Search Overhead Per Shard

Hi all

I have a vague memory and a question I'd like to seek clarity on!

ISTR reading somewhere that elasticsearch has settings to control how search across multiple shards (the usual case) is carried out: either all shards send full results to the node coordinating the search, which in turn is responsible for sorting the combined results from all shards and sending the top hits back to the client.

ISTR further that there is a setting that would allow each shard to send less data to the node coordinating the search (only document IDs and sort keys) and the coordinating node would request more data concerning the documents that actually make the cut.

Not sure if I imagined this setting or if it really exists!

Another point I'd like to seek clarity on is what data the coordinating node collects from each shard for each document that does make the cut. Does a full lucene document with all its fields get sent, or do only the requested fields get sent?

Thanks in advance

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.