flask 程序使用 websocket，部署使用 gunicorn， gunicorn 导致 websocket 运行有问题

V2EX = way to explore

V2EX 是一个关于分享和探索的地方

现在注册

已注册用户请登录

推荐学习书目

Learn Python the Hard Way

Python Sites

PyPI - Python Package Index

http://diveintopython.org/toc/index.html

Pocoo

值得关注的项目

PyPy

Celery

Jinja2

Read the Docs

gevent

pyenv

virtualenv

Sentry

Shovel

Pyflakes

pytest

Python 编程

pep8 Checker

Styles

PEP 8

Google Python Style Guide

Code Style from The Hitchhiker's Guide

这是一个创建于 2558 天前的主题，其中的信息可能已经有所发展或是发生改变。

#1 本地开发 flask 应用程序，里面使用了 websocket，异步模式使用得是 gevent，并在程序刚开始打了 gevent 猴子补丁，s 端主要监听 connect disconnect 两个事件用来做一些前端页面的初始化和离开动作。启动代码：

socketio.run(app=app, host='0.0.0.0', port=5000, debug=True)

#2 正常在本地 debug 或 run 都没问题，部署到测试环境遇到问题，测试环境采用得是 Nginx+supervisor+gunicorn+flask， nginx 配置：

 server { listen 8000; listen [::]:8000; server_name x.x.x.x; root /root/python/asset; location / { proxy_pass http://127.0.0.1:5000; proxy_redirect off; proxy_set_header Host $host; proxy_set_header X-Real-IP $remote_addr; proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; access_log /var/log/asset/access.log; error_log /var/log/asset/error.log; proxy_set_header Upgrade $http_upgrade; proxy_set_header Connection "Upgrade"; } }

supervisor 配置：

[program:xx] directory = /root/python/xx command = /root/.local/xxx/bin/gunicorn -b 127.0.0.1:5000 -k geventwebsocket.gunicorn.workers.GeventWebSocketWorker -w 1 module:app user = root autostart = true autorestart = true stopasgroup = true killasgroup = true startsecs = 5 startretries = 3 redirect_stderr = true stdout_logfile_maxbytes = 20MB stdout_logfile_backups = 20 stdout_logfile = /var/log/xx/stdout.log stderr_logfile = /var/log/xx/stderr.log

其中启动 gunicorn 的命令参考自
https://stackoverflow.com/questions/38624447/websockets-proxied-by-nginx-to-gunicorn-over-https-giving-400-bad-request
https://segmentfault.com/q/1010000007495163
https://flask-socketio.readthedocs.io/en/latest/#gunicorn-web-server
gunicorn 也在 pipenv 环境下安装

#3 不使用 gunicorn 的情况下没任何问题，如图：使用了 gunicorn 的情况，
首先会导致 402 错误，
402 详情，
其次，会一直 pending，一直在发请求，
最后是 402 详情在控制台的信息，
请各位帮帮忙，尝试了好长时间没办法解决，感谢各位。

30 条回复 2018-10-11 17:56:05 +08:00

greyli

&nbp;2018-10-11 12:18:59 +08:00

按照 Flask-SocketIO 的文档，使用 Gunicorn 的时候要使用 eventlet 或 gevent worker，所以启动命令应该是：
```
gunicorn --worker-class eventlet -w 1 module:app
```
或：
```
gunicorn -k gevent -w 1 module:app
```

greyli

2018-10-11 12:19:31 +08:00

原来回复不支持 Mardown ……

hanssx

2018-10-11 12:22:16 +08:00

对了，supervisor 有错误日志，一直在尝试连接，如下图：
![image]( http://pgf1p2i0h.bkt.clouddn.com/supervisor%E9%94%99%E8%AF%AF%E6%97%A5%E5%BF%9755.png)

hanssx

2018-10-11 12:23:38 +08:00

@greyli 辉哥，我改了启动命令了，参考的文档里面
gunicorn -k geventwebsocket.gunicorn.workers.GeventWebSocketWorker -w 1 module:app

