Here is what I discovered after dozens of reindexes. We have a tool that is pulling all of the documents' uniqueIds. This tools is causing the cache to fill up. We turned it off and the system was able to reindex.
Here is what is still puzzling to me about this entire scenario. When we had only 1 core active I was able to reindex the core even with the tool filling up the document cache. As soon as I added a second empty core the OOM stuff started. Could this be caused by the second core allowing the document cache to "leak" into it? Just seems strange that a second empty core allows the system to run out of heap. -- Jeff Newburn Software Engineer, Zappos.com jnewb...@zappos.com - 702-943-7562 > From: Mark Miller <markrmil...@gmail.com> > Reply-To: <solr-user@lucene.apache.org> > Date: Tue, 06 Oct 2009 17:21:47 -0400 > To: <solr-user@lucene.apache.org> > Subject: Re: Solr Trunk Heap Space Issues > > Mark Miller wrote: >> Jeff Newburn wrote: >> >>> So could that potentially explain our use of more ram on indexing? Or is >>> this a rare edge case. >>> >>> >> I think it could explain the JVM using more RAM while indexing - but it >> should be fairly easily recoverable from what I can tell - so no >> explanation on the OOM yet. Still looking at that one. >> >> Is you system basically stock, or do you have custom plugins in it? >> >> > No matter what I try with however many cores, I can't duplicate your > problem. > > -- > - Mark > > http://www.lucidimagination.com > > >