On 01/09/2013 05:32 AM, Lars Marowsky-Bree wrote: > On 2013-01-08T10:03:20, Digimer <[email protected]> wrote: > >> If you use two separate networks for RRP, and each network sits on top >> of a mode=1 bond, then you will have the highest redundancy. I do this >> with stacked switches where each leg of the bond is in a different >> switch, so I can survive total switch failures as well as individual >> link losses. > > This does not work for things that use the DLM (OCFS2/GFS2), since SCTP > is a notorious PITA. > > I still think a mode=3 bond for corosync should be even better, but the > rest of the system probably wouldn't appreciate the duplicated packets > ;-) > > > Regards, > Lars
It actually works just fine with DLM-using tools. Our clusters all use DLM for GFS2, clvmd and rgmanager. Part of our demo is to kill the power to the primary switch with the cluster fully operating and then recover the same switch. I will play with the bonding modes again in the next week if/when I get the time and I will publish my test results. Cheers -- Digimer Papers and Projects: https://alteeve.ca/w/ What if the cure for cancer is trapped in the mind of a person without access to education? _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
