Deploying telegraphy

[Brandl]

I'm not a Twisted expert and unfortunately I haven't found any explanation in the docs on which is the best setup to use telegraphy in production?

Your recommendation would really help me out.

Kind regards

[D3f0]

Sorry for the late answer.
Telegraphy, in its current state, must run in a different process than gunicorn/uWSGI, or whatever wsgi server your're using. The process is started as a django management process.
You must keep this process alive with some sort of process monitor (ie: superivosrd). Although is written on top of twisted, it does not use twisted's twistd.

Fornt end configuration could be done with something like http://nginx.org/en/docs/http/websocket.html.

[Brandl]

Thanks. We decided to run the telegraphy management command with supervisord, so this was a lucky guess. :)

But I've got another question: It's best practice to run websockets on port 80 to avoid getting blocked by a firewall. Which makes this kinda tricky, because (I hope I don't get this wrong) the telegraphy websocket locations have no particular prefix (for the nginx location)?
Another possibility to a prefixe would be a ws.example.com subdomain.

Which would be the simplest setup to run telegraphy behind nginx on port 80?

---

If anyone comes across the same question, this is our supervisord config file:

$ cat /etc/supervisor/conf.d/telegraphy.conf 

[program:run_telegraphy] 
command=/webapps/[app]/bin/python /webapps/[app]/Django/manage.py run_telegraph 
directory=/webapps/[app]/Django
environment=PYTHONPATH='/webapps/[app]/Django' 
user=[restricted_user]
numprocs=1 
stdout_logfile=/webapps/[app]/logs/telegraphy.log 
stderr_logfile=/webapps/[app]/logs/telegraphy.log 
autostart=true 
autorestart=true 
startsecs=10 
stopwaitsecs=30

[D3f0]

Brandl indeed, one of the this I have to change is to add a Prefix for telegraphy's WS entry point URLs.