Bug #7487
Updated by Robert Waltz over 8 years ago
After ripping hazelcast out of Log Aggregation, I substituted a ArrayBlockingQueue for what was an originally a hazelcast structure. Instead of sending these solrItem Beans directly to solr, I place them in batches on the ArrayBlockingQueue (the original hazelcast communication structure for all the CNs). Then a rather complex bit of code batches again all the received solrItem Beans and submitted them to an executor service. The executor service iss threaded pool that manages the connections to solr. The executor service appears to have failed somehow. I suspect solr may have been the sticking point, but I am not certain. At least I know this, if the executor service fails, then log aggregation is unable to recover.
Remove the executor service from logAggregation and allow the harvester to submit records directly to solr.
Remove the executor service from logAggregation and allow the harvester to submit records directly to solr.
