G.Eustace at massey.ac.nz
Tue May 6 20:00:20 UTC 2008
> I think you have about the optimum setup there. You can only have two
> peers in a failover relationship, so it's not really practical to add
> additional servers. To completely lose DHCP service then you'd have
> to lose both a site-local server AND the WAN link - at which point
> you probably more to worry about than just the DHCP.
> As long as you have reasonably reliable servers, and a reasonably
> reliable network, then I personally wouldn't be too concerned.
Thanks for the comments. The current design was intended to be optimum but an HA/DR project is just kicking off here and I have been asked to look at the fundamental network infrastructure again. Given your comments, I think the only change we might make is to virtualise the DHCP servers onto an HA VMware ESx cluster at each site. That makes them HA but also means we can move them quickly or restart them on an alternate platform if we do have issues with hardware.
More information about the dhcp-users