Oak Lucene Indexing issue or partial values removal from multivalued property

[shubhanshu_singh]

We are encountering an issue related to full-text search and indexing. Here's the observed behavior:

We have a custom multi-valued property indexed for full-text search. Whenever tags are added—either to cq:tags or custom properties—we store the localized tag titles in this property. This approach is intended to enable search functionality for localized tag titles in Omnisearch.

When new tags are added, the multi-valued property is updated via our workflow process, and the newly added values are searchable both in asset search and through queries.

However, when a tag is removed, its corresponding localized value is also removed. Despite this, the asset still appears in search results, which is incorrect. The issue seems to be that the index is not reflecting this change, and the updated asset data is not being properly indexed after the tag removal.

As per our analysis the problem is that the removal of a tag and its localized value (or any new multivalued property) does not trigger the correct update in the search index, leading to incorrect search results.

For example, suppose we have an asset with the tag "summer" (cq:tags). In the corresponding multi-valued property, we store the localized tag title: "summer" in English and "verano" in Spanish. These localized titles are then searchable through Omnisearch, enabling users to search for "summer" or "verano."

When new tags are added, the multi-valued property is updated via our workflow process, and the newly added values are searchable both in asset search and through queries.

For instance, if a new tag "vacation" is added, the property will now include localized values like "vacation" in English and "vacaciones" in Spanish.

However, when a tag is removed, its corresponding localized value is also removed. For example, if the "summer" tag is removed, the localized values "summer" and "verano" should also be removed from the multi-valued property. Despite this, the asset still appears in search results when searching for "summer" or "verano", which is incorrect.

[narendiran_ravi]

Please refer to this documentation to investigate why the reindexing is not happening when the multivalued property is updated. Also, if possible please share the custom index config added for this property

[shubhanshu_singh]

Thanks for the response @narendiran_ravi , I went through the documentation mentioned and everything seems OK. There are not a lot of assets on local as well and neither an overload of updates.

Index definition is

<<property>
        jcr:primaryType="nt:unstructured"
        analyzed="{Boolean}true"
        name="jcr:content/metadata/<property>"
        nodeScopeIndex="{Boolean}true"
        propertyIndex="{Boolean}true"/>

This is added to damAssetLucene-11-custom-xx on cloud