online incremental indexing

[zzzzz shalev <zz...@yahoo.com>]

hello all,
   
  i have an environment with a number of search instances (index searchers) running as rmi servers and a federator (a parallel multi searcher) combining the results of all the instances, this is working great, and allows us to load into memory millions of docs.
   
  my problem is as follows:
   
  i want to update the indexes but i do not want any down time,
  looking at the parallel multi searcher code i see that if one of the searchables throws an io exception the multisearcher doesnt bother dealing with that searchable's results (am i correct? if so this is great), 
   
  i would like to write my own remotesearchaer that adds a reload function that will close the indexsearcher and reload the indexes into memory after they have been updated.
   
  however this means that closing and then reopening/reloading an indexsearcher (running as a search instance) would result in a specific instance not returning results during the load time.
   
  is there a better solution?
   
  thanks

		
---------------------------------
Yahoo! Photos
 Ring in the New Year with Photo Calendars. Add photos, events, holidays, whatever.