This is a Django project that you can use as the starting point to develop your own and deploy it on an OpenShift cluster.
The steps in this document assume that you have access to an OpenShift deployment that you can deploy applications on.
This is a minimal Django 1.8 project. It was created with these steps:
- Create a virtualenv
- Manually install Django and other dependencies
pip freeze > requirements.txt
django-admin startproject project .
- Update
project/settings.py
to configureSECRET_KEY
,DATABASE
andSTATIC_ROOT
entries ./manage.py startapp welcome
, to create the welcome page's app
From this initial state you can:
- create new Django apps
- remove the
welcome
app - rename the Django project
- update settings to suit your needs
- install more Python libraries and add them to the
requirements.txt
file
Apart from the regular files created by Django (project/*
, welcome/*
, manage.py
), this repository contains:
openshift/ - OpenShift-specific files
├── scripts - helper scripts
└── templates - application templates
requirements.txt - list of dependencies
To run this project in your development machine, follow these steps:
-
(optional) Create and activate a virtualenv (you may want to use virtualenvwrapper).
-
Fork this repo and clone your fork:
git clone https://github.com/openshift/django-ex.git
-
Install dependencies:
pip install -r requirements.txt
-
Create a development database:
./manage.py migrate
-
If everything is alright, you should be able to start the Django development server:
./manage.py runserver
-
Open your browser and go to http://127.0.0.1:8000, you will be greeted with a welcome page.
To follow the next steps, you need to be logged in to an OpenShift cluster and have an OpenShift project where you can work on.
The directory openshift/templates/
contains OpenShift application templates that you can add to your OpenShift project with:
oc create -f openshift/templates/<TEMPLATE_NAME>.json
The template django.json
contains just a minimal set of components to get your Django application into OpenShift.
The template django-postgresql.json
contains all of the components from django.json
, plus a PostgreSQL database service and an Image Stream for the Python base image. For simplicity, the PostgreSQL database in this template uses ephemeral storage and, therefore, is not production ready.
After adding your templates, you can go to your OpenShift web console, browse to your project and click the create button. Create a new app from one of the templates that you have just added.
Adjust the parameter values to suit your configuration. Most times you can just accept the default values, however you will probably want to set the GIT_REPOSITORY
parameter to point to your fork and the DATABASE_*
parameters to match your database configuration.
Alternatively, you can use the command line to create your new app, assuming your OpenShift deployment has the default set of ImageStreams defined. Instructions for installing the default ImageStreams are available here. If you are defining the set of ImageStreams now, remember to pass in the proper cluster-admin credentials and to create the ImageStreams in the 'openshift' namespace:
oc new-app openshift/templates/django.json -p SOURCE_REPOSITORY_URL=<your repository location>
Your application will be built and deployed automatically. If that doesn't happen, you can debug your build:
oc get builds
# take build name from the command above
oc logs build/<build-name>
And you can see information about your deployment too:
oc describe dc/django-example
In the web console, the overview tab shows you a service, by default called "django-example", that encapsulates all pods running your Django application. You can access your application by browsing to the service's IP address and port. You can determine these by running
oc get svc
Templates give you full control of each component of your application. Sometimes your application is simple enough and you don't want to bother with templates. In that case, you can let OpenShift inspect your source code and create the required components automatically for you:
$ oc new-app centos/python-35-centos7~https://github.com/openshift/django-ex
imageStreams/python-35-centos7
imageStreams/django-ex
buildConfigs/django-ex
deploymentConfigs/django-ex
services/django-ex
A build was created - you can run `oc start-build django-ex` to start it.
Service "django-ex" created at 172.30.16.213 with port mappings 8080.
You can access your application by browsing to the service's IP address and port.
By default your Django application is served with gunicorn and configured to output its access log to stderr. You can look at the combined stdout and stderr of a given pod with this command:
oc get pods # list all pods in your project
oc logs <pod-name>
This can be useful to observe the correct functioning of your application.
You can fine tune the gunicorn configuration through the environment variable APP_CONFIG
that, when set, should point to a config file as documented here.
When using one of the templates provided in this repository, this environment variable has its value automatically generated. For security purposes, make sure to set this to a random string as documented here.
At times you might want to manually execute some command in the context of a running application in OpenShift. You can drop into a Python shell for debugging, create a new user for the Django Admin interface, or perform any other task.
You can do all that by using regular CLI commands from OpenShift.
To make it a little more convenient, you can use the script openshift/scripts/run-in-container.sh
that wraps some calls to oc
.
In the future, the oc
CLI tool might incorporate changes
that make this script obsolete.
Here is how you would run a command in a pod specified by label:
-
Inspect the output of the command below to find the name of a pod that matches a given label:
oc get pods -l <your-label-selector>
-
Open a shell in the pod of your choice. Because of how the images produced with CentOS and RHEL work currently, we need to wrap commands with
bash
to enable any Software Collections that may be used (done automatically inside every bash shell).oc exec -p <pod-name> -it -- bash
-
Finally, execute any command that you need and exit the shell.
Related GitHub issues:
The wrapper script combines the steps above into one. You can use it like this:
./run-in-container.sh ./manage.py migrate # manually migrate the database
# (done for you as part of the deployment process)
./run-in-container.sh ./manage.py createsuperuser # create a user to access Django Admin
./run-in-container.sh ./manage.py shell # open a Python shell in the context of your app
If your Django pods are labeled with a name other than "django", you can use:
POD_NAME=name ./run-in-container.sh ./manage.py check
If there is more than one replica, you can also specify a POD by index:
POD_INDEX=1 ./run-in-container.sh ./manage.py shell
Or both together:
POD_NAME=django-example POD_INDEX=2 ./run-in-container.sh ./manage.py shell
You can deploy this application without a configured database in your OpenShift project, in which case Django will use a temporary SQLite database that will live inside your application's container, and persist only until you redeploy your application.
After each deploy you get a fresh, empty, SQLite database. That is fine for a first contact with OpenShift and perhaps Django, but sooner or later you will want to persist your data across deployments.
To do that, you should add a properly configured database server or ask your OpenShift administrator to add one for you. Then use oc env
to update the DATABASE_*
environment variables in your DeploymentConfig to match your database settings.
Redeploy your application to have your changes applied, and open the welcome page again to make sure your application is successfully connected to the database server.
If you get stuck at some point, or think that this document needs further details or clarification, you can give feedback and look for help using the channels mentioned in the OpenShift Origin repo, or by filing an issue.
This code is dedicated to the public domain to the maximum extent permitted by applicable law, pursuant to CC0.