Author: Ian Huston @ianhuston
License: See LICENSE.txt in each example code folder
First presented at PyData London 2015.
This tutorial is an introduction to deploying Python applications to Cloud Foundry and covers:
- Pushing your first application to Cloud Foundry
- Using a custom buildpack to use PyData packages in your app
- Adding data services like Redis to your app
Each example app is self-contained and can be deployed to a Cloud Foundry instance with
$ cf push
The slides are available on Speakerdeck. There is also a full length video of this tutorial from PyData London 2015.
A shorter version of these notes, which can be used by assistants in the tutorial is also provided.
Update 03/10/2016: The official CF Python buildpack now includes support for the conda package manager. This means there is no need to use a separate community buildpack to run apps using PyData packages. The tutorial has been updated to reflect this.
Update 23/11/2015: These sample apps have been updated to use the new Diego runtime. The main difference is that app instances all run on the same port inside different containers. The first sample app has been changed to show the instance number rather than the port so that multiple instances can be recognised.
If you do not have an account on a Cloud Foundry installation you can register for a free trial at Pivotal Web Services (PWS).
Download the Cloud Foundry Command Line Interface from the CF management console
or the CF Github repo.
This provides the cf command which you will use to interact with a CF installation.
Target the CF installation API with cf api. For example for PWS:
$ cf api https://api.run.pivotal.io
The first app is in the folder 01-simple-python-app.
Have a look at the code and then deploy using cf push.
$ cd 01-simple-python-app
$ cf push
You can inspect the app's streaming logs using cf logs myapp.
You can scale your app by changing the number of instances and the available RAM:
$ cf scale myapp -i 5
$ cf scale myapp -m 256M
The second app is in the folder 02-pydata-spyre-app.
This app is built using Spyre by Adam Hajari.
This application uses a new feature of the official Python buildpack
to provision dependencies using conda.
The environment.yml file contains the conda environment specification.
The buildpack detects this file and uses conda instead of pip to install the dependencies.
Push the app using cf push and visit the app URL to see a simple interactive visualisation.
$ cd ../02-pydata-spyre-app
$ cf push
The third app is in the folder 03-services-redis.
Services provide persistent storage and other useful resources for your applications.
First, create a free Redis service provided by Rediscloud:
$ cf create-service rediscloud 30mb myredis
You can bind this to your apps making it available via the VCAP_SERVICES
environmental variable.
$ cf bind-service myapp myredis
Look at the environmental variables your app has with cf env myapp.
Now you can push the third app which uses Redis as its backing store.
$ cd ../03-services-redis
$ cf push
The manifest specifies that this app should be bound to the myredis service instance that you just created.
The fourth app is in the folder 04-learning-api.
We can put everything we've learned into practice by building our own simple machine learning API.
$ cd ../04-learning-api
$ cf push
This is a simplified version of a project by Alex Kagoshima.
Don't forget to stop all your apps when you are finished by doing the following:
$ cf stop myapp
$ cf stop sinewave
$ cf stop testredis
$ cf stop learning-api