🚝 multirail

Set up and configure multiple Rails app in the same DigitalOcean droplet / Linux (Ubuntu) server using nginx, Puma and Capistrano.

Stop paying hundreds of dollars to Heroku and similar services! Host multiple Rails apps in a single Linux server with ease and minimal config. All this with a free, ready-to-use SSL certificate! Multirail removes the frustration out of the process.

Designed to work with a DigitalOcean one-click Rails droplet.

[⚠️ WARNING]: This is still a really rough prototype – I only created it to help myself deploy my rails projects. I wouldn't recommend using it in real production environments cuz it might be buggy and not very secure and all that (even though I use it for my production apps) Please contribute to improve the project!

Pre-requisites

You need an already working Rails app, ready to deploy, and using the same Ruby version as your server. Also you need to be using a postgresql db and adapter.

You need a target production Linux server you can ssh into, with a user with nopasswd sudo access, and with the following stuff installed: ruby, rvm, bundler, certbot, nginx, psql, git (the recommended option is to spin up a Rails DigitalOcean droplet that has all this already installed). If you're using private Git repos, make sure your server has access to those via ssh (how-to in the following setup instructions).

Pre-requisites quick setup

1. Spin up a Rails DigitalOcean droplet using the Ruby on Rails one-click setup.

Sign up > Click the "Create" button on the upper right corner > Droplets > One-click apps > Select "Ruby-on-Rails on XX.XX" > Choose a droplet size > Choose a datacenter region > Create

Alternatively, and we highly recommend against this, set up your own Linux server with Ruby, rvm, Bundler, nginx, postgresql and all required packages using a guide like this one (not recommended, a DigitalOcean Rails one-click droplet removes all that hassle).

From now on, use your rails Linux user, not root.

2. Check your Rails project uses the exact same Ruby version as your server

Check the Gemfile on your Rails project and either change the Ruby version to the one on the server and verify everything still works or install that exact Ruby version on the server with rvm install x.y.z and then tell the server to use that version by default by running rvm use system x.y.z or rvm x.y.z --default

Also check you're using a postgresql adapter in your database.yml (in production, at least)

3. If you're using private Git repos: Make your server talk with your Git server via ssh. We'll assume you're using private GitHub repos:

   • Create a ssh key on your Linux server using the email you use to log in to GitHub.

   ssh-keygen -t rsa -b 4096 -C "your-github-email@email.com"

   • Make sure your ssh agent is running

   eval "$(ssh-agent -s)"

   • Add your new ssh key (assuming you didn't specify a custom key name, so the key name is the default id_rsa)

   ssh-add ~/.ssh/id_rsa

   • Show the public key content so you can copy it (again, assuming default key name id_rsa):

   cat ~/.ssh/id_rsa.pub

   • Go to GitHub.com > Settings > SSH and GPG keys > New SSH key. Add a human-readable title like "production server", paste in your public key you had just copied and hit "Add SSH key"

4. You'll need to configure nopasswd sudo access for the user you'll use for deploying (Capistrano 3 doesn't allow password prompting). Assuming your user is the default rails in DigitalOcean, execute: sudo nano /etc/sudoers And add rails ALL=(ALL) NOPASSWD:ALL right before #includedir /etc/sudoers.d

5. If you're getting weird LOCALE errors, you might need to stop accepting remote locale on the server by commenting out the AcceptEnv LANG LC_* line in the remote (server) /etc/ssh/sshd_config file.

6. Check bundler and the required gems are installed on the Linux server. Run gem to check. Run gem list bundler to check you have version >2 installed. If not, run gem install bundler.

7. Add x86_64-linux and ruby are added to the local Gemfile.lock, or the deployment script will complain and interrupt in the first deployment

bundle lock --add-platform x86_64-linux
bundle lock --add-platform ruby

Then commit and push

8. Make sure you're using puma ~> 4.0, this script doesn't work with versions of puma >=5 because we need to run the puma daemon, which is unsopported nowadays in favor of systemctl services

Usage

On your local development machine clone Multirail to the location you desire with git clone https://github.com/rameerez/multirail.git && cd multirail

Then just execute (substitute with your own values!):

./multirail -d example.com -i 123.123.123.123 -u rails -g git@github.com:rameerez/my-app.git -n my-app -f ~/git/my-app -v 2.4.0

Where:

• example.com is the domain you wish to deploy your app to.
• 123.123.123.123 the IP address of your server.
• rails is the Linux user in your server that will be used to deploy.
• git@github.com:rameerez/my-app.git is the Git SSH URL of your repository from which Capistrano will pull the code.
• my-app is just the name of your app (best to name it equal to your repo). [KNOWN BUG: due to a bug, please don't use any characters or spaces in the name. Just letters. Or else Postgresql won't create the user]
• ~/git/my-app is the path on your local machine where the Rails project is.
• 2.4.0 is the Ruby version both your project and your server are using.

For more info, execute ./multirail -h

---

You're done! Import your PostgreSQL data (if any) and you're ready to go! ✨ Now go to your domain and verify everything is working flawlessly and with a shiny 🔒 SSL cert!

Troubleshooting

If one of the servers is not working:

ssh into the server and check that puma is running

ps aux | grep puma

Then you should see at least one process that's running on an unix sock. It should be the same sock as configured in the deploy.rb script. Example output:

rails      848  0.0  6.5 803396 66228 ?        Ssl  May03   9:30 puma 3.12.0 (tcp://0.0.0.0:3000) [example]
rails     4683  0.0 13.0 864652 131388 ?       Sl   May14   5:13 puma 3.12.1 (unix:///home/rails/apps/app1/shared/tmp/sockets/app1-puma.sock)
rails    25487  3.6  9.2 847816 93244 ?        Sl   08:57   0:03 puma 3.12.0 (unix:///home/rails/apps/app2/shared/tmp/sockets/app2-puma.sock)
rails    25525  0.0  0.1  11464  1152 pts/0    S+   08:58   0:00 grep --color=auto puma

If the required app is not running, check if it has a valid pid (or any at all). cd to the app dir (cd ~/apps/APPNAME) and ls the pids folder in search of the puma.pid:

ls shared/tmp/pids/

If puma.pid is not in the output, the puma server is down.

From the local dev environment, cd to the rails project and spin it up with:

bundle exec cap production deploy:restart

This should create the puma.pid and ps aux | grep puma should now output our process. We may also try

bundle exec cap production puma:restart

if that's not enough.

To-do

• Don't create DB role and database if already created (from a failed installation, or from trying to redeploy)
• Revert unsuccessfull installations
• Completely remove an app from the server

How to uninstall sites

1. Delete the app /apps/THE_SITE
2. Delete Postgres databases

   sudo su - postgres
   psql
   \l
   DROP DATABASE THE_DATABASE;
   

3. See what's on nginx config ls -lah /etc/nginx/sites-enabled/
4. Remove the sites that are no longer present rm /etc/nginx/sites-enabled/THE_SITE
5. Restart nginx sudo service nginx restart
6. See what's on Letsencrypt sudo ls /etc/letsencrypt/live
7. Remove the site from Letsencrypt sudo rm -rf /etc/letsencrypt/live/THE_SITE