When enabling caching inside a query there is a small decrease in performance without any benefit.
query: bool: must_have: [term: gender: 'M']
40 shards, 5 million documents, 200 million child documents not part of this query(not sure if relevant).
10 servers, CPU load is very low, 30 gigs of ram each with 15 allocated to heap.
How do I go about debugging something like this?