Closed tarsil closed 3 years ago
Bogdanp
commented
3 years ago Looks like something is killing your processes. I'm not familiar with supervisord so I'm not sure what could be causing that. Does your system have enough memory? It could be that the process is being killed by the OOM killer. What happens if you run the app without supervisord?
tarsil
commented
3 years ago But I need supervisord to make sure ai can control the auto restarts and so on, this is for production
tarsil
commented
3 years ago I have a more powerful app running that uses the same memory, so might not be that
Bogdanp
commented
3 years ago But I need supervisord to make sure ai can control the auto restarts and so on, this is for production
Right, I understand that, but the reason I asked you to run it without was to see if the issue still occurs. I'd like to determine if it's supervisord that's killing it or if it's something else.
tarsil
commented
3 years ago Apologies, I didn't mention that. I've tried and still happens the same
cchacholiades
commented
3 years ago I am as well trying to figure out on how to use rundramatiq in production. @tarsil I thought of using supervisord to make sure it always runs, but in my case I think I need to use systemd. What is the recommended way to deploy django dramatiq in a production environment @Bogdanp, how can it be daemonized?
jcass77
commented
3 years ago We use supervisord to monitor dramatiq - no issues.
Configuring with systemd should also not present any special challenges.
vangale
commented
3 years ago We use systemd. Our service file looks something like:
Description=Dramatiq Task Queue Workers
After=syslog.target
[Service]
Type=simple
Environment=DJANGO_SETTINGS_MODULE=app.settings
WorkingDirectory=/srv/app
User=someone
Group=someone
ExecStart=/srv/app/bin/python3 manage.py rundq --log-file /srv/app/log/dramatiq.log --processes 2 --threads 2
Restart=always
KillSignal=SIGQUIT
[Install]
WantedBy=multi-user.targetSave that in /etc/systemd/system with a name like dramatiq.service then run systemctl daemon-reload and then systemctl start dramatiq.
In the above case, /srv/app is a virtualenv so we run python3 from that bin directory to "activate" the venv.
This is probably a better approach now than using supervisord, unless you are running dramatiq tasks in a container.
mangelozzi
commented
2 years ago We use systemd. Our service file looks something like:
Description=Dramatiq Task Queue Workers After=syslog.target [Service] Type=simple Environment=DJANGO_SETTINGS_MODULE=app.settings WorkingDirectory=/srv/app User=someone Group=someone ExecStart=/srv/app/bin/python3 manage.py rundq --log-file /srv/app/log/dramatiq.log --processes 2 --threads 2 Restart=always KillSignal=SIGQUIT [Install] WantedBy=multi-user.target
Refering to http://supervisord.org/configuration.html#fcgi-program-x-section-settings:
Trying to understand your config, but can't find help on some of the options, e.g. .After, Type, WorkingDirectory, KillSignal, ExecStart.
Where did these parameters come from?
Should ExecStart not be command? And KillSignal should be stopsignal?
mangelozzi
commented
2 years ago I had the same issue. Adding the directory parameter solved it, as when it start in the wrong dir, import were failing. Heres my config before its rendered:
; SYM LINK FILE INTO -> {{ SUPERVISOR_CONF_DRAMATIQ }}
[program:dramatiq]
command={{ PYTHON_VENV_BIN }} {{ SRC }}/manage.py rundramatiq --processes {{ CPU_CORES }} --threads 4
priority=3
directory={{ SRC }}
stdout_logfile={{ SUPERVISOR_LOG }}/dramatiq_stdout.log
stderr_logfile={{ SUPERVISOR_LOG }}/dramatiq_stderr.log
stdout_logfile_maxbytes=2MB
stderr_logfile_maxbytes=2MB
I'm trying to deploy using supervisord but it always gives me an error
Any suggestion here? I'm triggering the command by running
rundramatiqthat was provided but it doesn't work