Asset Sync

Synchronises Assets between Rails and S3.

Asset Sync is built to run with the new Rails Asset Pipeline feature introduced in Rails 3.1. After you run bundle exec rake assets:precompile your assets will be synchronised to your S3 bucket, optionally deleting unused files and only uploading the files it needs to.

This was initially built and is intended to work on Heroku but can work on any platform.

Upgraded from 1.x? Read UPGRADING.md

Since 2.x, Asset Sync depends on gem fog-core instead of fog.
This is due to fog is including many unused storage provider gems as its dependencies.

Asset Sync has no idea about what provider will be used,
so you are responsible for bundling the right gem for the provider to be used.

In your Gemfile:

gem "asset_sync"
gem "fog-aws"

Or, to use Azure Blob storage, configure as this.

gem "asset_sync"
gem "gitlab-fog-azure-rm"

# This gem seems unmaintianed
# gem "fog-azure-rm"

To use Backblaze B2, insert these.

gem "asset_sync"
gem "fog-backblaze"

It's possible to improve asset:precompile time if you are using Rails 3.2.x the main source of which being compilation of non-digest assets.

turbo-sprockets-rails3 solves this by only compiling digest assets. Thus cutting compile time in half.

NOTE: It will be deprecated in Rails 4 as sprockets-rails has been extracted out of Rails and will only compile digest assets by default.

Configure config/environments/production.rb to use Amazon S3 as the asset host and ensure precompiling is enabled.

  #config/environments/production.rb
  config.action_controller.asset_host = "//#{ENV['FOG_DIRECTORY']}.s3.amazonaws.com"

Or, to use Google Storage Cloud, configure as this.

  #config/environments/production.rb
  config.action_controller.asset_host = "//#{ENV['FOG_DIRECTORY']}.storage.googleapis.com"

Or, to use Azure Blob storage, configure as this.

  #config/environments/production.rb
  config.action_controller.asset_host = "//#{ENV['AZURE_STORAGE_ACCOUNT_NAME']}.blob.core.windows.net/#{ENV['FOG_DIRECTORY']}"

Or, to use Backblaze B2, configure as this.

  #config/environments/production.rb
  config.action_controller.asset_host = "//f000.backblazeb2.com/file/#{ENV['FOG_DIRECTORY']}"

On HTTPS: the exclusion of any protocol in the asset host declaration above will allow browsers to choose the transport mechanism on the fly. So if your application is available under both HTTP and HTTPS the assets will be served to match.

The only caveat with this is that your S3 bucket name must not contain any periods so, mydomain.com.s3.amazonaws.com for example would not work under HTTPS as SSL certificates from Amazon would interpret our bucket name as not a subdomain of s3.amazonaws.com, but a multi level subdomain. To avoid this don't use a period in your subdomain or switch to the other style of S3 URL.

  config.action_controller.asset_host = "//s3.amazonaws.com/#{ENV['FOG_DIRECTORY']}"

Or, to use Google Storage Cloud, configure as this.

  config.action_controller.asset_host = "//storage.googleapis.com/#{ENV['FOG_DIRECTORY']}"

Or, to use Azure Blob storage, configure as this.

  #config/environments/production.rb
  config.action_controller.asset_host = "//#{ENV['AZURE_STORAGE_ACCOUNT_NAME']}.blob.core.windows.net/#{ENV['FOG_DIRECTORY']}"

On non default S3 bucket region: If your bucket is set to a region that is not the default US Standard (us-east-1) you must use the first style of url //#{ENV['FOG_DIRECTORY']}.s3.amazonaws.com or amazon will return a 301 permanently moved when assets are requested. Note the caveat above about bucket names and periods.

If you wish to have your assets sync to a sub-folder of your bucket instead of into the root add the following to your production.rb file

  # store assets in a 'folder' instead of bucket root
  config.assets.prefix = "/production/assets"

Also, ensure the following are defined (in production.rb or application.rb)

• config.assets.digest is set to true.
• config.assets.enabled is set to true.

Additionally, if you depend on any configuration that is setup in your initializers you will need to ensure that

• config.assets.initialize_on_precompile is set to true

AssetSync supports the following methods of configuration.

