Gunicorn not responding -- Ignored premature client disconnection

Question

This is a cross-post with the GitHub issue

Given:

• gunicorn 19.7.1
• Python 2.7.9
• Debian 8
• Kernel 3.16.39-1+deb8u2
• On Google Compute Engine

And wsgi.py of:

def application(env, start_response):
    start_response('200 OK', [('Content-Type', 'text/html')])
    return ["Hello!"]

Running $ gunicorn wsgi --log-level=debug

Yields:

[2017-04-10 18:01:25 +0000] [31956] [INFO] Starting gunicorn 19.7.1

[2017-04-10 18:01:25 +0000] [31956] [DEBUG] Arbiter booted

[2017-04-10 18:01:25 +0000] [31956] [INFO] Listening at: http://127.0.0.1:8000 (31956)

[2017-04-10 18:01:25 +0000] [31956] [INFO] Using worker: sync

[2017-04-10 18:01:25 +0000] [31961] [INFO] Booting worker with pid: 31961

[2017-04-10 18:01:25 +0000] [31956] [DEBUG] 1 workers

Then using netcat to test it with e.g. echo -en "GET /ping HTTP/1.0\n\n\n" | nc localhost 8000 times out with:

[2017-04-10 18:07:36 +0000] [31977] [CRITICAL] WORKER TIMEOUT (pid:31982)

[2017-04-10 18:07:36 +0000] [31982] [INFO] Worker exiting (pid: 31982)

[2017-04-10 18:07:36 +0000] [31985] [INFO] Booting worker with pid: 31985

Alternatively, sending SIGINT (ctrl-c) to nc instance while waiting yields:

[2017-04-10 18:01:28 +0000] [31961] [DEBUG] Ignored premature client disconnection. No more data after: 'GET /ping HTTP/1.0\n\n\n'

Using --spew didn't illuminate anything obvious.

Answer 1

Although I haven't changed anything since I posted this question yesterday, it's working today.

Your guess is as good as mine, but I'm closing this on the basis that I can't reproduce it.

If it recurs, I'll re-open.