Re: DHCP server fail-over and accounting

[Nicolas CARTRON <nicolas () ncartron org>]

Hi,

On Wed, Feb 2, 2011 at 10:38 AM, John Adams <jna () retina net> wrote:

> 2011/2/1 Joe <sj_hznm () hotmail com>:
> > hi,
> >
> >    we plan to implement DHCP server farm in our network.   Currently ,
> there are there  problems burning my head. could anybody
>
>
> You're making this way, way too complicated.
>
> Run two DHCP servers. Allocate two different netblocks to each server.
> For Example, if your network is a /24, allocate a couple of /26's.
> Both will answer on a request.
> The client will ack to whatever address it decides to accept. Full
> redundancy.

Well, it also depends on the constraints: having such a configuration
implies that every scope will have to be declared twice, as well as the DHCP
options.
Plus, if the server who issued the lease is down, the client will get a new
DHCP lease - which maybe an issue for some people.


> >       To our experience, this needs to set up  DHCP  server on two sites
> and syncronize their content in real time.
> >      Beside this ,  we hope  there should be as less modification as
> possible  on edge router when one DHCP  server is down.
> >      should anycast architecture helpful ?   or should we just set up two
> dhcp servers on two sites and  sync. with ISC DHCPD?
>
> Don't even bother with the syncing, and anycast is the wrong protocol here.



Agree, anycast makes no sense.
ISC DHCPd sync works well, provided you know it and configured it
correctly.