On Thu, 2012-06-21 at 16:40 +0900, Yuichi SEINO wrote: > Hi Jiaju, > > I have a question about booth. > I structure 2 sites and 1 arbitrator. And each site consist of 2 > node(ACT node and STB node). > Firstly, I kill corosync in one node. This node have a booth and a ticket. > Then booth is fail-over to STB node. The site including this node > always have a ticket. > However, an another site is granted at the same time. Therefore, 2 > sites was granted. > > I guess that sites cause fail-over because of a ticket was out of > expire date in synchronized timing. > However, I not see any reason to be granted 2 sites. Are you correct > this behavior? > > A following information is each site of ticket information after this > case was caused. > > siteA:<ticket_state id="ticketA" owner="2" expires="1340255934" > ballot="2" granted="true" last-granted="1340255142"/> > siteB:<ticket_state id="ticketA" owner="2" expires="1340255933" > ballot="2" granted="true" last-granted="1340255441"/> > siteC(arbitrator):<ticket_state id="ticketA" owner="2" > expires="1340255934" ballot="2" granted="false"/>
I think I've not quite understood what happened there. From this ticket information, Both siteA and siteB think ticket owner is "2", which is consistent, so why you say ticket was granted on two sites? For siteC, it has not gotten the last-granted for the time being, it should get this information shortly, or even for some reason siteC cannot get it, because the algorithm is majority-based, so it is no harm and allowable;) Or I have not gotten the full picture of this problem? If so, you can report a bug to bugzilla.novell.com, attaching the full logs there, then it will be good for us to investigate;) Thanks, Jiaju _______________________________________________ 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
