[Pacemaker] cman multi-homed with udp-broadcast issues

[Andrew Beekhof]

This is getting into some pretty specific cman knowledge, you might
find more experts on that at linux-cluster at redhat.com

On Sat, Jan 7, 2012 at 4:00 AM, Patrick H. <pacemaker at feystorm.net> wrote:
> So I'm trying to setup a cluster with a secondary communication ring in case
> the first ring fails. The cluster operates fine, but doesnt seem to handle
> path failure properly. When I break the path between the 2 nodes on ring 1,
> I get the following in the logs:
>
> Jan  6 16:55:17 syslog02.cms.usa.net corosync[13931]:   [TOTEM ]
> Incrementing problem counter for seqid 202 iface 165.212.15.49 to [1 of 3]
> Jan  6 16:55:19 syslog02.cms.usa.net corosync[13931]:   [TOTEM ] ring 1
> active with no faults
> Jan  6 16:55:24 syslog02.cms.usa.net corosync[13931]:   [TOTEM ]
> Incrementing problem counter for seqid 204 iface 165.212.15.49 to [1 of 3]
> Jan  6 16:55:26 syslog02.cms.usa.net corosync[13931]:   [TOTEM ] ring 1
> active with no faults
> Jan  6 16:55:30 syslog02.cms.usa.net corosync[13931]:   [TOTEM ]
> Incrementing problem counter for seqid 206 iface 165.212.15.49 to [1 of 3]
> Jan  6 16:55:32 syslog02.cms.usa.net corosync[13931]:   [TOTEM ] ring 1
> active with no faults
>
> And it just repeats over and over. From notes I've found from others, it
> appears this might be because of each ring sharing the same broadcast
> address. Indeed this is the case as `cman_tool status` shows
> Multicast addresses: 255.255.255.255 255.255.255.255
> Node addresses: 165.212.64.49 165.212.15.49
>
> However I've tried changing this address in the cluster.conf and it seems to
> be completely ignored. I've also tried changing the port for the second ring
> and thats also ignored (tcpdump shows them still going to the same port as
> ring 0).
>
> So, is this indeed the cause of it not properly detecting ring failure? And
> if so, how can I fix it?
>
>
> cluster.conf:
> <?xml version="1.0" ?>
> <cluster name="syslog" config_version="6">
>     <logging to_logfile="no" syslog_facility="local2" debug="on" />
>     <cman expected_votes="1" two_node="1" transport="udpb" port="5408" />
>     <totem rrp_mode="active" secauth="off" />
>     <clusternodes>
>         <clusternode name="syslog01" nodeid="1">
>             <altname name="syslog01-cms" port="5406" mcast="165.212.15.255"
> />
>             <fence>
>                 <method name="pcmk-redirect">
>                     <device name="pcmk" port="pcmk-1" />
>                 </method>
>             </fence>
>         </clusternode>
>         <clusternode name="syslog02" nodeid="2">
>             <altname name="syslog02-cms" port="5406" mcast="165.212.15.255"
> />
>             <fence>
>                 <method name="pcmk-redirect">
>                     <device name="pcmk" port="pcmk-1" />
>                 </method>
>             </fence>
>         </clusternode>
>     </clusternodes>
>     <fencedevices>
>         <fencedevice name="pcmk" agent="fence_pcmk" />
>     </fencedevices>
> </cluster>
>
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at 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
>