Deploying an existing Flask App on Heroku
If you already have a Flask app running locally by just running it in IDLE, or at the command line (e.g. python hello.py), and you want to convert it to run on Heroku, you need to do three things. Each is very simple.
- It needs to be in a git repository.
- You need a Procfile—this is usually just one line of code (see below.)
- You need a requirements.txt file. This is generated automatically with one unix command (see below.)
- You use the command heroku create to set up a remote repository on heroku where you can deploy your application
- You use git push heroku master to deploy your application.
- You can see your application on the web, or use heroku logs to see the logs (if there are errors.)
(Note to instructors: The heroku create and heroku logs commands are provided by the “Heroku toolbelt”, and it was installed especially for SPIS. It requires installation of Ruby, then it is a simple download.)
First, to deploy on Heroku, we need to create two extra files. We need a file called Procfile in our git repo. This file tells Heroku what to do with our github repo when we push it to github. It should contain the following:
web: gunicorn hello:app --log-file=-
The part of this line that reads
hello:app assumes that the main python code for your web app is in
hello.py, and that the variable
app is the one that appears in the line of code
app = Flask(__name__).
If that is not the case, you may need to adjust either
app as needed.
Now that we have that file, you will want to do these commands to commit this file to our github repo.
git add Procfile git commit -m "Added Procfile needed by Heroku" git push origin master
We also need a file called
requirements.txt which is a list of the Python modules that are needed for our Heroku flask application.
This file will list all of the Python modules that we may have installed using
pip install --user blah, including
flask, and anything else that
flask might have required.
Note that before you do the next step, you should do the following
pip install command if you haven’t already. While this next line isn’t necessarily needed for running Flask applications locally, it is needed for Heroku.
pip install --user gunicorn
We can create the file
requirements.txt with this command:
pip freeze > requirements.txt
But we won’t do that! Because “pip freeze” outputs the installed packages in the requirements format; however, over the course of SPIS, we have installed many packages, and the list is very very long; also heroku does not like some of the packages.
Instead, go ahead and create a file called “requirements.txt”
(hint: you can do this by typing "idle requirements.txt" into the command line), and paste this into the file:
Flask==0.10.1 itsdangerous==0.24 Jinja2==2.8 MarkupSafe==0.23 Werkzeug==0.10.4 wheel==0.24.0 gunicorn==19.3.0
We now have a list of packages our program needs to run.
Go ahead and save that file, and now lets push that to github as well:
git add requirements.txt git commit -m "Added list of Python modules needed by Heroku" git push origin master
heroku create and notice the name of the application created.
- It will take the form word-word-number, e.g. flying-tomato-4321
The next step is to type:
git push heroku master
git push heroku master, you’ll probably see lots of output, showing that your webapp is now running on Heroku.
- To see your app, visit https://word-word-number.herokuapp.com,
- e.g. https://flying-tomato-4321.herokuapp.com
If there are errors, check them by typing
A side note about that “itsdangerous” thing
When I first saw that name show up in the modules we were downloading, I was a little taken aback. If you are worried about having something called “itsdangerous” in your account, this paragraph is to reassure you that its not dangerous.
I read the documentation for the itsdangerous module and realized that that the only thing dangerous here was the name. The name refers to the fact that sometimes data has to be passed from a “trusted environment” to an “untrusted environment” or vice-versa, and when that happens, you want to “sign” the data—that is, do some cryptography with it—to ensure that it isn’t modified enroute. There isn’t anything “dangerous” about the software itself. On the contrary—not using it would be dangerous.