Dynamically change active storage service on rails model?

Question

I'm using active storage with s3 to attach a file to a model called Document. I need to add support for users in the EU that want their document files stored in an s3 bucket in the EU.

I have my storage.yml configured like:

amazon:
    service: S3
    access_key_id: <%= ENV['S3_ACCESS_KEY_ID'] %>
    secret_access_key: <%= ENV['S3_SECRET_KEY_ACCESS'] %>
    region: <%= ENV['S3_REGION_EU'] %>
    bucket: <%= ENV['S3_BUCKET_NAME'] %>

amazon_eu:
    service: S3
    access_key_id: <%= ENV['S3_ACCESS_KEY_ID'] %>
    secret_access_key: <%= ENV['S3_SECRET_KEY_ACCESS'] %>
    region: <%= ENV['S3_REGION_EU'] %>
    bucket: <%= ENV['S3_BUCKET_NAME_EU'] %>

Is there a way to dynamically switch in the service in the Document model based on the region set in the Account? Something like:

class Document < ApplicationRecord
  belongs_to :account
  
  if account.region == 'eu'
    has_one_attached :file, service: amazon_eu
  else
    has_one_attached :file, service: amazon
  end

end

Or really anyway to specify the specific service that I want to use dynamically at runtime?

Answer 1

I figured out how to do it. I created a custom service for active storage that inherits from the s3 service and created a second bucket instance on the class for the eu region I want to use.

# lib/active_storage/service/dynamic_storage_service.rb
require "active_storage/service/s3_service"

module ActiveStorage
  class Service::DynamicStorageService < ActiveStorage::Service::S3Service
    # create attributes for another client and bucket
    attr_reader :client_eu, :bucket_eu
    

    # override the initializer with options to pass in an eu bucket from the storage.yml
    # this is where you would create whatever extra buckets you need
    def initialize(bucket:, bucket_eu:, upload: {}, public: false, **options)
      eu_options = options.except(:region)
      eu_options[:region] = eu_options[:region_eu]

      @client_eu = Aws::S3::Resource.new(eu_options.except(:region_eu))
      @bucket_eu = @client_eu.bucket(bucket_eu)

      super(bucket: bucket, upload: upload, public: public, **options.except(:region_eu))
    end

    private
      # override the method where the bucket is used
      # this is where you would add whatever logic you need to select the bucket
      # this is the implementation that works for me
      # the method just needs to return the S3 object and everything else will work
      def object_for(key)
        # this is how you get the record based on the key
        document_id = ActiveStorage::Attachment.find_by(blob_id: ActiveStorage::Blob.find_by(key: key).id).record_id

        document = Document.find(document_id)

        if document.account.region == 'eu'
          return bucket_eu.object(key)
        else
          return bucket.object(key)
        end
      end
  end
end

And then you would use it in the storage.yml instead of the regular S3 service:

# config/storage.yml
amazon:
    service: DynamicStorage
    access_key_id: <%= ENV['S3_ACCESS_KEY_ID'] %>
    secret_access_key: <%= ENV['S3_SECRET_KEY_ACCESS'] %>
    region: <%= ENV['S3_REGION'] %>
    bucket: <%= ENV['S3_BUCKET_NAME'] %>
    region_eu: <%= ENV['S3_REGION_EU'] %>
    bucket_eu: <%= ENV['S3_BUCKET_NAME_EU'] %>

And the Document.rb can work like normal

# document.rb
class Document < ApplicationRecord
  belongs_to :account
  
  has_one_attached :file

end

You could probably rewrite this better and make it more dynamic and able to handle as many buckets as you want but this is what worked for me