Hi Steve, 22.09.2011 20:10, Steven Dake wrote: > On 09/21/2011 10:23 PM, Andrew Beekhof wrote: >> Stupid question, but if quorum is being calculated in quorumd, >> shouldn't clients get it from there directly? >> What's the advantage to corosync being involved (other than as a client itself)? >> > > There are two problems with getting it directly from quorum daemon > 1) The quorum state and cpg timing are not synced > 2) more generically, it would be helpful to get all the low level > cluster info from one location > > 1 is the deal killer though. Note we have this problem today in > corosync because quorum and cpg are not synchronized. We plan to > synchronize quorum delivery into the cpg callback stream so that cpg > applications that need quorum will only obtain callbacks if they are > part of the primary component. May it be related to what I (am trying to) talk about in http://www.mail-archive.com/pacemaker@xxxxxxxxxxxxxxxxxxx/msg09788.html (I mean part which is related to simultaneous cman's kick_node_from_cluster() calls from two partitions - they are originated from callback)? Thanks, Vladislav _______________________________________________ discuss mailing list discuss@xxxxxxxxxxxx http://lists.corosync.org/mailman/listinfo/discuss
