|
Dual homing a production server for redundancy
I've got an IBM x335 server with Slackware 11.0 (2.6.20.x kernel) that comes with two GigE network interfaces that I wish to dual-home in the server co-location facility.
There are two main reasons for this. The first being redundancy in case one NIC or the switch it's attached to fails, and the second reason is that the box needs two IP addresses anyway, in order to be able to serve as an authorative DNS server for my domain (which requires to BINDs on different IPs) (and yes, in case the entire box or co-location facility fails, I have a 3rd and 4th DNS slave setup in a different continent)
My question is how to set this up in Slackware, both from a DNS point of view and from a Slackware network configuration point of view.
My guess is that I should assign the two interfaces separate IP addresses rather than use interface bonding, but I am not sure.
My other guess is that I should give the machine one primary, canonical name and bind that to the primary (eth0) interface on the box, so that for instance 192.168.1.101 is server.example.com.
Question: If 192.168.1.101 (eth0) is server.example.com which corresponds to the box hostname, then what about eth1 which may be 10.1.2.102 ? How should that be registered in DNS? as server-backup.example.com ? Or can both be registered both in PTR and A records as server.example.com ?
Is special configuration required in rc.inet1.conf?
I would greatly appreciate some feedback from others who have experience with this, as I am obviously quite new to the problem.
(PS. I know the easy thing would be to get someone else to configure it for me, but that's beside the point - the entire reason why I'm going through this exercise is to learn Slackware from the ground up, including getting BIND+postfix+iptables+apache+++ to work in a "production" environment)
Thanks much for any shared insight!
-Y1
|
