Wouldn’t it make sense to increase the WebSocket(WS) timeout to 5 or 10 minutes or to adjust it dynamically?
I noticed that it takes ~2 seconds when the packet was received by the TTIG till it starts a new WS connection and another second till the connection is established and the packet get’s send. If the packet is received by multiple gateways deduplication doesn’t work anymore and the packet shows up as “retry” in the TTN console. Also in conditions where rx packet frequency is between 1 to 10 minutes it might cause more load in the backend then leaving the session established…
TTIG WS established → immediate transmit of the packet to TTN
07:12:43.703939 IP LorixOne.44329 > 52.169.76.203.1700: RADIUS, Access-Request (1), id: 0x83 length: 243
07:12:43.711171 IP IC880a-RPi.41703 > 52.169.76.203.1700: RADIUS, Access-Request (1), id: 0x79 length: 243
07:12:43.734337 IP TTIG.17829 > 52.169.76.203.443: Flags [P.], seq 848:1221, ack 1814, win 5840, length 373#
TTIG WS NOT established → 2 seconds till tcp connect:
07:13:45.816602 IP IC880a-RPi.41703 > 52.169.76.203.1700: RADIUS, Access-Request (1), id: 0x0a length: 243
07:13:45.818058 IP LorixOne.44329 > 52.169.76.203.1700: RADIUS, Access-Request (1), id: 0x18 length: 243
07:13:47.912911 IP TTIG.54858 > 52.169.76.203.443: Flags [S], seq 73142392, win 5840, options [mss 1460], length 0
