[Bug]: Event listener won't stay connected after ~2mn of cloud vars inativity

[JloAu]

scratchattach version

2.2.1

What happened?

this only happens with a custom cloud server using a cloudflare tunnel, it works properly with TwCloud or CustomCloud with 127.0.0.1:8080

• I start my python script (init_cloud_server + cloud event listener)
• I load and run the turbowarp project -> the server responds to the requests
• I stay inactive on the turbowarp web page for ~2 minutes, then the issue happens and the server doesn't respond anymore to my requests
• I restart the python script, then everything works again (I did not reload the turbowarp tab)

I was wondering if this was fixable or not? Because the turbowarp tab seems to work correctly even after a long inactivity...

Your code.

server = sa.init_cloud_server("0.0.0.0", server_port, length_limit=None, allow_non_numeric=True, whitelisted_projects=None, allow_nonscratch_names=True, blocked_ips=[], sync_players=True, log_var_sets=False)

server.start(thread=True)

cloud = sa.CustomCloud(project_id=project_id, cloud_host="wss://[random name].trycloudflare.com", username = "EvtListener", length_limit = None, allow_non_numeric = True, _session = None, header = None, cookie = None, origin = None, print_connect_messages = False)

cloud_events=cloud.events()
cloud_events.start(thread=True, ignore_exceptions=True)

Traceback

127.0.0.1:48574 handshaked, project: 1332266638 user: EvtListener
127.0.0.1:48574 disconnected
127.0.0.1:48580 connected
Traceback (most recent call last):
  File "/home/jloau/.venv/lib/python3.12/site-packages/scratchattach/cloud/_base.py", line 180, in read
    self.receive_new(not recv_once, timeout = timeout_end - time.time() if has_timeout else None)
  File "/home/jloau/.venv/lib/python3.12/site-packages/scratchattach/cloud/_base.py", line 156, in receive_new
    received = self.source_cloud.websocket.recv().splitlines()
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jloau/.venv/lib/python3.12/site-packages/websocket/_core.py", line 395, in recv
    opcode, data = self.recv_data()
                   ^^^^^^^^^^^^^^^^
  File "/home/jloau/.venv/lib/python3.12/site-packages/websocket/_core.py", line 423, in recv_data
    opcode, frame = self.recv_data_frame(control_frame)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jloau/.venv/lib/python3.12/site-packages/websocket/_core.py", line 444, in recv_data_frame
    frame = self.recv_frame()
            ^^^^^^^^^^^^^^^^^
  File "/home/jloau/.venv/lib/python3.12/site-packages/websocket/_core.py", line 485, in recv_frame
    return self.frame_buffer.recv_frame()
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jloau/.venv/lib/python3.12/site-packages/websocket/_abnf.py", line 375, in recv_frame
    self.recv_header()
  File "/home/jloau/.venv/lib/python3.12/site-packages/websocket/_abnf.py", line 329, in recv_header
    header = self.recv_strict(2)
             ^^^^^^^^^^^^^^^^^^^
  File "/home/jloau/.venv/lib/python3.12/site-packages/websocket/_abnf.py", line 418, in recv_strict
    bytes_ = self.recv(min(16384, shortage))
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jloau/.venv/lib/python3.12/site-packages/websocket/_core.py", line 581, in _recv
    return recv(self.sock, bufsize)
           ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jloau/.venv/lib/python3.12/site-packages/websocket/_socket.py", line 125, in recv
    bytes_ = sock.recv(bufsize)
             ^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.12/ssl.py", line 1233, in recv
    return self.read(buflen)
           ^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.12/ssl.py", line 1106, in read
    return self._sslobj.read(len)
           ^^^^^^^^^^^^^^^^^^^^^^
ssl.SSLWantReadError: The operation did not complete (read) (_ssl.c:2559)

[TheCommCraft]

Is the error in the cloud server or the cloud client?

[TheCommCraft]

Seems like cloudflare terminates idle connections after some time

[JloAu]

Is the error in the cloud server or the cloud client?

Client I think, but not sure... I forgot to say that the traceback I sent above is spammed until I kill the script, it looks like the event listener tries to reconnect, but fails everytime...

Seems like cloudflare terminates idle connections after some time

That's possible... I noticed that Cloudflare closes idle connections after 125 seconds (after timing it, that's exactly what happens). Is there a way for the event listener to signal its presence every 120 seconds, for example?

[uukelele]

Is there a way for the event listener to signal its presence every 120 seconds, for example?

we could possibly send a heartbeat event down the websocket but it depends entirely on how the cloud server handles this e.g. if there heartbeats supported

[JloAu]

it depends entirely on how the cloud server handles this e.g. if there heartbeats supported

This happens when I use my own cloud server, using sa.init_cloud_server(), so I guess it would be possible to implement it... But idk how other servers would handle this...

[JloAu]

I was wondering if this was fixable or not? Because the turbowarp tab seems to work correctly even after a long inactivity...

I forgot to mention that when there are no cloud activity, I also see the Turbowarp client being disconnected, but it reconnects successfully. I don't know if this can help...

[TheCommCraft]

So the turbowarp website correctly handles the disconnection? Or do you mean the turbowarp client provided by scratchattach?

[JloAu]

The turbowarp website. I saw multiple times something like this while my custom cloud server is running:

127.0.0.1:55086 disconnected
127.0.0.1:32776 connected
127.0.0.1:32776 handshaked, project: 1332266638 user: playerFirefox

Also when I restart the python script (when event listener gets disconnected), the turbowarp website reconnects itself to the custom cloud server after a few seconds.

I noticed that the event listener spams reconnections tries (the traceback of the original post is spammed in the console). Turbowarp website seems not to do that, as it reconnects to the server between 1 and 10 seconds after the server started). Maybe cloudflare doesn't allow that spam? Event listener seems to connect to the custom cloud server, but it gets instantly disconnected

[TheCommCraft]

I am going to create a new release now that should fix this, please tell me if it does.

[TheCommCraft]

The new version is scratchattach==2.2.4

[JloAu]

The new version is scratchattach==2.2.4

I just tried it, unfortunately the event listener still gets disconnected. There are not errors anymore, but I see

127.0.0.1:53948 disconnected
127.0.0.1:53952 connected
127.0.0.1:53952 handshaked, project: 1332266638 user: EvtListener

every 2-5 seconds, and it doesn't receive turbowarp client requests :/

I also get this from cloudflared sometimes:

2026-09-03T14:20:06Z ERR error="Unable to reach the origin service. The service may be down or it may not be responding to traffic from cloudflared: dial tcp 127.0.0.1:35987: connect: connection refused" connIndex=0 event=1 ingressRule=0 originService=http://localhost:35987
2026-09-03T14:20:06Z ERR Request failed error="Unable to reach the origin service. The service may be down or it may not be responding to traffic from cloudflared: dial tcp 127.0.0.1:35987: connect: connection refused" connIndex=0 dest=https://cattle-der-procedure-brief.trycloudflare.com/ event=0 ip=[probably my ipv6] type=ws

I don't know if this can help? cloudflared says that the tunnel is healthy when I start it