Last Modified: Sat, 05 Nov 2011 15:18:34 +0000 ; Created: Wed, 19 Oct 2011 01:44:40 +0000
|Basic normal instructions work okay when you don't have special subnets or VLANs, but the scripts on both the master and peer nodes fail if you don't have a specific IP configured for eth0.
I don't want one on the default VLAN on eth0 so I got around the problem by typing in an address in /etc/network/interfaces for eth0. The script just parses this file and doesn't care if the interface actually has that IP. It is just an auto-probe reference limitation.
Something as simple as the following works fine for slave nodes:
iface eth0 inet manual address 172.16.0.121 netmask 255.255.255.0
Don't restart the networking, just modify the configuration file and run "pveca -a -h IP-ADDRESS-MASTER". You may see a few warnings about missing localip, but you can ignore those. Change the file back after you've added it.
You can also manually modify the /etc/pve/cluster.cfg and /etc/pve/master/cluster.cfg files if needed.