We have two environments. One environment has two ES hot data nodes, another has five ES hot data nodes. The nodes server type is similar, all other config also similar. However, we tried to ingest the same data set in those two environments. To my surprise, the performance with two hot data nodes is better than the performance with five hot data nodes! Is that normal? In another word, is it possible because more data nodes will cost more time to do the data allocation/copying? If not, what's the best practice/approaches to find the bottomneck?