Why 100% CPU usage when using elasticsearch?

Question

linux

I use on the server:
- ElasticSearch 2.3.2
- 3 indexes, 2,000,000 documents in total
When getting data via hot_threads, everything is OK - joxi.ru/EA4dpJIDRRwzAb is always less than 2% of the CPU.
But here in htop I see such a picture - joxi.ru/82Qj7RS1GG1xAd
Java is to blame for everything?! What is the best thing to do?

Answer 1

About caches: generally speaking, they are, they are simply called mapping. It looks like you didn't set it up. See the index mapping document, it is also created by default, but it is far from always optimal. Fields that are not searchable - do not index. Where it is possible to optimize the index itself, for example, the timestamp should be explicitly indexed as a date according to a template, and not just a string, multilingual search with declensions should be left only where it is really necessary, etc.
Elastic is not a silver bullet or AI, it can adapt to the data, but it's more efficient to do it yourself.
Also, don't forget to pick up the warmers.

Answer 2

Well, the software works, it does something there.
He should eat the processor, especially on such a volume of documents.
It doesn't stay at 100% all the time, does it?