[ClusterLabs] Corosync do not send traffic
Jan Friesse
jfriesse at redhat.com
Tue Apr 5 08:46:40 CEST 2016
Roberto Munoz Gomez napsal(a):
> 2016-03-30 13:42 GMT+02:00 Jan Friesse <jfriesse at redhat.com>:
>
>> Roberto Munoz Gomez napsal(a):
>>
>>> 2016-03-30 11:27 GMT+02:00 Jan Friesse <jfriesse at redhat.com>:
>>>
>>> Hello,
>>>>
>>>>>
>>>>> Due to a change in the switch in one of the datacenters now I have an
>>>>> odd
>>>>> behaviour in the cluster.
>>>>>
>>>>> I am using cman with corosync and pacemaker. The versions are:
>>>>>
>>>>> pacemaker-1.1.10-14.el6_5.3.x86_64
>>>>> corosync-1.4.1-15.el6_4.1.x86_64
>>>>> cman-3.0.12.1-49.el6.x86_64
>>>>>
>>>>> The problem is, when I launch /etc/init.d/pacemaker start and the
>>>>> cman_tool
>>>>> launch corosync, I don't see any UDP traffic, so the cluster is "broken"
>>>>>
>>>>> But if I launch manually the same command "corosync -f" I do see udp
>>>>> traffic and the totem is correctly sent between nodes.
>>>>>
>>>>> It all began with the change in the switch, but I set the tcpdump in the
>>>>> hosts and I do not see traffic.
>>>>>
>>>>> I have tried the multicast and unicast configuration, different network,
>>>>> but all with the same behaviour.
>>>>>
>>>>> What am I missing?
>>>>>
>>>>>
>>>> I'm pretty sure main problem is that:
>>>> - corosync -f uses /etc/corosync/corosync.conf as a config file
>>>> - cman_tool join uses /etc/cluster/cluster.conf as a config file
>>>>
>>> Make sure to "replicate" (specially udpu/udp, ...) configuration from
>>>> corosync.conf to cluster.conf and I'm pretty sure it will work.
>>>>
>>>
>
> It worked!
>
> I didn't knew cman_tool do not use corosync.conf. Setting udpu in
> cluster.conf made it work again.
>
> I am trying to migrate the bindnetaddr option from corosync.conf to
> cluster.conf but couldn't make it work. I want to use this interface for
> corosync traffic.
cluster.conf configuration file is more similar to corosync.conf with
nodelist defined than corosync.conf without it. For cluster.conf you
don't need to define bindnetaddr. Just put node names into /etc/hosts
and add clusternode tag into cluster.conf. That's it.
>
> inet addr:10.76.125.236 Bcast:10.76.125.239 Mask:255.255.255.240
>
> In man cluster.conf says the last octect must be 0. But despite ccs_config
> validate says it is ok, the cman_tool fails.
Do not use network address. Just put nodes into cluster.conf.
Regards,
Honza
>
>
> Regards.
>
>
>
> _______________________________________________
> Users mailing list: Users at clusterlabs.org
> http://clusterlabs.org/mailman/listinfo/users
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org
>
More information about the Users
mailing list