• Built-in Initializer (configured through environment variables)
• Rails Initializer
• A YAML config file

Using the Built-in Initializer is the default method and is supposed to be used with environment variables. It's the recommended approach for deployments on Heroku.

If you need more control over configuration you will want to use a custom rails initializer.

Configuration using a YAML file (a common strategy for Capistrano deployments) is also supported.

The recommend way to configure asset_sync is by using environment variables however it's up to you, it will work fine if you hard code them too. The main reason why using environment variables is recommended is so your access keys are not checked into version control.

The Built-in Initializer will configure AssetSync based on the contents of your environment variables.

Add your configuration details to heroku

heroku config:add AWS_ACCESS_KEY_ID=xxxx
heroku config:add AWS_SECRET_ACCESS_KEY=xxxx
heroku config:add FOG_DIRECTORY=xxxx
heroku config:add FOG_PROVIDER=AWS
# and optionally:
heroku config:add FOG_REGION=eu-west-1
heroku config:add ASSET_SYNC_COMPRESSION=gz
heroku config:add ASSET_SYNC_MANIFEST=true
heroku config:add ASSET_SYNC_EXISTING_REMOTE_FILES=keep

Or add to a traditional unix system

export AWS_ACCESS_KEY_ID=xxxx
export AWS_SECRET_ACCESS_KEY=xxxx
export FOG_DIRECTORY=xxxx

Rackspace configuration is also supported

heroku config:add RACKSPACE_USERNAME=xxxx
heroku config:add RACKSPACE_API_KEY=xxxx
heroku config:add FOG_DIRECTORY=xxxx
heroku config:add FOG_PROVIDER=Rackspace

Google Storage Cloud configuration is supported as well. The preferred option is using the GCS JSON API which requires that you create an appropriate service account, generate the signatures and make them accessible to asset sync at the prescribed location

heroku config:add FOG_PROVIDER=Google
heroku config:add GOOGLE_PROJECT=xxxx
heroku config:add GOOGLE_JSON_KEY_LOCATION=xxxx
heroku config:add FOG_DIRECTORY=xxxx

If using the S3 API the following config is required

heroku config:add FOG_PROVIDER=Google
heroku config:add GOOGLE_STORAGE_ACCESS_KEY_ID=xxxx
heroku config:add GOOGLE_STORAGE_SECRET_ACCESS_KEY=xxxx
heroku config:add FOG_DIRECTORY=xxxx

The Built-in Initializer also sets the AssetSync default for existing_remote_files to keep.

If you want to enable some of the advanced configuration options you will want to create your own initializer.

Run the included Rake task to generate a starting point.

rails g asset_sync:install --provider=Rackspace
rails g asset_sync:install --provider=AWS
rails g asset_sync:install --provider=AzureRM
rails g asset_sync:install --provider=Backblaze

The generator will create a Rails initializer at config/initializers/asset_sync.rb.

AssetSync.configure do |config|
  config.fog_provider = 'AWS'
  config.fog_directory = ENV['FOG_DIRECTORY']
  config.aws_access_key_id = ENV['AWS_ACCESS_KEY_ID']
  config.aws_secret_access_key = ENV['AWS_SECRET_ACCESS_KEY']
  config.aws_session_token = ENV['AWS_SESSION_TOKEN'] if ENV.key?('AWS_SESSION_TOKEN')

