Technical details

From wikidev.net

Daily full backup

The entire /var dir which holds all databases, mails, uploaded data etc is synced each night from the current primary server 'runner' to the secondary server in Berlin (sedna.wikidev.net). Ten days worth of full backups are kept there at any time, so even software errors on the primary could do little harm. In case the primary server disappears completely the secondary is ready to take over- it runs exactly the same software, and the data is already there. There's a heartbeat running between the two servers that sends text messages if runner doesn't respond to http requests so that i can react immediately.

I'm investgating automatic failover potentially involving DRBD (http://www.drbd.org), OpenSSI (http://www.opensi.org) and SuperSparrow (http://www.supersparrow.org). My older network concept for wikipedia (http://www.aulinx.de/oss/code/wikipedia/) that i since went on to implement (most of it) has some information on this as well. See meta:Wikimedia servers for the current wikipedia status and fancy images.

See also: Getting started - MediaWiki user documentation