On Mon, Oct 7, 2019 at 3:05 PM Kévin Bois <kevin.bois@biblissima-condorcet.fr> wrote:

Hello,
[...]
It looks like an index error, so I tried to rerun updateSearchIndexConfig.php maintenance script, but this one gives me the same kind of error too.
`
/mediawiki/extensions/Elastica/vendor/ruflin/elastica/lib/Elastica/Bulk.php: Error in one or more bulk request actions:

index: /mw_cirrus_metastore_first/mw_cirrus_metastore/namespace-mediawiki--2 caused blocked by: [FORBIDDEN/12/index read-only / allow delete (api)];
`
I'm a bit confused as I've managed to run this once when I first installed CirrusSearch, so I don't really understand why it wouldn't work now.

This error may appear when elasticsearch is running low on disk space, it can switch affected indices to read-only, please see https://www.elastic.co/guide/en/elasticsearch/reference/6.7/disk-allocator.html and the index.blocks.read_only_allow_delete property to reset the read-only flag.

In short: double check disk usage on the partition where Elasticsearch has its data directory, reset the read-only flag for all indices marked as read-only and rerun the maint scripts as described in the CirrusSearch README file.

Finally, autocompletion seems to be working even though the search itself doesn't, but it only autocompletes Item:QID instead of working on the item label, do you have any idea why ?

The reason you still see results for QID queries is that WikibaseCirrusSearch still performs a query on the mysql database when it detects QIDs in the search query it's why you may see search results even though the elasticsearch infrastructure is broken.