[ClusterLabs] Loss of quoram not detected. corosync 1.4.8 , pacemaker 1.1.14. CentOS 6

[neeraj ch]

Hi,

I have three node cluster set up. My corosync config is as follows.

compatibility: whitetank
totem {
    version: 2
    secauth: on
    threads: 0
    interface {
    member{
        memberaddr: ip
        }
    member{
        memberaddr:ip
         }
    member{
        memberaddr:ip
        }
    bindnetaddr:172.31.31.12
    ringnumber: 0
    mcastport: 5405
    ttl: 1
    }
    transport: udpu
}
logging {
    fileline: off
    to_stderr:yes
    to_logfile: yes
    to_syslog:yes
    logfile:/var/log/cluster/corosync.log
    debug:off
    timestamp: on
    logger_subsys {
    subsys: AMF
    debug:off
    }
}
service {
    name: pacemaker
    ver: 1
}
amf {
    mode: disabled
}



I went on to partition the master, the rest of the cluster considered the
master down and elected a new master.

But the corosync on the partitoned node didnt realize loss of quoram and
consequently pacemaker didnt shut down the resources. Giving me a
multi-master scenario.

I have tried adding , quorum tool as votequoram and it seems to give the
same result.

Why isn't corosync not detecting this scenario ? any ideas.

I used IP tables to block all UDP traffic in and out of the node . I
didn't block traffic on lo.

any help would be appreciated.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.clusterlabs.org/pipermail/users/attachments/20170413/3d2c2686/attachment.html>