[Pacemaker] Different Corosync Rings for Different Nodes in Same Cluster?

[Andrew Martin]

Hi Steve, 


Thanks for the clarification. Am I correct in understanding that in a complete network, corosync will automatically re-add nodes that drop out and reappear for any reason (e.g. maintenance, network connectivity loss, STONITH, etc)? 


Thanks, 


Andrew 

----- Original Message -----

From: "Steven Dake" <sdake at redhat.com> 
To: "The Pacemaker cluster resource manager" <pacemaker at oss.clusterlabs.org> 
Cc: discuss at corosync.org 
Sent: Friday, June 29, 2012 9:40:43 AM 
Subject: Re: [Pacemaker] Different Corosync Rings for Different Nodes in Same Cluster? 

On 06/29/2012 01:42 AM, Dan Frincu wrote: 
> Hi, 
> 
> On Thu, Jun 28, 2012 at 6:13 PM, Andrew Martin <amartin at xes-inc.com> wrote: 
>> Hi Dan, 
>> 
>> Thanks for the help. If I configure the network as I described - ring 0 as 
>> the network all 3 nodes are on, ring 1 as the network only 2 of the nodes 
>> are on, and using "passive" - and the ring 0 network goes down, corosync 
>> will start using ring 1. Does this mean that the quorum node will appear to 
>> be offline to the cluster? Will the cluster attempt to STONITH it? Once the 
>> ring 0 network is available again, will corosync transition back to using it 
>> as the communication ring, or will it continue to use ring 1 until it fails? 
>> 
>> The ideal behavior would be when ring 0 fails it then communicates over ring 
>> 1, but keeps periodically checking to see if ring 0 is working again. Once 
>> it is, it returns to using ring 0. Is this possible? 
> 
> Added corosync ML in CC as I think this is better asked here as well. 
> 
> Regards, 
> Dan 
> 
>> 
>> Thanks, 
>> 
>> Andrew 
>> 
>> ________________________________ 
>> From: "Dan Frincu" <df.cluster at gmail.com> 
>> To: "The Pacemaker cluster resource manager" <pacemaker at oss.clusterlabs.org> 
>> Sent: Wednesday, June 27, 2012 3:42:42 AM 
>> Subject: Re: [Pacemaker] Different Corosync Rings for Different Nodes 
>> in Same Cluster? 
>> 
>> 
>> Hi, 
>> 
>> On Tue, Jun 26, 2012 at 9:53 PM, Andrew Martin <amartin at xes-inc.com> wrote: 
>>> Hello, 
>>> 
>>> I am setting up a 3 node cluster with Corosync + Pacemaker on Ubuntu 12.04 
>>> server. Two of the nodes are "real" nodes, while the 3rd is in standby 
>>> mode 
>>> as a quorum node. The two "real" nodes each have two NICs, one that is 
>>> connected to a shared LAN and the other that is directly connected between 
>>> the two nodes (for DRBD replication). The quorum node is only connected to 
>>> the shared LAN. I would like to have multiple Corosync rings for 
>>> redundancy, 
>>> however I do not know if this would cause problems for the quorum node. Is 
>>> it possible for me to configure the shared LAN as ring 0 (which all 3 
>>> nodes 
>>> are connected to) and set the rrp_mode to passive so that it will use ring 
>>> 0 
>>> unless there is a failure, but to also configure the direct link between 
>>> the 
>>> two "real" nodes as ring 1? 
>> 

In general I think you cannot do what you describe. Let me repeat it so 
its clear: 

A B C - NET #1 
A B - Net #2 

Where A, B are your cluster nodes, and C is your quorum node. 

You want Net #1 and Net #2 to serve as redundant rings. Since C is 
missing, Net #2 will automatically be detected as faulty. 

The part about corosync automatically repairing nodes is correct, that 
would work (If you had a complete network). 

Regards 
-steve 

>> Short answer, yes. 
>> 
>> Longer answer. I have a setup with two nodes with two interfaces, one 
>> is connected via a switch to the other node and one is a back-to-back 
>> link for DRBD replication. In Corosync I have two rings, one that goes 
>> via the switch and one via the back-to-back link (rrp_mode: active). 
>> With rrp_mode: passive it should work the way you mentioned. 
>> 
>> HTH, 
>> Dan 
>> 
>>> 
>>> Thanks, 
>>> 
>>> Andrew 
>>> 
>>> _______________________________________________ 
>>> 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 
>>> 
>> 
>> 
>> 
>> -- 
>> Dan Frincu 
>> CCNA, RHCE 
>> 
>> _______________________________________________ 
>> 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 
>> 
>> 
>> _______________________________________________ 
>> 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 
>> 
> 
> 
> 



_______________________________________________ 
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 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.clusterlabs.org/pipermail/pacemaker/attachments/20120702/a5efc615/attachment-0002.html>