Wikibase/Suite
Introduction
A Wikibase install often consists of the following elements:
- An RDBMS (MariaDB/MySQL or SQLite) [required]
- MediaWiki [required]
- The Wikibase extension [required]
- Other related extensions [optional]
- The Wikidata query service (WDQS) [optional]
- The WDQS proxy
- The WDQS frontend
- QuickStatements [optional]
- Elasticsearch [optional]
The Docker install contains all these components and can be started with a single docker-compose
command.
This install meets the needs of many Wikibase users, especially those just starting out.
However, for those not wanting to run Docker (because of additional administrative overhead, scaling issues in production, or a desire to understand the components better), here's how to install each component manually.
Network configuration
Consult the network description and diagram: https://www.mediawiki.org/wiki/Wikibase/Maintaining#Architecture_overview
Before you dive in
There's a large amount of technical documentation to be read and enjoyed at the Wikibase technical doc site.
To successfully navigate this process, you should probably have:
- Familiarity with Linux and web server software
- One or more servers on which you have permissions to install software
Installing the Wikibase suite
MariaDB
Notes
- MediaWiki creates a database structure outlined in this document: MediaWiki database layout
- Wikibase's database schema is described here: Wikibase database schema
Instructions
The MediaWiki install linked in the next step relies on a running RDBMS with a database and user already created.
We recommend using MariaDB, and though a full installation guide is outside of the scope of this document, start with this excellent collection of documentation and this tutorial.
MediaWiki
Notes
- Note that while the vanilla install of MediaWiki (correctly) notes PostgreSQL as a working option, Wikibase does not work with PostgreSQL.
- For this install of MediaWiki, we recommend using nginx with php-fpm.
- This configuration assumes you have php-fpm running on the same host; to use a network port use
fastcgi_pass HOST:9000
. - You will need the following PHP extensions (install doc):
calendar ctype dom fileinfo iconv intl json mysql mbstring xml
Instructions
Install instructions for MediaWiki
Working MediaWiki nginx config
Wikibase extension
Notes
Install zip
and unzip
for Composer to work correctly.
The Wikibase data model is described in these documents:
Instructions
Installation guide for Wikibase
Installation guide for Composer
Optional sidebar
If you'd like to add some Wikibase-related links to your main page sidebar, you can add these configuration stanzas to your LocalSettings.php.
Extensions
There are many extensions often used with Wikibase. Only some require special install instructions for a Wikibase context; the list below includes information about all extensions included in our Docker images.
OAuth
Wikibase needs the OAuth extension for the Wikidata query service, as you'll see below.
Installation
The OAuth install instructions are comprehensive. Note that you'll benefit from a working understanding of User rights to finish this install.
Other extensions
These extensions can be installed simply using their own installation instructions.
- Extension:ConfirmEdit
- Extension:Nuke
- Extension:EntitySchema
- Extension:UniversalLanguageSelector
- Extension:CLDR
- Extension:VisualEditor
- Extension:Scribunto
- Extension:Babel
- Extension:Wikibase EDTF
- Extension:Wikibase Local Media
- Extension:WikibaseManifest
Native and non-native data types
Extending Wikibase's functionality beyond its native data types may require the installation of additional extensions.
Native data types
- commonsMedia = "string"
- entity-schema = "string"
- external-id = "string"
- geo-shape = "string"
- globe-coordinate = "globecoordinate"
- monolingualtext = "monolingualtext"
- quantity = "quantity"
- string = "string"
- tabular-data = "string"
- time = "time"
- url = "string"
- wikibase-item = "wikibase-entityid"
- wikibase-property = "wikibase-entityid"
Non-native data types
Many other data types exist and are implemented in some instances of Wikibase, such as Wikidata. Because these other instances are frequently highly customized, below is a list of officially supported data types and the extensions they require:
- mathematical expression ("string") -- requires Extension:Math
- musical notation ("string") -- requires Extension:Score
- lexeme ("wikibase-entityid") -- requires Extension:WikibaseLexeme
- local media ("string") -- requires Extension:Wikibase Local Media
Wikidata Query Service
The Wikidata Query Service (WDQS), which consists of the Blazegraph SPARQL query backend and its updater, can be and often is used in conjunction with Wikibase.
Instructions
While intrepid users may wish to compile their own instance of WDQS, we recommend using the precompiled version.
- Download the desired version from Archiva.
- 0.3.42 is the recommended version as of September 2020. However, expect this information to be updated with the impending MediaWiki release.
- You can see the changes in each version in the WDQS code repository browser.
- To configure the query service, consult both the Wikitech WDQS guide and the WDQS user manual.
- In particular, configure RWstore.properties as noted here: Wikidata Query Service/User Manual .
Elasticsearch
Overview
Elasticsearch is a Lucene-based search engine. To get it working with MediaWiki/Wikibase, you'll need the following software:
- Three MediaWiki extensions:
- Two Elasticsearch plugins:
Instructions
- Install Elasticsearch 6.5.4 (refer to its install instructions ).
- ES's install process will prompt you to install OpenJDK.
