[nylug-talk] Multi-NIC Redundancy On A Single Host

[H. G.]

I appreciate your advice, but this host is being used for simple core
services, and
I was only looking to add a measure of redundancy.

The services to be offered on this machine are support/monitoring and
infrastructure in nature,
(DHCP,DNS,monitoring,trending) and since we're running some core stuff here,
I just thought
that I wanted the redundancy at the link level (MAC redundance) instead of
at the
network layer (IP redundance).

Not only is it simpler in practice (1 IP address for all hosts to be aware
of and deal with),
I'm pretty sure that it will now allow me to have redundance at the switch
level, by having
both physical interfaces hook into different switches.  I've yet to test
this out,  but I think it
should work, and if it does, this means that I've knocked out the one switch
as a single point
while still keeping the same MAC address on both switches:  all my
applications need only
be aware aware of a single IP address, and the DHCP server running on the
machine can
provide the same DHCP configuration regardless of what interface that it's
receiving DHCP
requests on.  Steve Lembark suggested the dual switch config, though I don't
think he
meant using that method in combination with bonding.

It just makes things simpler.

Again, thank you all for your input.


On Sat, Feb 23, 2008 at 12:25 AM, Rodrick Brown <rodrick.brown at gmail.com>
wrote:

>
> Unless you're going multi nics to distribute i/o I would recommended
> building redundancy into your applications by going distributed if
> possible and not using bonding.
>
> I use to be a big fan of bonding and ipmp until I started to notice
> how much extra cabling and ports we're consumed for very little gain.
> I cant remember a single instance where bonding has truly been useful
> the only advantage I could see were during  switch upgrades where you
> could upgrade each switch in succession and theoretically not impact
> users, but seriously this type of work could easily be coordinated off
> hours or during low usage times.
>
> When it comes to server hardware the network adapter is the single
> most reliable component I've seen to date.  More and more systems
> these days have the network i/o's directly attached to the logic/main
> board so its less likely that you could every attribute a network
> adapter problem to a faulty adapter.
>
> Just my two cents.
>
> --
> Rodrick R. Brown
> http://www.rodrickbrown.com
> http://www.linkedin.com/in/rodrickbrown
>
> _____________________________________________________________________________
> Hire expert Linux talent by posting jobs here :: http://jobs.nylug.org
> The nylug-talk mailing list is at nylug-talk at nylug.org
> The list archive is at http://nylug.org/pipermail/nylug-talk
> To subscribe or unsubscribe: http://nylug.org/mailman/listinfo/nylug-talk
>