It seems to me that the more sane approach is to give all the machines descriptive names, such as "Dataserver", "webserver1", and "webserver2". As it is, all past references to the database server were named "pliny". Now pliny is going to be a webserver - this is confusing. It would be much less confusing IMO if we had stated "database server" all along and just noted upgrades as an aside when they happen. However, if total logic is ruled out, then please stabilize on some names and give each of those names specific tasks. For example the new name Diderot (or whatever is chosen) will always be the name of the fastest, most powerful database server (no matter what specific hardware it has). The name Pliny will always be used for webserver 1 and the nameLarousse will always be used for webserver 2 (both irrespective of what hardware they have). Adding more machines to the mix will add more names, but those names should be given as-yet-unfilled tasks such as database server 2, and webserver 3, 4, 5... So when we buy our next even more monstrous database server that would become the new Diderot and the current machine with that name would be assigned the new task of database server 2 (in which case a new name would be needed for it).
That way the name "Diderot" would always refer to the primary database server.
-- mav
__________________________________ Do you Yahoo!? Protect your identity with Yahoo! Mail AddressGuard http://antispam.yahoo.com/whatsnewfree
Daniel-
It seems to me that the more sane approach is to give all the machines descriptive names, such as "Dataserver", "webserver1", and "webserver2".
How dare you bring logic into this discussion! I vote that the new server should be called "Maverick", just to piss you off. ;-)
Regards,
Erik
On Dec 2, 2003, at 12:28, Daniel Mayer wrote:
It seems to me that the more sane approach is to give all the machines descriptive names, such as "Dataserver", "webserver1", and "webserver2". As it is, all past references to the database server were named "pliny". Now pliny is going to be a webserver - this is confusing.
The reason we give names to the machines is to tell the *machines* apart.
You'll notice we're not sending people to http://pliny.wikipedia.org or http://larousse.wikipedia.org, but giving them functional work names like en.wikipedia.org (English Wikipedia), download.wikipedia.org (download backups), mail.wikipedia.org (mailing lists)... These are "titles", or job positions, that don't change regardless of who's filling them; they could even be served by multiple machines.
But the machines still have to be able to be individually addressed for us to work on them, and changing their inherent names when we decide to change what they do is a real pain in the ass which would involve waiting for DNS changes to go through and invalidating SSH keys, which is totally insane.
-- brion vibber (brion @ pobox.com)
wikitech-l@lists.wikimedia.org