  # Don't delete files from the store
  # config.existing_remote_files = 'keep'
  #
  # Increase upload performance by configuring your region
  # config.fog_region = 'eu-west-1'
  #
  # Set `public` option when uploading file depending on value,
  # Setting to "default" makes asset sync skip setting the option
  # Possible values: true, false, "default" (default: true)
  # config.fog_public = true
  #
  # Change AWS signature version. Default is 4
  # config.aws_signature_version = 4
  #
  # Change canned ACL of uploaded object. Default is unset. Will override fog_public if set.
  # Choose from: private | public-read | public-read-write | aws-exec-read |
  #              authenticated-read | bucket-owner-read | bucket-owner-full-control
  # config.aws_acl = nil
  #
  # Change host option in fog (only if you need to)
  # config.fog_host = 's3.amazonaws.com'
  #
  # Change port option in fog (only if you need to)
  # config.fog_port = "9000"
  #
  # Use http instead of https.
  # config.fog_scheme = 'http'
  #
  # Extra fog options.
  # Overrides any existing value (even those set by AssetSync)
  # config.fog_options = {}
  #
  # Automatically replace files with their equivalent gzip/brotli compressed version
  # config.compression = 'gz'
  #
  # Use the Rails generated 'manifest.yml' file to produce the list of files to
  # upload instead of searching the assets directory.
  # config.manifest = true
  #
  # Upload the manifest file also.
  # config.include_manifest = false
  #
  # Upload files concurrently
  # config.concurrent_uploads = false
  #
  # Number of threads when concurrent_uploads is enabled
  # config.concurrent_uploads_max_threads = 10
  #
  # Path to cache file to skip scanning remote
  # config.remote_file_list_cache_file_path = './.asset_sync_remote_file_list_cache.json'
  #
  # Path on remote storage to persist remote file list file
  # config.remote_file_list_remote_path = '/remote/asset_sync_remote_file.json'
  #
  # Fail silently.  Useful for environments such as Heroku
  # config.fail_silently = true
  #
  # Log silently. Default is `true`. But you can set it to false if more logging message are preferred.
  # Logging messages are sent to `STDOUT` when `log_silently` is falsy
  # config.log_silently = true
  #
  # Allow custom assets to be cacheable. Note: The base filename will be matched
  # If you have an asset with name `app.0b1a4cd3.js`, only `app.0b1a4cd3` will need to be matched
  # only one of `cache_asset_regexp` or `cache_asset_regexps` is allowed.
  # config.cache_asset_regexp = /\.[a-f0-9]{8}$/i
  # config.cache_asset_regexps = [ /\.[a-f0-9]{8}$/i, /\.[a-f0-9]{20}$/i ]
end

Run the included Rake task to generate a starting point.

rails g asset_sync:install --use-yml --provider=Rackspace
rails g asset_sync:install --use-yml --provider=AWS
rails g asset_sync:install --use-yml --provider=AzureRM
rails g asset_sync:install --use-yml --provider=Backblaze

The generator will create a YAML file at config/asset_sync.yml.

defaults: &defaults
  fog_provider: "AWS"
  fog_directory: "rails-app-assets"
  aws_access_key_id: "<%= ENV['AWS_ACCESS_KEY_ID'] %>"
  aws_secret_access_key: "<%= ENV['AWS_SECRET_ACCESS_KEY'] %>"

  # To use AWS reduced redundancy storage.
  # aws_reduced_redundancy: true
  #
  # You may need to specify what region your storage bucket is in
  # fog_region: "eu-west-1"
  #
  # Change AWS signature version. Default is 4
  # aws_signature_version: 4
  #
  # Change canned ACL of uploaded object. Default is unset. Will override fog_public if set.
  # Choose from: private | public-read | public-read-write | aws-exec-read |
  #              authenticated-read | bucket-owner-read | bucket-owner-full-control
  # aws_acl: null
  #
  # Change host option in fog (only if you need to)
  # fog_host: "s3.amazonaws.com"
  #
  # Use http instead of https. Default should be "https" (at least for fog-aws)
  # fog_scheme: "http"

  existing_remote_files: keep # Existing pre-compiled assets on S3 will be kept
  # To delete existing remote files.
  # existing_remote_files: delete
  # To ignore existing remote files and overwrite.
  # existing_remote_files: ignore
  # Automatically replace files with their equivalent gzip/brotli compressed version
  # compression: "gz"
  # Fail silently.  Useful for environments such as Heroku
  # fail_silently: true
  # Always upload. Useful if you want to overwrite specific remote assets regardless of their existence
  #  eg: Static files in public often reference non-fingerprinted application.css
  #  note: You will still need to expire them from the CDN's edge cache locations
  # always_upload: ['application.js', 'application.css', !ruby/regexp '/application-/\d{32}\.css/']
  # Ignored files. Useful if there are some files that are created dynamically on the server and you don't want to upload on deploy.
  # ignored_files: ['ignore_me.js', !ruby/regexp '/ignore_some/\d{32}\.css/']
  # Public path. To change the public directory.
  # public_path: 'public'
  # Prefix. To change the directory to sync relative to public_path.
  # prefix: 'assets'
  # Allow custom assets to be cacheable. Note: The base filename will be matched
  # If you have an asset with name "app.0b1a4cd3.js", only "app.0b1a4cd3" will need to be matched
  # cache_asset_regexps: ['cache_me.js', !ruby/regexp '/cache_some\.\d{8}\.css/']

