I agree with Tim. There is on-going work to support building SOLR index using MapReduce. See http://issues.apache.org/jira/browse/SOLR-1045. The initial patch does what Tim described - build Lucene indexes the same way that SOLR would do. Later patches would be different as it will use the new mapreduce package when it comes out in 0.20.
Cheers, Ning On Wed, Mar 11, 2009 at 5:43 AM, tim robertson <[email protected]> wrote: > I am definitely not an expert in SOLR, and I would like to be > corrected, but to my knowledge unless you have built the Lucene > indexes the same way that SOLR would do so (handling types etc the > same way) then you won't be able to use SOLR search with the indexes > built. > > Assuming they are the same - once you have built the indexes, I think > you then need to copy them out of HDFS into a vanilla SOLR > installation - e.g. it has nothing to do with hadoop anymore. > > If your indexes are very large, you might consider Katta to search > across them. I don't think this plays well with SOLR though. > > I am also interested in using with SOLR indexes built from HBase so I > am interested in your findings. > > Cheers, > > Tim > > > > On Wed, Mar 11, 2009 at 10:37 AM, Trọng Trần Bình > <[email protected]> wrote: >> Hi all, >> I have Index repository by Lucene in Hadoop, I am using Solr for web search >> interface. How can I using those indexed repositories with Solr web search? >> >
