Less known Solaris features - IP Multipathing (Part 5): Prerequisites

At first you need a testbed. In this tutorial I will use a system with three interfaces. Two of them are Intel networking cards. They are named e1000g0 and e1000g0. The third interface is an onboard Realtek LAN adapter called rge0.
The configuration of the ip network is straight forward. The subnet in this test is \verb= I have a router at The physical network is a little bit more complex to demonstrate the limits of link-based failure detection. e1000g0 and e1000g1 are connected to a first switch called \verb=Switch A=. This switch connects to to a second switch called Switch B. The rge0 interface connects directly to Switch B. The router of this network is connected to Switch B as well.

To make the configuration a little bit more comfortable, we add a few hosts to our /etc/hosts file. We need four adresses while going through the tutorial. At first we need the name for the data address:

echo " hivemind-prod" >> /etc/hosts

Now we need names for our test addresses. It’s a good practice to use the name of the data address appended with the name of the physical address:

echo " hivemind-prod-e1000g0" >> /etc/hosts
echo " hivemind-prod-e1000g1" >> /etc/hosts
echo " hivemind-prod-rge0" >> /etc/hosts