Configuration of the number of shards , under ElasticSearch version 1.3.1

(Whu_cs_yang) #1


I have made a test,

12 nodes, each of them are the same Hardware configuration:

 cpu MHz         : 2099.998

 8 core and 16G Mem

The main configuration in ElasticSearch.yml, most of the other items use default value.

 node.master: true true

 index.number_of_shards: 8

 index.number_of_replicas: 2

 bootstrap.mlockall: true

 gateway.type: local

We have indexed 180 million documents, storage space: 2.2T(single shard), 6.7T(total shards). Then we make query test with the cluster and found even a single query with one phrase need average 200~300 seconds! This is unbelievable, is there something wrong with my configuration? Is the number of shards too small? How to decide the number of shards? What is the basis for this setting? And I want to know what the settings that impact the search performance greatly?

Thank you for your time very much!‍

(system) #2