You likely are running into problems by having 1000 cores and simply don't have enough memory to handle that much. You should look at what you are filtering, sorting, faceting on and see if there are ways you can tune these. Take a look at your cache configurations as well.
What are your index sizes? Also, take a close look at what field types you are using. Do you have "sortables" where you don't need them? More details about what's going on will help troubleshoot, but part of the answer may just be that you have too much on one machine. -Grant On Jan 21, 2010, at 10:02 AM, Matthieu Labour wrote: > Hi > > I have been requested to look at a solr instance that has been patched with > our own home grown patch to be able to handle 1000 cores on a solr instance > > The solr instance doesn't perform well. Within 12 hours, I can see the > garbage collection taking a lot of time and query & update requests are > timing out (see below ) > > [Full GC [PSYoungGen: 673152K->98800K(933888K)] [PSOldGen: > 2389375K->2389375K(2389376K)] 3062527K->2488176K(3323264K) [PSPermGen: > 23681K->23681K(23744K)], 4.0807080 secs] [Times: user=4.08 sys=0.00, > real=4.08 secs] > > org.apache.solr.client.solrj.SolrServerException: > java.net.SocketTimeoutException: Read timed out > at > org.apache.solr.client.solrj.impl.CommonsHttpSolrServer.request(CommonsHttpSolrServer.java:472) > at > org.apache.solr.client.solrj.impl.CommonsHttpSolrServer.request(CommonsHttpSolrServer.java:243) > at > org.apache.solr.client.solrj.request.AbstractUpdateRequest.process(AbstractUpdateRequest.java:105) > > > I used yourkit to track down eventual memory leaks but didn't succeed in > finding one > > The biggest objects using up the memory seem to be org.apache.lucene.Term > and org.apache.lucene.TermInfo > > The total size of the data directory in index is 46G with a typical big core > being 100000 documents and size of 103M > > There are lots of search requests and indexing happening > > I am posting to the mailing list hoping to hear that we must be doing > something completely wrong because it doesn't seem to me that we are pushing > the limit. I would appreciate any tips as where to look at etc... to > troubleshoot and solve the issue > > Thank you for your help ! > > matt
