unblock the wait loop under python 3.5

[benoitc]

in python 3.5 the select is blocking when waiting for it which prevent quick
exit on SIGTERM.

The problem is described:
https://www.python.org/dev/peps/pep-0475/#backward-compatibility

This change fix it by listening for signal event on the worker pipe. Once an
event is triggered it will forcefully wake up the select and return.

fix #1256

[aolefira]

In python 3.6 this bug appeared again

[benoitc]

@aolefira how do you produce it?

[aolefira]

I run script ./stop.sh with next content:
kill cat gunicorn.pid
and waiting a lot of seconds until gunicorn processes will be killed.
Software:
gunicorn 19.7.1
python 3.6.0
ubuntu 16.04.1 LTS
In settings worker_class = gthread
No '--timeout' argument used

[benoitc]

how can I rrepoduce it? Do you have any log?

…

On Fri, Sep 29, 2017 at 8:40 AM, aolefira ***@***.***> wrote: I run script ./stop.sh with next content: kill cat gunicorn.pid and waiting a lot of seconds until gunicorn processes will be killed. Software: gunicorn 19.7.1 python 3.6.0 ubuntu 16.04.1 LTS — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#1257 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAA4ol3Xqf2dSoZXo31EYmfTsDIcOF8xks5snJDEgaJpZM4Iaz4k> .

[aolefira]

There are no records in logs while shutting down.

Command for start:
gunicorn --paste pyramid.ini --daemon

[server:main] part of pyramid.ini:
[server:main]
use = egg:gunicorn#main
host = 0.0.0.0
port = 6543
threads = 60
worker_class = gthread
errorlog = logs/error.log
loglevel = warning
capture_output = True
pidfile = gunicorn.pid
statsd_host = localhost:8125

Any pyramid project should be used.

To reproduce bug just enter:
kill ‵cat gunicorn.pid‵

[tilgovi]

I cannot reproduce this.

MacOS X 10.13
Python 3.6.2
Gunicorn 19.7.1
Pyramid 1.9.1

Do you have a long connection open when you do this? What happens if you use kill -9?

[aolefira]

There are no long connections.
If kill -9 used, first process terminate, but second process remains hanging.

[aolefira]

Yes, on my macbook with MacOS X 10.11.6 this issue not reproduced too.
This issue reproduced on server with Ubuntu 16.04.1 LTS.

[aolefira]

Do you have a long connection open when you do this?

This bug reproduce (on Ubuntu 16.04.1) even if there are no connections.

[aolefira]

Problem is that pyramid runs thread at startup. In mac os x that thread die immediately on app shutdown but in ubuntu it's waiting for some time

[aolefira]

setDaemon(True) for threads resolve this problem

[tilgovi]

What thread does Pyramid run at startup? I'm not aware of it doing that.

[tilgovi]

As far as I know, the only thread Pyramid starts are started by pserve command, but you said you are using the gunicorn --paste command.

[aolefira]

What thread does Pyramid run at startup? I'm not aware of it doing that.

It's project specific threads. Two custom threads. First - scheduler and second - mailer.

[aolefira]

As far as I know, the only thread Pyramid starts are started by pserve command, but you said you are using the gunicorn --paste command.

pserve is good for development. but for production its better to use something more advanced, like gunicorn

https://docs.pylonsproject.org/projects/pyramid/en/latest/narr/project.html#using-an-alternate-wsgi-server

https://docs.pylonsproject.org/projects/pyramid-cookbook/en/latest/deployment/gunicorn.html

[Dringho]

Also experiencing this bug on Ubuntu 16.04 / Python 3.5.2 / Gunicorn 19.7.1