On Sat, Jan 5, 2013 at 2:21 AM, Imri Zvik <[email protected]> wrote: > Hello All, > > I am trying to compile a configuration for DRBD replication between two > geographical sites, in order to achieve MySQL HA. > According to my requirments, I only need one node at each site. > The solution should be able to cope with split brain scenarios. > > I was thinking about the following configuration: > > 1. Single node cluster at each site, with DRBD MS resource and a MySQL > resource set with colocation to start on the DRBD master. > 2. no-quorum-policy=ignore > 3. stonith set to suicide > 4. Boothd with a third site to run an arbitrator. DRBD and MySQL will depend > on the ticket from Boothd using rsc_ticket ( loss-policy=demote). > > My questions are: > 1. what are the cons of this configuration? Any caveats/pitfalls I am > missing? will it work?
I think stonith is going to be the catch. Traditionally pacemaker hasn't wanted to, nor been very good at having nodes fence themselves (because there is, by definition, no-one around to confirm it worked). > 2. I haven't seen any concrete examples/configuration references on how to > correctly configure pacemaker for single node cluster. Do I need corosync? Yes. > can it be disabled? No. > I've seen something about a "null comm module" but wasn't able to find any > detailed information on how to configure it. Not needed, just don't start up any other nodes for it to talk to. _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
