Hi Jiaju, Thank you for a reply. I could understand some ways that the arbitrator is redundant.
Sincerely, Jiaju 2012年3月27日22:38 Jiaju Zhang <jjzh...@suse.de>: > On Tue, 2012-03-27 at 12:22 +0900, Yuichi SEINO wrote: >> Hi Jiaju, >> >> Thank you for a reply. I understand the case that a arbitrator have to >> be redundant . >> And I want to ask two questions . >> >> 1. I think about a satisfied way. Its way is that we make a special >> site of arbitrator. >> Can this way satisfy? >> For example, if a running arbitrator node is stop, this node is >> fail-over to a other node. > > Ok, no problem. If the boothd on the original arbitrator node fails, it > can be started on another node to be functional, or just restarted on > the original node to proceed, but please ensure that the IP address will > be floating with the boothd as well. So, in this scenario, it seems that > you have two machines on the 3rd site to be as an arbitrator, you may > want to configure boothd as a primitive resource and let pacemaker > manage it. > If you have 3 machines on the 3rd site, you can also configure the 3 > machines as 3 different arbitrators. > If you have 3 different sites (besides that 2 cluster sites) to put one > machine each to be as arbitrator, it'll also be OK. > > If the running arbitrator node is stop, the ticket won't be failed over > from one site to the other if that two sites and the connection between > them are fine. However, the danger is: if the arbitrator is down for > quite a long time, just at that moment, one site is down or the network > connection between them is broken, the other site won't be able to get > the ticket. > >> >> 2. In the case " for example, you have 2 sites, and configure 3 arbitrators", >> if 1 arbitrators stop, the structure is 2 site and 2 arbitrators. >> In this case, because the number of instance is an even number, >> i think that the arbitrator can't stop anymore. >> Am I correct think? > > Site itself is kind of arbitrator as well. So, if you have 2 sites and 3 > arbitrators, totally is 5. We allow that 2 of 5 can be down to make the > booth work. Of course, the 2 should be two arbitrators or one site plus > one arbitrator. If the failed 2 are two sites, booth doesn't work;) > > Thanks, > Jiaju > > -- Yuichi SEINO METROSYSTEMS CORPORATION E-mail:seino.clust...@gmail.com _______________________________________________ Pacemaker mailing list: Pacemaker@oss.clusterlabs.org http://oss.clusterlabs.org/mailman/listinfo/pacemaker Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://bugs.clusterlabs.org
