 26 Aug 2003 00:48:56 -0400 From: Scott Lambert To: freebsd-performance@freebsd.org Message-ID: <20030826044855.GB8831@laptop.lambertfam.org> Mail-Followup-To: freebsd-performance@freebsd.org References: <200308250929.32143.paul@pathiakis.com> <20030826030527.GB1514@perrin.nxad.com> <1061869321.3193.9.camel@jester> <20030826042046.GC1514@perrin.nxad.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable In-Reply-To: <20030826042046.GC1514@perrin.nxad.com> User-Agent: Mutt/1.4.1i Subject: Re: Tuning Postgresql on FreeBSD 5.1 X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 26 Aug 2003 04:49:16 -0000 On Mon, Aug 25, 2003 at 09:20:46PM -0700, Sean Chittenden wrote: > > > while still maintaining reasonable file system performance (ext2 is > > > faster than ext3 by a wide margin, but ext2 is _not_ a reliable FS). > >=20 > > > > #fsync =3D true > > >=20 > > > Change this to false. > >=20 > > This is a funny thing to suggest after the filesystem comment. >=20 > :) Did you enjoy that? :) I should clarify that piece of "advice" > though. >=20 > If you're _sure_ power isn't going to fail (using a UPS) and make > regular backups (standard operating procedure), then fsync =3D false > can be a good thing in terms of performance (roughly speaking, it > > keep fsync =3D true. If, however, you have mitigated the risks to > levels that your business/operation is comfortable with (your data > center is on generators, your DB systems have UPSs, are using RAID, > and make frequent backups), then I would set fsync =3D false for most > noncritical apps (non-financial, non-medical). I've had more UPS failures than grid failures in the past year and a half. In the blackout, one of my APCs smoked rather than do it's job the second time the power went down. The UPS maintainance people at on of the colo facilities I use managed to drop power to everything while doing routine maintenance on their UPS. That left me with a lot of unexpected softupdates inconsistencies. And three machines that dropped to single user mode on restart. One machine's disk didn't survive. We lost a lot of files on the other two systems. The sad part being that the backups had just started for the day when they burped the UPS. Two hours later and it wouldn't have hurt as bad. =20 A UPS doesn't necesarily reduce your likelyhood of having a power failure. Two fully online UPS's and redundant power supplies in each machine might. =20 BTW, e-mail and web hosting for Grandma and Aunt Bessie and Uncle Joe are critical services too. No, really, I'm serious. --=20 Scott Lambert KC5MLE Unix SysAdmin lambert@lambertfam.org =20