In this project I explore making my own python package and installing it as a dependecy for this flask app, I also give a basic example of a complete continuous integration pipline from development to deployment to a docker swarm hosted on http://lydanny.com. for a live example(hopefully still running since I have student credits)

• http://lydanny.com:6969
• http://lydanny.com:6969/finviz
• http://lydanny.com:6969/hello
• http://lydanny.com:6969/REST

If you also want to see the condition of the swarm check out

• http://lydanny.com:8080

1. Docker
2. Docker swarm
3. Docker Hub
4. Travis CI
5. pytest
6. Flask
7. Jinja2
8. Sherperd
9. Git
10. redis

The pipline starts at our local development system , cloning the git repo we begin the development, as we develop we can branch to feature branches or stay with master, changing the code base and checking the features on our local machine with docker local host machine.

As we commit and push our code changes, this is sent to travis CI for unit testing and docker build confirmations, this is done with python pytest, to test our new feature builds. Travis will clone our entire repo as well as start its own virtual env on their services, following the defined .travis.yml instructions to set up the build enviroment.

Once the build environment is setup our unit test(pytest) scripts are called, any of the 4 stages of travis build stages that return a non-zero exit code are considered broken build: [before_install],[after_install],[install],[script]. if any of these return non-zero exit codes then travis will notify the users (defined within the travis.yml config) once a build has passed the predefined unittest code, it then moves to a script to test docker image build passes, sometimes unittest passes however the docker build image (in production may fail) so we test this here. If the building and running of the docker image does NOT return a non-zero exit code, we move to start building the actual image appending the latest build version tag to it. once the build has completed travis is instructed to upload/push the new image to our docker hub repo.

We use shepered (very small 4mb) docker image running on our docker swarm, to periodically check for updates on our containers currently running on the cluster, if there is a new digest in our dockerhub( which is all taken care of by docker hub itself) then sherpered will, gracefully perform a rolling update on the current running containers, pulling the new build-passed code updates to the running environment(production)

This example concludes there are alot of configurations to learn and undertand however please check out these tools.

Travis CI Sheperd Docker Flask Jinja2