development:
  <<: *defaults

test:
  <<: *defaults

production:
  <<: *defaults

Most AssetSync configuration can be modified directly using environment variables with the Built-in initializer. e.g.

AssetSync.config.fog_provider == ENV['FOG_PROVIDER']

Simply upcase the ruby attribute names to get the equivalent environment variable to set. The only exception to that rule are the internal AssetSync config variables, they must be prepended with ASSET_SYNC_* e.g.

AssetSync.config.compression == ENV['ASSET_SYNC_COMPRESSION']

• existing_remote_files: ('keep', 'delete', 'ignore') what to do with previously precompiled files. default: 'keep'
• gzip_compression: (deprecated, can still be used but use compression when possible)
• compression: ('gz', 'br') when enabled (non nil value set), will automatically replace files that have a gzip/brotli compressed equivalent with the compressed version. default: nil
• manifest: (true, false) when enabled, will use the manifest.yml generated by Rails to get the list of local files to upload. experimental. default: 'false'
• include_manifest: (true, false) when enabled, will upload the manifest.yml generated by Rails. default: 'false'
• concurrent_uploads: (true, false) when enabled, will upload the files in different Threads, this greatly improves the upload speed. default: 'false'
• concurrent_uploads_max_threads: when concurrent_uploads is enabled, this determines the number of threads that will be created. default: 10
• remote_file_list_cache_file_path: if present, use this path to cache remote file list to skip scanning remote default: nil
• remote_file_list_remote_path: if present, use this path to download remote file list file to cache file list in local to skip scanning remote. useful in container environment where you cannot maintain the local file, remote_file_list_cache_file_path also needed to make use of this option default: nil
• enabled: (true, false) when false, will disable asset sync. default: 'true' (enabled)
• ignored_files: an array of files to ignore e.g. ['ignore_me.js', %r(ignore_some/\d{32}\.css)] Useful if there are some files that are created dynamically on the server and you don't want to upload on deploy default: []
• cache_asset_regexps: an array of files to add cache headers e.g. ['cache_me.js', %r(cache_some\.\d{8}\.css)] Useful if there are some files that are added to sprockets assets list and need to be set as 'Cacheable' on uploaded server. Only rails compiled regexp is matched internally default: []

Adding local files by providing a block:

AssetSync.configure do |config|
  # The block should return an array of file paths
  config.add_local_file_paths do
    # Any code that returns paths of local asset files to be uploaded
    # Like Webpacker
    public_root = Rails.root.join("public")
    Dir.chdir(public_root) do
      packs_dir = Webpacker.config.public_output_path.relative_path_from(public_root)
      Dir[File.join(packs_dir, '/**/**')]
    end
  end
end

The blocks are run when local files are being scanned and uploaded

It's reported that mime-types 3.x returns application/ecmascript instead of application/javascript
Such change of mime type might cause some CDN to disable asset compression
So this gem has defined a default override for file ext js to be mapped to application/javascript by default

To customize the overrides:

AssetSync.configure do |config|
  # Clear the default overrides
  config.file_ext_to_mime_type_overrides.clear

  # Add/Edit overrides
  # Will call `#to_s` for inputs
  config.file_ext_to_mime_type_overrides.add(:js, :"application/x-javascript")
end

The blocks are run when local files are being scanned and uploaded

• fog_provider: your storage provider AWS (S3) or Rackspace (Cloud Files) or Google (Google Storage) or AzureRM (Azure Blob) or Backblaze (Backblaze B2)
• fog_directory: your bucket name

