How can I migrate CarrierWave files to a new storage mechanism?

Question

I have a Ruby on Rails site with models using CarrierWave for file handling, currently using local storage. I want to start using cloud storage and I need to migrate existing local files to the cloud. I am wondering if anyone can point out a method for doing this?

Bonus points for using a model attribute that would allow me to do this row-by-row in the background without interrupting my site for extended downtime (in other words, some model rows would still have local storage while others used cloud storage).

My first instinct is to create a new uploader for each model that uses cloud storage, so I have two uploaders on each model, then transferring the files from one to the other, setting an attribute to indicate which file should be used until they are all transferred, then removing the old uploader. That seems a little excessive.

Answer 1

I'd try the following steps:

1. Change the storage in the uploaders to :fog or what ever you want to use
2. Write a migration like rails g migration MigrateFiles to let carrierwave get the current files, process them and upload them to the cloud.

If your model looks like this:

class Video
  mount_uploader :attachment, VideoUploader
end

The migration would look like this:

@videos = Video.all
@videos.each do |video|
  video.remote_attachment_url = video.attachment_url
  video.save
end

If you execute this migration the following should happen:

Carrierwave downloads each image because you specified a remote url for the attachment(the current location, like http://test.com/images/1.jpg) and saves it to the cloud because you changed that in the uploader.

Edit:

Since San pointed out this will not work directly you should maybe create an extra column first, run a migration to copy the current attachment_urls from all the videos into that column, change the uploader after that and run the above migration using the copied urls in that new column. With another migration just delete the column again. Not that clean and easy but done in some minutes.

Answer 2

Minimal to Possibly Zero Donwtime Procedure

In my opinion, the easiest and fastest way to accomplish what you want with almost no downtime is this: (I will assume that you will use AWS cloud, but similar procedure is applicable to any cloud service)

1. Figure out and setup your assets bucket, bucket policies etc for making the assets publicly accessible.
2. Using s3cmd (command line tool for interacting with S3) or a GUI app, copy entire assets folder from file system to the appropriate folder in S3.
3. In your app, setup carrierwave and update your models/uploaders for :fog storage.

4. Do not restart your application yet. Instead bring up rails console and for your models, check that the new assets URL is correct and accessible as planned. For example, for a video model with picture asset, you can check this way:

   Video.first.picture.url
   

   This will give you a full cloud URL based on the updated settings. Copy the URL and paste in a browser to make sure that you can get to it fine.

5. If this works for at least one instance of each model that has assets, you are good to restart your application.

6. Upon restart, all your assets are being served from cloud, and you didn't need any migrations or multiple uploaders in your models.

7. (Based on comment by @Frederick Cheung): Using s3cmd (or something similar) rsync or sync the assets folder from the filesystem to S3 to account for assets that were uploaded between steps 2 and 5, if any.

PS: If you need help setting up carrierwave for cloud storage, let me know.

Answer 3

When we use Heroku, most of people suggest to use cloudinary. Free and simple setup. My case is when we use cloudinary service and need move into aws S3 for some reasons.

This is what i did with the uploader:

class AvatarUploader < CarrierWave::Uploader::Base

  def self.set_storage
    if ENV['UPLOADER_SERVICE'] == 'aws'
      :fog
    else
      nil
    end
  end

  if ENV['UPLOADER_SERVICE'] == 'aws'
     include CarrierWave::MiniMagick
  else
     include Cloudinary::CarrierWave
  end

  storage set_storage
end

also, setup the rake task:

task :migrate_cloudinary_to_aws do
    profile_image_old_url = []
    Profile.where("picture IS NOT NULL").each do |profile_image|
      profile_image_old_url << profile_image
    end

   ENV['UPLOADER_SERVICE'] = 'aws'
   load("#{Rails.root}/app/uploaders/avatar_uploader.rb")

   Profile.where("picture IS NOT NULL OR cover IS NOT NULL").each do |profile_image|
     old_profile_image = profile_image_old_url.detect { |image| image.id == profile_image.id }
     profile_image.remote_picture_url = old_profile_image.picture.url
     profile_image.save
   end
end

The trick is how to change the uploader provider by env variable. Good luck!