On Thu, Feb 5, 2009 at 2:02 PM, oleg_gnatovskiy <oleg_gnatovs...@citysearch.com> wrote: > Hello all. I am using distributed search over three servers, and found that a > shard query to one of the boxes sometimes results in two hits to each of the > other boxes. What is the purpose for having 2 hits? Do we need both?
The first hit is to simply gather ids of the top matching documents (and first-phase faceting). The second hit requests stored fields and does highlighting (and possible second-phase faceting). This architecture allows greater scalability and higher aggregate throughput by minimizing the amount of work done. -Yonik
