hign insanity_count

[Phil Hoy <ph...@friendsreunited.co.uk>]

Hi,

We have a large number of documents, around 200 million in a sharded multi-core solrcloud 4.0 instance. The box that hosts it is huge with 124gb of ram of which 100gb is reserved for solr. I have found that with less RAM available to the JVM we get out of memory exceptions and the solr crashes.
 
Having looked at the cache stats I see we have a number of insanity_counts in the FieldCache. We are running distributed queries with faceting using method=enum and ordering on the 'insane' CensusDate__int field.

Here is the info:

isanity_count:8
insanity#0:SUBREADER: Found caches for descendants of DirectoryReader(segments_kb:1327083993578 _n9b(4.0):C20457412)+CensusDate__int 'DirectoryReader(segments_kb:1327083993578 _n9b(4.0):C20457412)'=>'CensusDate__int',class org.apache.lucene.search.FieldCache$DocTermsIndex,true=>org.apache.lucene.search.FieldCacheImpl$DocTermsIndexImpl#2137364410 'SegmentCoreReader(owner=_n9b(4.0):C20457412)'=>'CensusDate__int',int,org.apache.lucene.search.FieldCache.NUMERIC_UTILS_INT_PARSER=>[I#1036639319
insanity#1:SUBREADER: Found caches for descendants of DirectoryReader(segments_52:1327083896176 _27q(4.0):C540614)+CensusDate__int 'DirectoryReader(segments_52:1327083896176 _27q(4.0):C540614)'=>'CensusDate__int',class org.apache.lucene.search.FieldCache$DocTermsIndex,true=>org.apache.lucene.search.FieldCacheImpl$DocTermsIndexImpl#2034361748 'SegmentCoreReader(owner=_27q(4.0):C540614)'=>'CensusDate__int',int,org.apache.lucene.search.FieldCache.NUMERIC_UTILS_INT_PARSER=>[I#265210210
etc.


What does it mean and how can I resolve it if it is an issue?

Phil