From owner-freebsd-cluster Tue Dec 10 5:26:44 2002 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 D4C3837B401 for ; Tue, 10 Dec 2002 05:26:42 -0800 (PST) Received: from gate.nentec.de (gate2.nentec.de [194.25.215.66]) by mx1.FreeBSD.org (Postfix) with ESMTP id 14C8543EC2 for ; Tue, 10 Dec 2002 05:26:41 -0800 (PST) (envelope-from sporner@nentec.de) Received: from nenny.nentec.de (root@nenny.nentec.de [153.92.64.1]) by gate.nentec.de (8.11.3/8.9.3) with ESMTP id gBADQcP04700 for ; Tue, 10 Dec 2002 14:26:38 +0100 Received: from nentec.de (andromeda.nentec.de [153.92.64.34]) by nenny.nentec.de (8.11.3/8.11.3) with ESMTP id gBADQWt30394 for ; Tue, 10 Dec 2002 14:26:32 +0100 Message-ID: <3DF5EB88.9090409@nentec.de> Date: Tue, 10 Dec 2002 14:26:32 +0100 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 Cc: freebsd-cluster@FreeBSD.ORG Subject: Re: sharing files within a cluster References: <200212101257.gBACvv609153@splat.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/) Sender: owner-freebsd-cluster@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Michael Grant wrote: >Bob Bishop wrote: > > >>For another alternative approach, have a look at >>http://www.complang.tuwien.ac.at/reisner/drbd/ >>Linux only at the moment, but lighter-weight than Coda and directly >>addresses this problem. >> >> > >Good try, but no dice, this is a fail-over solution only. You can >only mount the disk read-write on one of the boxes at a time. > >However, there's a link on the page you pointed me at to GFS the >Global File System by Sistina: > >http://www.sistina.com/products_gfs.htm > >This seems to be similar to Coda but for Linux. I hate to say this >but it does look like Linux has superior support for clustering. > >Michael Grant > >To Unsubscribe: send mail to majordomo@FreeBSD.org >with "unsubscribe freebsd-cluster" in the body of the message > > > > However, what's wrong with this if you also failover NFS??? I wrote this to another recipient privately, but I put this here too as a possible scenario... I am also looking into the porting of the drbd code that Bob Bishop mentioned earlier. I am not a real "Fan" of the idea of shared SCSI. I have seen it work and have worked on such systems, but it has some very significant drawbacks. I like the idea of a network distributed collection of raw devices that can be brought together in a VINUM kind of way so that redundancy is assured. I have no problem with NFS as long as the underlying backing store is redundant. Then the NFS server can fail to anywhere in the cluster. Take this example. Node 1, 2, 3 and 4 have a 2 GB slice of raw space that is network exported (via DRBD). These slices are present on all of the cluster nodes via the network interface. (I don't know Vinum terms so I use Sequent ones (varitas)). We create a subdisk on all or part of these raw slices. Now we create two plexes by concatenating the subdisks on nodes 1 and 3. We create a corresponding plex on nodes 2 and 4. Now we create a mirrored volume by attaching both plexes. Now Node 1 becomes the NFS server. He has half of the volume local and the other half remote. He can serve any file to any other node in the cluster and assure that the locking paradymns work correctly. If he should die, than any of the other nodes can recover the NFS server and continue to operate. For instance if Node 2 takes it over it has the mirrored half of what was on node 1 and the other half is available over the network. The NFS server than is configured on a virtual IP address that fails around the cluster. With a journaled filesystem underlying the volume, it is coherent if not all of the mirroring was completed when node 1 took the fall. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-cluster" in the body of the message