Ron - Gluster doesn't have a high availability solution built in but we test with CTDB for CIFS and NFS, or you could use UCARP for just NFS. CTDB is required for CIFS, and works well with NFS. Both projects are mature and easy to use, just set them up across a Gluster replica pair and failover will be very quick and you will not get a connection reset, no errors to your applications. If you have multiple replica pairs use RRDNS to load balance across the entire cluster. http://ctdb.samba.org/configuring.html http://www.ucarp.org/project/ucarp Thanks, Craig --> Craig Carl Senior Systems Engineer Gluster From: "ron gage" <ron at rongage.org> To: gluster-users at gluster.org Sent: Wednesday, November 10, 2010 8:57:52 AM Subject: Question about redundancy over ethernet Greetings: I am considering deploying Gluster using NFS primarily as the protocol of choice. In order to achieve failover redundancy, does Gluster utilize a virtual IP address like a load balancer would? Let's say I have a 4 node Gluster cluster set up, my client connects to node A. If node A goes off network for whatever reason, what happens to my client? Are sessions maintained so the client(s) don't register a server disconnect? In other words: with NFS (or even CIFS), how transparent is the failover process. Finally, what is the typical failover cutover timing like? Is it sub-second? Ron _______________________________________________ Gluster-users mailing list Gluster-users at gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
