Q) Virtual IP for fault-tolerant servers

Apologies in advance if this is OT.

We have a need to implement a fault tolerant server. I have only a
vague idea regarding how to proceed, so I would appreciate any
guidelines/pointers. Furthermore, I am somehat of a newbie. I know
that fault tolerant servers are pretty common and many smart people
have come up with many elegant solutions for fault tolerance.

I know that as per one approach, I need to use virtual IP addresses,
but I don't know the difference between a virtual IP address and
multiple IP addresses assigned to the same interface. Are they the
same thing? If not, what is the difference and how do you assign a
virtual IP address as opposed to the real IP address?

Also, do I need to implement a proprietary protocol to decide which
server actually uses the virtual IP address at any given point of time?
Or, is there some standard way of realizing this? For example, both
the servers could ping the virtual IP address, and if they get a
response, they know that the IP address is operational. Each server
then performs an ARP operation to resolve the IP address to the MAC
address. If the MAC address is one of its own, it knows that it is
currently providing service. Now my question is: is this a feasible
approach?

Also, assuming that my approach is feasible, what is the mechanism by
which a server "snatches" the virtual IP address for its own use
(either first time or upon the failure of the active server)? Is the
mechanism OS dependent (e.g. different for Linux, Windows and Solaris)?


We would like one of the servers to be the "preferred" server, meaning
that if it came down, and then came back on, it should assume control.
Any suggestions how to go about that?

Now that you know the problem I have on my hand, and my level of
ignorance in this subject, I would be eternally thankful if you could
show the lamp to my face and dispel this darkness that surrounds me.

-Baht

(E-Mail Removed) wrote:
> We have a need to implement a fault tolerant server. I have only
> a vague idea regarding how to proceed, so I would appreciate any
> guidelines/pointers.

Top hit on google for "high availability linux": http://www.linux-ha.org

I'm sure there will be descriptions and solutions for the Microsoft
world, too.

Chris [followups simplified]