Manual:Backing up a wiki

Translate this page

It is important to make regular backups of your wiki (data and files). This page provides an overview of the backup process for a typical MediaWiki wiki; you will probably want to devise your own backup scripts or schedule to suit the size of your wiki and your individual needs.

Help:Export is a quick and easy way to save all pages on your wiki.

Overview

MediaWiki stores important data in two places:

Database: Pages and their contents, users and their preferences, metadata, search index, etc.
File system: Software configuration files, custom skins, extensions, images (including deleted images), etc.

Consider making the wiki read-only before creating the backup - see $wgReadOnly . This makes sure all parts of your backup are consistent (some of your installed extensions may write data nonetheless).

File transfer

You will have to choose a method for transferring files from the server where they are:

Non-private data you can simply publish on archive.org and/or in a dumps/ directory of your webserver.
SCP (or WinSCP), SFTP/FTP or any other transfer protocol you choose.
The hosting company might provide a file manager interface via a web browser; check with your provider.

Database

Most of the critical data in the wiki is stored in the database. If your wiki is currently offline, its database can be backed up by simply copying the database file.

When using the default MySQL or MariaDB backend, the database can be dumped into a script file which can be used later to recreate the database and all the data in it from scratch.

MySQL

Automysqlbackup

See the package on Debian:

$ apt show automysqlbackup
[...]
Description: automysqlbackup creates backup every day, week and month for all of your MySQL database, to a configured folder. There's nothing to do but to install this package, and you'll rest assured that you have a way to go back in the history of your database.
[...]

Install the package:

# apt install automysqlbackup

All your databases will be saved in /var/lib/automysqlbackup/:

$ find /var/lib/automysqlbackup/
/var/lib/automysqlbackup/
/var/lib/automysqlbackup/weekly
/var/lib/automysqlbackup/weekly/my_wiki
/var/lib/automysqlbackup/weekly/my_wiki/my_wiki_week.18.2016-05-07_15h32m.sql.gz
/var/lib/automysqlbackup/monthly
/var/lib/automysqlbackup/daily
/var/lib/automysqlbackup/daily/my_wiki

Manual backup:

# automysqlbackup

Restore a database:

gunzip < /var/lib/automysqlbackup/weekly/my_wiki/my_wiki_week.18.2016-05-07_15h32m.sql.gz|mysql -uUSER -pPASSWORD my_wiki

For other distributions, see on Sourceforge.

Mysqldump from the command line

The most convenient way to create a dump file of the database you want to back up is to use the standard MySQL dump tool mysqldump from the command line. Be sure to get the parameters right or you may have difficulty restoring the database. Depending on database size, mysqldump could take a considerable amount of time.

First insert the following line into LocalSettings.php

$wgReadOnly = 'Dumping Database, Access will be restored shortly';

this can be removed as soon as the dump is completed.

Example of the command to run on the Linux/UNIX shell:

mysqldump -h hostname -u userid -p --default-character-set=charset dbname > backup.sql

Substituting hostname, userid, charset, and dbname as appropriate. All four may be found in your LocalSettings.php (LSP) file. hostname may be found under $wgDBserver ; by default it is localhost. userid may be found under $wgDBuser , charset may be found under $wgDBTableOptions , where it is listed after DEFAULT CHARSET=. If charset is not specified mysqldump will likely use the default of utf8, or if using an older version of MySQL, latin1. While dbname may be found under $wgDBname . After running this line from the command line mysqldump will prompt for the server password (which may be found under Manual:$wgDBpassword in LSP).

See mysqldump for a full list of command line parameters.

