[ClusterLabs] Corosync over dedicated interface?

Dave Withheld davewithheld at hotmail.com
Wed Sep 28 11:37:00 EDT 2022

I am setting up a high availability two-node, master/slave cluster using drbd for shared data using CentOS 8 stream. The two nodes are identical and either can be master or slave with the master having the drbd device mounted and the slave mirroring the data and standing ready to take on the master roll if any of the resources fail. All resources (except demoted drbd) are colocated on the master.  Each node has two network interfaces with one being the usual LAN interface and the other being a dedicated private network for drbd data/syncronization.

One of the resources assigns a virtual IP to the LAN interface for serving users. If that LAN interface should fail, I want the master node to be demoted and the slave to be promoted to master (assuming its LAN interface is still functioning). For that scenario to succeed, I would like the private network to also be used by corosync to maintain the communications between the nodes which will allow drbd to function AND corosync/pacemaker to coordinate, even if a LAN interface fails.

Unfortunately corosync does not seem to be using that private interface, but rather seems to use the LAN interface, based on node hostname, rather than the IP address defined in its config file (which is the private network IP). To test functionality, I unplugged the network cable from the master system. This left the slave system thinking the master system was offline and all resources stopped. The master system remained master but I had no access to it from the network (obviously). drbdadm status showed both systems happy with Primary/Secondary rolls still intact on the master/slave, resp, but pacemaker did not allow the promotion of the slave to master due, I believe, to its inability to communicate over the LAN interface and determine the drbd state (which was fine and working over the private network).

Is it possible to get corosync to use the private network and stop trying to use the LAN for cluster communications? Or am I totally off-base and am missing something in my drbd/pacemaker configuration?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/users/attachments/20220928/17a1923c/attachment.htm>

More information about the Users mailing list