Way back in the days when I was first building this Rails app, not knowing much about Amazon services, I decided to use Dropbox’s Public folder to store and serve all the images (it was bad, I know). As time passes by, Public folder is no more, and I had to do the grownup thing and migrate files to a real CDN/storage service. I had just finally done it, and it was surprisingly painless.
First I took the chance to upgrade Paperclip, cleaned up the path constructor to fix some unideal paths, and changed the storage settings to S3.
Next I found rclone, rclone syncs files across cloud storages. I used a spare computer to run the sync job from Dropbox to the S3 bucket, it took around 14 hours.
During this time, I made an inventory of all the file name differences, and wrote a simple script to rename the files.
Once the sync was completed, I pushed the changes to heroku as a review app, sharing the production database. Connecting to the review app console, I did a dry run of the renaming script, which uncovered a bunch of unexpected file names… I made a bunch more changes and did a few more dry-runs, and eventually successfully renamed all the files. Using the review app, I confirmed all the links were indeed working.
Then I tried syncing again with rclone to see what files had been uploaded since the last run. With
rclone sync --dry-run I found out because of the name changes, rclone will copy/remove all the changed files, so it obviously wouldn’t work. Well, worst case scenario, I can still manaully copy all the files. So, I logged into Dropbox and checked the recent file changes. Luckily there weren’t many, so I just copied and renamed them one by one.
Lastly, I set up Cloudfront and changed the s3 alias, and that was all!
I was really surprised at how smooth the process was, so this is a note to self that things aren’t always as tedious as I imagine them to be. 😂