How to reliably determine which PaaS is my project running on?

Question

I have a small project that I want to deploy to both Heroku and OpenShift, to try things out on both platforms. I'm using Python/Django So I want to take advantage of the dynamic import feature to define settings per environment, maintaining a settings module for those platforms I want to deploy to, like:

my_project/
    settings/
        __init__.py  <-- This is what's being imported during app init, here is where I want to detect which platform specific settings module to load
        dev_settings.py
        heroku_settings.py
        openshift_settings.py

So far I use this code:

ENVIRONMENT = the_function_or_code_that_returns_the_environment() # Defaults to "dev"
from django.utils.importlib import import_module

try:
    # Import any symbols that begin with A-Z. Append to lists any symbols that
    # begin with "EXTRA_".
    local_settings_module_name = '%s_settings' % ENVIRONMENT
    local_settings = import_module(local_settings_module_name)
    import re
    for attr in dir(local_settings):
        match = re.search('^EXTRA_(\w+)', attr)
        if match:
            name = match.group(1)
            value = getattr(local_settings, attr)
            try:
                globals()[name] += value
            except KeyError:
                globals()[name] = value
        elif re.search('^[A-Z]', attr):
            globals()[attr] = getattr(local_settings, attr)
except ImportError:
    pass

It would be prefered to detect this via default (not set by me/developer) environment variables. Heroku expose just a few:

DATABASE_URL
ENVIRONMENT
LANG
LD_LIBRARY_PATH
LIBRARY_PATH
PATH
PYTHONHASHSEED
PYTHONHOME
PYTHONPATH
PYTHONUNBUFFERED
SHARED_DATABASE_URL

And OpenShift's:

OPENSHIFT_DB_HOST
OPENSHIFT_LOG_DIR
OPENSHIFT_TMP_DIR
OPENSHIFT_DB_CTL_SCRIPT
OPENSHIFT_REPO_DIR
OPENSHIFT_HOMEDIR
OPENSHIFT_GEAR_NAME
OPENSHIFT_PHPMYADMIN_IP
OPENSHIFT_RUNTIME_DIR
OPENSHIFT_INTERNAL_PORT
OPENSHIFT_PHPMYADMIN_CTL_SCRIPT
OPENSHIFT_DB_MYSQL_51_RESTORE
OPENSHIFT_DB_MYSQL_51_DUMP
OPENSHIFT_DB_PASSWORD
OPENSHIFT_DB_USERNAME
OPENSHIFT_PHPMYADMIN_GEAR_DIR
OPENSHIFT_RUN_DIR
OPENSHIFT_INTERNAL_IP
OPENSHIFT_GEAR_DIR
OPENSHIFT_GEAR_CTL_SCRIPT
OPENSHIFT_APP_DNS
OPENSHIFT_GEAR_TYPE
OPENSHIFT_GEAR_DNS
OPENSHIFT_DB_MYSQL_51_EMBEDDED_TYPE
OPENSHIFT_DB_URL
OPENSHIFT_APP_NAME
OPENSHIFT_DATA_DIR
OPENSHIFT_DB_SOCKET
OPENSHIFT_GEAR_UUID
OPENSHIFT_APP_UUID
OPENSHIFT_DB_TYPE
OPENSHIFT_DB_MYSQL_51_DUMP_CLEANUP
OPENSHIFT_DB_PORT

OpenShift variables are prefixed with OPENSHIFT_ so there's a start, but that's not the case with Heroku.

How can I reliably detect the platform using environment variables?

score 2 · Answer 1 · answered May 19 '12 at 03:08

2

I'd recommend setting my own ENV configuration setting. This can be done easily using heroku.

heroku config:add MY_DEPLOYMENT_PLATFORM=heroku

Then, in your initializer (or wherever you need) you can access that value like so:

if ENV["MY_DEPLOYMENT_PLATFORM"] == "heroku"
 // perform heroku config
else 
 // perform other PAAS config
end

answered May 19 '12 at 03:08

Kevin Bedell

13,254
10
78
114

OpenShift doesn't allow you set custom environment variables (yet), that's why I'm looking for a "heuristic" approach. – Armando Pérez Marqués May 20 '12 at 13:04
You could still check for the presence (or lack off) of the variable. – Kevin Bedell May 20 '12 at 14:00
1

What about `len([k for k in os.environ.keys() if k.startswith('OPENSHIFT_')]) > 0 # Test for OpenShift`, `len([k for k in os.environ.values() if k.find('heroku') > -1]) > 0 # Test for Heroku` ? – Armando Pérez Marqués May 20 '12 at 14:12
1

Exactly -- something like that. You may have to experiment a bit, but yes exactly like that looks good. – Kevin Bedell May 20 '12 at 14:15

score 1 · Accepted Answer · answered May 20 '12 at 14:31

@Kevin, I'm posting my comment it as an answer, so people can see it quickly (or improve it!):

if len([k for k in os.environ.keys() if k.startswith('OPENSHIFT_')]):
    ENVIRONMENT = 'openshift'
elif len([k for k in os.environ.values() if k.find('heroku') > -1]):
    ENVIRONMENT = 'heroku'

local_settings_module_name = '%s_settings' % ENVIRONMENT
local_settings = import_module(local_settings_module_name)
# do the magic with the `local_settings` module

score 1 · Answer 3 · answered Jun 18 '12 at 15:01

This is how I detect the PaaS my app currently on. The overide_settings method isn't mine.

def override_settings(dottedpath):
    """Imports uppercase modules from an string based module.
    Example:
        override_settings('my.module.settings')
    """
    try:
        _m = import_module(dottedpath)
    except ImportError:
        warnings.warn("Failed to import %s" % dottedpath) # <-- will show up in your error log
    else:
        _thismodule = sys.modules[__name__]
        for _k in dir(_m): # <-- moved the block inside else
            if _k.isupper() and not _k.startswith('__'): setattr(_thismodule,
                _k, getattr(_m, _k))

ON_OPENSHIFT = False
if os.environ.has_key('OPENSHIFT_REPO_DIR'):
    ON_OPENSHIFT = True

ON_HEROKU = False
if os.environ.get('LD_LIBRARY_PATH', None) == '/app/.heroku/vendor/lib':
    ON_HEROKU = True

if ON_OPENSHIFT:
    override_settings('settings.openshift')

if ON_HEROKU:
    override_settings('settings.heroku')

How to reliably determine which PaaS is my project running on?

3 Answers3