• fog_region: the region your storage bucket is in e.g. eu-west-1 (AWS), ord (Rackspace), japanwest (Azure Blob)
• fog_path_style: To use buckets with dot in names, check fog/fog#2381 (comment)
• fog_options: For extra fog options. Overrides any existing value (even those set by AssetSync)

• aws_access_key_id: your Amazon S3 access key
• aws_secret_access_key: your Amazon S3 access secret
• aws_acl: set canned ACL of uploaded object, will override fog_public if set

• rackspace_username: your Rackspace username
• rackspace_api_key: your Rackspace API Key.

When using the JSON API

• google_project: your Google Cloud Project name where the Google Cloud Storage bucket resides
• google_json_key_location: path to the location of the service account key. The service account key must be a JSON type key

When using the S3 API

• google_storage_access_key_id: your Google Storage access key
• google_storage_secret_access_key: your Google Storage access secret

• azure_storage_account_name: your Azure Blob access key
• azure_storage_access_key: your Azure Blob access secret

• b2_key_id: Your Backblaze B2 key ID
• b2_key_token: Your Backblaze B2 key token
• b2_bucket_id: Your Backblaze B2 bucket ID

• rackspace_auth_url: Rackspace auth URL, for Rackspace London use: https://lon.identity.api.rackspacecloud.com/v2.0

If you are using anything other than the US buckets with S3 then you'll want to set the region. For example with an EU bucket you could set the following environment variable.

heroku config:add FOG_REGION=eu-west-1

Or via a custom initializer

AssetSync.configure do |config|
  # ...
  config.fog_region = 'eu-west-1'
end

Or via YAML

production:
  # ...
  fog_region: 'eu-west-1'

Amazon has switched to the more secure IAM User security policy model. When generating a user & policy for asset_sync you must ensure the policy has the following permissions, or you'll see the error:

Expected(200) <=> Actual(403 Forbidden)

IAM User Policy Example with minimum require permissions (replace bucket_name with your bucket):

{
  "Statement": [
    {
      "Action": "s3:ListBucket",
      "Effect": "Allow",
      "Resource": "arn:aws:s3:::bucket_name"
    },
    {
      "Action": "s3:PutObject*",
      "Effect": "Allow",
      "Resource": "arn:aws:s3:::bucket_name/*"
    }
  ]
}

If you want to use IAM roles you must set config.aws_iam_roles = true in your initializers.

AssetSync.configure do |config|
  # ...
  config.aws_iam_roles = true
end

With the compression option enabled, when uploading your assets. If a file has a gzip/brotli (only 1 of them, depending on the option value set) compressed equivalent we will replace that asset with the compressed version and sets the correct headers for S3 to serve it. For example, if you have a file master.css and it was compressed to master.css.gz we will upload the .gz file to S3 in place of the uncompressed file.

If the compressed file is actually larger than the uncompressed file we will ignore this rule and upload the standard uncompressed version.

Option compression supported values:

• nil => Disabled
• 'gz' => Only consider gzip compressed one
• 'br' => Only consider brotli compressed one

With the fail_silently option enabled, when running rake assets:precompile AssetSync will never throw an error due to missing configuration variables.

With the new user_env_compile feature of Heroku (see above), this is no longer required or recommended. Yet was added for the following reasons:

With Rails 3.1 on the Heroku cedar stack, the deployment process automatically runs rake assets:precompile. If you are using ENV variable style configuration. Due to the methods with which Heroku compile slugs, there will be an error raised by asset_sync as the environment is not available. This causes heroku to install the rails31_enable_runtime_asset_compilation plugin which is not necessary when using asset_sync and also massively slows down the first incoming requests to your app.

To prevent this part of the deploy from failing (asset_sync raising a config error), but carry on as normal set fail_silently to true in your configuration and ensure to run heroku run rake assets:precompile after deploy.