The output from mysqldump can instead be piped to gzip, for a smaller output file, as follows

mysqldump -h hostname -u userid -p dbname | gzip > backup.sql.gz

Some newer versions of MySQL might show an error about tablespaces and PROCESS privilege. MediaWiki does not use tablespaces. The solution is to add the --no-tablespaces option to the command:

mysqldump --no-tablespaces -h hostname -u userid -p dbname | gzip > backup.sql.gz

A similar mysqldump command can be used to produce XML output instead, by including the --xml parameter.

mysqldump -h hostname -u userid -p --xml dbname > backup.xml

and to compress the file with a pipe to gzip

mysqldump -h hostname -u userid -p --xml dbname | gzip > backup.xml.gz

Additional options you should consider using with mysqldump for a backup includes the following.

Additional Mysqldump Options
Option	Description
--default-character-set	Specify default character set
--no-tablespaces	Do not write CREATE LOGFILE GROUP or CREATE TABLESPACE statements in output
--single-transaction	Issue a BEGIN SQL statement before dumping data from server
--triggers	Dump triggers for each dumped table
--routines	Dump stored routines (procedures and functions) from dumped databases
--events	Dump events from dumped databases
--add-drop-table	Add DROP DATABASE statement before each CREATE DATABASE statement
--create-options	Include MySQL-specific table options in CREATE TABLE statements
--extended-insert	Use multiple-row INSERT syntax

If you do not use --single-transaction, then you should consider using the --lock-tables and --add-locks options.

Due to an unexpected change in MySQL versions 5.7.41 and 8.0.32 in February 2023, the --single-transaction option required the backup user to have RELOAD or FLUSH_TABLES privileges. The issue was fixed in MySQL versions 5.7.42 and 8.0.33. See MySQL Bug 109685 and Ubuntu Bug 2003866 for details.

Remember to backup the additional file system components used by the wiki that might be required during a restore, like images, logo, skins and extensions.

Running mysqldump with Cron

Cron is the time-based job scheduler in Unix-like computer operating systems. Cron enables users to schedule jobs (commands or shell scripts) to run periodically at certain times or dates.

A sample command that you may run from a crontab may look like this:

nice -n 19 mysqldump -u $USER --password=$PASSWORD $DATABASE -c | nice -n 19 gzip -9 > ~/backup/wiki-$DATABASE-$(date '+%Y%m%d').sql.gz

The nice -n 19 lowers the priority of the process.

Use valid values for $USER, $PASSWORD, and $DATABASE. This will write a backup file with the weekday in the filename so you would have a rolling set of backups. If you want to save the files and extensions as well, you might want to use this one.

Warning:

Do not attempt to back up your MediaWiki database using mysqlhotcopy. The table format used by MediaWiki cannot be backed up with this tool, and it will fail silently!

If you want to add this task in Cron through Cpanel then you must escape the character "%"

/usr/bin/mysqldump -u $USER --password=$PASSWORD $DATABASE -c | /bin/gzip > ~/backup/wiki-$DATABASE-$(date '+\%Y\%m\%d').sql.gz

or you will get an error:

/bin/sh: -c: line 0: unexpected EOF while looking for matching `''
/bin/sh: -c: line 1: syntax error: unexpected end of file

Running mysqldump with Systemd

Systemd unifies service configurations and control. Timers are systemd unit files that control service files or events. Timers can be used as an alternative to cron. An example of systemd unit files and backup script is shown below.

wiki-backup.timer

The following timer runs the wiki-backup service at 5:10 AM every morning.

$ cat /etc/systemd/system/wiki-backup.timer

[Unit]
Description=Run the backup service once a day
Documentation=...

[Timer]
OnCalendar=*-*-* 05:10:00
RandomizedDelaySec=600
Persistent=true

[Install]
WantedBy=timers.target

wiki-backup.service

When the wiki-backup timer fires, then the service is invoked. The service runs a script located in /sbin.

$ cat /etc/systemd/system/wiki-backup.service

[Unit]
Description=Run the backup service once a day
Documentation=...

