websocket does not resume

[AnthonyZJiang]

Summary

websocket does not resume after closure

Reproduction Steps

I wrote a simple code to monitor incoming messages from a few servers and pass them to a discord bot for reposting. The selfcord code only receives message and the discord bot only reposts messages. My code worked fine until recently (1-2 weeks ago).

Normally the bot runs fine for a few hours but at some point gateway stops responding and closes. Most of the time, it can resume but, as a rough estimate, 1 out of 5 times it does not resume and stops there with no further log output, and surely no longer connected. I had to restart the bot to get it working again.

There is no exception or all raised ones are captured in try/except block, so no traceback. I will see if I can log debug in a log file.

Code

import selfcord

from .sender import MessageSender

class MessageReceiver(selfcord.Client):
    def __init__(self, channels: list, sender: MessageSender):
        super().__init__()
        self.channels = channels
        self.sender = sender
            
    async def on_message(self, message: selfcord.Message):
        if message.channel.id not in self.channels:
            return
        self.sender.forward_message(message)

Expected Results

Closing websocket with code 4000
2025-04-21 22:16:46 INFO     selfcord.gateway Finished closing websocket
2025-04-21 22:16:47 INFO     selfcord.gateway Gateway has successfully RESUMED session *********************.

Actual Results

2025-04-22 00:15:43 WARNING  selfcord.gateway Gateway has stopped responding. Closing and restarting.
2025-04-22 00:15:43 INFO     selfcord.gateway Closing websocket with code 4000
2025-04-22 00:15:43 INFO     selfcord.gateway Finished closing websocket

System Information

• Python v3.11.2-final
• discord.py v2.5.2-final
• aiohttp v3.11.16
• system info: Linux 6.6.51+rpt-rpi-v8 ✨ add trusted users kwarg to bot class #1 SMP PREEMPT Debian 1:6.6.51-1+rpt3 (2024-10-08)

Checklist

• I have confirmed I am using unmodified discord.py-self and not the upstream discord.py.
• I am using a user token (and it isn't visible in the code).
• I have searched the open issues for duplicates.
• I have searched closed issues for resolved duplicates.
• I have confirmed the development version doesn't resolve the issue.
• I have shared the entire traceback.

Additional Information

I can confirm my internet is running 24x7 as the message sender is online all the time. The sender's websocket also closes occasionally, but oddly not the same time and way less frequent.

[AnthonyZJiang]

Added some lines to print debug messages around the websocket closure calls and got this:

2025-04-22 13:09:32 WARNING  selfcord.gateway Gateway has stopped responding. Closing and restarting.
2025-04-22 13:09:32 INFO     selfcord.gateway Closing websocket with code 4000
2025-04-22 13:09:32 INFO     selfcord.gateway Finished closing websocket
2025-04-22 13:09:32 DEBUG    selfcord.gateway Gateway received CURLE_GOT_NOTHING, ignoring...

Looks like the exception is ignored as I tracked it down to this line:

discord.py-self/discord/gateway.py

Line 657 in 9262ca7

_log.debug('Gateway received CURLE_GOT_NOTHING, ignoring...')

[acetheking987]

I have got the same error, however mine did raise an exception so i have a traceback, but it doesn't seem to recover on its own without restarting it.

2025-05-19 07:19:03,151 - WARNING - selfcord.gateway - Gateway has stopped responding. Closing and restarting.
2025-05-19 07:19:03,162 - INFO - selfcord.gateway - Closing websocket with code 4000
2025-05-19 07:19:03,170 - ERROR - selfcord.gateway - An error occurred while stopping the gateway. Ignoring.
Traceback (most recent call last):
  File "/home/alfie/Documents/programming-projects/python/glados/.venv/lib/python3.13/site-packages/selfcord/gateway.py", line 172, in run
    await self.ws.close(4000)
  File "/home/alfie/Documents/programming-projects/python/glados/.venv/lib/python3.13/site-packages/selfcord/gateway.py", line 864, in close
    await self.socket.close(code, reason)
  File "/home/alfie/Documents/programming-projects/python/glados/.venv/lib/python3.13/site-packages/curl_cffi/requests/websockets.py", line 730, in close
    await self.send(msg, CurlWsFlag.CLOSE)
  File "/home/alfie/Documents/programming-projects/python/glados/.venv/lib/python3.13/site-packages/curl_cffi/requests/websockets.py", line 675, in send
    return await self.loop.run_in_executor(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        None, self.curl.ws_send, payload, flags
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^
  File "/usr/lib/python3.13/concurrent/futures/thread.py", line 59, in run
    result = self.fn(*self.args, **self.kwargs)
  File "/home/alfie/Documents/programming-projects/python/glados/.venv/lib/python3.13/site-packages/curl_cffi/curl.py", line 463, in ws_send
    self._check_error(ret, "WS_SEND")
    ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^
  File "/home/alfie/Documents/programming-projects/python/glados/.venv/lib/python3.13/site-packages/curl_cffi/curl.py", line 154, in _check_error
    raise error
curl_cffi.curl.CurlError: Failed to WS_SEND, curl: (55) Send failure: Broken pipe. See https://curl.se/libcurl/c/libcurl-errors.html first for more details.

