Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Hey everyone, I have some questions about handling unresolvable split-brains in a dual- primary setup with pacemaker. This is the scenario: - DRBD and pacemaker are configured to start resource r0 in dual-primary mode - For test reasons, a split-brain between node1 and node2 is created, that is not automatically resolvable - Thus, DRDB disconnects the resource r0 after all configured automatic resolvers failed. So far so good. This is the problem: - Although DRDB disconnects the resource due to the split-brain, pacemaker doesn't get to know about this condition at all, because both nodes are in "cs:StandAlone ro:Primary/Unknown" state. My questions are: 1) Is resource level fencing the solution (also in a dual-primary setup)? 2) is there any documentation out there on the "crm-fence-peer.sh" script? 3) Is the "crm-fence-peer.sh" also working in a dual-primary setup? 4) What I would like to happen is that BOTH nodes stay in "slave" mode and do not get promoted to "primary" after the disconnect. I don't want pacemaker to continue starting other resources in disconnected DRBD state! Is there any way to do this with the "crm-fence-peer.sh" handler? Thanks for any help! -- Dipl.-Ing. Joschi Brauchle, M.Sc. Institute for Communications Engineering (LNT) Technische Universitaet Muenchen (TUM) 80290 Munich, Germany Tel (work): +49 89 289-23474 Fax (work): +49 89 289-23490 E-mail: joschi.brauchle at tum.de Web: http://www.lnt.ei.tum.de/