RE: FN-FORUM At Last Some Possible Work Available

date posted 12th December 2001 15:11

On Wed, 12 Dec 2001, Russell Foster wrote:

> > To avoid the single point of failure you could always add in a load
> balancer
> > machine with IP Chains which decides which Apache Server/Machine to hit,
> > thus rendering a failsafe methodology and hopefully some redundency.
>
> Actually from the way I read this there is still a single point of failure
> as what happenes if the IPChains box dies? There is actually a much better
> way in doing this which is a mixture of round robin DNS and 2 machines at
> the front acting as nothing more than packets forwards with some form of
> automatic failover. There fore if one dies the other picks up. Admitally
> some things like incoming connections might be disrupted but even this can
> be minimized with a carefully crafted application.
>
> I've done stuff like this before without any problems and as far as I know
> it is still working to this day

If you want to go a stage further, you can look into things like Fake
which uses ARP spoofing to make
one machine appear to be another with next-to-zero delay between failure
of the first and failover to the second host.

I've worked with a similar (but *much* more expensive) proprietary
solution for a couple of years and this sort of technique works /really/
well, providing there's no state issues (i.e. database commits and so on).

> Rdgs
> Rus Foster

Best Regards,
Alex.
--
Alex Butcher Brainbench MVP for Internet Security: www.brainbench.com
Berkshire, UK Is *your* company hiring UNIX/Security/Pen. testing folks?
PGP/GnuPG ID:0x271fd950 http://www.cocoa.demon.co.uk/cv/