Page MenuHomePhabricator

Make site rename requests possible on-wiki without low-level ops intervention
Open, MediumPublic

Description

Bugs like T28725 (Bug 26725) should really be something the community can handle itself (given enough consensus).

As I understand it (and there is no guarantee that I do at this point), this involves DNS changes at this point so we'd probably want some way for the old name to resolve for a while.

I'm thinking that it should be possible to get a list of names that the wiki serves and use that to populate, say, a BIND configuration file.

But I leave this here as a place for people to comment and shoot my idea full of holes.

Details

Reference
bz26994

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 21 2014, 11:15 PM
bzimport set Reference to bz26994.
bzimport added a subscriber: Unknown Object (MLST).

Full site renames are much harder than user renames; there's a few areas to consider:

  • changing the '''database name'''

This is really hard, and makes ops and toolserver admins hate anybody asking for it. Ideally we should never have to do it but we have a lot of stuff that maps language & site suffix directly to/from DB names, or uses the DB names directly for some cross-wiki operations.

If the mapping of '''site key name''' <-> '''database name''' can be abstracted, this at least some of that may become easier.

  • changing the '''language code prefix''' for the subdomain and interwiki links

Currently I think some manual mappings can be done to change the visible language & subdomain without changing the database name, but it leaves the "different" db name exposed in other places as above.

Note also that things like secure.wikimedia.org can need additional mappings. Ugh!

Clarified the summary to avoid confusion with user renames.

bugs wrote:

(In reply to comment #0)

Bugs like Bug#26725 should really be something the community can handle itself
(given enough consensus).

I disagree here, actually. Why should a community be able to rename itself? There aren't really enough cases that I could see this being necessary. Does it really happen enough that we need to take it out of the sysadmins' hands?

What it needs is to be doable without an entire company's worth of sysops going "gosh this is too hard, let's put it off another few years".

What it needs is to be doable without an entire company's worth of sysops going
"gosh this is too hard, let's put it off another few years".

Hence Bug#27002

(In reply to comment #1)

If the mapping of '''site key name''' <-> '''database name''' can be
abstracted, this at least some of that may become easier.

What do you mean by "site key name"? Something fairly closely related to the URL?

(In reply to comment #6)

(In reply to comment #1)

If the mapping of '''site key name''' <-> '''database name''' can be
abstracted, this at least some of that may become easier.

What do you mean by "site key name"? Something fairly closely related to the
URL?

Brion. ^