Hi there, I asked the same question some time ago and received no suitable answer so far.
DRBD [1] does no "proper" replication over three nodes; it's basically still a Two-Node-RAID-1 with a third node, which doesn't really take part in the cluster but receives replication data as kind of a "backup" node. What we would need was some kind of node-based RAID-5. As of now, I have no solution, but some ideas where to look for one: http://www.gluster.org/ I haven't spent much time with working out how it works, but as far as I figured out, glusterfs can keep data consistent over more than two nodes. See also: http://en.wikipedia.org/wiki/GlusterFS and: http://en.wikipedia.org/wiki/List_of_file_systems#Distributed_parallel_fault-tolerant_file_systems Cheers, Andreas [1] The DRBB folks might please correct me, if I'm wrong. -- CONET Solutions GmbH Andreas Stallmann, Theodor-Heuss-Allee 19, 53773 Hennef Tel.: +49 2242 939-677, Fax: +49 2242 939-393 Mobil: +49 172 2455051 Internet: http://www.conet.de, mailto: astallm...@conet.de -----Ursprüngliche Nachricht----- Von: linux-ha-boun...@lists.linux-ha.org [mailto:linux-ha-boun...@lists.linux-ha.org] Im Auftrag von Miles Fidelman Gesendet: Montag, 4. April 2011 03:55 An: General Linux-HA mailing list Betreff: [Linux-HA] 3+node clusters? Hi, I'm running a 2-node HA cluster - Xen, Pacemaker, Corosync, DRBD - fairly traditional. But... I'm getting ready to add a new node or two, and I'm starting to wonder how one might approach 3-node failover - replicating data across three nodes, with two levels of failover for a VM -- which should also make it a lot easier to replace hardware and upgrade software without sacrificing downtime. It sort of looks like DRBD stacking can replicate data across three nodes, but it looks a little baroque, and it's not at all clear how to set up a 3-node failover strategy. It doesn't look like the various configurations provide for the same flexibility in specifying spare nodes, akin to the way that md raid configuration gives flexibility in the number of active and spare drives in an array. Any suggestions, pointers, or comments on how to approach this? Should I be looking at some kind of cluster file system instead of DRBD? Thanks very much, Miles Fidelman -- In theory, there is no difference between theory and practice. In<fnord> practice, there is. .... Yogi Berra _______________________________________________ Linux-HA mailing list Linux-HA@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems ------------------------ CONET Solutions GmbH, Theodor-Heuss-Allee 19, 53773 Hennef. Registergericht/Registration Court: Amtsgericht Siegburg (HRB Nr. 9136) Geschäftsführer/Managing Directors: Jürgen Zender (Sprecher/Chairman), Anke Höfer CONET Business Consultants, Schwieberdinger Straße 52, 71636 Ludwigsburg. Registergericht/Registration Court: Amtsgericht Stuttgart (HRB Nr. 204667) Geschäftsführer/Managing Directors: Klaus Böhle, Rüdiger Zeyen CONET Technologies AG, Theodor-Heuss-Allee 19, 53773 Hennef. Registergericht/Registration Court: Amtsgericht Siegburg (HRB Nr. 10328 ) Vorstand/Member of the Managementboard: Rüdiger Zeyen (Sprecher/Chairman), Wilfried Pütz Vorsitzender des Aufsichtsrates/Chairman of the Supervisory Board: Dr. Gerd Jakob _______________________________________________ Linux-HA mailing list Linux-HA@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems