Warning: when using a load-balanced production configuration, this configuration can result in intermittently broken assets. The problem occurs when traffic arrives during a reload that includes fresh assets. If care isn't taken, it's possible for a browser to request assets that aren't yet available from every backend.
The CDN may then cache a 404 for that asset, with horrible results. A typical symptom is missing CSS and JS for all subsequent visitors. The recovery from this breakage may be painfully slow, because cache invalidation in Cloudfront is a) fiddly and b) takes minutes/hours to complete. You can fine-tune Cloudfront to not cache 404s, but serving them to any user at all is simply unacceptable in production.
To avoid all this, you must ensure that fully precompiled assets are available from all CDN origins prior to any reload. There's several ways to achieve that, but I'll generally configure a sync to S3 in the before_restart hook; the CDN then uses the bucket as origin. You could also use a CI/CD pipeline, or blue/green deploys, or a shared public/ folder, or carefully choreographed rolling reloads.
You can also put a CDN directly in front of your application for serving dynamic content, as well. Talking about CloudFront specifically, one benefit of doing this includes reduced latency because AWS maintains persistent connections in their internal network much longer. So I usually have two CDNs configured: one for static assets, and one for dynamic requests.
Join 24,647+ developers who get early access to new screencasts, articles, guides, updates, and more.