hanssx

2018-10-11 12:25:19 +08:00

这是 flask-socketio 官网给的例子，说了一大堆英文，反正就是要用 gunicorn 就要那么写，如果搭配 gevent 的话，
https://flask-socketio.readthedocs.io/en/latest/#gunicorn-web-server
最后命令是这么写的，
gunicorn -b 127.0.0.1:5000 -k geventwebsocket.gunicorn.workers.GeventWebSocketWorker -w 1 module:app

itertools

2018-10-11 13:41:05 +08:00

gunicorn 版本问题

itertools

2018-10-11 13:42:19 +08:00

使用 gunicorn==19.9.0 这个版本试试。

FiveDDD

2018-10-11 13:47:20 +08:00

启动命令：gunicorn -b :5000 -w 1 module:app -k eventlet

版本是 19.9.0 没问题

hanssx

2018-10-11 13:48:22 +08:00

@itertools 感谢关注，但是测试机上面的 gunicorn 版本就是你说的这个，
(asset-ar0OxIPP) [root@VM asset]# pipenv graph | grep gunicorn
gunicorn==19.9.0

hanssx

2018-10-11 13:49:47 +08:00

@FiveDDD 感谢回复，我用的 gevent，命令和你这个不一样

itertools

2018-10-11 13:52:32 +08:00

@hanssx 看日记，可能是之前启动的进程还在导致。

hanssx

2018-10-11 13:57:57 +08:00

@itertools
supervisorctl stop all
find / -name supervisor.sock 后 unlink 掉
再重新启动 supervisord -c /etc/supervisord.conf
还是不行。。。
supervisorctl status，显示
asset STARTING
没 RUNNING 起来。。。

hanssx

2018-10-11 13:59:40 +08:00

看日志，还是那些东西，我去 gg 一下日志相关的东西，
[2018-10-11 13:58:33 +0800] [19707] [INFO] Starting gunicorn 19.9.0
[2018-10-11 13:58:33 +0800] [19707] [ERROR] Connection in use: ('127.0.0.1', 5000)
[2018-10-11 13:58:33 +0800] [19707] [ERROR] Retrying in 1 second.
[2018-10-11 13:58:34 +0800] [19702] [ERROR] Connection in use: ('127.0.0.1', 5000)
[2018-10-11 13:58:34 +0800] [19702] [ERROR] Retrying in 1 second.
[2018-10-11 13:58:34 +0800] [19707] [ERROR] Connection in use: ('127.0.0.1', 5000)
[2018-10-11 13:58:34 +0800] [19707] [ERROR] Retrying in 1 second.
[2018-10-11 13:58:35 +0800] [19702] [ERROR] Can't connect to ('127.0.0.1', 5000)
[2018-10-11 13:58:35 +0800] [19707] [ERROR] Connection in use: ('127.0.0.1', 5000)
[2018-10-11 13:58:35 +0800] [19707] [ERROR] Retrying in 1 second.
[2018-10-11 13:58:36 +0800] [19707] [ERROR] Connection in use: ('127.0.0.1', 5000)
[2018-10-11 13:58:36 +0800] [19707] [ERROR] Retrying in 1 second.
[2018-10-11 13:58:37 +0800] [19716] [INFO] Starting gunicorn 19.9.0
[2018-10-11 13:58:37 +0800] [19716] [ERROR] Connection in use: ('127.0.0.1', 5000)
[2018-10-11 13:58:37 +0800] [19716] [ERROR] Retrying in 1 second.

hanssx

2018-10-11 14:06:19 +08:00

@itertools
我根据 https://stackoverflow.com/questions/16756624/gunicorn-connection-in-use-0-0-0-0-5000/40894414
[root@VM asset]# sudo fuser -k 5000/tcp
5000/tcp: 18762 18765
[root@VM asset]# supervisorctl status
asset RUNNING pid 20781, uptime 0:00:32
[root@VM asset]# netstat -antlp | grep 5000
tcp 0 0 127.0.0.1:5000 0.0.0.0:* LISTEN 20781/python3.7m
tcp 0 0 127.0.0.1:49678 127.0.0.1:5000 TIME_WAIT -
tcp 0 0 127.0.0.1:5000 127.0.0.1:49708 ESTABLISHED 20787/python3.7m
tcp 0 0 127.0.0.1:49650 127.0.0.1:5000 TIME_WAIT -
tcp 0 0 127.0.0.1:5000 127.0.0.1:49706 TIME_WAIT -
tcp 0 0 127.0.0.1:49708 127.0.0.1:5000 ESTABLISHED 25344/nginx: worker
tcp 0 0 127.0.0.1:5000 127.0.0.1:49680 TIME_WAIT -

