From owner-freebsd-current@FreeBSD.ORG Sat Sep 18 21:48:27 2004 Return-Path: Delivered-To: freebsd-current@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 70E6916A4CE; Sat, 18 Sep 2004 21:48:27 +0000 (GMT) Received: from jagor.srce.hr (jagor.srce.hr [161.53.2.130]) by mx1.FreeBSD.org (Postfix) with ESMTP id BD42243D2F; Sat, 18 Sep 2004 21:48:26 +0000 (GMT) (envelope-from ivoras@fer.hr) Received: from [193.198.135.199] (cmung1977.cmu.carnet.hr [193.198.135.199]) by jagor.srce.hr (8.12.10/8.12.10) with ESMTP id i8ILmNaC019592; Sat, 18 Sep 2004 23:48:24 +0200 (CEST) Message-ID: <414CAD21.9060409@fer.hr> Date: Sat, 18 Sep 2004 23:48:17 +0200 From: Ivan Voras User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7) Gecko/20040502 Thunderbird/0.6 Mnenhy/0.6.0.104 X-Accept-Language: en MIME-Version: 1.0 To: Pawel Jakub Dawidek References: <414C2E85.6090904@fer.hr> <20040918185841.GA30151@darkness.comp.waw.pl> In-Reply-To: <20040918185841.GA30151@darkness.comp.waw.pl> X-Enigmail-Version: 0.84.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime X-Face: L+\SC&qzD^$mE"W7='[pvNVqN~%w%yE$u=Z8>3^$16dK+jG[@H`; lFz^h,f=uzlw01 fajy]=lHAh(S@'EmM3FbC-`HqOh!,fJFeBS$2JU3w-3WQr{$ADS`,'xm8>G0/7I{p61vqy+RMCwQDg xh'z9&s0V:n Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-Scanned-By: MIMEDefang 2.42 X-Virus-Scanned: by amavisd-new at jagor.srce.hr cc: current@FreeBSD.org Subject: Re: gmirror: DIRTY flag? X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 18 Sep 2004 21:48:27 -0000 Pawel Jakub Dawidek wrote: > You use ggate as one of your mirror components? Nice:) Indeed very nice. Not a single problem so far. :) > If you export this device over slow link, you can considern changing > balance algorithm to 'prefer', so only local component will be used for > reading. Yes, I tried several algorithms. This is on a test bench machines, and both the disks and network are very slow (udma33, 100mbit E.). The 'load' algorithm actually seems to work best in this setup. In a month or so we're planning to implement this on some beefy hardware, and I'll test it more then. (If anybody's interested, I'll post some "benchmark" results ;) ). > Remember, that 'prefer' algorithm is based on priority, so in your case > ggate0 will be used for reading as component with bigger priority. I was wondering about that: is bigger priority assigned to smaller or bigger number? The way I've expected it, since command line goes like this: # gmirror label myname device1 [device2...] I've expected that device1 would have bigger priority (semantically speaking). > Priority is also important for synchronization. If you have power > failure, component with the biggest priority will be used as a master > component, and all the rest components synchronize to him. What if the master fails? :) I'm expecting this: When device1 fails, I deactivate it from the mirror, (at this time device2 becomes master) and insert another (will it automagically gain priority 3 or I'll need to specify it?). The mirror resynchronises (transferring everything from device2 to the newly inserted one, thus killing the bandwidth for applications?), and device2 keeps the master status. > Command for changing priority for running mirror is missing, but you > can increase priority for local component by doing: > > # gmirror remove netmirror ad2s1e > # gmirror insert -p 2 netmirror ad2s1e > Ok. Since this is a test setup, it really isn't a problem to tear down and rebuild the mirror. Which leads to another question: Is it ok to rebuild mirrors and change device orders/priorities? E.g. if I do: # gmirror label myname device1 device2 # gmirror stop myname # gmirror label myname device2 device1 ... nothing extraordinary is supposed to happen? -- What part of "Ph'nglui mglw'nath Cthulhu R'lyeh wgah'nagl fhtagn" don't you understand?