Django Broken pipe in Debug mode

Question

I have django 1.3 on the remote server behind Nginx.

If I run django with apache + mod_wsgi, I can watch errors in apache log files. It's ok but I'd like to have in console.

If I run django own development server, I get errors with stacktrace in console only when DEBUG = False. In DEBUG mode console outputs

Exception happened during processing of request from (..., ...)
Traceback (most recent call last):
  File "/usr/local/python/lib/python2.7/SocketServer.py", line 284, in _handle_request_noblock
    self.process_request(request, client_address)
  File "/usr/local/python/lib/python2.7/SocketServer.py", line 310, in process_request
    self.finish_request(request, client_address)
  File "/usr/local/python/lib/python2.7/SocketServer.py", line 323, in finish_request
    self.RequestHandlerClass(request, client_address, self)
  File "/usr/local/python/lib/python2.7/site-packages/django/core/servers/basehttp.py", line 570, in __init__
    BaseHTTPRequestHandler.__init__(self, *args, **kwargs)
  File "/usr/local/python/lib/python2.7/SocketServer.py", line 641, in __init__
    self.finish()
  File "/usr/local/python/lib/python2.7/SocketServer.py", line 694, in finish
    self.wfile.flush()
  File "/usr/local/python/lib/python2.7/socket.py", line 301, in flush
    self._sock.sendall(view[write_offset:write_offset+buffer_size])
error: [Errno 32] Broken pipe

I want to figure out why? Why does django just output unnamed Exception? Why does it depend on DEBUG variable.

This errors occurs mostly outside views when I have no access to request object. So I can't catch it in middleware or using logging handler.

UPDATE. I noticed if I request to django server directly I never get Broken pipe. So may the issue occur while Nginx proxy django?

Liston

I haven't had this issue on a live server, however it occurs from time to time on the django dev server on my local computer.. I haven't been able to resolve this either – Inhaler

It could just be NGINX timing out before django sends the response. What's the "proxy_read_timeout" set to in your nginx.conf file? – Vivacity

Not sure if it adds any valuable info, but I've been working on a local project (for practise purposes) for about a month without seeing the "broken pipe" message. As soon as I generated a new SECRET_KEY and added it to settings.py, I started getting this message. – Dewitt

Answer 1

Nginx directive proxy_intercept_errors off; (disabled by default) is what I needed

Liston

Hi San4ez, as you say this appears to be disabled by default. How did explicitly disabling this fix your problem? (I might have a related issue...) – Ablution

I'm not quite understand your question. Imagine, nginx proxies request to django upstream and gets some error status (400 or higher). If proxy_intercept_errors is disabled, nginx responses error page as is from django. Otherwise nginx stops reading django response and send to user custom error page. Django server (or any other) throws error that client doesn't read socket - broken pipe – Liston

Understood; and actually this worked for me. It is just odd because if proxy_intercept_errors is off by default, explicitly turning it off should have no effect. See wiki.nginx.org/HttpProxyModule#proxy_intercept_errors – Ablution

Yes, sure. But I had nginx config where this options was enabled by someone – Liston

Right. That makes sense then. Weirdly, it worked for me too, so I guess that key must be generated in some file somewhere in some builds. – Ablution

This does not answer the question. – Vey

Answer 2

This isn't really an issue with your site, more with the Django devserver: see this Django ticket. To put it bluntly, just ignore it as it is a known error, and won't be fixed.

In that ticket's comments a quite clear explanation is given:

According to many sources the 'Broken Pipe' is a normal browser quirk. For example, the browser reads from the socket and then decides that the image it's been reading apparently didn't change. The browser now this (forcefully) closes the connection because it does not need more data. The other end of this socket (the python runserver) now raises a socket exception telling the program that the client 'Broke the socket pipe'.

Ader

Answer 3

The nginx directive (checked answer) didn't work for me, but combining monkey patches from Igor Katson and Michael_Scharf did:

def patch_broken_pipe_error():
    """Monkey Patch BaseServer.handle_error to not write
    a stacktrace to stderr on broken pipe.
    https://mcmap.net/q/137436/-django-broken-pipe-in-debug-mode"""
    import sys
    from SocketServer import BaseServer
    from wsgiref import handlers

    handle_error = BaseServer.handle_error
    log_exception = handlers.BaseHandler.log_exception

    def is_broken_pipe_error():
        type, err, tb = sys.exc_info()
        return repr(err) == "error(32, 'Broken pipe')"

    def my_handle_error(self, request, client_address):
        if not is_broken_pipe_error():
            handle_error(self, request, client_address)

    def my_log_exception(self, exc_info):
        if not is_broken_pipe_error():
            log_exception(self, exc_info)

