Search on our internally hosted mediawiki installation is very poor. Has anyone found an easy way to get better search capabilities, like Solr integration or some internal spidering system?
We installed the Lucene search engine, as used on Wikipedia, and the MWSearch extension that integrates it with MediaWiki.
- Lucene: https://www.mediawiki.org/wiki/Extension:Lucene-search
- MWSearch: https://www.mediawiki.org/wiki/Extension:MWSearch
Installation is complex, but when it's working, you can basically forget about it and let it run.
An additional optimization is the OAIRepository extension that does incremental updates to your search index, rather than full updates all the time. The docs for OAIRepository are pretty bad, so read the discussion page as well. Again, once it's working, it works great.
Make sure your server has plenty of RAM for building the search index. I recommend 6-8GB for a wiki with tens of thousands of articles.