[ClusterLabs] Antw: Re: force RHEL6 and derivates to use corosync.conf

Andrew Beekhof andrew at beekhof.net
Mon Apr 13 03:13:28 UTC 2015


> On 8 Apr 2015, at 11:14 pm, Ulrich Windl <Ulrich.Windl at rz.uni-regensburg.de> wrote:
> 
>>>> Ionut Gogu <ionut.gogu at gmail.com> schrieb am 08.04.2015 um 15:01 in
> Nachricht
> <CAMESV9B8njhixNuBSqYBiUgdFbYJAxMm=rWaOVwm2JF5iOFAsQ at mail.gmail.com>:
>> Thanks for your answer Andrew,
>> i found very useful possibility to use corosync.conf for configuration of 2
>> redundant rings over 2 separate NICs, because i have the following
>> configuration:
>> 
>> *node01*
>> nic0 - productive network (82.77.xx.11 node01)
>> nic1 - corosync ring1     (10.1.1.11 node01hb0)
>> nic2 - corosync ring2     (10.1.2.11 node01hb1)
>> nic3 - backup network     (192.168.1.11 node01-back)
>> 
>> *node02*
>> nic0 - productive network (82.77.xx.12 node02)
>> nic1 - corosync ring1     (10.1.1.12 node02hb0)
>> nic2 - corosync ring2     (10.1.2.12 node02hb1)
>> nic3 - backup network        (192.168.1.12 node01-back)
>> 
>> I would like to use for <clusternode name> parameter, real hostnames of the
>> linux machines (node01 & node02) and 2 redundant rings over network
>> interfaces node01hb0 and node01hb0, but in RHEL6.5 6.6 only possibility
>> that i have founded to run 2 redundant rings over 2 different NICs is to
>> use for <clusternode name> parrameters: node01hb0 and node02hb0 in
>> combination with <altname> :

That sounds about right.

> 
> Hi!
> 
> We run a very similar configuration, except that everything is bond (and we
> use SLES). However I wonder what's the use of "altname": Does the cluster
> directly care about the corosync communication layer at all?

Pacemaker doesn’t.
But I agree that telling corosync about a single bonded device seems like the better approach these days.

> 
> Regards,
> Ulrich
> 
>> 
>>  <clusternodes>
>>    <clusternode name="node01hb0" nodeid="1">
>>      <altname mcast="226.10.1.1" name="node01hb1" port="5015"/>
>>    </clusternode>
>>    <clusternode name="node02hb0" nodeid="2">
>>        <altname mcast="226.10.1.1" name="node02hb1" port="5015"/>
>>    </clusternode>
>>  </clusternodes>
>> 
>>  <cman keyfile="/etc/cluster/corosync.key" expected_votes="1"
> two_node="1">
>>  <multicast addr="226.1.2.1" port="5010"/>
>> 
>> 
>> 
>> Mit freundlichen Grüßen/Kind regards,
>> 
>> *John Gogu*
>> 
>> M:       +49 (0) 152 569 485 26
>> Skype: ionut.gogu
>> 
>> On Wed, Apr 8, 2015 at 12:53 AM, Ionut Gogu <ionut.gogu at gmail.com> wrote:
>> 
>>> Hello,
>>> there is any possibility to force usage of /etc/corosync/corosync.conf in
>>> RHEL6 and derivates (corosync-1.4.7,pacemaker-1.1.12-4,cman-3.0.12.1-68)
>>> instead using parameters from cman tags from /etc/cluster/cluster.conf?
>>> 
>>>  <cman keyfile="/etc/cluster/corosync.key" expected_votes="1"
>>> two_node="1">
>>>  <multicast addr="226.1.2.1" port="5010"/>
>>>  </cman>
>>>   <totem rrp_mode="active" token="2000" consensus="24000">
>>>   </totem>
>>> 
>>> Mit freundlichen Grüßen/Kind regards,
>>> 
>>> *John Gogu*
>>> 
>>> M:       +49 (0) 152 569 485 26
>>> Skype: ionut.gogu
>>> 
> 
> 
> 
> 
> _______________________________________________
> 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