First, that resource leak is worrying. Is there any way you could take a stack trace and/or memory dump? I suppose it’d be easy enough to simulate. It’s particularly worrying because SolrJ is how Solr<->Solr communications happen so if there really is more than transitory leak that’d affect Solr as well.
Second, I don’t think there’s a method that does exactly what you want, anything like “isOneNodeHealthyForEachSlice()”, but there is DocCollection.getActiveSlices() so I guess you could ask if (docCollection.getActiveSlices().length == known_number_of_shards)… Best Erick > On Mar 1, 2019, at 2:24 PM, Webster Homer <webster.ho...@milliporesigma.com> > wrote: > > I am using the CloudSolrClient Solrj api for querying solr cloud collections. > For the most part it works well. However we recently experienced a series of > outages where our production cloud became unavailable. All the nodes were > down. That's a separate topic... The client application tried to launch > searches but always experienced a SolrServerException that there were no live > nodes available. After a few hundred such exceptions, the application ran out > of memory and failed when trying to allocate a thread... I'm not sure where > the resources are being leaked in exception handling. Is there a way to ask > the CloudSolrClient if there are enough replicas to execute the search. > > I'm using Solr 7.2
