From: Eric Sandeen Subject: Re: [PATCH 0/4] (RESEND) ext3[34] barrier changes Date: Wed, 21 May 2008 13:15:12 -0500 Message-ID: <483466B0.6000606@redhat.com> References: <482DDA56.6000301@redhat.com> <20080518211140.b29bee30.akpm@linux-foundation.org> <200805191316.27551.chris.mason@oracle.com> <200805191439.36577.chris.mason@oracle.com> <20080521112224.GD5028@ucw.cz> <20080521110324.668048e0.akpm@linux-foundation.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: Pavel Machek , Chris Mason , Theodore Tso , Andi Kleen , linux-ext4@vger.kernel.org, linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org To: Andrew Morton Return-path: Received: from mx1.redhat.com ([66.187.233.31]:33621 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756902AbYEUSUM (ORCPT ); Wed, 21 May 2008 14:20:12 -0400 In-Reply-To: <20080521110324.668048e0.akpm@linux-foundation.org> Sender: linux-ext4-owner@vger.kernel.org List-ID: Andrew Morton wrote: > On Wed, 21 May 2008 13:22:25 +0200 Pavel Machek wrote: >>> I tested this one with a larger FS (40GB instead of 2GB) and larger log (128MB >>> instead of 32MB). barrier-test -s 32 -p 1500 was still able to get a 50% >>> corruption rate on the larger FS. >> Ok, Andrew, is this enough to get barrier patch applied and stop >> corrupting data in default config, or do you want some more testing? >> >> I guess 20% benchmark regression is bad, but seldom and impossible to >> debug data corruption is worse... > > It is 20%? I recall 30% from a few years ago, but that's vague and it > might have changed. Has much quantitative testing been done recently? > I might have missed it. > > If we do make this change I think it should be accompanied by noisy > printks so that as many people as possible know about the decision > which we just made for them. > > afaik there is no need to enable this feature if the machine (actually > the disks) are on a UPS, yes? As long as your power supply (or your UPS) doesn't go boom, I suppose so. It is too bad that there is no way to determine no-barrier safety from software. (maybe apcupsd could do something... ;) I guess it's levels of confidence. I agree that a user education campaign is probably in order... maybe if this thread is long enough to make LWN it'll raise some awareness. :) -Eric