如何在Unix套接字上绑定Daphne？

Question

到目前为止，我使用过Gunicorn，并且使用了这个systemd配置文件：

[Unit]
Description=gunicorn daemon
After=network.target


[Service]
EnvironmentFile=/var/www/user/.envvars
User=user
Group=www-data
WorkingDirectory=/var/www/user
ExecStart=/var/www/venv/bin/gunicorn --workers 3 --bind unix:/var/www/user/user.sock config.wsgi:application --env DJANGO_SETTINGS_MODULE='config.settings.production'

[Install]
WantedBy=multi-user.target

并且Nginx代理请求到这个Unix socket。

当我尝试使用Daphne时，发现我无法指定--env DJANGO_SETTINGS_MODULE='config.settings.production'

但是我在asgi.py中指定了这个变量，一切正常。

当我运行gunicorn service时，套接字立即被创建，并处理来自Nginx的请求。

当我尝试使用Daphne进行相同的操作时，会出现错误：

Jun 26 11:48:32 p435061.kvmvps daphne[8447]: ValueError: invalid literal for int() with base 8: '/var/www/user/user.sock'

但是我不知道如何在尝试运行Daphne和创建我的systemd文件时使用unix.sock

更新

我更改了设置文件，获得了一些奇怪的东西，但它可以工作：

[Unit]
Description=daphne daemon
After=network.target


[Service]
EnvironmentFile=/var/www/gglobal/.envvars
User=gglobal
Group=www-data
WorkingDirectory=/var/www/gglobal
ExecStart=/var/www/venv/bin/daphne -b unix:/var/www/gglobal/gglobal.sock  config.asgi:channel_layer -v2

[Install]
WantedBy=multi-user.target

systemctl status daphne 返回：

Jun 26 17:53:46 p435061.kvmvps daphne[14183]: 2017-06-26 17:53:46,722 DEBUG    get 'RDS_HOSTNAME' casted as 'None' with default '<NoValue>'
Jun 26 17:53:46 p435061.kvmvps daphne[14183]: 2017-06-26 17:53:46,722 DEBUG    get 'RDS_PORT' casted as 'None' with default '<NoValue>'
Jun 26 17:53:46 p435061.kvmvps daphne[14183]: 2017-06-26 17:53:46,722 DEBUG    get 'REDIS_ENDPOINT_ADDRESS' casted as 'None' with default '<NoValue>'
Jun 26 17:53:46 p435061.kvmvps daphne[14183]: 2017-06-26 17:53:46,722 DEBUG    get 'REDIS_PORT' casted as 'None' with default '<NoValue>'
Jun 26 17:53:46 p435061.kvmvps daphne[14183]: 2017-06-26 17:53:46,723 DEBUG    get 'DJANGO_SENTRY_DSN' casted as 'None' with default 'https://56bb387ee6ef4e7b
Jun 26 17:53:46 p435061.kvmvps daphne[14183]: 2017-06-26 17:53:46,723 DEBUG    get 'DJANGO_SENTRY_CLIENT' casted as 'None' with default 'raven.contrib.django.
Jun 26 17:53:46 p435061.kvmvps daphne[14183]: 2017-06-26 17:53:46,723 DEBUG    get 'DJANGO_SENTRY_LOG_LEVEL' casted as '<class 'int'>' with default '20'
Jun 26 17:53:46 p435061.kvmvps daphne[14183]: 2017-06-26 17:53:46,724 DEBUG    get 'DJANGO_SENTRY_LOG_LEVEL' casted as '<class 'int'>' with default '20'
Jun 26 17:53:46 p435061.kvmvps daphne[14183]: 2017-06-26 17:53:46,724 DEBUG    get 'DJANGO_ADMIN_URL' casted as 'None' with default '<NoValue>'
Jun 26 17:53:48 p435061.kvmvps daphne[14183]: DEBUG 2017-06-26 17:53:48,186 base 14183 140286354425600 Configuring Raven for host: <raven.conf.remote.RemoteCo

但是nginx抱怨缺少 *.sock 文件:

2017/06/26 17:47:51 [crit] 12682#12682: *59 connect() to unix:/var/www/project/project.sock failed (2: No such file or directory) while connecting to upstream, client: 5.152.1.201, server: mydomain.com, request: "GET / HTTP/2.0", upstream: "http://unix:/var/www/project/project.sock:/", host: "mydomain.com"

我花了一整天的时间在这个问题上，但什么都没有发生。

- Narnik Gamarnik

---

看起来你在使用Daphne时参数有误。应该是 -u <path to socke> 而不是 -b <path to sock>。 - Zada Zorg

@ZadaZorg 应该同时指定 -b 和 -u 吗？ - Narnik Gamarnik

Answer 1

答案非常简单：

问题在于我指定了 -b unix:，而 unix: 前缀干扰了工作。当然，正确的标志应该是 -u。

[Unit]
Description=daphne daemon
After=network.target


[Service]
EnvironmentFile=/var/www/gglobal/.envvars
User=gglobal
Group=www-data
WorkingDirectory=/var/www/gglobal
ExecStart=/var/www/venv/bin/daphne -u /var/www/gglobal/gglobal.sock  config.asgi:channel_layer -v2

[Install]
WantedBy=multi-user.target

- Narnik Gamarnik