[Service]
Type=oneshot
User=root
ExecStart=/sbin/wiki-backup

wiki-backup script

$ cat /sbin/wiki-backup

#!/usr/bin/env bash

# Systemd adds random paths at times. Take full control of PATH.
PATH=/bin:/sbin:/usr/bin:/usr/sbin
export PATH

# Read the backup password from conf or ini Failed
wiki_password=...

# Fix the wiki tables just in case. This step produces a lot of noise,
# so send stdout to /dev/null.
if MYSQL_PWD="${wiki_password}" \
   mysqlcheck my_wiki --auto-repair --user=mwuser 1>/dev/null;
then
    echo "Repair wiki database ok"
else
    echo "Failed to repair wiki database"
    echo "Continuing anyways"
fi

# Disable the connection from Apache to MySQL for the dump
if ! systemctl stop apache2.service ;
then
    echo "Failed to stop Apache service"
    echo "Continuing anyways"
fi

# Lock option choice due to MySQL change at versions 5.7.41 and 8.0.32 in
# February 2023. See https://bugs.mysql.com/bug.php?id=109685 and
# https://bugs.launchpad.net/ubuntu/+source/mysql-5.7/+bug/2003866.
if mysql --version 2>&1 | grep -q -E 'mysql[[:space:]]+Ver 8\.0\.32'; then
   echo "Using MySQL --lock-tables --add-locks options"
   mysql_lock_opt="--lock-tables --add-locks"
else
   echo "Using MySQL --single-transaction option"
   mysql_lock_opt="--single-transaction"
fi

if MYSQL_PWD="${wiki_password}" \
   mysqldump --no-tablespaces \
     ${mysql_lock_opt} \
     --events --triggers --routines \
     --add-drop-table --create-options \
     --extended-insert \
     --default-character-set=utf8 \
     -u mwuser -h localhost my_wiki | gzip -q -v9 > /backup/wiki-backup.sql.gz ;
then
    echo "Dump wiki database ok"
else
    echo "Failed to dump wiki database"
    echo "Continuing anyways"
fi

# Re-enable connection from Apache to MySQL for the dump
if ! systemctl start apache2.service ;
then
    echo "Failed to start Apache service"
    echo "Continuing anyways"
fi

exit 0

Tables

Some of the tables dumped have different degrees of temporariness. So to save disk space (beyond just gziping), although those tables need to be present in a proper dump, their data does not. However, under certain circumstances the disadvantage of having to rebuild all this data may outweigh saving disk space (for example, on a large wiki where restoration speed is paramount).

See mailing list thread mysql5 binary schema about the topic.

Latin-1 to UTF-8 conversion

See the relevant section of the upgrading page for information about this process. Also see the talk page for more information about working with character sets in general.

PostgreSQL

You can use the pg_dump tool to back up a MediaWiki PostgreSQL database. For example:

pg_dump mywiki > mywikidump.sql

will dump the mywiki database to mywikidump.sql.

To restore the dump:

psql mywiki -f mywikidump.sql

You may also want to dump the global information, e.g. the database users:

pg_dumpall --globals > postgres_globals.sql

SQLite

If your wiki is currently offline, its database can be backed up by simply copying the database file. Otherwise, you should use a maintenance script: php maintenance/SqliteMaintenance.php --backup-to <backup file name>, which will make sure that operation is atomic and there are no inconsistencies. If your database is not really huge and server is not under heavy load, users editing the wiki will notice nothing but a short lag. Users who are just reading will not notice anything in any case.

phpMyAdmin

Turn your wiki to read only by adding $wgReadOnly = 'Site Maintenance'; to LocalSettings.php.

Find the wiki database in LocalSettings.php. Here is an example of what this looks like in LocalSettings.php:

## Database settings
$wgDBtype           = "mysql";
$wgDBserver         = "localhost";
$wgDBname           = "sashtmax_mw19999";
$wgDBuser           = "sashtmax_mw19999";
$wgDBpassword       = "S7[88p]jJJ";

Open the browser to your phpadmin link, login, choose the wiki database.
Select Export. Make sure all items under Export are highlighted, and make sure Structure is highlighted (it's important to maintain the table structure). Optionally check Add DROP TABLE to delete existing references when importing. Make sure Data is checked.
Select zipped.
Click on GO and save the backup file.^[1]
Remove $wgReadOnly = 'Site Maintenance'; from LocalSettings.php

Remember to also backup the file system components of the wiki that might be required, e.g. images, logo, and extensions.

