On 6/13/05 12:56 AM, Petr Klíma 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! :-)