The following is my experience with this issue in the 2.x branch, which I believe is related.
In order to remove a dependency on lucene, a change was made at some point: https://issues.apache.org/jira/browse/NUTCH-930 However, the change inadvertently messed up the tstamp field - although the field was still a long in the schema, what was being written wasn't actually a long as you observed. Instead of fixing this change to store a long correctly, the schema used by nutch was updated to use "date" instead of "long" and necessary changes made to allow this. This means the updated nutch expects to be reading a "date". Therefore one option that should work (if you have the latest nutch) is to start using "date" for the "tstamp" field as opposed to "long" in your schema - this is what nutch is expecting. If you can change your schema this is what I would recommend trying. Take a look at the schema that comes with the latest nutch to see what I'm talking about. Since I wasn't sure at the time if anyone else was using the existing "long" field for Solr I didn't want to change the schema, so what I did was change the code to use "long" instead. Another option, but then you have to maintain code separately. To do this I had to change the SolrDeleteDuplicates code and also the code that actually indexes the field has to be changed to long too from date (BasicIndexingFilter). -- View this message in context: http://lucene.472066.n3.nabble.com/updating-from-nutch-1-2-to-nutch-1-7-with-Solr-1-4-1-dedup-crashes-tp4077567p4077612.html Sent from the Nutch - User mailing list archive at Nabble.com.
