Re: [Archive-access-discuss] Nutch-wax upgrade problem

[Michael S. <st...@du...>]

I do not believe such an index convertion tool exists (Check the nutch 
list).   Even if it did, I'd suggest you'd spend so much CPU running the 
convertion of index and supporting segments, you might as well start 
over (New nutch/hadoop runs much faster.. about X4 times faster).  
Starting over, you can be sure of the process, more sure than you can be 
of a little-tested transform, and you will pick up improvements made 
since old nutch.

The ClassCastException in the below is because old nutchwax used an UTF8 
class to represent Strings, a class since replaced by the Text class 
(Your new nutch frontend is trying to use Text to represent a UTF8 class 
read from segment directories I'm guessing).

St.Ack


Xavier Torelló wrote:
> Hi,
>
> First of all, thanks for your quick response :)
>
> The re-index option is not viable, since it is a expensive process 
> considering that we have about 150gb in indices.
>
> John talk about the option of convert the indexs. Somebody knows how to 
> do this process?
>
>
> Finally, the exception that appears when we try to make a request to 
> nutchwax (via opensearch):
>
> java.lang.RuntimeException: java.lang.ClassCastException: 
> org.apache.hadoop.io.Text
>
> 	org.apache.nutch.searcher.FetchedSegments.getSummary(FetchedSegments.java:204)
> 	org.apache.nutch.searcher.NutchBean.getSummary(NutchBean.java:344)
> 	org.archive.access.nutch.NutchwaxBean.getSummary(NutchwaxBean.java:52)
> 	org.apache.nutch.searcher.OpenSearchServlet.doGet(OpenSearchServlet.java:156)
> 	org.archive.access.nutch.NutchwaxOpenSearchServlet.doGet(NutchwaxOpenSearchServlet.java:76)
> 	javax.servlet.http.HttpServlet.service(HttpServlet.java:689)
> 	javax.servlet.http.HttpServlet.service(HttpServlet.java:802)
>
>
> Thanks,
>
>