so1n

2018-10-11 14:15:33 +08:00

有没有内存不够的可能呢？

hanssx

2018-10-11 14:31:45 +08:00

@itertools
@so1n
@greyli 大概率是我 gunicorn 进程的问题，现在正在学习咋清除干净。。。

hanssx

2018-10-11 14:52:21 +08:00

@itertools 打扰，我已经重新清理了之后使用 sudo kill -9，确认网站访问不了，然后重新 supervisord -c /etc/supervisord.conf
问题依然还在，日志中是这样，请问这是 work 吗？这和我在 supervior conf 中配置的不一样啊，我用了参数-w 1 呀。
[2018-10-11 14:28:36 +0800] [25625] [ERROR] Connection in use: ('127.0.0.1', 5000)
[2018-10-11 14:28:36 +0800] [25625] [ERROR] Retrying in 1 second.
[2018-10-11 14:28:37 +0800] [25625] [INFO] Listening at: http://127.0.0.1:5000 (25625)
[2018-10-11 14:28:37 +0800] [25625] [INFO] Using worker: geventwebsocket.gunicorn.workers.GeventWebSocketWorker
[2018-10-11 14:28:37 +0800] [25635] [INFO] Booting worker with pid: 25635
[2018-10-11 14:32:15 +0800] [26148] [INFO] Starting gunicorn 19.9.0
[2018-10-11 14:32:15 +0800] [26148] [INFO] Listening at: http://127.0.0.1:5000 (26148)
[2018-10-11 14:32:15 +0800] [26148] [INFO] Using worker: geventwebsocket.gunicorn.workers.GeventWebSocketWorker
[2018-10-11 14:32:15 +0800] [26151] [INFO] Booting worker with pid: 26151
[2018-10-11 14:33:32 +0800] [26345] [INFO] Starting gunicorn 19.9.0
[2018-10-11 14:33:32 +0800] [26345] [INFO] Listening at: http://127.0.0.1:5000 (26345)
[2018-10-11 14:33:32 +0800] [26345] [INFO] Using worker: geventwebsocket.gunicorn.workers.GeventWebSocketWorker
[2018-10-11 14:33:32 +0800] [26348] [INFO] Booting worker with pid: 26348
[2018-10-11 14:37:35 +0800] [26931] [INFO] Starting gunicorn 19.9.0
[2018-10-11 14:37:35 +0800] [26931] [INFO] Listening at: http://127.0.0.1:5000 (26931)
[2018-10-11 14:37:35 +0800] [26931] [INFO] Using worker: geventwebsocket.gunicorn.workers.GeventWebSocketWorker
[2018-10-11 14:37:35 +0800] [26934] [INFO] Booting worker with pid: 26934
[2018-10-11 14:39:18 +0800] [27180] [INFO] Starting gunicorn 19.9.0
[2018-10-11 14:39:18 +0800] [27180] [INFO] Listening at: http://127.0.0.1:5000 (27180)
[2018-10-11 14:39:18 +0800] [27180] [INFO] Using worker: geventwebsocket.gunicorn.workers.GeventWebSocketWorker
[2018-10-11 14:39:18 +0800] [27183] [INFO] Booting worker with pid: 27183

