I got same issue when using 4.10.2. I suspected this issue will cause trouble when using too many reducers. Then I tried to use less reducers, and made it work. I do not think map-reduce contrib in this version is stable... Anyway it is free.
On Tue, Apr 21, 2015 at 10:56 PM, ralph tice <ralph.t...@gmail.com> wrote: > Hello list, > > I'm using mapreduce from contrib and I get this stack trace: > https://gist.github.com/ralph-tice/b1e84bdeb64532c7ecab > > Whenever I specify <luceneMatchVersion>4.10</luceneMatchVersion> in my > solrconfig.xml. 4.9 works fine. I'm using 4.10.4 artifacts for both map > reduce runs. I tried raising maxWarmingSearchers to 20 and set > openSearcher to false in my configs with no difference. > > I have started studying the code, but why would BatchWriter invoke warming > (autowarming?) on a close, let alone opening a new searcher? Should I be > looking in Lucene or Solr code to investigate this regression? I also > notice there are interesting defaults for FaultTolerance in SolrReducer > that don't appear to be documented: > > https://github.com/apache/lucene-solr/blob/trunk/solr/contrib/map-reduce/src/java/org/apache/solr/hadoop/SolrReducer.java#L70-L73 > but reading https://issues.apache.org/jira/browse/SOLR-5758 sounds like > they are either unimportant or overlooked? > > Also, we will probably be testing mapreduce contrib with 5.x, has anyone > been successful with this yet or are there any known issues? I don't see a > lot of changes on contrib/map-reduce... > > Regards, > > --Ralph Tice > ralph.t...@gmail.com > -- Regards, Shenghua (Daniel) Wan