External links

Backing up the Database

HeidiSQL (alternative to phpMyAdmin)

HeidiSQL is similar to phpMyAdmin, but without any restrictions of phpMyAdmin's free version. HeidiSQL requires a direct database connection, where some hosts may only offer web interfaces (phpMyAdmin) to firewalled databases.

File system

MediaWiki stores other components of the wiki in the file system.

The most important of these are:

LocalSettings.php
uploaded files in the images/ directory (including deleted files, thumbnails, and rendered math and SVG images, if applicable).

The best method to back these up is to place them into an archive file, such as a .tar file, which can then be compressed if desired. On Windows, applications such as WinZip or 7-zip can be used.

For Linux variants, assuming the wiki is stored in /srv/www/htdocs/wiki

tar zcvhf wikidata.tgz /srv/www/htdocs/wiki

It should be possible to backup the entire "wiki" folder in "htdocs" if using XAMPP.

Configuration files

LocalSettings.php is the most important of these, but a wiki might also have things like .htaccess or other web server configuration files that should be backed up.

Uploaded files

Files uploaded to the wiki are by default put into the images/ directory, separated into subdirectories such as images/8/8f. There are also other directories such as images/archive/ and images/deleted/. These should all be backed up.

The images/thumb/ can be backed up along with everything else, but can optionally be excluded in order to save backup space. This directory stores the derived thumbnails of images and other files; generally multiple thumbnails per wiki file. After restoring from backup, these thumbnails will be recreated as required (although depending on $wgGenerateThumbnailOnParse this may need to be a manual process).

Backup the content of the wiki (XML dump)

It is also a good idea to create an XML dump in addition to the database dump. XML dumps contain the content of the wiki (wiki pages with all their revisions), without the site-related data (they do not contain user accounts, image metadata, logs, etc).^[2]

XML dumps are less likely to cause problems with character encoding, as a means of transferring large amounts of content quickly, and can easily be used by third party tools, which makes XML dumps a good fallback should your main database dump become unusable.

To create an XML dump, use the command-line tool dumpBackup.php , located in the maintenance directory of your MediaWiki installation. See Manual:DumpBackup.php for more details.

You can also create an XML dump for a specific set of pages online, using Special:Export, although attempting to dump large quantities of pages through this interface will usually time out.

To import an XML dump into a wiki, use the command-line tool importDump.php . For a small set of pages, you can also use the Special:Import page via your browser (by default, this is restricted to the sysop group).

See Manual:Importing XML dumps for more information.

Without shell access to the server

See m:Data dumps about Wikimedia database dumps.

WikiTeam3

If you have no server shell access, use Save the Web Project's WikiTeam3 Python 3 script, (full instructions are at that link).

Windows: When using --images, because NTFS does not allow characters such as :*?"<>| in filenames, some files may not be downloaded, please check the errors.log file.

Example usage

--curonly dumps only the latest revision of pages
--xml exports an XML dump, uses Special:Export by default when no other xmldump method is specified.
--xmlrevisions uses API:Allrevisions (MediaWiki 1.27+) xmldump method. Recommended as it's quicker and puts almost no pressure on the MediaWiki backend compared to Special:Export.
--images generates an image dump
--force generates a dump even if there is one already at Internet Archive

Public wikis: wikiteam3dumpgenerator <WIKI_URL> --xml --xmlrevisions
Private wikis: To dump a private wiki you will have to use a login that has at least read permission on the wiki.; wikiteam3dumpgenerator <WIKI_URL> --xml --xmlrevisions --force --user <USER> --pass <PASSWORD>; If that doesn't work. Login with a web browser, save the site cookies in Netscape format to cookies.txt, and add option --cookies cookies.txt

Scripts

With server shell access