Tristan Dunn

Deploying Jekyll to a VPS


Part 2: Security, Monitoring, and Local Deployment

Continuing on part one of the series we’ll increase the server security by disabling password authentication for SSH, add Monit to oversee services, and deploy to the local Vagrant server with Capistrano.

SSH Security

It’s generally recommended that you disable password authentication for SSH to help prevent common brute force attacks. You could also whitelist IP addresses, but key authentication is reasonable enough for us.

We already created and authorized a deploy user in the previous article. If you’re already deploying remotely and using DigitalOcean you should have authorized your key for the root user, or received a root password. If you are using another provider, please ensure you have authorized your key or have a strong root password.

To disable password authentication we can add the sshd cookbook and customize SSH options in the node file. We just need to add it to our Cheffile and install it with bundle exec librarian-chef install.

site ""

cookbook "nginx"
cookbook "rbenv"
cookbook "sshd"
Adding the sshd cookbook dependency the Cheffile.

Now we can add the recipe to the run list and disable password authentication. Note that I’m excluding the other settings from the previous articles.

  "run_list" : [

  "sshd" : {
    "sshd_config" : {
      "PasswordAuthentication" : "no"
Adding attributes to nodes/vagrant.json for the sshd recipe and settings.

And we can run the new recipe with the standard bundle exec knife solo cook vagrant process.

We can confirm that password authentication is not allowed by attempting to authentication with a password using ssh deploy:@vagrant. If all went as planned we should see a Permission denied (publickey). message.


While a Jekyll website may not be mission critical, monitoring is still important to keep it functioning when the unexpected happens. We’re going to use Monit for basic monitoring, a simple and popular solution for system monitoring and error recovery.

We can of course use an existing cookbook, the monit-ng cookbook, to add some basic checks. To start add it to the Cheffile and install it with bundle exec librarian-chef install.

site ""

cookbook "monit-ng"
cookbook "nginx"
cookbook "rbenv"
cookbook "sshd"
Adding the monit-ng cookbook dependency the Cheffile.

First we need to define our custom cookbook name and dependency.

name    "monit"
depends "monit-ng"
Creating the site-cookbooks/monit/metadata.rb file.

And the default recipe will define basic process ID checks for the nginx and sshd services.

include_recipe "monit-ng::default"

monit_check "nginx" do
  stop     "/etc/init.d/nginx stop"
  start    "/etc/init.d/nginx start"
  check_id "/var/run/"

monit_check "sshd" do
  stop     "/etc/init.d/ssh stop"
  start    "/etc/init.d/ssh start"
  check_id "/var/run/"
Defining checks in the site-cookbooks/monit/recipes/default.rb file.

And we need to add our new recipe to the run list. We’re running it last to ensure the processes we’re monitoring are available.

  "run_list" : [
Adding monit recipe to nodes/vagrant.json.

After running the recurring bundle exec knife solo cook vagrant command to install, we can double check that it’s monitoring properly. Just SSH into the server to stop the web server with sudo /etc/init.d/nginx stop and it should restart automatically within about 30 seconds. The current status of monitored services are available by running sudo monit status.

Deploying to Vagrant

We of course need a Jekyll website to be able to deploy. If you don’t already have one, you can generate one by running jekyll new jekyll-vps-website, with the last argument being whatever name you would like.

Next we need to install Capistrano and a couple of dependencies. We also add jekyll to install it on the server and therubyracer for a JavaScript environment. At the time of writing Jekyll requires a JS environment for the CoffeeScript dependency, but it will no longer be a required dependency in the future.

gem "jekyll",       "2.5.3"
gem "therubyracer", "0.12.2"

group :development do
  gem "capistrano",         "3.4.0"
  gem "capistrano-bundler", "1.1.4"
  gem "capistrano-rbenv",   "2.0.3"
Adding Capistrano dependencies Gemfile.

After running bundle install we can generate the Capistrano structure, including our local stage, with bundle exec cap install STAGES=local.

Our other dependencies aren’t included by default, so we’ll require them in the Capfile.

require "capistrano/setup"
require "capistrano/deploy"

require "capistrano/bundler"
require "capistrano/rbenv"
Setting up the base Capfile for Capistrano.

We should also exclude some files and folders from the Jekyll output to prevent them from being publicly accessible in the future.

