When you create an index collection with the type Apache Solr, the resulting search index is stored in your Apache Solr infrastructure, not within the system instances.
Hitachi Content Intelligence supports external indexes running on Solr server versions 5 and 6.
These considerations apply to how the system interacts with external indexes in your Solr infrastructure:
- When you register an existing external Apache Solr index, the existing fields from that index are not automatically made visible to your users. To make them visible:
- In the query settings for the index collection, on the Fields tab, click Actions > Remove All Fields to remove the default fields.
- Click Actions > Add All Fields to add all existing fields from the index schema to the Fields tab.
- The 1.1 version of the Solr plugin sends SolrJ 5.5.3 requests (with JSON, when available). Solr calls that don't have explicit SolrJ support are implemented directly using the SolrJ HTTP clients.
- Your system uses the Solr commitWithin feature for indexing documents. This means that documents are quickly made available in searches, but are not guaranteed to be committed to disk. When a workflow task reaches an internal checkpoint, the systemsends a request to Solr to commit these documents to disk.
- After sending a commit request during a workflow task checkpoint, the system assumes that all documents have been successfully committed. If you change the commit behavior for an index in Apache Solr, any node outage on the Solr server can cause you to silently lose indexed documents. In this situation, the system does not retry indexing the failed documents unless you clear the workflow task and start it over.
- The default internal solrconfig.xml is configured to autoCommit to disk every 3 min without a searcher being opened
- SolrJ add/edit requests are made specifying a commitWithinMs parameter of 30 seconds. This setting is not configurable.
For information on creating index collections, see Adding index collections.