From owner-freebsd-fs@FreeBSD.ORG Wed Apr 13 15:35:42 2011 Return-Path: Delivered-To: freebsd-fs@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id A27F41065672 for ; Wed, 13 Apr 2011 15:35:42 +0000 (UTC) (envelope-from mckusick@mckusick.com) Received: from chez.mckusick.com (chez.mckusick.com [64.81.247.49]) by mx1.freebsd.org (Postfix) with ESMTP id 59CC18FC13 for ; Wed, 13 Apr 2011 15:35:40 +0000 (UTC) Received: from chez.mckusick.com (localhost [127.0.0.1]) by chez.mckusick.com (8.14.3/8.14.3) with ESMTP id p3DF5tPa015944; Wed, 13 Apr 2011 08:05:56 -0700 (PDT) (envelope-from mckusick@chez.mckusick.com) Message-Id: <201104131505.p3DF5tPa015944@chez.mckusick.com> To: Sergi Seira In-reply-to: <4DA45042.5060204@cdmon.com> Date: Wed, 13 Apr 2011 08:05:55 -0700 From: Kirk McKusick X-Spam-Status: No, score=0.0 required=5.0 tests=MISSING_MID, UNPARSEABLE_RELAY autolearn=failed version=3.2.5 X-Spam-Checker-Version: SpamAssassin 3.2.5 (2008-06-10) on chez.mckusick.com Cc: freebsd-fs@freebsd.org Subject: Re: background fsck high load on 8.1 X-BeenThere: freebsd-fs@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Filesystems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 13 Apr 2011 15:35:42 -0000 > Date: Tue, 12 Apr 2011 15:14:42 +0200 > From: Sergi Seira > To: freebsd-current@freebsd.org > Subject: background fsck high load on 8.1 > > Hello, > > We've experienced that background fsck on 8.1 degrades server > performance on a higher degree than in previous fbsd versions (6.3, > 7.3; amd64). > > We've noticed it after upgrading - same hardware - to a 8.1-RELEASE. > Now, performance of other services (i.e. apache, mysql) during a > background fsck falls miserably. > > Is there any way to calm fsck down?, nice(1)?, some sysctl? > > We have also gmirror, but we prevent to rebuild it if there is a > fsck running in background. > > Thanks for your help, > regards, > Sergi This is the first report that I have heard of the higher level of slowdown by background fsck on 8.X systems. My guess is that it comes about from work done to make the I/O subsystem faster which in turn allows fsck to have a higher impact. For a period we were working on a kernel feature (associated with nice) that would allow the system to throttle I/O activity for a given process. It turns out to be a much harder problem than it at first appears. We set this aside when we got journaled soft updates working as they eliminate the need for background fsck. While journaled soft updates are too big of a change to be merged back into the 8.X tree, we do expect them to be used by default on 9.X systems. Kirk McKusick