hanssx

2018-10-11 15:29:56 +08:00

开启了 flask-socketio 的日志，现在重新启动，ps -ef | grep guni 出现 2 行，其中 1 行是 work，work 数应该没错，日志现在报错，继续搜索解决方案，
[2018-10-11 15:19:04 +0800] [697] [INFO] Starting gunicorn 19.9.0
[2018-10-11 15:19:04 +0800] [697] [INFO] Listening at: http://127.0.0.1:5000 (697)
[2018-10-11 15:19:04 +0800] [697] [INFO] Using worker: geventwebsocket.gunicorn.workers.GeventWebSocketWorker
[2018-10-11 15:19:04 +0800] [700] [INFO] Booting worker with pid: 700
Server initialized for gevent.
bae5c5f54be848fcabd9e5cfe8ce59fa: Sending packet OPEN data {'sid': 'bae5c5f54be848fcabd9e5cfe8ce59fa', 'upgrades': ['websocket'], 'pingTimeout': 60000, 'pingInterval': 25000}
bae5c5f54be848fcabd9e5cfe8ce59fa: Sending packet MESSAGE data 0
bae5c5f54be848fcabd9e5cfe8ce59fa: Received packet MESSAGE data 0/process_upload_csv
bae5c5f54be848fcabd9e5cfe8ce59fa: Sending packet MESSAGE data 0/process_upload_csv
Traceback (most recent call last):
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/pywsgi.py", line 975, in handle_one_response
self.run_application()
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/geventwebsocket/handler.py", line 82, in run_application
self.process_result()
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/pywsgi.py", line 909, in process_result
self.write(data)
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/pywsgi.py", line 756, in write
self._write_with_headers(data)
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/pywsgi.py", line 777, in _write_with_headers
self._write(data)
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/pywsgi.py", line 741, in _write
self._sendall(data)
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/pywsgi.py", line 703, in _sendall
self.socket.sendall(data)
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/_socket3.py", line 457, in sendall
data_memory = _get_memory(data)
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/_socket3.py", line 45, in _get_memory
mv = memoryview(data)
TypeError: memoryview: a bytes-like object is required, not 'str'
2018-10-11T07:20:48Z {'REMOTE_ADDR': '127.0.0.1', 'REMOTE_PORT': '54310', 'HTTP_HOST': '10.26.15.222', (hidden keys: 31)} failed with TypeError

bae5c5f54be848fcabd9e5cfe8ce59fa: Client is gone, closing socket
bae5c5f54be848fcabd9e5cfe8ce59fa: Client is gone, closing socket
ca996671e95d4e54834a745c9129caa8: Sending packet OPEN data {'sid': 'ca996671e95d4e54834a745c9129caa8', 'upgrades': ['websocket'], 'pingTimeout': 60000, 'pingInterval': 25000}
ca996671e95d4e54834a745c9129caa8: Sending packet MESSAGE data 0
ca996671e95d4e54834a745c9129caa8: Received packet MESSAGE data 0/process_upload_csv
ca996671e95d4e54834a745c9129caa8: Sending packet MESSAGE data 0/process_upload_csv
Traceback (most recent call last):
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/pywsgi.py", line 975, in handle_one_response
self.run_application()
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/geventwebsocket/handler.py", line 82, in run_application
self.process_result()
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/pywsgi.py", line 909, in process_result
self.write(data)
File "/root/.local/share/virtualenvs/asset-ar0OxIPP/lib/python3.7/site-packages/gevent/pywsgi.py", line 756, in write

hanssx

2018-10-11 15:55:43 +08:00

![image]( http://pgf1p2i0h.bkt.clouddn.com/%E6%9C%80%E5%90%8E%E8%A6%81%E8%A7%A3%E5%86%B3%E7%9A%84%E9%97%AE%E9%A2%98.png)

itertools

2018-10-11 16:21:27 +08:00

@hanssx https://github.com/miguelgrinberg/Flask-SocketIO/issues/272 gevent-websocket 貌似不支持 python3

hanssx

2018-10-11 16:44:18 +08:00

清理了 gunicorn 进程，抛弃了 websocket-gevent，使用了 gevent，gunicorn -b 127.0.0.1:5000 -k gevent -w 1 module:app，不再报 402 错误了，但是程序还是会一直 pending，一直发请求。。。

