MetricRegistry instances for transient cores are not getting GC'd

[nandakishorek <na...@gmail.com>]

I took a heap dump when Solr heap usage kept growing to 32GB and comes down
to 15GB after GC.
Heap dump analysis shows there are MetricRegistry instances for transient
cores.i.e., cores with "isLoaded" as false in STATUS output.

<http://lucene.472066.n3.nabble.com/file/t492836/Screen_Shot_2018-07-10_at_10.png> 
<http://lucene.472066.n3.nabble.com/file/t492836/Screen_Shot_2018-07-10_at_10.png> 

I validated that this problem happens on the default configset as well (I
tested it with 5 cores and transientCache size set to 2).

Please help.

My environment is as follows 
 * Solr 7.2.1 in standalone mode. 
 * 32GB heap 
 * 150 cores with data getting continuously ingested to ~10 cores and all of 
the cores queried. 
 * transient cache size is set to 30. 

The solr.xml is as follows 
<?xml version="1.0" encoding="UTF-8" ?> 
<solr> 

  <int name="transientCacheSize">32</int> 
  <str name="shareSchema">true</str> 
  <str name="configSetBaseDir">${configSetBaseDir:configsets}</str> 

  <shardHandlerFactory name="shardHandlerFactory" 
                       class="HttpShardHandlerFactory"> 
    <int name="socketTimeout">${socketTimeout:600000}</int> 
    <int name="connTimeout">${connTimeout:60000}</int> 
  </shardHandlerFactory> 
</solr>



--
Sent from: http://lucene.472066.n3.nabble.com/Solr-User-f472068.html

Re: MetricRegistry instances for transient cores are not getting GC'd

[nandakishorek <na...@gmail.com>]

This is a duplicate of https://issues.apache.org/jira/browse/SOLR-11882.
The problem is fixed in 7.4.



--
Sent from: http://lucene.472066.n3.nabble.com/Solr-User-f472068.html