I am trying to run Gunicorn from Supervisor in an Ubuntu 12.04 system. Gunicorn runs a Flask app (simple REST web service tested with Flask's embedded server). I have installed Gunicorn by clonning GIT repo, trying to avoid 'apt-get install' because it runs Gunicorn server when installs it. I do not want it running, it will be run by Supervisor only.

So after install it, if I try:

cd /usr/local/bin
gunicorn my_app:app -c /path/to/gu_config_file

Gunicorn works. Then I kill it. Note config file without extension, because with '.py' extension does not work for me. So I edit Supervisor's config file like:

[program:gunicorn]
command=/usr/local/bin/gunicorn my_app:app -c /path/to/.gu_setup
directory=/usr/local/bin/
autostart=true
autorestart=true
redirect_stderr=True

And update changes in Supervisor:

supervisorctl reread
# gunicorn: changed
supervisorctl update
# gunicorn: stopped
# gunicorn: updated process group

Detects changes in file and works for Gunicorn program. Ok, but then I try to start it:

supervisorctl start gunicorn

Getting an annoying:

gunicorn: ERROR (abnormal termination)

Checking supervisor's log:

2013-03-08 13:07:22,378 INFO spawned: 'gunicorn' with pid 3355
2013-03-08 13:07:22,916 INFO exited: gunicorn (exit status 3; not expected)
2013-03-08 13:07:23,918 INFO spawned: 'gunicorn' with pid 3361
2013-03-08 13:07:24,492 INFO exited: gunicorn (exit status 3; not expected)
2013-03-08 13:07:26,496 INFO spawned: 'gunicorn' with pid 3367
2013-03-08 13:07:27,078 INFO exited: gunicorn (exit status 3; not expected)
2013-03-08 13:07:30,085 INFO spawned: 'gunicorn' with pid 3373
2013-03-08 13:07:30,628 INFO exited: gunicorn (exit status 3; not expected)
2013-03-08 13:07:31,630 INFO gave up: gunicorn entered FATAL state, too many start retries too quickly

I do not know what to do right now... Can you help me? Thx a lot!

EDIT: sorry I forgot to say I have exported PYTHONPATH variable as:

export PYTHONPATH=/usr/local/bin:/usr/local/lib/project

'my_app' is in /usr/local/bin. The lib path is needed for other modules. I have edited also Supervisor config file to indicate environmental variable, like:

environment=PYTHONPATH=/usr/local/bin:/usr/local/lib/project/

But did not work.

EDIT 2: as @robertklep suggest in his comment, this is log's output:

Traceback (most recent call last):File "/tmp/gunicorn/gunicorn/arbiter.py", line 485, in spawn_workerworker.init_process()File "/tmp/gunicorn/gunicorn/workers/base.py", line 100, in init_processself.wsgi = self.app.wsgi()File "/tmp/gunicorn/gunicorn/app/base.py", line 103, in wsgiself.callable = self.load()File "/tmp/gunicorn/gunicorn/app/wsgiapp.py", line 25, in loadreturn util.import_app(self.app_uri)File "/tmp/gunicorn/gunicorn/util.py", line 369, in import_app__import__(module)File "/usr/local/bin/my_app.py", line 4, in <module>import const
ImportError: No module named const
2013-03-08 13:29:35 [3670] [INFO] Worker exiting (pid: 3670)
2013-03-08 13:29:36 [3665] [INFO] Shutting down: Master
2013-03-08 13:29:36 [3665] [INFO] Reason: Worker failed to boot.

'const' module is in /usr/local/lib/project...

I don't see you setting the environment in your supervisor config file:

[program:gunicorn]
environment=PYTHONPATH=/usr/local/bin:/usr/local/lib/project
command=/usr/local/bin/gunicorn my_app:app -c /path/to/.gu_setup
...

If that doesn't work, try starting gunicorn in debug mode:

command=/usr/local/bin/gunicorn --debug --log-level debug my_app:app -c /path/to/.gu_setup

Or pass the path directly to gunicorn:

command=/usr/local/bin/gunicorn --pythonpath /usr/local/bin,/usr/local/lib/project my_app:app -c /path/to/.gu_setup

EDIT: gunicorn's --pythonpath is broken, you can only pass one directory:

command=/usr/local/bin/gunicorn --pythonpath /usr/local/lib/project my_app:app -c /path/to/.gu_setup