Hi,

We are observing below error in Nginx Error Log.
At same time we see sudden spike in Active & Waiting Connection from 3k to 13-15k for a while which comes back to normal within a minute.

We have also observed that some time during this sudden spike , server goes into Hung State with no helpful message in OS Kernel messages to debug or find out the reason for hung.


Below is the actual message:
Also attached is the graph of connection behaviour.
2017/05/09 16:01:49 [alert] 23370#23370: accept4() failed (107: Transport endpoint is not connected)

We are using CentOS 7.2 and Nginx 1.10.2 on our system.
Please help , what could be the probable reason for sudden spike. We have analyzed number of requests coming on the server at time of Spike, that doesnt increase.

Regards,
Anish