README - Overview for developers

More documentation

We're starting to add more documentation into the doc folder.

If you would prefer to read formatted HTML (highly recommended), simply make doc and then look under doc/_build/html

This README is automatically generated and placed at the top level as part of that documentation build. If you're editing this at the top level, your changes will shortly be clobbered. Edit doc/README.rst instead.

You can browse the latest checked-in version of the docs at http://slug-usergroup.rtfd.org

You can check the status of the latest checking thanks to our free CI service from Travis

Initial Configuration

To get the code and dependencies:

git clone [email protected]:sydney-linux-user-group/slug.git
cd slug
make install

Running a test server

Simply make serve; this will configure a virtualenv, download and install dependencies (inside the virtualenv; your system will not be touched); and a test server will be started.

If this is your first time running make serve you'll be prompted to provide a username and password for an admin account.

Running tests

Simply make test

Coding standards

In general, we follow PEP-8. make lint will tell you in detail about all the things we need to fix.

Production Deployment

1. To account for differences between the dev and prod infrastructure, we have a private repo which needs to be checked out. Exactly where the private repo comes from will be specific to your deployment.

   To see where this is used and get an idea of what you can override, search for private in settings.py

2. The SLUG deployment uses one user to deploy the code, and another user to run the code:

   zhasper@tridge:~$ grep slug /etc/passwd /etc/group
   /etc/passwd:slug:x:1001:1001::/home/slug:/bin/bash
   /etc/passwd:slug-run:x:1005:1005::/home/slug-run:/bin/sh
   /etc/group:slug:x:1001:
   /etc/group:slug-run:x:1005:slug

3. Usually, the run user should only need read access to the files you've checked out. If there are specific files or directories that the run user needs to write to [1], simply use chgrp slug-run $FILE; chmod g+w $FILE to make them accessible to both the deploy and run users. If this needs to be persisted across deployments, you may have to take care of this in your deploy script.

4. For deployment, I simply ssh slug@localhost -A; this turns on agent forwarding so that my usual SSH keys are used to pull the code from bitbucket. For automated deployments, you can create a passwordless key stored for the slug user to use and upload it to Github as a deploy key

5. To do the deployment, I use this simple script, which would benefit from a ton more checking:

   #!/bin/bash
   
   TIMESTAMP=$(date +%Y%m%d-%H%M)
   cd ~/django
   git clone -b django [email protected]:sydney-linux-user-group/slug.git ${TIMESTAMP}
   cd ${TIMESTAMP}
   make install
   make private
   make prepare-serve

   This makes it relatively easy to revert to an earlier version of the code.

   Note

   Depending on your database setup you may have to do some extra work here. If using sqlite you will want to point the new checkout at your existing sqlite database somewhow. If you're using mysql or postgres (configured in private/settings.py) you'll probably want to take a backup of the existing database before running make prepare-serve in case you need to roll back the database changes.

6. We've chosen to run the app inside Green Unicorn, and have it started by upstart:

   slug@tridge:~/django/current$ cat /etc/init/slug.conf
   description "SLUG Django instance"
   start on runlevel [2345]
   stop on runlevel [06]
   respawn
   respawn limit 10 5
   script
     cd /home/slug/django/current
     bin/gunicorn_django -u slug-run -g slug-run
   end script
   slug@tridge:~/django/current$ ls -l /etc/init.d/ | grep slug
   lrwxrwxrwx 1 root root    21 2012-02-04 23:24 slug -> /lib/init/upstart-job

   This solution is not perfect. upstart doesn't kill gunicorn properly, so a restart involves killing a few processes before using sudo service slug start. I need to find time to figure out how to improve this.

7. We've chosen to use nginx as a frontend, and to serve static files. Only a few changes from the default config are needed to accomplish this:

   # path for static files
   root /home/slug/django/current/usergroup/;
   
   location /static/ {
           alias /tmp/slug-static/;
   }
   
   location /admin/media/ {
           root /home/slug/django/current/lib/python2.6/site-packages/django/contrib;
   }
   
   location / {
       proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
       proxy_set_header Host $http_host;
       proxy_redirect off;
   
       proxy_pass   http://localhost:8000/;

   /tmp/slug-static is stipulated as the STATIC_ROOT in settings.py. We should really get around to fixing this - it just needs to be a location that the deploy user can write to and the user running nginx can read from.

Footnotes

[1]	For instance, if you're using sqlite as the database, the run user will need permission to write to the sqlite file