Elasticsearch off-heap memory usage and behavior on Windows Server

I'm using Elasticsearh 6.5.2 on Windows Server 2016.
This memory of this server sometimes reaches 99% and I'm investigating the reason and risk of this phenomenon.
The OS memory is 8GB, -Xmx4096m, and -Xms4096m.

The allocated heap memory is 4 GB, so I'm thinking the other memory usage is caused by off-heap memory.
https://www.elastic.co/guide/en/elasticsearch/reference/current/heap-size.html

The heap memory usage and it's up and down is under the heap max and sync with gavage collection.
This is normal behavior and I think there is no risk.
However, sometimes OS memory alerts it's 99% usage and there is no high usage on heap memory.

I reproduced and collect performance data for a similar and simple case.
After server reboot, all of the memory indexes have been stable until query execution from Kibana.

What I can't understand is the reason for jumping not only heap memory and also OS memory.
The server has 8GB memory, so it's almost 100%.
In addition, the increase of Java working set and decreasing of java private working set means a surge of shareable memory.
And most The most mysterious thing is sudden decrease of Java working set after several minutes.

All of this phenomenon seems to be happening outside the heap area, but I can't explain rationally.

Please help me to understand what is happening.
And let me know whether or not this 99% of memory usage is safe or risky.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.