[Linux-HA] Secondary Interfaces, and Host Names

Karl W. Lewis karl.w.lewis at gmail.com
Tue Jul 21 06:55:04 MDT 2009


So, I'm setting up a web server on an HA-Linux Cluster using heartbeat v2.99
and Pacemaker.

I'm setting it up on a set of three Egenera blades.   These blades have no
actual network interfaces, that is handled by the frame.  So, there are the
interfaces over which the web server will talk to the world, and a virtual
IP Address managed by the cluster, and I have set up a "private" network
that he cluster can use to chat that is not connected to any other network.
I've assigned IP addresses for that virtual interface on each blade, but now
I have a question that I fear will reveal my ignorance...

How do I relate the private network IP Addresses to each host?  Currently,
the public addresses for each host are associated with the hosts in the
/etc/hosts file.  And, those names are the ones that you get if you do a
`uname -n` on any of the blades.  Should I use the private network addresses
instead?  My ha.cf file contains the line: bcast eth1 eth0. eth1 is for the
private network, but it occurs to me that heartbeat can't know how to get to
the other nodes over eth1 because I've not told it what those addresses
are.  Should I re-write my hosts file such that the private addresses are
the ones associated with the proper host names?

Since these are egenera blades, I can't use serial/null modem cables to
connect the cluster as the blades don't have any of those, either.

My google skilz, such as they are, have failed me... I have been unable to
find any details of a setup like this.  I'd be most appreciative if anyone
could provide any tips, pointers, or practical examples.

Thanks,

Karl



More information about the Linux-HA mailing list