We are running a small cluster with 6 machines with 1TB of load data (Elasticsearch v5.6). We are running performance tests using 10 specific queries in a bash file calling with curl.
In the first test running the average time of the 10 queries are 1m30s, if I execute the same test the times goes down getting values bellow 500ms. I'd tried to delete cache node by node using the command bellow but it doesnt worked. Any suggestion about how to get the correctly time response? I want to perform the test without any cache data.
curl -XPOST "http://<node_IP>:9200/_cache/clear"
Clearly the elasticseach is doing some mechanism to store and retrieve the data after the first query.
Thanks in advance.