To ensure a redundant installation with redundant network interfaces, and to work with the installation scripts, all servers in the cluster need to have multipathing configured. We recommend static IP addresses.
The general steps required to set up IP multipathing on a Solaris machine are:
Ensure that the system detects both network interface cards and they have different MAC addresses.
Group the network interface cards.
Add a test IP address to the first network interface card.
Add a test IP address to the second network interface card.
Edit each server’s /etc/hosts file.
Change the hostname.* files to keep these settings after a reboot.
When activating IP multipathing on a server, a total of three IP addresses are used. If the private address space is a Class C space, only 254 usable addresses are usually available.
To keep the IP address space uncluttered, a convention for IP assignments has evolved—assign the main failover IP to be in the range 1–54, the first test address 100 higher than the main address, and the second test address 200 higher than the main address. So for a main address of 192.168.0.41, the first test address is 192.168.0.141 and the second 192.168.0.241.
Create DNS entries for the test addresses although the addresses are not used. The convention for this is to use the hostname of the server with -test1 or -test2 appended to the end.
Going back to the sample worksheet (Table 8), hostname MyNode_1 uses a multipath IP of 192.168.0.1 and its interfaces are MyNode_1-test1 at 192.168.0.101 and MyNode_1-test2 at 192.168.0.201.