A rake task is included within the asset_sync gem to perform the sync:

  namespace :assets do
    desc "Synchronize assets to S3"
    task :sync => :environment do
      AssetSync.sync
    end
  end

If AssetSync.config.run_on_precompile is true (default), then assets will be uploaded to S3 automatically after the assets:precompile rake task is invoked:

  if Rake::Task.task_defined?("assets:precompile:nondigest")
    Rake::Task["assets:precompile:nondigest"].enhance do
      Rake::Task["assets:sync"].invoke if defined?(AssetSync) && AssetSync.config.run_on_precompile
    end
  else
    Rake::Task["assets:precompile"].enhance do
      Rake::Task["assets:sync"].invoke if defined?(AssetSync) && AssetSync.config.run_on_precompile
    end
  end

You can disable this behavior by setting AssetSync.config.run_on_precompile = false.

You can use the gem with any Rack application, but you must specify two additional options; prefix and public_path.

AssetSync.configure do |config|
  config.fog_provider = 'AWS'
  config.fog_directory = ENV['FOG_DIRECTORY']
  config.aws_access_key_id = ENV['AWS_ACCESS_KEY_ID']
  config.aws_secret_access_key = ENV['AWS_SECRET_ACCESS_KEY']
  config.prefix = 'assets'
  # Can be a `Pathname` or `String`
  # Will be converted into an `Pathname`
  # If relative, will be converted into an absolute path
  # via `::Rails.root` or `::Dir.pwd`
  config.public_path = Pathname('./public')
end

Then manually call AssetSync.sync at the end of your asset precompilation task.

namespace :assets do
  desc 'Precompile assets'
  task :precompile do
    target = Pathname('./public/assets')
    manifest = Sprockets::Manifest.new(sprockets, './public/assets/manifest.json')

    sprockets.each_logical_path do |logical_path|
      if (!File.extname(logical_path).in?(['.js', '.css']) || logical_path =~ /application\.(css|js)$/) && asset = sprockets.find_asset(logical_path)
        filename = target.join(logical_path)
        FileUtils.mkpath(filename.dirname)
        puts "Write asset: #{filename}"
        asset.write_to(filename)
        manifest.compile(logical_path)
      end
    end

    AssetSync.sync
  end
end

1. Add webpacker files and disable run_on_precompile:

AssetSync.configure do |config|
  # Disable automatic run on precompile in order to attach to webpacker rake task
  config.run_on_precompile = false
  # The block should return an array of file paths
  config.add_local_file_paths do
    # Support webpacker assets
    public_root = Rails.root.join("public")
    Dir.chdir(public_root) do
      packs_dir = Webpacker.config.public_output_path.relative_path_from(public_root)
      Dir[File.join(packs_dir, '/**/**')]
    end
  end
end

2. Add a asset_sync.rake in your lib/tasks directory that enhances the correct task, otherwise asset_sync runs before webpacker:compile does:

if defined?(AssetSync)
  Rake::Task['webpacker:compile'].enhance do
    Rake::Task["assets:sync"].invoke
  end
end

By adding local files outside the normal Rails assets directory, the uploading part works, however checking that the asset was previously uploaded is not working because asset_sync is only fetching the files in the assets directory on the remote bucket. This will mean additional time used to upload the same assets again on every precompilation.

Make sure you have a .env file with these details:-

# for AWS provider
AWS_ACCESS_KEY_ID=<yourkeyid>
AWS_SECRET_ACCESS_KEY=<yoursecretkey>
FOG_DIRECTORY=<yourbucket>
FOG_REGION=<youbucketregion>

# for AzureRM provider
AZURE_STORAGE_ACCOUNT_NAME=<youraccountname>
AZURE_STORAGE_ACCESS_KEY=<youraccesskey>
FOG_DIRECTORY=<yourcontainer>
FOG_REGION=<yourcontainerregion>

Make sure the bucket has read/write permissions. Then to run the tests:-

foreman run rake

1. Add some before and after filters for deleting and uploading
2. Support more cloud storage providers
3. Better test coverage
4. Add rake tasks to clean old assets from a bucket

Inspired by:

• https://github.com/moocode/asset_id
• https://gist.github.com/1053855

MIT License. Copyright 2011-2013 Rumble Labs Ltd. rumblelabs.com