drift-base

Base Services for Drift micro-framework.

Installation:

Run the following commands to install this project in developer mode:

pip install --user pipenv
pipenv install --dev

Run the following commands to enable drift and drift-config in developer mode for this project:

pipenv shell  # Make sure the virtualenv is active

pip install -e "../drift[aws,test]"
pip install -e "../drift-config[s3-backend,redis-backend]"

Run localserver

This starts a server on port 5000:

pipenv shell  # Make sure the virtualenv is active

export FLASK_APP=drift.devapplocal:app
flask run

Try it out here: http://localhost:5000/

Modifying library dependencies

Python package dependencies are maintained in Pipfile. If you make any changes there, update the Pipfile.lock file as well using the following command:

pipenv --rm && pipenv lock --verbose

Working with AWS

Note! For any of the following commands to work, make sure the virtualenv is active and the proper configuration database and tier is selected:

pipenv shell
export DRIFT_CONFIG_URL=somecfg && export DRIFT_TIER=SOME_NAME

Building drift-base

Drift-base runs in docker. To build and push a docker image run the following:

make build

This will create a docker image called directivegames/drift-base:<branch-name> and push it to dockerhub here: https://hub.docker.com/repository/docker/directivegames/drift-base/tags?page=1

You can run the container locally with the following command:

make run

Note that you must have the following environment variables set up: DRIFT_TIER and DRIFT_CONFIG_URL. See example.env.

drift-base docker images are automatically built by travis on all branches and tagged by the branch name. If any git tag is pushed, a docker image will be built with that tag as well.

Versioned images are created in this way. Simply add a version tag to git and an image with correct version will be built. Any image built after this version tag push will export the same version in its root endpoint.

To create a new version of drift-base run:

git tag 1.2.3
git push --tags

Note that tiers will typically run a branch-tagged version of drift-base. A development tier will run :develop and a live tier :master. However, this behavior can be set per tier when the launch config and autoscaling group is set up (see next section).

The Watchtower https://hub.docker.com/r/v2tec/watchtower/ service is running on the ec2 machines that serve drift-base and will automatically update the service every 5 minutes. This means that if you push to develop the development tier should have that version running within 5 minutes. You can see when this has been completed by inspecting the build_info key in the root endpoint of the service.

Create an AMI

Since drift-base is running in docker it is not necessary to create new AMI's for every deployment. In fact, new ami's only need to be created when there are updates to the infrastructure or basic dependencies like datadog and watchtower.

To build an AMI you will need to have the current AWS profile set correctly so that you can create instances in the eu-west-1 region.

make buildami

This will build a new AMI image and make it available in the following regions:

eu-west-1
us-east-1
ap-southeast-1

The built AMI will however not be used until it is added to autoscaling. For testing you can always launch it yourself however, and I would recommend using 'launch more like this' on an existing drift-base instance to get all the required settings but that is an exercise for the user.

AMI's are not versioned but their name contains the timestamp when they were created.

There is no service or tier configuration baked into the AMI. All that information is added via user data when the AMI is launched (new launch config created) below.

Launching AMI

Once a new AMI has been created it needs to be added into an autoscaling group in a tier. To do this you can run the following command:

make launchami

Before you do this however, you will need to set up your environment. We recommend copying example.env to .env and running pipenv shell, which will source the .env file and makes things easier, but it's up to you how you set up the environment.

There are a few things you will need to add into the .env file such as the tier name, redis url of the drift config database, etc. You will need to gather these pieces of information from elsewhere.

Since building AMI's is a rare occurence there is no CI process for this and the AMI's will need to be built manually using this process.

If you want to change the docker tag that should be deployed to a tier you can use this process without needing to create a new AMI.

The command will only take a few seconds to run and a new AMI should be running in the tier via autoscaling within 5 minutes.

dgnorth/drift-base