Valdis.Kletnieks@vt.edu wrote:
On Mon, 17 Sep 2007 17:15:38 EDT, John Curran said:
In addition, if the AAAA record is added for the node, instead of service as recommended, all the services of the node should be IPv6- enabled prior to adding the resource record. "
Not a problem for names which are single services (www.foo.com), but caution is required when the name has multiple services running.
My favorite shoot-self-in-foot on that topic - I stuck a quad-A in for a host that *was* IPv6-enabled on the production service, but it didn't have (at the time) an IPv6-ready ssh daemon. Hilarity ensued when using an IPv6-enabled ssh client - you'd get back an RST packet real fast and it was Game Over.
So remember - there's probably more services you need to worry about. ;)
Indeed, which is why a good policy to have for 'servers' is to have: - a hostname, generally I bind these to the EUI-64 address - a servicename, eg 'www' or 'imap', which are bound to ::80 and ::993 Then when the box dies or you want to move the service to another box, you just move the alias, or actually just kill the quagga on the box and let another instance handle it ;) Still the maintainance of the box can be done by directly accessing it. Of course one should simply have that all integrated into the service deployment system and not care about the boxes themselves, you just want <n> of them to provide service X and <m> of them to handle service Z, or to use as many of them so that service Y is running topnotch with capacity to spare. All depends on your size of course ;) Greets, Jeroen