hanssx

2018-10-11 16:47:12 +08:00

@itertools 是的，不支持 py3，真是我的锅，第一次部署花了很多时间，使用了 gevent 之后没有 402 错误了，但是会一直 pending，gunicorn 日志显示好像是心跳包，正在找办法关了它。。。
[2018-10-11 16:40:27 +0800] [12128] [INFO] Shutting down: Master
[2018-10-11 16:41:47 +0800] [13014] [INFO] Starting gunicorn 19.9.0
[2018-10-11 16:41:47 +0800] [13014] [INFO] Listening at: http://127.0.0.1:5000 (13014)
[2018-10-11 16:41:47 +0800] [13014] [INFO] Using worker: gevent
[2018-10-11 16:41:47 +0800] [13017] [INFO] Booting worker with pid: 13017
Server initialized for gevent.
94207eaf37114c118b05480ba12679ef: Sending packet OPEN data {'sid': '94207eaf37114c118b05480ba12679ef', 'upgrades': [], 'pingTimeout': 60000, 'pingInterval': 25000}
94207eaf37114c118b05480ba12679ef: Sending packet MESSAGE data 0
94207eaf37114c118b05480ba12679ef: Received packet MESSAGE data 0/process_upload_csv
94207eaf37114c118b05480ba12679ef: Sending packet MESSAGE data 0/process_upload_csv
94207eaf37114c118b05480ba12679ef: Received packet PING data None
94207eaf37114c118b05480ba12679ef: Sending packet PONG data None
94207eaf37114c118b05480ba12679ef: Received packet PING data None
94207eaf37114c118b05480ba12679ef: Sending packet PONG data None
94207eaf37114c118b05480ba12679ef: Received packet PING data None

hanssx

2018-10-11 16:50:01 +08:00

@itertools 前辈，我这一直发请求，这个有啥好办法解决吗？
![image]( http://pgf1p2i0h.bkt.clouddn.com/%E4%B8%80%E7%9B%B4pending%EF%BC%8C%E4%B8%80%E7%9B%B4%E5%8F%91%E8%AF%B7%E6%B1%82.png)

hanssx

2018-10-11 16:53:37 +08:00

@itertools 我 flask 应用程序的启动代码：socketio.run(app=app, host='0.0.0.0', port=5000, debug=True)
其中 socketio 初始化时，指定了 async_mode='gevent'，是这个原因吗？然后 gunicorn 也使用了 gevent。

hanssx

2018-10-11 17:26:37 +08:00

@itertools 我晕，我本地不加 gunicorn，socketio.run(app=app, host='0.0.0.0', port=5000, debug=True)，指定 async_mode='gevent'，也会一直 pending，一直请求。。。

itertools

2018-10-11 17:43:29 +08:00

@hanssx 目前最解决问题的办法是使用 gunicorn + eventlet 组合跑，卸载 gevent 安装 eventlet 然后把 async_mode="gevent"改成 async_mode="eventlet " 启动命令为 gunicorn --worker-class eventlet -w 1 module:app。
如果一定要用 gevent 请尝试使用 uWSGI + gevent

hanssx

2018-10-11 17:46:27 +08:00

@itertools 确实，老哥，能加您个 QQ 吗？想感谢您一下，gevent+flask-socketio 就会一直发请求并且 pending，eventlet 则不会，怎么会这样我也不明白，我的扣扣 9614 六 2392，期待老哥，哈哈。

hanssx

2018-10-11 17:48:46 +08:00

已解决，感谢各位的帮助，特别是 itertools。

daya

2018-10-11 17:53:42 +08:00 via iPhone

怎么解决的？

hanssx

2018-10-11 17:56:05 +08:00

@daya 首先清除 gunicorn 多余进程，使用 netstat -antlp | grep port 或者 ps -ef | grep guni，然后 sudo kill -9 pid
然后再把 gevent-websocket 换成 gevetn，因为前者不支持 py3
最后把 flask-socketio 的 async_mode 换成 eventlet 而不是 gevent，gevent 会一直发送类似心跳包的东西，不知道有没有选项可以清除。