Attached to this message is an entry from our internal issue tracking system which discusses (well, as you can see, it's mostly me discussing with myself ;-)) what the best way of configuring a DHCP client under Red Hat Linux is. I'm submitting it here to ask for other people's opinions on this issue. Does anyone with a similar setup, i.e. network with multiple hosts configured via DHCP, some of which may be disconnected and run stand-alone, or even connected to a different net, possibly without even rebooting, have anything to say?
-- Toralf Lund <[EMAIL PROTECTED]> +47 66 85 51 22 Kongsberg Scanners AS +47 66 85 51 00 (switchboard) http://www.kscanners.no/~toralf +47 66 85 51 01 (fax)Title: Full Text Bug Listing
Need to decide how to configure DHCP client on hosts booting via DHCP, in particluar the ones that are often disconnected from our LAN, and possibly connected to alternative networks, notably laptops. In particular, we need to decide whether to use dynamic or static host names. Disadvantages of dynamic names: * Name may change in the middle of a session, which could be *really* bad. Disadvantages of static names: * IP address lookup will fail when network is not connected. A workaround is to list <hostname> as alias for "localhost". IP address lookup will then return "loopback" address, and official hostname will be "localhost", which may or may not have side effects. * Host name is not fully qualified. ------- Additional Comments From Toralf Lund 2001-09-19 11:36 ------- Another drawback of having static names is of course that it may easlily lead to inconstencies between the (local) hostname and the official name from the host database's point of view; if the network card is changed, the DHCP server will assign a different IP address, and host name lookup thus returns a different name. Conversely, IP address lookup using the local name will _not_ return the hosts own address, which is when the fun really starts... ------- Additional Comments From Toralf Lund 2001-09-19 11:48 ------- Of course, dynamic naming leads to problems only when you want to start/stop/restart the network while the machine is up-and-running. This something we do a lot on laptops. That's also where a change of (PCMCIA) network cards is likely to occur. ------- Additional Comments From Toralf Lund 2001-10-02 09:59 ------- Moving bug to "Laptop". This is mainly a laptop issue as dynamic names or even static ones with no default address will work just fine on stationary hosts. ------- Additional Comments From Toralf Lund 2001-10-02 13:16 ------- Note that name as returned by 'hostname' command or 'gethostname()' call isn't qualified in the dynamic case either, at least not when using dhcpcd with our current setup (it may be possible to configure the server and/or client to qualify the name.) As for the static case, we should test what happens if we use a fully private name, i.e. one that isn't known by external hosts. ------- Additional Comments From Toralf Lund 2001-12-18 21:40 ------- A better solution than the one we've tried so far may be to use a static hostname, and create an entry with /etc/hosts for the name and the current address every time an IP address is received from a DHCP server. This entry should be replaced with a new one when a new address is set, but NOT removed when the network is disconnected. ------- Additional Comments From Toralf Lund 2001-12-18 22:46 ------- Our dhcpcd change script has now been updated as suggested in the note prior to this. ------- Additional Comments From Toralf Lund 2002-06-11 20:44 ------- Reopening but as /etc/hosts update doesn't help a lot. It is still not possible for the host to contact itself via its hostname when the network isn't running. Aliasing to localhost may be the only way if we want static name (but it doesn't feel right.) | |||||||||||||||||||||||||||||||||||
