Need help with rails-k8s-demo?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

lewagon
296 Stars 27 Forks 99 Commits 1 Opened issues

Description

Easy Rails on Kubernetes approach for Digital Ocean and a classic Rails stack: Puma, Redis, Sidekiq, Postgres, Action Cable, Webpacker. Helm 3 in production and docker-compose + dip (https://github.com/bibendi/dip) in development. Issues and PRs welcome.

Services available

!
?

Need anything else?

Contributors list

# 166,733
Ruby
microfr...
messeng...
faceboo...
67 commits
# 163,827
ebay
PHP
rubydat...
Rails
1 commit
# 48,318
digital...
Go
github-...
golang
1 commit

Le Wagon x Digital Ocean Rails-on-Kubernetes demo

:construction: This setup is a work in progress :construction:

An attempt at creating a standardized Docker/Kubernetes-Helm/GitHub Actions setup for use with Rails projects. A version of this setup (that is little bit further ahead) is now used in production for services available to Le Wagon teachers, students, and alumni.

Demo application

A standard Action Cable chat application (Rails 6 + Stimulus) that uses twitter-text gem to extract free-formed URLs out of text and generate Open Graph previews for links. The dependency was chosen consciously, as it requires an external system library to function (

libidn11
), which makes it a perfect case for Docker.

It requires Sidekiq, Redis, and PostgreSQL to run, which represents the standard deployment stack of production Rails apps.

The demo application at https://demo.lewagon.co

Goals

  • Create a sensible approach to running fairly complex Rails projects in development with Docker
  • Generate reproducible and customizable configuration for production
  • Store everything related to deployments in the same repo as application code. Rely on GitHub one-way encrtypred secrets to store sensitive data like database connection strings.
  • Cater to a "classic Rails stack": Rails, Puma, Sidekiq, Redis, PostgreSQL, Webpacker, Action Cable
  • Allow for only a certain amount of complexity: enough to fit inside a head of a single full-stack developer who may end up maintaining the projects.
  • Hide ugly details of Kubernetes management by relying on 1 environment/1 cluster approach and letting Digital Ocean manage the master node.
  • Deploy with
    git push origin master
    or a PR merge on GitHub
  • Use any external Postgres database
  • Provide the easy migration from "Heroku way of doing things", once the application is past the MVP and needs to scale for a reasonable cost.

Current status

By constantly migrating legacy applications to this setup and starting new ones with the same setup in mind we make sure that all common production bugs will be squashed before we can fully automate the generation of charts and configuration.

Running locally

We use dip by Misha Merkushin at Evil Martians to achieve a "classic CLI" feel while working with Docker under the hood.

  • gem install dip
  • dip provision
    to build all necessary images and setup the local DB
  • dip up -d
    to run all services in the background:
    app
    ,
    rails
    ,
    sidekiq
    ,
    webpacker
    (runs
    webpack-dev-server
    for JS live reload),
    postgres
    ,
    redis
    . Go to
    localhost:3000
    :tada:
  • dip rails c
    to visit console.
  • dip down
    to stop the music and go home.

If you need to change project dependencies—run

dip bundle
and/or
dip yarn
after any changes to
Gemfile
or
package.json
.

Check out

dip.yml
for other common shortcuts.

Alternatively, you can launch

dip rails s
,
dip sidekiq
, and
dip webpacker
in respective tabs. This will allow for easier debugging with
byebug
(no need to attach to running Docker processes).

Many-tabs workflow

Running in production

  • Provision a cluster from Digital Ocean Managed Kubernetes, install Nginx Ingress Controller, create DNS A record, set up jetstack/certmanager for SSL. Expect futher explanation of those steps once the setup is out of internal testing.
  • Install Helm 3 on your machine. Run
    make upgrade-dev
    from project repo to deploy the release into the cluster.
  • Once you verified that the deployment from local machine works — feel free to fork the repo and set up the following secrets:

Secrets to be set

:warning: You don't have to use Quay as Docker image backend, in that case you can alter

Makefile
push commands to your own image hosting provider.
  • After you're done—a push to master will trigger a GitHub Action for deployment. See
    deploy.yaml
    for details.

:pray: Suggestions and PRs are welcome!

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.