Hi, I'm currently running DRBD over NFS in a primary/secondary - secondary/primary set up. Two nodes with one primary on each. It does most of what I want, but there are certain conditions that would lead to split brain situations that I'd like to cater for. Since I'm running a primary per server I can't really use ordinary fencing measures.
There are two things I need to do 1) If crossover fails (pingd1), adjust drbd to another route, if crossover and lan fail (pingd2), outdate the secondary so it doesn't get promoted before resync. 2) Node 1 goes down, node 2 runs everything. Node 2 goes down, node 1 comes up and must not take things over since node 2 is the up to date one. The first case I would assume needs an agent (new agent or modified drbd agent) that can outdate and adjust drbd when called by a constraint? The second would need an agent that can check a file on the server (only way I can think of that's persistant if nodes are out of contact but running at different times) and based on that decide whether it was the previous master and can promote drbd or not? I'm thinking an agent similar to pingd, but checking a file instead of interfaces. If anyone has better ways of doing this or suggestions on how to go about the constraints etc I'd really appreciate it. It may grow a little more complex in that my live servers have crossover cable between them, a LAN interface each and an external interface each, so drbd could go through any one of them, even though I'm using the crossover currently. Thanks for any help or insight anyone can provide. Guy -- Don't just do something...sit there! _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
