This project is meant for you to make awesome software, with the skills in your Continuous Delivery and DevOps toolbox. Have fun, automate and be awesome!
The project can be run by 1-3 people in a team.
Fork this repository so you can make changes to it:
On your repository, enable issues:
Congratulations, you just scored your first job. Fresh out of college. It pays decently and the interviewers seemed nice. There are all sorts of perks. It even sounded like you would work with some bleeding edge technologies.
So now you're ready for whatever tasks they'll throw at you. You started yesterday together with a few college grads. You spent the day yesterday getting access to the building. You know now how to both obtain lunch and relieve any pressures such a feast might impose on you.
You've got your state-of-the-art laptop booted and are ready to make the world a better place, To innovate! To use all the skills you've gathered while studying at the university.
Your project manager calls you for a meeting and let's you in on the details on the first task that you'll complete.
He truly knows that you are very eager to get started and do cool stuff. But unfortunately they have a Python Flask Web application that runs in production at a bunch of customers, and no one quite remembers how it works. It's called CoDeChan, just like 4chan but for CoDe. You get it, right, don't you? Ahh never mind then....
The application is CoDe Chan an anonymous posting tool. No one really knows why or for what the customers use it. But none the less, it brings in some cash.
It is obviously important to keep this application running well and good, otherwise Initrode might end up looking very bad indeed.
Congratulations, on your second day you became the proud owner of a piece of Legacy Code.
Now, where to start?
You've been handed a huge pile of .. Well you do not quite know of what.
You know that you need two things in order for you to be a success.
You need to know what is happening to the code and who is working on what.
The first step is obvious. Put your code under version control and set up a task management system for your project.
An easy way to create an agile task management setup for your project is to use GitHub issues together with their projects tab. This allows you to have a full swim lane and track your tasks in a nice visual way that is completely integrated with GitHub.
Go over to the Projects tab and create your project.
- Use your fork of the Git repository containing the code
- Create issues in your repository
- See
github_issues.pyfor the issues to create - To run the Python script, edit
github_issues.pyto setREPO_OWNERandUSERNAMEwith your own GitHub account name - Then set an environment variable called
GITHUB_PASSWORDe.g.export GITHUB_PASSWORD=yourSuperSecretPassword
- See
Python hints for the issues script:
- Linux:
sudo apt-get install python-pipsudo pip install requests
- Mac:
sudo easy_install pipsudo pip install requests
From now on you should consider every step in a task as a story, also created in your GH repository. Remember to groom them together, breaking down the stories where it makes sense.
In some ancient documentation you can see that you can run the application with the command python run.py.
Try that and see how that flies for you.
Of course dependencies are not explicitly managed, so you can't just start up the application.
Figure out the dependencies, put them in requirements.txt so you can install them with pip install -r requirements.txt.
Now that you've fixed that hurdle and you've actually got an application running, spend a few minutes familiarizing yourself with it.
- Investigate dependencies
- Put dependencies in
requirements.txt - Install dependencies with
pip install -r requirements.txt - Run application with
python run.py - Familiarize yourself with the application
You've got your dependencies handled and all is good. Or sort of good.
It is tedious to set up dependencies on each system that you want to run on.
Luckily you know how to use Docker.
- Create a
Dockerfilecontaining your application code and requirements - Build the Docker image
- Run the application in Docker with a docker-compose file
- Check Dockerfile and compose file into your repository.
The ability to run your project in Docker has made many things easier for you. You even have the ability to track code changes and manage your self-organizing team. Things are looking up for you.
You keep stepping a bit on each others toes, so of course you look into some sort of automated tests.
Luckily the project was not created by entities of pure evil. There are some tests that can be run. You need to automate them. You can run the current set of tests with python tests.py.
Continuous Integration is the goal, and you look to your good old friend CircleCI for some needed support.
- Setup a CircleCI project
- Setup a Continuous Integration pipeline (the config CircleCI is running)
- Run the tests in the pipeline
- Make sure you maintain mainline integrity meaning all commits to master needs to be tested by CI before merging.
Hint: A good starting point could be GitHub Flow
Every time that release time comes around, you get uncomfortable. You do not feel safe about deploying to production, and it always seems to be the case that nothing is quite the same from time to time.
It is now time to script your way to deployment.
Hint: If you want to run a dockerfile on your server, you can ssh into the server from CircleCI, just like you SSH into the server from your machine. For more into, look here
- Create a script that runs the application on your server.
- Augment your script such that you can deploy to multiple targets ( eg. local, staging, production ).
- If you are more people in the team, try to push your code to one of the other servers, using its private IP (You can get that by issuing the command
ifconfig).
You realise that your process is actually quite mature now, but you don't exactly feel you're quite done.
Even though you are testing the internal quality of your code - you do not have any means of functional testing.
You just want the bare minimum test. So you add a test in your pipeline that will deploy your Dockerized application temporarily and test its availability.
Even a very simple thing as being able to reach your server with curl or wget would be much better than what you have now. Nothing.
- Use your automated deploy to deploy in testing
- Do functional testing (You know your servers IP)
- Display result in CircleCI
- This might not be a tollgate criteria, but it is important information for you to gain.
You've reached a very mature point for your software development. You have a containerized, automatically testable and deployable application.
You know who is working on what and why. You have full faith that you can refactor your legacy project without destroying anything.
As you dare to change the code, it is no longer legacy, it is just your software.
Now it is time to tread new paths.
The following are suggestions to explorations that you can take.
- Run a linter as part of the pipeline
- Stress the application, using for instance the docker image
rufus/siege-engine(more info about using the application can be found here) - Refactor to remove the unused User and login code from the application
- Add another machine as a build node, and make it run some of the steps in the pipeline.
- Deploy to production if functional tests pass
- Add something to do your buildtasks for you. Gradle, Rake, Grunt, Make, etc.
- Allow rollbacks to a previous version
- Setup an ELK-stack for monitoring
- Use for instance HAProxy to have multiple containers running in production, through a single interface
- Use persistent storage for your SQLite database
- Setup a database in a separate Docker container, change application to use it
- Investigate the usage of docker-compose for your multi-container setup
- Investitage the usage of Kubernetes for your multi-container setup
- Do some TDD on the application
- Make sure that you can't do an empty post
- Make links in a post clickable
- Make the layout prettier
- Add additional pages
- Add fields to the post