Skip to main content

Resolve "SolrCore is loading" errors in solr-exporter metrics collection

Updated

Issue

When the Solr Prometheus Exporter (solr-exporter pod) attempts to scrape metrics from Solr, it may return an HTTP 503 error with the message SolrCore is loading. This indicates that the targeted Solr core has not yet finished initializing and is unavailable for metrics collection.

 

Diagnosis

The error can appear intermittently, particularly after Solr pods restart, indexes are large, or recovery operations are in progress.
To confirm, check the exporter logs for entries similar to:

org.apache.solr.client.solrj.impl.BaseHttpSolrClient$RemoteSolrException: Error from server ... "msg":"SolrCore is loading", "code":503

You can also verify the status of the core by running:

kubectl exec -n <fusion-namespace> <solr-pod> -- curl "http://localhost:8983/solr/<core-name>/admin/ping?wt=json"

If the response shows "status":"loading", the core is still initializing.

 

Environment

Fusion 5.x on Kubernetes (EKS, AKS, GKE, or other certified distributions)
Solr 9.x with Solr Prometheus Exporter enabled for metrics collection

 

Cause

The SolrCore is loading response occurs when a request is made to a Solr core before it has fully loaded its index into memory. Common triggers include:

  • Pod restarts or redeployments

  • Large index sizes causing extended load times

  • Solr recovering replicas after cluster changes

  • Index merges or segment warm-up

 

Resolution

Monitor Solr core load progress

  • Use the Solr Admin UI or the admin/cores API to check if the core has transitioned from loading to active.

  • Example API call:

kubectl exec -n <fusion-namespace> <solr-pod> -- curl "http://localhost:8983/solr/admin/cores?action=STATUS&wt=json"

 

Verify Solr pod health

  • Inspect pod logs for core load completion messages:

kubectl logs <solr-pod> -n <fusion-namespace>

  • Look for log entries indicating the core has finished loading.

 

Allow time before metrics scraping

If exporter pods attempt to scrape during core loading, the request will fail. In environments with frequent Solr restarts or large indexes, consider:

  • Increasing scrape intervals in Prometheus configuration

  • Introducing an initial delay after Solr pod startup before scraping begins

 

Investigate recurring long load times

If the loading state consistently lasts an unusually long time, investigate index size, replica recovery configuration, and Solr heap allocation.

 

 

 

 

Powered by Zendesk