The Signpost

Technology report

Bugs, Repairs, and Internal Operational News

September Engineering Report published

The Wikimedia Foundation's Engineering Report for September was published last week on the Wikimedia Techblog and on the MediaWiki wiki, giving an overview of all Foundation-sponsored technical operations in that month. Many of the projects mentioned have been covered in The Signpost, including the switchover to protocol-relative URLs, the release of the new mobile site, and the ongoing deployment of 1.18 to Wikimedia wikis. The report's writers also chose to highlight a new roadmap of what the development team will be working on over the coming months (although the map appears considerably incomplete at time of writing).

Also announced were the successful replication of article text data from the WMF's main bank of servers, in Tampa, Florida, to the new data centre in Ashburn, Virginia; the first trials of basic Wikimedia Labs functionality; that the WMF was looking into ways of accepting text-based reviews of articles in addition to the current system of star rankings; and a recent overhaul to the system of gadgets (which will, it is hoped, allow for a WMF shared gadget repository). A test wiki that ran the very latest MediaWiki revisions (to emulate a process known as continuous integration, which Wikimedia hopes to adopt as a standard in the near future) is expected in early October, with full https support later in the month.

It was not all positive news, however. Also mentioned in the report, under the heading "lowlights", was the results of an investigation into three short outages that occurred on 26 September. The investigation concluded that the first was caused by an important cable being "accidentally knocked loose" during separate maintenance work (proposed solution: add more redundancy to the older server racks), that the second was caused by the 1.18 upgrade affecting the database cluster responsible for the CentralAuth login functionality (solution: potentially give it its own cluster), and that the third was caused by a combination of the 1.18 upgrade and a series of particularly expensive database queries being run at the time (solution: kill queries more effectively in future).

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 many weeks.


















Wikipedia:Wikipedia Signpost/2011-10-03/Technology_report