daphne: KeyError from create_application

Since upgrading to Channels 2, I get a number of these errors. I suspect this happens when restarting the deployed instance:

Traceback (most recent call last):
  File "/usr/local/lib/python3.6/site-packages/daphne/http_protocol.py", line 160, in process
    "server": self.server_addr,
  File "/usr/local/lib/python3.6/site-packages/twisted/internet/defer.py", line 1386, in _inlineCallbacks
    result = g.send(result)
  File "/usr/local/lib/python3.6/site-packages/daphne/server.py", line 191, in create_application
    ), loop=asyncio.get_event_loop())
KeyError: <WebRequest at 0x7ff738677710 method=GET uri=/api/internal/projects/21/managed_objects/ clientproto=HTTP/1.1>

Unfortunately, it doesn’t help that there is a lot happening on this line, and I am not sure what part causes this error.

About this issue

Original URL
State: closed
Created 6 years ago
Comments: 26 (21 by maintainers)

Commits related to this issue

Don't try and read requests that are closed already (#205) — committed to django/daphne by deleted user 6 years ago
Don't crash if connection closed before application started (#213) Fixes #205. — committed to django/daphne by brianmay 6 years ago

Most upvoted comments

Reference to commit that (hopefully) fixed this: ece52b8e7958b41c86fd89830bc7814368e42bde

brianmay on Jun 12, 2018

Traceback (most recent call last):
  File "/home/tarek/Documents/python/lms-mt/env/lib/python3.5/site-packages/daphne/http_protocol.py", line 166, in process
    "body": self.content.read(),
ValueError: I/O operation on closed file.

Twisted closes “content” when the connection is lost, i.e. request is “finished” internally in Twisted. I fixed it by checking if self.content.closed: before initiating the request handling where content is read.

Similar error only happened for me when using nginx as reverse proxy and I was able to reproduce it consistently when I found a working way to trigger it.

JohnDoee on May 31, 2018