I would start by reviewing the logs to see if there are any issues/error immediately prior to this occuring.  It's worrth paying particular attention to the performance snapshot log in the lead up to the issue.

Also, at the point of failure, look at the number of connections to port 8443, and possibly things like open files.  Since it sounds like this happens after a period of time, then perhaps you are maxing something out or reaching some limit imposed by the OS.

This is my lab server, EPS is about 200-400 

when version on 8.6.0...it could work fine about 3 or 4 months......but upgrade to 8.6.1,...this issue seem occur in one month

I have collect log folder as a gz file...whether I should create a case for it ?

Wencheng