From owner-freebsd-questions@FreeBSD.ORG Sun Feb 1 20:07:52 2009 Return-Path: Delivered-To: freebsd-questions@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 3F6E7106564A for ; Sun, 1 Feb 2009 20:07:52 +0000 (UTC) (envelope-from AHamiltonWright@MtA.ca) Received: from mailserv.mta.ca (mailserv.mta.ca [138.73.1.1]) by mx1.freebsd.org (Postfix) with ESMTP id 0F1898FC27 for ; Sun, 1 Feb 2009 20:07:51 +0000 (UTC) (envelope-from AHamiltonWright@MtA.ca) Received: from [138.73.29.51] (helo=qemg.org) by mailserv.mta.ca with esmtp (Exim 4.61) (envelope-from ) id 1LTibd-0006sB-R1; Sun, 01 Feb 2009 16:07:50 -0400 Date: Sun, 1 Feb 2009 16:08:47 -0400 (AST) From: Andrew Hamilton-Wright Sender: andrew@qemg.org To: RW In-Reply-To: <20090201183128.524f16db@gumby.homeunix.com> Message-ID: References: <20090201183128.524f16db@gumby.homeunix.com> User-Agent: Alpine 2.00 (BSF 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Cc: freebsd-questions@freebsd.org Subject: Re: dump(8) using snapshot + "recommended" cache X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 01 Feb 2009 20:07:52 -0000 On Sun, 1 Feb 2009, RW wrote: >> ***It is recommended that you always use this option when >> dumping a snapshot.*** > > > When you dump a snapshot there are, by definition, no changes between > passes. So it's saying that in that case there in no reason not to > cache. Ah, that makes sense. That being the case, perhaps we can update the text to: If dumping from a snapshot, the filesystem is already frozen, therefore using a cache with a snapshot will ensure that consistency is maintained while also providing best performance. If that sounds good, I'll make a doc patch. Out of curiosity, under what circumstances is the improved performance the most likely? I dump from cron when the system usage is low, and haven't noticed any significant difference in time with or without cacheing -- but I haven't done any testing under heavy load, nor with limited RAM, so there are many mbufs available in any case. Thanks for the info, Andrew.