# ...

  - Capfile
  - Gemfile
  - Gemfile.lock
  - config
Adding exclusions to the _config.yml for Jekyll.

Now we can customize the config/deploy.rb file with our custom settings and actions for building and deploying the website. It’s a decent chunk of code, so I explain each section in comments.

# Lock the Capistrano version to ensure we're running the version we expect.
lock "3.4.0"

# Application name and deployment location.
# The repository URL is not used locally, so no need to change it yet. The
# deployment location and application name are from the name used in part one
# of the series, so be sure to update if you used a different name.
set :repo_url,    ""
set :deploy_to,   "/var/www/"
set :application, "example"

# Ensure bundler runs for the web role.
set :bundle_roles, :web

# Location and settings for rbenv environment.
set :rbenv_type,        :system
set :rbenv_ruby,        "2.1.5"
set :rbenv_roles,       :all
set :rbenv_map_bins,    %w(bundle gem rake ruby)
set :rbenv_custom_path, "/opt/rbenv"

# Don't keep any previous releases.
set :keep_releases, 1

# Avoid UTF-8 issues when building Jekyll.
set :default_env, { "LC_ALL" => "en_US.UTF-8" }

# Define a custom Jekyll build task and run it before publishing the website. It
# allows for a custom configuration setting per environment, which is helpful
# for customizing settings in production.
namespace :deploy do
  desc "Build the website with Jekyll"
  task :build do
    on roles(:web) do
      within release_path do
        execute :bundle, "exec", "jekyll", "build", "--config",
          fetch(:configuration, "_config.yml")

  before :publishing, :build

# Don't log revisions.
Adding the core settings, actions, and customization in config/deploy.rb.

Instead of having to commit to a branch, push to a remote, and then deploy a branch on a local server we’re just going to package and upload the directory content to the local server. It allows you to test changes in a “production” environment much faster. To do so we need to define a custom strategy. It’s a rather large class so I’ve commented the code heavily.

module FileStrategy
  # Pretend we don't have a repository cache.
  def test

  # Ensure the repository path exists.
  def check
    context.execute :mkdir, "-p", repo_path

  # Pretend we've cloned the repository.
  def clone

  # Create and upload a package of the local directory as an update.
  def update
    # Ensure a local `tmp` directory exists.
    `mkdir -p #{File.dirname(path)}`

    # Package the local directory, ignoring unnecessary files.
    `tar -zcf #{path} --exclude .git --exclude _site --exclude tmp .`

    # Upload the package to the server.
    context.upload! path, "/#{path}"

    # Remove the package locally.
    `rm #{path}`

  # Extract the uploaded package to the release path and remove.
  def release
    # Ensure the release directory exists on the server.
    context.execute :mkdir, "-p", release_path

    # Extract the uploaded package to the release directory.
    context.execute :tar, "-xmf" "/#{path}", "-C", release_path

    # Remove the package from the server.
    context.execute :rm, "/#{path}"

  # Use the latest repository SHA as the revision.
  def fetch_revision
    `git log --pretty=format:'%h' -n 1 HEAD`


  # Helper method for the directory package path.
  def path
Creating a custom deployment strategy in config/deploy/local/file_strategy.rb.

Lastly we’ll update our local stage to define the server, use the file strategy for deployment, and optionally include custom Jekyll configuration.

# Require our custom deployment strategy.
require "./config/deploy/local/file_strategy"

# Define a web server, where "vagrant" is our local SSH host and "deploy" is our
# server user created in part one.
server "vagrant", user: "deploy", roles: %w(web)

# Set our custom strategy.
set :git_strategy, FileStrategy

# Optionally define custom configuration files, where the staging version will
# overwrite the global version.
# set :configuration, "_config.yml,_config_staging.yml"
Defining the host, deployment strategy, and custom configuration in config/deploy/local.rb.

We should now be able to deploy by running cap local deploy. It will take a minute the first time as it needs to install dependencies. After the website generation completes you should see your website at localhost:8080.


We now how the minimal components needed to deploy a Jekyll website to a Vagrant box. See the jekyll-vps-server repository for the complete Chef source code, with the part-2 branch being specific to this article. The website source code is available in the jekyll-vps-website repository, with the part-2 branch being relevant.

In the next part we’ll create and deploy to a DigitalOcean server to have a production version available. E-mail me if you have any tips, comments, or questions.