If you use virtual environments on your django servers, then getting manage.py
commands to run from cron is a little tricky. You need to activate the
virtualenv before running the command and so you might think that the following
*/10 * * * * root source /var/www/mysite/virtualenvs/dev/bin/activate && /var/www/mysite/build/dev/manage.py some_custom_command > /dev/null
It doesn't, although it's tricky to spot why as /var/log/syslog doesn't give much away (Debian-specific of course).
A good trick for cronjob debugging is to alias yourself as root within
postmaster: root root: email@example.com
and run sendmail -bi to initialise the aliases. As errors from cronjobs are emailed to root, you will also get a copy. Doing this reveals the above cron file fails as the default shell for cron is /bin/sh which doesn't support the source command.
The solution is to set the $SHELL variable within the cron file:
SHELL=/bin/bash */10 * * * * root source /var/www/mysite/virtualenvs/dev/bin/activate && /var/www/mysite/build/dev/manage.py some_custom_command > /dev/null
Update - have been informed of a much simpler technique that works for most cases: simply run manage.py using the python executable of your virtualenv:
*/10 * * * * root /var/www/mysite/virtualenvs/dev/bin/python /var/www/mysite/build/dev/manage.py some_custom_command > /dev/null
I didn't spot this one to start with as our settings configuration required a environmental variable to be used to indicate which settings file to use. This variable was set within the activate script, hence why the source command was needed. It turns out that this can just be set in the cron file too:
DJANGO_CONF=conf.dev */10 * * * * root /var/www/mysite/virtualenvs/dev/bin/python /var/www/mysite/build/dev/manage.py some_custom_command > /dev/null