cron.rake

task :cron => :environment do
  DatabaseBackup.back_up_to_s3!
end

database_backup.rb

require 'open-uri'
require 'heroku'
require 'heroku/command'
require 'heroku/command/auth'
require 'heroku/command/pgbackups'

class DatabaseBackup
  def self.back_up_to_s3!
    app_name = ENV.fetch('HEROKU_APP_NAME')
    Heroku::Command.run 'pgbackups:capture', ['--expire', '--app', app_name]
    url = capture_stdout do
      Heroku::Command.run 'pgbackups:url', ['--app', app_name]
    end

    AWS::S3::Base.establish_connection!(
      :access_key_id => ENV.fetch('AMAZON_ACCESS_KEY_ID'),
      :secret_access_key => ENV.fetch('AMAZON_SECRET_ACCESS_KEY'))
    url.gsub! /^"|"$/, ''
    AWS::S3::S3Object.store(
      "pgdb-#{Time.now.strftime('%y%m%d_%H%M')}.dump",
      open(url),
      ENV.fetch('S3_BACKUP_BUCKET'))
  end

  def self.capture_stdout
    out = StringIO.new
    $stdout = out
    yield
    return out.string
  ensure
    $stdout = STDOUT
  end
end

class Heroku::Auth
  def self.client
    Heroku::Client.new ENV.fetch('HEROKU_LOGIN'), ENV.fetch('HEROKU_PASSWORD')
  end
end

Is AWS required by Heroku?

-- EDIT

Just realized by question was horribly vague. I noticed you didn't require the aws gem. Does Heroku do it for you, or was it required by some other component of your system?

For me with the 2.1.4 version of the heroku gem, I had to require it aws-s3 gem.

Adding: gem 'aws-s3', '0.6.2', :require => 'aws/s3' to my Gemfile did the trick, although I am still on the fence about the require in the Gemfile instead of just in the actual database class.

Going forward on Heroku's cedar stack, I am sure you will be on your own for this gem either way.

The AWS gem is required by Bundler. I only explicitly require the Heroku gem because I had to for some reason, even though it was in my Gemfile. It seems that requiring 'heroku' doesn't also require those other bits.