How to handle server disconnect

[Joe-ChenZ]

During redeployment, the server sends disconnect() to some clients but not all clients. I am using socketio reconnection mechanism to reconnect the client, but the issue with a server disconnect is that it seems like the client would not reconnect after. When I try to use socketio.connect() again for the client, i got an exception for Already Connected. Using python socketio as the client. I need the client to stay connected with the server. What is the best way to handle this?

flask socketio version: 5.5.1
python-socketio version: 5.13.0
python-engineio version: 4.12.3

[Joe-ChenZ]

disconnect, and reconnect doesn't work either. EngineIO drops right after the reconnect attempt. Putting a sleep in the middle doesn't work either

[miguelgrinberg]

First of all, it's really difficult to do a graceful shutdown of a server that has active WebSocket connections. My guess is that some clients do not receive the disconnect because you are killing the server(s) before they get a chance to properly disconnect everyone.

What I don't understand is what you say about your clients. The clients should disconnect anyway. It should take a few more seconds for them to realize that the server is gone, but they eventually will, and at that point they'll attempt to reconnect. That's all I can say from your high level description. If you need more details, you need to share logs from the client and the server that show the problem clearly.

[Joe-ChenZ]

say about your clients. The clients should

My issue with the client is that they don't seem to reconnect, even after a long time after they receive the server disconnect. It seems like the server client state might be out of sync. The client thinks it is connected to the server but when it emits, it gets namespace is not a connected namespace. The server thinks the client is not connected, but the client thinks it is connected to the server.

[miguelgrinberg]

A verbal description of the situation is not very helpful. Please provide the logs that I requested if you want me to review the issue.