Make a database dump

Revision as of 21:41, 28 November 2005

This explains how to make a wiki dump, which is not the same thing as a full database backup. The xml dumps will then be available on http://download.wikimedia.org/ .

XML dumps generation

The backup should be run on the dump holder (benet as of August 26th 2005). As root simply start the backup script:

benet# /home/wikipedia/bin/backup-all

Dumps are currently placed in /var/backup with public and private directories below that.

how it works

backup-all call backup-site with the correct parameters needed for each WikiMedia project (quote, tionary, books, pedia, news).

The real jobs is handled by backup-wiki that dump several tables (see 'included' section below). The xml dump itself is generated with MediaWiki dumpBackup.php maintenance script.

A list of titles in the MAIN namespace is generated for each wiki (all_titles_in_ns0.gz).

Once done, symbolic links are set as:

pages_current.xml.gz
pages_full.xml.gz

included tables

public

site_stats, image, oldimage, pagelinks, categorylinks, imagelinks

private

user, watchlist, ipblocks, logging, user_rights, archive

not included

• The database configuration tables in the mysql database
• Any non-wiki tables (OTRS, bugzilla and such)
• Some useful wiki tables:
  • recentchanges
  • searchindex
  • imagelinks (so all use info for images will be lost from image *description pages)
  • interwiki
  • linkscc (so DB load will be high while it's rebuilt)
  • math
  • objectcache
  • querycache
  • hitcounter
  • user_newtalk

Template:PDDev