Skip to content
This repository has been archived by the owner on Aug 31, 2021. It is now read-only.

replicatedhq/replicated-starter-ship

Repository files navigation

Replicated Ship Starter

Starter project for managing a Replicated Airgapped Kubernetes Appliance or Ship application in a GitHub repo.

Prequisites

  • make
  • node
  • ship installed. On macOS, you can use brew install ship to install with Homebrew
  • A GitHub repository created to store your Ship application assets.

Note: While Ship supports any git repository, this example leverages features that are GitHub-only (for now). The repo you create can be private or public.

Project overview

This project contains an example application that can be deployed with ship. The main pieces are

  • base -- Kubernetes YAML that defines the application
  • scripts -- Shell scripts to be distributed with the application, primarily for testing and troubleshooting the application
  • ship.yaml -- ties these pieces together into a deployable application
  • Makefile -- Workflows for testing the application installation experience
  • CI integration starters for testing changes to your application.

Get started

This repo is a GitHub Template Repository. You can create a private copy by using the "Use this Template" link in the repo:

Template Repo

You should use the template to create a new private repo in your org, for example mycompany/replicated or mycompany/replicated-starter-ship.

Once you've created a repository from the template, you'll want to git clone your new repo and cd into it locally.

Hello, World!

To get started, you'll want to update the following fields in Makefile and ship.yaml:

  • Makefile REPO -- update this to the <owner>/<repo> of your repository on GitHub
  • ship.yaml assets.v1.*.github.repo -- this should match the REPO value in Makefile

You can test this out by launching ship with

make run-local

This will open a browser and walk you through configuring the application defined in ship.yaml. The test application creates a small Kubernetes Deployment to run Nginx, but it's a good way to get a sense of how ship works.

You can inspect the YAML at tmp/rendered.yaml, and deploy the app using kubectl by running

make deploy-ship

or

kubectl apply -f tmp/rendered.yaml

Iterate on your App

From here, you can add messaging and configuration options in the config and lifecycle sections of ship.yaml, and modify YAML in base to match your kubernetes YAML.

The above

make run-local

task can be run again to see the new changes. To iterate without using the UI, you can use

make run-local-headless

to regenerate assets. State will be stored in tmp/.ship/state.json between runs, and will persist any changes to config options or Kustomize patches. To deploy it after running, you can

make run-local-headless deploy

Create releases

To create a release of a Replicated Embedded Kubernetes or Ship app to vendor.replicated.com, you'll need to configure a few additional fields.

  1. Export your REPLICATED_API_TOKEN as described in the kubernetes starter project
  2. Set the following values in your Makefile. See the kubernetes starter project for information on how to get your application ID.
APPLIANCE_APP_ID := CHANGEME
APPLIANCE_CHANNEL := Unstable

SHIP_APP_ID := CHANGEME
SHIP_CHANNEL := Nightly

Then, you can create appliance releases with

make release-appliance

and create ship releases with

make release-ship

Integrate with CI

The project includes CI configs for Travis CI and CircleCI.

Both configs will lint your ship.yaml and replicated.yaml for syntax and logic errors. Once a PR is merged to master, a release will be promoted to the channels designated in your Makefile.

Note about appliance

While the Makefile includes a task for release-appliance, this is still very much a ship-focused starter project.

Example workflows

The optimal workflow for delivering the same Kubernetes manifests to both Ship and Appliance apps will depend greatly on how your team ships code. An example workflow is shown below.

Note: This example asumes you'll be updating your kubernetes manifests in this ship repo manually. Depending on your scale and your process, its possible that your team will instead keep this infrastructure repository up to date automatically. For example, if you have many service repos, you may have core CI jobs for each service push updates to Kubernetes Manifests and image tags in this infrastructure repo as part of your delivery process

Tools reference

License

MIT

Troubleshooting

make run-local fails with github asset returned no files

  1. Double check the assets.v1.github entries in your ship.yaml match the --set-github-contents flags in your Makefile.

  2. Note that the make run-local and make run-local-headless tasks don't handle symlinks well. If you have symlinks in your repo, or you've symlinked the repo root, this can cause issues. To determine if this is the cause, you can temporarily replace symlinks with the content they point to.