From owner-freebsd-cluster@FreeBSD.ORG Wed Jun 18 00:16:48 2003 Return-Path: Delivered-To: freebsd-cluster@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id C28A737B401 for ; Wed, 18 Jun 2003 00:16:48 -0700 (PDT) Received: from moutng.kundenserver.de (moutng.kundenserver.de [212.227.126.189]) by mx1.FreeBSD.org (Postfix) with ESMTP id 9A9BE43FAF for ; Wed, 18 Jun 2003 00:16:47 -0700 (PDT) (envelope-from sporner@nentec.de) Received: from [212.227.126.155] (helo=mrelayng.kundenserver.de) by moutng.kundenserver.de with esmtp (Exim 3.35 #1) id 19SXAH-0003o3-00; Wed, 18 Jun 2003 09:15:29 +0200 Received: from [80.131.138.80] (helo=gate.nentec.de) (TLSv1:EDH-RSA-DES-CBC3-SHA:168) (Exim 3.35 #1) id 19SXAH-0005RH-00; Wed, 18 Jun 2003 09:15:29 +0200 Received: from nenny.nentec.de (nenny.nentec.de [153.92.64.1]) by gate.nentec.de (8.11.3/) with ESMTP id h5I7FQi27642; Wed, 18 Jun 2003 09:15:26 +0200 Received: from nentec.de (andromeda.nentec.de [153.92.64.34]) by nenny.nentec.de (8.11.3/8.11.3) with ESMTP id h5I7FK023331; Wed, 18 Jun 2003 09:15:20 +0200 Message-ID: <3EF01187.7010709@nentec.de> Date: Wed, 18 Jun 2003 09:15:19 +0200 From: Andy Sporner User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2a) Gecko/20020910 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Michael Grant References: <20030617184938.GA1078@grant.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by AMaViS-perl11-milter (http://amavis.org/) cc: freebsd-cluster@freebsd.org Subject: Re: iSCSI and clustering with FreeBSD X-BeenThere: freebsd-cluster@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Clustering FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 18 Jun 2003 07:16:49 -0000 Michael Grant wrote: >It looks like no matter what I do, I need a second box next to the >first one to redirect packets to the other box if one of the boxes >goes down. I'd probably do this with NAT or an ip tunnel. This >second box almost makes it seem not worthwhile to put the other boxes >in different ISPs. Anyone have better ideas? > > > This is only a problem with migrating processes. A long time ago (and if you do a search in the IBM "linux" knowledgebase you might find it (ca 1996)) that a resolver protocol would be a good idea that is application based. Kind of like a DNS for applications. This would solve the problem. I had the idea to actually make a patch that would put this into the "connect" primitive in the socket layer so that it would be transparent to the user. HTTP redirect does most of it, but not every thing. For process Migration it is a problem however, since each machine at a minimum has a unique IP address. Therefore some sort of translation is needed to balance the traffic. If there be risk seeking beta testers there, I have somewhere handy a version of my software that has a load balancing NAT in the kernel. It is limited to TCP in the moment. There are some other minor limitations as well which are in the process of being fixed. But if anyone wishes to try it, send me a mail and I will send a small (~200K) tar file with code. In short, I think as long as you wish to redirect applications you are sort of stuck with a middle box. Either that or get a special kind of ethernet switch that does this (ahem ahem ahem... ;-) in hardware and then you don't waste a computer. Andy