From owner-freebsd-questions Fri Mar 8 1:23: 3 2002 Delivered-To: freebsd-questions@freebsd.org Received: from beta.netcraft.com (beta.netcraft.com [195.92.95.67]) by hub.freebsd.org (Postfix) with ESMTP id 4C01A37B404 for ; Fri, 8 Mar 2002 01:22:51 -0800 (PST) Received: from localhost (jon@localhost) by beta.netcraft.com (8.11.6/8.11.4) with ESMTP id g289MgX78458; Fri, 8 Mar 2002 09:22:42 GMT (envelope-from jon@beta.netcraft.com) Date: Fri, 8 Mar 2002 09:22:42 +0000 (GMT) From: Jon Wilson To: Mike Meyer Cc: questions@FreeBSD.ORG Subject: Re: questions about dump on live filesystems In-Reply-To: <15496.5730.638025.424864@guru.mired.org> Message-ID: <20020308091152.X74454-100000@beta.netcraft.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG On Thu, 7 Mar 2002, Mike Meyer wrote: > > My question: what happens if an inode is freed and assigned to a different > > file during the dump process? Will I end up with an erroneous bit of data > > in my restored file? > > I haven't checked the code, you have. Not very well! (rusty C, laziness, etc!) > It sure sounds like that's what will happen. Y. > > Or can I at least rely on dump to leave things in a state such that (for > > instance) I will get a valid backup of the file at the next level-$n++ > > dump? A backup strategy of level-0 multi-user r/w, followed by a level-1 > > r/o is what I would like to be doing, but various people here have doubts > > about this producing valid backups 100% of the time. > > I think you've got those backwards. I admint that it's odd, but I don't see how it causes any major problems. The amount of data we are talking about is large, and on a high use system, so level-0 single-user takes some time. Minimising user disruption is a priority. > The thing to watch for is if the level 1 dump is done at the same time > as some critical file being updated on a daily basis. That means that > those files will be bad on every single daily dump, which is probably > not acceptable. This is my reasoning for doing the level 1 dump single-user / r.o. Even if my level-0 dump of the active fs has corrupt files due to reassigned inodes, the single user level-1 will catch these. Of course the level-0 multi-user backup can not be considered a "good" backup until the level-1 r/o is done, so it should be done asap. The same does not apply to your suggesuted level-0 r/o, level-1 r/w strategy, which is why mine sounds "backwards". > Actually, I recommend that you not do the level above level 0 as level > 1, but as at least level 2. The exact level will depend on your > system. Personally, I do level 0, 4, and 8, with 8's happening daily > and 4's weekly. I'm not following you here. According to my understanding, a level-($n+1) backup will include all files changed since the last level-$n. What is the point in incrementing by more than one in this way? Thanks for the comments. Jon -- Jon Wilson Netcraft Ltd. Tel: +44 (0)1225 867975 jon@netcraft.com Fax: +44 (0)1225 867700 http://www.netcraft.com Mob: +44 (0)7776 137939 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message