Re: mirrored machines via network

Timothy Peters (tim.peters@nene.ac.uk)
Wed, 06 Nov 1996 10:27:55 +0000

Peter Rival wrote:
>
> -----BEGIN PGP SIGNED MESSAGE-----
>
> On Mon, 4 Nov 1996, Ian Main wrote:
> > On Fri, 1 Nov 1996, Ray Lehtiniemi wrote:
> > >On Fri, 1 Nov 1996, James R. Leu wrote:
> > >> operations of a small ISP. I am looking into ways of implementing redundant
> > >> servers.
> > >[snip]
> > >> The idea I have, necessitates two main developments. The first of which is
> > >> a "intelligent" name-server. I have begun development of this already. I have
> > >> started by using th bind-4.x.x source. The end result will hopefully be a
> > >> name-server that checks if a machine is responding, before it gives out the IP
> <snip simplification reasoning>
> > ping master
> > if no response ifconfig eth0 down
> > ifconfig eth0 <master_ip> up
> >
> > so long as they're on the same subnet, it should work no?
>
> Yes, it (could) work. However, I see two potential probs with that
> idea. One - how do you decide how often (and how much) to ping the master?
> Granted ICMP echo requests aren't all that large for packets, but in this day
> and age, we need all the spare bandwith we can get. And two - what happens if
> you lose the entire segment? (I know, scream bloody murder and turn on the
> coffee pot...;) If you could set up the master DNS server to check on the
> state of the server (I think the newest version of lbnamed does this) before
> returning the IP, you could put them on separate segments and hopefully have
> a little more redundancy.
>
> > Now all you have to deal with is dropped connections, and the mirroring of
> > the HD.
>
> Well, that's where this idea kinda falls apart. Unless you can write
> some really schnazzy clustering software that works _reliably_ through a
> router. I know there are things out there like this...unfortunately, I've
> gotta go install an AIX (eeww...hate that word) box, so I don't have the time
> to try to research it. Quick thought - Auspex does do something like this,
> but I don't think it solves the separate segment thing.
>
> - Pete
>
> -----BEGIN PGP SIGNATURE-----
> Version: 2.6.2
>
> iQCVAwUBMnvMHjkUd9WlV5D5AQFYmwP/UQoFYzBFQJigdlpQecj4Ns6n9Z5aRaba
> +JKuZW9kuxnJktpZUMjz4YAiHbpHIsbct19/055FilWSPuGdwsM3AOIILRb7IkTE
> nbp98a2R40uxb2cgoOHlsXdei7rNgdBzcps+dibdpRMZBetJQQ+PCVSn6BxXd+cF
> 3n9cMD/wNBU=
> =+cLp
> -----END PGP SIGNATURE-----
This situitation is something taht i will be looking into as part of my
research project what I hope to have is two machines which will share a
common memory image (don't ask how I don't know yet :-) ). So that if
one fails the other can assume it responsabities.
--
Timothy Peters
Senior Software Engineering Techinician
School of Engineering & Technology
Nene College
St Georges Ave
Northampton
NN2 6JD
(01604) 735500 Ext. 3112
tim.peters@nene.ac.uk