[mudkipdev]

curl_cffi.curl.CurlError: Failed to WS_SEND, curl: (55) Send failure: Broken pipe. See https://curl.se/libcurl/c/libcurl-errors.html first for more details.

Can confirm I have the same issue, same as #818.

[EchoQuill]

curl_cffi.curl.CurlError: Failed to WS_SEND, curl: (55) Send failure: Broken pipe. See https://curl.se/libcurl/c/libcurl-errors.html first for more details.

I get this as well, after I get this all of the on_message() becomes unresponsive. I have to force a restart for it to work again.

[qonyx1]

Same issue here.

[KhabarovaNina]

+1
get same issue

[dolfies]

From my local testing, the new pre-release version of curl_cffi resolves this issue. Please install it from the releases page and test. Make sure you install it in the correct venv!

Example install commands:

# Windows
$ pip install https://github.com/lexiforest/curl_cffi/releases/download/v0.11.4/curl_cffi-0.11.4-cp39-abi3-win_amd64.whl
# macOS
$ pip install https://github.com/lexiforest/curl_cffi/releases/download/v0.11.4/curl_cffi-0.11.4-cp39-abi3-macosx_11_0_arm64.whl
# Linux
$ pip install https://github.com/lexiforest/curl_cffi/releases/download/v0.11.4/curl_cffi-0.11.4-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl

[KhabarovaNina]

From my local testing, the new pre-release version of curl_cffi resolves this issue. Please install it from the releases page and test. Make sure you install it in the correct venv!

Example install commands:

Hi! Could you please tell what exactly needs to be changed? Is it enough to just install curl_cffi in my environment or do I need to apply it somewhere in the code as well?

[AnthonyZJiang]

From my local testing, the new pre-release version of curl_cffi resolves this issue. Please install it from the releases page and test. Make sure you install it in the correct venv!
Example install commands:

Hi! Could you please tell what exactly needs to be changed? Is it enough to just install curl_cffi in my environment or do I need to apply it somewhere in the code as well?

I think we just need to install the new curl_cffi release

[AnthonyZJiang]

From my local testing, the new pre-release version of curl_cffi resolves this issue. Please install it from the releases page and test. Make sure you install it in the correct venv!

Example install commands:

Windows

$ pip install https://github.com/lexiforest/curl_cffi/releases/download/v0.11.4/curl_cffi-0.11.4-cp39-abi3-win_amd64.whl

macOS

$ pip install https://github.com/lexiforest/curl_cffi/releases/download/v0.11.4/curl_cffi-0.11.4-cp39-abi3-macosx_11_0_arm64.whl

Linux

$ pip install https://github.com/lexiforest/curl_cffi/releases/download/v0.11.4/curl_cffi-0.11.4-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl

Thank you. I will test this when I have time.

[KhabarovaNina]

From my local testing, the new pre-release version of curl_cffi resolves this issue. Please install it from the releases page and test. Make sure you install it in the correct venv!
Example install commands:

Hi! Could you please tell what exactly needs to be changed? Is it enough to just install curl_cffi in my environment or do I need to apply it somewhere in the code as well?

I think we just need to install the new curl_cffi release

Unfortunately, it didn't work for me :(

After 3 hours of the session I get logs like this:

INFO - Closing websocket with code 4000
INFO - Finished closing websocket

my Dockerfile:

FROM python:3.10

WORKDIR /build

RUN pip install --upgrade pip

COPY backend/dependencies/requirements.txt requirements.txt

RUN pip install -r requirements.txt
RUN pip install git+https://github.com/dolfies/discord.py-self.git
RUN pip install https://github.com/lexiforest/curl_cffi/releases/download/v0.11.4/curl_cffi-0.11.4-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl


WORKDIR /src

CMD ["python", "main.py"]

[AnthonyZJiang]

From my local testing, the new pre-release version of curl_cffi resolves this issue. Please install it from the releases page and test. Make sure you install it in the correct venv!
Example install commands:

Hi! Could you please tell what exactly needs to be changed? Is it enough to just install curl_cffi in my environment or do I need to apply it somewhere in the code as well?

I think we just need to install the new curl_cffi release

Unfortunately, it didn't work for me :(

After 3 hours of the session I get logs like this:

INFO - Closing websocket with code 4000 INFO - Finished closing websocket

my Dockerfile:

FROM python:3.10

WORKDIR /build

RUN pip install --upgrade pip

COPY backend/dependencies/requirements.txt requirements.txt

RUN pip install -r requirements.txt
RUN pip install git+https://github.com/dolfies/discord.py-self.git
RUN pip install https://github.com/lexiforest/curl_cffi/releases/download/v0.11.4/curl_cffi-0.11.4-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl


WORKDIR /src

CMD ["python", "main.py"]

Same, it didn't work for me either.

[outmaneuver]

this is a bump cause it's still occurring and it also does the same for the voice websocket

(also you can't really test this when it happens at random times)
oh yeah also it's happening on the main development branch

[mikasa]

Hi @outmaneuver

This is solved in #871