netstat

You can check network connection information on the server.

Since the states of the TCP 3-way handshake and TCP 4-way handshake processes are displayed, it's crucial to understand them well.

You will most frequently encounter the LISTEN, TIME_WAIT, and ESTABLISHED states, and each state can be interpreted as follows.

• LISTEN: The state where a process is listening for requests through a socket.
• ESTABLISHED: The state where a connection has been established.
• TIME_WAIT: The state where a connection has been terminated, and the active closer (the side that initiated the disconnection) waits briefly before cleaning up the socket.
  • This is to allow the active closer to resend the last ACK packet gracefully to terminate the connection if it gets lost during the FIN process of the 4-way handshake. (If the last ACK is lost, the passive closer, not having received the ACK, resends the previously sent FIN. Since the active closer is in the TIME_WAIT state, it can receive the FIN again before cleaning up the socket. The active closer can then send an ACK to gracefully terminate the connection.)

keepalive-timeout: One of the settings in Nginx that maintains connections to prevent network performance degradation that would occur if a TCP 3-way handshake were performed every time.

In HTTP/1.0, this was implemented by explicitly adding a Connection: keep-alive header, but from 1.1 onwards, the concept of persistent connections was introduced into the specification itself, so no separate header needs to be added.

CLOSE_WAIT: This is the state after receiving a FIN packet and before creating an ACK request. If this state persists, it indicates a situation where the application cannot generate a response, requiring root cause analysis and corrective action.

In summary, you can check network connection status using the netstat command.

You can check endpoint IP and port information for connections, as well as socket states like LISTEN, ESTABLISHED, and TIME_WAIT.

The CLOSE_WAIT state should be interpreted as a situation where the application cannot clean up the socket and fails to generate a response, thus requiring identification of the cause and corrective action.