Note that there is a MediaWiki-RfCs project and a documentation one. (If that makes sense, don't know, description of this task is short and a bit cryptic. :)

https://www.mediawiki.org/wiki/Requests_for_comment/Master/slave_datacenter_strategy_for_MediaWiki

Related patch set, implementing a cache that is replicated across data centers: https://gerrit.wikimedia.org/r/#/c/187074/

RFC meeting was held 2015-03-18 (Meetbot log) , conclusion is:

AGREED: yes we want this, AaronSchulz please continue with designing and prototyping (TimStarling, 21:58:22)

@aaron, can you update the RFC with

ACTION: AaronSchulz to expand "Code that uses caches must be aware of whether it needs to do explicit purges (use the WAN cache) or can use the local cache" to explain how developers know which to use

and other concerns from the discussion, e.g. purging of elasticsearch and Swift (out of scope?), duration of datacenter cookie, "tombstone"? etc.

Then it feels like one more IRC discussion should get the gold stamp of approval :)

The page was updated a bit. The swift/elastic stuff isn't really the scope of the RfC. Of course, I am interesting in getting swift replicated at "live speed" through some means or another. Media doesn't have to be included in the first phase though, and it's mostly a swift-specific discussion. The same goes for elastic.

@aaron thanks. If you think you're done then move this to "under discussion" (per Architecture committee process rule 432-T35 :) ).

We'll schedule this for discussion on IRC on September 2. @ori, @aaron, is that OK with you?

IRC meeting scheduled for 17 hours from now. See the Phab event board (E54) for more

2015-09-02 IRC office hour summary, in which this was approved with several action items.