Issue
After upgrading Fusion and Solr, Solr index datasources may process documents at a significantly reduced rate compared to previous versions. This can occur even when CPU and memory usage remain below specified limits, and all other configurations appear unchanged.
Diagnosis
-
Indexing jobs process documents at a slower rate after upgrading Fusion (and Solr), with throughput reduced by approximately half compared to the previous version.
-
The job configuration, CPU, and memory allocations remain unchanged from the previous deployment.
-
No errors are observed in resource utilization monitoring.
-
Solr logs display errors similar to:
Error instantiating cache => org.apache.solr.common.SolrException: Error loading class 'solr.LRUCache'
Environment
-
Fusion 5.9.x (Self-Hosted on Kubernetes)
-
Solr 9.1.x (as included with Fusion 5.9.x)
-
Amazon EKS or other supported Kubernetes platforms
Note: The specific Solr and Fusion versions in use must be confirmed, but this guidance applies to all recent Fusion 5.x releases using Solr 9.x.
Cause
Solr’s LRUCache implementation has been deprecated and removed in recent versions. If the cache type in solrconfig.xml remains set to LRUCache after upgrading, Solr cannot instantiate the cache, which can negatively impact performance for indexing and query operations.
Resolution
Update your Solr configuration to use CaffeineCache instead of the deprecated LRUCache.
Update solrconfig.xml
-
Open the
solrconfig.xmlfile for the affected Solr collections. -
Locate any cache configuration entries referencing
LRUCache. For example: -
Change the
classattribute fromsolr.LRUCachetosolr.CaffeineCache: -
Repeat this update for all cache entries referencing
LRUCache(such asqueryResultCache,documentCache, etc.). -
Save your changes and apply the updated
solrconfig.xmlto the relevant Solr cores or collections. -
Restart the Solr pods as necessary for the new configuration to take effect.
Note: For further details on CaffeineCache, refer to the Solr documentation.