Hi, is there any way to implement failnode pairs without shared storage? I mean, two nodes with two syncronized separated hard disk and if one node fails, the other one starts to work. Thanks in avant, Jos?
This is a interesting question, Has DRBD ever been explored on this list? It wouldn''t work for very large clusters, but for moderate size? Brock Palen Center for Advanced Computing brockp at umich.edu (734)936-1985 On Oct 25, 2007, at 6:05 PM, Jos? M. Mart?n wrote:> Hi, > > is there any way to implement failnode pairs without shared storage? I > mean, two nodes with two syncronized separated hard disk and if one > node > fails, the other one starts to work. > > Thanks in avant, > > Jos? > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss at clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > >
On Oct 26, 2007 00:05 +0200, Jos? M. Mart?n wrote:> is there any way to implement failnode pairs without shared storage? I > mean, two nodes with two syncronized separated hard disk and if one node > fails, the other one starts to work.Some people have reported using GNBD or other replicating block devices in the list. You may find some information in the list archives. Cheers, Andreas -- Andreas Dilger Sr. Software Engineer, Lustre Group Sun Microsystems of Canada, Inc.