HEARTBEATS

Quickie overview of how 'heartbeats' work for freeha:

You should set up multiple networks between the machines in the cluster 
(2 networks minimum) , on at least 2 separate interfaces per machine, so 
that they can tell each other what is going on.

freeha takes advantage of standard IP 'broadcast' definitions, to send a UDP
broadcast packet to all the machines on the heartbeat subnet, with a single
packet. This 'heartbeat' packet is sent by each cluster machine every second.
That is good, in that it easily allows for multiple machines in standby mode 
in the cluster. It is "bad", if you happen to be sharing the network with
other machines.

Ideally, you will use dedicated interfaces with private switches, or
direct-connect crossover cables, for this. However, it is just as easy to use
existing network interfaces, and allocate additional virtual interfaces for
them.

Remember that every machine plugged into the same switch will still see all
broadcast traffic, unless you configure the switch to keep the private address
range traffic to the appropriate slots on the switch.



There are standard 'private' network address ranges that can be used, so that
you can allocate networks solely for use between cluster machines.

Examples of 'reserved networks' appropriate for this use:

192.168.1.0, 192.168.2.0, 192.168.3.0, and so on.


It is up to YOU, the sysadmin, to ensure that only cluster machines are on
that subnet. If there are other machines on the same subnet, they will see a
lot of 'garbage' traffic. That is why it is best to use 'private' addresses 
for heatbeats.


If you are running out of secondary network interfaces, you might choose
to use a free serial port in a two-machine cluster, and set up PPP across
the serial ports. This should work just as well.


See the 'INSTALL' text file for a specific sample network configuration.