Wikipedia:Wikipedia Signpost/2013-01-14/Technology report
Intermittent outages planned, first Wikidata client deployment
Wikidata deployed to Hungarian Wikipedia
The Wikidata client extension was successfully deployed to the Hungarian Wikipedia on 14 January, its team reports. The interwiki language links can now come from Contact the development team page on Wikidata, on #wikimedia-wikidata on IRC, or on the test2wiki where interested users who do not speak Hungarian can test it.
Development of phase 2, relating to infobox-style data entries is progressing well, the Wikidata team reports, though again code review is likely to be a slow and arduous process and no deployment dates have yet been set. In related news, the 3 millionth Wikidata item was created on January 13: "List of mayors of Westdorpe", a large village in the Netherlands.
Intermittent outages expected during primary data centre switchover
The Wikimedia operations team is busy preparing to switch over the master database servers and other key infrastructure to use the Ashburn, Virginia data centre as the primary data centre. The Ashburn data centre is already serving about 90% of traffic, but this is mainly the result of it hosting caching servers which serve pages for logged-out users, as well as images, JavaScript and CSS. When Wikimedians edit a wiki, the regeneration of the pages is still handled by master database servers in the Tampa data centre. The switchover is part of an ongoing project to enable redundancy for key infrastructure; in addition, the change in "primary" centre from Tampa to Ashburn revolved around the quality of the facilities at both locations.
CT Woo, the Foundation's Director of Operations, announced the change on the wikitech-l mailing list:
| “ | The team has scheduled the week of 22nd January, 2013 to perform the switchover. We are going to block a 8-hour migration window on the 22nd, 23rd and 24th. During those periods, 17:00 UTC to 01:00 UTC (9am to 5pm PST), there will be intermittent blackouts. ... We will make the final Go/No decision on 18th January. | ” |
In the event of an outage, Wikimedians can get updates on IRC in the #wikimedia-tech channel, or via Wikimedia's Twitter accounts @wikimediatech (detailed server logs).
As the operations and core platform team are setting up new servers and infrastructure, they are taking it as an opportunity to also switchover to new deployment tools (git-deploy) and a new internal workflow for the regular code updates to the Wikimedia projects.
Echo project holds office hours
The Echo team held IRC office hours on January 8 where they discussed development progress on a Facebook-style notification system for editors. As previously reported, Echo, currently deployed on MediaWiki.org, provides notifications when someone edits your talk page, creates a link to an article you created, nominates it for deletion, adds maintenance tags, or reverts your edit. There will be a notifications "badge" at the top of the page, next to your user name, which can replace the "yellow" bar that users see now when you have a new talk page message. The team described how they have a "'user mention' notification in the works" which could work similar to Twitter mentions and notify you when someone mentions you on another page. The team welcomes feedback on the types of notifications to provide, although they initially have in mind new users, who are more likely to miss important events than established users. Echo may also provide a public notifications API that could be used by bots and scripts, the team said, though they, together with the community, would need to figure out how to balance ease of use of the tool while minimizing abuse and spamming with the notifications API.
Echo is editor engagement mailing list.
- Disclaimer: User:Aude is an employee of Wikimedia Germany, working on the Wikidata project.
In brief
Not all fixes may have gone live to WMF sites at the time of writing; some may not be scheduled to go live for several weeks.
- Experimental features pushed to mobile: Following the release of a new mobile Wikimedia experience in 2012, the Wikimedia Foundation has further extended their beta to include new features. This update is intended to "unleash the full creativity of our engineers and designers without disrupting the user experience for millions of readers". Specifically, logged-in mobile users can now upload pictures to the Wikimedia Commons directly from their browser, while Wikimedia editors will be interested in the new watchlist and section editing tools. These features, while functional, are described as experimental at this time.
- Monthly engineering report published: The Wikimedia Foundation has released their December 2012 technical report, highlighting important developments including the deployment of VisualEditor to the English Wikipedia, as well as progress on many of the Foundation's less-well-known projects.
- Universal Language Selector deployment broadens: The Internationalisation team at the Wikimedia Foundation is deploying the Universal Language Selector on wikis with the Translate extension currently enabled (Meta-Wiki, MediaWiki.org, Outreach wiki, Incubator and the Wikimania wikis). The Universal Language Selector is replacing the similar Narayam and WebFonts extensions, which will be disabled on those wikis. The Selector, which enables easier switching between interface languages and input scripts, will be available only to logged-in users, as there are still caching issues for logged-out users; the plan is to release it to all users later in the year.
- Staff promotion at Foundation: Patrick Reilly has been promoted to the role of Site Performance Engineer and Senior Technical Advisor at the Wikimedia Foundation. Reilly has previously worked on the MobileFrontend extension and as technical lead for the mobile development team.
- Why isn't there a way to switch from the Tampa to the Ashburn servers without causing any outages? Whoop whoop pull up Bitching Betty | Averted crashes 18:44, 16 January 2013 (UTC)[reply]
- I'm sure operations folks are aiming for no outages and there might well be none. (hopefully) But there's some risk involved in this sort of switchover and maybe the transition isn't 100% perfect. --Aude (talk) 19:56, 16 January 2013 (UTC)[reply]
- Repeating what Aude said for the most part. There isn't some gigantic, Frankenstein-style switch that can just be levered. There are a lot of moving parts and things that can go wrong. As the sayings go, "The map is not the terrain," and "No plan, however well-conceived, survives contact with the enemy."--Jorm (WMF) (talk) 20:57, 16 January 2013 (UTC)[reply]
- Well if both clusters are running and synchronising before the switch over, then careful DNS migration should provide a seamless transition, provided synchronisation can be maintained. However most architectures have weak points, and atomicity is probably ours. Rich Farmbrough, 19:12, 19 January 2013 (UTC).[reply]
- I think Echo can only reach its potential if people reply your message on your user talk page, not on their own talk page. OhanaUnitedTalk page 03:20, 17 January 2013 (UTC)[reply]
- Good point, that. MikeLynch (talk) 17:53, 21 January 2013 (UTC)[reply]
- Good point, that. MikeLynch (talk) 17:53, 21 January 2013 (UTC)[reply]