Greetings,

I've been out of touch for a while and need to touch base with the boys
from the Lucy 'hood :)

1.  We're using r1167124 on a largish cluster with ~200m pages - and
cluster search performance has always been terrible.
2.  I see from http://www.apache.org/dist/lucy/CHANGES-0.3.2.txt that:

 * [LUCY-204] - Process ClusterSearcher RPCs in parallel
 * [LUCY-205] - Parallel processing for SearchServer
      This introduces an incompatible API change: The port argument has moved
      from the constructor to the serve method and the password argument has
      been removed.

Which frankly looks mouth watering.

So:

1.  If we svn up to latest, I take it we'll need to reindex all?

2.  Is the take-away from LUCY-204/5 that cluster searching now operates in
parallel?  ie, none of the old serial-by-search-node which leads to 10s -
60s+ search times (instead of sub-second if using a single node)?

Pray tell!

Thanks

Reply via email to