    BaseServer.handle_error = my_handle_error
    handlers.BaseHandler.log_exception = my_log_exception

patch_broken_pipe_error()

Alcus

You can also enable proxy_ignore_client_abort option in your nginx conf. In this case nginx will not break connection with backend if client (web-browser) does this. – Liston

This code needs to be put in settings.py - using Django 1.5. – Phil

Answer 4

Here is a way to prevent the to print the message to stderr. Just monkey patch the BaseServer.handle_error function. This is how I do it:

def patch_broken_pipe_error():
    """Monkey Patch BaseServer.handle_error to not write
    a stacktrace to stderr on broken pipe.
    https://stackoverflow.com/a/7913160"""
    import sys
    from SocketServer import BaseServer

    handle_error = BaseServer.handle_error

    def my_handle_error(self, request, client_address):
        type, err, tb = sys.exc_info()
        # there might be better ways to detect the specific erro
        if repr(err) == "error(32, 'Broken pipe')":
            # you may ignore it...
            logging.getLogger('mylog').warn(err)
        else:
            handle_error(self, request, client_address)

    BaseServer.handle_error = my_handle_error


patch_broken_pipe_error()

Vendee

Answer 5

In Django 3.2+ the "error" is an INFO reported by logging similar to:

INFO - Broken pipe from ('127.0.0.1', 59895)

@igor-katson has a good quick and dirty monkey patch. Updated for 3.2+ it can be:

# Suppress broken pipe errors
from django.core.servers.basehttp import WSGIServer
WSGIServer.handle_error = lambda *args, **kwargs: None

If you have a conditional check for being in development (runserver) mode in your settings.py, you could put this code there. It will squelch that annoying log entry.

Keep in mind, WSGIServer.handle_error also reports non-socket errors. If you had some other connectivity issue to debug this may get in the way.

Jokester

Answer 6

I was able to get rid of this by

proxy_buffering off;

This stops response buffering of proxied server. This leads to other issues of the back-end application being locked for long if the client is on a extremely slow connection.

To make it conditional for particular requests, use X-Accel-Buffering=no in the response header.

Finding

Answer 7

I came up with a quick and dirty monkey patch (i don't know if it supresses any useful errors), that gets rid of this annoying error when using "./manage.py runserver" or running LiveServerTestCase tests.

Just insert it anywhere in your code, where you need that:

# Monkeypatch python not to print "Broken Pipe" errors to stdout.
import SocketServer
from wsgiref import handlers
SocketServer.BaseServer.handle_error = lambda *args, **kwargs: None
handlers.BaseHandler.log_exception = lambda *args, **kwargs: None

Ariew

Answer 8

I fixed it. If you use links i.e, anchor tag, inside the page, you have to face the "Borken Pipe" problem. Just use inside the link tag href='#'. Don't leave the href attribute blank. It will avoid that type of error.

Microcyte

Answer 9

My BrokenPipe error disappears when I remove the "/" in the path in "urls.py"

urlpatterns = [
    path('', views.index)
]

Privacy

Answer 10

I turned

DEBUG = True 

and added

if os.getcwd() == '/app':
    DEBUG = False

now it is working

Haugh

This is a nice answer because it provides new alternative solution. In which django version you test this solution ? is it on Django 3.x?. I mean it's a really old question (django 1.3) – Regardant

I tested with django version 4.0.4 – Hema

Answer 11

I came across this issue as well while using tilelite. It's actually caused by a known, and now fixed, bug in python. You can resolve this issue by applying the following patch:

http://bugs.python.org/issue14574

Otherwise, you can download one of the more recent builds of python.

Kerek

Answer 12

Now currently it should be (in settings.py):

# Miscellaneous patch code

def patch_broken_pipe_error():
    """Monkey Patch BaseServer.handle_error to not write
    a stacktrace to stderr on broken pipe.
    https://mcmap.net/q/137436/-django-broken-pipe-in-debug-mode"""
    import sys
    from socketserver import BaseServer
    from wsgiref import handlers

    handle_error = BaseServer.handle_error
    log_exception = handlers.BaseHandler.log_exception

    def is_broken_pipe_error():
        type, err, tb = sys.exc_info()
        return repr(err) == "error(32, 'Broken pipe')"

    def my_handle_error(self, request, client_address):
        if not is_broken_pipe_error():
            handle_error(self, request, client_address)

    def my_log_exception(self, exc_info):
        if not is_broken_pipe_error():
            log_exception(self, exc_info)

    BaseServer = my_handle_error
    handlers.BaseHandler.log_exception = my_log_exception

patch_broken_pipe_error()

Note the lowercase socketserver import line.

Invagination

Should be BaseServer.handle_error = my_handle_error – Narcoma

django 3.2, this hook does not works – Ammoniate