- Install the aforementioned extensions on your MediaWiki instance:
- Add the following lines to
LocalSettings.php
:
wfLoadExtension( 'Elastica' );
wfLoadExtension( 'CirrusSearch' );
wfLoadExtension( 'WikibaseCirrusSearch' );
$wgCirrusSearchServers = [ '<name/IP of host running ElasticSearch>' ];
$wgSearchType = 'CirrusSearch';
$wgCirrusSearchExtraIndexSettings['index.mapping.total_fields.limit'] = 5000;
$wgWBCSUseCirrus = true;
- Install the aforementioned Elasticsearch plugins using these commands:
> $PATH_TO_ELASTICSEARCH/bin/elasticsearch-plugin install org.wikimedia.search:extra:6.5.4
> $PATH_TO_ELASTICSEARCH/bin/elasticsearch-plugin install org.wikimedia.search.highlighter:experimental-highlighter-elasticsearch-plugin:6.5.4
- Modify
elasticsearch.yml
as needed:- Name your cluster with the
cluster.name
directive. - Make a single-node cluster:
discovery.zen.minimum_master_nodes: 1
- Set
network.host
as needed
- Name your cluster with the
- Modify
jvm.options
as needed (above all, set-Xms
and-Xmx
based on the amount of memory available to your host) - Note the "flood stage" settings for your disk size.
- You can either ensure that at least 5% of the disk is available at all times, or you can alter the
cluster.routing.allocation.disk.watermark.flood_stage
setting. For more information, consult this StackOverflow article.
- You can either ensure that at least 5% of the disk is available at all times, or you can alter the
- Run
UpdateSearchIndexConfig
for CirrusSearch:
php extensions/CirrusSearch/maintenance/UpdateSearchIndexConfig.php --startOver
- Start jobs running in the background:
php /var/www/html/maintenance/runJobs.php --wiki my_wiki --wait &
- Run
ForceSearchIndex
:
php extensions/CirrusSearch/maintenance/ForceSearchIndex.php --queue --maxJobs 10000 --pauseForJobs 1000 --skipLinks --indexOnSkip
QuickStatements
Instructions
- Using Git, clone the two repositories into your web root (hereafter
www_root
):- QuickStatements.
- MagnusTools.
(cd MagnusTools; git checkout e71b214d56d6f68a097871a510ba39ffc6fc96ee)
- In
www_root/quickstatements
, runcomposer install
. - For the files that follow, determine the appropriate values in your installation for the following environment variables, and set them. They can also be easily substituted in place using envsubst.
MW_SITE_NAME="site name"
WB_PUBLIC_SCHEME_HOST_AND_PORT=<public ip>
WIKIBASE_SCHEME_AND_HOST=<localhost>
WB_PROPERTY_NAMESPACE=122
WB_ITEM_NAMESPACE=120
WB_PROPERTY_PREFIX="Property:"
WB_ITEM_PREFIX="Item:"
- Create the following files with the contents specified.
/www_root/quickstatements/public_html/config.json
:
{
"site" : "${MW_SITE_NAME}" ,
"bot_config_file" : "/var/www/html/bot.ini" ,
"logfile" : "/var/log/quickstatements/tool.log" ,
"sites" : {
"${MW_SITE_NAME}" : {
"oauth" : {
"language":"${MW_SITE_LANG}" ,
"project":"${MW_SITE_NAME}" ,
"ini_file":"/quickstatements/data/oauth.ini" ,
"publicMwOAuthUrl":"${WB_PUBLIC_SCHEME_HOST_AND_PORT}/w/index.php?title=Special:OAuth" ,
"mwOAuthUrl":"${WB_PUBLIC_SCHEME_HOST_AND_PORT}/w/index.php?title=Special:OAuth" ,
"mwOAuthIW":"mw"
} ,
"server" : "${WB_PUBLIC_HOST_AND_PORT}" ,
"api" : "${WIKIBASE_SCHEME_AND_HOST}/w/api.php" ,
"pageBase" : "${WB_PUBLIC_SCHEME_HOST_AND_PORT}/wiki/" ,
"toolBase" : "${QS_PUBLIC_SCHEME_HOST_AND_PORT}/" ,
"types" : {
"P" : { "type":"property" , "ns":"${WB_PROPERTY_NAMESPACE}" , "ns_prefix":"${WB_PROPERTY_PREFIX}" } ,
"Q" : { "type":"item" , "ns":"${WB_ITEM_NAMESPACE}" , "ns_prefix":"${WB_ITEM_PREFIX}" }
}
}
}
}
/quickstatements/data/oauth.ini
:
; HTTP User-Agent header
agent = 'Wikibase QuickStatements'
; assigned by Special:OAuthConsumerRegistration (request modelled after https://www.wikidata.org/wiki/Special:OAuthListConsumers/view/77b4ae5506dd7dbb0bb07f80e3ae3ca9)
consumerKey = '<OAUTH_CONSUMER_KEY>'
consumerSecret = '<OAUTH_CONSUMER_SECRET>'
- Set these environment variables according to your installation:
MW_ADMIN_NAME
: MediaWiki username with administrator rightsQS_PUBLIC_SCHEME_HOST_AND_PORT
: External IP of your query service instance
- With the variables set, create an OAuth consumer by executing the following php script on your MediaWiki installation where OAuth is installed:
php /var/www/html/extensions/OAuth/maintenance/createOAuthConsumer.php \
--approve \
--callbackUrl $QS_PUBLIC_SCHEME_HOST_AND_PORT/api.php \
--callbackIsPrefix true \
--user $MW_ADMIN_NAME \
--name QuickStatements \
--description QuickStatements \
--version 1.0.1 \
--grants createeditmovepage \
--grants editpage \
--grants highvolume \
--jsonOnSuccess
- A successful response will have the "created" flag set to "true", as in this example:
{"created":true,"id":1,"name":"QuickStatements","key":"30e532f04e1bdf63ac281fcbc819170c","secret":"f60f31ad4196af40bb0598e1d4d3435a3515604e","approved":1}
For more information, consult the Docker image README.