Wikimedia Developer Support

Some content (in extra Namespaces) refuses to index



Is there a way to add specific content to Elasticsearch’s indexes?

I’ve created some new namespaces to hold content that was previously held in the main namespace. I ran the namespacedupes.php maintenance script to correct things. These namespaces are added to $wgContentNamespaces[] However, content in these extra namespaces doesn’t seem to get picked up by Elasticsearch. I’ve recreated the indexes from scratch to no avail.


If you recreated the indices from scratch then it should have worked.
How do you verify that the content in these extra namespaces is not picked up by Elasticsearch?

If you open such a page and add ?action=cirrusDump, is there any content displayed, could you share if possible some example output?

namespacedupes.php will only work at the database level leaving the elasticsearch index in a inconsistent state. The fact you recreated everything should have fixed this inconsistency. Could you share with us what are the commands you’ve run to recreate the indices?

Thank you.


cat /opt/.deploy-meza/


if [ -z "$1" ]; then


cd "$wiki_dir"
for d in $do_wikis; do

        if [ -z "$1" ]; then

        if [ ! -d "$wiki_dir/$wiki_id" ]; then
                echo "\"$wiki_id\" not a valid wiki ID"

        timestamp=$(date +"%F_%T")


        echo "Rebuilding index for $wiki_id"
        echo "  Output log:"
        echo "    $out_log"

        wiki_id="$wiki_id" bash "/opt/.deploy-meza/" > "$out_log"  2>&1

        endtimestamp=$(date +"%F_%T")

        # If the above command had a failing exit code
        if [[ $? -ne 0 ]]; then

                # FIXME #577 #681: add notification/warning system here
                echo "elastic-build-index FAILED for \"$wiki_id\" at $endtimestamp"

        #if the above command had a passing exit code (e.g. zero)
                echo "elastic-build-index completed for \"$wiki_id\" at $endtimestamp"


cat /opt/.deploy-meza/

echo "******* Generating elasticsearch index *******"


# disable search update in wiki-specific settings
echo -e "<?php\n\$wgDisableSearchUpdate = true;\n" > "$disable_search_file"

# Run script to generate elasticsearch index
cd "/opt/htdocs/mediawiki"
WIKI="$wiki_id" php "/opt/htdocs/mediawiki/extensions/CirrusSearch/maintenance/updateSearchIndexConfig.php" --startOver

# Remove search-update disable in wiki-specific settings
rm -f "$disable_search_file"

# Bootstrap the search index
# Note that this can take some time
# For large wikis read "Bootstrapping large wikis" in
WIKI="$wiki_id" php "/opt/htdocs/mediawiki/extensions/CirrusSearch/maintenance/forceSearchIndex.php" --skipLinks --indexOnSkip
WIKI="$wiki_id" php "/opt/htdocs/mediawiki/extensions/CirrusSearch/maintenance/forceSearchIndex.php" --skipParse

echo "******* Elastic Search build index complete! *******"

examples of content not indexed

Error text

There are no reported errors, but looking at the indexing log, I get this:

[             wiki_en] Indexed 7 pages ending at 28517 at 8/second
[837a724cd8e9691cd1bedabc] [no req]   MWException from line 333 of /opt/htdocs/mediawiki/includes/parser/ParserOutput.php: Bad parser output text.
#0 [internal function]: ParserOutput->{closure}(array)
#1 /opt/htdocs/mediawiki/includes/parser/ParserOutput.php(344): preg_replace_callback(string, Closure, string)
#2 /opt/htdocs/mediawiki/includes/content/WikiTextStructure.php(152): ParserOutput->getText(array)
#3 /opt/htdocs/mediawiki/includes/content/WikiTextStructure.php(221): WikiTextStructure->extractWikitextParts()
#4 /opt/htdocs/mediawiki/includes/content/WikitextContentHandler.php(152): WikiTextStructure->getOpeningText()
#5 /opt/htdocs/mediawiki/extensions/CirrusSearch/includes/Updater.php(351): WikitextContentHandler->getDataForSearchIndex(WikiPage, ParserOutput, CirrusSearch)
#6 /opt/htdocs/mediawiki/extensions/CirrusSearch/includes/Updater.php(407): CirrusSearch\Updater::buildDocument(CirrusSearch, WikiPage, CirrusSearch\Connection, integer, integer, integer)
#7 /opt/htdocs/mediawiki/extensions/CirrusSearch/includes/Updater.php(205): CirrusSearch\Updater->buildDocumentsForPages(array, integer)
#8 /opt/htdocs/mediawiki/extensions/CirrusSearch/maintenance/forceSearchIndex.php(218): CirrusSearch\Updater->updatePages(array, integer)
#9 /opt/htdocs/mediawiki/maintenance/doMaintenance.php(94): CirrusSearch\ForceSearchIndex->execute()
#10 /opt/htdocs/mediawiki/extensions/CirrusSearch/maintenance/forceSearchIndex.php(680): require_once(string)
#11 {main}
[             wiki_en] Indexed 50 pages ending at 100 at 120/second


[             wiki_en] Indexed 49 pages ending at 290919 at 134/second
[             wiki_en] Indexed 23 pages ending at 290942 at 134/second
Indexed a total of 191458 pages at 134/second
******* Elastic Search build index complete! *******

Notice how the count starts over, and it completes eventually without complaint. I don’t know how to find the article that is causing a parse problem.


I’ve checked your scripts but at a glance I don’t see anything wrong.
What happens if you try to edit the problematic pages on-wiki to go through the CirrusSearch update process, will they get indexed in the end?

Could you confirm that the Parser issues are related to the pages not being indexed?
forceSearchIndex.php can be given a list of ids using the --ids e.g.:
php extensions/CirrusSearch/maintenance/forceSearchIndex.php --ids 290930

If the only output is the Parser failure then there are good chances that the indexing issues are related to this Parser problem and I’d suggest investigating in this direction.

Thank you.


I used the --ids parameter to successively index pages by ID until I found the malformed content (a page with no title!!). I fixed that page (in User:Talk namespace) by giving it a title and it seems that I was able to index more content by default. However, it did not seem to want to index all the content in the ‘GuidedResearch’ extra namespace. In the end, I used the --namespace 3100 argument to index all the pages in GuidedResearch. I seem to be missing something in configuration to get these high number extra namespaces indexed as content by default. nb. These extra namespaces are added to wgContentNamespaces