On 6/13/05 12:56 AM, Petr Kl?ma wrote:
> Hello
>
> A few (or a lot) of us supports larger corporate networks. I would like
> know your opinion/solution based on CentOS. No Win auto/dynamic/smart
> (WI|D)NS/DHCP
>
> How do you manage DNS&DHCP (the informations filled in, not the server
> software) for hundreds or thousands of computers?
The problem with automating DHCP entries, from my view in the cheap
seats, is that policies tend to differ widely enough from site to site
to make generalizations difficult. Even within a site, it can get dicey.
I've managed networks, for instance, where
a. some hosts get a static address on their "home" subnet (typically,
the one available at the user's desk) but are able to get a dynamic
address on other subnets, like those available in conference rooms.
b. some hosts only get dynamically allocated addresses, and only on a
limited set of subnets.
c. some hosts, mostly rack-mount machines, are treated like hosts in the
first group above, but they get long leases on the home subnet and short
leases on testing nets.
d. some hosts, these of the dying breed of stationary desktop
workstations, actually lived only on one subnet.
As you know, only the last type had a single host entry in the final
config files; the rest required at least two. We tried really hard to
script all these policies, but our users were so creative at finding new
DHCP niches where their machines "had" to live.
In the end, vi[m] was the best tool, because it was just too hard (for
us, anyway) to abstract DHCP usage policies well enough to encapusate
them in scripts. Plus, in the end, the configuration entries didn't even
resemble a DHCP config file, so all admins had to know their way around
both production DHCP config files *and* our funky pre-production
script-friendly configs. Feh! :-)
--
Paul Heinlein <> heinlein@xxxxxxxxxx <> www.madboa.com
