The designation of the Heartbeat network interface in Linux RHCS

In the process of building RHCs, one place is easily forgotten, but it has unintended consequences: no specific heartbeat network interface is specified, then production network traffic and heartbeat network traffic run on the same network interface.

I often see the following configuration:

/etc/hosts file is:

192.168.10.10 Node1

192.168.10.11 Node2

10.0.0.10 NODE1-HBT

10.0.0.11 NODE2-HBT

/etc/cluster/cluster.conf is like this:

# .... omitted

<clusternodes>
<clusternode name= "Node1" nodeid= "1" votes= "1" >

# .... omitted
</clusternode>
< Clusternode name= "Node2" nodeid= "2" votes= "1" >
# .... omitted
</clusternode>
</clusternodes>

# .... omitted

Finally, according to this configuration of the cluster, production and heartbeat are running on the 192.168.10.x interface. The 10.0.0.x network interface is completely fake. You can see the flow with Tcpdump–i [interface].

Based on my experience, the exact configuration method is as follows:

The/etc/cluster/cluster.conf should be like this:

# .... omitted

<clusternodes>
<clusternode name= "NODE1-HBT" nodeid= "1" votes= "1" >

# .... omitted
</clusternode>
<clusternode name= "NODE2-HBT" nodeid= "2" votes= "1" >
# .... omitted
</clusternode>
</clusternodes>

# .... omitted

It is especially easy to make this mistake when using the graphical configuration tool Luci and Ricci configuration, because when adding nodes in the graphical interface, we subconsciously fill in the node name: NODE1/2, not NODE1/2-HBT

Note: The above only applies to Rhcs/centos 6.x Conga

The designation of the Heartbeat network interface in Linux RHCS