dshaneg / triple-play

Multi-repository version of double-tap. Experimenting with different styles of dealing with charts and config.

triple play

Proof of concept for creating a "workspace" container to be published along with the application's main container. Alternate multi-git-repository option based on double-tap

The workspace container can be used to run deploy and test scripts against the main container.

To Do

• proper smoke test (there's a dummy in place now)
• proper automated acceptance tests on a running container
• automated rollback if smoke fails
• delete deployment for volatile stages (aat, etc)
• include non-helm deployment steps (e.g. s3 bucket)
• alternate example with pieces fully broken apart into separate repos
  • app
  • chart
  • config/deploy?

Building the application

make build

Running the application locally (not in k8s)

Once built, you can run the application in the container.

make run

This will start the application and leave the container running in the background. You'll have to stop it before you can run it again.

make kill

Publishing the docker images

On a build server, need to set environment variable APP_VERSION to a valid semantic version value. This will become the tag of the docker image as well as the version of the helm chart.

If the APP_VERSION variable is not set (as when running the build locally), the build will use version 0.0.0, and 'latest' for the docker image tag.

make publish

Deploying the application

A build server should have the APP_VERSION environment variable set as for the Publish step.

On a build server, you need to make sure the environment variable STAGE is set to a value that matches one of the filenames in the config folder, excluding the suffix: (e.g. cert, local, prod).

Both variables use default values if not set, but any automation should explicitly set values for these variables--the defaults are intended for local development.

BE CAREFUL when running this step locally if you have kubernetes credentials to any higher environment clusters (like production). Make sure your current context is set to a local or dev cluster. The process copies your kube config into the deploy container and lets it fly.

make deploy