Record a problem now.
Application of new on-line, on-line for the first time deployed two nodes, mapped out by the NGINX DMZ.
After on-line, the third access suddenly found very slow, there is a 50% chance maintained at about seven seconds, the log observation platform code handling time is about 40ms.
Open F12, found TTFB time consumed 6 seconds. Guess nginx configuration problems, but we write nginx configuration is sent to the operation and maintenance colleagues do not go wrong.
Guess is the DMZ to the application server's firewall does not get through. The last position is the reason.
Phenomenon, the occasional visit for too long, NGINX hash routing consuming too long.
The cause, NGINX Hash routing in which a station can not access, resulting in circuitous access, time-consuming.